On 6/7/21 2:30 PM, Martin K. Petersen wrote:
>
> Hannes,
>
>>> Any ideas?
>
>>> Can you enable SCSI logging via
>>
>> scsi.scsi_logging_level=216
>>
>> on the kernel commandline and send me the output?
>
> You now effectively set SAM_STAT_CHECK_CONDITION if the scsi_cmnd has a
> sense buffer.
>
> The original code only set DRIVER_SENSE if the adapter response actually
> contained sense information:
>
> @@ -161,8 +161,7 @@ static void virtscsi_complete_cmd(struct virtio_scsi *vscsi, void *buf)
> min_t(u32,
> virtio32_to_cpu(vscsi->vdev, resp->sense_len),
> VIRTIO_SCSI_SENSE_SIZE));
> - if (resp->sense_len)
> - set_driver_byte(sc, DRIVER_SENSE);
> + set_status_byte(sc, SAM_STAT_CHECK_CONDITION);
> }
>
Oh, I know. But we're checking for a valid sense code during scanning:
if (scsi_status_is_check_condition(result) &&
scsi_sense_valid(&sshdr)) {
so if that makes a difference it would mean that the virtio driver has
some stale sense data which then gets copied over.
Anyway.
Can you test with this patch?
diff --git a/drivers/scsi/virtio_scsi.c b/drivers/scsi/virtio_scsi.c
index fd69a03d6137..0cb1182fd734 100644
--- a/drivers/scsi/virtio_scsi.c
+++ b/drivers/scsi/virtio_scsi.c
@@ -161,7 +161,8 @@ static void virtscsi_complete_cmd(struct virtio_scsi
*vscsi, void *buf)
min_t(u32,
virtio32_to_cpu(vscsi->vdev, resp->sense_len),
VIRTIO_SCSI_SENSE_SIZE));
- set_status_byte(sc, SAM_STAT_CHECK_CONDITION);
+ if (resp->sense_len)
+ set_status_byte(sc, SAM_STAT_CHECK_CONDITION);
}
sc->scsi_done(sc);
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@xxxxxxx +49 911 74053 688
SUSE Software Solutions Germany GmbH, 90409 Nürnberg
GF: F. Imendörffer, HRB 36809 (AG Nürnberg)
