XFS Status Updates
(July 2010 update)
|Line 1:||Line 1:|
== XFS status update for July 2010 ==
== XFS status update for July 2010 ==
Revision as of 14:59, 2 September 2010
XFS status update for August 2010
At the first of August we finally saw the release of Linux 2.6.35, which includes a large XFS update. The most prominent feature in Linux 2.6.35 is the new delayed logging code which provides massive speedups for metadata-intensive workloads, but there has been a large amount of other fixes and cleanups, leading to the following diffstat:
67 files changed, 4426 insertions(+), 3835 deletions(-)
Given the early release of Linux 2.6.35 the merge window for the next release fully fell into the month of August. The XFS updates for Linux 2.6.36 include various additional performance improvements in the delayed logging code, for direct I/O writes and for avoiding synchronous transactions, as well as various fixed and large amount of cleanups, including the removal of the remaining dead DMAPI code.
On the userspace side we saw the 3.1.3 release of xfsprogs, which includes various smaller fixes, support for the new XFS_IOC_ZERO_RANGE ioctl and Debian packaging updates. The xfstests package saw one new test case and a couple of smaller patches, and xfsdump has not seen any updates at all.
The XMLified versions of the XFS users guide, training labs and filesystem structure documentation are now available as on the fly generated html on the xfs.org website and can be found at: http://www.xfs.org/index.php/XFS_Papers_and_Documentation
XFS status update for July 2010
July saw three more release candidates for the Linux 2.6.35 kernel, which included a relatively large number of XFS updates. There were two security fixes, a small one to prevent swapext to operate on write-only file descriptors, and a much larger one to properly validate inode numbers coming from NFS clients or userspace applications using the bulkstat or the open-by-handle interfaces. In addition to that another relatively large patch fixes the way inodes get reclaimed in the background, and avoids inode caches growing out of bounds.
In the meantime the code for the Linux 2.6.36 got the last touches before the expected opening of the merge window, by merging a few more last minute fixes and cleanups. The most notable one is a patch series that fixes in-memory corruption when concurrently accessing unwritten extents using the in-kernel AIO code.
The userspace side was still quite slow, but some a bit more activity than June. In xfsprogs the xfs_db code grew two bug fixes, as did the xfs_io script. The xfstests package saw one new test cases and various fixes to existing code. Last but not least a few patches affecting the build system for all userspace tools were committed.
XFS status update for June 2010
The month of June saw a few important bug fixes for the Linux 2.6.35 release candidates. That includes ensuring that files used for the swapext ioctl are writable to the user, and doing proper validation of inodes coming from untrusted sources, such as NFS exporting and the open by handle system calls. The main work however has been focused on development for the Linux 2.6.36 merge window, including merging various patches that have been out on the mainline list for a long time. Highlights include further performance improvements for sync heavy metadata workloads, stack space reduction in the writeback path and improvements of the XFS tracing infrastructure. Also after some discussion the remaining hooks for DMAPI are going to be dropped in mainline. As a replacement a tree containing full DMAPI support with a slightly cleaner XFS interaction will be hosted by SGI.
On the userspace side June was a rather slow month, with no updates to xfsprogs and xfsdump at all, and just one new test case and a cleanup applied to xfstests.
XFS status update for May 2010
In May 2010 we saw the long awaited release of Linux 2.6.34, which includes a large XFS update. The most important features appearing in 2.6.34 was the new inode and quota flushing code, which leads to much better I/O patterns for metadata-intensive workloads. Additionally support for synchronous NFS exports has been improved to give much better performance, and performance for the fsync, fdatasync and sync system calls has been improved slightly. A bug when resizing extremely busy filesystems has been fixed, which required extensive modification to the data structure used for looking up the per-allocation group data. Last but not least there was a steady flow of minor bug fixes and cleanups, leading to the following diffstat from 2.6.33 to 2.6.34:
86 files changed, 3209 insertions(+), 3178 deletions(-)
Meanwhile active development aimed at 2.6.35 merge progressed. The major feature for this window is the merge of the delayed logging code, which adds a new logging mode that dramatically reduces the bandwidth required for log I/O. See the documentation for details. Testers for this new code are welcome.
In userland xfsprogs saw the long awaited 3.1.2 release, which can be considered a bug fix release for xfs_repair, xfs_fsr and mkfs.xfs. After the release a few more fixes were merged into the development tree. The xfstests package saw various new tests, including many tests to exercise the quota code, and a few fixes to existing tests.
XFS status update for April 2010
In April 2.6.34 still was in the release candidate phase, with a hand full of XFS fixes making it into mainline. Development for the 2.6.35 merge window went ahead full steam at the same time.
While a fair amount of patches hit the development tree these were largely cleanups, with the real development activity happening on the mailing list. There was another round of patches and following discussion on the scalable busy extent tracking and delayed logging features mentioned last month. They are expected to be merged in May and queue up for the Linux 2.6.35 window. Last but not least April saw a large number of XFS fixes backported to the 2.6.32 and 2.6.33 -stable series.
In user land xfsprogs has seen few but important updates, preparing for a new release next month. The xfs_repair tool saw a fix to correctly enable the lazy superblock counters on an existing filesystem, and xfs_fsr saw updates to better deal with dynamic attribute forks. Last but not a least a port to Debian GNU/kFreeBSD got merged. The xfstests test suite saw two new test cases and various smaller fixes.
XFS status update for March 2010
The merge window for Linux 2.6.34 closed in the first week of March, with the important XFS features already landing in February. Not surprisingly the XFS merge activity in March has been rather slow, with only about a dozen bug fixes patches making it towards Linus' tree in that time.
On the other hand active development for the 2.6.35 merge window has been very active. Most importantly there was a lot of work on the transaction and log subsystems. Starting with a large patchset to clean up and refactor the transaction subsystem and introducing more flexible I/O containers in the low-level logging code work is progressing to a new, more efficient logging implementation. While this preparatory work has already been merged in the development tree, the actual delayed logging implementation still needs more work after the initial public posting. The delayed logging implementation which is very briefly modeled after the journaling mode in the ext3/4 and reiserfs filesystems allows to accumulated multiple asynchronous transactions in memory instead of possibly writing them out many times. Using the new delayed logging mechanism I/O bandwidth used for the log decreases by orders of magnitude and performance on metadata intensive workloads increases massively.
In addition to that a new version of the discard (aka TRIM) support has been posted, this time entirely contained in kernel space and without the need of a userspace utility to drive it. Last but not least the usual steady stream of cleanups and bug fixes has not ceased this month either.
Besides the usual flow of fixes and new test cases in the xfstests test suite development on the userspace side has been rather slow. Xfsprogs has only seen a single fix for SMP locking in xfs_repair and support for building on Debian GNU/kFreeBSD, and xfsdump has seen no commit at all.
XFS status update for February 2010
February saw the release of the Linux 2.6.33 kernel, which includes a large XFS update. The biggest user-visible change in Linux 2.6.33 is that XFS now support the generic Linux trace event infrastructure, which allows tracing lots of XFS behavior with a normal production built kernel. Except for this Linux 2.6.33 has been mostly a bug-fix release, fixing various user reported bugs in previous releases. The total diffstat for XFS in Linux 2.6.33 looks like:
84 files changed, 3023 insertions(+), 3550 deletions(-)
In addition to that the merge window for Linux 2.6.34 opened and the first merge of the XFS tree made it into Linus tree. Unlike Linux 2.6.33 this merge window includes major feature work. The most important change for users is a new algorithm for inode and quota writeback that leads to better I/O locality and improved metadata performance. The second big change is a rewrite of the per-allocation group data lookup which fixes a long-standing problem in the code to grow a life filesystem and will also ease future filesystem shrinking support. Not merged through the XFS tree, but of great importance for embedded users is a new API that allows XFS to properly flush cache lines on it's log and large directory buffers, making XFS work properly on architectures with virtually indexed caches, such as parisc and various arm and mips variants. Last but not least there is an above average amount of cleanups that went into Linus tree in this cycle.
There have been more patches on the mailing list that haven't made it to Linus tree yet, including an optimized implementation of fdatasync(2) and massive speedups for metadata workloads on NFS exported XFS filesystems.
On the userspace side February has been a relatively quite month. Lead by xfstests only a moderate amount of fixes made it into the respective trees.
XFS status update for January 2010
January saw additional release candidates of the Linux 2.6.33 kernel, including a couple of bug fixes for XFS. In the meantime the XFS tree has been growing a large number of patches destined for the Linux 2.6.34 merge window: a large rework of the handling of per-AG data, support for the quota netlink interface, and better power saving behavior of the XFS kernel threads, and of course various cleanups.
A large patch series to replace the current asynchronous inode writeback with a new scheme that uses the delayed write buffers was posted to the list. The new scheme, which allows archive better I/O locality by dispatching meta-data I/O from a single place has been discussed extensively and is expected to be merged in February.
On the userspace side January saw the 3.1.0 and 3.1.1 releases of xfsprogs, as well as the 3.0.4 release of xfsdump. The biggest changes in xfsprogs 3.1.0 were optimizations in xfs_repair that lead to a much lower memory usage, and optional use of the blkid library for filesystem detection and retrieving storage topology information. The 3.1.1 release contained various important bug fixes for these changes and a various improvements to the build system. The major feature of xfsdump 3.0.4 were fixes for time stamp handling on 64-bit systems.
The xfstests package also lots of activity including various new testcases and an improved build system.
XFS status update for December 2009
December finally saw the long awaited release of Linux 2.6.32, which for XFS is mostly a bug fix release, with the major changes being various improvement to the sync path, including working around the expectation from the grub boot loader where metadata is supposed to be after a sync() system call. Together with a refactoring of the inode allocator this gives a nice diffstat for this kernel release:
46 files changed, 767 insertions(+), 1048 deletions(-)
In the meantime development for the 2.6.33 has been going strong. The new event tracing code that allows to observe the inner workings of XFS in production systems has finally been merged, with another patch to reduce the size of the tracing code by using new upstream kernel features posted for review. Also a large patch series has been posted which changes per-AG data to be looked up by a radix tree instead of the existing array. This works around possible deadlocks and user after free issues during growfs, and prepares for removing a global (shared) lock from the free space allocators. In addition to that a wide range of fixes has been posted and applied.
Work on the userspace packages has been just as busy. In mkfs.xfs the lazy superblock counter feature has now been enabled by default for the upcoming xfsprogs 3.1.0 release, which will require kernel 2.6.22 for the default mkfs invocation. Also for mkfs.xfs as patch was posted to correct the automatic detection of 4 kilobyte sector drivers which are expected to show up in large quantities the real work soon. The norepair mode in xfs_repair has been enhanced with additional freespace btree correction checks from xfs_db and is now identical to xfs_check in filesystem consistency checking coverage. A temporary file permission problems has been fixed in xfs_fsr, and the libhandle library has been fixed to better deal with symbolic links. In xfs_io a few commands that were added years ago have finally been wired up to actually be usable. And last but not least xfsdump saw a fix to the time stamp handling in the backup format and some usability and documentation improvements to xfsinvutil.
XFS status update for November 2009
November was a relatively slow month for XFS development. The XFS tree that is destined for the Linux 2.6.33 merge window saw a few fixes and cleanups applied to it, and few important fixes still made it into the last Linux 2.6.32 release candidates. A few more patches including a final version of the event tracing support for XFS were posted but not reviewed yet.
On the userspace side there has been a fair amount of xfsprogs activity. The repair speedup patches have finally been merged into the main development branch and a couple of other fixes to the various utilities made it in, too. The xfstests test suite saw another new regression test suite and a build system fix up.
XFS status update for October 2009
In October we saw the Linux 2.6.32 merge window with a major XFS update. This update includes a refactoring of the inode allocator which also allows for speedups for very large filesystems, major sync fixes, updates to the fsync and O_SYNC handling which merge the two code paths into a single and more efficient one, a workaround for the VFS time stamp behavior, and of course various smaller fixes. A couple of additional fixes have been queued up for the next merge window.
On the userspace side there has been a healthy activity on xfsprogs: mkfs can now discard unused sectors on SSDs and thinly provisioned storage devices and use the more generic libblkid for topology information and filesystems detection instead of the older libdisk, and the build system gained some updates to make the source package generation simpler and shared for different package types. A patch has been out to the list but yet committed to add symbol versioning to the libhandle library to make future ABI additions easier. The xfstests package only saw some minor activity with a new test case and small build system fixes.
New minor releases of xfsprogs and xfsdump were tagged but not formally released after additional discussion. Instead a new major xfsprogs release is planned for next month.
XFS status update for September 2009
In September the Linux 2.6.31 kernel was finally released, including another last minute XFS fix for the swapext (defragmentation) compat ioctl handler. The final patch from 2.6.30 to 2.6.31 shows the following impressive diffstat for XFS:
55 files changed, 1476 insertions(+), 2269 deletions(-)
The 2.6.32 merge window started with a large XFS merge that included changes to the inode allocator, and a few smaller fixes. New versions of the sync and time stamp fixes as well as the event tracing support have been posted in September but not yet merged into the XFS development tree and/or mainline.
On the userspace side a large patch series to reduce the memory usage in xfs_repair to acceptable levels was posted, but not yet merged. A new xfs_df shell script to measure use of the on disk space was posted but not yet merged pending some minor review comments and a missing man page. In addition we saw the usual amount of smaller fixes and cleanups.
Also this month Felix Blyakher resigned from his post as XFS maintainer and handed off to Alex Elder.
XFS status update for August 2009
In August the Linux 2.6.31 kernel has still been in the release candidate stage, but a couple of important XFS fixes made it in time for the release, including a fix for the inode cache races with NFS workloads that have plagued us for a long time.
The list saw various patches destined for the Linux 2.6.32 merge window, including a merge of the fsync and O_SYNC handling code to address various issues with the latter, a workaround for deficits in the timestamp handling interface between the VFS and filesystems, a repost of the sync improvements patch series and various smaller patches.
August also saw the minor 3.0.3 release of xfsprogs which collects smaller fixes to the various tools and most importantly a fix to allow xfsprogs to work again on SPARC and other strict alignment handling which regressed a few releases ago. The xfstests repository saw a few new test cases and a various small improvements.
XFS status update for July 2009
As a traditional summer vacation month July has not seen a lot of XFS activity. The mainline 2.6.31 kernel made it to the 5th release candidate but besides a few kernel-wide patches touching XFS the only activity were two small patches fixing a bug in FIEMAP and working around writeback performance problems in the VM.
A few more patches were posted to the list but haven't been merged yet. Two big patch series deal with theoretically possible deadlocks due to locks taken in reclaim contexts, which are now detected by lockdep.
The pace on the userspace side has been slow. There have been a couple of fixes to xfs_repair and xfs_db, and xfstests grew a few more testcases.
XFS status update for June 2009
On June 9th we finally saw the release of Linux 2.6.30. For XFS this release mostly contains the improved ENOSPC handling, but also various smaller bugfixes and lots of cleanups. The code size of XFS decreased again by 500 lines of code in this release.
The Linux 2.6.31 merge opened in the mid of the month and some big XFS changes have been pushed: A removal of the quotaops infrastructure which simplifies the quota implementation, the switch from XFS's own Posix ACL implementation to the generic one shared by various other filesystems which also supports in-memory caching of ACLs and another incremental refactoring of the sync code.
A patch to better track dirty inodes and work around issues in the way the VFS updates the access time stamp on inodes has been reposted and discussed. Another patch to converting the existing XFS tracing infrastructure to use the ftrace even tracer has been posted.
On the userspace side there have been a few updates to xfsprogs, including some repair fixes and a new fallocate command for xfs_io. There were major updates for xfstests: The existing aio-dio-regress testsuite has been merged into xfstests, and various changes went into the tree to make xfstests better suitable for use with other filesystems.
The attr and acl projects which have been traditionally been hosted as part of the XFS userspace utilities have now been split into a separate project maintained by Andreas Gruenbacher, who has been doing most of the work on it, and moved to the Savannah hosting platform.
XFS status update for May 2009
In May Linux 2.6.30 was getting close to be released, and various important XFS fixes made it during the latest release candidates. In the meantime some big patch series to rework the sync code and the inode allocator have been posted for the next merge window.
On the userspace side xfsprogs and xfsdump 3.0.1 were finally released, quickly followed by 3.0.2 releases with updated Debian packaging. After that various small patches that were held back made it into xfsprogs. A patch to add the xfs_reno tool which allows to move inodes around to fit into 32 bit inode number space has been posted which is also one central aspect of future online shrinking support.
There has been major activity on xfstests including adding generic filesystems support to allow running tests that aren't XFS-specific on any Linux filesystems.
XFS status update for April 2009
In April development for Linux 2.6.30 was in full motion. A patchset to correct flushing of delayed allocations with near full filesystems has been committed in early April, as well as various smaller fixes. A patch series to improve the behavior of sys_sync has been posted but is waiting for VFS changes queued for Linux 2.6.31.
On the userspace side xfsprogs and xfsdump 3.0.1 have managed to split their release dates into May again after a lot of last-minute build system updates.
XFS status update for March 2009
Linux 2.6.29 has been released which includes major XFS updates like the new generic btree code, a fully functional 32bit compat ioctl implementation and the new combined XFS and Linux inode. (See previous status reports for more details). A patch series to improve correctness and performance has been posted but not yet applied. Various minor fixes and cleanups have been sent to Linus for 2.6.30 which looks like it will be a minor release for XFS after the big churn in 2.6.29.
On userspace a lot of time has been spent on fixing and improving the build system shared by the various XFS utilities as well as various smaller improvements leading to the xfsprogs and xfsdump 3.0.1 releases which are still outstanding.
XFS status update for February 2009
In February various smaller fixes have been sent to Linus for 2.6.29, including a revert of the faster vmap APIs which don't seem to be quite ready yet on the VM side. At the same time various patches have been queued up for 2.6.30, with another big batch pending. There also has been a repost of the CRC patch series, including support for a new, larger inode core.
SGI released various bits of work in progress from former employees that will be extremely helpful for the future development of XFS, thanks a lot to Mark Goodwin for making this happen.
On the userspace side the long awaited 3.0.0 releases of xfsprogs and xfsdump finally happened early in the month, accompanied by a 2.2.9 release of the dmapi userspace. There have been some issues with packaging so a new minor release might follow soon.
The xfs_irecover tool has been relicensed so that it can be merged into the GPLv2 codebase of xfsprogs, but the actual integration work hasn't happened yet.
Important bits of XFS documentation that have been available on the XFS website in PDF form have been released in the document source form under the Creative Commons license so that they can be updated as a community effort, and checked into a public git tree.
XFS status update for January 2009
January has been an extremely busy month on the userspace front. Many smaller and medium updates went into xfsprogs, xfstests and to a lesser extent xfsdump. xfsprogs and xfsdump are ramping up for getting a 3.0.0 release out in early February which will include the first major re-sync with the kernel code in libxfs, a cleanup of the exported library interfaces and the move of two tools (xfs_fsr and xfs_estimate) from the xfsdump package to xfsprogs. After this the xfsprogs package will contain all tools that use internal libxfs interfaces which fortunately equates to those needed for normal administration. The xfsdump package now only contains the xfsdump/xfsrestore tools needed for backing up and restoring XFS filesystems. In addition it grew a fix to support dump/restore on systems with a 64k page size. A large number of acl/attr package patches was posted to the list, but pending a possible split of these packages from the XFS project these weren't processed yet.
On the kernel side the big excitement in January was an in-memory corruption introduced in the btree refactoring which hit people running 32bit platforms without support for large block devices. This issue was fixed and pushed to the 2.6.29 development tree after a long collaborative debugging effort at linux.conf.au. Besides that about a dozen minor fixes were pushed to 2.6.29 and the first batch of misc patches for the 2.6.30 release cycle was sent out.
At the end of December the SGI group in Melbourne which the previous XFS maintainer and some other developers worked for has been closed down and they will be missed greatly. As a result maintainership has been passed on in a way that has been slightly controversial in the community, and the first patchset of work in progress in Melbourne have been posted to the list to be picked up by others.
The xfs.org wiki has gotten a little facelift on it's front page making it a lot easier to read.
XFS status update for December 2008
On Christmas Eve the 2.6.28 mainline kernel was release, with only minor XFS bug fixes over 2.6.27.
On the development side December has been busy but unspectacular month. All lot of misc fixes and improvements have been sent out, tested and committed especially on the user land side.
XFS status update for November 2008
The mainline kernel is now at 2.6.28-rc6 and includes a small number of XFS fixes. There have been no updates to the XFS development tree during November. Without new regressions that large number of changes that missed 2.6.28 has thus stabilized to be ready for 2.6.29. In the meantime kernel-side development has been slow, with the only major patch set being a wide number of fixes to the compatibility for 32 bit ioctls on a 64 bit kernel.
In the meantime there has been a large number of commits to the user space tree, which mostly consist of smaller fixes. xfsprogs is getting close to have the 3.0.0 release which will be the first full resync with the kernel sources since the year 2005.
XFS status update for October 2008
Linux 2.6.27 released with all the bits covered in last month's report. It did however miss two important fixes for regressions that a few people hit. 188.8.131.52 or later are recommended for use with XFS.
In the meantime the generic btree implementation, the sync reorganization and after a lot of merge pain the XFS and VFS inode unification hit the development tree during the time allocated for the merge window. No XFS updates other than the two regression fixes also in 184.108.40.206 have made it into mainline as of 2.6.28-rc3.
The only new feature on the list in October is support for the fiemap interface that has been added to the VFS during the 2.6.28 merge window. However there was lot of patch traffic consisting of fixes and respun versions of previously known patches. There still is a large backlog of patches on the list that is not applied to the development tree yet.
XFS status update for September 2008
With Linux 2.6.27 still not released but only making slow progress from 2.6.27-rc5 to 2.6.27-rc8 XFS changes in mainline have been minimal in September with only about half a dozen bug fixes patches.
In the meantime the generic btree patch set has been committed to the development tree, but not many other updates yet. On the user space side xfsprogs 2.10.1 has been released on September 5th with a number of important bug fixes. Following the release of xfsprogs 2.10.1 open season for development of the user space code has started. The first full update of the shared kernel / user space code in libxfs since 2005 has been committed. In addition to that the number of headers installed for the regular devel package has been reduced to the required minimum and support for checking the source code for endianess errors using sparse has been added.
The patch sets to unify the XFS and Linux inode structures, and rewrite various bits of the sync code have seen various iterations on the XFS list, but haven't been committed yet. A first set of patches implementing CRCs for various metadata structures has been posted to the list.
XFS status update for August 2008
With the 2.6.27-rc5 release the 2.6.27 cycle is nearing it's end. The major XFS feature in 2.6.27-rc5 is support for case-insensitive file names. At this point it is still limited to 7bit ASCII file names, with updates for utf8 file names expected to follow later. In addition to that 2.6.27-rc5 fixes a long-standing problem with non-EABI arm compiler which pack some XFS data structures wrongly. Besides this 2.6.27-rc5 also contains various cleanups, most notably the removal of the last bhv_vnode_t instances, and most uses of semaphores. As usual the diffstat for XFS from 2.6.26 to 2.6.26-rc5 is negative:
100 files changed, 3819 insertions(+), 4409 deletions(-)
On the user space front a new minor xfsprogs version is about to be released containing various fixes including the user space part of arm packing fix.
Work in progress on the XFS mailing list are a large patch set to unify the alloc, inobt and bmap btree implementation into a single that supports arbitrarily pluggable key and record formats. These btree changes are the first major preparation for adding CRC checks to all metadata structures in XFS, and an even larger patch set to unify the XFS and Linux inode structures, and perform all inode write back from the btree uses instead of an inode cache in XFS.