Skip to contentSkip to conversation

writeback changes for 3.2-rc1

GIT PULL Wu Fengguang 2 participants 2 messages Nov 2, 2011 – Nov 4, 2011 VFS

lore
Share & tools
lore
#0 lore raw

Hi Linus,

The following changes since commit 9b13776977d45505469edc6decc93e9e3799afe2:

Merge branch 'for-linus' of git://git.infradead.org/users/sameo/mfd-2.6 (2011-10-02 19:23:44 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux.git writeback-for-linus

There are two patch series:

1) IO-less dirty throttling

By not submitting writeback IO from the dirtier tasks' context, IO seeks _and_ CPU contentions could be greatly reduced.

Users may also notice much smoothed pause times in balance_dirty_pages() and hence in the write() syscall, especially in cases NFS, JBOD and 2+ concurrent dd's. This will in general reduce stalls in the "pipeline" type workloads that have the write() syscall inside its loop, such as

loop {
read() from network
write() to disk
}

The change will inevitably trigger regressions in some FS and/or unknown workloads. However I've managed to address the known issues and built the tracing/visualization scripts for debugging new ones in the easy way :-)

2) report writeback reason in tracing events

I apologize for rebasing the writeback reason patches, which would otherwise cost a merge commit. Either way is not good and I'll take care next time to make the multiple topic branches either fast-forwardable or base one series on top of the other.

Curt Wohlgemuth (2):
writeback: send work item to queue_io, move_expired_inodes
writeback: Add a 'reason' to wb_writeback_work

Wu Fengguang (14):
writeback: account per-bdi accumulated dirtied pages
writeback: dirty position control
writeback: add bg_threshold parameter to __bdi_update_bandwidth()
writeback: dirty rate control
writeback: stabilize bdi->dirty_ratelimit
writeback: per task dirty rate limit
writeback: IO-less balance_dirty_pages()
writeback: limit max dirty pause time
writeback: control dirty pause time
writeback: dirty position control - bdi reserve area
writeback: per-bdi background threshold
writeback: fix ppc compile warnings on do_div(long long, unsigned long)
writeback: trace event bdi_dirty_ratelimit
writeback: trace event balance_dirty_pages

 fs/btrfs/extent-tree.c           |    3 +-
 fs/buffer.c                      |    2 +-
 fs/ext4/inode.c                  |    2 +-
 fs/fs-writeback.c                |   84 +++--
 fs/quota/quota.c                 |    2 +-
 fs/sync.c                        |    4 +-
 fs/ubifs/budget.c                |    2 +-
 include/linux/backing-dev.h      |   14 +-
 include/linux/sched.h            |    7 +
 include/linux/writeback.h        |   33 ++-
 include/trace/events/writeback.h |  161 +++++++--
 kernel/fork.c                    |    3 +
 mm/backing-dev.c                 |    7 +-
 mm/page-writeback.c              |  704 ++++++++++++++++++++++++++++++--------
 mm/vmscan.c                      |    3 +-
 15 files changed, 806 insertions(+), 225 deletions(-)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Commit 0e175a1835ffc979e55787774e58ec79e41957d7
from git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux.git
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.18 (GNU/Linux)

iQIcBAEBAgAGBQJOr4+qAAoJECvKgwp+S8JafoYP/39roT/zsKkrm1wi0wb22x/P 8EcRZAzZLCk/sxu0Hdzqvs2rcf5kqHrWOXX4r+9KI5IvrI1NiTdhvNdn68ui5RT8 iRXC3F+3pIBDIjToEfYlgDuzjucD86mHowIIKIylc7KzC6hJS9EUy39+xFET5moY gfvKsSVWtkVb61sci9rvKl/n416PLnKTBzdMZSiQEKf6mh3/kXrSxv/mUvdzeo2n pHA+P3h7pSvXRDsrVo+SA+8e0jwuZc+UoHVCgYaVrRVffP8TNvw4WEyRQHBo0fu7 S3dUtXqMNGAyYvtZ2wIPTJ6TOLaDIm+AcsllywozjGK+mFMUTeXH5XbFvYiMnnJj N8YHTTZwlpjUfEdVihgSEFMTYBRt4QfaShHiJ+oFo/98uU4INzJyQZ5hqWsvDmcy cuZrXZoyJY5dG+HHUspS93jFjD5XuArcbDDckKnwkWcejSgHd/vX8fFoTm5C/xyy NQjPSZrRui/za/2CWkmkHu1OA+WWhSqoRWm5LxXA00IWSmFaMaUL8+XHnPGb2vQq aCpH42zm7bVvWD9NqXIfmlNysuhgKTparFFald1MwmE20wCIEvZ5wqm1NcGI9LbI 1S6Mus6s/lBEtrzLh1YRbW2PzA57wlVfRjRz3+v0R1hlrT64r8Jr4fV+3Vbi4b+W WVxc0Qpw9HVuOJ+uFjhs
=W8+G
-----END PGP SIGNATURE-----

#1 lore raw

Linus,

any reason this hasn't been pulled yet? The changes are fairly important for any kind of buffered I/O workloads on medium and large machines.