linux/drivers/target/target_core_tmr.c
Christoph Hellwig 35e0e75753 target: use a workqueue for I/O completions
Instead of abusing the target processing thread for offloading I/O
completion in the backends to user context add a new workqueue.  This means
completions can be processed as fast as available CPU time allows it,
including in parallel with other completions and more importantly I/O
submission or QUEUE FULL retries.  This should give much better performance
especially on loaded systems.

As a fallout we can merge all the completed states into a single
one.

On the downside this change complicates lun reset handling a bit by
requiring us to cancel a work item only for those states that have it
initialized.  The alternative would be to either always initialize the work
item to a dummy handler, or always use the same handler and do a switch on
the state. The long term solution will be a flag that says that the command
has an initialized work item, but that's only going to be useful once we
have more users.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
2011-10-24 03:21:52 +00:00

435 lines
13 KiB
C

/*******************************************************************************
* Filename: target_core_tmr.c
*
* This file contains SPC-3 task management infrastructure
*
* Copyright (c) 2009,2010 Rising Tide Systems
* Copyright (c) 2009,2010 Linux-iSCSI.org
*
* Nicholas A. Bellinger <nab@kernel.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
******************************************************************************/
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/list.h>
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <target/target_core_base.h>
#include <target/target_core_device.h>
#include <target/target_core_tmr.h>
#include <target/target_core_transport.h>
#include <target/target_core_fabric_ops.h>
#include <target/target_core_configfs.h>
#include "target_core_alua.h"
#include "target_core_pr.h"
struct se_tmr_req *core_tmr_alloc_req(
struct se_cmd *se_cmd,
void *fabric_tmr_ptr,
u8 function,
gfp_t gfp_flags)
{
struct se_tmr_req *tmr;
tmr = kmem_cache_zalloc(se_tmr_req_cache, gfp_flags);
if (!tmr) {
pr_err("Unable to allocate struct se_tmr_req\n");
return ERR_PTR(-ENOMEM);
}
tmr->task_cmd = se_cmd;
tmr->fabric_tmr_ptr = fabric_tmr_ptr;
tmr->function = function;
INIT_LIST_HEAD(&tmr->tmr_list);
return tmr;
}
EXPORT_SYMBOL(core_tmr_alloc_req);
void core_tmr_release_req(
struct se_tmr_req *tmr)
{
struct se_device *dev = tmr->tmr_dev;
unsigned long flags;
if (!dev) {
kmem_cache_free(se_tmr_req_cache, tmr);
return;
}
spin_lock_irqsave(&dev->se_tmr_lock, flags);
list_del(&tmr->tmr_list);
spin_unlock_irqrestore(&dev->se_tmr_lock, flags);
kmem_cache_free(se_tmr_req_cache, tmr);
}
static void core_tmr_handle_tas_abort(
struct se_node_acl *tmr_nacl,
struct se_cmd *cmd,
int tas,
int fe_count)
{
if (!fe_count) {
transport_cmd_finish_abort(cmd, 1);
return;
}
/*
* TASK ABORTED status (TAS) bit support
*/
if ((tmr_nacl &&
(tmr_nacl == cmd->se_sess->se_node_acl)) || tas)
transport_send_task_abort(cmd);
transport_cmd_finish_abort(cmd, 0);
}
static void core_tmr_drain_tmr_list(
struct se_device *dev,
struct se_tmr_req *tmr,
struct list_head *preempt_and_abort_list)
{
LIST_HEAD(drain_tmr_list);
struct se_tmr_req *tmr_p, *tmr_pp;
struct se_cmd *cmd;
unsigned long flags;
/*
* Release all pending and outgoing TMRs aside from the received
* LUN_RESET tmr..
*/
spin_lock_irqsave(&dev->se_tmr_lock, flags);
list_for_each_entry_safe(tmr_p, tmr_pp, &dev->dev_tmr_list, tmr_list) {
/*
* Allow the received TMR to return with FUNCTION_COMPLETE.
*/
if (tmr && (tmr_p == tmr))
continue;
cmd = tmr_p->task_cmd;
if (!cmd) {
pr_err("Unable to locate struct se_cmd for TMR\n");
continue;
}
/*
* If this function was called with a valid pr_res_key
* parameter (eg: for PROUT PREEMPT_AND_ABORT service action
* skip non regisration key matching TMRs.
*/
if (preempt_and_abort_list &&
(core_scsi3_check_cdb_abort_and_preempt(
preempt_and_abort_list, cmd) != 0))
continue;
spin_lock(&cmd->t_state_lock);
if (!atomic_read(&cmd->t_transport_active)) {
spin_unlock(&cmd->t_state_lock);
continue;
}
if (cmd->t_state == TRANSPORT_ISTATE_PROCESSING) {
spin_unlock(&cmd->t_state_lock);
continue;
}
spin_unlock(&cmd->t_state_lock);
list_move_tail(&tmr->tmr_list, &drain_tmr_list);
}
spin_unlock_irqrestore(&dev->se_tmr_lock, flags);
while (!list_empty(&drain_tmr_list)) {
tmr = list_entry(drain_tmr_list.next, struct se_tmr_req, tmr_list);
list_del(&tmr->tmr_list);
cmd = tmr_p->task_cmd;
pr_debug("LUN_RESET: %s releasing TMR %p Function: 0x%02x,"
" Response: 0x%02x, t_state: %d\n",
(preempt_and_abort_list) ? "Preempt" : "", tmr,
tmr->function, tmr->response, cmd->t_state);
transport_cmd_finish_abort(cmd, 1);
}
}
static void core_tmr_drain_task_list(
struct se_device *dev,
struct se_cmd *prout_cmd,
struct se_node_acl *tmr_nacl,
int tas,
struct list_head *preempt_and_abort_list)
{
LIST_HEAD(drain_task_list);
struct se_cmd *cmd;
struct se_task *task, *task_tmp;
unsigned long flags;
int fe_count;
/*
* Complete outstanding struct se_task CDBs with TASK_ABORTED SAM status.
* This is following sam4r17, section 5.6 Aborting commands, Table 38
* for TMR LUN_RESET:
*
* a) "Yes" indicates that each command that is aborted on an I_T nexus
* other than the one that caused the SCSI device condition is
* completed with TASK ABORTED status, if the TAS bit is set to one in
* the Control mode page (see SPC-4). "No" indicates that no status is
* returned for aborted commands.
*
* d) If the logical unit reset is caused by a particular I_T nexus
* (e.g., by a LOGICAL UNIT RESET task management function), then "yes"
* (TASK_ABORTED status) applies.
*
* Otherwise (e.g., if triggered by a hard reset), "no"
* (no TASK_ABORTED SAM status) applies.
*
* Note that this seems to be independent of TAS (Task Aborted Status)
* in the Control Mode Page.
*/
spin_lock_irqsave(&dev->execute_task_lock, flags);
list_for_each_entry_safe(task, task_tmp, &dev->state_task_list,
t_state_list) {
if (!task->task_se_cmd) {
pr_err("task->task_se_cmd is NULL!\n");
continue;
}
cmd = task->task_se_cmd;
/*
* For PREEMPT_AND_ABORT usage, only process commands
* with a matching reservation key.
*/
if (preempt_and_abort_list &&
(core_scsi3_check_cdb_abort_and_preempt(
preempt_and_abort_list, cmd) != 0))
continue;
/*
* Not aborting PROUT PREEMPT_AND_ABORT CDB..
*/
if (prout_cmd == cmd)
continue;
list_move_tail(&task->t_state_list, &drain_task_list);
atomic_set(&task->task_state_active, 0);
/*
* Remove from task execute list before processing drain_task_list
*/
if (!list_empty(&task->t_execute_list))
__transport_remove_task_from_execute_queue(task, dev);
}
spin_unlock_irqrestore(&dev->execute_task_lock, flags);
while (!list_empty(&drain_task_list)) {
task = list_entry(drain_task_list.next, struct se_task, t_state_list);
list_del(&task->t_state_list);
cmd = task->task_se_cmd;
pr_debug("LUN_RESET: %s cmd: %p task: %p"
" ITT/CmdSN: 0x%08x/0x%08x, i_state: %d, t_state: %d"
"cdb: 0x%02x\n",
(preempt_and_abort_list) ? "Preempt" : "", cmd, task,
cmd->se_tfo->get_task_tag(cmd), 0,
cmd->se_tfo->get_cmd_state(cmd), cmd->t_state,
cmd->t_task_cdb[0]);
pr_debug("LUN_RESET: ITT[0x%08x] - pr_res_key: 0x%016Lx"
" t_task_cdbs: %d t_task_cdbs_left: %d"
" t_task_cdbs_sent: %d -- t_transport_active: %d"
" t_transport_stop: %d t_transport_sent: %d\n",
cmd->se_tfo->get_task_tag(cmd), cmd->pr_res_key,
cmd->t_task_list_num,
atomic_read(&cmd->t_task_cdbs_left),
atomic_read(&cmd->t_task_cdbs_sent),
atomic_read(&cmd->t_transport_active),
atomic_read(&cmd->t_transport_stop),
atomic_read(&cmd->t_transport_sent));
/*
* If the command may be queued onto a workqueue cancel it now.
*
* This is equivalent to removal from the execute queue in the
* loop above, but we do it down here given that
* cancel_work_sync may block.
*/
if (cmd->t_state == TRANSPORT_COMPLETE)
cancel_work_sync(&cmd->work);
spin_lock_irqsave(&cmd->t_state_lock, flags);
target_stop_task(task, &flags);
if (!atomic_dec_and_test(&cmd->t_task_cdbs_ex_left)) {
spin_unlock_irqrestore(&cmd->t_state_lock, flags);
pr_debug("LUN_RESET: Skipping task: %p, dev: %p for"
" t_task_cdbs_ex_left: %d\n", task, dev,
atomic_read(&cmd->t_task_cdbs_ex_left));
continue;
}
fe_count = atomic_read(&cmd->t_fe_count);
if (atomic_read(&cmd->t_transport_active)) {
pr_debug("LUN_RESET: got t_transport_active = 1 for"
" task: %p, t_fe_count: %d dev: %p\n", task,
fe_count, dev);
atomic_set(&cmd->t_transport_aborted, 1);
spin_unlock_irqrestore(&cmd->t_state_lock, flags);
core_tmr_handle_tas_abort(tmr_nacl, cmd, tas, fe_count);
continue;
}
pr_debug("LUN_RESET: Got t_transport_active = 0 for task: %p,"
" t_fe_count: %d dev: %p\n", task, fe_count, dev);
atomic_set(&cmd->t_transport_aborted, 1);
spin_unlock_irqrestore(&cmd->t_state_lock, flags);
core_tmr_handle_tas_abort(tmr_nacl, cmd, tas, fe_count);
}
}
static void core_tmr_drain_cmd_list(
struct se_device *dev,
struct se_cmd *prout_cmd,
struct se_node_acl *tmr_nacl,
int tas,
struct list_head *preempt_and_abort_list)
{
LIST_HEAD(drain_cmd_list);
struct se_queue_obj *qobj = &dev->dev_queue_obj;
struct se_cmd *cmd, *tcmd;
unsigned long flags;
/*
* Release all commands remaining in the struct se_device cmd queue.
*
* This follows the same logic as above for the struct se_device
* struct se_task state list, where commands are returned with
* TASK_ABORTED status, if there is an outstanding $FABRIC_MOD
* reference, otherwise the struct se_cmd is released.
*/
spin_lock_irqsave(&qobj->cmd_queue_lock, flags);
list_for_each_entry_safe(cmd, tcmd, &qobj->qobj_list, se_queue_node) {
/*
* For PREEMPT_AND_ABORT usage, only process commands
* with a matching reservation key.
*/
if (preempt_and_abort_list &&
(core_scsi3_check_cdb_abort_and_preempt(
preempt_and_abort_list, cmd) != 0))
continue;
/*
* Not aborting PROUT PREEMPT_AND_ABORT CDB..
*/
if (prout_cmd == cmd)
continue;
/*
* Skip direct processing of TRANSPORT_FREE_CMD_INTR for
* HW target mode fabrics.
*/
spin_lock(&cmd->t_state_lock);
if (cmd->t_state == TRANSPORT_FREE_CMD_INTR) {
spin_unlock(&cmd->t_state_lock);
continue;
}
spin_unlock(&cmd->t_state_lock);
atomic_set(&cmd->t_transport_queue_active, 0);
atomic_dec(&qobj->queue_cnt);
list_move_tail(&cmd->se_queue_node, &drain_cmd_list);
}
spin_unlock_irqrestore(&qobj->cmd_queue_lock, flags);
while (!list_empty(&drain_cmd_list)) {
cmd = list_entry(drain_cmd_list.next, struct se_cmd, se_queue_node);
list_del_init(&cmd->se_queue_node);
pr_debug("LUN_RESET: %s from Device Queue: cmd: %p t_state:"
" %d t_fe_count: %d\n", (preempt_and_abort_list) ?
"Preempt" : "", cmd, cmd->t_state,
atomic_read(&cmd->t_fe_count));
/*
* Signal that the command has failed via cmd->se_cmd_flags,
*/
transport_new_cmd_failure(cmd);
core_tmr_handle_tas_abort(tmr_nacl, cmd, tas,
atomic_read(&cmd->t_fe_count));
}
}
int core_tmr_lun_reset(
struct se_device *dev,
struct se_tmr_req *tmr,
struct list_head *preempt_and_abort_list,
struct se_cmd *prout_cmd)
{
struct se_node_acl *tmr_nacl = NULL;
struct se_portal_group *tmr_tpg = NULL;
int tas;
/*
* TASK_ABORTED status bit, this is configurable via ConfigFS
* struct se_device attributes. spc4r17 section 7.4.6 Control mode page
*
* A task aborted status (TAS) bit set to zero specifies that aborted
* tasks shall be terminated by the device server without any response
* to the application client. A TAS bit set to one specifies that tasks
* aborted by the actions of an I_T nexus other than the I_T nexus on
* which the command was received shall be completed with TASK ABORTED
* status (see SAM-4).
*/
tas = dev->se_sub_dev->se_dev_attrib.emulate_tas;
/*
* Determine if this se_tmr is coming from a $FABRIC_MOD
* or struct se_device passthrough..
*/
if (tmr && tmr->task_cmd && tmr->task_cmd->se_sess) {
tmr_nacl = tmr->task_cmd->se_sess->se_node_acl;
tmr_tpg = tmr->task_cmd->se_sess->se_tpg;
if (tmr_nacl && tmr_tpg) {
pr_debug("LUN_RESET: TMR caller fabric: %s"
" initiator port %s\n",
tmr_tpg->se_tpg_tfo->get_fabric_name(),
tmr_nacl->initiatorname);
}
}
pr_debug("LUN_RESET: %s starting for [%s], tas: %d\n",
(preempt_and_abort_list) ? "Preempt" : "TMR",
dev->transport->name, tas);
core_tmr_drain_tmr_list(dev, tmr, preempt_and_abort_list);
core_tmr_drain_task_list(dev, prout_cmd, tmr_nacl, tas,
preempt_and_abort_list);
core_tmr_drain_cmd_list(dev, prout_cmd, tmr_nacl, tas,
preempt_and_abort_list);
/*
* Clear any legacy SPC-2 reservation when called during
* LOGICAL UNIT RESET
*/
if (!preempt_and_abort_list &&
(dev->dev_flags & DF_SPC2_RESERVATIONS)) {
spin_lock(&dev->dev_reservation_lock);
dev->dev_reserved_node_acl = NULL;
dev->dev_flags &= ~DF_SPC2_RESERVATIONS;
spin_unlock(&dev->dev_reservation_lock);
pr_debug("LUN_RESET: SCSI-2 Released reservation\n");
}
spin_lock_irq(&dev->stats_lock);
dev->num_resets++;
spin_unlock_irq(&dev->stats_lock);
pr_debug("LUN_RESET: %s for [%s] Complete\n",
(preempt_and_abort_list) ? "Preempt" : "TMR",
dev->transport->name);
return 0;
}