[GIT PULL] block/IO bits for 2.6.35-rc1

[GIT PULL] block/IO bits for 2.6.35-rc1

[Jens Axboe]

Hi,

These are a bunch of updates for 2.6.35-rc1. The major components:

- Revert of the two writeback fixes that made WB_SYNC_NONE
  writeback from umount actually work. They apparently cause
  problems for Christoph, so revert them until we figure this
  out.

- Updates to the pipe fcntl F_{GET,SET}PIPE_SZ interface to
  reflect the discussion with Michael. The major bits there
  are allowing for just 1 page in the pipe, change of capability
  required to enlarge a pipe beyond the system setting, and
  change of the /proc/sys/fs/pipe-max-pages to be in bytes
  like the fcntl API. The latter also renames the file to
  pipe-max-size to reflect that.

- A drbd update.

- The rest are minor fixes, like discard support for brd to
  ease testing of that and some memory consumption improvements
  for CFQ.

Please pull.

  git://git.kernel.dk/linux-2.6-block.git for-linus

Andrea Gelmini (1):
      drbd: removed duplicated #includes

Dan Carpenter (1):
      cciss: call BUG() earlier

Jens Axboe (9):
      pipe: F_SETPIPE_SZ should return -EPERM for non-root
      pipe: make F_{GET,SET}PIPE_SZ deal with byte sizes
      Revert "writeback: ensure that WB_SYNC_NONE writeback with sb pinned is sync"
      Revert "writeback: fix WB_SYNC_NONE writeback from umount"
      block: disable preemption before using sched_clock()
      Merge branch 'master' into for-linus
      pipe: adjust minimum pipe size to 1 page
      pipe: change the privilege required for growing a pipe beyond system max
      pipe: change /proc/sys/fs/pipe-max-pages to byte sized interface

Konstantin Khlebnikov (2):
      cfq-iosched: remove dead_key from cfq_io_context
      cfq-iosched: compact io_context radix_tree

Lars Ellenberg (5):
      drbd: improve network latency, TCP_QUICKACK
      drbd: need to set socket bufsize early to take effect
      drbd: improve usage of MSG_MORE
      drbd: fix hang on local read errors while disconnected
      drbd: use drbd specific ratelimit instead of global printk_ratelimit

Mike Snitzer (3):
      block: Adjust elv_iosched_show to return "none" for bio-based DM
      block: avoid unconditionally freeing previously allocated request_queue
      block: make blk_init_free_list and elevator_init idempotent

Nick Piggin (2):
      fs/splice.c: fix mapping_gfp_mask usage
      brd: support discard

Philipp Reisner (4):
      drbd: Revert "drbd: Create new current UUID as late as possible"
      drbd: Removed the now empty w_io_error() function
      drbd: Reduce verbosity
      Preparing 8.3.8rc2

Shaohua Li (1):
      cfq-iosched: fix an oops caused by slab leak

 block/blk-core.c                   |   20 +++++--
 block/cfq-iosched.c                |  101 ++++++++++++++++++++++++++++--------
 block/elevator.c                   |    8 ++-
 drivers/block/brd.c                |   53 ++++++++++++++++++-
 drivers/block/cciss_scsi.c         |    2 +-
 drivers/block/drbd/drbd_int.h      |   14 +----
 drivers/block/drbd/drbd_main.c     |   68 +++++++++---------------
 drivers/block/drbd/drbd_receiver.c |   45 ++++++++--------
 drivers/block/drbd/drbd_req.c      |   54 +++++--------------
 drivers/block/drbd/drbd_req.h      |    1 +
 drivers/block/drbd/drbd_worker.c   |   24 +--------
 fs/fs-writeback.c                  |   64 ++++++-----------------
 fs/pipe.c                          |   77 +++++++++++++++++++---------
 fs/splice.c                        |    2 +-
 fs/sync.c                          |    2 +-
 include/linux/backing-dev.h        |    2 +-
 include/linux/blkdev.h             |    9 +++
 include/linux/drbd.h               |    2 +-
 include/linux/iocontext.h          |    1 -
 include/linux/pipe_fs_i.h          |    4 +-
 include/linux/writeback.h          |   10 ----
 kernel/sysctl.c                    |    8 ++--
 mm/page-writeback.c                |    4 +-
 23 files changed, 311 insertions(+), 264 deletions(-)

-- 
Jens Axboe

Re: [GIT PULL] block/IO bits for 2.6.35-rc1

[Linus Torvalds]

On Fri, 4 Jun 2010, Jens Axboe wrote:
>
> - A drbd update.

Can we agree to just stop this? 

During the merge window, please. The drbd fiasco has been going on too 
long for me to believe in the whole "a few more fixes" thing.

If it's a security issue, or a major oops, take it. But don't send me 
"updates" or stuff "just to make it work", since it never has before.

It's not a regression that drbd is in bad shape. So don't try to make it 
out to be one. And if it's not a regression or a major oops/security 
issue, it goes into the NEXT MERGE WINDOW.

Yes, I'm annoyed. Somebody needs to call out the drbd crud. I really don't 
want _any_ more drbd crap after the merge window. If I see any more, I'll 
stop pulling. Comprende?

			Linus

Re: [GIT PULL] block/IO bits for 2.6.35-rc1

[Jens Axboe]

On 2010-06-05 00:37, Linus Torvalds wrote:
> 
> 
> On Fri, 4 Jun 2010, Jens Axboe wrote:
>>
>> - A drbd update.
> 
> Can we agree to just stop this? 
> 
> During the merge window, please. The drbd fiasco has been going on too 
> long for me to believe in the whole "a few more fixes" thing.
> 
> If it's a security issue, or a major oops, take it. But don't send me 
> "updates" or stuff "just to make it work", since it never has before.
> 
> It's not a regression that drbd is in bad shape. So don't try to make it 
> out to be one. And if it's not a regression or a major oops/security 
> issue, it goes into the NEXT MERGE WINDOW.
> 
> Yes, I'm annoyed. Somebody needs to call out the drbd crud. I really don't 
> want _any_ more drbd crap after the merge window. If I see any more, I'll 
> stop pulling. Comprende?

Understood, I should have been a bit more strict on accepting
late drbd chages.

-- 
Jens Axboe