On 4/26/21 5:23 PM, Christoph Hellwig wrote:
> On Fri, Apr 23, 2021 at 01:39:19PM +0200, Hannes Reinecke wrote:
>> The message byte can take only two values, COMMAND_COMPLETE and ABORT.
>> So we can easily map ABORT to DID_ABORT and do not set the message byte.
>
> The real question is rather: did anyone care about the message byte,
> and why can't this assignment be dropped entirely?
>
We need to handle this to be compatible with current SCSI EH. There we do
if (msg_byte(scmd->result) != COMMAND_COMPLETE)
return FAILED;
And as it's extremely hard to validate the various stages which the
driver can find itself I couldn't exclude that the ABORT message never
ended up present by the time the command completed.
Hence I'd opted for the careful approach and mapped it to DID_ABORT;
that way I'm sure I don't mess up SCSI EH.
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)
