Martin, > Introduce a new blist flag that indicates the device may return certain > sense code/ASC/ASCQ combinations that indicate different treatment than > normal. In particular, some devices need unconditional retry (aka > ADD_TO_MLQUEUE) under certain conditions; otherwise path failures may be > falsely detected in the "maybe_retry" case.
I'm still not convinced of the merits of all this complexity to save one bit in a per-device data structure. Introducing a BLIST_RETRY_ITF entry for 44/xx might be generally useful for devices other than Symmetrix. So let's just do that. As far as the vendor-specific Fujitsu ASC/ASCQ combo, I'm also OK with a quirk flag. If we end up with many devices that needs vendor-specific treatment we can revisit the bigger patch. But for now I think it's an overkill. Only caveat is that I think you need to tweak the blist_flags_t stuff to accommodate the extra bits... -- Martin K. Petersen Oracle Linux Engineering
