On 8/10/22 12:06 PM, Mike Christie wrote: > I think because the cases scsi_noretry_cmd is used for are really specific cases > (scsi_decide_disposition sees NEEDS_RETRY, retries < allowed, and REQ_FAILFAST_DEV > is not set) that might not be very useful. I'm not sure we want to add a bunch of > cases specific to scsi_execute callers to scsi_check_sense? I don't think we do. To be clear, I mean for this approach to be generally useful we would have to add the cases scsi_execute callers are retrying in scsi_check_sense. We could then remove the scsi_execute caller driven retries and only have scsi_decide_disposition retry. It sounded really nice at first, but to do that I would have a bunch of cases that might be specific to pr_ops, alua or scanning, etc. Then I also had cases where user1 does not want to retry but user2 did, so I have to add more SCMD bits. So, in the end it will get messier than you initially think.
