[SCSI] Avoid dangling pointer in scsi_requeue_command()
authorBart Van Assche <bvanassche@acm.org>
Fri, 29 Jun 2012 15:34:26 +0000 (15:34 +0000)
committerJames Bottomley <JBottomley@Parallels.com>
Fri, 20 Jul 2012 07:58:40 +0000 (08:58 +0100)
When we call scsi_unprep_request() the command associated with the request
gets destroyed and therefore drops its reference on the device.  If this was
the only reference, the device may get released and we end up with a NULL
pointer deref when we call blk_requeue_request.

Reported-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Reviewed-by: Tejun Heo <tj@kernel.org>
Cc: <stable@kernel.org>
[jejb: enhance commend and add commit log for stable]
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
drivers/scsi/scsi_lib.c

index 4acf5c284c2e54e0de2b4bf5195b0ce6cb583d42..0e52ff03513585a00b904c045343403fbb5c40a2 100644 (file)
@@ -479,15 +479,26 @@ void scsi_requeue_run_queue(struct work_struct *work)
  */
 static void scsi_requeue_command(struct request_queue *q, struct scsi_cmnd *cmd)
 {
+       struct scsi_device *sdev = cmd->device;
        struct request *req = cmd->request;
        unsigned long flags;
 
+       /*
+        * We need to hold a reference on the device to avoid the queue being
+        * killed after the unlock and before scsi_run_queue is invoked which
+        * may happen because scsi_unprep_request() puts the command which
+        * releases its reference on the device.
+        */
+       get_device(&sdev->sdev_gendev);
+
        spin_lock_irqsave(q->queue_lock, flags);
        scsi_unprep_request(req);
        blk_requeue_request(q, req);
        spin_unlock_irqrestore(q->queue_lock, flags);
 
        scsi_run_queue(q);
+
+       put_device(&sdev->sdev_gendev);
 }
 
 void scsi_next_command(struct scsi_cmnd *cmd)