Merge patch series "scsi: fixes for targets with many LUNs, and scsi_target_block rework"

Martin Wilck <mwilck@suse.com> says: This patch series addresses some issues we saw in a test setup with a large number of SCSI LUNs. The first two patches simply increase the number of available sg and bsg devices. 3-5 fix a large delay we encountered between blocking a Fibre Channel remote port and the dev_loss_tmo. 6 renames scsi_target_block() to scsi_block_targets(), and makes additional changes to this API, as suggested in the review of the v2 series. 7 improves a warning message. Link: https://lore.kernel.org/r/20230614103616.31857-1-mwilck@suse.com Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2024-11-21 19:41:42 +00:00 · 2023-06-16 12:21:04 -04:00 · 2023-06-16 12:21:04 -04:00 · af92c02fb2
commit af92c02fb2
parent 20fce500b2 6d7160c7da
8 changed files with 50 additions and 49 deletions
--- a/block/bsg.c
+++ b/block/bsg.c
@ -36,7 +36,7 @@ static inline struct bsg_device *to_bsg_device(struct inode *inode)
 }

 #define BSG_DEFAULT_CMDS	64
-#define BSG_MAX_DEVS		32768
+#define BSG_MAX_DEVS		(1 << MINORBITS)

 static DEFINE_IDA(bsg_minor_ida);
 static struct class *bsg_class;
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@ -2731,24 +2731,16 @@ void scsi_start_queue(struct scsi_device *sdev)
 		blk_mq_unquiesce_queue(sdev->request_queue);
 }

-static void scsi_stop_queue(struct scsi_device *sdev, bool nowait)
+static void scsi_stop_queue(struct scsi_device *sdev)
 {
 	/*
 	 * The atomic variable of ->queue_stopped covers that
 	 * blk_mq_quiesce_queue* is balanced with blk_mq_unquiesce_queue.
 	 *
-	 * However, we still need to wait until quiesce is done
-	 * in case that queue has been stopped.
+	 * The caller needs to wait until quiesce is done.
 	 */
-	if (!cmpxchg(&sdev->queue_stopped, 0, 1)) {
-		if (nowait)
-			blk_mq_quiesce_queue_nowait(sdev->request_queue);
-		else
-			blk_mq_quiesce_queue(sdev->request_queue);
-	} else {
-		if (!nowait)
-			blk_mq_wait_quiesce_done(sdev->request_queue->tag_set);
-	}
+	if (!cmpxchg(&sdev->queue_stopped, 0, 1))
+		blk_mq_quiesce_queue_nowait(sdev->request_queue);
 }

 /**
@ -2775,19 +2767,19 @@ int scsi_internal_device_block_nowait(struct scsi_device *sdev)
 	 * request queue.
 	 */
 	if (!ret)
-		scsi_stop_queue(sdev, true);
+		scsi_stop_queue(sdev);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(scsi_internal_device_block_nowait);

 /**
- * scsi_internal_device_block - try to transition to the SDEV_BLOCK state
+ * scsi_device_block - try to transition to the SDEV_BLOCK state
 * @sdev: device to block
+ * @data: dummy argument, ignored
 *
- * Pause SCSI command processing on the specified device and wait until all
- * ongoing scsi_request_fn() / scsi_queue_rq() calls have finished. May sleep.
- *
- * Returns zero if successful or a negative error code upon failure.
+ * Pause SCSI command processing on the specified device. Callers must wait
+ * until all ongoing scsi_queue_rq() calls have finished after this function
+ * returns.
 *
 * Note:
 * This routine transitions the device to the SDEV_BLOCK state (which must be
@ -2795,17 +2787,26 @@ EXPORT_SYMBOL_GPL(scsi_internal_device_block_nowait);
 * is paused until the device leaves the SDEV_BLOCK state. See also
 * scsi_internal_device_unblock().
 */
-static int scsi_internal_device_block(struct scsi_device *sdev)
+static void scsi_device_block(struct scsi_device *sdev, void *data)
 {
 	int err;
+	enum scsi_device_state state;

 	mutex_lock(&sdev->state_mutex);
 	err = __scsi_internal_device_block_nowait(sdev);
+	state = sdev->sdev_state;
 	if (err == 0)
-		scsi_stop_queue(sdev, false);
+		/*
+		 * scsi_stop_queue() must be called with the state_mutex
+		 * held. Otherwise a simultaneous scsi_start_queue() call
+		 * might unquiesce the queue before we quiesce it.
+		 */
+		scsi_stop_queue(sdev);
+
 	mutex_unlock(&sdev->state_mutex);

-	return err;
+	WARN_ONCE(err, "%s: failed to block %s in state %d\n",
+		  __func__, dev_name(&sdev->sdev_gendev), state);
 }

 /**
@ -2888,36 +2889,35 @@ static int scsi_internal_device_unblock(struct scsi_device *sdev,
 	return ret;
 }

-static void
-device_block(struct scsi_device *sdev, void *data)
-{
-	int ret;
-
-	ret = scsi_internal_device_block(sdev);
-
-	WARN_ONCE(ret, "scsi_internal_device_block(%s) failed: ret = %d\n",
-		  dev_name(&sdev->sdev_gendev), ret);
-}
-
 static int
 target_block(struct device *dev, void *data)
 {
 	if (scsi_is_target_device(dev))
 		starget_for_each_device(to_scsi_target(dev), NULL,
-					device_block);
+					scsi_device_block);
 	return 0;
 }

+/**
+ * scsi_block_targets - transition all SCSI child devices to SDEV_BLOCK state
+ * @dev: a parent device of one or more scsi_target devices
+ * @shost: the Scsi_Host to which this device belongs
+ *
+ * Iterate over all children of @dev, which should be scsi_target devices,
+ * and switch all subordinate scsi devices to SDEV_BLOCK state. Wait for
+ * ongoing scsi_queue_rq() calls to finish. May sleep.
+ *
+ * Note:
+ * @dev must not itself be a scsi_target device.
+ */
 void
-scsi_target_block(struct device *dev)
+scsi_block_targets(struct Scsi_Host *shost, struct device *dev)
 {
-	if (scsi_is_target_device(dev))
-		starget_for_each_device(to_scsi_target(dev), NULL,
-					device_block);
-	else
-		device_for_each_child(dev, NULL, target_block);
+	WARN_ON_ONCE(scsi_is_target_device(dev));
+	device_for_each_child(dev, NULL, target_block);
+	blk_mq_wait_quiesce_done(&shost->tag_set);
 }
-EXPORT_SYMBOL_GPL(scsi_target_block);
+EXPORT_SYMBOL_GPL(scsi_block_targets);

 static void
 device_unblock(struct scsi_device *sdev, void *data)
--- a/drivers/scsi/scsi_transport_fc.c
+++ b/drivers/scsi/scsi_transport_fc.c
@ -3451,7 +3451,7 @@ fc_remote_port_delete(struct fc_rport  *rport)

 	spin_unlock_irqrestore(shost->host_lock, flags);

-	scsi_target_block(&rport->dev);
+	scsi_block_targets(shost, &rport->dev);

 	/* see if we need to kill io faster than waiting for device loss */
 	if ((rport->fast_io_fail_tmo != -1) &&
--- a/drivers/scsi/scsi_transport_iscsi.c
+++ b/drivers/scsi/scsi_transport_iscsi.c
@ -1943,13 +1943,14 @@ static void __iscsi_block_session(struct work_struct *work)
 	struct iscsi_cls_session *session =
 			container_of(work, struct iscsi_cls_session,
 				     block_work);
+	struct Scsi_Host *shost = iscsi_session_to_shost(session);
 	unsigned long flags;

 	ISCSI_DBG_TRANS_SESSION(session, "Blocking session\n");
 	spin_lock_irqsave(&session->lock, flags);
 	session->state = ISCSI_SESSION_FAILED;
 	spin_unlock_irqrestore(&session->lock, flags);
-	scsi_target_block(&session->dev);
+	scsi_block_targets(shost, &session->dev);
 	ISCSI_DBG_TRANS_SESSION(session, "Completed SCSI target blocking\n");
 	if (session->recovery_tmo >= 0)
 		queue_delayed_work(session->workq,
--- a/drivers/scsi/scsi_transport_srp.c
+++ b/drivers/scsi/scsi_transport_srp.c
@ -396,7 +396,7 @@ static void srp_reconnect_work(struct work_struct *work)
 }

 /*
- * scsi_target_block() must have been called before this function is
+ * scsi_block_targets() must have been called before this function is
 * called to guarantee that no .queuecommand() calls are in progress.
 */
 static void __rport_fail_io_fast(struct srp_rport *rport)
@ -480,7 +480,7 @@ static void __srp_start_tl_fail_timers(struct srp_rport *rport)
 	    srp_rport_set_state(rport, SRP_RPORT_BLOCKED) == 0) {
 		pr_debug("%s new state: %d\n", dev_name(&shost->shost_gendev),
 			 rport->state);
-		scsi_target_block(&shost->shost_gendev);
+		scsi_block_targets(shost, &shost->shost_gendev);
 		if (fast_io_fail_tmo >= 0)
 			queue_delayed_work(system_long_wq,
 					   &rport->fast_io_fail_work,
@ -548,7 +548,7 @@ int srp_reconnect_rport(struct srp_rport *rport)
 		 * later is ok though, scsi_internal_device_unblock_nowait()
 		 * treats SDEV_TRANSPORT_OFFLINE like SDEV_BLOCK.
 		 */
-		scsi_target_block(&shost->shost_gendev);
+		scsi_block_targets(shost, &shost->shost_gendev);
 	res = rport->state != SRP_RPORT_LOST ? i->f->reconnect(rport) : -ENODEV;
 	pr_debug("%s (state %d): transport.reconnect() returned %d\n",
 		 dev_name(&shost->shost_gendev), rport->state, res);
--- a/drivers/scsi/sg.c
+++ b/drivers/scsi/sg.c
@ -71,7 +71,7 @@ static int sg_proc_init(void);

 #define SG_ALLOW_DIO_DEF 0

-#define SG_MAX_DEVS 32768
+#define SG_MAX_DEVS (1 << MINORBITS)

 /* SG_MAX_CDB_SIZE should be 260 (spc4r37 section 3.1.30) however the type
 * of sg_io_hdr::cmd_len can only represent 255. All SCSI commands greater
--- a/drivers/scsi/snic/snic_disc.c
+++ b/drivers/scsi/snic/snic_disc.c
@ -214,7 +214,7 @@ snic_tgt_del(struct work_struct *work)
 		scsi_flush_work(shost);

 	/* Block IOs on child devices, stops new IOs */
-	scsi_target_block(&tgt->dev);
+	scsi_block_targets(shost, &tgt->dev);

 	/* Cleanup IOs */
 	snic_tgt_scsi_abort_io(tgt);
--- a/include/scsi/scsi_device.h
+++ b/include/scsi/scsi_device.h
@ -456,7 +456,7 @@ extern void scsi_scan_target(struct device *parent, unsigned int channel,
 			     unsigned int id, u64 lun,
 			     enum scsi_scan_mode rescan);
 extern void scsi_target_reap(struct scsi_target *);
-extern void scsi_target_block(struct device *);
+void scsi_block_targets(struct Scsi_Host *shost, struct device *dev);
 extern void scsi_target_unblock(struct device *, enum scsi_device_state);
 extern void scsi_remove_target(struct device *);
 extern const char *scsi_device_state_name(enum scsi_device_state);