>>>>> "Mike" == Mike Snitzer <snitzer@xxxxxxxxxx> writes: Mike> The SCSI, or block layer, should return a non-retryable error for Mike> this case. But we only have the differentiated IO errors for SCSI Mike> cmds that are issued, so it seems we still need to train SCSI (and Mike> block by association/dependency) to return permanent errors that Mike> are identified during request preparation. There currently isn't any way to make BLKPREP_KILL return anything other than -EIO. I'll take a look tomorrow... -- Martin K. Petersen Oracle Linux Engineering -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
