[PATCH 0/3] SG v4 support

[PATCH 0/3] SG v4 support

[FUJITA Tomonori]

The following patches add SG v4 support to bsg. They are over Jens'
bsg branch.

I just replaced bsg's SG v3 support with SG v4 and haven't added
anything new yet. So it can only handle SCSI commands.

I've uploaded simple programs to just read/write by using SG v4:

http://www.kernel.org/pub/linux/kernel/people/tomo/sg4/

# They haven't showed up yet. kernel.org is very slow today. please
# wait for a while.

I've not heavily tested this but seems to work. We can begin to do fun
stuff like extending the current code paths for blk_pc (or creates new
paths) for non SCSI commands.

Re: [PATCH 0/3] SG v4 support

[Jens Axboe]

On Sat, Dec 16 2006, FUJITA Tomonori wrote:
> The following patches add SG v4 support to bsg. They are over Jens'
> bsg branch.
> 
> I just replaced bsg's SG v3 support with SG v4 and haven't added
> anything new yet. So it can only handle SCSI commands.

Good start! Just one comment before I look over this and merge it - I'd
prefer keeping this out of sg.c. One of the problems we have right now
are dual pieces of code for sg v3, and I think it would be silly to
continue down this path. Lets keep sg.c as a legacy sg v3 interface
(it'll be the only one except SG_IO in the block layer), and let bsg
take sg v4 and forward.

There's really zero gain in having it in two places.

IOW, put the sg v4 structure in bsg.h.

> I've uploaded simple programs to just read/write by using SG v4:
> 
> http://www.kernel.org/pub/linux/kernel/people/tomo/sg4/

Great!

I'll get this reviewed and merged as soon as I can, probably monday.

-- 
Jens Axboe

Re: [PATCH 0/3] SG v4 support

[James Bottomley]

On Fri, 2006-12-15 at 19:57 +0100, Jens Axboe wrote:
> Good start! Just one comment before I look over this and merge it -
> I'd
> prefer keeping this out of sg.c. One of the problems we have right now
> are dual pieces of code for sg v3, and I think it would be silly to
> continue down this path. Lets keep sg.c as a legacy sg v3 interface
> (it'll be the only one except SG_IO in the block layer), and let bsg
> take sg v4 and forward.
> 
> There's really zero gain in having it in two places.

There is a slight complexity here in that sg is the only thing that
attaches to things like processor devices, or other devices we have no
ULD for; so if we want sg v4 exposed to them, we need to at least route
sg though this ioctl.

James

Re: [PATCH 0/3] SG v4 support

[Jens Axboe]

On Fri, Dec 15 2006, James Bottomley wrote:
> On Fri, 2006-12-15 at 19:57 +0100, Jens Axboe wrote:
> > Good start! Just one comment before I look over this and merge it -
> > I'd
> > prefer keeping this out of sg.c. One of the problems we have right now
> > are dual pieces of code for sg v3, and I think it would be silly to
> > continue down this path. Lets keep sg.c as a legacy sg v3 interface
> > (it'll be the only one except SG_IO in the block layer), and let bsg
> > take sg v4 and forward.
> > 
> > There's really zero gain in having it in two places.
> 
> There is a slight complexity here in that sg is the only thing that
> attaches to things like processor devices, or other devices we have no
> ULD for; so if we want sg v4 exposed to them, we need to at least route
> sg though this ioctl.

Rather a little pain for something like that, than the pain of
duplicating and maintaining that code.

-- 
Jens Axboe

Re: [PATCH 0/3] SG v4 support

[FUJITA Tomonori]

From: Jens Axboe <jens.axboe@oracle.com>
Subject: Re: [PATCH 0/3] SG v4 support
Date: Fri, 15 Dec 2006 19:57:05 +0100

> On Sat, Dec 16 2006, FUJITA Tomonori wrote:
> > The following patches add SG v4 support to bsg. They are over Jens'
> > bsg branch.
> > 
> > I just replaced bsg's SG v3 support with SG v4 and haven't added
> > anything new yet. So it can only handle SCSI commands.
> 
> Good start! Just one comment before I look over this and merge it - I'd
> prefer keeping this out of sg.c. One of the problems we have right now
> are dual pieces of code for sg v3, and I think it would be silly to
> continue down this path. Lets keep sg.c as a legacy sg v3 interface
> (it'll be the only one except SG_IO in the block layer), and let bsg
> take sg v4 and forward.

Yeah, that's what I want to do.


> There's really zero gain in having it in two places.
> 
> IOW, put the sg v4 structure in bsg.h.

That's fine by me. I put the sg v4 structure in sg.h simply I thought
bsg.h is for kernel space and some user-space programs need to include
sg.h for some old ioctl stuff even if they use only sg v4.


> > I've uploaded simple programs to just read/write by using SG v4:
> > 
> > http://www.kernel.org/pub/linux/kernel/people/tomo/sg4/
> 
> Great!
> 
> I'll get this reviewed and merged as soon as I can, probably monday.

Thanks. The patchset is on the top of the two minor patches:

http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265732083&w=2
http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265731963&w=2

The latter needs small modifications if you will change
blk_rq_unmap_user interface.

Re: [PATCH 0/3] SG v4 support

[Jens Axboe]

On Tue, Dec 19 2006, FUJITA Tomonori wrote:
> From: Jens Axboe <jens.axboe@oracle.com>
> Subject: Re: [PATCH 0/3] SG v4 support
> Date: Fri, 15 Dec 2006 19:57:05 +0100
> 
> > On Sat, Dec 16 2006, FUJITA Tomonori wrote:
> > > The following patches add SG v4 support to bsg. They are over Jens'
> > > bsg branch.
> > > 
> > > I just replaced bsg's SG v3 support with SG v4 and haven't added
> > > anything new yet. So it can only handle SCSI commands.
> > 
> > Good start! Just one comment before I look over this and merge it - I'd
> > prefer keeping this out of sg.c. One of the problems we have right now
> > are dual pieces of code for sg v3, and I think it would be silly to
> > continue down this path. Lets keep sg.c as a legacy sg v3 interface
> > (it'll be the only one except SG_IO in the block layer), and let bsg
> > take sg v4 and forward.
> 
> Yeah, that's what I want to do.
> 
> 
> > There's really zero gain in having it in two places.
> > 
> > IOW, put the sg v4 structure in bsg.h.
> 
> That's fine by me. I put the sg v4 structure in sg.h simply I thought
> bsg.h is for kernel space and some user-space programs need to include
> sg.h for some old ioctl stuff even if they use only sg v4.

Ah ok, bsg.h is just a generic header, there's no reason it cannot
contain user-mapped bits as well.

> > > I've uploaded simple programs to just read/write by using SG v4:
> > > 
> > > http://www.kernel.org/pub/linux/kernel/people/tomo/sg4/
> > 
> > Great!
> > 
> > I'll get this reviewed and merged as soon as I can, probably monday.
> 
> Thanks. The patchset is on the top of the two minor patches:
> 
> http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265732083&w=2
> http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265731963&w=2
> 
> The latter needs small modifications if you will change
> blk_rq_unmap_user interface.

I'll rebase the bsg tree as soon as Linus pulls the for-linus branch,
which has a number of SG_IO related changes and fixes. Ah I'll just
rebase bsg now and send you a note, then please resend the pending
patches against that and I'll include it.

-- 
Jens Axboe

Re: [PATCH 0/3] SG v4 support

[Jens Axboe]

On Tue, Dec 19 2006, Jens Axboe wrote:
> > Thanks. The patchset is on the top of the two minor patches:
> > 
> > http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265732083&w=2
> > http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265731963&w=2
> > 
> > The latter needs small modifications if you will change
> > blk_rq_unmap_user interface.
> 
> I'll rebase the bsg tree as soon as Linus pulls the for-linus branch,
> which has a number of SG_IO related changes and fixes. Ah I'll just
> rebase bsg now and send you a note, then please resend the pending
> patches against that and I'll include it.

bsg branch rebased, please send the patch against the current one. HEAD
of tree is fc68dbc4630ffef6dce5198fd90bdec37a9bbf44.

-- 
Jens Axboe

Re: [PATCH 0/3] SG v4 support

[FUJITA Tomonori]

From: Jens Axboe <jens.axboe@oracle.com>
Subject: Re: [PATCH 0/3] SG v4 support
Date: Tue, 19 Dec 2006 15:09:24 +0100

> On Tue, Dec 19 2006, Jens Axboe wrote:
> > > Thanks. The patchset is on the top of the two minor patches:
> > > 
> > > http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265732083&w=2
> > > http://marc.theaimsgroup.com/?l=linux-scsi&m=116621265731963&w=2
> > > 
> > > The latter needs small modifications if you will change
> > > blk_rq_unmap_user interface.
> > 
> > I'll rebase the bsg tree as soon as Linus pulls the for-linus branch,
> > which has a number of SG_IO related changes and fixes. Ah I'll just
> > rebase bsg now and send you a note, then please resend the pending
> > patches against that and I'll include it.
> 
> bsg branch rebased, please send the patch against the current one. HEAD
> of tree is fc68dbc4630ffef6dce5198fd90bdec37a9bbf44.

Thanks. I will update and resend the patches tomorrow.