system/freebsd-src
system/freebsd-src
1
0
Fork 0
mirror of https://github.com/freebsd/freebsd-src synced 2024-09-21 01:03:42 +00:00
Code Issues Packages Projects Releases Wiki Activity
freebsd-src/sys/dev/esp/ncr53c9x.c

3259 lines
85 KiB
C
Raw Normal View History

Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*-
* Copyright (c) 2004 Scott Long
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* Copyright (c) 2005, 2008 Marius Strobl <marius@FreeBSD.org>
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
/* $NetBSD: ncr53c9x.c,v 1.145 2012/06/18 21:23:56 martin Exp $ */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*-
* Copyright (c) 1998, 2002 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Charles M. Hannum.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* Copyright (c) 1994 Peter Galbavy
* Copyright (c) 1995 Paul Kranenburg
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Peter Galbavy
* 4. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Based on aic6360 by Jarle Greipsland
*
* Acknowledgements: Many of the algorithms used in this driver are
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* inspired by the work of Julian Elischer (julian@FreeBSD.org) and
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* Charles Hannum (mycroft@duality.gnu.ai.mit.edu). Thanks a million!
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bus.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/lock.h>
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
#include <sys/module.h>
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#include <sys/mutex.h>
#include <sys/queue.h>
#include <sys/time.h>
#include <sys/callout.h>
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_debug.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt_sim.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
#include <dev/esp/ncr53c9xreg.h>
#include <dev/esp/ncr53c9xvar.h>
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
devclass_t esp_devclass;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
MODULE_DEPEND(esp, cam, 1, 1, 1);
#ifdef NCR53C9X_DEBUG
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
int ncr53c9x_debug =
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
NCR_SHOWMISC /* | NCR_SHOWPHASE | NCR_SHOWTRAC | NCR_SHOWCMDS */;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#endif
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static void ncr53c9x_abort(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void ncr53c9x_action(struct cam_sim *sim, union ccb *ccb);
static void ncr53c9x_async(void *cbarg, uint32_t code,
struct cam_path *path, void *arg);
static void ncr53c9x_callout(void *arg);
static void ncr53c9x_clear(struct ncr53c9x_softc *sc, cam_status result);
static void ncr53c9x_clear_target(struct ncr53c9x_softc *sc, int target,
cam_status result);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static void ncr53c9x_dequeue(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
static void ncr53c9x_done(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
static void ncr53c9x_free_ecb(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
static void ncr53c9x_msgin(struct ncr53c9x_softc *sc);
static void ncr53c9x_msgout(struct ncr53c9x_softc *sc);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void ncr53c9x_init(struct ncr53c9x_softc *sc, int doreset);
static void ncr53c9x_intr1(struct ncr53c9x_softc *sc);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static void ncr53c9x_poll(struct cam_sim *sim);
static int ncr53c9x_rdfifo(struct ncr53c9x_softc *sc, int how);
static int ncr53c9x_reselect(struct ncr53c9x_softc *sc, int message,
int tagtype, int tagid);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void ncr53c9x_reset(struct ncr53c9x_softc *sc);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static void ncr53c9x_sense(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
static void ncr53c9x_sched(struct ncr53c9x_softc *sc);
static void ncr53c9x_select(struct ncr53c9x_softc *sc,
struct ncr53c9x_ecb *ecb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
static void ncr53c9x_watch(void *arg);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
static void ncr53c9x_wrfifo(struct ncr53c9x_softc *sc, uint8_t *p,
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
int len);
static struct ncr53c9x_ecb *ncr53c9x_get_ecb(struct ncr53c9x_softc *sc);
static struct ncr53c9x_linfo *ncr53c9x_lunsearch(struct ncr53c9x_tinfo *sc,
int64_t lun);
static inline void ncr53c9x_readregs(struct ncr53c9x_softc *sc);
static inline void ncr53c9x_setsync(struct ncr53c9x_softc *sc,
struct ncr53c9x_tinfo *ti);
static inline int ncr53c9x_stp2cpb(struct ncr53c9x_softc *sc,
int period);
#define NCR_RDFIFO_START 0
#define NCR_RDFIFO_CONTINUE 1
#define NCR_SET_COUNT(sc, size) do { \
NCR_WRITE_REG((sc), NCR_TCL, (size)); \
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG((sc), NCR_TCM, (size) >> 8); \
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
if ((sc->sc_features & NCR_F_LARGEXFER) != 0) \
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG((sc), NCR_TCH, (size) >> 16); \
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (sc->sc_rev == NCR_VARIANT_FAS366) \
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG(sc, NCR_RCH, 0); \
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} while (/* CONSTCOND */0)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#ifndef mstohz
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#define mstohz(ms) \
(((ms) < 0x20000) ? \
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
((ms +0u) / 1000u) * hz : \
((ms +0u) * hz) /1000u)
#endif
/*
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
* Names for the NCR53c9x variants, corresponding to the variant tags
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* in ncr53c9xvar.h.
*/
static const char *ncr53c9x_variant_names[] = {
"ESP100",
"ESP100A",
"ESP200",
"NCR53C94",
"NCR53C96",
"ESP406",
"FAS408",
"FAS216",
"AM53C974",
"FAS366/HME",
"NCR53C90 (86C01)",
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
"FAS100A",
"FAS236",
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
};
/*
* Search linked list for LUN info by LUN id.
*/
static struct ncr53c9x_linfo *
ncr53c9x_lunsearch(struct ncr53c9x_tinfo *ti, int64_t lun)
{
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
LIST_FOREACH(li, &ti->luns, link)
if (li->lun == lun)
return (li);
return (NULL);
}
/*
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* Attach this instance, and then all the sub-devices.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
int
ncr53c9x_attach(struct ncr53c9x_softc *sc)
{
struct cam_devq *devq;
struct cam_sim *sim;
struct cam_path *path;
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
struct ncr53c9x_ecb *ecb;
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
int error, i;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (NCR_LOCK_INITIALIZED(sc) == 0) {
device_printf(sc->sc_dev, "mutex not initialized\n");
return (ENXIO);
}
callout_init_mtx(&sc->sc_watchdog, &sc->sc_lock, 0);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* Note, the front-end has set us up to print the chip variation.
*/
if (sc->sc_rev >= NCR_VARIANT_MAX) {
device_printf(sc->sc_dev, "unknown variant %d, devices not "
"attached\n", sc->sc_rev);
return (EINVAL);
}
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
device_printf(sc->sc_dev, "%s, %d MHz, SCSI ID %d\n",
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_variant_names[sc->sc_rev], sc->sc_freq, sc->sc_id);
sc->sc_ntarg = (sc->sc_rev == NCR_VARIANT_FAS366) ? 16 : 8;
/*
* Allocate SCSI message buffers.
* Front-ends can override allocation to avoid alignment
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* handling in the DMA engines. Note that ncr53c9x_msgout()
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* can request a 1 byte DMA transfer.
*/
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (sc->sc_omess == NULL) {
sc->sc_omess_self = 1;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_omess = malloc(NCR_MAX_MSG_LEN, M_DEVBUF, M_NOWAIT);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (sc->sc_omess == NULL) {
device_printf(sc->sc_dev,
"cannot allocate MSGOUT buffer\n");
return (ENOMEM);
}
} else
sc->sc_omess_self = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (sc->sc_imess == NULL) {
sc->sc_imess_self = 1;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imess = malloc(NCR_MAX_MSG_LEN + 1, M_DEVBUF, M_NOWAIT);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (sc->sc_imess == NULL) {
device_printf(sc->sc_dev,
"cannot allocate MSGIN buffer\n");
error = ENOMEM;
goto fail_omess;
}
} else
sc->sc_imess_self = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_tinfo = malloc(sc->sc_ntarg * sizeof(sc->sc_tinfo[0]),
M_DEVBUF, M_NOWAIT | M_ZERO);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (sc->sc_tinfo == NULL) {
device_printf(sc->sc_dev,
"cannot allocate target info buffer\n");
error = ENOMEM;
goto fail_imess;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
* Treat NCR53C90 with the 86C01 DMA chip exactly as ESP100
* from now on.
*/
if (sc->sc_rev == NCR_VARIANT_NCR53C90_86C01)
sc->sc_rev = NCR_VARIANT_ESP100;
sc->sc_ccf = FREQTOCCF(sc->sc_freq);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* The value *must not* be == 1. Make it 2. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_ccf == 1)
sc->sc_ccf = 2;
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* The recommended timeout is 250ms. This register is loaded
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* with a value calculated as follows, from the docs:
*
s/timout/timeout
2011-12-15 06:29:13 +00:00
* (timeout period) x (CLK frequency)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* reg = -------------------------------------
* 8192 x (Clock Conversion Factor)
*
* Since CCF has a linear relation to CLK, this generally computes
* to the constant of 153.
*/
sc->sc_timeout = ((250 * 1000) * sc->sc_freq) / (8192 * sc->sc_ccf);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* The CCF register only has 3 bits; 0 is actually 8. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_ccf &= 7;
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Register with CAM.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
devq = cam_simq_alloc(sc->sc_ntarg);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (devq == NULL) {
device_printf(sc->sc_dev, "cannot allocate device queue\n");
error = ENOMEM;
goto fail_tinfo;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sim = cam_sim_alloc(ncr53c9x_action, ncr53c9x_poll, "esp", sc,
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_get_unit(sc->sc_dev), &sc->sc_lock, 1, NCR_TAG_DEPTH, devq);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sim == NULL) {
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
device_printf(sc->sc_dev, "cannot allocate SIM entry\n");
error = ENOMEM;
goto fail_devq;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK(sc);
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
if (xpt_bus_register(sim, sc->sc_dev, 0) != CAM_SUCCESS) {
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
device_printf(sc->sc_dev, "cannot register bus\n");
error = EIO;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto fail_lock;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
if (xpt_create_path(&path, NULL, cam_sim_path(sim),
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
device_printf(sc->sc_dev, "cannot create path\n");
error = EIO;
goto fail_bus;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (xpt_register_async(AC_LOST_DEVICE, ncr53c9x_async, sim, path) !=
CAM_REQ_CMP) {
device_printf(sc->sc_dev, "cannot register async handler\n");
error = EIO;
goto fail_path;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_sim = sim;
sc->sc_path = path;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Reset state and bus. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#if 0
sc->sc_cfflags = sc->sc_dev.dv_cfdata->cf_flags;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#else
sc->sc_cfflags = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#endif
sc->sc_state = 0;
ncr53c9x_init(sc, 1);
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
TAILQ_INIT(&sc->free_list);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
if ((sc->ecb_array =
malloc(sizeof(struct ncr53c9x_ecb) * NCR_TAG_DEPTH, M_DEVBUF,
M_NOWAIT | M_ZERO)) == NULL) {
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
device_printf(sc->sc_dev, "cannot allocate ECB array\n");
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
error = ENOMEM;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto fail_async;
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
}
for (i = 0; i < NCR_TAG_DEPTH; i++) {
ecb = &sc->ecb_array[i];
ecb->sc = sc;
ecb->tag_id = i;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
callout_init_mtx(&ecb->ch, &sc->sc_lock, 0);
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
TAILQ_INSERT_HEAD(&sc->free_list, ecb, free_links);
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_reset(&sc->sc_watchdog, 60 * hz, ncr53c9x_watch, sc);
NCR_UNLOCK(sc);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
return (0);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
fail_async:
xpt_register_async(0, ncr53c9x_async, sim, path);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
fail_path:
xpt_free_path(path);
fail_bus:
xpt_bus_deregister(cam_sim_path(sim));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
fail_lock:
NCR_UNLOCK(sc);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
cam_sim_free(sim, TRUE);
fail_devq:
cam_simq_free(devq);
fail_tinfo:
free(sc->sc_tinfo, M_DEVBUF);
fail_imess:
if (sc->sc_imess_self)
free(sc->sc_imess, M_DEVBUF);
fail_omess:
if (sc->sc_omess_self)
free(sc->sc_omess, M_DEVBUF);
return (error);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
int
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
ncr53c9x_detach(struct ncr53c9x_softc *sc)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
struct ncr53c9x_linfo *li, *nextli;
int t;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
callout_drain(&sc->sc_watchdog);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK(sc);
if (sc->sc_tinfo) {
/* Cancel all commands. */
ncr53c9x_clear(sc, CAM_REQ_ABORTED);
/* Free logical units. */
for (t = 0; t < sc->sc_ntarg; t++) {
for (li = LIST_FIRST(&sc->sc_tinfo[t].luns); li;
li = nextli) {
nextli = LIST_NEXT(li, link);
free(li, M_DEVBUF);
}
}
}
xpt_register_async(0, ncr53c9x_async, sc->sc_sim, sc->sc_path);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
xpt_free_path(sc->sc_path);
xpt_bus_deregister(cam_sim_path(sc->sc_sim));
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cam_sim_free(sc->sc_sim, TRUE);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_UNLOCK(sc);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
free(sc->ecb_array, M_DEVBUF);
free(sc->sc_tinfo, M_DEVBUF);
if (sc->sc_imess_self)
free(sc->sc_imess, M_DEVBUF);
if (sc->sc_omess_self)
free(sc->sc_omess, M_DEVBUF);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
return (0);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* This is the generic ncr53c9x reset function. It does not reset the SCSI
* bus, only this controller, but kills any on-going commands, and also stops
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* and resets the DMA.
*
* After reset, registers are loaded with the defaults from the attach
* routine above.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_reset(struct ncr53c9x_softc *sc)
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Reset DMA first. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRDMA_RESET(sc);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Reset SCSI chip. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_RSTCHIP);
NCRCMD(sc, NCRCMD_NOP);
DELAY(500);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Do these backwards, and fall through. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
switch (sc->sc_rev) {
case NCR_VARIANT_ESP406:
case NCR_VARIANT_FAS408:
NCR_WRITE_REG(sc, NCR_CFG5, sc->sc_cfg5 | NCRCFG5_SINT);
NCR_WRITE_REG(sc, NCR_CFG4, sc->sc_cfg4);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* FALLTHROUGH */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_VARIANT_AM53C974:
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
case NCR_VARIANT_FAS100A:
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_VARIANT_FAS216:
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
case NCR_VARIANT_FAS236:
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_VARIANT_NCR53C94:
case NCR_VARIANT_NCR53C96:
case NCR_VARIANT_ESP200:
sc->sc_features |= NCR_F_HASCFG3;
NCR_WRITE_REG(sc, NCR_CFG3, sc->sc_cfg3);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* FALLTHROUGH */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_VARIANT_ESP100A:
sc->sc_features |= NCR_F_SELATN3;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
if ((sc->sc_cfg2 & NCRCFG2_FE) != 0)
sc->sc_features |= NCR_F_LARGEXFER;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG(sc, NCR_CFG2, sc->sc_cfg2);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* FALLTHROUGH */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_VARIANT_ESP100:
NCR_WRITE_REG(sc, NCR_CFG1, sc->sc_cfg1);
NCR_WRITE_REG(sc, NCR_CCF, sc->sc_ccf);
NCR_WRITE_REG(sc, NCR_SYNCOFF, 0);
NCR_WRITE_REG(sc, NCR_TIMEOUT, sc->sc_timeout);
break;
case NCR_VARIANT_FAS366:
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
sc->sc_features |= NCR_F_HASCFG3 | NCR_F_FASTSCSI |
NCR_F_SELATN3 | NCR_F_LARGEXFER;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_cfg3 = NCRFASCFG3_FASTCLK | NCRFASCFG3_OBAUTO;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (sc->sc_id > 7)
sc->sc_cfg3 |= NCRFASCFG3_IDBIT3;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_cfg3_fscsi = NCRFASCFG3_FASTSCSI;
NCR_WRITE_REG(sc, NCR_CFG3, sc->sc_cfg3);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_cfg2 = NCRCFG2_HMEFE | NCRCFG2_HME32;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG(sc, NCR_CFG2, sc->sc_cfg2);
NCR_WRITE_REG(sc, NCR_CFG1, sc->sc_cfg1);
NCR_WRITE_REG(sc, NCR_CCF, sc->sc_ccf);
NCR_WRITE_REG(sc, NCR_SYNCOFF, 0);
NCR_WRITE_REG(sc, NCR_TIMEOUT, sc->sc_timeout);
break;
default:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
device_printf(sc->sc_dev,
"unknown revision code, assuming ESP100\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG(sc, NCR_CFG1, sc->sc_cfg1);
NCR_WRITE_REG(sc, NCR_CCF, sc->sc_ccf);
NCR_WRITE_REG(sc, NCR_SYNCOFF, 0);
NCR_WRITE_REG(sc, NCR_TIMEOUT, sc->sc_timeout);
}
if (sc->sc_rev == NCR_VARIANT_AM53C974)
NCR_WRITE_REG(sc, NCR_AMDCFG4, sc->sc_cfg4);
#if 0
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
device_printf(sc->sc_dev, "%s: revision %d\n", __func__, sc->sc_rev);
device_printf(sc->sc_dev, "%s: cfg1 0x%x, cfg2 0x%x, cfg3 0x%x, ccf "
"0x%x, timeout 0x%x\n", __func__, sc->sc_cfg1, sc->sc_cfg2,
sc->sc_cfg3, sc->sc_ccf, sc->sc_timeout);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#endif
}
/*
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* Clear all commands.
*/
static void
ncr53c9x_clear(struct ncr53c9x_softc *sc, cam_status result)
{
struct ncr53c9x_ecb *ecb;
int r;
NCR_LOCK_ASSERT(sc, MA_OWNED);
/* Cancel any active commands. */
sc->sc_state = NCR_CLEANING;
sc->sc_msgify = 0;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
ecb = sc->sc_nexus;
if (ecb != NULL) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ecb->ccb->ccb_h.status = result;
ncr53c9x_done(sc, ecb);
}
/* Cancel outstanding disconnected commands. */
for (r = 0; r < sc->sc_ntarg; r++)
ncr53c9x_clear_target(sc, r, result);
}
/*
* Clear all commands for a specific target.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
static void
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ncr53c9x_clear_target(struct ncr53c9x_softc *sc, int target,
cam_status result)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
struct ncr53c9x_ecb *ecb;
struct ncr53c9x_linfo *li;
int i;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/* Cancel outstanding disconnected commands on each LUN. */
LIST_FOREACH(li, &sc->sc_tinfo[target].luns, link) {
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
ecb = li->untagged;
if (ecb != NULL) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
li->untagged = NULL;
/*
* XXX should we terminate a command
* that never reached the disk?
*/
li->busy = 0;
ecb->ccb->ccb_h.status = result;
ncr53c9x_done(sc, ecb);
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
for (i = 0; i < NCR_TAG_DEPTH; i++) {
ecb = li->queued[i];
if (ecb != NULL) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
li->queued[i] = NULL;
ecb->ccb->ccb_h.status = result;
ncr53c9x_done(sc, ecb);
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
li->used = 0;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Initialize ncr53c9x state machine.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_init(struct ncr53c9x_softc *sc, int doreset)
{
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
int r;
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_MISC(("[NCR_INIT(%d) %d] ", doreset, sc->sc_state));
if (sc->sc_state == 0) {
/* First time through; initialize. */
TAILQ_INIT(&sc->ready_list);
sc->sc_nexus = NULL;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
memset(sc->sc_tinfo, 0, sizeof(*sc->sc_tinfo));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
for (r = 0; r < sc->sc_ntarg; r++) {
LIST_INIT(&sc->sc_tinfo[r].luns);
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
} else
ncr53c9x_clear(sc, CAM_CMD_TIMEOUT);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Reset the chip to a known state.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
ncr53c9x_reset(sc);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
sc->sc_flags = 0;
sc->sc_msgpriq = sc->sc_msgout = sc->sc_msgoutq = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_phase = sc->sc_prevphase = INVALID_PHASE;
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/*
* If we're the first time through, set the default parameters
* for all targets. Otherwise we only clear their current transfer
* settings so we'll renegotiate their goal settings with the next
* command.
*/
if (sc->sc_state == 0) {
for (r = 0; r < sc->sc_ntarg; r++) {
ti = &sc->sc_tinfo[r];
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/* XXX - config flags per target: low bits: no reselect; high bits: no synch */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->flags = ((sc->sc_minsync != 0 &&
(sc->sc_cfflags & (1 << ((r & 7) + 8))) == 0) ?
0 : T_SYNCHOFF) |
((sc->sc_cfflags & (1 << (r & 7))) == 0 ?
0 : T_RSELECTOFF);
ti->curr.period = ti->goal.period = 0;
ti->curr.offset = ti->goal.offset = 0;
ti->curr.width = ti->goal.width =
MSG_EXT_WDTR_BUS_8_BIT;
}
} else {
for (r = 0; r < sc->sc_ntarg; r++) {
ti = &sc->sc_tinfo[r];
ti->flags &= ~(T_SDTRSENT | T_WDTRSENT);
ti->curr.period = 0;
ti->curr.offset = 0;
ti->curr.width = MSG_EXT_WDTR_BUS_8_BIT;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
if (doreset) {
sc->sc_state = NCR_SBR;
NCRCMD(sc, NCRCMD_RSTSCSI);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/* Give the bus a fighting chance to settle. */
DELAY(250000);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
} else {
sc->sc_state = NCR_IDLE;
ncr53c9x_sched(sc);
}
}
/*
* Read the NCR registers, and save their contents for later use.
* NCR_STAT, NCR_STEP & NCR_INTR are mostly zeroed out when reading
* NCR_INTR - so make sure it is the last read.
*
* I think that (from reading the docs) most bits in these registers
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* only make sense when the DMA CSR has an interrupt showing. Call only
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* if an interrupt is pending.
*/
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static inline void
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_readregs(struct ncr53c9x_softc *sc)
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_espstat = NCR_READ_REG(sc, NCR_STAT);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Only the step bits are of interest. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_espstep = NCR_READ_REG(sc, NCR_STEP) & NCRSTEP_MASK;
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
if (sc->sc_rev == NCR_VARIANT_FAS366)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_espstat2 = NCR_READ_REG(sc, NCR_STAT2);
sc->sc_espintr = NCR_READ_REG(sc, NCR_INTR);
/*
* Determine the SCSI bus phase, return either a real SCSI bus phase
* or some pseudo phase we use to detect certain exceptions.
*/
sc->sc_phase = (sc->sc_espintr & NCRINTR_DIS) ?
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
BUSFREE_PHASE : sc->sc_espstat & NCRSTAT_PHASE;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_INTS(("regs[intr=%02x,stat=%02x,step=%02x,stat2=%02x] ",
sc->sc_espintr, sc->sc_espstat, sc->sc_espstep, sc->sc_espstat2));
}
/*
* Convert Synchronous Transfer Period to chip register Clock Per Byte value.
*/
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static inline int
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_stp2cpb(struct ncr53c9x_softc *sc, int period)
{
int v;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
v = (sc->sc_freq * period) / 250;
if (ncr53c9x_cpb2stp(sc, v) < period)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Correct round-down error. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
v++;
return (v);
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
static inline void
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_setsync(struct ncr53c9x_softc *sc, struct ncr53c9x_tinfo *ti)
{
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t cfg3, syncoff, synctp;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
cfg3 = sc->sc_cfg3;
if (ti->curr.offset != 0) {
syncoff = ti->curr.offset;
synctp = ncr53c9x_stp2cpb(sc, ti->curr.period);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_features & NCR_F_FASTSCSI) {
/*
* If the period is 200ns or less (ti->period <= 50),
* put the chip in Fast SCSI mode.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.period <= 50)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* There are (at least) 4 variations of the
* configuration 3 register. The drive attach
* routine sets the appropriate bit to put the
* chip into Fast SCSI mode so that it doesn't
* have to be figured out here each time.
*/
cfg3 |= sc->sc_cfg3_fscsi;
}
/*
* Am53c974 requires different SYNCTP values when the
* FSCSI bit is off.
*/
if (sc->sc_rev == NCR_VARIANT_AM53C974 &&
(cfg3 & NCRAMDCFG3_FSCSI) == 0)
synctp--;
} else {
syncoff = 0;
synctp = 0;
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.width != MSG_EXT_WDTR_BUS_8_BIT) {
if (sc->sc_rev == NCR_VARIANT_FAS366)
cfg3 |= NCRFASCFG3_EWIDE;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_features & NCR_F_HASCFG3)
NCR_WRITE_REG(sc, NCR_CFG3, cfg3);
NCR_WRITE_REG(sc, NCR_SYNCOFF, syncoff);
NCR_WRITE_REG(sc, NCR_SYNCTP, synctp);
}
/*
* Send a command to a target, set the driver state to NCR_SELECTING
* and let the caller take care of the rest.
*
* Keeping this as a function allows me to say that this may be done
* by DMA instead of programmed I/O soon.
*/
static void
ncr53c9x_select(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
{
struct ncr53c9x_tinfo *ti;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t *cmd;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
size_t dmasize;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
int clen, error, selatn3, selatns;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
int lun = ecb->ccb->ccb_h.target_lun;
int target = ecb->ccb->ccb_h.target_id;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s(t%d,l%d,cmd:%x,tag:%x,%x)] ", __func__, target, lun,
ecb->cmd.cmd.opcode, ecb->tag[0], ecb->tag[1]));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[target];
sc->sc_state = NCR_SELECTING;
/*
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* Schedule the callout now, the first time we will go away
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* expecting to come back due to an interrupt, because it is
* always possible that the interrupt may never happen.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_reset(&ecb->ch, mstohz(ecb->timeout), ncr53c9x_callout, ecb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* The docs say the target register is never reset, and I
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* can't think of a better place to set it.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
if (sc->sc_rev == NCR_VARIANT_FAS366) {
NCRCMD(sc, NCRCMD_FLUSH);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_WRITE_REG(sc, NCR_SELID, target | NCR_BUSID_HMEXC32 |
NCR_BUSID_HMEENCID);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} else
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_WRITE_REG(sc, NCR_SELID, target);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/*
* If we are requesting sense, force a renegotiation if we are
* currently using anything different from asynchronous at 8 bit
* as the target might have lost our transfer negotiations.
*/
if ((ecb->flags & ECB_SENSE) != 0 && (ti->curr.offset != 0 ||
ti->curr.width != MSG_EXT_WDTR_BUS_8_BIT)) {
ti->curr.period = 0;
ti->curr.offset = 0;
ti->curr.width = MSG_EXT_WDTR_BUS_8_BIT;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ncr53c9x_setsync(sc, ti);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
selatn3 = selatns = 0;
if (ecb->tag[0] != 0) {
if (sc->sc_features & NCR_F_SELATN3)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Use SELATN3 to send tag messages. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
selatn3 = 1;
else
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* We don't have SELATN3; use SELATNS to send tags. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
selatns = 1;
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.period != ti->goal.period ||
ti->curr.offset != ti->goal.offset ||
ti->curr.width != ti->goal.width) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* We have to use SELATNS to send sync/wide messages. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
selatn3 = 0;
selatns = 1;
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
cmd = (uint8_t *)&ecb->cmd.cmd;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (selatn3) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* We'll use tags with SELATN3. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
clen = ecb->clen + 3;
cmd -= 3;
cmd[0] = MSG_IDENTIFY(lun, 1); /* msg[0] */
cmd[1] = ecb->tag[0]; /* msg[1] */
cmd[2] = ecb->tag[1]; /* msg[2] */
} else {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* We don't have tags, or will send messages with SELATNS. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
clen = ecb->clen + 1;
cmd -= 1;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
cmd[0] = MSG_IDENTIFY(lun, (ti->flags & T_RSELECTOFF) == 0);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
if ((sc->sc_features & NCR_F_DMASELECT) && !selatns) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Setup DMA transfer for command. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
dmasize = clen;
sc->sc_cmdlen = clen;
sc->sc_cmdp = cmd;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
error = NCRDMA_SETUP(sc, &sc->sc_cmdp, &sc->sc_cmdlen, 0,
&dmasize);
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
if (error != 0)
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
goto cmd;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Program the SCSI counter. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_SET_COUNT(sc, dmasize);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Load the count in. */
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
NCRCMD(sc, NCRCMD_NOP | NCRCMD_DMA);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* And get the target's attention. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (selatn3) {
sc->sc_msgout = SEND_TAG;
sc->sc_flags |= NCR_ATN;
NCRCMD(sc, NCRCMD_SELATN3 | NCRCMD_DMA);
} else
NCRCMD(sc, NCRCMD_SELATN | NCRCMD_DMA);
NCRDMA_GO(sc);
return;
}
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cmd:
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* Who am I? This is where we tell the target that we are
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* happy for it to disconnect etc.
*/
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Now get the command into the FIFO. */
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
sc->sc_cmdlen = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_wrfifo(sc, cmd, clen);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* And get the target's attention. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (selatns) {
NCR_MSGS(("SELATNS \n"));
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Arbitrate, select and stop after IDENTIFY message. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_SELATNS);
} else if (selatn3) {
sc->sc_msgout = SEND_TAG;
sc->sc_flags |= NCR_ATN;
NCRCMD(sc, NCRCMD_SELATN3);
} else
NCRCMD(sc, NCRCMD_SELATN);
}
static void
ncr53c9x_free_ecb(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->flags = 0;
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
TAILQ_INSERT_TAIL(&sc->free_list, ecb, free_links);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
static struct ncr53c9x_ecb *
ncr53c9x_get_ecb(struct ncr53c9x_softc *sc)
{
struct ncr53c9x_ecb *ecb;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
ecb = TAILQ_FIRST(&sc->free_list);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (ecb) {
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
if (ecb->flags != 0)
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
panic("%s: ecb flags not cleared", __func__);
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
TAILQ_REMOVE(&sc->free_list, ecb, free_links);
ecb->flags = ECB_ALLOC;
bzero(&ecb->ccb, sizeof(struct ncr53c9x_ecb) -
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
offsetof(struct ncr53c9x_ecb, ccb));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
return (ecb);
}
/*
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* DRIVER FUNCTIONS CALLABLE FROM HIGHER LEVEL DRIVERS:
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Start a SCSI-command.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* This function is called by the higher level SCSI-driver to queue/run
* SCSI-commands.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_action(struct cam_sim *sim, union ccb *ccb)
{
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ccb_pathinq *cpi;
struct ccb_scsiio *csio;
struct ccb_trans_settings *cts;
struct ccb_trans_settings_scsi *scsi;
struct ccb_trans_settings_spi *spi;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
struct ncr53c9x_ecb *ecb;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_softc *sc;
struct ncr53c9x_tinfo *ti;
int target;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc = cam_sim_softc(sim);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s %d]", __func__, ccb->ccb_h.func_code));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
switch (ccb->ccb_h.func_code) {
case XPT_RESET_BUS:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ncr53c9x_init(sc, 1);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_REQ_CMP;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_CALC_GEOMETRY:
Make the XPT_CALC_GEOMETRY op dependent on the sc_extended_geometry flag that is set in the bus front-end.
2004-06-11 15:33:20 +00:00
cam_calc_geometry(&ccb->ccg, sc->sc_extended_geom);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
case XPT_PATH_INQ:
cpi = &ccb->cpi;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
cpi->version_num = 1;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
cpi->hba_inquiry = PI_SDTR_ABLE | PI_TAG_ABLE;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
cpi->hba_inquiry |=
(sc->sc_rev == NCR_VARIANT_FAS366) ? PI_WIDE_16 : 0;
cpi->target_sprt = 0;
cpi->hba_misc = 0;
cpi->hba_eng_cnt = 0;
cpi->max_target = sc->sc_ntarg - 1;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
cpi->max_lun = 7;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
cpi->initiator_id = sc->sc_id;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
strncpy(cpi->hba_vid, "NCR", HBA_IDLEN);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->unit_number = cam_sim_unit(sim);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cpi->bus_id = 0;
cpi->base_transfer_speed = 3300;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_2;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cpi->transport = XPORT_SPI;
cpi->transport_version = 2;
cpi->maxio = sc->sc_maxxfer;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_REQ_CMP;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_GET_TRAN_SETTINGS:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
cts = &ccb->cts;
ti = &sc->sc_tinfo[ccb->ccb_h.target_id];
scsi = &cts->proto_specific.scsi;
spi = &cts->xport_specific.spi;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_SPI;
cts->transport_version = 2;
if (cts->type == CTS_TYPE_CURRENT_SETTINGS) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
spi->sync_period = ti->curr.period;
spi->sync_offset = ti->curr.offset;
spi->bus_width = ti->curr.width;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
if ((ti->flags & T_TAG) != 0) {
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
} else {
spi->flags &= ~CTS_SPI_FLAGS_DISC_ENB;
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
}
} else {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ti->flags & T_SYNCHOFF) != 0) {
spi->sync_period = 0;
spi->sync_offset = 0;
} else {
spi->sync_period = sc->sc_minsync;
spi->sync_offset = sc->sc_maxoffset;
}
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
spi->bus_width = sc->sc_maxwidth;
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
}
spi->valid =
CTS_SPI_VALID_BUS_WIDTH |
CTS_SPI_VALID_SYNC_RATE |
CTS_SPI_VALID_SYNC_OFFSET |
CTS_SPI_VALID_DISC;
scsi->valid = CTS_SCSI_VALID_TQ;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_REQ_CMP;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_ABORT:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "XPT_ABORT called\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_TERM_IO:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "XPT_TERM_IO called\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_RESET_DEV:
case XPT_SCSI_IO:
if (ccb->ccb_h.target_id < 0 ||
ccb->ccb_h.target_id >= sc->sc_ntarg) {
ccb->ccb_h.status = CAM_PATH_INVALID;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
goto done;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/* Get an ECB to use. */
ecb = ncr53c9x_get_ecb(sc);
/*
* This should never happen as we track resources
* in the mid-layer.
*/
if (ecb == NULL) {
xpt_freeze_simq(sim, 1);
ccb->ccb_h.status = CAM_REQUEUE_REQ;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "unable to allocate ecb\n");
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
goto done;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Initialize ecb. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->ccb = ccb;
ecb->timeout = ccb->ccb_h.timeout;
When autosense is retrieved, tell CAM about it instead of juust pretending that the command succeeded. Sheesh! This makes CDROMs no longer cause an instant panic at boot. Thanks to Jake Burkholder for providing a remote test setup. Also make device resets work, thanks to another typo.
2004-06-12 05:19:17 +00:00
if (ccb->ccb_h.func_code == XPT_RESET_DEV) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->flags |= ECB_RESET;
ecb->clen = 0;
ecb->dleft = 0;
} else {
csio = &ccb->csio;
if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0)
bcopy(csio->cdb_io.cdb_ptr, &ecb->cmd.cmd,
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
csio->cdb_len);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
else
bcopy(csio->cdb_io.cdb_bytes, &ecb->cmd.cmd,
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
csio->cdb_len);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->clen = csio->cdb_len;
ecb->daddr = csio->data_ptr;
ecb->dleft = csio->dxfer_len;
}
ecb->stat = 0;
TAILQ_INSERT_TAIL(&sc->ready_list, ecb, chain);
ecb->flags |= ECB_READY;
if (sc->sc_state == NCR_IDLE)
ncr53c9x_sched(sc);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case XPT_SET_TRAN_SETTINGS:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
cts = &ccb->cts;
target = ccb->ccb_h.target_id;
ti = &sc->sc_tinfo[target];
scsi = &cts->proto_specific.scsi;
spi = &cts->xport_specific.spi;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
if ((scsi->valid & CTS_SCSI_VALID_TQ) != 0) {
if ((sc->sc_cfflags & (1<<((target & 7) + 16))) == 0 &&
(scsi->flags & CTS_SCSI_FLAGS_TAG_ENB)) {
NCR_MISC(("%s: target %d: tagged queuing\n",
device_get_nameunit(sc->sc_dev), target));
ti->flags |= T_TAG;
} else
ti->flags &= ~T_TAG;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) != 0) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_MISC(("%s: target %d: wide negotiation\n",
device_get_nameunit(sc->sc_dev), target));
ti->goal.width = spi->bus_width;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
}
if ((spi->valid & CTS_SPI_VALID_SYNC_RATE) != 0) {
NCR_MISC(("%s: target %d: sync period negotiation\n",
device_get_nameunit(sc->sc_dev), target));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->goal.period = spi->sync_period;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) != 0) {
NCR_MISC(("%s: target %d: sync offset negotiation\n",
device_get_nameunit(sc->sc_dev), target));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->goal.offset = spi->sync_offset;
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_REQ_CMP;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
default:
device_printf(sc->sc_dev, "Unhandled function code %d\n",
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ccb->ccb_h.func_code);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_PROVIDE_FAIL;
}
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
done:
xpt_done(ccb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* Used when interrupt driven I/O is not allowed, e.g. during boot.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
static void
ncr53c9x_poll(struct cam_sim *sim)
{
struct ncr53c9x_softc *sc;
sc = cam_sim_softc(sim);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s] ", __func__));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (NCRDMA_ISINTR(sc))
ncr53c9x_intr1(sc);
}
/*
* Asynchronous notification handler
*/
static void
ncr53c9x_async(void *cbarg, uint32_t code, struct cam_path *path, void *arg)
{
struct ncr53c9x_softc *sc;
struct ncr53c9x_tinfo *ti;
int target;
sc = cam_sim_softc(cbarg);
NCR_LOCK_ASSERT(sc, MA_OWNED);
switch (code) {
case AC_LOST_DEVICE:
target = xpt_path_target_id(path);
if (target < 0 || target >= sc->sc_ntarg)
break;
/* Cancel outstanding disconnected commands. */
ncr53c9x_clear_target(sc, target, CAM_REQ_ABORTED);
/* Set the default parameters for the target. */
ti = &sc->sc_tinfo[target];
/* XXX - config flags per target: low bits: no reselect; high bits: no synch */
ti->flags = ((sc->sc_minsync != 0 &&
(sc->sc_cfflags & (1 << ((target & 7) + 8))) == 0) ?
0 : T_SYNCHOFF) |
((sc->sc_cfflags & (1 << (target & 7))) == 0 ?
0 : T_RSELECTOFF);
ti->curr.period = ti->goal.period = 0;
ti->curr.offset = ti->goal.offset = 0;
ti->curr.width = ti->goal.width = MSG_EXT_WDTR_BUS_8_BIT;
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
}
/*
* LOW LEVEL SCSI UTILITIES
*/
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Schedule a SCSI operation. This has now been pulled out of the interrupt
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* handler so that we may call it from ncr53c9x_action and ncr53c9x_done.
* This may save us an unnecessary interrupt just to get things going.
* Should only be called when state == NCR_IDLE and with sc_lock held.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
static void
ncr53c9x_sched(struct ncr53c9x_softc *sc)
{
struct ncr53c9x_ecb *ecb;
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
int lun, tag;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s] ", __func__));
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_state != NCR_IDLE)
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
panic("%s: not IDLE (state=%d)", __func__, sc->sc_state);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* Find first ecb in ready queue that is for a target/lunit
* combinations that is not busy.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
TAILQ_FOREACH(ecb, &sc->ready_list, chain) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
lun = ecb->ccb->ccb_h.target_lun;
/* Select type of tag for this command */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ti->flags & (T_RSELECTOFF | T_TAG)) != T_TAG)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
tag = 0;
else if ((ecb->flags & ECB_SENSE) != 0)
tag = 0;
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
else if ((ecb->ccb->ccb_h.flags & CAM_TAG_ACTION_VALID) == 0)
tag = 0;
else if (ecb->ccb->csio.tag_action == CAM_TAG_ACTION_NONE)
tag = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
else
tag = ecb->ccb->csio.tag_action;
li = TINFO_LUN(ti, lun);
if (li == NULL) {
/* Initialize LUN info and add to list. */
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
li = malloc(sizeof(*li), M_DEVBUF, M_NOWAIT | M_ZERO);
if (li == NULL)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
continue;
li->lun = lun;
LIST_INSERT_HEAD(&ti->luns, li, link);
if (lun < NCR_NLUN)
ti->lun[lun] = li;
}
li->last_used = time_second;
if (tag == 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Try to issue this as an untagged command. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (li->untagged == NULL)
li->untagged = ecb;
}
if (li->untagged != NULL) {
tag = 0;
if ((li->busy != 1) && li->used == 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/*
* We need to issue this untagged command
* now.
*/
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb = li->untagged;
} else {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* not ready, yet */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
continue;
}
}
ecb->tag[0] = tag;
if (tag != 0) {
Get rid of UMA zones and instead allocate all ecb's up front and track them in a TAILQ. Re-arrange some of the ecb elements so that they can stay stable through alloc/free cycles while the rest get bzero'd. - Use the tag_id from the ecb rather than fro the ccb. The latter is only for target mode. - Honor the ccb flags for tag_action when deciding whether to do a tagged or untagged transaction. - Re-arrange autosense completion so that it works correctly in failure cases. - Turn on the PI_TAG_ABLE flag so that CAM will send us tagged transactions. This enables tagged queueing in the driver.
2004-06-13 09:08:44 +00:00
li->queued[ecb->tag_id] = ecb;
ecb->tag[1] = ecb->tag_id;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
li->used++;
}
if (li->untagged != NULL && (li->busy != 1)) {
li->busy = 1;
TAILQ_REMOVE(&sc->ready_list, ecb, chain);
ecb->flags &= ~ECB_READY;
sc->sc_nexus = ecb;
ncr53c9x_select(sc, ecb);
break;
}
if (li->untagged == NULL && tag != 0) {
TAILQ_REMOVE(&sc->ready_list, ecb, chain);
ecb->flags &= ~ECB_READY;
sc->sc_nexus = ecb;
ncr53c9x_select(sc, ecb);
break;
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
} else
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s %d:%d busy] \n", __func__,
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->ccb->ccb_h.target_id,
ecb->ccb->ccb_h.target_lun));
}
}
static void
ncr53c9x_sense(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
{
union ccb *ccb = ecb->ccb;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_linfo *li;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
struct ncr53c9x_tinfo *ti;
struct scsi_request_sense *ss = (void *)&ecb->cmd.cmd;
int lun;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s] ", __func__));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
lun = ccb->ccb_h.target_lun;
ti = &sc->sc_tinfo[ccb->ccb_h.target_id];
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Next, setup a REQUEST SENSE command block. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
memset(ss, 0, sizeof(*ss));
ss->opcode = REQUEST_SENSE;
ss->byte2 = ccb->ccb_h.target_lun << SCSI_CMD_LUN_SHIFT;
ss->length = sizeof(struct scsi_sense_data);
ecb->clen = sizeof(*ss);
Merge from r225950: Set the sense residual properly. Reviewed by: ken
2011-10-07 08:59:54 +00:00
memset(&ccb->csio.sense_data, 0, sizeof(ccb->csio.sense_data));
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
ecb->daddr = (uint8_t *)&ccb->csio.sense_data;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->dleft = sizeof(struct scsi_sense_data);
ecb->flags |= ECB_SENSE;
ecb->timeout = NCR_SENSE_TIMEOUT;
ti->senses++;
li = TINFO_LUN(ti, lun);
if (li->busy)
li->busy = 0;
ncr53c9x_dequeue(sc, ecb);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
li->untagged = ecb; /* Must be executed first to fix C/A. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
li->busy = 2;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (ecb == sc->sc_nexus)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_select(sc, ecb);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
else {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
TAILQ_INSERT_HEAD(&sc->ready_list, ecb, chain);
ecb->flags |= ECB_READY;
if (sc->sc_state == NCR_IDLE)
ncr53c9x_sched(sc);
}
}
/*
* POST PROCESSING OF SCSI_CMD (usually current)
*/
static void
ncr53c9x_done(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
{
union ccb *ccb = ecb->ccb;
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
Merge from r225950: Set the sense residual properly. Reviewed by: ken
2011-10-07 08:59:54 +00:00
int lun, sense_returned;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s(status:%x)] ", __func__, ccb->ccb_h.status));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ccb->ccb_h.target_id];
lun = ccb->ccb_h.target_lun;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
li = TINFO_LUN(ti, lun);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_stop(&ecb->ch);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* Now, if we've come here with no error code, i.e. we've kept the
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* initial CAM_REQ_CMP, and the status code signals that we should
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* check sense, we'll need to set up a request sense cmd block and
* push the command back into the ready queue *before* any other
* commands for this target/lunit, else we lose the sense info.
* We don't support chk sense conditions for the request sense cmd.
*/
if (ccb->ccb_h.status == CAM_REQ_CMP) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ccb->csio.scsi_status = ecb->stat;
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
if ((ecb->flags & ECB_ABORT) != 0)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->ccb_h.status = CAM_CMD_TIMEOUT;
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
else if ((ecb->flags & ECB_SENSE) != 0 &&
(ecb->stat != SCSI_STATUS_CHECK_COND)) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ccb->csio.scsi_status = SCSI_STATUS_CHECK_COND;
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR |
CAM_AUTOSNS_VALID;
Merge from r225950: Set the sense residual properly. Reviewed by: ken
2011-10-07 08:59:54 +00:00
sense_returned = sizeof(ccb->csio.sense_data) -
ecb->dleft;
if (sense_returned < ccb->csio.sense_len)
ccb->csio.sense_resid = ccb->csio.sense_len -
sense_returned;
else
ccb->csio.sense_resid = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
} else if (ecb->stat == SCSI_STATUS_CHECK_COND) {
if ((ecb->flags & ECB_SENSE) != 0)
ccb->ccb_h.status = CAM_AUTOSENSE_FAIL;
else {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* First, save the return values. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->csio.resid = ecb->dleft;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ccb->ccb_h.flags & CAM_DIS_AUTOSENSE) ==
0) {
ncr53c9x_sense(sc, ecb);
return;
}
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
} else
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ccb->csio.resid = ecb->dleft;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ecb->stat == SCSI_STATUS_QUEUE_FULL)
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR;
else if (ecb->stat == SCSI_STATUS_BUSY)
ccb->ccb_h.status = CAM_SCSI_BUSY;
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
} else if ((ccb->ccb_h.status & CAM_DEV_QFRZN) == 0) {
ccb->ccb_h.status |= CAM_DEV_QFRZN;
xpt_freeze_devq(ccb->ccb_h.path, 1);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
#ifdef NCR53C9X_DEBUG
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if ((ncr53c9x_debug & NCR_SHOWTRAC) != 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (ccb->csio.resid != 0)
printf("resid=%d ", ccb->csio.resid);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ccb->ccb_h.status & CAM_AUTOSNS_VALID) != 0)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
printf("sense=0x%02x\n",
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ccb->csio.sense_data.error_code);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
else
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
printf("status SCSI=0x%x CAM=0x%x\n",
ccb->csio.scsi_status, ccb->ccb_h.status);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
#endif
/*
* Remove the ECB from whatever queue it's on.
*/
ncr53c9x_dequeue(sc, ecb);
if (ecb == sc->sc_nexus) {
sc->sc_nexus = NULL;
if (sc->sc_state != NCR_CLEANING) {
sc->sc_state = NCR_IDLE;
ncr53c9x_sched(sc);
}
}
Freeze the device queue if a request didn't complete without error and isn't already frozen.
2011-11-06 22:28:13 +00:00
if ((ccb->ccb_h.status & CAM_SEL_TIMEOUT) != 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Selection timeout -- discard this LUN if empty. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (li->untagged == NULL && li->used == 0) {
if (lun < NCR_NLUN)
ti->lun[lun] = NULL;
LIST_REMOVE(li, link);
free(li, M_DEVBUF);
}
}
ncr53c9x_free_ecb(sc, ecb);
ti->cmds++;
xpt_done(ccb);
}
static void
ncr53c9x_dequeue(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
{
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
int64_t lun;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
lun = ecb->ccb->ccb_h.target_lun;
li = TINFO_LUN(ti, lun);
#ifdef DIAGNOSTIC
if (li == NULL || li->lun != lun)
Fix build: - Use %ll instead of %q for explicit long long casts; - Use %j instead of %q in XFS and cast to intmax_t. Tested with: make universe
2012-05-23 06:49:50 +00:00
panic("%s: lun %llx for ecb %p does not exist", __func__,
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
(long long)lun, ecb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#endif
if (li->untagged == ecb) {
li->busy = 0;
li->untagged = NULL;
}
if (ecb->tag[0] && li->queued[ecb->tag[1]] != NULL) {
#ifdef DIAGNOSTIC
if (li->queued[ecb->tag[1]] != NULL &&
(li->queued[ecb->tag[1]] != ecb))
Fix build: - Use %ll instead of %q for explicit long long casts; - Use %j instead of %q in XFS and cast to intmax_t. Tested with: make universe
2012-05-23 06:49:50 +00:00
panic("%s: slot %d for lun %llx has %p instead of ecb "
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
"%p", __func__, ecb->tag[1], (long long)lun,
li->queued[ecb->tag[1]], ecb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#endif
li->queued[ecb->tag[1]] = NULL;
li->used--;
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ecb->tag[0] = ecb->tag[1] = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((ecb->flags & ECB_READY) != 0) {
ecb->flags &= ~ECB_READY;
TAILQ_REMOVE(&sc->ready_list, ecb, chain);
}
}
/*
* INTERRUPT/PROTOCOL ENGINE
*/
/*
* Schedule an outgoing message by prioritizing it, and asserting
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* attention on the bus. We can only do this when we are the initiator
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* else there will be an illegal command interrupt.
*/
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#define ncr53c9x_sched_msgout(m) do { \
NCR_MSGS(("ncr53c9x_sched_msgout %x %d", m, __LINE__)); \
NCRCMD(sc, NCRCMD_SETATN); \
sc->sc_flags |= NCR_ATN; \
sc->sc_msgpriq |= (m); \
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} while (/* CONSTCOND */0)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
static void
ncr53c9x_flushfifo(struct ncr53c9x_softc *sc)
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s] ", __func__));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_FLUSH);
if (sc->sc_phase == COMMAND_PHASE ||
sc->sc_phase == MESSAGE_OUT_PHASE)
DELAY(2);
}
static int
ncr53c9x_rdfifo(struct ncr53c9x_softc *sc, int how)
{
int i, n;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t *ibuf;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
switch (how) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_RDFIFO_START:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ibuf = sc->sc_imess;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imlen = 0;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_RDFIFO_CONTINUE:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ibuf = sc->sc_imess + sc->sc_imlen;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
default:
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
panic("%s: bad flag", __func__);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* NOTREACHED */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
* XXX buffer (sc_imess) size for message
*/
n = NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF;
if (sc->sc_rev == NCR_VARIANT_FAS366) {
n *= 2;
for (i = 0; i < n; i++)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ibuf[i] = NCR_READ_REG(sc, NCR_FIFO);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_espstat2 & NCRFAS_STAT2_ISHUTTLE) {
NCR_WRITE_REG(sc, NCR_FIFO, 0);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ibuf[i++] = NCR_READ_REG(sc, NCR_FIFO);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_READ_REG(sc, NCR_FIFO);
ncr53c9x_flushfifo(sc);
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} else
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
for (i = 0; i < n; i++)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ibuf[i] = NCR_READ_REG(sc, NCR_FIFO);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imlen += i;
#if 0
#ifdef NCR53C9X_DEBUG
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
NCR_TRACE(("\n[rdfifo %s (%d):",
(how == NCR_RDFIFO_START) ? "start" : "cont", (int)sc->sc_imlen));
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if ((ncr53c9x_debug & NCR_SHOWTRAC) != 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
for (i = 0; i < sc->sc_imlen; i++)
printf(" %02x", sc->sc_imess[i]);
printf("]\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
#endif
#endif
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
return (sc->sc_imlen);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
static void
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
ncr53c9x_wrfifo(struct ncr53c9x_softc *sc, uint8_t *p, int len)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
int i;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#ifdef NCR53C9X_DEBUG
NCR_MSGS(("[wrfifo(%d):", len));
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if ((ncr53c9x_debug & NCR_SHOWMSGS) != 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
for (i = 0; i < len; i++)
printf(" %02x", p[i]);
printf("]\n");
}
#endif
for (i = 0; i < len; i++) {
NCR_WRITE_REG(sc, NCR_FIFO, p[i]);
if (sc->sc_rev == NCR_VARIANT_FAS366)
NCR_WRITE_REG(sc, NCR_FIFO, 0);
}
}
static int
ncr53c9x_reselect(struct ncr53c9x_softc *sc, int message, int tagtype,
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
int tagid)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
struct ncr53c9x_ecb *ecb = NULL;
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t lun, selid, target;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (sc->sc_rev == NCR_VARIANT_FAS366)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
target = sc->sc_selid;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
else {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* The SCSI chip made a snapshot of the data bus
* while the reselection was being negotiated.
* This enables us to determine which target did
* the reselect.
*/
selid = sc->sc_selid & ~(1 << sc->sc_id);
if (selid & (selid - 1)) {
device_printf(sc->sc_dev, "reselect with invalid "
"selid %02x; sending DEVICE RESET\n", selid);
goto reset;
}
target = ffs(selid) - 1;
}
lun = message & 0x07;
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Search wait queue for disconnected command.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* The list should be short, so I haven't bothered with
* any more sophisticated structures than a simple
* singly linked list.
*/
ti = &sc->sc_tinfo[target];
li = TINFO_LUN(ti, lun);
/*
* We can get as far as the LUN with the IDENTIFY
* message. Check to see if we're running an
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* untagged command. Otherwise ack the IDENTIFY
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* and wait for a tag message.
*/
if (li != NULL) {
if (li->untagged != NULL && li->busy)
ecb = li->untagged;
else if (tagtype != MSG_SIMPLE_Q_TAG) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Wait for tag to come by. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_state = NCR_IDENTIFIED;
return (0);
} else if (tagtype)
ecb = li->queued[tagid];
}
if (ecb == NULL) {
device_printf(sc->sc_dev, "reselect from target %d lun %d "
"tag %x:%x with no nexus; sending ABORT\n",
target, lun, tagtype, tagid);
goto abort;
}
/* Make this nexus active again. */
sc->sc_state = NCR_CONNECTED;
sc->sc_nexus = ecb;
ncr53c9x_setsync(sc, ti);
if (ecb->flags & ECB_RESET)
ncr53c9x_sched_msgout(SEND_DEV_RESET);
else if (ecb->flags & ECB_ABORT)
ncr53c9x_sched_msgout(SEND_ABORT);
/* Do an implicit RESTORE POINTERS. */
sc->sc_dp = ecb->daddr;
sc->sc_dleft = ecb->dleft;
return (0);
reset:
ncr53c9x_sched_msgout(SEND_DEV_RESET);
return (1);
abort:
ncr53c9x_sched_msgout(SEND_ABORT);
return (1);
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* From NetBSD; these should go into CAM at some point. */
#define MSG_ISEXTENDED(m) ((m) == MSG_EXTENDED)
#define MSG_IS1BYTE(m) \
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
((!MSG_ISEXTENDED(m) && (m) < 0x20) || MSG_ISIDENTIFY(m))
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#define MSG_IS2BYTE(m) (((m) & 0xf0) == 0x20)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
static inline int
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
__verify_msg_format(uint8_t *p, int len)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
if (len == 1 && MSG_IS1BYTE(p[0]))
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
return (1);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (len == 2 && MSG_IS2BYTE(p[0]))
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
return (1);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (len >= 3 && MSG_ISEXTENDED(p[0]) &&
len == p[1] + 2)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
return (1);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
return (0);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
* Get an incoming message as initiator.
*
* The SCSI bus must already be in MESSAGE_IN_PHASE and there is a
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* byte in the FIFO.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
static void
ncr53c9x_msgin(struct ncr53c9x_softc *sc)
{
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_ecb *ecb;
struct ncr53c9x_linfo *li;
struct ncr53c9x_tinfo *ti;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t *pb;
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
int len, lun;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s(curmsglen:%ld)] ", __func__, (long)sc->sc_imlen));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_imlen == 0) {
device_printf(sc->sc_dev, "msgin: no msg byte available\n");
return;
}
/*
* Prepare for a new message. A message should (according
* to the SCSI standard) be transmitted in one single
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* MESSAGE_IN_PHASE. If we have been in some other phase,
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* then this is a new message.
*/
if (sc->sc_prevphase != MESSAGE_IN_PHASE &&
sc->sc_state != NCR_RESELECTED) {
device_printf(sc->sc_dev, "phase change, dropping message, "
"prev %d, state %d\n", sc->sc_prevphase, sc->sc_state);
sc->sc_flags &= ~NCR_DROP_MSGI;
sc->sc_imlen = 0;
}
/*
* If we're going to reject the message, don't bother storing
* the incoming bytes. But still, we need to ACK them.
*/
if ((sc->sc_flags & NCR_DROP_MSGI) != 0) {
NCRCMD(sc, NCRCMD_MSGOK);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "<dropping msg byte %x>",
sc->sc_imess[sc->sc_imlen]);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
return;
}
if (sc->sc_imlen >= NCR_MAX_MSG_LEN) {
ncr53c9x_sched_msgout(SEND_REJECT);
sc->sc_flags |= NCR_DROP_MSGI;
} else {
switch (sc->sc_state) {
/*
* if received message is the first of reselection
* then first byte is selid, and then message
*/
case NCR_RESELECTED:
pb = sc->sc_imess + 1;
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
len = sc->sc_imlen - 1;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
default:
pb = sc->sc_imess;
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
len = sc->sc_imlen;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
if (__verify_msg_format(pb, len))
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto gotit;
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Acknowledge what we have so far. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_MSGOK);
return;
gotit:
NCR_MSGS(("gotmsg(%x) state %d", sc->sc_imess[0], sc->sc_state));
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/*
* We got a complete message, flush the imess.
* XXX nobody uses imlen below.
*/
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imlen = 0;
/*
* Now we should have a complete message (1 byte, 2 byte
* and moderately long extended messages). We only handle
* extended messages which total length is shorter than
* NCR_MAX_MSG_LEN. Longer messages will be amputated.
*/
switch (sc->sc_state) {
case NCR_CONNECTED:
ecb = sc->sc_nexus;
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
switch (sc->sc_imess[0]) {
case MSG_CMDCOMPLETE:
NCR_MSGS(("cmdcomplete "));
if (sc->sc_dleft < 0) {
xpt_print_path(ecb->ccb->ccb_h.path);
printf("got %ld extra bytes\n",
-(long)sc->sc_dleft);
sc->sc_dleft = 0;
}
ecb->dleft = (ecb->flags & ECB_TENTATIVE_DONE) ?
0 : sc->sc_dleft;
if ((ecb->flags & ECB_SENSE) == 0)
ecb->ccb->csio.resid = ecb->dleft;
sc->sc_state = NCR_CMDCOMPLETE;
break;
case MSG_MESSAGE_REJECT:
NCR_MSGS(("msg reject (msgout=%x) ", sc->sc_msgout));
switch (sc->sc_msgout) {
case SEND_TAG:
/*
* Target does not like tagged queuing.
* - Flush the command queue
* - Disable tagged queuing for the target
* - Dequeue ecb from the queued array.
*/
device_printf(sc->sc_dev, "tagged queuing "
"rejected: target %d\n",
ecb->ccb->ccb_h.target_id);
NCR_MSGS(("(rejected sent tag)"));
NCRCMD(sc, NCRCMD_FLUSH);
DELAY(1);
ti->flags &= ~T_TAG;
lun = ecb->ccb->ccb_h.target_lun;
li = TINFO_LUN(ti, lun);
if (ecb->tag[0] &&
li->queued[ecb->tag[1]] != NULL) {
li->queued[ecb->tag[1]] = NULL;
li->used--;
}
ecb->tag[0] = ecb->tag[1] = 0;
li->untagged = ecb;
li->busy = 1;
break;
case SEND_SDTR:
device_printf(sc->sc_dev, "sync transfer "
"rejected: target %d\n",
ecb->ccb->ccb_h.target_id);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->flags &= ~T_SDTRSENT;
ti->curr.period = ti->goal.period = 0;
ti->curr.offset = ti->goal.offset = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_setsync(sc, ti);
break;
case SEND_WDTR:
device_printf(sc->sc_dev, "wide transfer "
"rejected: target %d\n",
ecb->ccb->ccb_h.target_id);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->flags &= ~T_WDTRSENT;
ti->curr.width = ti->goal.width =
MSG_EXT_WDTR_BUS_8_BIT;
ncr53c9x_setsync(sc, ti);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
case SEND_INIT_DET_ERR:
goto abort;
}
break;
case MSG_NOOP:
NCR_MSGS(("noop "));
break;
case MSG_HEAD_OF_Q_TAG:
case MSG_SIMPLE_Q_TAG:
case MSG_ORDERED_Q_TAG:
NCR_MSGS(("TAG %x:%x",
sc->sc_imess[0], sc->sc_imess[1]));
break;
case MSG_DISCONNECT:
NCR_MSGS(("disconnect "));
ti->dconns++;
sc->sc_state = NCR_DISCONNECT;
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Mark the fact that all bytes have moved. The
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* target may not bother to do a SAVE POINTERS
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* at this stage. This flag will set the residual
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* count to zero on MSG COMPLETE.
*/
if (sc->sc_dleft == 0)
ecb->flags |= ECB_TENTATIVE_DONE;
break;
case MSG_SAVEDATAPOINTER:
NCR_MSGS(("save datapointer "));
ecb->daddr = sc->sc_dp;
ecb->dleft = sc->sc_dleft;
break;
case MSG_RESTOREPOINTERS:
NCR_MSGS(("restore datapointer "));
sc->sc_dp = ecb->daddr;
sc->sc_dleft = ecb->dleft;
break;
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
case MSG_IGN_WIDE_RESIDUE:
NCR_MSGS(("ignore wide residue (%d bytes)",
sc->sc_imess[1]));
if (sc->sc_imess[1] != 1) {
xpt_print_path(ecb->ccb->ccb_h.path);
printf("unexpected MESSAGE IGNORE WIDE "
"RESIDUE (%d bytes); sending REJECT\n",
sc->sc_imess[1]);
goto reject;
}
/*
* If there was a last transfer of an even number of
* bytes, wipe the "done" memory and adjust by one
* byte (sc->sc_imess[1]).
*/
len = sc->sc_dleft - ecb->dleft;
if (len != 0 && (len & 1) == 0) {
ecb->flags &= ~ECB_TENTATIVE_DONE;
sc->sc_dp = (char *)sc->sc_dp - 1;
sc->sc_dleft--;
}
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case MSG_EXTENDED:
NCR_MSGS(("extended(%x) ", sc->sc_imess[2]));
switch (sc->sc_imess[2]) {
case MSG_EXT_SDTR:
NCR_MSGS(("SDTR period %d, offset %d ",
sc->sc_imess[3], sc->sc_imess[4]));
if (sc->sc_imess[1] != 3)
goto reject;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.period = sc->sc_imess[3];
ti->curr.offset = sc->sc_imess[4];
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_minsync == 0 ||
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.offset == 0 ||
ti->curr.period > 124) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
#if 0
#ifdef NCR53C9X_DEBUG
xpt_print_path(ecb->ccb->ccb_h.path);
printf("async mode\n");
#endif
#endif
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ti->flags & T_SDTRSENT) == 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* target initiated negotiation
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.offset = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_sched_msgout(
SEND_SDTR);
}
} else {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.period =
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ncr53c9x_cpb2stp(sc,
ncr53c9x_stp2cpb(sc,
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.period));
if ((ti->flags & T_SDTRSENT) == 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* target initiated negotiation
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.period <
sc->sc_minsync)
ti->curr.period =
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_minsync;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.offset >
sc->sc_maxoffset)
ti->curr.offset =
sc->sc_maxoffset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_sched_msgout(
SEND_SDTR);
}
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->flags &= ~T_SDTRSENT;
ti->goal.period = ti->curr.period;
ti->goal.offset = ti->curr.offset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_setsync(sc, ti);
break;
case MSG_EXT_WDTR:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_MSGS(("wide mode %d ", sc->sc_imess[3]));
ti->curr.width = sc->sc_imess[3];
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (!(ti->flags & T_WDTRSENT))
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/*
* target initiated negotiation
*/
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_sched_msgout(SEND_WDTR);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->flags &= ~T_WDTRSENT;
ti->goal.width = ti->curr.width;
ncr53c9x_setsync(sc, ti);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
default:
xpt_print_path(ecb->ccb->ccb_h.path);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
printf("unrecognized MESSAGE EXTENDED 0x%x;"
" sending REJECT\n", sc->sc_imess[2]);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto reject;
}
break;
default:
NCR_MSGS(("ident "));
xpt_print_path(ecb->ccb->ccb_h.path);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
printf("unrecognized MESSAGE 0x%x; sending REJECT\n",
sc->sc_imess[0]);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* FALLTHROUGH */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
reject:
ncr53c9x_sched_msgout(SEND_REJECT);
break;
}
break;
case NCR_IDENTIFIED:
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* IDENTIFY message was received and queue tag is expected
* now.
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
*/
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((sc->sc_imess[0] != MSG_SIMPLE_Q_TAG) ||
(sc->sc_msgify == 0)) {
device_printf(sc->sc_dev, "TAG reselect without "
"IDENTIFY; MSG %x; sending DEVICE RESET\n",
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
sc->sc_imess[0]);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto reset;
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
(void)ncr53c9x_reselect(sc, sc->sc_msgify,
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imess[0], sc->sc_imess[1]);
break;
case NCR_RESELECTED:
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (MSG_ISIDENTIFY(sc->sc_imess[1]))
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_msgify = sc->sc_imess[1];
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
else {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "reselect without IDENTIFY;"
" MSG %x; sending DEVICE RESET\n", sc->sc_imess[1]);
goto reset;
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
(void)ncr53c9x_reselect(sc, sc->sc_msgify, 0, 0);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
default:
device_printf(sc->sc_dev, "unexpected MESSAGE IN; "
"sending DEVICE RESET\n");
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* FALLTHROUGH */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
reset:
ncr53c9x_sched_msgout(SEND_DEV_RESET);
break;
abort:
ncr53c9x_sched_msgout(SEND_ABORT);
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* If we have more messages to send set ATN. */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (sc->sc_msgpriq) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_SETATN);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_flags |= NCR_ATN;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Acknowledge last message byte. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_MSGOK);
/* Done, reset message pointer. */
sc->sc_flags &= ~NCR_DROP_MSGI;
sc->sc_imlen = 0;
}
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Send the highest priority, scheduled message.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
static void
ncr53c9x_msgout(struct ncr53c9x_softc *sc)
{
struct ncr53c9x_tinfo *ti;
struct ncr53c9x_ecb *ecb;
size_t size;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
int error;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#ifdef NCR53C9X_DEBUG
int i;
#endif
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
NCR_TRACE(("[%s(priq:%x, prevphase:%x)]", __func__, sc->sc_msgpriq,
sc->sc_prevphase));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* XXX - the NCR_ATN flag is not in sync with the actual ATN
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* condition on the SCSI bus. The 53c9x chip
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* automatically turns off ATN before sending the
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* message byte. (See also the comment below in the
* default case when picking out a message to send.)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
if (sc->sc_flags & NCR_ATN) {
if (sc->sc_prevphase != MESSAGE_OUT_PHASE) {
new:
NCRCMD(sc, NCRCMD_FLUSH);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
#if 0
DELAY(1);
#endif
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_msgoutq = 0;
sc->sc_omlen = 0;
}
} else {
if (sc->sc_prevphase == MESSAGE_OUT_PHASE) {
ncr53c9x_sched_msgout(sc->sc_msgoutq);
goto new;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} else
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "at line %d: unexpected "
"MESSAGE OUT phase\n", __LINE__);
}
if (sc->sc_omlen == 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Pick up highest priority message. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_msgout = sc->sc_msgpriq & -sc->sc_msgpriq;
sc->sc_msgoutq |= sc->sc_msgout;
sc->sc_msgpriq &= ~sc->sc_msgout;
sc->sc_omlen = 1; /* "Default" message len */
switch (sc->sc_msgout) {
case SEND_SDTR:
ecb = sc->sc_nexus;
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
sc->sc_omess[0] = MSG_EXTENDED;
sc->sc_omess[1] = MSG_EXT_SDTR_LEN;
sc->sc_omess[2] = MSG_EXT_SDTR;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_omess[3] = ti->goal.period;
sc->sc_omess[4] = ti->goal.offset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_omlen = 5;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_WDTR:
ecb = sc->sc_nexus;
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
sc->sc_omess[0] = MSG_EXTENDED;
sc->sc_omess[1] = MSG_EXT_WDTR_LEN;
sc->sc_omess[2] = MSG_EXT_WDTR;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_omess[3] = ti->goal.width;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_omlen = 4;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
case SEND_IDENTIFY:
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (sc->sc_state != NCR_CONNECTED)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
device_printf(sc->sc_dev, "at line %d: no "
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
"nexus\n", __LINE__);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ecb = sc->sc_nexus;
sc->sc_omess[0] =
MSG_IDENTIFY(ecb->ccb->ccb_h.target_lun, 0);
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_TAG:
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (sc->sc_state != NCR_CONNECTED)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "at line %d: no "
"nexus\n", __LINE__);
ecb = sc->sc_nexus;
sc->sc_omess[0] = ecb->tag[0];
sc->sc_omess[1] = ecb->tag[1];
sc->sc_omlen = 2;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_DEV_RESET:
sc->sc_flags |= NCR_ABORTING;
sc->sc_omess[0] = MSG_BUS_DEV_RESET;
ecb = sc->sc_nexus;
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->curr.period = 0;
ti->curr.offset = 0;
ti->curr.width = MSG_EXT_WDTR_BUS_8_BIT;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_PARITY_ERROR:
sc->sc_omess[0] = MSG_PARITY_ERROR;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_ABORT:
sc->sc_flags |= NCR_ABORTING;
sc->sc_omess[0] = MSG_ABORT;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_INIT_DET_ERR:
sc->sc_omess[0] = MSG_INITIATOR_DET_ERR;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case SEND_REJECT:
sc->sc_omess[0] = MSG_MESSAGE_REJECT;
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
default:
/*
* We normally do not get here, since the chip
* automatically turns off ATN before the last
* byte of a message is sent to the target.
* However, if the target rejects our (multi-byte)
* message early by switching to MSG IN phase
* ATN remains on, so the target may return to
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* MSG OUT phase. If there are no scheduled messages
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* left we send a NO-OP.
*
* XXX - Note that this leaves no useful purpose for
* the NCR_ATN flag.
*/
sc->sc_flags &= ~NCR_ATN;
sc->sc_omess[0] = MSG_NOOP;
}
sc->sc_omp = sc->sc_omess;
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#ifdef NCR53C9X_DEBUG
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if ((ncr53c9x_debug & NCR_SHOWMSGS) != 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_MSGS(("<msgout:"));
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
for (i = 0; i < sc->sc_omlen; i++)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_MSGS((" %02x", sc->sc_omess[i]));
NCR_MSGS(("> "));
}
#endif
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
if (sc->sc_rev != NCR_VARIANT_FAS366) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* (Re)send the message. */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
size = ulmin(sc->sc_omlen, sc->sc_maxxfer);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
error = NCRDMA_SETUP(sc, &sc->sc_omp, &sc->sc_omlen, 0, &size);
if (error != 0)
goto cmd;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Program the SCSI counter. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_SET_COUNT(sc, size);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Load the count in and start the message-out transfer. */
NCRCMD(sc, NCRCMD_NOP | NCRCMD_DMA);
NCRCMD(sc, NCRCMD_TRANS | NCRCMD_DMA);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRDMA_GO(sc);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cmd:
/*
* XXX FIFO size
*/
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
sc->sc_cmdlen = 0;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
ncr53c9x_flushfifo(sc);
ncr53c9x_wrfifo(sc, sc->sc_omp, sc->sc_omlen);
NCRCMD(sc, NCRCMD_TRANS);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
void
ncr53c9x_intr(void *arg)
{
struct ncr53c9x_softc *sc = arg;
if (!NCRDMA_ISINTR(sc))
return;
NCR_LOCK(sc);
ncr53c9x_intr1(sc);
NCR_UNLOCK(sc);
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* This is the most critical part of the driver, and has to know
* how to deal with *all* error conditions and phases from the SCSI
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* bus. If there are no errors and the DMA was active, then call the
* DMA pseudo-interrupt handler. If this returns 1, then that was it
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* and we can return from here without further processing.
*
* Most of this needs verifying.
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
static void
ncr53c9x_intr1(struct ncr53c9x_softc *sc)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
struct ncr53c9x_ecb *ecb;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_linfo *li;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
struct ncr53c9x_tinfo *ti;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct timeval cur, wait;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
size_t size;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
int error, i, nfifo;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
uint8_t msg;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_INTS(("[ncr53c9x_intr: state %d]", sc->sc_state));
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
again:
/* and what do the registers say... */
ncr53c9x_readregs(sc);
/*
* At the moment, only a SCSI Bus Reset or Illegal
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Command are classed as errors. A disconnect is a
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* valid condition, and we let the code check is the
* "NCR_BUSFREE_OK" flag was set before declaring it
* and error.
*
* Also, the status register tells us about "Gross
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Errors" and "Parity errors". Only the Gross Error
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* is really bad, and the parity errors are dealt
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* with later.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*
* TODO
* If there are too many parity error, go to slow
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* cable mode?
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
if ((sc->sc_espintr & NCRINTR_SBR) != 0) {
if ((NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF) != 0) {
NCRCMD(sc, NCRCMD_FLUSH);
DELAY(1);
}
if (sc->sc_state != NCR_SBR) {
device_printf(sc->sc_dev, "SCSI bus reset\n");
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ncr53c9x_init(sc, 0); /* Restart everything. */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
#if 0
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/*XXX*/ device_printf(sc->sc_dev, "<expected bus reset: "
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
"[intr %x, stat %x, step %d]>\n",
sc->sc_espintr, sc->sc_espstat, sc->sc_espstep);
#endif
if (sc->sc_nexus != NULL)
panic("%s: nexus in reset state",
device_get_nameunit(sc->sc_dev));
goto sched;
}
ecb = sc->sc_nexus;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#define NCRINTR_ERR (NCRINTR_SBR | NCRINTR_ILL)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_espintr & NCRINTR_ERR ||
sc->sc_espstat & NCRSTAT_GE) {
if ((sc->sc_espstat & NCRSTAT_GE) != 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Gross Error; no target? */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF) {
NCRCMD(sc, NCRCMD_FLUSH);
DELAY(1);
}
if (sc->sc_state == NCR_CONNECTED ||
sc->sc_state == NCR_SELECTING) {
ecb->ccb->ccb_h.status = CAM_SEL_TIMEOUT;
ncr53c9x_done(sc, ecb);
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
if ((sc->sc_espintr & NCRINTR_ILL) != 0) {
if ((sc->sc_flags & NCR_EXPECT_ILLCMD) != 0) {
/*
* Eat away "Illegal command" interrupt
* on a ESP100 caused by a re-selection
* while we were trying to select
* another target.
*/
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
#ifdef NCR53C9X_DEBUG
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "ESP100 work-around "
"activated\n");
#endif
sc->sc_flags &= ~NCR_EXPECT_ILLCMD;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Illegal command, out of sync? */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "illegal command: 0x%x "
"(state %d, phase %x, prevphase %x)\n",
sc->sc_lastcmd,
sc->sc_state, sc->sc_phase, sc->sc_prevphase);
if (NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF) {
NCRCMD(sc, NCRCMD_FLUSH);
DELAY(1);
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
}
sc->sc_flags &= ~NCR_EXPECT_ILLCMD;
/*
* Call if DMA is active.
*
* If DMA_INTR returns true, then maybe go 'round the loop
* again in case there is no more DMA queued, but a phase
* change is expected.
*/
if (NCRDMA_ISACTIVE(sc)) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
if (NCRDMA_INTR(sc) == -1) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "DMA error; resetting\n");
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/* If DMA active here, then go back to work... */
if (NCRDMA_ISACTIVE(sc))
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((sc->sc_espstat & NCRSTAT_TC) == 0) {
/*
* DMA not completed. If we can not find a
* acceptable explanation, print a diagnostic.
*/
if (sc->sc_state == NCR_SELECTING)
/*
* This can happen if we are reselected
* while using DMA to select a target.
*/
/*void*/;
else if (sc->sc_prevphase == MESSAGE_OUT_PHASE) {
/*
* Our (multi-byte) message (eg SDTR) was
* interrupted by the target to send
* a MSG REJECT.
* Print diagnostic if current phase
* is not MESSAGE IN.
*/
if (sc->sc_phase != MESSAGE_IN_PHASE)
device_printf(sc->sc_dev,"!TC on MSGOUT"
" [intr %x, stat %x, step %d]"
" prevphase %x, resid %lx\n",
sc->sc_espintr,
sc->sc_espstat,
sc->sc_espstep,
sc->sc_prevphase,
(u_long)sc->sc_omlen);
} else if (sc->sc_dleft == 0) {
/*
* The DMA operation was started for
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* a DATA transfer. Print a diagnostic
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* if the DMA counter and TC bit
* appear to be out of sync.
If we get interrupted during a data phase and the DMA engine is still pumping data despite our scsi data counters being at 0, something has gone massively wrong. The consequence of happily ignoring this is more DMA phase errors and a disk full of spammed sectors. Instead, panic on the first occurance to hopefully limit the damage. MFC After: 3 days
2005-04-22 03:37:10 +00:00
*
* XXX This is fatal and usually means that
* the DMA engine is hopelessly out of
* sync with reality. A disk is likely
* getting spammed at this point.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
device_printf(sc->sc_dev, "!TC on DATA XFER"
" [intr %x, stat %x, step %d]"
" prevphase %x, resid %x\n",
sc->sc_espintr,
sc->sc_espstat,
sc->sc_espstep,
sc->sc_prevphase,
ecb ? ecb->dleft : -1);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
}
}
/*
* Check for less serious errors.
*/
if ((sc->sc_espstat & NCRSTAT_PE) != 0) {
device_printf(sc->sc_dev, "SCSI bus parity error\n");
if (sc->sc_prevphase == MESSAGE_IN_PHASE)
ncr53c9x_sched_msgout(SEND_PARITY_ERROR);
else
ncr53c9x_sched_msgout(SEND_INIT_DET_ERR);
}
if ((sc->sc_espintr & NCRINTR_DIS) != 0) {
sc->sc_msgify = 0;
NCR_INTS(("<DISC [intr %x, stat %x, step %d]>",
sc->sc_espintr,sc->sc_espstat,sc->sc_espstep));
if (NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF) {
NCRCMD(sc, NCRCMD_FLUSH);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
#if 0
DELAY(1);
#endif
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
* This command must (apparently) be issued within
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* 250mS of a disconnect. So here you are...
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
NCRCMD(sc, NCRCMD_ENSEL);
switch (sc->sc_state) {
case NCR_RESELECTED:
goto sched;
case NCR_SELECTING:
ecb->ccb->ccb_h.status = CAM_SEL_TIMEOUT;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Selection timeout -- discard all LUNs if empty. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
li = LIST_FIRST(&ti->luns);
while (li != NULL) {
if (li->untagged == NULL && li->used == 0) {
if (li->lun < NCR_NLUN)
ti->lun[li->lun] = NULL;
LIST_REMOVE(li, link);
free(li, M_DEVBUF);
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Restart the search at the beginning.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
li = LIST_FIRST(&ti->luns);
continue;
}
li = LIST_NEXT(li, link);
}
goto finish;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_CONNECTED:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ecb != NULL) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if ((ti->flags & T_SDTRSENT) != 0) {
xpt_print_path(ecb->ccb->ccb_h.path);
printf("sync nego not completed!\n");
ti->flags &= ~T_SDTRSENT;
ti->curr.period = ti->goal.period = 0;
ti->curr.offset = ti->goal.offset = 0;
ncr53c9x_setsync(sc, ti);
}
if ((ti->flags & T_WDTRSENT) != 0) {
xpt_print_path(ecb->ccb->ccb_h.path);
printf("wide nego not completed!\n");
ti->flags &= ~T_WDTRSENT;
ti->curr.width = ti->goal.width =
MSG_EXT_WDTR_BUS_8_BIT;
ncr53c9x_setsync(sc, ti);
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* It may be OK to disconnect. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((sc->sc_flags & NCR_ABORTING) == 0) {
/*
* Section 5.1.1 of the SCSI 2 spec
* suggests issuing a REQUEST SENSE
* following an unexpected disconnect.
* Some devices go into a contingent
* allegiance condition when
* disconnecting, and this is necessary
* to clean up their state.
*/
device_printf(sc->sc_dev, "unexpected "
"disconnect [state %d, intr %x, stat %x, "
"phase(c %x, p %x)]; ", sc->sc_state,
sc->sc_espintr, sc->sc_espstat,
sc->sc_phase, sc->sc_prevphase);
Apply a torniquet to the problem of the drive unexpectedly disconnecting during a data phase. Before, we would try to recover the autosense, but the DMA engine would still be active with interrupted transfer, and we'd quickly spiral out of control and cause massive data corruption. For now, just reset the chip and cancel everything. The better solution is to cancel the DMA operation, but there is no clear way to do that right now. The data corruption problem is severe enough to warrant this fix in the interim. Thanks to Kris Kenneway to sacrificing countless filesystems to this bug. MFC After: 3 days
2005-04-25 22:11:43 +00:00
/*
* XXX This will cause a chip reset and will
* prevent us from finding out the real
* problem with the device. However, it's
Correct spelling in comments. Submitted by: brucec
2011-04-22 13:56:21 +00:00
* necessary until a way can be found to
Apply a torniquet to the problem of the drive unexpectedly disconnecting during a data phase. Before, we would try to recover the autosense, but the DMA engine would still be active with interrupted transfer, and we'd quickly spiral out of control and cause massive data corruption. For now, just reset the chip and cancel everything. The better solution is to cancel the DMA operation, but there is no clear way to do that right now. The data corruption problem is severe enough to warrant this fix in the interim. Thanks to Kris Kenneway to sacrificing countless filesystems to this bug. MFC After: 3 days
2005-04-25 22:11:43 +00:00
* safely cancel the DMA that is in
* progress.
*/
if (1 || (ecb->flags & ECB_SENSE) != 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
printf("resetting\n");
goto reset;
}
printf("sending REQUEST SENSE\n");
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_stop(&ecb->ch);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_sense(sc, ecb);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
} else if (ecb != NULL &&
(ecb->flags & ECB_RESET) != 0) {
ecb->ccb->ccb_h.status = CAM_REQ_CMP;
goto finish;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
ecb->ccb->ccb_h.status = CAM_CMD_TIMEOUT;
goto finish;
case NCR_DISCONNECT:
sc->sc_nexus = NULL;
goto sched;
case NCR_CMDCOMPLETE:
ecb->ccb->ccb_h.status = CAM_REQ_CMP;
goto finish;
}
}
switch (sc->sc_state) {
case NCR_SBR:
device_printf(sc->sc_dev, "waiting for Bus Reset to happen\n");
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case NCR_RESELECTED:
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* We must be continuing a message?
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
device_printf(sc->sc_dev, "unhandled reselect continuation, "
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
"state %d, intr %02x\n", sc->sc_state, sc->sc_espintr);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
case NCR_IDENTIFIED:
ecb = sc->sc_nexus;
if (sc->sc_phase != MESSAGE_IN_PHASE) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
i = NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF;
/*
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
* Things are seriously screwed up.
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Pull the brakes, i.e. reset.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
device_printf(sc->sc_dev, "target didn't send tag: %d "
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
"bytes in FIFO\n", i);
/* Drain and display FIFO. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
while (i-- > 0)
printf("[%d] ", NCR_READ_REG(sc, NCR_FIFO));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
} else
goto msgin;
case NCR_IDLE:
case NCR_SELECTING:
ecb = sc->sc_nexus;
if (sc->sc_espintr & NCRINTR_RESEL) {
sc->sc_msgpriq = sc->sc_msgout = sc->sc_msgoutq = 0;
sc->sc_flags = 0;
/*
* If we're trying to select a
* target ourselves, push our command
* back into the ready list.
*/
if (sc->sc_state == NCR_SELECTING) {
NCR_INTS(("backoff selector "));
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_stop(&ecb->ch);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_dequeue(sc, ecb);
TAILQ_INSERT_HEAD(&sc->ready_list, ecb, chain);
ecb->flags |= ECB_READY;
ecb = sc->sc_nexus = NULL;
}
sc->sc_state = NCR_RESELECTED;
if (sc->sc_phase != MESSAGE_IN_PHASE) {
/*
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
* Things are seriously screwed up.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* Pull the brakes, i.e. reset
*/
device_printf(sc->sc_dev, "target didn't "
"identify\n");
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
/*
* The C90 only inhibits FIFO writes until reselection
Add some style(9) touch ups; style(9) states that new code should follow these conventions and, well, this is a new driver. Tested on: i386, sparc64 Reviewed by: scottl
2004-08-02 22:55:23 +00:00
* is complete instead of waiting until the interrupt
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* status register has been read. So, if the reselect
* happens while we were entering command bytes (for
* another target) some of those bytes can appear in
* the FIFO here, after the interrupt is taken.
*
* To remedy this situation, pull the Selection ID
* and Identify message from the FIFO directly, and
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* ignore any extraneous FIFO contents. Also, set
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* a flag that allows one Illegal Command Interrupt
* to occur which the chip also generates as a result
* of writing to the FIFO during a reselect.
*/
if (sc->sc_rev == NCR_VARIANT_ESP100) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
nfifo =
NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_imess[0] = NCR_READ_REG(sc, NCR_FIFO);
sc->sc_imess[1] = NCR_READ_REG(sc, NCR_FIFO);
sc->sc_imlen = 2;
if (nfifo != 2) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Flush the rest. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_FLUSH);
}
sc->sc_flags |= NCR_EXPECT_ILLCMD;
if (nfifo > 2)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
nfifo = 2; /* We fixed it... */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
} else
nfifo = ncr53c9x_rdfifo(sc, NCR_RDFIFO_START);
if (nfifo != 2) {
device_printf(sc->sc_dev, "RESELECT: %d bytes "
"in FIFO! [intr %x, stat %x, step %d, "
"prevphase %x]\n",
nfifo,
sc->sc_espintr,
sc->sc_espstat,
sc->sc_espstep,
sc->sc_prevphase);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
sc->sc_selid = sc->sc_imess[0];
NCR_INTS(("selid=%02x ", sc->sc_selid));
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Handle IDENTIFY message. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_msgin(sc);
if (sc->sc_state != NCR_CONNECTED &&
sc->sc_state != NCR_IDENTIFIED) {
/* IDENTIFY fail?! */
device_printf(sc->sc_dev, "identify failed, "
"state %d, intr %02x\n", sc->sc_state,
sc->sc_espintr);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
goto reset;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
goto shortcut; /* i.e. next phase expected soon */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
#define NCRINTR_DONE (NCRINTR_FC | NCRINTR_BS)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((sc->sc_espintr & NCRINTR_DONE) == NCRINTR_DONE) {
/*
* Arbitration won; examine the `step' register
* to determine how far the selection could progress.
*/
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
if (ecb == NULL) {
/*
* When doing path inquiry during boot
* FAS100A trigger a stray interrupt which
* we just ignore instead of panicing.
*/
if (sc->sc_state == NCR_IDLE &&
sc->sc_espstep == 0)
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
panic("%s: no nexus", __func__);
- Try to not leak resources in the attach functions of the esp(4) SBus front-end and the LSI64854 and NCR53C9x code in case one of these functions fails. Add detach functions to these parts and make esp(4) detachable. - Revert rev. 1.7 of esp_sbus.c, since rev. 1.34 of sbus.c the clockfreq IVAR defaults to the per-child values. - Merge ncr53c9x.c rev. 1.111 from NetBSD (partial): On reset, clear state flags and the msgout queue. In NetBSD code to notify the upper layer (i.e. CAM in FreeBSD) on reset was also added with this revision. This is believed to be not necessary in FreeBSD and was not merged. This makes ncr53c9x.c to be in sync with NetBSD up to rev. 1.114. - Conditionalize the LSI64854 support on sbus(4) only instead of sbus(4) and esp(4) as it's also required for the 'dma', 'espdma' and 'ledma' busses/devices as well as the 'SUNW,bpp' device (printer port) which all hang off of sbus(4). - Add a driver for the 'dma', 'espdma' and 'ledma' (pseudo-)busses/ devices. These busses and devices actually represent the LSI64854 DMA engines for the ESP SCSI and LANCE Ethernet controllers found on the SBus of Ultra 1 and SBus add-on cards. With 'espdma' and 'ledma' the 'esp' and 'le' devices hang off of the respective DMA bus instead of directly from the SBus. The 'dma' devices are either also used in this manner or on some add-on cards also as a companion device to an 'esp' device which also hangs off directly from the SBus. With the latter variant it's a bit tricky to glue the DMA engine to the core logic of the respective 'esp' device. With rev. 1.35 of sbus.c we are however guaranteed that such a 'dma' device is probed before the respective 'esp' device which simplifies things a lot. [1] - In the esp(4) SBus front-end read the part-unique ID code of Fast-SCSI capable chips the right way. This fixes erroneously detecting some chips as FAS366 when in fact they are not. Add explicit checks for the FAS100A, FAS216 and FAS236 variants instead treating all of these as ESP200. That way we can correctly set the respective Fast-SCSI config bits instead of driving them out of specs. This includes adding the FAS100A and FAS236 variants to the NCR53C9x core code. We probably still subsume some chip variants as ESP200 while in fact they are another variant which however shouldn't really matter as this will only happen when these chips are driven at 25MHz or less which implies not being able to run Fast-SCSI. [3] - Add a workaround to the NCR53C9x interrupt handler which ignores the stray interrupt generated by FAS100A when doing path inquiry during boot and which otherwiese would trigger a panic. - Add support for the 'esp' devices hanging off of a 'dma' or 'espdma' busses or which are companions of 'dma' devices to esp(4). In case of the variants that hang off of a DMA device this is a bit hackish as esp(4) then directly uses the softc of the respective parent to talk to the DMA engine. It might make sense to add an interface for this in order to implement this in a cleaner way however it's not yet clear how the requirements for the LANCE Ethernet controllers are and the hack works for now. [2] This effectively adds support for the onboard SCSI controller in Ultra 1 as well as most of the ESP-based SBus add-on cards to esp(4). With this the code for supporting the Performance Technologies SBS430 SBus SCSI add-on cards is also largely in place the remaining bits were however omitted as it's unclear from the NetBSD how to couple the DMA engine and the core logic together for these cards. Obtained from: OpenBSD [1] Obtained from: NetBSD [2] Clue from: BSD/OS [3] Reviewed by: scottl (earlier version) Tested with: FSBE/S add-on card (FAS236), SSHA add-on card (ESP100A), Ultra 1 (onboard FAS100A), Ultra 2 (onboard FAS366)
2005-05-19 14:51:10 +00:00
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ti = &sc->sc_tinfo[ecb->ccb->ccb_h.target_id];
switch (sc->sc_espstep) {
case 0:
/*
* The target did not respond with a
* message out phase - probably an old
* device that doesn't recognize ATN.
* Clear ATN and just continue, the
* target should be in the command
* phase.
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* XXX check for command phase?
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
NCRCMD(sc, NCRCMD_RSTATN);
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case 1:
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.period == ti->goal.period &&
ti->curr.offset == ti->goal.offset &&
ti->curr.width == ti->goal.width &&
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->tag[0] == 0) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "step 1 "
"and no negotiation to perform "
"or tag to send\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto reset;
}
if (sc->sc_phase != MESSAGE_OUT_PHASE) {
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
device_printf(sc->sc_dev, "step 1 "
"but not in MESSAGE_OUT_PHASE\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto reset;
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_prevphase = MESSAGE_OUT_PHASE; /* XXX */
if (ecb->flags & ECB_RESET) {
/*
* A DEVICE RESET was scheduled and
* ATNS used. As SEND_DEV_RESET has
* the highest priority, the target
* will reset and disconnect and we
* will end up in ncr53c9x_done w/o
* negotiating or sending a TAG. So
* we just break here in order to
* avoid warnings about negotiation
* not having completed.
*/
ncr53c9x_sched_msgout(SEND_DEV_RESET);
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
if (ti->curr.width != ti->goal.width) {
ti->flags |= T_WDTRSENT | T_SDTRSENT;
ncr53c9x_sched_msgout(SEND_WDTR |
SEND_SDTR);
}
if (ti->curr.period != ti->goal.period ||
ti->curr.offset != ti->goal.offset) {
ti->flags |= T_SDTRSENT;
ncr53c9x_sched_msgout(SEND_SDTR);
}
if (ecb->tag[0] != 0)
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Could not do ATN3 so send TAG. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ncr53c9x_sched_msgout(SEND_TAG);
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case 3:
/*
* Grr, this is supposed to mean
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* "target left command phase prematurely".
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* It seems to happen regularly when
* sync mode is on.
* Look at FIFO to see if command went out.
* (Timing problems?)
*/
if (sc->sc_features & NCR_F_DMASELECT) {
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
if (sc->sc_cmdlen == 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Hope for the best... */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
}
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} else if ((NCR_READ_REG(sc, NCR_FFLAG) &
NCRFIFO_FF) == 0) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Hope for the best... */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
xpt_print_path(ecb->ccb->ccb_h.path);
printf("selection failed; %d left in FIFO "
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
"[intr %x, stat %x, step %d]\n",
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF,
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_espintr, sc->sc_espstat,
sc->sc_espstep);
NCRCMD(sc, NCRCMD_FLUSH);
ncr53c9x_sched_msgout(SEND_ABORT);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case 2:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Select stuck at Command Phase. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_FLUSH);
break;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case 4:
if (sc->sc_features & NCR_F_DMASELECT &&
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
sc->sc_cmdlen != 0) {
xpt_print_path(ecb->ccb->ccb_h.path);
printf("select; %lu left in DMA buffer "
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
"[intr %x, stat %x, step %d]\n",
(u_long)sc->sc_cmdlen,
sc->sc_espintr,
sc->sc_espstat,
sc->sc_espstep);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* So far, everything went fine. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
break;
}
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
sc->sc_prevphase = INVALID_PHASE; /* ??? */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/* Do an implicit RESTORE POINTERS. */
sc->sc_dp = ecb->daddr;
sc->sc_dleft = ecb->dleft;
sc->sc_state = NCR_CONNECTED;
break;
} else {
device_printf(sc->sc_dev, "unexpected status after "
"select: [intr %x, stat %x, step %x]\n",
sc->sc_espintr, sc->sc_espstat, sc->sc_espstep);
NCRCMD(sc, NCRCMD_FLUSH);
DELAY(1);
goto reset;
}
if (sc->sc_state == NCR_IDLE) {
device_printf(sc->sc_dev, "stray interrupt\n");
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
break;
case NCR_CONNECTED:
if ((sc->sc_flags & NCR_ICCS) != 0) {
/* "Initiate Command Complete Steps" in progress */
sc->sc_flags &= ~NCR_ICCS;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if ((sc->sc_espintr & NCRINTR_DONE) == 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "ICCS: "
": [intr %x, stat %x, step %x]\n",
sc->sc_espintr, sc->sc_espstat,
sc->sc_espstep);
}
ncr53c9x_rdfifo(sc, NCR_RDFIFO_START);
if (sc->sc_imlen < 2)
device_printf(sc->sc_dev, "can't get status, "
"only %d bytes\n", (int)sc->sc_imlen);
ecb->stat = sc->sc_imess[sc->sc_imlen - 2];
msg = sc->sc_imess[sc->sc_imlen - 1];
NCR_PHASE(("<stat:(%x,%x)>", ecb->stat, msg));
if (msg == MSG_CMDCOMPLETE) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ecb->dleft =
(ecb->flags & ECB_TENTATIVE_DONE) ?
0 : sc->sc_dleft;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if ((ecb->flags & ECB_SENSE) == 0)
ecb->ccb->csio.resid = ecb->dleft;
sc->sc_state = NCR_CMDCOMPLETE;
} else
device_printf(sc->sc_dev, "STATUS_PHASE: "
"msg %d\n", msg);
sc->sc_imlen = 0;
NCRCMD(sc, NCRCMD_MSGOK);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
goto shortcut; /* i.e. wait for disconnect */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
break;
default:
device_printf(sc->sc_dev, "invalid state: %d [intr %x, "
"phase(c %x, p %x)]\n", sc->sc_state,
sc->sc_espintr, sc->sc_phase, sc->sc_prevphase);
goto reset;
}
/*
* Driver is now in state NCR_CONNECTED, i.e. we
* have a current command working the SCSI bus.
*/
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
if (sc->sc_state != NCR_CONNECTED || ecb == NULL)
panic("%s: no nexus", __func__);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
switch (sc->sc_phase) {
case MESSAGE_OUT_PHASE:
NCR_PHASE(("MESSAGE_OUT_PHASE "));
ncr53c9x_msgout(sc);
sc->sc_prevphase = MESSAGE_OUT_PHASE;
break;
case MESSAGE_IN_PHASE:
msgin:
NCR_PHASE(("MESSAGE_IN_PHASE "));
if ((sc->sc_espintr & NCRINTR_BS) != 0) {
if ((sc->sc_rev != NCR_VARIANT_FAS366) ||
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
(sc->sc_espstat2 & NCRFAS_STAT2_EMPTY) == 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_FLUSH);
}
sc->sc_flags |= NCR_WAITI;
NCRCMD(sc, NCRCMD_TRANS);
} else if ((sc->sc_espintr & NCRINTR_FC) != 0) {
if ((sc->sc_flags & NCR_WAITI) == 0) {
device_printf(sc->sc_dev, "MSGIN: unexpected "
"FC bit: [intr %x, stat %x, step %x]\n",
sc->sc_espintr, sc->sc_espstat,
sc->sc_espstep);
}
sc->sc_flags &= ~NCR_WAITI;
ncr53c9x_rdfifo(sc,
Merge some style and minor changes from NetBSD: - ncr53c9x.c: 1.108: Remove unreachable break after return and goto statements. 1.109: avoid strong words; use 'screw' instead 1.110: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.114: nuke trailing whitespace 1.107 was already merged, 1.112 and 1.113 are not relevant for FreeBSD. 1.111 is a functional change and will be merged later. - ncr53c9xreg.h: 1.12: DMA, not dma nor Dma. 1.13: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.14: nuke trailing whitespace - ncr53c9xvar.h: 1.43: Fix some typos. From Tom Cosgrove via jmc@openbsd. 1.44: Constify. 1.42 and 1.46 were already merged, 1.45 is not relevant for FreeBSD.
2005-04-17 17:44:01 +00:00
(sc->sc_prevphase == sc->sc_phase) ?
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_RDFIFO_CONTINUE : NCR_RDFIFO_START);
ncr53c9x_msgin(sc);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
} else
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
device_printf(sc->sc_dev, "MSGIN: weird bits: "
"[intr %x, stat %x, step %x]\n",
sc->sc_espintr, sc->sc_espstat, sc->sc_espstep);
sc->sc_prevphase = MESSAGE_IN_PHASE;
goto shortcut; /* i.e. expect data to be ready */
case COMMAND_PHASE:
/*
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Send the command block. Normally we don't see this
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* phase because the SEL_ATN command takes care of
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* all this. However, we end up here if either the
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* target or we wanted to exchange some more messages
* first (e.g. to start negotiations).
*/
NCR_PHASE(("COMMAND_PHASE 0x%02x (%d) ",
ecb->cmd.cmd.opcode, ecb->clen));
if (NCR_READ_REG(sc, NCR_FFLAG) & NCRFIFO_FF) {
NCRCMD(sc, NCRCMD_FLUSH);
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
#if 0
DELAY(1);
#endif
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
/*
* If we have more messages to send, e.g. WDTR or SDTR
* after we've sent a TAG, set ATN so we'll go back to
* MESSAGE_OUT_PHASE.
*/
if (sc->sc_msgpriq) {
NCRCMD(sc, NCRCMD_SETATN);
sc->sc_flags |= NCR_ATN;
}
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_features & NCR_F_DMASELECT) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Setup DMA transfer for command. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
size = ecb->clen;
sc->sc_cmdlen = size;
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
sc->sc_cmdp = (void *)&ecb->cmd.cmd;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
error = NCRDMA_SETUP(sc, &sc->sc_cmdp, &sc->sc_cmdlen,
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
0, &size);
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
if (error != 0)
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
goto cmd;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Program the SCSI counter. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_SET_COUNT(sc, size);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Load the count in. */
NCRCMD(sc, NCRCMD_NOP | NCRCMD_DMA);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Start the command transfer. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_TRANS | NCRCMD_DMA);
NCRDMA_GO(sc);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
sc->sc_prevphase = COMMAND_PHASE;
break;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
cmd:
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
sc->sc_cmdlen = 0;
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
ncr53c9x_wrfifo(sc, (uint8_t *)&ecb->cmd.cmd, ecb->clen);
NCRCMD(sc, NCRCMD_TRANS);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sc->sc_prevphase = COMMAND_PHASE;
break;
case DATA_OUT_PHASE:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
NCR_PHASE(("DATA_OUT_PHASE [%ld] ", (long)sc->sc_dleft));
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
sc->sc_prevphase = DATA_OUT_PHASE;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCRCMD(sc, NCRCMD_FLUSH);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
size = ulmin(sc->sc_dleft, sc->sc_maxxfer);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
error = NCRDMA_SETUP(sc, &sc->sc_dp, &sc->sc_dleft, 0, &size);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto setup_xfer;
case DATA_IN_PHASE:
NCR_PHASE(("DATA_IN_PHASE "));
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
sc->sc_prevphase = DATA_IN_PHASE;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
if (sc->sc_rev == NCR_VARIANT_ESP100)
NCRCMD(sc, NCRCMD_FLUSH);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
size = ulmin(sc->sc_dleft, sc->sc_maxxfer);
- Use device_t rather than the NetBSDish struct device. - Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends to use it. - Use KOBJMETHOD_END. - Remove the gl_clear_latched_intr hook as it's not needed for any of the chips nor the front-ends supported in FreeBSD and likely never will be. - Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't limited to 32-bit DMA. - The ESP200 also only supports up to 64k transfers. - Don't let the DMA and SBus front-end supply a maximum transfer size larger than MAXPHYS as that's the maximum the upper layers use and we otherwise just waste resources unnecessarily. - Initialize the ECB callout and don't zero the handle when returning ECBs to the free list so that ncr53c9x_callout() actually is called with the driver lock held. - On detach the driver lock should be held across cam_sim_free() according to isp(4) and a panic received. - Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try to handle failures gracefully. - In ncr53c9x_action() replace N calls to xpt_done() in a switch with just one at the end. - On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former is somewhat more correct as well as the maximum supported transfer size via maxio in order to take advantage of controllers that that can handle more than DFLTPHYS. - Print the number of MESSAGE (EXTENDED) rejected. - Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h. - Correct the DMA constraints used in the LSI64854 core to not exceed the maximum supported transfer size and include the boundary so we don't need to check on every setup of a DMA transfer. - Let the bus DMA map callbacks do nothing in case of an error. - Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature flag NCR_F_LARGEXFER was introduced so we just need to check for this one and not for individual controllers supporting large transfers in several places. - Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers support, sparc64 doesn't actually use EINPROGRESS and likely never will, as an example for writing additional front-ends for the NCR53C9x core it makes sense to set BUS_DMA_NOWAIT anyway though. - Some minor cleanup.
2011-10-30 21:17:42 +00:00
error = NCRDMA_SETUP(sc, &sc->sc_dp, &sc->sc_dleft, 1, &size);
setup_xfer:
if (error != 0) {
switch (error) {
case EFBIG:
ecb->ccb->ccb_h.status |= CAM_REQ_TOO_BIG;
break;
case EINPROGRESS:
panic("%s: cannot deal with deferred DMA",
__func__);
case EINVAL:
ecb->ccb->ccb_h.status |= CAM_REQ_INVALID;
break;
case ENOMEM:
ecb->ccb->ccb_h.status |= CAM_REQUEUE_REQ;
break;
default:
ecb->ccb->ccb_h.status |= CAM_REQ_CMP_ERR;
}
goto finish;
}
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
/* Target returned to data phase: wipe "done" memory. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
ecb->flags &= ~ECB_TENTATIVE_DONE;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Program the SCSI counter. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
NCR_SET_COUNT(sc, size);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Load the count in. */
NCRCMD(sc, NCRCMD_NOP | NCRCMD_DMA);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/*
* Note that if `size' is 0, we've already transceived
* all the bytes we want but we're still in DATA PHASE.
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* Apparently, the device needs padding. Also, a
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* transfer size of 0 means "maximum" to the chip
* DMA logic.
*/
NCRCMD(sc,
(size == 0 ? NCRCMD_TRPAD : NCRCMD_TRANS) | NCRCMD_DMA);
NCRDMA_GO(sc);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
case STATUS_PHASE:
NCR_PHASE(("STATUS_PHASE "));
sc->sc_flags |= NCR_ICCS;
NCRCMD(sc, NCRCMD_ICCS);
sc->sc_prevphase = STATUS_PHASE;
goto shortcut; /* i.e. expect status results soon */
case INVALID_PHASE:
break;
default:
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
device_printf(sc->sc_dev,
"unexpected bus phase; resetting\n");
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
goto reset;
}
return;
reset:
ncr53c9x_init(sc, 1);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
finish:
ncr53c9x_done(sc, ecb);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
sched:
sc->sc_state = NCR_IDLE;
ncr53c9x_sched(sc);
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
return;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
shortcut:
/*
* The idea is that many of the SCSI operations take very little
* time, and going away and getting interrupted is too high an
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
* overhead to pay. For example, selecting, sending a message
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
* and command and then doing some work can be done in one "pass".
*
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
* The delay is a heuristic. It is 2 when at 20 MHz, 2 at 25 MHz and
* 1 at 40 MHz. This needs testing.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
microtime(&wait);
wait.tv_usec += 50 / sc->sc_freq;
if (wait.tv_usec > 1000000) {
wait.tv_sec++;
wait.tv_usec -= 1000000;
}
do {
if (NCRDMA_ISINTR(sc))
goto again;
microtime(&cur);
} while (cur.tv_sec <= wait.tv_sec && cur.tv_usec <= wait.tv_usec);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
}
static void
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
ncr53c9x_abort(struct ncr53c9x_softc *sc, struct ncr53c9x_ecb *ecb)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/* 2 secs for the abort */
ecb->timeout = NCR_ABORT_TIMEOUT;
ecb->flags |= ECB_ABORT;
if (ecb == sc->sc_nexus) {
/*
* If we're still selecting, the message will be scheduled
* after selection is complete.
*/
if (sc->sc_state == NCR_CONNECTED)
ncr53c9x_sched_msgout(SEND_ABORT);
/*
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
* Reschedule callout.
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
*/
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
callout_reset(&ecb->ch, mstohz(ecb->timeout),
ncr53c9x_callout, ecb);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
} else {
/*
* Just leave the command where it is.
* XXX - what choice do we have but to reset the SCSI
* eventually?
*/
if (sc->sc_state == NCR_IDLE)
ncr53c9x_sched(sc);
}
}
static void
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ncr53c9x_callout(void *arg)
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
{
struct ncr53c9x_ecb *ecb = arg;
union ccb *ccb = ecb->ccb;
struct ncr53c9x_softc *sc = ecb->sc;
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
struct ncr53c9x_tinfo *ti;
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti = &sc->sc_tinfo[ccb->ccb_h.target_id];
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
xpt_print_path(ccb->ccb_h.path);
device_printf(sc->sc_dev, "timed out [ecb %p (flags 0x%x, dleft %x, "
"stat %x)], <state %d, nexus %p, phase(l %x, c %x, p %x), "
"resid %lx, msg(q %x,o %x) %s>",
ecb, ecb->flags, ecb->dleft, ecb->stat,
sc->sc_state, sc->sc_nexus,
NCR_READ_REG(sc, NCR_STAT),
sc->sc_phase, sc->sc_prevphase,
(long)sc->sc_dleft, sc->sc_msgpriq, sc->sc_msgout,
NCRDMA_ISACTIVE(sc) ? "DMA active" : "");
Fix -Wundef from compiling the amd64 LINT.
2005-12-04 10:06:06 +00:00
#if defined(NCR53C9X_DEBUG) && NCR53C9X_DEBUG > 1
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
printf("TRACE: %s.", ecb->trace);
#endif
if (ecb->flags & ECB_ABORT) {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Abort timed out. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
printf(" AGAIN\n");
ncr53c9x_init(sc, 1);
} else {
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Abort the operation that has timed out. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
printf("\n");
ccb->ccb_h.status = CAM_CMD_TIMEOUT;
ncr53c9x_abort(sc, ecb);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Disable sync mode if stuck in a data phase. */
- Merge from NetBSD: When issuing a non-DMA command, make sure to set the "remaining length of command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise we might get confused on command competition interrupt (no DMA active but still data left to transfer). - Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as just rejecting these leads to a resend and disconnect loop. Reported and tested by: mjacob MFC after: 3 days
2012-08-06 08:58:54 +00:00
if (ecb == sc->sc_nexus && ti->curr.offset != 0 &&
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
(sc->sc_phase & (MSGI | CDI)) == 0) {
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/* XXX ASYNC CALLBACK! */
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
ti->goal.offset = 0;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
xpt_print_path(ccb->ccb_h.path);
printf("sync negotiation disabled\n");
}
}
}
static void
ncr53c9x_watch(void *arg)
{
Merge from NetBSD: - Remove clause 3 and 4 from TNF licenses. - Fix memset usage. - Various cleanup. - Kill caddr_t.
2011-10-15 09:29:43 +00:00
struct ncr53c9x_softc *sc = arg;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
struct ncr53c9x_linfo *li;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
struct ncr53c9x_tinfo *ti;
time_t old;
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
int t;
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
o Move the MODULE_DEPEND() for cam(4) from the esp_sbus.c front-end to the ncr53c9x.c core where it actually belongs so future front-ends don't need to add it. o Use the correct OFW property when looking for the initiator ID of the SBus device. o Don't specify an alignment when creating the parent DMA tag for SUNW,fas; their DMA engine doesn't require an alignment constraint and it's no inherited by the child DMA tags anyway (which probably is a bug though). o Drop the superfluous sc_maxsync and use sc_minsync instead. The former apparently was added due to a confusion with the maximum frequency used in cam(4), which basically corresponds to the inverse of minimum sync period. o Merge ncr53c9x.c from NetBSD: 1.116: NCRDMA_SETUP() should be called before NCR_SET_COUNT() and NCRCMD_DMA command in ncr53c9x_select(). 1.125: free allocated resources on detach. o Static'ize ncr53c9x_action(), ncr53c9x_init() and ncr53c9x_reset() as these are not required outside of ncr53c9x.c. o In ncr53c9x_attach() don't leak the device mutex in case attaching fails. o Register an asynchronous notification handler so in case cam(4) reports a lost device we can cancel outstanding commands and restore the default parameters for the target in question. o For FAS366 correctly support 16-bit target IDs and let it know that we use 32-bit transfers. o Overhaul the negotiation of transfer settings. This includes distinguishing between current and goal transfer settings of the target so we can renegotiate their goal settings when necessary and correcting the order in which tagged, wide and synchronous transfers are negotiated. o If we are requesting sense, force a renegotiation if we are currently using anything different from asynchronous at 8 bit as the target might have lost our transfer negotiations. o In case of an XPT_RESET_BUS just directly call ncr53c9x_init() instead of issuing a NCRCMD_RSTSCSI, which in turn will issue an interrupt that is treated as an unexpected SCSI bus reset by ncr53c9x_intr() and thus calls ncr53c9x_init(). Remove the now no longer used ncr53c9x_scsi_reset(). o Correct an off-by-one error when setting cpi->max_lun. o In replace printf(9) with device_printf(9) calls where appropriate and in ncr53c9x_action() remove some unnecessarily verbose messages. o In ncr53c9x_sched() use TAILQ_FOREACH() instead of reimplementing it and consolidate two tagging-related target info checks into one. o In ncr53c9x_done() set the CAM status to CAM_SCSI_STATUS_ERROR when appropriate, respect CAM_DIS_AUTOSENSE and teach it to return SCSI status information. o In ncr53c9x_dequeue() ensure the tags are cleared. o Use ulmin() instead of min() where appropriate. o In ncr53c9x_msgout() consistently use the reset label. o When we're interrupted during a data phase and the DMA engine is still active, don't panic but reset the core and the DMA engine as this should be sufficient. Also, the typical problem for triggering this was the lack of renegotiation when requesting sense. o Correctly handle DEVICE RESETs. o Adapt the locking of esp(4) to MPSAFE cam(4). This includes moving the calls of lsi64854_attach() to the bus front-ends so it can pass the esp(4) mutex to bus_dma_tag_create(9). o Change the LSI64854 driver to not create a DMA tag and map for the Ethernet channel as le(4) will handle these on its own as well as sync and unload the DMA maps for the SCSI and parallel port channel after a DMA transfer. o Cam(4)'ify some NetBSD-centric comments. o Use bus_{read,write}_*(9) instead of bus_space_{read,write}_*(9) and take advantage of rman_get_rid(9) in order to save some softc members. Reviewed by: scottl MFC after: 1 month
2008-09-08 20:20:44 +00:00
NCR_LOCK_ASSERT(sc, MA_OWNED);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
/* Delete any structures that have not been used in 10min. */
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
old = time_second - (10 * 60);
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
for (t = 0; t < sc->sc_ntarg; t++) {
ti = &sc->sc_tinfo[t];
li = LIST_FIRST(&ti->luns);
while (li) {
if (li->last_used < old &&
li->untagged == NULL &&
li->used == 0) {
if (li->lun < NCR_NLUN)
ti->lun[li->lun] = NULL;
LIST_REMOVE(li, link);
free(li, M_DEVBUF);
o Fix style(9) bugs and similar nits. o Merge ncr53c9x.c from NetBSD: 1.115: fix variable shadowing 1.118: __inline -> inline 1.121: fix empty if
2008-07-22 13:51:21 +00:00
/* Restart the search at the beginning. */
Port the NetBSD esp(4) driver. This only includes the sbus front-end, so its primary use is for the FEPS/FAS366 SCSI found in Sun Ultra 1e and 2 machines. Once the pci front-end is ported, this driver can replace the amd(4) driver. The code as-is is fairly stable. I've disabled tagged-queueing until I can figure out a corruption bug related to it. I'm importing it now so that people with these machines can (finally) stop netbooting and report bugs before 5.3.
2004-06-10 05:11:39 +00:00
li = LIST_FIRST(&ti->luns);
continue;
}
li = LIST_NEXT(li, link);
}
}
callout_reset(&sc->sc_watchdog, 60 * hz, ncr53c9x_watch, sc);
}
Reference in a new issue Copy permalink