On 10/19/23 22:30, Mike Christie wrote:
On 10/16/23 7:15 AM, Hannes Reinecke wrote:
@@ -1671,7 +1682,8 @@ static int scsi_eh_target_reset(struct Scsi_Host *shost,
if (rtn == SUCCESS)
list_move_tail(&scmd->eh_entry, &check_list);
else if (rtn == FAST_IO_FAIL)
- scsi_eh_finish_cmd(scmd, done_q);
+ __scsi_eh_finish_cmd(scmd, done_q,
+ DID_TRANSPORT_DISRUPTED);
else
/* push back on work queue for further processing */
list_move(&scmd->eh_entry, work_q);
@@ -1736,8 +1748,9 @@ static int scsi_eh_bus_reset(struct Scsi_Host *shost,
list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
if (channel == scmd_channel(scmd)) {
if (rtn == FAST_IO_FAIL)
- scsi_eh_finish_cmd(scmd,
- done_q);
+ __scsi_eh_finish_cmd(scmd,
+ done_q,
+ DID_TRANSPORT_DISRUPTED);
else
list_move_tail(&scmd->eh_entry,
&check_list);
@@ -1780,9 +1793,9 @@ static int scsi_eh_host_reset(struct Scsi_Host *shost,
if (rtn == SUCCESS) {
list_splice_init(work_q, &check_list);
} else if (rtn == FAST_IO_FAIL) {
- list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
- scsi_eh_finish_cmd(scmd, done_q);
- }
+ list_for_each_entry_safe(scmd, next, work_q, eh_entry)
+ __scsi_eh_finish_cmd(scmd, done_q,
+ DID_TRANSPORT_DISRUPTED);
} else {
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
For FAST_IO_FAIL I think you want to use DID_TRANSPORT_FAILFAST because when
drivers return that, they normally have hit their fast io fail timer or have
hit a hard transport issue like the port is offline. For example for FC drivers
they do:
Ah, yes, you are right. Will be modifying that.
err = fc_block_rport(rport);
if (err)
return err;
where fc_block_rport does:
if (rport->flags & FC_RPORT_FAST_FAIL_TIMEDOUT)
return FAST_IO_FAIL;
and then for fc_remote_port_chkready we return DID_TRANSPORT_FAILFAST
when FC_RPORT_FAST_FAIL_TIMEDOUT is set.
So using DID_TRANSPORT_FAILFAST would align the return values for that
state.
One question I had is why you added those checks for target and host
reset but not scsi_eh_bus_device_reset because drivers will do something
similar to above where they call fc_block_rport for that callout as
well.
That's done in one of the later patches where I convert the loop in
scsi_eh_bus_device_reset().
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@xxxxxxx +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Ivo Totev, Andrew
Myers, Andrew McDonald, Martje Boudien Moerman
