[PATCH] scsi: Keep interrupts disabled while submitting requests

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



scsi_request_fn() can be called from softirq context during IO
completion. If it enables interrupts there, HW interrupts can interrupt
softirq processing and queue more IO completion work which can
eventually lead to softlockup reports because IO completion softirq runs
for too long. Keep interrupts disabled in scsi_request_fn().

Signed-off-by: Jan Kara <jack@xxxxxxx>
---
 drivers/scsi/scsi_lib.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index f7e316368c99..44b867e9adc9 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1481,7 +1481,8 @@ static void scsi_softirq_done(struct request *rq)
  *
  * Returns:     Nothing
  *
- * Lock status: IO request lock assumed to be held when called.
+ * Lock status: IO request lock assumed to be held when called, interrupts
+ * must be disabled.
  */
 static void scsi_request_fn(struct request_queue *q)
 	__releases(q->queue_lock)
@@ -1563,7 +1564,7 @@ static void scsi_request_fn(struct request_queue *q)
 		 * XXX(hch): This is rather suboptimal, scsi_dispatch_cmd will
 		 *		take the lock again.
 		 */
-		spin_unlock_irq(shost->host_lock);
+		spin_unlock(shost->host_lock);
 
 		/*
 		 * Finally, initialize any error handling parameters, and set up
@@ -1575,7 +1576,7 @@ static void scsi_request_fn(struct request_queue *q)
 		 * Dispatch the command to the low-level driver.
 		 */
 		rtn = scsi_dispatch_cmd(cmd);
-		spin_lock_irq(q->queue_lock);
+		spin_lock(q->queue_lock);
 		if (rtn)
 			goto out_delay;
 	}
@@ -1583,7 +1584,7 @@ static void scsi_request_fn(struct request_queue *q)
 	return;
 
  not_ready:
-	spin_unlock_irq(shost->host_lock);
+	spin_unlock(shost->host_lock);
 
 	/*
 	 * lock q, handle tag, requeue req, and decrement device_busy. We
@@ -1593,7 +1594,7 @@ static void scsi_request_fn(struct request_queue *q)
 	 * cases (host limits or settings) should run the queue at some
 	 * later time.
 	 */
-	spin_lock_irq(q->queue_lock);
+	spin_lock(q->queue_lock);
 	blk_requeue_request(q, req);
 	sdev->device_busy--;
 out_delay:
-- 
1.8.1.4

_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs




[Index of Archives]     [Linux XFS Devel]     [Linux Filesystem Development]     [Filesystem Testing]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux