Skip to contentSkip to conversation

Rework runtime suspend and SCSI domain validation

PATCH Superseded Bart Van Assche 4 participants 17 messages Nov 23, 2020 – Nov 24, 2020

Subsystems (3)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover Rework runtime suspend and SCSI domain validation

Hi Martin,

The SCSI runtime suspend and domain validation mechanisms both use scsi_device_quiesce(). scsi_device_quiesce() restricts blk_queue_enter() to BLK_MQ_REQ_PREEMPT requests. There is a conflict between the requirements of runtime suspend and SCSI domain validation: no requests must be sent to runtime suspended devices that are in the state RPM_SUSPENDED while BLK_MQ_REQ_PREEMPT requests must be processed during SCSI domain validation. This conflict is resolved by reworking the SCSI domain validation implementation.

Hybernation, runtime suspend and SCSI domain validation have been retested.

Please consider this patch series for kernel v5.11.

Thanks,

Bart.

Changes between v2 and v3:
- Inlined scsi_mq_alloc_queue() into scsi_alloc_sdev() as requested by Christoph.

Changes between v1 and v2:
- Rebased this patch series on top of kernel v5.10-rc1.

Alan Stern (1):
block: Do not accept any requests while suspended

Bart Van Assche (8):
block: Fix a race in the runtime power management code
ide: Do not set the RQF_PREEMPT flag for sense requests
scsi: Pass a request queue pointer to __scsi_execute()
scsi: Inline scsi_mq_alloc_queue()
scsi: Do not wait for a request in scsi_eh_lock_door()
scsi_transport_spi: Make spi_execute() accept a request queue pointer
scsi_transport_spi: Freeze request queues instead of quiescing block, scsi, ide: Only process PM requests if rpm_status != RPM_ACTIVE

 block/blk-core.c                  |  12 +--
 block/blk-mq-debugfs.c            |   1 -
 block/blk-mq.c                    |   4 +-
 block/blk-pm.c                    |  15 +--
 block/blk-pm.h                    |  14 ++-
 drivers/ide/ide-atapi.c           |   1 -
 drivers/ide/ide-io.c              |   3 +-
 drivers/ide/ide-pm.c              |   2 +-
 drivers/scsi/scsi_error.c         |   7 +-
 drivers/scsi/scsi_lib.c           |  72 ++++++--------
 drivers/scsi/scsi_priv.h          |   3 +-
 drivers/scsi/scsi_scan.c          |  12 ++-
 drivers/scsi/scsi_transport_spi.c | 151 ++++++++++++++++++------------
 include/linux/blk-mq.h            |   4 +-
 include/linux/blkdev.h            |   6 +-
 include/scsi/scsi_device.h        |  14 ++-
 16 files changed, 175 insertions(+), 146 deletions(-)
#1 lore raw

PATCH v3 1/9 block: Fix a race in the runtime power management code

With the current implementation the following race can happen:
* blk_pre_runtime_suspend() calls blk_freeze_queue_start() and blk_mq_unfreeze_queue().
* blk_queue_enter() calls blk_queue_pm_only() and that function returns true.
* blk_queue_enter() calls blk_pm_request_resume() and that function does not call pm_request_resume() because the queue runtime status is RPM_ACTIVE.
* blk_pre_runtime_suspend() changes the queue status into RPM_SUSPENDING.

Fix this race by changing the queue runtime status into RPM_SUSPENDING before switching q_usage_counter to atomic mode.

Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
Cc stable <[email protected]>
Fixes 986d413b7c15 ("blk-mq: Enable support for runtime power management")
---
 block/blk-pm.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
block/blk-pm.c +9 -6
index b85234d758f7..17bd020268d4 100644
@@ -67,6 +67,10 @@ int blk_pre_runtime_suspend(struct request_queue *q)
 
 	WARN_ON_ONCE(q->rpm_status != RPM_ACTIVE);
 
+	spin_lock_irq(&q->queue_lock);
+	q->rpm_status = RPM_SUSPENDING;
+	spin_unlock_irq(&q->queue_lock);
+
 	/*
 	 * Increase the pm_only counter before checking whether any
 	 * non-PM blk_queue_enter() calls are in progress to avoid that any
@@ -89,15 +93,14 @@ int blk_pre_runtime_suspend(struct request_queue *q)
 	/* Switch q_usage_counter back to per-cpu mode. */
 	blk_mq_unfreeze_queue(q);
 
-	spin_lock_irq(&q->queue_lock);
-	if (ret < 0)
+	if (ret < 0) {
+		spin_lock_irq(&q->queue_lock);
+		q->rpm_status = RPM_ACTIVE;
 		pm_runtime_mark_last_busy(q->dev);
-	else
-		q->rpm_status = RPM_SUSPENDING;
-	spin_unlock_irq(&q->queue_lock);
+		spin_unlock_irq(&q->queue_lock);
 
-	if (ret)
 		blk_clear_pm_only(q);
+	}
 
 	return ret;
 }
Reviewed-by on v3 1/9
#2 lore raw

Re: [PATCH v3 1/9] block: Fix a race in the runtime power management code

Quoted 22 lines · Bart Van Assche

With the current implementation the following race can happen:
* blk_pre_runtime_suspend() calls blk_freeze_queue_start() and blk_mq_unfreeze_queue().
* blk_queue_enter() calls blk_queue_pm_only() and that function returns true.
* blk_queue_enter() calls blk_pm_request_resume() and that function does not call pm_request_resume() because the queue runtime status is RPM_ACTIVE.
* blk_pre_runtime_suspend() changes the queue status into RPM_SUSPENDING.

Fix this race by changing the queue runtime status into RPM_SUSPENDING before switching q_usage_counter to atomic mode.

Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
Cc stable <[email protected]>
Fixes 986d413b7c15 ("blk-mq: Enable support for runtime power management")
Quoted 4 lines
---
  block/blk-pm.c | 15 +++++++++------
  1 file changed, 9 insertions(+), 6 deletions(-)

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke                Kernel Storage Architect
[email protected]                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
#3 lore raw

PATCH v3 2/9 ide: Do not set the RQF_PREEMPT flag for sense requests

In reply to Bart Van Assche

RQF_PREEMPT is used for two different purposes in the legacy IDE code:
1. To mark power management requests.
2. To mark requests that should preempt another request. An (old) explanation of that feature is as follows:
"The IDE driver in the Linux kernel normally uses a series of busywait delays during its initialization. When the driver executes these busywaits, the kernel does nothing for the duration of the wait. The time spent in these waits could be used for other initialization activities, if they could be run concurrently with these waits.

More specifically, busywait-style delays such as udelay() in module init functions inhibit kernel preemption because the Big Kernel Lock is held, while yielding APIs such as schedule_timeout() allow preemption. This is true because the kernel handles the BKL specially and releases and reacquires it across reschedules allowed by the current thread.

This IDE-preempt specification requires that the driver eliminate these busywaits and replace them with a mechanism that allows other work to proceed while the IDE driver is initializing."

Since I haven't found an implementation of (2), do not set the PREEMPT flag for sense requests. This patch causes sense requests to be postponed while a drive is suspended instead of being submitted to ide_queue_rq().

If it would ever be necessary to restore the IDE PREEMPT functionality, that can be done by introducing a new flag in struct ide_request.

This patch is a first step towards removing the PREEMPT flag from the block layer.

Cc David S. Miller <[email protected]>
Cc Alan Stern <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
 drivers/ide/ide-atapi.c | 1 -
 1 file changed, 1 deletion(-)
drivers/ide/ide-atapi.c +0 -1
index 2162bc80f09e..013ad33fbbc8 100644
@@ -223,7 +223,6 @@ void ide_prep_sense(ide_drive_t *drive, struct request *rq)
 	sense_rq->rq_disk = rq->rq_disk;
 	sense_rq->cmd_flags = REQ_OP_DRV_IN;
 	ide_req(sense_rq)->type = ATA_PRIV_SENSE;
-	sense_rq->rq_flags |= RQF_PREEMPT;
 
 	req->cmd[0] = GPCMD_REQUEST_SENSE;
 	req->cmd[4] = cmd_len;
Reviewed-by on v3 2/9
#4 lore raw

Re: [PATCH v3 2/9] ide: Do not set the RQF_PREEMPT flag for sense requests

Quoted 49 lines · Bart Van Assche

RQF_PREEMPT is used for two different purposes in the legacy IDE code:
1. To mark power management requests.
2. To mark requests that should preempt another request. An (old) explanation of that feature is as follows:
"The IDE driver in the Linux kernel normally uses a series of busywait delays during its initialization. When the driver executes these busywaits, the kernel does nothing for the duration of the wait. The time spent in these waits could be used for other initialization activities, if they could be run concurrently with these waits.

More specifically, busywait-style delays such as udelay() in module init functions inhibit kernel preemption because the Big Kernel Lock is held, while yielding APIs such as schedule_timeout() allow preemption. This is true because the kernel handles the BKL specially and releases and reacquires it across reschedules allowed by the current thread.

This IDE-preempt specification requires that the driver eliminate these busywaits and replace them with a mechanism that allows other work to proceed while the IDE driver is initializing."

Since I haven't found an implementation of (2), do not set the PREEMPT flag for sense requests. This patch causes sense requests to be postponed while a drive is suspended instead of being submitted to ide_queue_rq().

If it would ever be necessary to restore the IDE PREEMPT functionality, that can be done by introducing a new flag in struct ide_request.

This patch is a first step towards removing the PREEMPT flag from the block layer.

Cc David S. Miller <[email protected]>
Cc Alan Stern <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
  drivers/ide/ide-atapi.c | 1 -
  1 file changed, 1 deletion(-)
drivers/ide/ide-atapi.c +0 -0
index 2162bc80f09e..013ad33fbbc8 100644

drivers/ide/ide-atapi.c void ide_prep_sense(ide_drive_t *drive, struct request *rq) · line 228

Quoted 7 lines
drivers/ide/ide-atapi.c +0 -1
@@ -223,7 +223,6 @@ void ide_prep_sense(ide_drive_t *drive, struct request *rq)
  	sense_rq->rq_disk = rq->rq_disk;
  	sense_rq->cmd_flags = REQ_OP_DRV_IN;
  	ide_req(sense_rq)->type = ATA_PRIV_SENSE;
-	sense_rq->rq_flags |= RQF_PREEMPT;
 
  	req->cmd[0] = GPCMD_REQUEST_SENSE;
  	req->cmd[4] = cmd_len;

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke                Kernel Storage Architect
[email protected]                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
#5 lore raw

PATCH v3 3/9 scsi: Pass a request queue pointer to __scsi_execute()

In reply to Bart Van Assche

This patch does not change any functionality but makes a later patch easier to read.

---
 drivers/scsi/scsi_lib.c    | 12 +++++-------
 include/scsi/scsi_device.h |  8 ++++----
 2 files changed, 9 insertions(+), 11 deletions(-)
drivers/scsi/scsi_lib.c +5 -7
index b7ac14571415..a7252df74c7b 100644
@@ -221,7 +221,7 @@ void scsi_queue_insert(struct scsi_cmnd *cmd, int reason)
 
 /**
  * __scsi_execute - insert request and wait for the result
- * @sdev:	scsi device
+ * @q:		queue to insert the request into
  * @cmd:	scsi command
  * @data_direction: data direction
  * @buffer:	data buffer
@@ -237,7 +237,7 @@ void scsi_queue_insert(struct scsi_cmnd *cmd, int reason)
  * Returns the scsi_cmnd result field if a command was executed, or a negative
  * Linux error code if we didn't get that far.
  */
-int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
+int __scsi_execute(struct request_queue *q, const unsigned char *cmd,
 		 int data_direction, void *buffer, unsigned bufflen,
 		 unsigned char *sense, struct scsi_sense_hdr *sshdr,
 		 int timeout, int retries, u64 flags, req_flags_t rq_flags,
@@ -247,15 +247,13 @@ int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 	struct scsi_request *rq;
 	int ret = DRIVER_ERROR << 24;
 
-	req = blk_get_request(sdev->request_queue,
-			data_direction == DMA_TO_DEVICE ?
+	req = blk_get_request(q, data_direction == DMA_TO_DEVICE ?
 			REQ_OP_SCSI_OUT : REQ_OP_SCSI_IN, BLK_MQ_REQ_PREEMPT);
 	if (IS_ERR(req))
 		return ret;
 	rq = scsi_req(req);
 
-	if (bufflen &&	blk_rq_map_kern(sdev->request_queue, req,
-					buffer, bufflen, GFP_NOIO))
+	if (bufflen && blk_rq_map_kern(q, req, buffer, bufflen, GFP_NOIO))
 		goto out;
 
 	rq->cmd_len = COMMAND_SIZE(cmd[0]);
@@ -268,7 +266,7 @@ int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 	/*
 	 * head injection *required* here otherwise quiesce won't work
 	 */
-	blk_execute_rq(req->q, NULL, req, 1);
+	blk_execute_rq(q, NULL, req, 1);
 
 	/*
 	 * Some devices (USB mass-storage in particular) may transfer
include/scsi/scsi_device.h +4 -4
index 1a5c9a3df6d6..f47fdf9cf788 100644
@@ -438,7 +438,7 @@ extern const char *scsi_device_state_name(enum scsi_device_state);
 extern int scsi_is_sdev_device(const struct device *);
 extern int scsi_is_target_device(const struct device *);
 extern void scsi_sanitize_inquiry_string(unsigned char *s, int len);
-extern int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
+extern int __scsi_execute(struct request_queue *q, const unsigned char *cmd,
 			int data_direction, void *buffer, unsigned bufflen,
 			unsigned char *sense, struct scsi_sense_hdr *sshdr,
 			int timeout, int retries, u64 flags,
@@ -449,9 +449,9 @@ extern int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 ({									\
 	BUILD_BUG_ON((sense) != NULL &&					\
 		     sizeof(sense) != SCSI_SENSE_BUFFERSIZE);		\
-	__scsi_execute(sdev, cmd, data_direction, buffer, bufflen,	\
-		       sense, sshdr, timeout, retries, flags, rq_flags,	\
-		       resid);						\
+	__scsi_execute(sdev->request_queue, cmd, data_direction,	\
+		       buffer, bufflen, sense, sshdr, timeout, retries,	\
+		       flags, rq_flags, resid);				\
 })
 static inline int scsi_execute_req(struct scsi_device *sdev,
 	const unsigned char *cmd, int data_direction, void *buffer,
#6 lore raw

PATCH v3 4/9 scsi: Inline scsi_mq_alloc_queue()

In reply to Bart Van Assche

Since scsi_mq_alloc_queue() only has one caller, inline it. This change was suggested by Christoph Hellwig.

Cc Christoph Hellwig <[email protected]>
Cc Alan Stern <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
 drivers/scsi/scsi_lib.c  | 12 ------------
 drivers/scsi/scsi_priv.h |  1 -
 drivers/scsi/scsi_scan.c | 12 ++++++++----
 3 files changed, 8 insertions(+), 17 deletions(-)
drivers/scsi/scsi_lib.c +0 -12
index a7252df74c7b..b5449efc7283 100644
@@ -1881,18 +1881,6 @@ static const struct blk_mq_ops scsi_mq_ops = {
 	.map_queues	= scsi_map_queues,
 };
 
-struct request_queue *scsi_mq_alloc_queue(struct scsi_device *sdev)
-{
-	sdev->request_queue = blk_mq_init_queue(&sdev->host->tag_set);
-	if (IS_ERR(sdev->request_queue))
-		return NULL;
-
-	sdev->request_queue->queuedata = sdev;
-	__scsi_init_queue(sdev->host, sdev->request_queue);
-	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, sdev->request_queue);
-	return sdev->request_queue;
-}
-
 int scsi_mq_setup_tags(struct Scsi_Host *shost)
 {
 	unsigned int cmd_size, sgl_size;
drivers/scsi/scsi_priv.h +0 -1
index 180636d54982..e34755986b47 100644
@@ -90,7 +90,6 @@ extern void scsi_queue_insert(struct scsi_cmnd *cmd, int reason);
 extern void scsi_io_completion(struct scsi_cmnd *, unsigned int);
 extern void scsi_run_host_queues(struct Scsi_Host *shost);
 extern void scsi_requeue_run_queue(struct work_struct *work);
-extern struct request_queue *scsi_mq_alloc_queue(struct scsi_device *sdev);
 extern void scsi_start_queue(struct scsi_device *sdev);
 extern int scsi_mq_setup_tags(struct Scsi_Host *shost);
 extern void scsi_mq_destroy_tags(struct Scsi_Host *shost);
drivers/scsi/scsi_scan.c +8 -4
index f2437a7570ce..43416e7259a7 100644
@@ -216,6 +216,7 @@ static struct scsi_device *scsi_alloc_sdev(struct scsi_target *starget,
 					   u64 lun, void *hostdata)
 {
 	struct scsi_device *sdev;
+	struct request_queue *q;
 	int display_failure_msg = 1, ret;
 	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
 
@@ -265,16 +266,19 @@ static struct scsi_device *scsi_alloc_sdev(struct scsi_target *starget,
 	 */
 	sdev->borken = 1;
 
-	sdev->request_queue = scsi_mq_alloc_queue(sdev);
-	if (!sdev->request_queue) {
+	q = blk_mq_init_queue(&sdev->host->tag_set);
+	if (IS_ERR(q)) {
 		/* release fn is set up in scsi_sysfs_device_initialise, so
 		 * have to free and put manually here */
 		put_device(&starget->dev);
 		kfree(sdev);
 		goto out;
 	}
-	WARN_ON_ONCE(!blk_get_queue(sdev->request_queue));
-	sdev->request_queue->queuedata = sdev;
+	sdev->request_queue = q;
+	q->queuedata = sdev;
+	__scsi_init_queue(sdev->host, q);
+	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, q);
+	WARN_ON_ONCE(!blk_get_queue(q));
 
 	scsi_change_queue_depth(sdev, sdev->host->cmd_per_lun ?
 					sdev->host->cmd_per_lun : 1);
Reviewed-by on v3 4/9
#8 lore raw

Re: [PATCH v3 4/9] scsi: Inline scsi_mq_alloc_queue()

In reply to Bart Van Assche
Quoted 10 lines · Bart Van Assche

Since scsi_mq_alloc_queue() only has one caller, inline it. This change was suggested by Christoph Hellwig.

Cc Christoph Hellwig <[email protected]>
Cc Alan Stern <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
Quoted 81 lines
---
 drivers/scsi/scsi_lib.c  | 12 ------------
 drivers/scsi/scsi_priv.h |  1 -
 drivers/scsi/scsi_scan.c | 12 ++++++++----
 3 files changed, 8 insertions(+), 17 deletions(-)
drivers/scsi/scsi_lib.c +0 -12
index a7252df74c7b..b5449efc7283 100644
@@ -1881,18 +1881,6 @@ static const struct blk_mq_ops scsi_mq_ops = {
 	.map_queues	= scsi_map_queues,
 };
 
-struct request_queue *scsi_mq_alloc_queue(struct scsi_device *sdev)
-{
-	sdev->request_queue = blk_mq_init_queue(&sdev->host->tag_set);
-	if (IS_ERR(sdev->request_queue))
-		return NULL;
-
-	sdev->request_queue->queuedata = sdev;
-	__scsi_init_queue(sdev->host, sdev->request_queue);
-	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, sdev->request_queue);
-	return sdev->request_queue;
-}
-
 int scsi_mq_setup_tags(struct Scsi_Host *shost)
 {
 	unsigned int cmd_size, sgl_size;
drivers/scsi/scsi_priv.h +0 -1
index 180636d54982..e34755986b47 100644
@@ -90,7 +90,6 @@ extern void scsi_queue_insert(struct scsi_cmnd *cmd,
 int reason);
 extern void scsi_io_completion(struct scsi_cmnd *, unsigned int);
 extern void scsi_run_host_queues(struct Scsi_Host *shost);
 extern void scsi_requeue_run_queue(struct work_struct *work);
-extern struct request_queue *scsi_mq_alloc_queue(struct scsi_device 
 *sdev);
 extern void scsi_start_queue(struct scsi_device *sdev);

extern int scsi_mq_setup_tags(struct Scsi_Host *shost); extern void scsi_mq_destroy_tags(struct Scsi_Host *shost);

drivers/scsi/scsi_scan.c +8 -4
index f2437a7570ce..43416e7259a7 100644
@@ -216,6 +216,7 @@ static struct scsi_device *scsi_alloc_sdev(struct
 scsi_target *starget,
 					   u64 lun, void *hostdata)
 {
 	struct scsi_device *sdev;
+	struct request_queue *q;
 	int display_failure_msg = 1, ret;
 	struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
@@ -265,16 +266,19 @@ static struct scsi_device
 *scsi_alloc_sdev(struct scsi_target *starget,
 	 */
 	sdev->borken = 1;
 
-	sdev->request_queue = scsi_mq_alloc_queue(sdev);
-	if (!sdev->request_queue) {
+	q = blk_mq_init_queue(&sdev->host->tag_set);
+	if (IS_ERR(q)) {
 		/* release fn is set up in scsi_sysfs_device_initialise, so
 		 * have to free and put manually here */
 		put_device(&starget->dev);
 		kfree(sdev);
 		goto out;
 	}
-	WARN_ON_ONCE(!blk_get_queue(sdev->request_queue));
-	sdev->request_queue->queuedata = sdev;
+	sdev->request_queue = q;
+	q->queuedata = sdev;
+	__scsi_init_queue(sdev->host, q);
+	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, q);
+	WARN_ON_ONCE(!blk_get_queue(q));
 
 	scsi_change_queue_depth(sdev, sdev->host->cmd_per_lun ?

sdev->host->cmd_per_lun : 1);

#9 lore raw

PATCH v3 5/9 scsi: Do not wait for a request in scsi_eh_lock_door()

In reply to Bart Van Assche

scsi_eh_lock_door() is the only function in the SCSI error handler that calls blk_get_request(). It is not guaranteed that a request is available when scsi_eh_lock_door() is called. Hence pass the BLK_MQ_REQ_NOWAIT flag to blk_get_request(). This patch has a second purpose, namely preventing that scsi_eh_lock_door() deadlocks if sdev->request_queue is frozen and if a SCSI command is submitted to a SCSI device through a second request queue that has not been frozen. A later patch will namely modify the SPI DV code such that sdev->requeust_queue is frozen during domain validation.

---
 drivers/scsi/scsi_error.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/scsi/scsi_error.c +6 -1
index d94449188270..6de6e1bf3dcb 100644
@@ -1993,7 +1993,12 @@ static void scsi_eh_lock_door(struct scsi_device *sdev)
 	struct request *req;
 	struct scsi_request *rq;
 
-	req = blk_get_request(sdev->request_queue, REQ_OP_SCSI_IN, 0);
+	/*
+	 * It is not guaranteed that a request is available nor that
+	 * sdev->request_queue is unfrozen. Hence the BLK_MQ_REQ_NOWAIT below.
+	 */
+	req = blk_get_request(sdev->request_queue, REQ_OP_SCSI_IN,
+			      BLK_MQ_REQ_NOWAIT);
 	if (IS_ERR(req))
 		return;
 	rq = scsi_req(req);
#10 lore raw

PATCH v3 6/9 scsi_transport_spi: Make spi_execute() accept a request queue pointer

In reply to Bart Van Assche

Passing a request queue pointer to spi_execute() instead of a SCSI device pointer will allow a later patch to associate two request queues with a SCSI device. Additionally, instead of assuming that the device state is SDEV_QUIESCE before domain validation starts, read the device state. This patch does not change any functionality but makes a later patch easier to read.

---
 drivers/scsi/scsi_transport_spi.c | 69 ++++++++++++++++---------------
 1 file changed, 36 insertions(+), 33 deletions(-)
drivers/scsi/scsi_transport_spi.c +36 -33
index f3d5b1bbd5aa..959990f66865 100644
@@ -104,7 +104,7 @@ static int sprint_frac(char *dest, int value, int denom)
 	return result;
 }
 
-static int spi_execute(struct scsi_device *sdev, const void *cmd,
+static int spi_execute(struct request_queue *q, const void *cmd,
 		       enum dma_data_direction dir,
 		       void *buffer, unsigned bufflen,
 		       struct scsi_sense_hdr *sshdr)
@@ -117,7 +117,7 @@ static int spi_execute(struct scsi_device *sdev, const void *cmd,
 		sshdr = &sshdr_tmp;
 
 	for(i = 0; i < DV_RETRIES; i++) {
-		result = scsi_execute(sdev, cmd, dir, buffer, bufflen, sense,
+		result = __scsi_execute(q, cmd, dir, buffer, bufflen, sense,
 				      sshdr, DV_TIMEOUT, /* retries */ 1,
 				      REQ_FAILFAST_DEV |
 				      REQ_FAILFAST_TRANSPORT |
@@ -620,13 +620,14 @@ enum spi_compare_returns {
 /* This is for read/write Domain Validation:  If the device supports
  * an echo buffer, we do read/write tests to it */
 static enum spi_compare_returns
-spi_dv_device_echo_buffer(struct scsi_device *sdev, u8 *buffer,
-			  u8 *ptr, const int retries)
+spi_dv_device_echo_buffer(struct scsi_device *sdev, struct request_queue *q,
+			  u8 *buffer, u8 *ptr, const int retries)
 {
 	int len = ptr - buffer;
 	int j, k, r, result;
 	unsigned int pattern = 0x0000ffff;
 	struct scsi_sense_hdr sshdr;
+	enum scsi_device_state sdev_state = sdev->sdev_state;
 
 	const char spi_write_buffer[] = {
 		WRITE_BUFFER, 0x0a, 0, 0, 0, 0, 0, len >> 8, len & 0xff, 0
@@ -671,11 +672,10 @@ spi_dv_device_echo_buffer(struct scsi_device *sdev, u8 *buffer,
 	}
 
 	for (r = 0; r < retries; r++) {
-		result = spi_execute(sdev, spi_write_buffer, DMA_TO_DEVICE,
+		result = spi_execute(q, spi_write_buffer, DMA_TO_DEVICE,
 				     buffer, len, &sshdr);
 		if(result || !scsi_device_online(sdev)) {
-
-			scsi_device_set_state(sdev, SDEV_QUIESCE);
+			scsi_device_set_state(sdev, sdev_state);
 			if (scsi_sense_valid(&sshdr)
 			    && sshdr.sense_key == ILLEGAL_REQUEST
 			    /* INVALID FIELD IN CDB */
@@ -693,9 +693,9 @@ spi_dv_device_echo_buffer(struct scsi_device *sdev, u8 *buffer,
 		}
 
 		memset(ptr, 0, len);
-		spi_execute(sdev, spi_read_buffer, DMA_FROM_DEVICE,
-			    ptr, len, NULL);
-		scsi_device_set_state(sdev, SDEV_QUIESCE);
+		spi_execute(q, spi_read_buffer, DMA_FROM_DEVICE, ptr, len,
+			    NULL);
+		scsi_device_set_state(sdev, sdev_state);
 
 		if (memcmp(buffer, ptr, len) != 0)
 			return SPI_COMPARE_FAILURE;
@@ -706,11 +706,12 @@ spi_dv_device_echo_buffer(struct scsi_device *sdev, u8 *buffer,
 /* This is for the simplest form of Domain Validation: a read test
  * on the inquiry data from the device */
 static enum spi_compare_returns
-spi_dv_device_compare_inquiry(struct scsi_device *sdev, u8 *buffer,
-			      u8 *ptr, const int retries)
+spi_dv_device_compare_inquiry(struct scsi_device *sdev, struct request_queue *q,
+			      u8 *buffer, u8 *ptr, const int retries)
 {
 	int r, result;
 	const int len = sdev->inquiry_len;
+	enum scsi_device_state sdev_state = sdev->sdev_state;
 	const char spi_inquiry[] = {
 		INQUIRY, 0, 0, 0, len, 0
 	};
@@ -718,11 +719,11 @@ spi_dv_device_compare_inquiry(struct scsi_device *sdev, u8 *buffer,
 	for (r = 0; r < retries; r++) {
 		memset(ptr, 0, len);
 
-		result = spi_execute(sdev, spi_inquiry, DMA_FROM_DEVICE,
-				     ptr, len, NULL);
+		result = spi_execute(q, spi_inquiry, DMA_FROM_DEVICE, ptr, len,
+				     NULL);
 		
 		if(result || !scsi_device_online(sdev)) {
-			scsi_device_set_state(sdev, SDEV_QUIESCE);
+			scsi_device_set_state(sdev, sdev_state);
 			return SPI_COMPARE_FAILURE;
 		}
 
@@ -742,9 +743,10 @@ spi_dv_device_compare_inquiry(struct scsi_device *sdev, u8 *buffer,
 }
 
 static enum spi_compare_returns
-spi_dv_retrain(struct scsi_device *sdev, u8 *buffer, u8 *ptr,
-	       enum spi_compare_returns 
-	       (*compare_fn)(struct scsi_device *, u8 *, u8 *, int))
+spi_dv_retrain(struct scsi_device *sdev, struct request_queue *q, u8 *buffer,
+	       u8 *ptr, enum spi_compare_returns
+	       (*compare_fn)(struct scsi_device *, struct request_queue *,
+			     u8 *, u8 *, int))
 {
 	struct spi_internal *i = to_spi_internal(sdev->host->transportt);
 	struct scsi_target *starget = sdev->sdev_target;
@@ -754,7 +756,7 @@ spi_dv_retrain(struct scsi_device *sdev, u8 *buffer, u8 *ptr,
 
 	for (;;) {
 		int newperiod;
-		retval = compare_fn(sdev, buffer, ptr, DV_LOOPS);
+		retval = compare_fn(sdev, q, buffer, ptr, DV_LOOPS);
 
 		if (retval == SPI_COMPARE_SUCCESS
 		    || retval == SPI_COMPARE_SKIP_TEST)
@@ -800,7 +802,8 @@ spi_dv_retrain(struct scsi_device *sdev, u8 *buffer, u8 *ptr,
 }
 
 static int
-spi_dv_device_get_echo_buffer(struct scsi_device *sdev, u8 *buffer)
+spi_dv_device_get_echo_buffer(struct scsi_device *sdev,
+			      struct request_queue *q, u8 *buffer)
 {
 	int l, result;
 
@@ -824,8 +827,8 @@ spi_dv_device_get_echo_buffer(struct scsi_device *sdev, u8 *buffer)
 	 * (reservation conflict, device not ready, etc) just
 	 * skip the write tests */
 	for (l = 0; ; l++) {
-		result = spi_execute(sdev, spi_test_unit_ready, DMA_NONE, 
-				     NULL, 0, NULL);
+		result = spi_execute(q, spi_test_unit_ready, DMA_NONE, NULL, 0,
+				     NULL);
 
 		if(result) {
 			if(l >= 3)
@@ -836,8 +839,8 @@ spi_dv_device_get_echo_buffer(struct scsi_device *sdev, u8 *buffer)
 		}
 	}
 
-	result = spi_execute(sdev, spi_read_buffer_descriptor, 
-			     DMA_FROM_DEVICE, buffer, 4, NULL);
+	result = spi_execute(q, spi_read_buffer_descriptor, DMA_FROM_DEVICE,
+			     buffer, 4, NULL);
 
 	if (result)
 		/* Device has no echo buffer */
@@ -847,7 +850,8 @@ spi_dv_device_get_echo_buffer(struct scsi_device *sdev, u8 *buffer)
 }
 
 static void
-spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
+spi_dv_device_internal(struct scsi_device *sdev, struct request_queue *q,
+		       u8 *buffer)
 {
 	struct spi_internal *i = to_spi_internal(sdev->host->transportt);
 	struct scsi_target *starget = sdev->sdev_target;
@@ -859,7 +863,7 @@ spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
 	DV_SET(offset, 0);
 	DV_SET(width, 0);
 
-	if (spi_dv_device_compare_inquiry(sdev, buffer, buffer, DV_LOOPS)
+	if (spi_dv_device_compare_inquiry(sdev, q, buffer, buffer, DV_LOOPS)
 	    != SPI_COMPARE_SUCCESS) {
 		starget_printk(KERN_ERR, starget, "Domain Validation Initial Inquiry Failed\n");
 		/* FIXME: should probably offline the device here? */
@@ -875,9 +879,8 @@ spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
 	if (i->f->set_width && max_width) {
 		i->f->set_width(starget, 1);
 
-		if (spi_dv_device_compare_inquiry(sdev, buffer,
-						   buffer + len,
-						   DV_LOOPS)
+		if (spi_dv_device_compare_inquiry(sdev, q, buffer, buffer + len,
+						  DV_LOOPS)
 		    != SPI_COMPARE_SUCCESS) {
 			starget_printk(KERN_ERR, starget, "Wide Transfers Fail\n");
 			i->f->set_width(starget, 0);
@@ -946,7 +949,7 @@ spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
 	DV_SET(width, max_width);
 
 	/* Do the read only INQUIRY tests */
-	spi_dv_retrain(sdev, buffer, buffer + sdev->inquiry_len,
+	spi_dv_retrain(sdev, q, buffer, buffer + sdev->inquiry_len,
 		       spi_dv_device_compare_inquiry);
 	/* See if we actually managed to negotiate and sustain DT */
 	if (i->f->get_dt)
@@ -958,7 +961,7 @@ spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
 	 * negotiated DT */
 
 	if (len == -1 && spi_dt(starget))
-		len = spi_dv_device_get_echo_buffer(sdev, buffer);
+		len = spi_dv_device_get_echo_buffer(sdev, q, buffer);
 
 	if (len <= 0) {
 		starget_printk(KERN_INFO, starget, "Domain Validation skipping write tests\n");
@@ -970,7 +973,7 @@ spi_dv_device_internal(struct scsi_device *sdev, u8 *buffer)
 		len = SPI_MAX_ECHO_BUFFER_SIZE;
 	}
 
-	if (spi_dv_retrain(sdev, buffer, buffer + len,
+	if (spi_dv_retrain(sdev, q, buffer, buffer + len,
 			   spi_dv_device_echo_buffer)
 	    == SPI_COMPARE_SKIP_TEST) {
 		/* OK, the stupid drive can't do a write echo buffer
@@ -1030,7 +1033,7 @@ spi_dv_device(struct scsi_device *sdev)
 
 	starget_printk(KERN_INFO, starget, "Beginning Domain Validation\n");
 
-	spi_dv_device_internal(sdev, buffer);
+	spi_dv_device_internal(sdev, sdev->request_queue, buffer);
 
 	starget_printk(KERN_INFO, starget, "Ending Domain Validation\n");
 
#11 lore raw

PATCH v3 7/9 scsi_transport_spi: Freeze request queues instead of quiescing

In reply to Bart Van Assche

Instead of quiescing the request queues involved in domain validation, freeze these. As a result, the struct request_queue pm_only member is no longer set during domain validation. That will allow to modify scsi_execute() such that it stops setting the BLK_MQ_REQ_PREEMPT flag. Three additional changes in this patch are that scsi_mq_alloc_queue() is exported, that scsi_device_quiesce() is no longer exported and that scsi_target_{quiesce,resume}() have been changed into scsi_target_{freeze,unfreeze}().

Cc James Bottomley <[email protected]>
Cc Woody Suwalski <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
 drivers/scsi/scsi_lib.c           | 21 ++++----
 drivers/scsi/scsi_priv.h          |  2 +
 drivers/scsi/scsi_transport_spi.c | 84 +++++++++++++++++++++----------
 include/scsi/scsi_device.h        |  6 +--
 4 files changed, 72 insertions(+), 41 deletions(-)
drivers/scsi/scsi_lib.c +10 -11
index b5449efc7283..fef4708f3778 100644
@@ -2555,7 +2555,6 @@ scsi_device_quiesce(struct scsi_device *sdev)
 
 	return err;
 }
-EXPORT_SYMBOL(scsi_device_quiesce);
 
 /**
  *	scsi_device_resume - Restart user issued commands to a quiesced device.
@@ -2584,30 +2583,30 @@ void scsi_device_resume(struct scsi_device *sdev)
 EXPORT_SYMBOL(scsi_device_resume);
 
 static void
-device_quiesce_fn(struct scsi_device *sdev, void *data)
+device_freeze_fn(struct scsi_device *sdev, void *data)
 {
-	scsi_device_quiesce(sdev);
+	blk_mq_freeze_queue(sdev->request_queue);
 }
 
 void
-scsi_target_quiesce(struct scsi_target *starget)
+scsi_target_freeze(struct scsi_target *starget)
 {
-	starget_for_each_device(starget, NULL, device_quiesce_fn);
+	starget_for_each_device(starget, NULL, device_freeze_fn);
 }
-EXPORT_SYMBOL(scsi_target_quiesce);
+EXPORT_SYMBOL(scsi_target_freeze);
 
 static void
-device_resume_fn(struct scsi_device *sdev, void *data)
+device_unfreeze_fn(struct scsi_device *sdev, void *data)
 {
-	scsi_device_resume(sdev);
+	blk_mq_unfreeze_queue(sdev->request_queue);
 }
 
 void
-scsi_target_resume(struct scsi_target *starget)
+scsi_target_unfreeze(struct scsi_target *starget)
 {
-	starget_for_each_device(starget, NULL, device_resume_fn);
+	starget_for_each_device(starget, NULL, device_unfreeze_fn);
 }
-EXPORT_SYMBOL(scsi_target_resume);
+EXPORT_SYMBOL(scsi_target_unfreeze);
 
 /**
  * scsi_internal_device_block_nowait - try to transition to the SDEV_BLOCK state
drivers/scsi/scsi_priv.h +2 -0
index e34755986b47..18485595762a 100644
@@ -95,6 +95,8 @@ extern int scsi_mq_setup_tags(struct Scsi_Host *shost);
 extern void scsi_mq_destroy_tags(struct Scsi_Host *shost);
 extern void scsi_exit_queue(void);
 extern void scsi_evt_thread(struct work_struct *work);
+extern int scsi_device_quiesce(struct scsi_device *sdev);
+extern void scsi_device_resume(struct scsi_device *sdev);
 struct request_queue;
 struct request;
 
drivers/scsi/scsi_transport_spi.c +58 -26

123 lines not on this page · raw message

include/scsi/scsi_device.h +2 -4
index f47fdf9cf788..dc193d7f479a 100644
@@ -423,10 +423,8 @@ extern struct scsi_event *sdev_evt_alloc(enum scsi_device_event evt_type,
 extern void sdev_evt_send(struct scsi_device *sdev, struct scsi_event *evt);
 extern void sdev_evt_send_simple(struct scsi_device *sdev,
 			  enum scsi_device_event evt_type, gfp_t gfpflags);
-extern int scsi_device_quiesce(struct scsi_device *sdev);
-extern void scsi_device_resume(struct scsi_device *sdev);
-extern void scsi_target_quiesce(struct scsi_target *);
-extern void scsi_target_resume(struct scsi_target *);
+extern void scsi_target_freeze(struct scsi_target *);
+extern void scsi_target_unfreeze(struct scsi_target *);
 extern void scsi_scan_target(struct device *parent, unsigned int channel,
 			     unsigned int id, u64 lun,
 			     enum scsi_scan_mode rescan);
#12 lore raw

Re: [PATCH v3 7/9] scsi_transport_spi: Freeze request queues instead of quiescing

Quoted 9 lines · Bart Van Assche

Instead of quiescing the request queues involved in domain validation, freeze these. As a result, the struct request_queue pm_only member is no longer set during domain validation. That will allow to modify scsi_execute() such that it stops setting the BLK_MQ_REQ_PREEMPT flag. Three additional changes in this patch are that scsi_mq_alloc_queue() is exported, that scsi_device_quiesce() is no longer exported and that scsi_target_{quiesce,resume}() have been changed into scsi_target_{freeze,unfreeze}().

The description is now partially wrong, as scsi_mq_alloc_queue() is certainly not exported (why would it?).

And it also glosses over the main idea of this patch, namely that not only sdev->request_queue is frozen, but also a completely _new_ request queue is allocated to send SPI DV requests over.

Please modify the description.

Quoted 165 lines
Cc James Bottomley <[email protected]>
Cc Woody Suwalski <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
  drivers/scsi/scsi_lib.c           | 21 ++++----
  drivers/scsi/scsi_priv.h          |  2 +
  drivers/scsi/scsi_transport_spi.c | 84 +++++++++++++++++++++----------
  include/scsi/scsi_device.h        |  6 +--
  4 files changed, 72 insertions(+), 41 deletions(-)
drivers/scsi/scsi_lib.c +10 -11
index b5449efc7283..fef4708f3778 100644
@@ -2555,7 +2555,6 @@ scsi_device_quiesce(struct scsi_device *sdev)
 
  	return err;
  }
-EXPORT_SYMBOL(scsi_device_quiesce);
 
  /**
   *	scsi_device_resume - Restart user issued commands to a quiesced device.
@@ -2584,30 +2583,30 @@ void scsi_device_resume(struct scsi_device *sdev)
  EXPORT_SYMBOL(scsi_device_resume);
 
  static void
-device_quiesce_fn(struct scsi_device *sdev, void *data)
+device_freeze_fn(struct scsi_device *sdev, void *data)
  {
-	scsi_device_quiesce(sdev);
+	blk_mq_freeze_queue(sdev->request_queue);
  }
 
  void
-scsi_target_quiesce(struct scsi_target *starget)
+scsi_target_freeze(struct scsi_target *starget)
  {
-	starget_for_each_device(starget, NULL, device_quiesce_fn);
+	starget_for_each_device(starget, NULL, device_freeze_fn);
  }
-EXPORT_SYMBOL(scsi_target_quiesce);
+EXPORT_SYMBOL(scsi_target_freeze);
 
  static void
-device_resume_fn(struct scsi_device *sdev, void *data)
+device_unfreeze_fn(struct scsi_device *sdev, void *data)
  {
-	scsi_device_resume(sdev);
+	blk_mq_unfreeze_queue(sdev->request_queue);
  }
 
  void
-scsi_target_resume(struct scsi_target *starget)
+scsi_target_unfreeze(struct scsi_target *starget)
  {
-	starget_for_each_device(starget, NULL, device_resume_fn);
+	starget_for_each_device(starget, NULL, device_unfreeze_fn);
  }
-EXPORT_SYMBOL(scsi_target_resume);
+EXPORT_SYMBOL(scsi_target_unfreeze);
 
  /**
   * scsi_internal_device_block_nowait - try to transition to the SDEV_BLOCK state
drivers/scsi/scsi_priv.h +2 -0
index e34755986b47..18485595762a 100644
@@ -95,6 +95,8 @@ extern int scsi_mq_setup_tags(struct Scsi_Host *shost);
  extern void scsi_mq_destroy_tags(struct Scsi_Host *shost);
  extern void scsi_exit_queue(void);
  extern void scsi_evt_thread(struct work_struct *work);
+extern int scsi_device_quiesce(struct scsi_device *sdev);
+extern void scsi_device_resume(struct scsi_device *sdev);
  struct request_queue;
  struct request;
 
drivers/scsi/scsi_transport_spi.c +31 -17
index 959990f66865..2352c441302f 100644
@@ -983,6 +983,18 @@ spi_dv_device_internal(struct scsi_device *sdev, struct request_queue *q,
  	}
  }
 
+static struct request_queue *scsi_mq_alloc_queue(struct scsi_device *sdev)
+{
+	struct request_queue *q = blk_mq_init_queue(&sdev->host->tag_set);
+
+	if (IS_ERR(q))
+		return NULL;
+
+	q->queuedata = sdev;
+	__scsi_init_queue(sdev->host, q);
+	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, q);
+	return q;
+}
 
  /**	spi_dv_device - Do Domain Validation on the device
   *	@sdev:		scsi device to validate
@@ -997,59 +1009,79 @@ void
  spi_dv_device(struct scsi_device *sdev)
  {
  	struct scsi_target *starget = sdev->sdev_target;
+	struct request_queue *q2;
  	u8 *buffer;
  	const int len = SPI_MAX_ECHO_BUFFER_SIZE*2;
 
  	/*
-	 * Because this function and the power management code both call
-	 * scsi_device_quiesce(), it is not safe to perform domain validation
-	 * while suspend or resume is in progress. Hence the
-	 * lock/unlock_system_sleep() calls.
+	 * Because this function creates a new request queue that is not
+	 * visible to the rest of the system, this function must be serialized
+	 * against suspend, resume and runtime power management. Hence the
+	 * lock/unlock_system_sleep() and scsi_autopm_{get,put}_device()
+	 * calls.
  	 */
  	lock_system_sleep();
 
+	if (scsi_autopm_get_device(sdev))
+		goto unlock_system_sleep;
+
  	if (unlikely(spi_dv_in_progress(starget)))
-		goto unlock;
+		goto put_autopm;
 
  	if (unlikely(scsi_device_get(sdev)))
-		goto unlock;
-
-	spi_dv_in_progress(starget) = 1;
+		goto put_autopm;
 
  	buffer = kzalloc(len, GFP_KERNEL);
 
  	if (unlikely(!buffer))
-		goto out_put;
-
-	/* We need to verify that the actual device will quiesce; the
-	 * later target quiesce is just a nice to have */
-	if (unlikely(scsi_device_quiesce(sdev)))
-		goto out_free;
-
-	scsi_target_quiesce(starget);
+		goto put_sdev;
 
  	spi_dv_pending(starget) = 1;
+
  	mutex_lock(&spi_dv_mutex(starget));
+	if (unlikely(spi_dv_in_progress(starget)))
+		goto clear_pending;
+
+	spi_dv_in_progress(starget) = 1;
 
  	starget_printk(KERN_INFO, starget, "Beginning Domain Validation\n");
 
-	spi_dv_device_internal(sdev, sdev->request_queue, buffer);
+	q2 = scsi_mq_alloc_queue(sdev);
+

drivers/scsi/scsi_transport_spi.c void · line 1060

Quoted 10 lines
drivers/scsi/scsi_transport_spi.c +10 -0
+	if (q2) {
+		/*
+		 * Freeze the target such that no other subsystem can submit
+		 * SCSI commands to 'sdev'. Submitting SCSI commands through
+		 * q2 may trigger the SCSI error handler. The SCSI error
+		 * handler must be able to handle a frozen sdev->request_queue
+		 * and must also use blk_mq_rq_from_pdu(q2)->q instead of
+		 * sdev->request_queue if it would be necessary to access q2
+		 * directly.
+		 */

... 'it would be necessary' indicates that it doesn't do so, currently. And the SPI DV code most certainly _is_ submitting SCSI commands. So doesn't the above imply that SCSI EH will fail to work correctly? And if so, why isn't it fixed with some later patch in this series? Or how do you plan to address it?
Hmm?

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke                Kernel Storage Architect
[email protected]                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
author Reviewed-by Alan Stern Tested-by Stan Johnson
#13 lore raw

Re: [PATCH v3 7/9] scsi_transport_spi: Freeze request queues instead of quiescing

Quoted 10 lines · Hannes Reinecke
Quoted 9 lines · Bart Van Assche

Instead of quiescing the request queues involved in domain validation, freeze these. As a result, the struct request_queue pm_only member is no longer set during domain validation. That will allow to modify scsi_execute() such that it stops setting the BLK_MQ_REQ_PREEMPT flag. Three additional changes in this patch are that scsi_mq_alloc_queue() is exported, that scsi_device_quiesce() is no longer exported and that scsi_target_{quiesce,resume}() have been changed into scsi_target_{freeze,unfreeze}().

Quoted 8 lines

The description is now partially wrong, as scsi_mq_alloc_queue() is certainly not exported (why would it?).

And it also glosses over the main idea of this patch, namely that not only sdev->request_queue is frozen, but also a completely _new_ request queue is allocated to send SPI DV requests over.

Please modify the description.

Hi Hannes,

Thanks for the feedback. Please take a look at the attached patch.

Quoted 13 lines
Quoted 4 lines

+    q2 = scsi_mq_alloc_queue(sdev);
+
+    if (q2) {
+        /*

Quoted 8 lines

+         * Freeze the target such that no other subsystem can submit +         * SCSI commands to 'sdev'. Submitting SCSI commands through +         * q2 may trigger the SCSI error handler. The SCSI error +         * handler must be able to handle a frozen sdev->request_queue +         * and must also use blk_mq_rq_from_pdu(q2)->q instead of +         * sdev->request_queue if it would be necessary to access q2 +         * directly.
+         */

Quoted 6 lines

... 'it would be necessary' indicates that it doesn't do so, currently. And the SPI DV code most certainly _is_ submitting SCSI commands. So doesn't the above imply that SCSI EH will fail to work correctly? And if so, why isn't it fixed with some later patch in this series? Or how do you plan to address it?
Hmm?

Before SPI domain validation starts, sdev->request_queue is frozen which means that DV will only start after all pending SCSI commands have finished, including potential error handling for these commands.

As far as I know there is only one sdev->request_queue access in the SCSI error handler, namely in scsi_eh_lock_door(). Patch 5/9 from this series makes the blk_get_request() call in that function use BLK_MQ_REQ_NOWAIT so the SCSI error handler should still work fine if a DV command fails. Does this answer your question?

Thanks,

Bart.

From 66156c468f418da09512ef321913627ddf0da20d Mon Sep 17 00:00:00 2001
From: Bart Van Assche <[email protected]>
Date: Sun, 30 Aug 2020 09:35:23 -0700
Subject: [PATCH] scsi_transport_spi: Freeze request queues instead of
quiescing

Instead of quiescing the request queues involved in domain validation, freeze these. As a result, the struct request_queue pm_only member is no longer set during domain validation. That will allow to modify scsi_execute() such that it stops setting the BLK_MQ_REQ_PREEMPT flag.

Attach a secondary request queue for submitting DV requests while the primary request queue is frozen.

Change scsi_target_{quiesce,resume}() into scsi_target_{freeze,unfreeze}().

Unexport scsi_device_quiesce().

Cc James Bottomley <[email protected]>
Cc Woody Suwalski <[email protected]>
Cc Can Guo <[email protected]>
Cc Stanley Chu <[email protected]>
Cc Ming Lei <[email protected]>
Cc Rafael J. Wysocki <[email protected]>
---
 drivers/scsi/scsi_lib.c           | 21 ++++----
 drivers/scsi/scsi_priv.h          |  2 +
 drivers/scsi/scsi_transport_spi.c | 88 ++++++++++++++++++++++---------
 include/scsi/scsi_device.h        |  6 +--
 4 files changed, 76 insertions(+), 41 deletions(-)
drivers/scsi/scsi_lib.c +10 -11
index b5449efc7283..fef4708f3778 100644
@@ -2555,7 +2555,6 @@ scsi_device_quiesce(struct scsi_device *sdev)
 
 	return err;
 }
-EXPORT_SYMBOL(scsi_device_quiesce);
 
 /**
  *	scsi_device_resume - Restart user issued commands to a quiesced device.
@@ -2584,30 +2583,30 @@ void scsi_device_resume(struct scsi_device *sdev)
 EXPORT_SYMBOL(scsi_device_resume);
 
 static void
-device_quiesce_fn(struct scsi_device *sdev, void *data)
+device_freeze_fn(struct scsi_device *sdev, void *data)
 {
-	scsi_device_quiesce(sdev);
+	blk_mq_freeze_queue(sdev->request_queue);
 }
 
 void
-scsi_target_quiesce(struct scsi_target *starget)
+scsi_target_freeze(struct scsi_target *starget)
 {
-	starget_for_each_device(starget, NULL, device_quiesce_fn);
+	starget_for_each_device(starget, NULL, device_freeze_fn);
 }
-EXPORT_SYMBOL(scsi_target_quiesce);
+EXPORT_SYMBOL(scsi_target_freeze);
 
 static void
-device_resume_fn(struct scsi_device *sdev, void *data)
+device_unfreeze_fn(struct scsi_device *sdev, void *data)
 {
-	scsi_device_resume(sdev);
+	blk_mq_unfreeze_queue(sdev->request_queue);
 }
 
 void
-scsi_target_resume(struct scsi_target *starget)
+scsi_target_unfreeze(struct scsi_target *starget)
 {
-	starget_for_each_device(starget, NULL, device_resume_fn);
+	starget_for_each_device(starget, NULL, device_unfreeze_fn);
 }
-EXPORT_SYMBOL(scsi_target_resume);
+EXPORT_SYMBOL(scsi_target_unfreeze);
 
 /**
  * scsi_internal_device_block_nowait - try to transition to the SDEV_BLOCK state
drivers/scsi/scsi_priv.h +2 -0
index e34755986b47..18485595762a 100644
@@ -95,6 +95,8 @@ extern int scsi_mq_setup_tags(struct Scsi_Host *shost);
 extern void scsi_mq_destroy_tags(struct Scsi_Host *shost);
 extern void scsi_exit_queue(void);
 extern void scsi_evt_thread(struct work_struct *work);
+extern int scsi_device_quiesce(struct scsi_device *sdev);
+extern void scsi_device_resume(struct scsi_device *sdev);
 struct request_queue;
 struct request;
 
drivers/scsi/scsi_transport_spi.c +62 -26

127 lines not on this page · raw message

include/scsi/scsi_device.h +2 -4
index f47fdf9cf788..dc193d7f479a 100644
@@ -423,10 +423,8 @@ extern struct scsi_event *sdev_evt_alloc(enum scsi_device_event evt_type,
 extern void sdev_evt_send(struct scsi_device *sdev, struct scsi_event *evt);
 extern void sdev_evt_send_simple(struct scsi_device *sdev,
 			  enum scsi_device_event evt_type, gfp_t gfpflags);
-extern int scsi_device_quiesce(struct scsi_device *sdev);
-extern void scsi_device_resume(struct scsi_device *sdev);
-extern void scsi_target_quiesce(struct scsi_target *);
-extern void scsi_target_resume(struct scsi_target *);
+extern void scsi_target_freeze(struct scsi_target *);
+extern void scsi_target_unfreeze(struct scsi_target *);
 extern void scsi_scan_target(struct device *parent, unsigned int channel,
 			     unsigned int id, u64 lun,
 			     enum scsi_scan_mode rescan);
#14 lore raw

Re: [PATCH v3 7/9] scsi_transport_spi: Freeze request queues instead of quiescing

Quoted 48 lines · Bart Van Assche
Quoted 10 lines · Hannes Reinecke
Quoted 9 lines · Bart Van Assche

Instead of quiescing the request queues involved in domain validation, freeze these. As a result, the struct request_queue pm_only member is no longer set during domain validation. That will allow to modify scsi_execute() such that it stops setting the BLK_MQ_REQ_PREEMPT flag. Three additional changes in this patch are that scsi_mq_alloc_queue() is exported, that scsi_device_quiesce() is no longer exported and that scsi_target_{quiesce,resume}() have been changed into scsi_target_{freeze,unfreeze}().

Quoted 8 lines

The description is now partially wrong, as scsi_mq_alloc_queue() is certainly not exported (why would it?).

And it also glosses over the main idea of this patch, namely that not only sdev->request_queue is frozen, but also a completely _new_ request queue is allocated to send SPI DV requests over.

Please modify the description.

Hi Hannes,

Thanks for the feedback. Please take a look at the attached patch.

Quoted 13 lines
Quoted 4 lines

+    q2 = scsi_mq_alloc_queue(sdev);
+
+    if (q2) {
+        /*

Quoted 8 lines

+         * Freeze the target such that no other subsystem can submit +         * SCSI commands to 'sdev'. Submitting SCSI commands through +         * q2 may trigger the SCSI error handler. The SCSI error +         * handler must be able to handle a frozen sdev->request_queue +         * and must also use blk_mq_rq_from_pdu(q2)->q instead of +         * sdev->request_queue if it would be necessary to access q2 +         * directly.
+         */

Quoted 6 lines

... 'it would be necessary' indicates that it doesn't do so, currently. And the SPI DV code most certainly _is_ submitting SCSI commands. So doesn't the above imply that SCSI EH will fail to work correctly? And if so, why isn't it fixed with some later patch in this series? Or how do you plan to address it?
Hmm?

Before SPI domain validation starts, sdev->request_queue is frozen which means that DV will only start after all pending SCSI commands have finished, including potential error handling for these commands.

Quoted 6 lines

As far as I know there is only one sdev->request_queue access in the SCSI error handler, namely in scsi_eh_lock_door(). Patch 5/9 from this series makes the blk_get_request() call in that function use BLK_MQ_REQ_NOWAIT so the SCSI error handler should still work fine if a DV command fails. Does this answer your question?

My argument wasn't so much about scsi_eh_lock_door() (which will be sent only in rare cases), but rather scsi_eh_try_stu() and scsi_eh_tur() as invoked from scsi_eh_test_devices() at the end of each SCSI EH step. Both of which would need similar treatment.

Plus one maybe should revisit scsi_end_request(); that does refer to

	struct request_queue *q = sdev->request_queue;

despite having a 'struct request' as argument; maybe one should modify that to

	struct request_queue *q = req->q;

to not start accounting on the wrong queue.

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke                Kernel Storage Architect
[email protected]                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
#15 lore raw

PATCH v3 8/9 block, scsi, ide: Only process PM requests if rpm_status != RPM_ACTIVE

In reply to Bart Van Assche

Instead of submitting all SCSI commands submitted with scsi_execute() to a SCSI device if rpm_status != RPM_ACTIVE, only submit RQF_PM (power management requests) if rpm_status != RPM_ACTIVE. Remove flag RQF_PREEMPT since it is no longer necessary.

---
 block/blk-core.c        |  6 +++---
 block/blk-mq-debugfs.c  |  1 -
 block/blk-mq.c          |  4 ++--
 drivers/ide/ide-io.c    |  3 +--
 drivers/ide/ide-pm.c    |  2 +-
 drivers/scsi/scsi_lib.c | 27 ++++++++++++++-------------
 include/linux/blk-mq.h  |  4 ++--
 include/linux/blkdev.h  |  6 +-----
 8 files changed, 24 insertions(+), 29 deletions(-)
block/blk-core.c +3 -3
index 2db8bda43b6e..a00bce9f46d8 100644
@@ -424,11 +424,11 @@ EXPORT_SYMBOL(blk_cleanup_queue);
 /**
  * blk_queue_enter() - try to increase q->q_usage_counter
  * @q: request queue pointer
- * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PREEMPT
+ * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PM
  */
 int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 {
-	const bool pm = flags & BLK_MQ_REQ_PREEMPT;
+	const bool pm = flags & BLK_MQ_REQ_PM;
 
 	while (true) {
 		bool success = false;
@@ -630,7 +630,7 @@ struct request *blk_get_request(struct request_queue *q, unsigned int op,
 	struct request *req;
 
 	WARN_ON_ONCE(op & REQ_NOWAIT);
-	WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PREEMPT));
+	WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PM));
 
 	req = blk_mq_alloc_request(q, op, flags);
 	if (!IS_ERR(req) && q->mq_ops->initialize_rq_fn)
block/blk-mq-debugfs.c +0 -1
index 3094542e12ae..9336a6f8d6ef 100644
@@ -297,7 +297,6 @@ static const char *const rqf_name[] = {
 	RQF_NAME(MIXED_MERGE),
 	RQF_NAME(MQ_INFLIGHT),
 	RQF_NAME(DONTPREP),
-	RQF_NAME(PREEMPT),
 	RQF_NAME(FAILED),
 	RQF_NAME(QUIET),
 	RQF_NAME(ELVPRIV),
block/blk-mq.c +2 -2
index 1b25ec2fe9be..d50504888b68 100644
@@ -292,8 +292,8 @@ static struct request *blk_mq_rq_ctx_init(struct blk_mq_alloc_data *data,
 	rq->mq_hctx = data->hctx;
 	rq->rq_flags = 0;
 	rq->cmd_flags = data->cmd_flags;
-	if (data->flags & BLK_MQ_REQ_PREEMPT)
-		rq->rq_flags |= RQF_PREEMPT;
+	if (data->flags & BLK_MQ_REQ_PM)
+		rq->rq_flags |= RQF_PM;
 	if (blk_queue_io_stat(data->q))
 		rq->rq_flags |= RQF_IO_STAT;
 	INIT_LIST_HEAD(&rq->queuelist);
drivers/ide/ide-io.c +1 -2
index 1a53c7a75224..beb850679fa9 100644
@@ -522,8 +522,7 @@ blk_status_t ide_issue_rq(ide_drive_t *drive, struct request *rq,
 		 * state machine.
 		 */
 		if ((drive->dev_flags & IDE_DFLAG_BLOCKED) &&
-		    ata_pm_request(rq) == 0 &&
-		    (rq->rq_flags & RQF_PREEMPT) == 0) {
+		    ata_pm_request(rq) == 0) {
 			/* there should be no pending command at this point */
 			ide_unlock_port(hwif);
 			goto plug_device;
drivers/ide/ide-pm.c +1 -1
index 192e6c65d34e..82ab308f1aaf 100644
@@ -77,7 +77,7 @@ int generic_ide_resume(struct device *dev)
 	}
 
 	memset(&rqpm, 0, sizeof(rqpm));
-	rq = blk_get_request(drive->queue, REQ_OP_DRV_IN, BLK_MQ_REQ_PREEMPT);
+	rq = blk_get_request(drive->queue, REQ_OP_DRV_IN, BLK_MQ_REQ_PM);
 	ide_req(rq)->type = ATA_PRIV_PM_RESUME;
 	ide_req(rq)->special = &rqpm;
 	rqpm.pm_step = IDE_PM_START_RESUME;
drivers/scsi/scsi_lib.c +14 -13
index fef4708f3778..08f99573c383 100644
@@ -248,7 +248,8 @@ int __scsi_execute(struct request_queue *q, const unsigned char *cmd,
 	int ret = DRIVER_ERROR << 24;
 
 	req = blk_get_request(q, data_direction == DMA_TO_DEVICE ?
-			REQ_OP_SCSI_OUT : REQ_OP_SCSI_IN, BLK_MQ_REQ_PREEMPT);
+			      REQ_OP_SCSI_OUT : REQ_OP_SCSI_IN,
+			      rq_flags & RQF_PM ? BLK_MQ_REQ_PM : 0);
 	if (IS_ERR(req))
 		return ret;
 	rq = scsi_req(req);
@@ -1204,6 +1205,8 @@ static blk_status_t
 scsi_device_state_check(struct scsi_device *sdev, struct request *req)
 {
 	switch (sdev->sdev_state) {
+	case SDEV_CREATED:
+		return BLK_STS_OK;
 	case SDEV_OFFLINE:
 	case SDEV_TRANSPORT_OFFLINE:
 		/*
@@ -1230,18 +1233,18 @@ scsi_device_state_check(struct scsi_device *sdev, struct request *req)
 		return BLK_STS_RESOURCE;
 	case SDEV_QUIESCE:
 		/*
-		 * If the devices is blocked we defer normal commands.
+		 * If the device is blocked we only accept power management
+		 * commands.
 		 */
-		if (req && !(req->rq_flags & RQF_PREEMPT))
+		if (req && WARN_ON_ONCE(!(req->rq_flags & RQF_PM)))
 			return BLK_STS_RESOURCE;
 		return BLK_STS_OK;
 	default:
 		/*
 		 * For any other not fully online state we only allow
-		 * special commands.  In particular any user initiated
-		 * command is not allowed.
+		 * power management commands.
 		 */
-		if (req && !(req->rq_flags & RQF_PREEMPT))
+		if (req && !(req->rq_flags & RQF_PM))
 			return BLK_STS_IOERR;
 		return BLK_STS_OK;
 	}
@@ -2503,15 +2506,13 @@ void sdev_evt_send_simple(struct scsi_device *sdev,
 EXPORT_SYMBOL_GPL(sdev_evt_send_simple);
 
 /**
- *	scsi_device_quiesce - Block user issued commands.
+ *	scsi_device_quiesce - Block all commands except power management.
  *	@sdev:	scsi device to quiesce.
  *
  *	This works by trying to transition to the SDEV_QUIESCE state
  *	(which must be a legal transition).  When the device is in this
- *	state, only special requests will be accepted, all others will
- *	be deferred.  Since special requests may also be requeued requests,
- *	a successful return doesn't guarantee the device will be
- *	totally quiescent.
+ *	state, only power management requests will be accepted, all others will
+ *	be deferred.
  *
  *	Must be called with user context, may sleep.
  *
@@ -2572,12 +2573,12 @@ void scsi_device_resume(struct scsi_device *sdev)
 	 * device deleted during suspend)
 	 */
 	mutex_lock(&sdev->state_mutex);
+	if (sdev->sdev_state == SDEV_QUIESCE)
+		scsi_device_set_state(sdev, SDEV_RUNNING);
 	if (sdev->quiesced_by) {
 		sdev->quiesced_by = NULL;
 		blk_clear_pm_only(sdev->request_queue);
 	}
-	if (sdev->sdev_state == SDEV_QUIESCE)
-		scsi_device_set_state(sdev, SDEV_RUNNING);
 	mutex_unlock(&sdev->state_mutex);
 }
 EXPORT_SYMBOL(scsi_device_resume);
include/linux/blk-mq.h +2 -2
index b23eeca4d677..1fa350592830 100644
@@ -444,8 +444,8 @@ enum {
 	BLK_MQ_REQ_NOWAIT	= (__force blk_mq_req_flags_t)(1 << 0),
 	/* allocate from reserved pool */
 	BLK_MQ_REQ_RESERVED	= (__force blk_mq_req_flags_t)(1 << 1),
-	/* set RQF_PREEMPT */
-	BLK_MQ_REQ_PREEMPT	= (__force blk_mq_req_flags_t)(1 << 3),
+	/* set RQF_PM */
+	BLK_MQ_REQ_PM		= (__force blk_mq_req_flags_t)(1 << 3),
 };
 
 struct request *blk_mq_alloc_request(struct request_queue *q, unsigned int op,
include/linux/blkdev.h +1 -5
index 639cae2c158b..7d4b746f7e6a 100644
@@ -79,9 +79,6 @@ typedef __u32 __bitwise req_flags_t;
 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
 /* don't call prep for this one */
 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
-/* set for "ide_preempt" requests and also for requests for which the SCSI
-   "quiesce" state must be ignored. */
-#define RQF_PREEMPT		((__force req_flags_t)(1 << 8))
 /* vaguely specified driver internal error.  Ignored by the block layer */
 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
 /* don't warn about errors */
@@ -430,8 +427,7 @@ struct request_queue {
 	unsigned long		queue_flags;
 	/*
 	 * Number of contexts that have called blk_set_pm_only(). If this
-	 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
-	 * processed.
+	 * counter is above zero then only RQF_PM requests are processed.
 	 */
 	atomic_t		pm_only;
 
#16 lore raw

PATCH v3 9/9 block: Do not accept any requests while suspended

In reply to Bart Van Assche

From: Alan Stern <[email protected]>

blk_queue_enter() accepts BLK_MQ_REQ_PREEMPT independent of the runtime power management state. Since SCSI domain validation no longer depends on this behavior, modify the behavior of blk_queue_enter() as follows:
- Do not accept any requests while suspended.
- Only process power management requests while suspending or resuming.

Submitting BLK_MQ_REQ_PREEMPT requests to a device that is runtime- suspended causes runtime-suspended block devices not to resume as they should. The request which should cause a runtime resume instead gets issued directly, without resuming the device first. Of course the device can't handle it properly, the I/O fails, and the device remains suspended.

The problem is fixed by checking that the queue's runtime-PM status isn't RPM_SUSPENDED before allowing a request to be issued, and queuing a runtime-resume request if it is. In particular, the inline blk_pm_request_resume() routine is renamed blk_pm_resume_queue() and the code is unified by merging the surrounding checks into the routine. If the queue isn't set up for runtime PM, or there currently is no restriction on allowed requests, the request is allowed. Likewise if the BLK_MQ_REQ_PREEMPT flag is set and the status isn't RPM_SUSPENDED. Otherwise a runtime resume is queued and the request is blocked until conditions are more suitable.

[ bvanassche: modified commit message and removed Cc: stable because without
the previous patches from this series this patch would break parallel SCSI domain validation ]

---
 block/blk-core.c |  6 +++---
 block/blk-pm.h   | 14 +++++++++-----
 2 files changed, 12 insertions(+), 8 deletions(-)
block/blk-core.c +3 -3
index a00bce9f46d8..230880cbf8c8 100644
@@ -440,7 +440,8 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 			 * responsible for ensuring that that counter is
 			 * globally visible before the queue is unfrozen.
 			 */
-			if (pm || !blk_queue_pm_only(q)) {
+			if ((pm && q->rpm_status != RPM_SUSPENDED) ||
+			    !blk_queue_pm_only(q)) {
 				success = true;
 			} else {
 				percpu_ref_put(&q->q_usage_counter);
@@ -465,8 +466,7 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
 
 		wait_event(q->mq_freeze_wq,
 			   (!q->mq_freeze_depth &&
-			    (pm || (blk_pm_request_resume(q),
-				    !blk_queue_pm_only(q)))) ||
+			    blk_pm_resume_queue(pm, q)) ||
 			   blk_queue_dying(q));
 		if (blk_queue_dying(q))
 			return -ENODEV;
block/blk-pm.h +9 -5
index ea5507d23e75..a2283cc9f716 100644
@@ -6,11 +6,14 @@
 #include <linux/pm_runtime.h>
 
 #ifdef CONFIG_PM
-static inline void blk_pm_request_resume(struct request_queue *q)
+static inline int blk_pm_resume_queue(const bool pm, struct request_queue *q)
 {
-	if (q->dev && (q->rpm_status == RPM_SUSPENDED ||
-		       q->rpm_status == RPM_SUSPENDING))
-		pm_request_resume(q->dev);
+	if (!q->dev || !blk_queue_pm_only(q))
+		return 1;	/* Nothing to do */
+	if (pm && q->rpm_status != RPM_SUSPENDED)
+		return 1;	/* Request allowed */
+	pm_request_resume(q->dev);
+	return 0;
 }
 
 static inline void blk_pm_mark_last_busy(struct request *rq)
@@ -44,8 +47,9 @@ static inline void blk_pm_put_request(struct request *rq)
 		--rq->q->nr_pending;
 }
 #else
-static inline void blk_pm_request_resume(struct request_queue *q)
+static inline int blk_pm_resume_queue(const bool pm, struct request_queue *q)
 {
+	return 1;
 }
 
 static inline void blk_pm_mark_last_busy(struct request *rq)