mbox series

[0/3] Btrfs: make ranged fsyncs always respect the given range

Message ID 20200306115329.16947-1-fdmanana@kernel.org (mailing list archive)
Headers show
Series Btrfs: make ranged fsyncs always respect the given range | expand

Message

Filipe Manana March 6, 2020, 11:53 a.m. UTC 
From: Filipe Manana <fdmanana@suse.com>

Btrfs doesn't respect the given range for a fsync when the inode's has the
"need full sync" bit set - it treats the fsync as a full ranged one, operating
on the whole file, doing more IO and cpu work then needed.

That behaviour was needed to fix a corruption bug. Commit 0c713cbab6200b
("Btrfs: fix race between ranged fsync and writeback of adjacent ranges")
fixed that bug by turning the ranged fsync into a full one.

Later the hole detection code of fsync was simplified a lot in order to
fix another bug when using the NO_HOLES feature - done by commit
0e56315ca147b3 ("Btrfs: fix missing hole after hole punching and fsync when
using NO_HOLES").

That commit now makes it easy to avoid turning the ranged fsyncs into
non-ranged fsyncs. This patchset does that. The first two patches are
just cleanups to make the fix easier and less verbose.

A special note is that test generic/456 now fails as fsck reports a missing
file extent item due to an implicit hole (when not using the NO_HOLES feature).
This wasn't supposed to happen anymore after Josef's recent patches that
changes how we update the on disk i_size of an inode. That patchset was
supposed to fix all case of missing file extent items for holes, but there's
more cases which I uncovered while working on this. I'm moving the discussion
of those particular cases under the e-mail thread of Josef's patch that has
the subject:

 "btrfs: use the file extent tree infrastructure"

That has the following address on lore:

https://lore.kernel.org/linux-btrfs/CAL3q7H69_tEsV2WGu9Y6ZgB_1gy9WKPB5iR9XiWaUGiU6C871A@mail.gmail.com/T/#u

Thanks.

Filipe Manana (3):
  Btrfs: add helper to get the end offset of a file extent item
  Btrfs: factor out inode items copy loop from btrfs_log_inode()
  Btrfs: make ranged full fsyncs more efficient

 fs/btrfs/ctree.h     |   1 +
 fs/btrfs/file-item.c |  40 ++++--
 fs/btrfs/file.c      |  13 --
 fs/btrfs/inode.c     |  10 +-
 fs/btrfs/send.c      |  44 +-----
 fs/btrfs/tree-log.c  | 379 +++++++++++++++++++++++++++++----------------------
 6 files changed, 251 insertions(+), 236 deletions(-)