[PATCH 0/4] Call scsi_device_put() from non-atomic context

[PATCH 0/4] Call scsi_device_put() from non-atomic context

[Bart Van Assche]

Hi Martin,

As reported by Dan Carpenter, a recent change may cause scsi_device_put() to
sleep while a few callers remain that call scsi_device_put() from atomic
context. This patch series converts those callers. Please consider this patch
series for the 6.0 kernel.

Thanks,

Bart.

Bart Van Assche (4):
  scsi: alua: Move a scsi_device_put() call out of alua_check_vpd()
  scsi: alua: Move a scsi_device_put() call out of
    alua_rtpg_select_sdev()
  scsi: bfa: Convert bfad_reset_sdev_bflags() from a macro into a
    function
  scsi: bfa: Rework bfad_reset_sdev_bflags()

 drivers/scsi/bfa/bfad_bsg.c                | 29 ++++++++++
 drivers/scsi/bfa/bfad_im.h                 | 26 ---------
 drivers/scsi/device_handler/scsi_dh_alua.c | 61 ++++++++++++++--------
 3 files changed, 68 insertions(+), 48 deletions(-)

[PATCH 1/4] scsi: alua: Move a scsi_device_put() call out of alua_check_vpd()

[Bart Van Assche]

This patch fixes the following smatch warning:

drivers/scsi/device_handler/scsi_dh_alua.c:1013 alua_rtpg_queue() warn: sleeping in atomic context

alua_check_vpd() <- disables preempt
-> alua_rtpg_queue()
   -> scsi_device_put()

Cc: Hannes Reinecke <hare@suse.de>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
---
 drivers/scsi/device_handler/scsi_dh_alua.c | 23 ++++++++++++++--------
 1 file changed, 15 insertions(+), 8 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh_alua.c b/drivers/scsi/device_handler/scsi_dh_alua.c
index 610a51538f03..f7bc81cc59ab 100644
--- a/drivers/scsi/device_handler/scsi_dh_alua.c
+++ b/drivers/scsi/device_handler/scsi_dh_alua.c
@@ -324,6 +324,7 @@ static int alua_check_vpd(struct scsi_device *sdev, struct alua_dh_data *h,
 	struct alua_port_group *pg, *old_pg = NULL;
 	bool pg_updated = false;
 	unsigned long flags;
+	bool put_sdev;
 
 	group_id = scsi_vpd_tpg_id(sdev, &rel_port);
 	if (group_id < 0) {
@@ -373,11 +374,14 @@ static int alua_check_vpd(struct scsi_device *sdev, struct alua_dh_data *h,
 		list_add_rcu(&h->node, &pg->dh_list);
 	spin_unlock_irqrestore(&pg->lock, flags);
 
-	alua_rtpg_queue(rcu_dereference_protected(h->pg,
+	put_sdev = alua_rtpg_queue(rcu_dereference_protected(h->pg,
 						  lockdep_is_held(&h->pg_lock)),
 			sdev, NULL, true);
 	spin_unlock(&h->pg_lock);
 
+	if (put_sdev)
+		scsi_device_put(sdev);
+
 	if (old_pg)
 		kref_put(&old_pg->kref, release_port_group);
 
@@ -968,9 +972,10 @@ static void alua_rtpg_work(struct work_struct *work)
  *         RTPG already has been scheduled.
  *
  * Returns true if and only if alua_rtpg_work() will be called asynchronously.
- * That function is responsible for calling @qdata->fn().
+ * That function is responsible for calling @qdata->fn(). If this function
+ * returns true, the caller is responsible for invoking scsi_device_put(@sdev).
  */
-static bool alua_rtpg_queue(struct alua_port_group *pg,
+static bool __must_check alua_rtpg_queue(struct alua_port_group *pg,
 			    struct scsi_device *sdev,
 			    struct alua_queue_data *qdata, bool force)
 {
@@ -1009,8 +1014,6 @@ static bool alua_rtpg_queue(struct alua_port_group *pg,
 		else
 			kref_put(&pg->kref, release_port_group);
 	}
-	if (sdev)
-		scsi_device_put(sdev);
 
 	return true;
 }
@@ -1117,10 +1120,12 @@ static int alua_activate(struct scsi_device *sdev,
 	rcu_read_unlock();
 	mutex_unlock(&h->init_mutex);
 
-	if (alua_rtpg_queue(pg, sdev, qdata, true))
+	if (alua_rtpg_queue(pg, sdev, qdata, true)) {
+		scsi_device_put(sdev);
 		fn = NULL;
-	else
+	} else {
 		err = SCSI_DH_DEV_OFFLINED;
+	}
 	kref_put(&pg->kref, release_port_group);
 out:
 	if (fn)
@@ -1146,7 +1151,9 @@ static void alua_check(struct scsi_device *sdev, bool force)
 		return;
 	}
 	rcu_read_unlock();
-	alua_rtpg_queue(pg, sdev, NULL, force);
+
+	if (alua_rtpg_queue(pg, sdev, NULL, force))
+		scsi_device_put(sdev);
 	kref_put(&pg->kref, release_port_group);
 }
 

Re: [PATCH 1/4] scsi: alua: Move a scsi_device_put() call out of alua_check_vpd()

[Hannes Reinecke]

On 10/31/22 23:47, Bart Van Assche wrote:
> This patch fixes the following smatch warning:
> 
> drivers/scsi/device_handler/scsi_dh_alua.c:1013 alua_rtpg_queue() warn: sleeping in atomic context
> 
> alua_check_vpd() <- disables preempt
> -> alua_rtpg_queue()
>     -> scsi_device_put()
> 
> Cc: Hannes Reinecke <hare@suse.de>
> Cc: Dan Carpenter <dan.carpenter@oracle.com>
> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
> Signed-off-by: Bart Van Assche <bvanassche@acm.org>
> ---
>   drivers/scsi/device_handler/scsi_dh_alua.c | 23 ++++++++++++++--------
>   1 file changed, 15 insertions(+), 8 deletions(-)
> 
Reviewed-by: Hannes Reinecke <hare@suse.de>

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		           Kernel Storage Architect
hare@suse.de			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer

[PATCH 2/4] scsi: alua: Move a scsi_device_put() call out of alua_rtpg_select_sdev()

[Bart Van Assche]

Move a scsi_device_put() call from alua_rtpg_select_sdev() to its
callers. This patch fixes the following smatch complaint:

drivers/scsi/device_handler/scsi_dh_alua.c:853 alua_rtpg_select_sdev() warn: sleeping in atomic context

alua_rtpg_work() <- disables preempt
-> alua_rtpg_select_sdev()
   -> scsi_device_put()

Cc: Hannes Reinecke <hare@suse.de>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
---
 drivers/scsi/device_handler/scsi_dh_alua.c | 38 ++++++++++++++--------
 1 file changed, 24 insertions(+), 14 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh_alua.c b/drivers/scsi/device_handler/scsi_dh_alua.c
index f7bc81cc59ab..693cd827e138 100644
--- a/drivers/scsi/device_handler/scsi_dh_alua.c
+++ b/drivers/scsi/device_handler/scsi_dh_alua.c
@@ -815,14 +815,19 @@ static unsigned alua_stpg(struct scsi_device *sdev, struct alua_port_group *pg)
 	return SCSI_DH_RETRY;
 }
 
-static bool alua_rtpg_select_sdev(struct alua_port_group *pg)
+/*
+ * The caller must call scsi_device_put() on the returned pointer if it is not
+ * NULL.
+ */
+static struct scsi_device * __must_check
+alua_rtpg_select_sdev(struct alua_port_group *pg)
 {
 	struct alua_dh_data *h;
-	struct scsi_device *sdev = NULL;
+	struct scsi_device *sdev = NULL, *prev_sdev;
 
 	lockdep_assert_held(&pg->lock);
 	if (WARN_ON(!pg->rtpg_sdev))
-		return false;
+		return NULL;
 
 	/*
 	 * RCU protection isn't necessary for dh_list here
@@ -849,22 +854,22 @@ static bool alua_rtpg_select_sdev(struct alua_port_group *pg)
 		pr_warn("%s: no device found for rtpg\n",
 			(pg->device_id_len ?
 			 (char *)pg->device_id_str : "(nameless PG)"));
-		return false;
+		return NULL;
 	}
 
 	sdev_printk(KERN_INFO, sdev, "rtpg retry on different device\n");
 
-	scsi_device_put(pg->rtpg_sdev);
+	prev_sdev = pg->rtpg_sdev;
 	pg->rtpg_sdev = sdev;
 
-	return true;
+	return prev_sdev;
 }
 
 static void alua_rtpg_work(struct work_struct *work)
 {
 	struct alua_port_group *pg =
 		container_of(work, struct alua_port_group, rtpg_work.work);
-	struct scsi_device *sdev;
+	struct scsi_device *sdev, *prev_sdev = NULL;
 	LIST_HEAD(qdata_list);
 	int err = SCSI_DH_OK;
 	struct alua_queue_data *qdata, *tmp;
@@ -905,7 +910,7 @@ static void alua_rtpg_work(struct work_struct *work)
 
 		/* If RTPG failed on the current device, try using another */
 		if (err == SCSI_DH_RES_TEMP_UNAVAIL &&
-		    alua_rtpg_select_sdev(pg))
+		    (prev_sdev = alua_rtpg_select_sdev(pg)))
 			err = SCSI_DH_IMM_RETRY;
 
 		if (err == SCSI_DH_RETRY || err == SCSI_DH_IMM_RETRY ||
@@ -917,9 +922,7 @@ static void alua_rtpg_work(struct work_struct *work)
 				pg->interval = ALUA_RTPG_RETRY_DELAY;
 			pg->flags |= ALUA_PG_RUN_RTPG;
 			spin_unlock_irqrestore(&pg->lock, flags);
-			queue_delayed_work(kaluad_wq, &pg->rtpg_work,
-					   pg->interval * HZ);
-			return;
+			goto queue_rtpg;
 		}
 		if (err != SCSI_DH_OK)
 			pg->flags &= ~ALUA_PG_RUN_STPG;
@@ -934,9 +937,7 @@ static void alua_rtpg_work(struct work_struct *work)
 			pg->interval = 0;
 			pg->flags &= ~ALUA_PG_RUNNING;
 			spin_unlock_irqrestore(&pg->lock, flags);
-			queue_delayed_work(kaluad_wq, &pg->rtpg_work,
-					   pg->interval * HZ);
-			return;
+			goto queue_rtpg;
 		}
 	}
 
@@ -950,6 +951,9 @@ static void alua_rtpg_work(struct work_struct *work)
 	pg->rtpg_sdev = NULL;
 	spin_unlock_irqrestore(&pg->lock, flags);
 
+	if (prev_sdev)
+		scsi_device_put(prev_sdev);
+
 	list_for_each_entry_safe(qdata, tmp, &qdata_list, entry) {
 		list_del(&qdata->entry);
 		if (qdata->callback_fn)
@@ -961,6 +965,12 @@ static void alua_rtpg_work(struct work_struct *work)
 	spin_unlock_irqrestore(&pg->lock, flags);
 	scsi_device_put(sdev);
 	kref_put(&pg->kref, release_port_group);
+	return;
+
+queue_rtpg:
+	if (prev_sdev)
+		scsi_device_put(prev_sdev);
+	queue_delayed_work(kaluad_wq, &pg->rtpg_work, pg->interval * HZ);
 }
 
 /**

[PATCH 3/4] scsi: bfa: Convert bfad_reset_sdev_bflags() from a macro into a function

[Bart Van Assche]

Before modifying bfad_reset_sdev_bflags(), convert it from a macro into a
function.

Cc: Anil Gurumurthy <anil.gurumurthy@qlogic.com>
Cc: Sudarsana Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
---
 drivers/scsi/bfa/bfad_bsg.c | 27 +++++++++++++++++++++++++++
 drivers/scsi/bfa/bfad_im.h  | 26 --------------------------
 2 files changed, 27 insertions(+), 26 deletions(-)

diff --git a/drivers/scsi/bfa/bfad_bsg.c b/drivers/scsi/bfa/bfad_bsg.c
index be8dfbe13e90..73754032e25c 100644
--- a/drivers/scsi/bfa/bfad_bsg.c
+++ b/drivers/scsi/bfa/bfad_bsg.c
@@ -2540,6 +2540,33 @@ bfad_iocmd_vf_clr_stats(struct bfad_s *bfad, void *cmd)
 	return 0;
 }
 
+/*
+ * Set the SCSI device sdev_bflags - sdev_bflags are used by the
+ * SCSI mid-layer to choose LUN Scanning mode REPORT_LUNS vs. Sequential Scan
+ *
+ * Internally iterates over all the ITNIM's part of the im_port & sets the
+ * sdev_bflags for the scsi_device associated with LUN #0.
+ */
+static void bfad_reset_sdev_bflags(struct bfad_im_port_s *im_port,
+				   int lunmask_cfg)
+{
+	const u32 scan_flags = BLIST_NOREPORTLUN | BLIST_SPARSELUN;
+	struct bfad_itnim_s *itnim;
+	struct scsi_device *sdev;
+
+	list_for_each_entry(itnim, &im_port->itnim_mapped_list, list_entry) {
+		sdev = scsi_device_lookup(im_port->shost, itnim->channel,
+					  itnim->scsi_tgt_id, 0);
+		if (sdev) {
+			if (lunmask_cfg == BFA_TRUE)
+				sdev->sdev_bflags |= scan_flags;
+			else
+				sdev->sdev_bflags &= ~scan_flags;
+			scsi_device_put(sdev);
+		}
+	}
+}
+
 /* Function to reset the LUN SCAN mode */
 static void
 bfad_iocmd_lunmask_reset_lunscan_mode(struct bfad_s *bfad, int lunmask_cfg)
diff --git a/drivers/scsi/bfa/bfad_im.h b/drivers/scsi/bfa/bfad_im.h
index c03b225ea1ba..4353feedf76a 100644
--- a/drivers/scsi/bfa/bfad_im.h
+++ b/drivers/scsi/bfa/bfad_im.h
@@ -198,30 +198,4 @@ irqreturn_t bfad_intx(int irq, void *dev_id);
 int bfad_im_bsg_request(struct bsg_job *job);
 int bfad_im_bsg_timeout(struct bsg_job *job);
 
-/*
- * Macro to set the SCSI device sdev_bflags - sdev_bflags are used by the
- * SCSI mid-layer to choose LUN Scanning mode REPORT_LUNS vs. Sequential Scan
- *
- * Internally iterate's over all the ITNIM's part of the im_port & set's the
- * sdev_bflags for the scsi_device associated with LUN #0.
- */
-#define bfad_reset_sdev_bflags(__im_port, __lunmask_cfg) do {		\
-	struct scsi_device *__sdev = NULL;				\
-	struct bfad_itnim_s *__itnim = NULL;				\
-	u32 scan_flags = BLIST_NOREPORTLUN | BLIST_SPARSELUN;		\
-	list_for_each_entry(__itnim, &((__im_port)->itnim_mapped_list),	\
-			    list_entry) {				\
-		__sdev = scsi_device_lookup((__im_port)->shost,		\
-					    __itnim->channel,		\
-					    __itnim->scsi_tgt_id, 0);	\
-		if (__sdev) {						\
-			if ((__lunmask_cfg) == BFA_TRUE)		\
-				__sdev->sdev_bflags |= scan_flags;	\
-			else						\
-				__sdev->sdev_bflags &= ~scan_flags;	\
-			scsi_device_put(__sdev);			\
-		}							\
-	}								\
-} while (0)
-
 #endif

[PATCH 4/4] scsi: bfa: Rework bfad_reset_sdev_bflags()

[Bart Van Assche]

Since commit f93ed747e2c7 ("scsi: core: Release SCSI devices
synchronously") it is no longer allowed to call scsi_device_put() from
atomic context. Hence this patch that reworks bfad_reset_sdev_bflags()
such that scsi_device_put() is no longer called. This patch fixes the
following smatch warning:

drivers/scsi/bfa/bfad_bsg.c:2551 bfad_iocmd_lunmask_reset_lunscan_mode() warn: sleeping in atomic context

bfad_iocmd_lunmask() <- disables preempt
-> bfad_iocmd_lunmask_reset_lunscan_mode()
   -> scsi_device_put()

Cc: Anil Gurumurthy <anil.gurumurthy@qlogic.com>
Cc: Sudarsana Kalluru <sudarsana.kalluru@qlogic.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
---
 drivers/scsi/bfa/bfad_bsg.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/bfa/bfad_bsg.c b/drivers/scsi/bfa/bfad_bsg.c
index 73754032e25c..79d4f7ee5bcb 100644
--- a/drivers/scsi/bfa/bfad_bsg.c
+++ b/drivers/scsi/bfa/bfad_bsg.c
@@ -2553,18 +2553,20 @@ static void bfad_reset_sdev_bflags(struct bfad_im_port_s *im_port,
 	const u32 scan_flags = BLIST_NOREPORTLUN | BLIST_SPARSELUN;
 	struct bfad_itnim_s *itnim;
 	struct scsi_device *sdev;
+	unsigned long flags;
 
+	spin_lock_irqsave(im_port->shost->host_lock, flags);
 	list_for_each_entry(itnim, &im_port->itnim_mapped_list, list_entry) {
-		sdev = scsi_device_lookup(im_port->shost, itnim->channel,
-					  itnim->scsi_tgt_id, 0);
+		sdev = __scsi_device_lookup(im_port->shost, itnim->channel,
+					    itnim->scsi_tgt_id, 0);
 		if (sdev) {
 			if (lunmask_cfg == BFA_TRUE)
 				sdev->sdev_bflags |= scan_flags;
 			else
 				sdev->sdev_bflags &= ~scan_flags;
-			scsi_device_put(sdev);
 		}
 	}
+	spin_unlock_irqrestore(im_port->shost->host_lock, flags);
 }
 
 /* Function to reset the LUN SCAN mode */

Re: [PATCH 0/4] Call scsi_device_put() from non-atomic context

[Martin K. Petersen]

Bart,

> As reported by Dan Carpenter, a recent change may cause
> scsi_device_put() to sleep while a few callers remain that call
> scsi_device_put() from atomic context. This patch series converts
> those callers. Please consider this patch series for the 6.0 kernel.

Applied to 6.2/scsi-staging, thanks!

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH 0/4] Call scsi_device_put() from non-atomic context

[Martin K. Petersen]

On Mon, 31 Oct 2022 15:47:24 -0700, Bart Van Assche wrote:

> As reported by Dan Carpenter, a recent change may cause scsi_device_put() to
> sleep while a few callers remain that call scsi_device_put() from atomic
> context. This patch series converts those callers. Please consider this patch
> series for the 6.0 kernel.
> 
> Thanks,
> 
> [...]

Applied to 6.2/scsi-queue, thanks!

[1/4] scsi: alua: Move a scsi_device_put() call out of alua_check_vpd()
      https://git.kernel.org/mkp/scsi/c/0b25e17e9018
[2/4] scsi: alua: Move a scsi_device_put() call out of alua_rtpg_select_sdev()
      https://git.kernel.org/mkp/scsi/c/379e2554e3d1
[3/4] scsi: bfa: Convert bfad_reset_sdev_bflags() from a macro into a function
      https://git.kernel.org/mkp/scsi/c/2e5a6c3baccd
[4/4] scsi: bfa: Rework bfad_reset_sdev_bflags()
      https://git.kernel.org/mkp/scsi/c/2e79cf37b15b

-- 
Martin K. Petersen	Oracle Linux Engineering