Handle the unexpected condition where the tape drive reports that tape is rewinding. Patch one in this series was designed to handle an unexpected third party reset condition on the tape device by setting pos_unknown following a POR Unit Attention. Because we do not have access to an Amazon VTL application Laurance and I tried to repoduce the aforementioned POR data corruption problem by using a physical tape drive with a multi-initiator iSCSI gateway. We were easily able to issue the third party reset from initiator 1 while initiator 2 had a backup in progress. We saw the tape drive automatically rewind following the reset, and the st driver on initiator 2 attempt to write a filemark with MTEOM. However, we discovered our tape drive (an HP Ultrium 5-SCSI Z64D) never sends a Unit Attention of any kind. Instead, following the third party reset, the tape drive continually returned "No Sense, Rewind operation in progress". Here are the test results w/out this patch. <<< Rest by other initiator st 33:0:0:0: [st0] Error: 2, cmd: a 0 0 28 0 0 st 33:0:0:0: [st0] Sense Key : No Sense [current] st 33:0:0:0: [st0] Add. Sense: Rewind operation in progress st 33:0:0:0: [st0] Error on write: st 33:0:0:0: [st0] Number of r/w requests 35913, dio used in 35913... st 33:0:0:0: [st0] Async write waits 0, finished 0. st 33:0:0:0: [st0] Error: 2, cmd: 10 0 0 0 1 0 <<< write filemark st 33:0:0:0: [st0] Sense Key : No Sense [current] st 33:0:0:0: [st0] Add. Sense: Rewind operation in progress st 33:0:0:0: [st0] Error on write filemark. st 33:0:0:0: [st0] Buffer flushed, 1 EOF(s) written <<< flush buffer st 33:0:0:0: [st0] Rewinding tape. st 33:0:0:0: [st0] Error: 2, cmd: 1 0 0 0 0 0 st 33:0:0:0: [st0] Sense Key : No Sense [current] st 33:0:0:0: [st0] Add. Sense: Rewind operation in progress With the patch: <<< Rest by other initiator st 32:0:0:0: [st0] Error: 8000002, cmd: a 0 0 28 0 0 st 32:0:0:0: [st0] Sense Key : No Sense [current] st 32:0:0:0: [st0] Add. Sense: Rewind operation in progress st 32:0:0:0: [st0] Error on write: <<< no write filemark or flush buffer >>> st 32:0:0:0: [st0] Number of r/w requests 1624, dio used in 1624... st 32:0:0:0: [st0] Rewinding tape. st 32:0:0:0: [st0] Error: 8000002, cmd: 1 0 0 0 0 0 st 32:0:0:0: [st0] Sense Key : No Sense [current] st 32:0:0:0: [st0] Add. Sense: Rewind operation in progress I'm providing this patch because I think it's valuable for testing purposes and it should be safe. Any time the device unexpectedly reports "Rewind is in progress", it should be safe to set pos_unknown in the driver. Tested-by: Laurence Oberman <loberman@xxxxxxxxxx> Signed-off-by: John Meneghini <jmeneghi@xxxxxxxxxx> --- drivers/scsi/st.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/scsi/st.c b/drivers/scsi/st.c index 338aa8c42968..b641490ed9d1 100644 --- a/drivers/scsi/st.c +++ b/drivers/scsi/st.c @@ -416,6 +416,9 @@ static int st_chk_result(struct scsi_tape *STp, struct st_request * SRpnt) STp->cleaning_req = 1; /* ASC and ASCQ => cleaning requested */ if (cmdstatp->have_sense && scode == UNIT_ATTENTION && cmdstatp->sense_hdr.asc == 0x29) STp->pos_unknown = 1; /* ASC => power on / reset */ + if (cmdstatp->have_sense && cmdstatp->sense_hdr.asc == 0 + && cmdstatp->sense_hdr.ascq == 0x1a) + STp->pos_unknown = 1; /* ASCQ => rewind in progress */ STp->pos_unknown |= STp->device->was_reset; -- 2.39.3
