On Sat, 2005-07-09 at 19:06 -0400, Ben Collins wrote: > Alright, I need some explanation on these changes to sbp2. Lots of things > ripped out. The patch seems reasonably explanatory: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=631e8a1398ce4cfef8b30678d51daf0c64313a09 > I can understand that TYPE_RDC is what we had as TYPE_SDAD. Now, in our > tree for TYPE_RDC, we converted it to TYPE_DISK. We also did a lot of mode > conversions for DISK/RDC/ROM types. > > My question is, why were the conversions all removed? The conversions, as > far as I know, are related to SBP protocol, and not SCSI, so why would the > SCSI maintainers feel the need to rip out an important part of the SBP2 > driver? Note, this isn't really a SCSI device, or a scsi host controller. > It's a protocol translation layer. TYPE_RBC is essentially a TYPE_DISK but with a few weird quirks, mainly in the mode page setups. The object of the changes is to make sd.c treat TYPE_RBC correctly without any need of translations. I'm not sure where you get the idea that SBP isn't a SCSI protocol from: it was defined and ratified by the T10 SCSI committee (http://www.t10.org/ftp/t10/drafts/sbp2/sbp2r04.pdf). The RBC command set was designed for the SBP protocol, but not for exclusive use by it. The idea of the changes is to treat SBP for what it is (a transport protocol) and have the ULDs understand the actual command protocol (RBC) so there's no need for a protocol translation layer at all (this is the bit that got removed). This has the added benefit that the sd ULD is ready for the day USB or other device vendors adopt RBC. James - : 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
