Re: [PATCH] always handle REQ_BLOCK_PC requests in common code

[Christoph Hellwig <hch@lst.de>]

On Sat, Jan 07, 2006 at 04:26:43PM +1000, Douglas Gilbert wrote:
> Christoph Hellwig wrote:
> > LLDDs should never see REQ_BLOCK_PC requests, we can handle them just
> > fine in the core code.  There is a small behaviour change in that some
> > check in sr's rw_intr are bypassed, but I consider the old behaviour
> > a bug.
> > 
> > Mike found this cleanup ooportunity and provdided early patches, so all
> > the credit goes to him, even if I redid the patches from scratch beause
> > that was easier than forward-porting the old patches.
> 
> The point of the hacks in the sd driver was to make
> sure SG_IO ioctl commands issued via a sd device node
> would not get caught up in the sd driver's error
> processing (e.g. retries on MEDIUM ERRORs). Is there
> a clean path back for such errors to the SG_IO ioctl
> and hence the user space?

Yeah.  scsi_io_completion deals specially with REQ_BLOCK_PC requests,
so we don't do these retries on MEDIUM ERRORs.  sg even before my patch
directly handed off these requests to scsi_io_completion, and didn't do
any own handling of errors for them.  There's zero change in behaviour
for sd and st.  Only sr did some error processesing previously which
was wrong.