[GIT PULL] Block drivers for 3.3-rc

[GIT PULL] Block drivers for 3.3-rc

[Jens Axboe]

Hi Linus,

Please see below for the 3.3-rc1 block driver pull request.

Essentially two different categories here:

- Adds the driver for the Micron RealSSD P320 pci-express flash card.

- Discard improvements for xen block front/back driver

This sits "on top of" the for-3.3/core pull request. Please pull.


  git://git.kernel.dk/linux-block.git for-3.3/drivers


Asai Thambi S P (2):
      mtip32xx: updates based on feedback
      mtip32xx: do rebuild monitoring asynchronously

Dan Carpenter (1):
      mtip32xx: uninitialized variable in mtip_quiesce_io()

Jens Axboe (8):
      mtip32xx: fix warnings/errors on 32-bit compiles
      mtip32xx: cleanup compat ioctl handling
      mtip32xx: ensure that all local functions are static
      mtip32xx: mark a few more items static
      mtip32xx: add module.h include to avoid conflict with moduleh tree
      mtip32xx: update to new ->make_request() API
      Merge branch 'stable/for-jens-3.3' of git://git.kernel.org/.../konrad/xen into for-3.3/drivers
      Merge branch 'for-3.3/mtip32xx' into for-3.3/drivers

Konrad Rzeszutek Wilk (3):
      xen/blk[front|back]: Squash blkif_request_rw and blkif_request_discard together
      xen/blk[front|back]: Enhance discard support with secure erasing support.
      xen/blkback: Move processing of BLKIF_OP_DISCARD from dispatch_rw_block_io

Li Dongyang (1):
      xen-blkback: convert hole punching to discard request on loop devices

Sam Bradshaw (1):
      block: Add driver for Micron RealSSD pcie flash cards

Thomas Meyer (1):
      xen-blkfront: Use kcalloc instead of kzalloc to allocate array

 drivers/block/Kconfig               |    2 +
 drivers/block/Makefile              |    1 +
 drivers/block/mtip32xx/Kconfig      |    9 +
 drivers/block/mtip32xx/Makefile     |    5 +
 drivers/block/mtip32xx/mtip32xx.c   | 3651 +++++++++++++++++++++++++++++++++++
 drivers/block/mtip32xx/mtip32xx.h   |  423 ++++
 drivers/block/xen-blkback/blkback.c |   84 +-
 drivers/block/xen-blkback/common.h  |   67 +-
 drivers/block/xen-blkback/xenbus.c  |   12 +
 drivers/block/xen-blkfront.c        |   79 +-
 include/xen/interface/io/blkif.h    |   40 +-
 11 files changed, 4261 insertions(+), 112 deletions(-)
 create mode 100644 drivers/block/mtip32xx/Kconfig
 create mode 100644 drivers/block/mtip32xx/Makefile
 create mode 100644 drivers/block/mtip32xx/mtip32xx.c
 create mode 100644 drivers/block/mtip32xx/mtip32xx.h

-- 
Jens Axboe

Re: [GIT PULL] Block drivers for 3.3-rc

[Christoph Hellwig]

On Sun, Jan 15, 2012 at 10:52:12AM +0100, Jens Axboe wrote:
> Hi Linus,
> 
> Please see below for the 3.3-rc1 block driver pull request.
> 
> Essentially two different categories here:
> 
> - Adds the driver for the Micron RealSSD P320 pci-express flash card.

The REQ_FUA and REQ_FLUSH handling in it is still completely broken,
with no attempt to fix it since it last came up.

So a pretty clear NAK on that driver from me.

Re: [GIT PULL] Block drivers for 3.3-rc

[Jens Axboe]

On 2012-01-15 11:33, Christoph Hellwig wrote:
> On Sun, Jan 15, 2012 at 10:52:12AM +0100, Jens Axboe wrote:
>> Hi Linus,
>>
>> Please see below for the 3.3-rc1 block driver pull request.
>>
>> Essentially two different categories here:
>>
>> - Adds the driver for the Micron RealSSD P320 pci-express flash card.
> 
> The REQ_FUA and REQ_FLUSH handling in it is still completely broken,
> with no attempt to fix it since it last came up.
> 
> So a pretty clear NAK on that driver from me.

Good point, I had forgotten about the ordered writes, I believe the rest
of the items have been handled. The driver/hw, as it stands, does not
reorder writes and it doesn't have a write back cache. So it should
actually be safe. Any acked write is stable. Sam, correct me if I'm
wrong.

-- 
Jens Axboe

Re: [GIT PULL] Block drivers for 3.3-rc

[Christoph Hellwig]

On Sun, Jan 15, 2012 at 01:22:05PM +0100, Jens Axboe wrote:
> Good point, I had forgotten about the ordered writes, I believe the rest
> of the items have been handled. The driver/hw, as it stands, does not
> reorder writes and it doesn't have a write back cache. So it should
> actually be safe. Any acked write is stable. Sam, correct me if I'm
> wrong.

I tried to get that out of the micron folks a while ago but don't
remember getting an answer.  If that's actually tree please remove
the (misnamed) barrier argument from mtip_hw_submit_io, and maybe add
a comment that the current hardware can't have a volatile write cache.

Btw, does the hardware claim to have writeback caches in the ATA
IDENTIFY words related to it?

Re: [GIT PULL] Block drivers for 3.3-rc

[Jens Axboe]

On 2012-01-15 13:44, Christoph Hellwig wrote:
> On Sun, Jan 15, 2012 at 01:22:05PM +0100, Jens Axboe wrote:
>> Good point, I had forgotten about the ordered writes, I believe the rest
>> of the items have been handled. The driver/hw, as it stands, does not
>> reorder writes and it doesn't have a write back cache. So it should
>> actually be safe. Any acked write is stable. Sam, correct me if I'm
>> wrong.
> 
> I tried to get that out of the micron folks a while ago but don't
> remember getting an answer.  If that's actually tree please remove
> the (misnamed) barrier argument from mtip_hw_submit_io, and maybe add
> a comment that the current hardware can't have a volatile write cache.

Sure, will do post pull to not mess up the current pull request.

> Btw, does the hardware claim to have writeback caches in the ATA
> IDENTIFY words related to it?

I'll let Sam answer this one and the question above, it's better if it
comes from the source.

-- 
Jens Axboe

RE: [GIT PULL] Block drivers for 3.3-rc

[Sam Bradshaw (sbradshaw)]

> On 2012-01-15 13:44, Christoph Hellwig wrote:
> > On Sun, Jan 15, 2012 at 01:22:05PM +0100, Jens Axboe wrote:
> >> Good point, I had forgotten about the ordered writes, I believe the rest
> >> of the items have been handled. The driver/hw, as it stands, does not
> >> reorder writes and it doesn't have a write back cache. So it should
> >> actually be safe. Any acked write is stable. Sam, correct me if I'm
> >> wrong.
> >
> > I tried to get that out of the micron folks a while ago but don't
> > remember getting an answer.  If that's actually tree please remove
> > the (misnamed) barrier argument from mtip_hw_submit_io, and maybe add
> > a comment that the current hardware can't have a volatile write cache.
> 
> Sure, will do post pull to not mess up the current pull request.
> 
> > Btw, does the hardware claim to have writeback caches in the ATA
> > IDENTIFY words related to it?
> 
> I'll let Sam answer this one and the question above, it's better if it
> comes from the source.

Apologies for not communicating this clearly to the mailing list.  The P320 does not have and will not claim writeback cache support in the IDENTIFY.

We will promptly submit a patch that renames the barrier argument and describes the current state of writeback cache support.

Thanks,
-Sam

RE: [GIT PULL] Block drivers for 3.3-rc

[Sam Bradshaw (sbradshaw)]

> On 2012-01-15 11:33, Christoph Hellwig wrote:
> > On Sun, Jan 15, 2012 at 10:52:12AM +0100, Jens Axboe wrote:
> >> Hi Linus,
> >>
> >> Please see below for the 3.3-rc1 block driver pull request.
> >>
> >> Essentially two different categories here:
> >>
> >> - Adds the driver for the Micron RealSSD P320 pci-express flash card.
> >
> > The REQ_FUA and REQ_FLUSH handling in it is still completely broken,
> > with no attempt to fix it since it last came up.
> >
> > So a pretty clear NAK on that driver from me.
> 
> Good point, I had forgotten about the ordered writes, I believe the rest
> of the items have been handled. The driver/hw, as it stands, does not
> reorder writes and it doesn't have a write back cache. So it should
> actually be safe. Any acked write is stable. Sam, correct me if I'm
> wrong.

That's correct, Jens.  Neither the driver nor the hardware reorders commands so we choose to ignore barriers.  I see now that the "barrier" parameter in the function is rather unfortunately named.  We'll fix that.

-Sam