From: Dave Chinner <david@fromorbit.com>
To: xfs@oss.sgi.com
Subject: [PATCH 00/17] xfs: current patch queue for 3.11
Date: Fri, 7 Jun 2013 15:45:12 +1000 [thread overview]
Message-ID: <1370583929-14276-1-git-send-email-david@fromorbit.com> (raw)
Hi folks,
This is my current patch queue for the 3.11 release cycle. There is
an update to the mount option/sysctl documentation that includes
date for removal of deprecated mount options and sysctls in this, so
make sure you read it.
There are a couple of readahead plugging fixes to make sure we merge
adjacent readahead buffers before dispatching them. This makes a big
difference to directory and bulkstat performance when the IO
subsystem is IOPS bound.
There's a specualtive preallocation tweak to avoid speculative
prealloc for small files. This is primarily designed for workloads
that use small files exclusively like certain gluster/swift
workloads that are becoming more common.
Following this are a couple of patches to reduce the number of
buffer cache lookups we need to do when creating and unlinking
inodes. The inode create path now does not need to do IO to allocate
a new inode, which means it is more deterministic and less likely to
get blocked behind other IO operations.
Then there is a series of patches to introduce a new logical inode
create transaction for CRC enabld filesystems. This just logs a
small structure that indicates where the inodes being created exist,
and uses a new concept of "ordered buffers" to ensure that the
correct transactional life cycle is maintained for the inodes that
were allocated. This requires special ordered log vector support so
this information tracks correct through the CIL. This significantly
reduces the amoutn of log traffic needed by inode creation, as well
as removes one of the nasty warts around inode buffer handling in
log recovery.
Next is a patch to move all inode fork local to extent conversion
code into the callers that need the coversion, rather than having it
done implicitly by xfs_bmapi_write(). This clarifies the code paths
that the directory and attribute code are using to do this
conversion by moving it directly into the functions that are doing
the short-form to block conversion.
Finally, there are a series of patches that move directory and
attribute code around to separate kerel only code from shared
userspace/kernel code. This greatly reduces the differences between
the shared files, and will make future userspace code
synchronisation much simpler and easy to do. There's more work to do
here, but this is the low hanging fruit....
Flames, comments, reviews all welcome.
-Dave.
$ git diff --stat --summary -C -M 6cb25a8..
Documentation/filesystems/xfs.txt | 282 +++++++++++------
fs/xfs/Makefile | 4 +
fs/xfs/xfs_attr.c | 388 +----------------------
fs/xfs/xfs_attr.h | 1 +
fs/xfs/xfs_attr_inactive.c | 453 +++++++++++++++++++++++++++
fs/xfs/xfs_attr_leaf.c | 643 +------------------------------------
fs/xfs/xfs_attr_list.c | 655 ++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_bmap.c | 199 +++++-------
fs/xfs/xfs_bmap.h | 1 +
fs/xfs/xfs_buf_item.c | 87 ++++--
fs/xfs/xfs_buf_item.h | 4 +-
fs/xfs/xfs_dir2.c | 34 --
fs/xfs/xfs_dir2.h | 71 +++++
fs/xfs/xfs_dir2_block.c | 120 +------
fs/xfs/xfs_dir2_data.c | 1 +
fs/xfs/xfs_dir2_format.h | 24 --
fs/xfs/xfs_dir2_leaf.c | 389 +----------------------
fs/xfs/xfs_dir2_node.c | 1 +
fs/xfs/xfs_dir2_priv.h | 33 +-
fs/xfs/xfs_dir2_readdir.c | 661 +++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_dir2_sf.c | 103 +-----
fs/xfs/xfs_file.c | 3 +-
fs/xfs/xfs_ialloc.c | 67 +++-
fs/xfs/xfs_ialloc.h | 8 +
fs/xfs/xfs_icreate_item.c | 195 ++++++++++++
fs/xfs/xfs_icreate_item.h | 52 +++
fs/xfs/xfs_inode.c | 68 ++--
fs/xfs/xfs_iomap.c | 13 +
fs/xfs/xfs_itable.c | 3 +
fs/xfs/xfs_log.c | 21 +-
fs/xfs/xfs_log.h | 5 +-
fs/xfs/xfs_log_cil.c | 75 +++--
fs/xfs/xfs_log_recover.c | 116 ++++++-
fs/xfs/xfs_super.c | 8 +
fs/xfs/xfs_trace.h | 4 +
fs/xfs/xfs_trans.c | 118 ++++---
fs/xfs/xfs_trans.h | 5 +-
fs/xfs/xfs_trans_buf.c | 34 +-
38 files changed, 2865 insertions(+), 2084 deletions(-)
create mode 100644 fs/xfs/xfs_attr_inactive.c
create mode 100644 fs/xfs/xfs_attr_list.c
create mode 100644 fs/xfs/xfs_dir2_readdir.c
create mode 100644 fs/xfs/xfs_icreate_item.c
create mode 100644 fs/xfs/xfs_icreate_item.h
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
next reply other threads:[~2013-06-07 5:46 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-06-07 5:45 Dave Chinner [this message]
2013-06-07 5:45 ` [PATCH 01/17] xfs: update mount options documentation Dave Chinner
2013-06-07 5:45 ` [PATCH 02/17] xfs: add pluging for bulkstat readahead Dave Chinner
2013-06-07 5:45 ` [PATCH 03/17] xfs: plug directory buffer readahead Dave Chinner
2013-06-07 5:45 ` [PATCH 04/17] xfs: don't use speculative prealloc for small files Dave Chinner
2013-06-07 5:45 ` [PATCH 05/17] xfs: don't do IO when creating an new inode Dave Chinner
2013-06-07 5:45 ` [PATCH 06/17] xfs: xfs_ifree doesn't need to modify the inode buffer Dave Chinner
2013-06-07 5:45 ` [PATCH 07/17] xfs: Introduce ordered log vector support Dave Chinner
2013-06-07 5:45 ` [PATCH 08/17] xfs: Introduce an ordered buffer item Dave Chinner
2013-06-07 5:45 ` [PATCH 09/17] xfs: Inode create log items Dave Chinner
2013-06-07 5:45 ` [PATCH 10/17] xfs: Inode create transaction reservations Dave Chinner
2013-06-07 5:45 ` [PATCH 11/17] xfs: Inode create item recovery Dave Chinner
2013-06-07 5:45 ` [PATCH 12/17] xfs: Use inode create transaction Dave Chinner
2013-06-07 5:45 ` [PATCH 13/17] xfs: remove local fork format handling from xfs_bmapi_write() Dave Chinner
2013-06-07 5:45 ` [PATCH 14/17] xfs: move getdents code into it's own file Dave Chinner
2013-06-07 5:45 ` [PATCH 15/17] xfs: reshuffle dir2 definitions around for userspace Dave Chinner
2013-06-07 5:45 ` [PATCH 16/17] xfs: split out attribute listing code into separate file Dave Chinner
2013-06-07 5:45 ` [PATCH 17/17] xfs: split out attribute fork truncation " Dave Chinner
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1370583929-14276-1-git-send-email-david@fromorbit.com \
--to=david@fromorbit.com \
--cc=xfs@oss.sgi.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox