On Wed, Mar 23 2005, Tejun Heo wrote:
>
> Hello, James.
>
> James Bottomley wrote:
> >On Wed, 2005-03-23 at 11:14 +0900, Tejun Heo wrote:
> >
> >>01_scsi_remove_scsi_release_buffers.patch
> >>
> >> Buffer bouncing hasn't been done inside the scsi midlayer for
> >> quite sometime now, but bounce-buffer release paths are still
> >> around. This patch removes these unused paths.
> >
> >
> >Yes, but scsi_release_buffers isn't referring to bounce buffers anymore,
> >it's simply releasing the sg buffers.
> >
>
> That's what I did. Replacing scsi_release_buffers() calls with calls
> to scsi_free_sgtable(). The only logic removed is bounce-buffer
> release/copy-back.
>
> >[...]
> >
> >>- else if (cmd->buffer != req->buffer) {
> >>- if (rq_data_dir(req) == READ) {
> >>- unsigned long flags;
> >>- char *to = bio_kmap_irq(req->bio, &flags);
> >>- memcpy(to, cmd->buffer, cmd->bufflen);
> >>- bio_kunmap_irq(to, &flags);
> >>- }
> >>- kfree(cmd->buffer);
> >>- }
> >
> >
> >I'll defer to Jens here, but I don't thing you can just remove this ...
> >sg_io with a misaligned buffer will fail without this.
>
> AFAIK, those are done by blk_rq_map_user() and blk_rq_unmap_user(),
> both of which are invoked directly by sg_io().
>
> >That rather nasty code freeing cmd->buffer needs to be in there as
> >well ... so it does make sense to keep this API
>
> That code is invoked only for REQ_BLOCK_PC requests without bio, and I
> digged pretty hard but, in those cases, AFAICT, the callers are
> responsible for supplying dma-able buffers and nothing seems to alter
> cmd->buffer after the cmd gets initialized, but I might be missing
> things here. If so, please point out.
That did not use to be true - eg request coming from the CDROM layer to
sr had to be bounced in the scsi layer for isa host adapters. I bet that
is still true.
--
Jens Axboe
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
