With this feature, the DASD device driver more robustly handles DASDs
that are attached via multiple channel paths and are subject to
constant Interface-Control-Checks (IFCCs) and Channel-Control-Checks
(CCCs) or loss of High-Performance-FICON (HPF) functionality on one or
more of these paths.
If a channel path does not work correctly, it is removed from normal
operation as long as other channel paths are available. All extended
error recovery states can be queried and reset via user space
interfaces.
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Reviewed-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Jan Hoeppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Store flags and path_data per channel path.
Implement get/set functions for various path masks.
The patch does not add functional changes.
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Reviewed-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Jan Hoeppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Before we set a device offline, the open_count for the block device is
checked and certain flags are checked and set as well.
However, this is all done without holding any lock. Potentially, if the
open_count was checked but the DASD_FLAG_OFFLINE wasn't set yet, a
different process might want to increase the open_count depending on
whether DASD_FLAG_OFFLINE is set or not in the meanwhile.
This is quite racy and can lead to the loss of the device for that
process and subsequently lead to a panic.
Fix this by checking the open_count and setting the offline flags while
holding the ccwdev lock.
Reviewed-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
block->request_queue is used many times in dasd_setup_queue. Define a
separate variable to increase readability a bit and to make it better
reusable.
Reviewed-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Currently the block queue value max_segments is set to -1L, which
is then implicitly casted to unsigned short in blk_queue_max_segments.
This results in 65535 (64k) max_segments.
Even though the resulting value is correct, setting it implicitly using
-1L is rather confusing. Set the value explicitly using the USHRT_MAX
macro instead.
Reviewed-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
A DASD device consists of the device itself and a discipline with a
corresponding private structure. These fields are set up during online
processing right after the device is created and before it is processed by
the state machine and made available for I/O.
During offline processing the discipline pointer and the private data gets
freed within the state machine and without protection of the existing
reference count. This might lead to a kernel panic because a function might
have taken a device reference and accesses the discipline pointer and/or
private data of the device while this is already freed.
Fix by freeing the discipline pointer and the private data after ensuring
that there is no reference to the device left.
Reviewed-by: Peter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Internal I/O is processed by the _sleep_on_function which might wait for a
device to get operational. During offline processing this will never happen
and therefore the refcount of the device will not drop to zero and the
offline processing blocks as well.
Fix by letting requests fail in the _sleep_on function during offline
processing. No further handling of the requests is necessary since this is
internal I/O and the device is thrown away afterwards.
Reviewed-by: Peter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The DASD device driver throws change events for the DASD blockdevice
after the online processing is done so that udev rules can take
actions after it.
The change event was missing for unformatted devices.
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
When a device is in a status where CIO has killed all I/O by itself the
interrupt for a clear request may not contain an irb to determine the
clear function. Instead it contains an error pointer -EIO.
This was ignored by the DASD int_handler leading to a hanging device
waiting for a clear interrupt.
Handle -EIO error pointer correctly for requests that are clear pending and
treat the clear as successful.
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Reviewed-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Implement new DASD IOCTL BIODASDCHECKFMT to check a range of tracks on a
DASD volume for correct formatting. The following characteristics are
checked:
- Block size
- ECKD key length
- ECKD record ID
- Number of records per track
Signed-off-by: Jan Höppner <hoeppner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
With this feature, applications can query if a DASD volume is online
to another operating system instances by checking the online status of
all attached hosts from the storage server.
Reviewed-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Commit ca369d51b ("sd: Fix device-imposed transfer length limits")
introduced a new queue limit max_dev_sectors which limits the maximum
sectors for requests. The default value leads to small dasd requests
and therefor to a performance drop.
Set the max_dev_sectors value to the same value as the max_hw_sectors
to use the maximum available request size for DASD devices.
Signed-off-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org # 4.4+
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Enabling failfast should let request fail immediately if either an
error occurred or the device gets disconnected.
For disconnected devices new requests are not fetches from the block
queue and therefore failfast is not triggered.
Fix by letting the DASD driver fetch requests for disconnected devices
with failfast active.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
We were able to reduce the CPU overhead of big paging scenarios
when announcing our paging disks as non-rotational.
Almost all dasd devices are implemented in storage servers with
cache, raid, striping and lots of magic. There is no point in
optimizing the disk schedulers and swap code for a single platter
moving arm rotational disks. Given the complexity of the setup
and the fact that this change is mostly to disable the additional
overhead in swap code, lets keep the other functionality unchanged
and do not disable the this device as entropy source - unlike other
non-rotational devices.
Suggested-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The dasd device driver selects which (alias or base) device is used
for a given requests when the request is build. If the chosen alias
device is set offline before the request gets queued to the device
queue the starting function may use device structures that are
already freed. This might lead to a hanging offline process or a
kernel panic.
Add a check to the starting function that returns the request to the
upper layer if the device is already in offline processing.
In addition to that prevent that an alias device that's already in
offline processing gets chosen as start device.
Reviewed-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Peter Oberparleiter <peter.oberparleiter@linux.vnet.ibm.com>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Enabling a DASD that was configured to use the DIAG250 access method
while the corresponding kernel module dasd_diag_mod has not been loaded
fails with an error message. To fix this, users need to manually load
the dasd_diag_mod module.
This procedure can be simplified by automatically loading the
dasd_diag_mod from within the kernel when a DASD configured for DIAG250
is set online.
Signed-off-by: Peter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The DASD device driver prevents I/O from being started on stopped
devices. This also prevented channel paths to be verified and so
the device was unable to be resumed.
Fix by allowing path verification requests on stopped devices.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The DASD device driver only has a limited amount of memory to build
I/O requests.
This memory was used by blocklayer requests leading to an inability
to build needed internal requests to resume the device.
Fix by preventing the DASD driver to fetch requests for a stopped
device.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Reference-ID: RQM 2520
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix ref counting for DASD devices leading to an inability to set a
DASD device offline.
Before a worker is scheduled the DASD device driver takes a reference
to the device. If the worker was already scheduled this reference was
never freed.
Fix by giving the reference to the DASD device free when
schedule_work() returns false.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the hard coded scheduler for the DASD device driver to enable
change of the scheduler during runtime. Set recommended deadline
scheduler via additional udev rule.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the 31 bit support in order to reduce maintenance cost and
effectively remove dead code. Since a couple of years there is no
distribution left that comes with a 31 bit kernel.
The 31 bit kernel also has been broken since more than a year before
anybody noticed. In addition I added a removal warning to the kernel
shown at ipl for 5 minutes: a960062e58 ("s390: add 31 bit warning
message") which let everybody know about the plan to remove 31 bit
code. We didn't get any response.
Given that the last 31 bit only machine was introduced in 1999 let's
remove the code.
Anybody with 31 bit user space code can still use the compat mode.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The dasd driver has a lot of duplicated code to handle
dasd_global_profile. With this patch we use the same code for the
global and the per device profiling data. Note that dasd_stats_write
had to change slightly to maintain some odd differences between
A) per device and global profile and B) proc and sysfs interface
usage.
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Access to DASDs global statistics is done without locking which
can lead to inconsistent data. Add locking to fix this. Also move
the relevant structs in a global dasd_profile struct.
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix race for sleep_on requests leading to list corruption.
The SLEEP_ON_END_TAG is set during CQR clean up. Remove it from
interrupt handler to avoid the CQR from being cleared when it is
still in the device_queue.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
During device activation all paths could be lost and since the device
is not active it has no indication of this fact - hence the CQR will
time-out. The following cancelation might fail with -EINVAL because
CIO took over control and started path verification. In this case mark
the CQR as being CLEARED since it could not be running any more.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Add support for Control Unit Initiated Reconfiguration (CUIR) to
Linux, a storage server interface to reconcile concurrent hardware
changes between storage and host.
Reviewed-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Error recovery requests may not be cleaned up correctly so that other
needed erp requests can not be build because of insufficient memory.
This would lead to an infinite loop trying to build erp requests.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Kernel panic or a hanging device during format if an alias device is
set offline or I/O errors occur.
Omit the error recovery procedure for alias devices and do retries on
the base device with full erp.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If I/O errors occur during format a kernel panic with a list_del
corruption may occur.
Stop error recovery procedure after an erp action was taken.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If path events occur the formatting process stucks because path
events may flush format requests from the queue.
Kick the format process after path events are handled.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Formatting of a previously formatted device is slower than newly
format a device when alias devices are available.
For already formatted devices the alias devices are not used for
formatting.
Fix the alias handling for already formatted devices.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix checkpatch warnings:
"WARNING: Prefer seq_puts to seq_printf"
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix checkpatch warnings:
"WARNING: EXPORT_SYMBOL(foo); should immediately follow its function/variable"
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This fixes checkpatch warning:
"WARNING: debugfs_remove(NULL) is safe this check is probably not required"
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Pull block driver updates from Jens Axboe:
"This is the block driver pull request for 3.13. As with the core pull
request just sent out, this was rebased on top of the core branch
again after the immutable series was pulled. This also means that
bcache gets to sit the initial pull over. I will send a second driver
pull request in the merge window to get those fixes in, once they have
been rebased and tested on top of the non-immutable stack.
This pull request contains:
- Add support for the sTec Kronos pci-e flash card from sTec. Also
has various cleanups for this driver, from myself, Bart, Mike
Snizter, and Wei Yongjun.
- Add surprise removal support for the micron mtip32xx driver from
Micron.
- Floppy documentation fix from Ben Harris.
- debugfs bug fix for pktcdvd from Dan Carpenter.
- Fix for the mtip32xx driver stack usage in the debugfs path,
dynamically allocating those buffers instead. From David Milburn.
- Disable cpqarray in Kconfig. The plan is to remove it on request
of HP, but lets disable it for a few revisions just to see if
anyone yells.
- drbd fixes from Lars Ellenberg and Philipp Reisner.
- Elevator switch fix for the s390 block driver from Heiko Carstens.
- loop crash fix on IO to unassigned device from Mikulas Patocka.
- A series of bug fixes for the IBM rsxx pci-e flash driver from
Philip J Kelleher.
- cciss probe fix from Stephen Cameron.
- Xen block front/back fixes from Roger Pau Monne and Vegard Nossum"
* 'for-3.13/drivers' of git://git.kernel.dk/linux-block: (41 commits)
floppy: Correct documentation of driver options when used as a module.
pktcdvd: debugfs functions return NULL on error
xen-blkfront: restore the non-persistent data path
skd: fix formatting in skd_s1120.h
skd: reorder construct/destruct code
skd: cleanup skd_do_inq_page_da()
skd: remove SKD_OMIT_FROM_SRC_DIST ifdefs
skd: remove redundant skdev->pdev assignment from skd_pci_probe()
skd: use <asm/unaligned.h>
skd: remove SCSI subsystem specific includes
skd: register block device only if some devices are present
skd: fix error messages in skd_init()
skd: fix error paths in skd_init()
skd: fix unregister_blkdev() placement
skd: more removal of bio-based code
skd: cleanup the skd_*() function block wrapping
skd: rip out bio path
skd: fix error return code in skd_pci_probe()
s390/dasd: hold request queue sysfs lock when calling elevator_init()
cciss: return 0 from driver probe function on success, not 1
...
"elevator: Fix a race in elevator switching and md device initialization"
changed the semantics of elevator_init() in a way that now enforces to hold
the corresponding request queue's sysfs_lock when calling elevator_init()
to fix a race.
The patch did not convert the s390 dasd device driver which is the only
device driver which also calls elevator_init(). So add the missing locking.
Cc: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We check for the existence of block->profile.data before we write to
it, but the dependent code block misses braces.
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The processing of the dasd_block tasklet may have been interrupted
by a path event.
Restart the dasd tasklets in sleep_on_immediately function.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Whenever a DASD request encounters a timeout we might
need to abort all outstanding requests on this or
even other devices.
This is especially useful if one wants to fail all
devices on one side of a RAID10 configuration, even
though only one device exhibited an error.
To handle this I've introduced a new device flag
DASD_FLAG_ABORTIO.
This flag is evaluated in __dasd_process_request_queue()
and will invoke blk_abort_request() for all
outstanding requests with DASD_CQR_FLAGS_FAILFAST set.
This will cause any of these requests to be aborted
immediately if the blk_timeout function is activated.
The DASD_FLAG_ABORTIO is also evaluated in
__dasd_process_request_queue to abort all
new request which would have the
DASD_CQR_FLAGS_FAILFAST bit set.
The flag can be set with the new ioctls 'BIODASDABORTIO'
and removed with 'BIODASDALLOWIO'.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch adds a 'timeout' attibute to the DASD driver.
When set to non-zero, the blk_timeout function will
be enabled with the timeout specified in the attribute.
Setting 'timeout' to '0' will disable block timeouts.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The DASD driver is using FASTFAIL as an equivalent to the
transport errors in SCSI. And the 'steal lock' function maps
roughly to a reservation error. So we should be returning the
appropriate error codes when completing a request.
Acked-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch implements generic block layer timeout handling
callbacks for DASDs. When the timeout expires the respective
cqr is aborted.
With this timeout handler time-critical request abort
is guaranteed as the abort does not depend on the internal
state of the various DASD driver queues.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Originally the DASD device tasklet would process the entries on
the ccw_queue until the first non-final request was found.
Which was okay as long as all requests have the same retries and
expires parameter.
However, as we're now allowing to modify both it is possible to
have requests _after_ the first request which already have expired.
So we need to check all requests in the device tasklet.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
dasd_cancel_req will never return 1, only 0.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
When a path is gone and dasd_generic_path_event is called with a
PE_PATH_GONE event, we must assume that any I/O request on that
subchannel is still running. This is unlike the dasd_generic_notify
handler and the CIO_NO_PATH event, which implies that the subchannel
has been cleared.
If dasd_generic_path_event finds that the path has been the last
usable path, it must not call dasd_generic_last_path_gone (which would
reset the state of running requests), but just set the
DASD_STOPPED_DC_WAIT bit.
Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The value passed is 0 in all but "it can never happen" cases (and those
only in a couple of drivers) *and* it would've been lost on the way
out anyway, even if something tried to pass something meaningful.
Just don't bother.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If too many ccw requests are pre-build before a suspend/resume cycle
the device might not get enough memory to do path verification
during resume.
Requeue requests to the block device request queue on suspend and free
pre-build ccw requests.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
