On Thu, 2014-12-11 at 22:12 -0500, Martin K. Petersen wrote: > >>>>> "Vlad" == Vladislav Bolkhovitin <vst@xxxxxxxx> writes: > > Vlad, > > Vlad> We are currently developing a SCSI target system with T10-PI. We > Vlad> are using block integrity interface and found a problem that this > Vlad> interface fundamentally can not pass Oracle T10-PI certification > Vlad> tests. Those tests require to receive on the initiator side > Vlad> information about which particular tag failed the target checks, > Vlad> but the block integrity interface does not preserve this > Vlad> information, hence the target can not deliver it to the initiator > Vlad> => certification failure. The storage provides the right sense, > Vlad> but then in scsi_io_completion() it is dropped and replaced by a > Vlad> single EILSEQ. > > Vlad> What would be the best way to fix that? By making a patch > Vlad> introducing new -EXXXXXX error codes for the PI errors? > > I posted such a patch a while back. We use that in our qualification > tooling to ensure that the right things are reported when a PI error is > injected at various places in the stack. > > One thing that needs to be done is to make returning these new errors to > userland conditional on !BIP_BLOCK_INTEGRITY. I'll put that on my list. > Any chance this patch will be merged up soon..? I still need the same logic for IBLOCK backends to propagate up the correct sense codes from bio completion in the event of GUARD check or REF tag failure. Also, AFAICT this patch still doesn't set the failed LBA, right..? Any thoughts about the most sane way to get the failed LBA information back to a caller like IBLOCK performing submission with bio + bip..? --nab -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
