[GIT PULL] MD update for 4.15-rc1

[GIT PULL] MD update for 4.15-rc1

[Shaohua Li]

Hi,

Please pull MD patches for 4.15. This update mostly includes bug fixes:
- md-cluster now supports raid10 from Guoqing
- raid5 PPL fixes from Artur
- badblock regression fix from Bo
- suspend hang related fixes from Neil
- raid5 reshape fixes from Neil
- raid1 freeze deadlock fix from Nate
- memleak fixes from Zdenek
- bitmap related fixes from Me and Tao
- other fixes and cleanup

Thanks,
Shaohua

The following changes since commit 9e66317d3c92ddaab330c125dfe9d06eee268aff:

  Linux 4.14-rc3 (2017-10-01 14:54:54 -0700)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/shli/md.git for-next

for you to fetch changes up to 0868b99c214a3d55486c700de7c3f770b7243e7c:

  md: free unused memory after bitmap resize (2017-11-10 11:45:13 -0800)

----------------------------------------------------------------
Artur Paszkiewicz (3):
      raid5-ppl: don't resync after rebuild
      raid5-ppl: check recovery_offset when performing ppl recovery
      md: don't check MD_SB_CHANGE_CLEAN in md_allow_write

Colin Ian King (3):
      md-cluster: make function cluster_check_sync_size static
      md: raid10: remove a couple of redundant variables and initializations
      md: remove redundant variable q

Guoqing Jiang (7):
      md: always set THREAD_WAKEUP and wake up wqueue if thread existed
      md-cluster: fix wrong condition check in raid1_write_request
      md-cluster/raid10: set "do_balance = 0" if area is resyncing
      md-cluster: Suspend writes in RAID10 if within range
      md-cluster: Use a small window for raid10 resync
      raid1: remove obsolete code in raid1_write_request
      md-cluster: update document for raid10

Hou Tao (1):
      md/bitmap: clear BITMAP_WRITE_ERROR bit before writing it to sb

Liu Bo (1):
      badblocks: fix wrong return value in badblocks_set if badblocks are disabled

Matthias Kaehlcke (1):
      md: raid10: remove VLAIS

Mike Snitzer (1):
      md: rename some drivers/md/ files to have an "md-" prefix

Mikulas Patocka (1):
      md: use TASK_IDLE instead of blocking signals

Nate Dailey (1):
      raid1: prevent freeze_array/wait_all_barriers deadlock

NeilBrown (10):
      md: fix deadlock error in recent patch.
      raid5: Set R5_Expanded on parity devices as well as data.
      md: forbid a RAID5 from having both a bitmap and a journal.
      md: always hold reconfig_mutex when calling mddev_suspend()
      md: don't call bitmap_create() while array is quiesced.
      md: move suspend_hi/lo handling into core md code
      md: use mddev_suspend/resume instead of ->quiesce()
      md: allow metadata update while suspending.
      md: remove special meaning of ->quiesce(.., 2)
      md: be cautious about using ->curr_resync_completed for ->recovery_offset

Shaohua Li (2):
      md/bitmap: revert a patch
      md: use lockdep_assert_held

Zdenek Kabelac (2):
      md: release allocated bitset sync_set
      md: free unused memory after bitmap resize

 Documentation/md/md-cluster.txt            |   3 +-
 MAINTAINERS                                |   7 +-
 block/badblocks.c                          |   2 +-
 drivers/md/Kconfig                         |   5 +-
 drivers/md/Makefile                        |   5 +-
 drivers/md/dm-raid.c                       |  12 +-
 drivers/md/{bitmap.c => md-bitmap.c}       |  27 ++++-
 drivers/md/{bitmap.h => md-bitmap.h}       |   0
 drivers/md/md-cluster.c                    |  12 +-
 drivers/md/{faulty.c => md-faulty.c}       |   0
 drivers/md/{linear.c => md-linear.c}       |   2 +-
 drivers/md/{linear.h => md-linear.h}       |   0
 drivers/md/{multipath.c => md-multipath.c} |   4 +-
 drivers/md/{multipath.h => md-multipath.h} |   0
 drivers/md/md.c                            | 147 ++++++++++++++++---------
 drivers/md/md.h                            |  20 ++--
 drivers/md/raid0.c                         |   2 +-
 drivers/md/raid1.c                         |  78 ++++---------
 drivers/md/raid10.c                        | 169 +++++++++++++++++++++++++----
 drivers/md/raid10.h                        |   6 +
 drivers/md/raid5-cache.c                   |  44 +++++---
 drivers/md/raid5-log.h                     |   2 +-
 drivers/md/raid5-ppl.c                     |   6 +-
 drivers/md/raid5.c                         |  79 +++++++-------
 24 files changed, 409 insertions(+), 223 deletions(-)
 rename drivers/md/{bitmap.c => md-bitmap.c} (99%)
 rename drivers/md/{bitmap.h => md-bitmap.h} (100%)
 rename drivers/md/{faulty.c => md-faulty.c} (100%)
 rename drivers/md/{linear.c => md-linear.c} (99%)
 rename drivers/md/{linear.h => md-linear.h} (100%)
 rename drivers/md/{multipath.c => md-multipath.c} (99%)
 rename drivers/md/{multipath.h => md-multipath.h} (100%)

Re: [GIT PULL] MD update for 4.15-rc1

[Linus Torvalds]

So I've pulled this, but have a question:

On Tue, Nov 14, 2017 at 11:56 AM, Shaohua Li <shli@kernel.org> wrote:
>
> Mike Snitzer (1):
>       md: rename some drivers/md/ files to have an "md-" prefix

is this really sensible? I know I absolutely hate what things like
this does to tab-completion, because everything now has the same
prefix.

And it's not like it adds any value - the "md" is already there in the
directory name.

So honestly, if this was code that I regularly looked at (it isn't),
I'd probably not be happy. As it is, I don't really care all that
much, just wanted to check.

Why is it "md-faulty,c", but "raid5.c", for example?

And if it's to try to separate the "dm-xyz" vs "md-xyz" ones, maybe
making this an actual directory structure rather than a prefix would
be better?

Just throwing this out.

              Linus

Re: [GIT PULL] MD update for 4.15-rc1

[Mike Snitzer]

On Tue, Nov 14 2017 at  7:14pm -0500,
Linus Torvalds <torvalds@linux-foundation.org> wrote:

> So I've pulled this, but have a question:
> 
> On Tue, Nov 14, 2017 at 11:56 AM, Shaohua Li <shli@kernel.org> wrote:
> >
> > Mike Snitzer (1):
> >       md: rename some drivers/md/ files to have an "md-" prefix
> 
> is this really sensible? I know I absolutely hate what things like
> this does to tab-completion, because everything now has the same
> prefix.
> 
> And it's not like it adds any value - the "md" is already there in the
> directory name.
> 
> So honestly, if this was code that I regularly looked at (it isn't),
> I'd probably not be happy. As it is, I don't really care all that
> much, just wanted to check.
> 
> Why is it "md-faulty,c", but "raid5.c", for example?
> 
> And if it's to try to separate the "dm-xyz" vs "md-xyz" ones, maybe
> making this an actual directory structure rather than a prefix would
> be better?
> 
> Just throwing this out.

I see you already took the change; but regardless the idea of splitting
MD and DM files into separate directories was explored briefly (by me).
I stopped short of that because dm-raid.c does share code with md (see
all the md includes at the top of dm-raid.c).

Would require more extensive work to formalize the MD interfaces through
include/linux/ which wasn't a priority to tackle.  I was just after
stopping the various janitor DM patch submissions from spamming
linux-raid and shli.  So went with the quickest way to express discrete
MD and DM files within the MAINTAINERS file.

Mike