On Fri, 2012-11-09 at 16:33 +0000, Elliott, Robert (Server Storage) wrote: > I recommend broadening this patch. T10 is discussing making READ > (10), WRITE (10), etc. obsolete in SBC-4 in favor of their 16-byte CDB > counterparts. > > The algorithm should be: > 1. During discovery, determine if 16-byte CDBs are supported. There > are several ways to determine this: > a) REPORT SUPPORTED OPERATION CODES command succeeds and reports that > READ (16) et al are supported. > b) READ (16) command specifying a Transfer Length of zero succeeds. > c) READ CAPACITY (16) command succeeds and reports that the capacity > is > 2 TiB. > d) (future) INQUIRY command succeeds fetching the Block Device > Characteristics VPD page and notices a new field added by the SBC-4 > simplified SCSI feature set proposal.
When you consider the problem that we must support all devices (which means the older ones as well), including the annoying USB ones which crash on unexpected commands, you can see that d) is about the only viable option. We can also force RC16 if the capacity is over 2^32 blocks, because the command will be required, so that's probably the place to start. James > Since REPORT SUPPORTED OPERATION CODES is optional, it won't always > work. READ CAPACITY (16) used to be optional for < 2 TiB drives, so it > won't always work either. READ (16) will always work, but requires the > drive to be spun up beforehand (e.g., with START STOP UNIT). > > 2. if 16-byte CDBs are supported, then use them; only drop down to > 10-byte CDBs if 16-byte CDBs are unavailable. Don't make the decision > by comparing the LBA on every IO. > -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
