Re: [PATCH] add use_sg_chaining option to scsi_host_template

[FUJITA Tomonori <tomof@acm.org>]

On Mon, 20 Aug 2007 15:05:22 +0200
Jens Axboe <jens.axboe@oracle.com> wrote:

> On Mon, Aug 20 2007, FUJITA Tomonori wrote:
> > On Mon, 20 Aug 2007 09:10:31 +0200
> > Jens Axboe <jens.axboe@oracle.com> wrote:
> > 
> > > On Sat, Aug 18 2007, FUJITA Tomonori wrote:
> > > > On Fri, 17 Aug 2007 01:47:59 +0900
> > > > FUJITA Tomonori <tomof@acm.org> wrote:
> > > > 
> > > > > This is for Jens' sglist branch in the block git tree.
> > > > > 
> > > > > It enables sg chaining support for the LLDs that use scsi_for_each_sg
> > > > > accessor properly.
> > > > > 
> > > > > ---
> > > > > >From a6e50a3b476bc193de103e8c1d95877ced38918e Mon Sep 17 00:00:00 2001
> > > > > From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> > > > > Date: Fri, 17 Aug 2007 01:35:41 +0900
> > > > > Subject: [PATCH] add use_sg_chaining option to scsi_host_template
> > > > > 
> > > > > This option is true if a low-level driver can support sg
> > > > > chaining. This will be removed eventually when all the drivers are
> > > > > converted to support sg chaining. q->max_phys_segments is set to
> > > > > SCSI_MAX_SG_SEGMENTS if false.
> > > > > 
> > > > > Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> > > > > ---
> > > > >  arch/ia64/hp/sim/simscsi.c            |    1 +
> > > > >  drivers/scsi/3w-9xxx.c                |    1 +
> > > > >  drivers/scsi/3w-xxxx.c                |    1 +
> > > > >  drivers/scsi/BusLogic.c               |    1 +
> > > > >  drivers/scsi/NCR53c406a.c             |    3 ++-
> > > > >  drivers/scsi/a100u2w.c                |    1 +
> > > > >  drivers/scsi/aacraid/linit.c          |    1 +
> > > > >  drivers/scsi/aha1740.c                |    1 +
> > > > >  drivers/scsi/aic7xxx/aic79xx_osm.c    |    1 +
> > > > >  drivers/scsi/aic7xxx/aic7xxx_osm.c    |    1 +
> > > > >  drivers/scsi/aic7xxx_old.c            |    1 +
> > > > >  drivers/scsi/arcmsr/arcmsr_hba.c      |    1 +
> > > > >  drivers/scsi/dc395x.c                 |    1 +
> > > > >  drivers/scsi/dpt_i2o.c                |    1 +
> > > > >  drivers/scsi/eata.c                   |    3 ++-
> > > > >  drivers/scsi/hosts.c                  |    1 +
> > > > >  drivers/scsi/hptiop.c                 |    1 +
> > > > >  drivers/scsi/ibmmca.c                 |    1 +
> > > > >  drivers/scsi/ibmvscsi/ibmvscsi.c      |    1 +
> > > > >  drivers/scsi/initio.c                 |    1 +
> > > > >  drivers/scsi/ipr.c                    |    1 +
> > > > 
> > > > I should have dropped ipr since we haven't converted libata yet.
> > > 
> > > But we have, are there still bits missing?
> > 
> > I'm just a bit nervous about possible bugs.
> 
> I hope there aren't any left in libata :-)
> 
> > Probably we need to remove blk_queue_max_phys_segments() in libata
> > for a possible lower sg list size (< 128).
> 
> Indeed, that should go, even in mainline now.
> 
> > BTW, I think that it would be good to test the sglist with a lower sg
> > list size (like 32) to test all the llds.
> 
> That's a good idea, then we'll get the chaining stuff exercised much
> more easily!

Yeah, if you push sglist with setting SCSI_MAX_SG_SEGMENTS to 32,
sglist can get tons of tests. I didn't found the ppc sglist bugs and
didn't lose my data until I tried a small sg list size. :)