libata: make WARN_ON conditions in ata_sff_hsm_move() more strict
authorTejun Heo <htejun@gmail.com>
Wed, 23 Apr 2008 11:48:36 +0000 (20:48 +0900)
committerJeff Garzik <jgarzik@redhat.com>
Fri, 25 Apr 2008 04:45:09 +0000 (00:45 -0400)
WARN_ON()'s in ata_hsm_move() was too liberal and got triggerred when
it shouldn't (e.g. hotplug events at the right moment).  As the HSM
only deals with device errors and state machine violations, make it
check only against them.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Mark Lord <liml@rtr.ca>
Cc: Albert Lee <albertcc@tw.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
drivers/ata/libata-sff.c

index 15499522e6427392fe80f5fe47f35573d9479d81..2ec65a8fda79ecc46986ed0da6de7d886c71ccf5 100644 (file)
@@ -1208,7 +1208,7 @@ fsm_start:
                DPRINTK("ata%u: dev %u command complete, drv_stat 0x%x\n",
                        ap->print_id, qc->dev->devno, status);
 
-               WARN_ON(qc->err_mask);
+               WARN_ON(qc->err_mask & (AC_ERR_DEV | AC_ERR_HSM));
 
                ap->hsm_task_state = HSM_ST_IDLE;
 
@@ -1222,7 +1222,7 @@ fsm_start:
                /* make sure qc->err_mask is available to
                 * know what's wrong and recover
                 */
-               WARN_ON(qc->err_mask == 0);
+               WARN_ON(!(qc->err_mask & (AC_ERR_DEV | AC_ERR_HSM)));
 
                ap->hsm_task_state = HSM_ST_IDLE;