On Wed, Jan 28, 2004 at 10:55:23PM +0000, Joe Thornber wrote: > On Wed, Jan 28, 2004 at 12:47:56PM -0800, Patrick Mansfield wrote: > > [cc-ing dm-devel] > > If we end up passing all the scsi information up to dm, and it does the > > same things that we already do in scsi (or in block), what is the point of > > putting the code into a separate layer? > > I don't know anything about the scsi layer, so I'm not sure exactly > what the information is that you're after. Is this information just > used for path selction ? If so a scsi specific path-selector could be > written that plugs into the core multipath code. Yes. Primarily whether or not we have a host (per-path) or device specific failure. The solution I have in mind would be non-scsi specific: pass up -errno values, or special BLK_ERR_FOO values. Currently, blk always gives back a -EIO on failure, and no even checks the result (as far as I've audited the code). -- Patrick Mansfield