[PATCH 00/11] scsi_debug: Some minor improvements

[PATCH 00/11] scsi_debug: Some minor improvements

[John Garry]

This series contains a bunch of minor improvements to the driver. I have
another bunch waiting with more major changes.

Most of the changes are quite straightforward, and the only patches of note are
as follows:
- Fix the command abort feature, enabled with host option SDEBUG_OPT_CMD_ABORT
- Drop driver count of queued commands per device
- Add poll mode completions to statistics. We already have poll mode callback
  call count, so maybe it was intentional to omit poll mode from the
  statistics.
  
Based on scsi-staging 6.4 @ commit ffba85603c62 ("Merge patch series "lpfc: Update lpfc to revision 14.2.0.11")

John Garry (11):
  scsi: scsi_debug: Don't hold driver host struct pointer in
    host->hostdata[]
  scsi: scsi_debug: Stop setting devip->sdbg_host twice
  scsi: scsi_debug: Drop scsi_debug_abort() NULL pointer checks
  scsi: scsi_debug: Drop scsi_debug_device_reset() NULL pointer checks
  scsi: scsi_debug: Drop scsi_debug_target_reset() NULL pointer checks
  scsi: scsi_debug: Drop scsi_debug_bus_reset() NULL pointer checks
  scsi: scsi_debug: Drop scsi_debug_host_reset() device NULL pointer
    check
  scsi: scsi_debug: Drop check for num_in_q exceeding queue depth
  scsi: scsi_debug: Drop sdebug_dev_info.num_in_q
  scsi: scsi_debug: Get command abort feature working again
  scsi: scsi_debug: Add poll mode deferred completions to statistics

 drivers/scsi/scsi_debug.c | 201 ++++++++++++++------------------------
 1 file changed, 74 insertions(+), 127 deletions(-)

-- 
2.35.3

[PATCH 01/11] scsi: scsi_debug: Don't hold driver host struct pointer in host->hostdata[]

[John Garry]

This driver stores just a pointer to the driver host structure in
host->hostdata[]. Most other drivers actually have the driver host
structure allocated in host->hostdata[], but this driver is different as
we allocate that memory separately.

However there is no need to allocate this memory only in host->hostdata[]
when we can already look up the driver host structure from shost->dma_dev.

Also remove a check for !sdbg_host in find_build_dev_info(), as this cannot
be true. Other similar checks will be later removed.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 16 +++++++---------
 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 226139a8aaac..f713117a5f2e 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -327,6 +327,9 @@ struct sdeb_store_info {
 #define to_sdebug_host(d)	\
 	container_of(d, struct sdebug_host_info, dev)
 
+#define to_sdebug_host_from_shost(shost)	\
+	to_sdebug_host(shost->dma_dev)
+
 enum sdeb_defer_type {SDEB_DEFER_NONE = 0, SDEB_DEFER_HRT = 1,
 		      SDEB_DEFER_WQ = 2, SDEB_DEFER_POLL = 3};
 
@@ -5168,11 +5171,7 @@ static struct sdebug_dev_info *find_build_dev_info(struct scsi_device *sdev)
 	struct sdebug_dev_info *open_devip = NULL;
 	struct sdebug_dev_info *devip;
 
-	sdbg_host = *(struct sdebug_host_info **)shost_priv(sdev->host);
-	if (!sdbg_host) {
-		pr_err("Host info NULL\n");
-		return NULL;
-	}
+	sdbg_host = to_sdebug_host_from_shost(sdev->host);
 
 	list_for_each_entry(devip, &sdbg_host->dev_info_list, dev_list) {
 		if ((devip->used) && (devip->channel == sdev->channel) &&
@@ -5409,7 +5408,7 @@ static int scsi_debug_target_reset(struct scsi_cmnd *SCpnt)
 	hp = sdp->host;
 	if (!hp)
 		goto lie;
-	sdbg_host = *(struct sdebug_host_info **)shost_priv(hp);
+	sdbg_host = to_sdebug_host_from_shost(hp);
 	if (sdbg_host) {
 		list_for_each_entry(devip,
 				    &sdbg_host->dev_info_list,
@@ -5442,7 +5441,7 @@ static int scsi_debug_bus_reset(struct scsi_cmnd *SCpnt)
 		sdev_printk(KERN_INFO, sdp, "%s\n", __func__);
 	hp = sdp->host;
 	if (hp) {
-		sdbg_host = *(struct sdebug_host_info **)shost_priv(hp);
+		sdbg_host = to_sdebug_host_from_shost(hp);
 		if (sdbg_host) {
 			list_for_each_entry(devip,
 					    &sdbg_host->dev_info_list,
@@ -7821,7 +7820,7 @@ static int sdebug_driver_probe(struct device *dev)
 	if (!sdebug_clustering)
 		sdebug_driver_template.dma_boundary = PAGE_SIZE - 1;
 
-	hpnt = scsi_host_alloc(&sdebug_driver_template, sizeof(sdbg_host));
+	hpnt = scsi_host_alloc(&sdebug_driver_template, 0);
 	if (NULL == hpnt) {
 		pr_err("scsi_host_alloc failed\n");
 		error = -ENODEV;
@@ -7864,7 +7863,6 @@ static int sdebug_driver_probe(struct device *dev)
 		hpnt->nr_maps = 3;
 
 	sdbg_host->shost = hpnt;
-	*((struct sdebug_host_info **)hpnt->hostdata) = sdbg_host;
 	if ((hpnt->this_id >= 0) && (sdebug_num_tgts > hpnt->this_id))
 		hpnt->max_id = sdebug_num_tgts + 1;
 	else
-- 
2.35.3

[PATCH 02/11] scsi: scsi_debug: Stop setting devip->sdbg_host twice

[John Garry]

In sdebug_device_create(), the devip->sdbg_host pointer is needlessly set
twice, so stop doing that.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index f713117a5f2e..6d1c483a0a3b 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5157,7 +5157,6 @@ static struct sdebug_dev_info *sdebug_device_create(
 		} else {
 			devip->zmodel = BLK_ZONED_NONE;
 		}
-		devip->sdbg_host = sdbg_host;
 		devip->create_ts = ktime_get_boottime();
 		atomic_set(&devip->stopped, (sdeb_tur_ms_to_ready > 0 ? 2 : 0));
 		list_add_tail(&devip->dev_list, &sdbg_host->dev_info_list);
-- 
2.35.3

[PATCH 03/11] scsi: scsi_debug: Drop scsi_debug_abort() NULL pointer checks

[John Garry]

The SCSI cmnd pointer arg would never be NULL, so drop the check. In
addition, its SCSI device pointer would never be NULL.

The only caller is scsi_send_eh_cmnd() -> scsi_abort_eh_cmnd() ->
scsi_try_to_abort_cmd() -> scsi_try_to_abort_cmd(), and in the origin of
that chain those pointers cannot be NULL.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 6d1c483a0a3b..4f971e8bfebc 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5362,13 +5362,13 @@ static int scsi_debug_abort(struct scsi_cmnd *SCpnt)
 	bool ok;
 
 	++num_aborts;
-	if (SCpnt) {
-		ok = stop_queued_cmnd(SCpnt);
-		if (SCpnt->device && (SDEBUG_OPT_ALL_NOISE & sdebug_opts))
-			sdev_printk(KERN_INFO, SCpnt->device,
-				    "%s: command%s found\n", __func__,
-				    ok ? "" : " not");
-	}
+
+	ok = stop_queued_cmnd(SCpnt);
+	if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
+		sdev_printk(KERN_INFO, SCpnt->device,
+			    "%s: command%s found\n", __func__,
+			    ok ? "" : " not");
+
 	return SUCCESS;
 }
 
-- 
2.35.3

[PATCH 04/11] scsi: scsi_debug: Drop scsi_debug_device_reset() NULL pointer checks

[John Garry]

The SCSI cmnd pointer arg would never be NULL, so drop the check. In
addition, its SCSI device pointer would never be NULL (so drop that check
also).

The only caller is scsi_try_bus_device_reset(), and the command and its
device pointer could not be NULL when calling eh_device_reset_handler()
there.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 4f971e8bfebc..b67d0c72118c 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5374,17 +5374,16 @@ static int scsi_debug_abort(struct scsi_cmnd *SCpnt)
 
 static int scsi_debug_device_reset(struct scsi_cmnd *SCpnt)
 {
+	struct scsi_device *sdp = SCpnt->device;
+	struct sdebug_dev_info *devip = sdp->hostdata;
+
 	++num_dev_resets;
-	if (SCpnt && SCpnt->device) {
-		struct scsi_device *sdp = SCpnt->device;
-		struct sdebug_dev_info *devip =
-				(struct sdebug_dev_info *)sdp->hostdata;
 
-		if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
-			sdev_printk(KERN_INFO, sdp, "%s\n", __func__);
-		if (devip)
-			set_bit(SDEBUG_UA_POR, devip->uas_bm);
-	}
+	if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
+		sdev_printk(KERN_INFO, sdp, "%s\n", __func__);
+	if (devip)
+		set_bit(SDEBUG_UA_POR, devip->uas_bm);
+
 	return SUCCESS;
 }
 
-- 
2.35.3

[PATCH 05/11] scsi: scsi_debug: Drop scsi_debug_target_reset() NULL pointer checks

[John Garry]

The checks for SCSI cmnd, SCSI device, and SCSI host are unnecessary, so
drop them. Likewise, drop the NULL check for sdbg_host.

The only caller is scsi_try_target_reset() -> eh_target_reset_handler(),
and there those pointers cannot be NULL.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 31 +++++++++++--------------------
 1 file changed, 11 insertions(+), 20 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index b67d0c72118c..33c0df6a3275 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5389,37 +5389,28 @@ static int scsi_debug_device_reset(struct scsi_cmnd *SCpnt)
 
 static int scsi_debug_target_reset(struct scsi_cmnd *SCpnt)
 {
+	struct scsi_device *sdp = SCpnt->device;
 	struct sdebug_host_info *sdbg_host;
 	struct sdebug_dev_info *devip;
-	struct scsi_device *sdp;
-	struct Scsi_Host *hp;
 	int k = 0;
 
 	++num_target_resets;
-	if (!SCpnt)
-		goto lie;
-	sdp = SCpnt->device;
-	if (!sdp)
-		goto lie;
 	if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
 		sdev_printk(KERN_INFO, sdp, "%s\n", __func__);
-	hp = sdp->host;
-	if (!hp)
-		goto lie;
-	sdbg_host = to_sdebug_host_from_shost(hp);
-	if (sdbg_host) {
-		list_for_each_entry(devip,
-				    &sdbg_host->dev_info_list,
-				    dev_list)
-			if (devip->target == sdp->id) {
-				set_bit(SDEBUG_UA_BUS_RESET, devip->uas_bm);
-				++k;
-			}
+
+	sdbg_host = to_sdebug_host_from_shost(sdp->host);
+
+	list_for_each_entry(devip, &sdbg_host->dev_info_list, dev_list) {
+		if (devip->target == sdp->id) {
+			set_bit(SDEBUG_UA_BUS_RESET, devip->uas_bm);
+			++k;
+		}
 	}
+
 	if (SDEBUG_OPT_RESET_NOISE & sdebug_opts)
 		sdev_printk(KERN_INFO, sdp,
 			    "%s: %d device(s) found in target\n", __func__, k);
-lie:
+
 	return SUCCESS;
 }
 
-- 
2.35.3

[PATCH 06/11] scsi: scsi_debug: Drop scsi_debug_bus_reset() NULL pointer checks

[John Garry]

The checks for SCSI cmnd, SCSI device, and SCSI host are unnecessary, so
drop them. Likewise, drop the NULL check for sdbg_host.

The only caller is scsi_try_bus_reset() -> eh_bus_reset_handler(),
and there those pointers cannot be NULL.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 26 +++++++++-----------------
 1 file changed, 9 insertions(+), 17 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 33c0df6a3275..6f1dbd8dc37f 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5418,32 +5418,24 @@ static int scsi_debug_bus_reset(struct scsi_cmnd *SCpnt)
 {
 	struct sdebug_host_info *sdbg_host;
 	struct sdebug_dev_info *devip;
-	struct scsi_device *sdp;
-	struct Scsi_Host *hp;
+	struct scsi_device *sdp = SCpnt->device;
 	int k = 0;
 
 	++num_bus_resets;
-	if (!(SCpnt && SCpnt->device))
-		goto lie;
-	sdp = SCpnt->device;
+
 	if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
 		sdev_printk(KERN_INFO, sdp, "%s\n", __func__);
-	hp = sdp->host;
-	if (hp) {
-		sdbg_host = to_sdebug_host_from_shost(hp);
-		if (sdbg_host) {
-			list_for_each_entry(devip,
-					    &sdbg_host->dev_info_list,
-					    dev_list) {
-				set_bit(SDEBUG_UA_BUS_RESET, devip->uas_bm);
-				++k;
-			}
-		}
+
+	sdbg_host = to_sdebug_host_from_shost(sdp->host);
+
+	list_for_each_entry(devip, &sdbg_host->dev_info_list, dev_list) {
+		set_bit(SDEBUG_UA_BUS_RESET, devip->uas_bm);
+		++k;
 	}
+
 	if (SDEBUG_OPT_RESET_NOISE & sdebug_opts)
 		sdev_printk(KERN_INFO, sdp,
 			    "%s: %d device(s) found in host\n", __func__, k);
-lie:
 	return SUCCESS;
 }
 
-- 
2.35.3

[PATCH 07/11] scsi: scsi_debug: Drop scsi_debug_host_reset() device NULL pointer check

[John Garry]

The check for device pointer for the SCSI command is unnecessary, so drop
it.

The only caller is scsi_try_host_reset() -> eh_host_reset_handler(),
and there that pointer cannot be NULL.

Indeed, there is already code later in the same function which does not
check the device pointer for the SCSI command.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 6f1dbd8dc37f..6707ccdc07c9 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5446,7 +5446,7 @@ static int scsi_debug_host_reset(struct scsi_cmnd *SCpnt)
 	int k = 0;
 
 	++num_host_resets;
-	if ((SCpnt->device) && (SDEBUG_OPT_ALL_NOISE & sdebug_opts))
+	if (SDEBUG_OPT_ALL_NOISE & sdebug_opts)
 		sdev_printk(KERN_INFO, SCpnt->device, "%s\n", __func__);
 	spin_lock(&sdebug_host_list_lock);
 	list_for_each_entry(sdbg_host, &sdebug_host_list, host_list) {
-- 
2.35.3

[PATCH 08/11] scsi: scsi_debug: Drop check for num_in_q exceeding queue depth

[John Garry]

The per-device num_in_q value cannot exceed the device queue depth, so drop
the check.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 6707ccdc07c9..fb7434d09cb9 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -5599,15 +5599,8 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 	}
 	num_in_q = atomic_read(&devip->num_in_q);
 	qdepth = cmnd->device->queue_depth;
-	if (unlikely((qdepth > 0) && (num_in_q >= qdepth))) {
-		if (scsi_result) {
-			spin_unlock_irqrestore(&sqp->qc_lock, iflags);
-			goto respond_in_thread;
-		} else
-			scsi_result = device_qfull_result;
-	} else if (unlikely(sdebug_every_nth &&
-			    (SDEBUG_OPT_RARE_TSF & sdebug_opts) &&
-			    (scsi_result == 0))) {
+	if (unlikely(sdebug_every_nth && (SDEBUG_OPT_RARE_TSF & sdebug_opts) &&
+		     (scsi_result == 0))) {
 		if ((num_in_q == (qdepth - 1)) &&
 		    (atomic_inc_return(&sdebug_a_tsf) >=
 		     abs(sdebug_every_nth))) {
-- 
2.35.3

[PATCH 09/11] scsi: scsi_debug: Drop sdebug_dev_info.num_in_q

[John Garry]

In schedule_resp(), under certain conditions we check whether the
per-device queue is full (num_in_q == queue depth - 1) and we may inject a
"task set full" (TSF) error if it is.

However how we read num_in_q is racy - many threads may see the same
"queue is full" value (and also issue a TSF).

There is per-queue locking in reading per-device num_in_q, but that would
not help.

Replace how we read num_in_q at this location with a call to
scsi_device_busy(). Calling scsi_device_busy() is likewise racy (as reading
num_in_q), so nothing lost or gained. Calling scsi_device_busy() is also
slow as it needs to read all bits in the per-device budget bitmap, but we
can live with that since we're just a simulator and it's only under
a certain configs which we would see this.

Also move the "task set full" print earlier as it would only be called
now under this conditions. However, previously it may not have been
called - like returning early - but keep it simple and always call it.

At this point we can drop sdebug_dev_info.num_in_q - it is difficult to
maintain properly and adds extra normal case command processing.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 63 ++++++++++-----------------------------
 1 file changed, 16 insertions(+), 47 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index fb7434d09cb9..cf3745d7b8f9 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -288,7 +288,6 @@ struct sdebug_dev_info {
 	uuid_t lu_name;
 	struct sdebug_host_info *sdbg_host;
 	unsigned long uas_bm[1];
-	atomic_t num_in_q;
 	atomic_t stopped;	/* 1: by SSU, 2: device start */
 	bool used;
 
@@ -4933,7 +4932,6 @@ static void sdebug_q_cmd_complete(struct sdebug_defer *sd_dp)
 	struct sdebug_queue *sqp;
 	struct sdebug_queued_cmd *sqcp;
 	struct scsi_cmnd *scp;
-	struct sdebug_dev_info *devip;
 
 	if (unlikely(aborted))
 		sd_dp->aborted = false;
@@ -4958,11 +4956,7 @@ static void sdebug_q_cmd_complete(struct sdebug_defer *sd_dp)
 		       sd_dp->sqa_idx, qc_idx, sd_dp->hc_idx);
 		return;
 	}
-	devip = (struct sdebug_dev_info *)scp->device->hostdata;
-	if (likely(devip))
-		atomic_dec(&devip->num_in_q);
-	else
-		pr_err("devip=NULL\n");
+
 	if (unlikely(atomic_read(&retired_max_queue) > 0))
 		retiring = 1;
 
@@ -5194,7 +5188,6 @@ static struct sdebug_dev_info *find_build_dev_info(struct scsi_device *sdev)
 	open_devip->target = sdev->id;
 	open_devip->lun = sdev->lun;
 	open_devip->sdbg_host = sdbg_host;
-	atomic_set(&open_devip->num_in_q, 0);
 	set_bit(SDEBUG_UA_POOCCUR, open_devip->uas_bm);
 	open_devip->used = true;
 	return open_devip;
@@ -5265,7 +5258,6 @@ static bool stop_queued_cmnd(struct scsi_cmnd *cmnd)
 	enum sdeb_defer_type l_defer_t;
 	struct sdebug_queue *sqp;
 	struct sdebug_queued_cmd *sqcp;
-	struct sdebug_dev_info *devip;
 	struct sdebug_defer *sd_dp;
 
 	for (j = 0, sqp = sdebug_q_arr; j < submit_queues; ++j, ++sqp) {
@@ -5280,10 +5272,6 @@ static bool stop_queued_cmnd(struct scsi_cmnd *cmnd)
 				if (cmnd != sqcp->a_cmnd)
 					continue;
 				/* found */
-				devip = (struct sdebug_dev_info *)
-						cmnd->device->hostdata;
-				if (devip)
-					atomic_dec(&devip->num_in_q);
 				sqcp->a_cmnd = NULL;
 				sd_dp = sqcp->sd_dp;
 				if (sd_dp) {
@@ -5310,7 +5298,6 @@ static void stop_all_queued(void)
 	enum sdeb_defer_type l_defer_t;
 	struct sdebug_queue *sqp;
 	struct sdebug_queued_cmd *sqcp;
-	struct sdebug_dev_info *devip;
 	struct sdebug_defer *sd_dp;
 
 	for (j = 0, sqp = sdebug_q_arr; j < submit_queues; ++j, ++sqp) {
@@ -5320,10 +5307,6 @@ static void stop_all_queued(void)
 				sqcp = &sqp->qc_arr[k];
 				if (sqcp->a_cmnd == NULL)
 					continue;
-				devip = (struct sdebug_dev_info *)
-					sqcp->a_cmnd->device->hostdata;
-				if (devip)
-					atomic_dec(&devip->num_in_q);
 				sqcp->a_cmnd = NULL;
 				sd_dp = sqcp->sd_dp;
 				if (sd_dp) {
@@ -5571,9 +5554,8 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 			 int delta_jiff, int ndelay)
 {
 	bool new_sd_dp;
-	bool inject = false;
 	bool polled = scsi_cmd_to_rq(cmnd)->cmd_flags & REQ_POLLED;
-	int k, num_in_q, qdepth;
+	int k;
 	unsigned long iflags;
 	u64 ns_from_boot = 0;
 	struct sdebug_queue *sqp;
@@ -5597,16 +5579,21 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 		spin_unlock_irqrestore(&sqp->qc_lock, iflags);
 		return SCSI_MLQUEUE_HOST_BUSY;
 	}
-	num_in_q = atomic_read(&devip->num_in_q);
-	qdepth = cmnd->device->queue_depth;
+
 	if (unlikely(sdebug_every_nth && (SDEBUG_OPT_RARE_TSF & sdebug_opts) &&
 		     (scsi_result == 0))) {
+		int num_in_q = scsi_device_busy(sdp);
+		int qdepth = cmnd->device->queue_depth;
+
 		if ((num_in_q == (qdepth - 1)) &&
 		    (atomic_inc_return(&sdebug_a_tsf) >=
 		     abs(sdebug_every_nth))) {
 			atomic_set(&sdebug_a_tsf, 0);
-			inject = true;
 			scsi_result = device_qfull_result;
+
+			if (unlikely(SDEBUG_OPT_Q_NOISE & sdebug_opts))
+				sdev_printk(KERN_INFO, sdp, "%s: num_in_q=%d +1, <inject> status: TASK SET FULL\n",
+					    __func__, num_in_q);
 		}
 	}
 
@@ -5622,7 +5609,6 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 		goto respond_in_thread;
 	}
 	set_bit(k, sqp->in_use_bm);
-	atomic_inc(&devip->num_in_q);
 	sqcp = &sqp->qc_arr[k];
 	sqcp->a_cmnd = cmnd;
 	cmnd->host_scribble = (unsigned char *)sqcp;
@@ -5632,7 +5618,6 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 	if (!sd_dp) {
 		sd_dp = kzalloc(sizeof(*sd_dp), GFP_ATOMIC);
 		if (!sd_dp) {
-			atomic_dec(&devip->num_in_q);
 			clear_bit(k, sqp->in_use_bm);
 			return SCSI_MLQUEUE_HOST_BUSY;
 		}
@@ -5692,7 +5677,6 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 				if (kt <= d) {	/* elapsed duration >= kt */
 					spin_lock_irqsave(&sqp->qc_lock, iflags);
 					sqcp->a_cmnd = NULL;
-					atomic_dec(&devip->num_in_q);
 					clear_bit(k, sqp->in_use_bm);
 					spin_unlock_irqrestore(&sqp->qc_lock, iflags);
 					if (new_sd_dp)
@@ -5768,9 +5752,7 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 			sd_dp->aborted = false;
 		}
 	}
-	if (unlikely((SDEBUG_OPT_Q_NOISE & sdebug_opts) && scsi_result == device_qfull_result))
-		sdev_printk(KERN_INFO, sdp, "%s: num_in_q=%d +1, %s%s\n", __func__,
-			    num_in_q, (inject ? "<inject> " : ""), "status: TASK SET FULL");
+
 	return 0;
 
 respond_in_thread:	/* call back to mid-layer using invocation thread */
@@ -7375,17 +7357,12 @@ static void sdebug_do_remove_host(bool the_end)
 
 static int sdebug_change_qdepth(struct scsi_device *sdev, int qdepth)
 {
-	int num_in_q = 0;
-	struct sdebug_dev_info *devip;
+	struct sdebug_dev_info *devip = sdev->hostdata;
 
-	block_unblock_all_queues(true);
-	devip = (struct sdebug_dev_info *)sdev->hostdata;
-	if (NULL == devip) {
-		block_unblock_all_queues(false);
+	if (!devip)
 		return	-ENODEV;
-	}
-	num_in_q = atomic_read(&devip->num_in_q);
 
+	block_unblock_all_queues(true);
 	if (qdepth > SDEBUG_CANQUEUE) {
 		qdepth = SDEBUG_CANQUEUE;
 		pr_warn("%s: requested qdepth [%d] exceeds canqueue [%d], trim\n", __func__,
@@ -7396,10 +7373,8 @@ static int sdebug_change_qdepth(struct scsi_device *sdev, int qdepth)
 	if (qdepth != sdev->queue_depth)
 		scsi_change_queue_depth(sdev, qdepth);
 
-	if (SDEBUG_OPT_Q_NOISE & sdebug_opts) {
-		sdev_printk(KERN_INFO, sdev, "%s: qdepth=%d, num_in_q=%d\n",
-			    __func__, qdepth, num_in_q);
-	}
+	if (SDEBUG_OPT_Q_NOISE & sdebug_opts)
+		sdev_printk(KERN_INFO, sdev, "%s: qdepth=%d\n", __func__, qdepth);
 	block_unblock_all_queues(false);
 	return sdev->queue_depth;
 }
@@ -7501,7 +7476,6 @@ static int sdebug_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num)
 	struct sdebug_queue *sqp;
 	struct sdebug_queued_cmd *sqcp;
 	struct scsi_cmnd *scp;
-	struct sdebug_dev_info *devip;
 	struct sdebug_defer *sd_dp;
 
 	sqp = sdebug_q_arr + queue_num;
@@ -7539,11 +7513,6 @@ static int sdebug_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num)
 
 		} else		/* ignoring non REQ_POLLED requests */
 			continue;
-		devip = (struct sdebug_dev_info *)scp->device->hostdata;
-		if (likely(devip))
-			atomic_dec(&devip->num_in_q);
-		else
-			pr_err("devip=NULL from %s\n", __func__);
 		if (unlikely(atomic_read(&retired_max_queue) > 0))
 			retiring = true;
 
-- 
2.35.3

[PATCH 10/11] scsi: scsi_debug: Get command abort feature working again

[John Garry]

The command abort feature allows us to test aborting a command which has
timed-out.

The idea is that for specific commands we just don't call scsi_done() and
allow the request to timeout, which ensures SCSI EH kicks-in we try to
abort the command.

Since commit 4a0c6f432d15 ("scsi: scsi_debug: Add new defer type for
mq_poll") this does not seem to work. The issue is that we clear the
sd_dp->aborted flag in schedule_resp() before the completion callback has
run. When the completion callback actually runs, it calls scsi_done() as
normal as sd_dp->aborted unset. This is all very racy.

Fix by not clearing sd_dp->aborted in schedule_resp(). Also move the call
to blk_abort_request() from schedule_resp() to sdebug_q_cmd_complete(),
which makes the code have a more logical sequence.

I also note that this feature only works for commands which are classed
as "SDEG_RES_IMMED_MASK", but only practically triggered with prior RW
commands. So for my experiment I need to run fio to trigger the error on
the "nth" command (see inject_on_this_cmd()), and then run something like
sg_sync to queue a command to actually trigger the abort.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index cf3745d7b8f9..6cf30fceab78 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -4985,7 +4985,8 @@ static void sdebug_q_cmd_complete(struct sdebug_defer *sd_dp)
 	spin_unlock_irqrestore(&sqp->qc_lock, iflags);
 	if (unlikely(aborted)) {
 		if (sdebug_verbose)
-			pr_info("bypassing scsi_done() due to aborted cmd\n");
+			pr_info("bypassing scsi_done() due to aborted cmd, kicking-off EH\n");
+		blk_abort_request(scsi_cmd_to_rq(scp));
 		return;
 	}
 	scsi_done(scp); /* callback to mid level */
@@ -5718,8 +5719,13 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 			sd_dp->issuing_cpu = raw_smp_processor_id();
 	} else {	/* jdelay < 0, use work queue */
 		if (unlikely((sdebug_opts & SDEBUG_OPT_CMD_ABORT) &&
-			     atomic_read(&sdeb_inject_pending)))
+			     atomic_read(&sdeb_inject_pending))) {
 			sd_dp->aborted = true;
+			atomic_set(&sdeb_inject_pending, 0);
+			sdev_printk(KERN_INFO, sdp, "abort request tag=%#x\n",
+				    blk_mq_unique_tag_to_tag(get_tag(cmnd)));
+		}
+
 		if (polled) {
 			sd_dp->cmpl_ts = ns_to_ktime(ns_from_boot);
 			spin_lock_irqsave(&sqp->qc_lock, iflags);
@@ -5744,13 +5750,6 @@ static int schedule_resp(struct scsi_cmnd *cmnd, struct sdebug_dev_info *devip,
 		}
 		if (sdebug_statistics)
 			sd_dp->issuing_cpu = raw_smp_processor_id();
-		if (unlikely(sd_dp->aborted)) {
-			sdev_printk(KERN_INFO, sdp, "abort request tag %d\n",
-				    scsi_cmd_to_rq(cmnd)->tag);
-			blk_abort_request(scsi_cmd_to_rq(cmnd));
-			atomic_set(&sdeb_inject_pending, 0);
-			sd_dp->aborted = false;
-		}
 	}
 
 	return 0;
-- 
2.35.3

[PATCH 11/11] scsi: scsi_debug: Add poll mode deferred completions to statistics

[John Garry]

Currently commands completed via poll mode are not included in the
statistics gathering for deferred completions and missed CPUs.

Poll mode completions should be treated the same as other deferred
completion types, so add poll mode completions to the statistics.

Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/scsi_debug.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index 6cf30fceab78..76ff8ca6070d 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -7537,6 +7537,13 @@ static int sdebug_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num)
 		}
 		WRITE_ONCE(sd_dp->defer_t, SDEB_DEFER_NONE);
 		spin_unlock_irqrestore(&sqp->qc_lock, iflags);
+
+		if (sdebug_statistics) {
+			atomic_inc(&sdebug_completions);
+			if (raw_smp_processor_id() != sd_dp->issuing_cpu)
+				atomic_inc(&sdebug_miss_cpus);
+		}
+
 		scsi_done(scp); /* callback to mid level */
 		num_entries++;
 		spin_lock_irqsave(&sqp->qc_lock, iflags);
-- 
2.35.3

Re: [PATCH 01/11] scsi: scsi_debug: Don't hold driver host struct pointer in host->hostdata[]

[Bart Van Assche]

On 3/7/23 02:55, John Garry wrote:
> +#define to_sdebug_host_from_shost(shost)	\
> +	to_sdebug_host(shost->dma_dev)

Please make the name of this macro shorter and following naming 
conventions used elsewhere in the SCSI core. How about the following 
name: shost_to_sdebug_host()?

Thanks,

Bart.

Re: [PATCH 01/11] scsi: scsi_debug: Don't hold driver host struct pointer in host->hostdata[]

[John Garry]

On 07/03/2023 14:58, Bart Van Assche wrote:
> On 3/7/23 02:55, John Garry wrote:
>> +#define to_sdebug_host_from_shost(shost)    \
>> +    to_sdebug_host(shost->dma_dev)
> 
> Please make the name of this macro shorter and following naming 
> conventions used elsewhere in the SCSI core. How about the following 
> name: shost_to_sdebug_host()?

ok, fine, I can change that. Maybe then I'll also change 
to_sdebug_host() -> dev_to_sdebug_host() to avoid ambiguity.

Thanks,
John