[PATCH] libata-eh-fw: update ata_exec_internal() for new EH
authorTejun Heo <htejun@gmail.com>
Mon, 15 May 2006 11:58:14 +0000 (20:58 +0900)
committerTejun Heo <htejun@gmail.com>
Mon, 15 May 2006 11:58:14 +0000 (20:58 +0900)
Update ata_exec_internal() such that it uses new EH framework.
->post_internal_cmd() is always invoked regardless of completion
status.  Also, when ata_exec_internal() detects a timeout condition
and new EH is in place, it freezes the port as timeout for normal
commands would do.

Note that ata_port_flush_task() is called regardless of
wait_for_completion status.  This is necessary as exceptions unrelated
to the qc can abort the qc, in which case PIO task could still be
running after the wait for completion returns.

Signed-off-by: Tejun Heo <htejun@gmail.com>
drivers/scsi/libata-core.c

index 35ae5b41f662f32d2c1b78180ba86cabb8561d3d..4def48ed6f46c8ff4d7f7472bfc156d20e73f4d9 100644 (file)
@@ -984,6 +984,7 @@ unsigned ata_exec_internal(struct ata_device *dev,
        DECLARE_COMPLETION(wait);
        unsigned long flags;
        unsigned int err_mask;
+       int rc;
 
        spin_lock_irqsave(&ap->host_set->lock, flags);
 
@@ -1036,20 +1037,25 @@ unsigned ata_exec_internal(struct ata_device *dev,
 
        spin_unlock_irqrestore(&ap->host_set->lock, flags);
 
-       if (!wait_for_completion_timeout(&wait, ATA_TMOUT_INTERNAL)) {
-               ata_port_flush_task(ap);
+       rc = wait_for_completion_timeout(&wait, ATA_TMOUT_INTERNAL);
+
+       ata_port_flush_task(ap);
 
+       if (!rc) {
                spin_lock_irqsave(&ap->host_set->lock, flags);
 
                /* We're racing with irq here.  If we lose, the
                 * following test prevents us from completing the qc
-                * again.  If completion irq occurs after here but
-                * before the caller cleans up, it will result in a
-                * spurious interrupt.  We can live with that.
+                * twice.  If we win, the port is frozen and will be
+                * cleaned up by ->post_internal_cmd().
                 */
                if (qc->flags & ATA_QCFLAG_ACTIVE) {
-                       qc->err_mask = AC_ERR_TIMEOUT;
-                       ata_qc_complete(qc);
+                       qc->err_mask |= AC_ERR_TIMEOUT;
+
+                       if (ap->ops->error_handler)
+                               ata_port_freeze(ap);
+                       else
+                               ata_qc_complete(qc);
 
                        ata_dev_printk(dev, KERN_WARNING,
                                       "qc timeout (cmd 0x%x)\n", command);
@@ -1058,6 +1064,16 @@ unsigned ata_exec_internal(struct ata_device *dev,
                spin_unlock_irqrestore(&ap->host_set->lock, flags);
        }
 
+       /* do post_internal_cmd */
+       if (ap->ops->post_internal_cmd)
+               ap->ops->post_internal_cmd(qc);
+
+       if (qc->flags & ATA_QCFLAG_FAILED && !qc->err_mask) {
+               ata_dev_printk(dev, KERN_WARNING, "zero err_mask for failed "
+                              "internal command, assuming AC_ERR_OTHER\n");
+               qc->err_mask |= AC_ERR_OTHER;
+       }
+
        /* finish up */
        spin_lock_irqsave(&ap->host_set->lock, flags);