On Tue, Aug 09, 2022 at 10:41:53PM -0500, Mike Christie wrote: > This adds 2 new SUBMITTED types so we know if a command was queued > because of a scsi_execute user vs SG IO/tape/cd. > > In the next patch we can then handle errors differently based on what > submitted the cmd. For scsi_execute we can let the scsi error handler > retry like normal if the user has requested retries. And for other users > we can fail fast for device errors like is expected by users like SG IO. But do we really want to handle errors differently based on the submitter, or based on what the submitter asks us to do? I.e. why should this be based on the callchain vs the intention?
