USB: UAS: fix disconnect by unplugging a hub

commit 325b008723b2dd31de020e85ab9d2e9aa4637d35 upstream.

The SCSI layer can go into an ugly loop if you ignore that a device is
gone. You need to report an error in the command rather than in the
return value of the queue method.

We need to specifically check for ENODEV. The issue goes back to the
introduction of the driver.

Fixes: 115bb1f ("USB: Add UAS driver")
Signed-off-by: Oliver Neukum <[email protected]>
Cc: stable <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

drivers/usb/storage/uas.c

@@ -670,8 +670,7 @@ static int uas_queuecommand_lck(struct scsi_cmnd *cmnd,
 	if (devinfo->resetting) {
 		cmnd->result = DID_ERROR << 16;
 		cmnd->scsi_done(cmnd);
-		spin_unlock_irqrestore(&devinfo->lock, flags);
-		return 0;
+		goto zombie;
 	}
 
 	/* Find a free uas-tag */
@@ -706,6 +705,16 @@ static int uas_queuecommand_lck(struct scsi_cmnd *cmnd,
 		cmdinfo->state &= ~(SUBMIT_DATA_IN_URB | SUBMIT_DATA_OUT_URB);
 
 	err = uas_submit_urbs(cmnd, devinfo);
+	/*
+	 * in case of fatal errors the SCSI layer is peculiar
+	 * a command that has finished is a success for the purpose
+	 * of queueing, no matter how fatal the error
+	 */
+	if (err == -ENODEV) {
+		cmnd->result = DID_ERROR << 16;
+		cmnd->scsi_done(cmnd);
+		goto zombie;
+	}
 	if (err) {
 		/* If we did nothing, give up now */
 		if (cmdinfo->state & SUBMIT_STATUS_URB) {
@@ -716,6 +725,7 @@ static int uas_queuecommand_lck(struct scsi_cmnd *cmnd,
 	}
 
 	devinfo->cmnd[idx] = cmnd;
+zombie:
 	spin_unlock_irqrestore(&devinfo->lock, flags);
 	return 0;
 }