[PATCH] USB: storage: Fix messed-up locking

This is patch as550 from Alan Stern.

Apparently someone changed the SCSI core so that it no longer holds the
host lock when doing a device or bus reset.  usb-storage was updated at
the time, but the change was done carelessly.  Some of the code depends
on that lock being held.

This patch reintroduces the host lock where needed and tries to clarify
the comments explaining why the lock is necessary.  It also moves the
code that clears the TIMED_OUT and ABORTING bitflags so that it executes
as soon as the timed-out command has completed (and while the host lock
is held).
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Matthew Dharm <mdharm-usb@one-eyed-alien.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drivers/usb/storage/scsiglue.c

@@ -227,42 +227,42 @@ static int queuecommand(struct scsi_cmnd *srb,
  ***********************************************************************/
 
 /* Command timeout and abort */
-/* This is always called with scsi_lock(host) held */
 static int command_abort(struct scsi_cmnd *srb)
 {
 	struct us_data *us = host_to_us(srb->device->host);
 
 	US_DEBUGP("%s called\n", __FUNCTION__);
 
+	/* us->srb together with the TIMED_OUT, RESETTING, and ABORTING
+	 * bits are protected by the host lock. */
+	scsi_lock(us_to_host(us));
+
 	/* Is this command still active? */
 	if (us->srb != srb) {
+		scsi_unlock(us_to_host(us));
 		US_DEBUGP ("-- nothing to abort\n");
 		return FAILED;
 	}
 
 	/* Set the TIMED_OUT bit.  Also set the ABORTING bit, but only if
 	 * a device reset isn't already in progress (to avoid interfering
-	 * with the reset).  To prevent races with auto-reset, we must
-	 * stop any ongoing USB transfers while still holding the host
-	 * lock. */
+	 * with the reset).  Note that we must retain the host lock while
+	 * calling usb_stor_stop_transport(); otherwise it might interfere
+	 * with an auto-reset that begins as soon as we release the lock. */
 	set_bit(US_FLIDX_TIMED_OUT, &us->flags);
 	if (!test_bit(US_FLIDX_RESETTING, &us->flags)) {
 		set_bit(US_FLIDX_ABORTING, &us->flags);
 		usb_stor_stop_transport(us);
 	}
+	scsi_unlock(us_to_host(us));
 
 	/* Wait for the aborted command to finish */
 	wait_for_completion(&us->notify);
-
-	/* Reacquire the lock and allow USB transfers to resume */
-	clear_bit(US_FLIDX_ABORTING, &us->flags);
-	clear_bit(US_FLIDX_TIMED_OUT, &us->flags);
 	return SUCCESS;
 }
 
 /* This invokes the transport reset mechanism to reset the state of the
  * device */
-/* This is always called with scsi_lock(host) held */
 static int device_reset(struct scsi_cmnd *srb)
 {
 	struct us_data *us = host_to_us(srb->device->host);
@@ -279,7 +279,6 @@ static int device_reset(struct scsi_cmnd *srb)
 }
 
 /* Simulate a SCSI bus reset by resetting the device's USB port. */
-/* This is always called with scsi_lock(host) held */
 static int bus_reset(struct scsi_cmnd *srb)
 {
 	struct us_data *us = host_to_us(srb->device->host);
@@ -291,7 +290,6 @@ static int bus_reset(struct scsi_cmnd *srb)
 	result = usb_stor_port_reset(us);
 	up(&(us->dev_semaphore));
 
-	/* lock the host for the return */
 	return result < 0 ? FAILED : SUCCESS;
 }
 

drivers/usb/storage/usb.c

@@ -392,11 +392,16 @@ static int usb_stor_control_thread(void * __us)
 		/* If an abort request was received we need to signal that
 		 * the abort has finished.  The proper test for this is
 		 * the TIMED_OUT flag, not srb->result == DID_ABORT, because
-		 * a timeout/abort request might be received after all the
-		 * USB processing was complete. */
-		if (test_bit(US_FLIDX_TIMED_OUT, &us->flags))
+		 * the timeout might have occurred after the command had
+		 * already completed with a different result code. */
+		if (test_bit(US_FLIDX_TIMED_OUT, &us->flags)) {
 			complete(&(us->notify));
 
+			/* Allow USB transfers to resume */
+			clear_bit(US_FLIDX_ABORTING, &us->flags);
+			clear_bit(US_FLIDX_TIMED_OUT, &us->flags);
+		}
+
 		/* finished working on this command */
 		us->srb = NULL;
 		scsi_unlock(host);