The handling of REQUEST SENSE is incorrect in sbd_scsi.c . There are certain
BIOSes which are very particular about how request sense is handled. I saw some
failures recently in booting and upon investigating, discovered this issue. In
response to the command, the LU is suppose to return sense data in a data-in
phase (even if there is no additional sense information). So basically this is
how this code should look like (inside sbd_new_task() ):
if (cdb0 == SCMD_REQUEST_SENSE) {
if ((cdb1 & ~1) || task->task_cdb[2] || task->task_cdb[3] ||
task->task_cdb[5]) {
stmf_scsilib_send_status(task, STATUS_CHECK,
STMF_SAA_INVALID_FIELD_IN_CDB);
} else {
sbd_handle_request_sense(task, initial_dbuf);
}
return;
}
And the sbd_handle_request_sense should look like this:
static uint8_t sbd_request_sense_payload[] = {
/* Fixed format payload to say no additional sense */
0x70, 0, 0, 0, 0, 0, 0, 10,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
void
sbd_handle_request_sense(struct scsi_task *task,
struct stmf_data_buf *initial_dbuf)
{
uint8_t *cdbp = (uint8_t *)&task->task_cdb[0];
uint32_t cmd_size;
uint32_t xfer_size = sizeof (scd_request_sense_payload);
cmd_size = cdbp[4];
if (cmd_size == 0) {
task->task_cmd_xfer_length = 0;
if (task->task_additional_flags &
TASK_AF_NO_EXPECTED_XFER_LENGTH) {
task->task_expected_xfer_length = 0;
}
stmf_scsilib_send_status(task, STATUS_GOOD, 0);
return;
}
sbd_handle_short_read_transfers(task, initial_dbuf,
sbd_request_sense_payload, cmd_size, min(cmd_size, xfer_size));
}
Sumit
--
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
storage-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
