Commit 42e22cac4e57f3e0b4b631c9489effe97f7d7d6c

Authored by K. Y. Srinivasan
Committed by James Bottomley
1 parent c62eef0d1b

[SCSI] storvsc: Properly handle errors from the host

Hyper-V cannot process some commands like ATA_12 and ATA_16. It also returns a
very generic error when this happens (SRB_STATUS_ERROR).  Most of the time we
treat SRB_STATUS_ERROR as DID_TARGET_FAILURE which causes error handler retry,
but in the case of pass through commands, they'll never succeed (and the error
handler will offline the device), so put a discriminating block in the command
completion routing and send the SRB_STATUS_ERROR upwards with DID_PASSTHROUGH
for commands we know should not be retried.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Showing 1 changed file with 15 additions and 5 deletions Side-by-side Diff

drivers/scsi/storvsc_drv.c
... ... @@ -785,12 +785,22 @@
785 785 /*
786 786 * If there is an error; offline the device since all
787 787 * error recovery strategies would have already been
788   - * deployed on the host side.
  788 + * deployed on the host side. However, if the command
  789 + * were a pass-through command deal with it appropriately.
789 790 */
790   - if (vm_srb->srb_status == SRB_STATUS_ERROR)
791   - scmnd->result = DID_TARGET_FAILURE << 16;
792   - else
793   - scmnd->result = vm_srb->scsi_status;
  791 + scmnd->result = vm_srb->scsi_status;
  792 +
  793 + if (vm_srb->srb_status == SRB_STATUS_ERROR) {
  794 + switch (scmnd->cmnd[0]) {
  795 + case ATA_16:
  796 + case ATA_12:
  797 + set_host_byte(scmnd, DID_PASSTHROUGH);
  798 + break;
  799 + default:
  800 + set_host_byte(scmnd, DID_TARGET_FAILURE);
  801 + }
  802 + }
  803 +
794 804  
795 805 /*
796 806 * If the LUN is invalid; remove the device.