[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200805153214.GA6090@magnolia>
Date: Wed, 5 Aug 2020 08:32:14 -0700
From: "Darrick J. Wong" <djwong@...nel.org>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: "Darrick J. Wong" <djwong@...nel.org>,
linux-fsdevel@...r.kernel.org, linux-xfs@...r.kernel.org,
david@...morbit.com, linux-kernel@...r.kernel.org,
sandeen@...deen.net, hch@....de,
linux-ext4 <linux-ext4@...r.kernel.org>,
Theodore Ts'o <tytso@....edu>, riteshh@...ux.ibm.com,
rgoldwyn@...e.de, agruenba@...hat.com, linux-btrfs@...r.kernel.org
Subject: [GIT PULL] iomap: new code for 5.9-rc1
Hi Linus,
Please pull these new changes to the iomap code for 5.9. The most
notable changes are:
1) iomap no longer invalidate the page cache when performing a direct
read, since doing so is unnecessary and the old directio code doesn't do
that either.
2) iomap embraced the use of returning ENOTBLK from a direct write to
trigger falling back to a buffered write since ext4 already did this and
btrfs wants it for their port.
3) iomap falls back to buffered writes if we're doing a direct write and
the page cache invalidation after the flush fails; this was necessary to
handle a corner case in the btrfs port.
The branch merges cleanly with your HEAD branch as of 15m ago. Please
let me know if there are any strange problems.
--D
The following changes since commit dcb7fd82c75ee2d6e6f9d8cc71c52519ed52e258:
Linux 5.8-rc4 (2020-07-05 16:20:22 -0700)
are available in the Git repository at:
git://git.kernel.org/pub/scm/fs/xfs/xfs-linux.git tags/iomap-5.9-merge-4
for you to fetch changes up to 66143873c6a0b24242f24f65fde70c643b26b4e7:
iomap: fall back to buffered writes for invalidation failures (2020-07-23 22:45:59 -0700)
----------------------------------------------------------------
New code for 5.9:
- Make sure we call ->iomap_end with a failure code if ->iomap_begin
failed in any way; some filesystems need to try to undo things.
- Don't invalidate the page cache during direct reads since we already
sync'd the cache with disk.
- Make direct writes fall back to the page cache if the pre-write
cache invalidation fails. This avoids a cache coherency problem.
----------------------------------------------------------------
Andreas Gruenbacher (1):
iomap: Make sure iomap_end is called after iomap_begin
Christoph Hellwig (2):
xfs: use ENOTBLK for direct I/O to buffered I/O fallback
iomap: fall back to buffered writes for invalidation failures
Dave Chinner (1):
iomap: Only invalidate page cache pages on direct IO writes
fs/ext4/file.c | 2 ++
fs/gfs2/file.c | 3 ++-
fs/iomap/apply.c | 13 +++++++++----
fs/iomap/direct-io.c | 37 +++++++++++++++++++++----------------
fs/iomap/trace.h | 1 +
fs/xfs/xfs_file.c | 8 ++++----
fs/zonefs/super.c | 7 +++++--
7 files changed, 44 insertions(+), 27 deletions(-)
Powered by blists - more mailing lists