Just a quick update - the performance regression fix has been merged into 2.6.23.7 and 2.6.22.14. For everyone's convenience, I have spun a new series of patches which can be applied atop either of those kernels without conflicts:
http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/backports/2.6.23.7/ http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/backports/2.6.22.14/ Additionally, the 2.6.23.7 series gets some new fixes. Full shortlog below: Mark Fasheh (4): ocfs2: Reset journal parameters after s_mount_opt update ocfs2: Filter -ENOSPC in mlog_errno() ocfs2: log valid inode # on bad inode ocfs2: Remove bug statement in ocfs2_dentry_iput() Roel Kluin (1): Fix priority mistakes in fs/ocfs2/{alloc.c, dlmglue.c} Srinivas Eeda (1): ocfs2: fix rename vs unlink race Thanks, --Mark On Mon, Nov 12, 2007 at 11:36:04AM -0800, Mark Fasheh wrote: > Anyone running Ocfs2 on the 2.6.23 kernel should consider downloading and > applying the patches under the '2.6.23' directory at the following URL: > > http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/backports/ > > A direct link to the combined patch is: > > http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/backports/2.6.23/ocfs2_2.6.23_fixes-20071106-1.patch > > > This series includes a patch which fixes a serious performance regression > which was recently introduced into Ocfs2. Anyone experiencing poor write > performance on 2.6.23 is urged to install these patches. > > > The write performance fix has also been backported to the 2.6.22.6 series of > Ocfs2 patches. I anticipate it to be the last patch I will backport to the > 2.6.22.y series. The specific patch against 2.6.22.6 can be found at: > > http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/backports/2.6.22.6/broken-out/0008-ocfs2-fix-write-performance-regression.patch > > > An explanation of what all this is follows. If all you're doing is running > ocfs2 1.2, feel free to ignore this e-mail. If you run a mainline kernel > with Ocfs2 or are considering it, please read on. > > > All Ocfs2 bug fixes are sent upstream at the earliest possible date. Some > times however, they don't make it upstream until after a stable kernel is > released. > > If a bug is severe enough, I will send the fix to the -stable team which > will review it and put it in an update kernel, if it passes their criteria - > which are rather stringent. > > Most fixes fall short of the criteria for -stable, but might still be > useful to folks running Ocfs2 on mainline kernels. This is where my Ocfs2 > patch series comes into play. > > My goal is to provide a reliable series of bugfix patches for those folks > running "bleeding edge" mainline kernels. This way they can get the most > stable mainline Ocfs2 version possible while also getting access to the > latest file system features. > > The way I plan to achieve this is by backporting bug fixes from the current > development kernel into the most recent released kernel. Not all fixes will > be backported, only those which I feel are straightforward and relatively > small. > > Once Linus tags the current development kernel as a stable one, it takes > about a month before my first set of patches is ready. Once those have had > some testing and are up on kernel.org, I'll send an e-mail to this list. > Additional patches will be added as they become available. > > Typically, any Ocfs2 patches for the initial release of a kernel will apply > to all subsequent versions without any problems. So for example, my '2.6.22' > series should apply fine against 2.6.22.1, 2.6.22.2 and so on. If there is > ever a situation where they _won't_ apply, I'll respin the series under a > directory named for the earliest kernel version on which they should be > used. For example, if a fix from the 2.6.22 series makes it into 2.6.22.4, > the 2.6.22 set of patches will cause a reject when applied to 2.6.22.4. In > that case, I'll add another directory, 2.6.22.4 with a patch series that no > longer includes the conflicting fix (as it will have been included already > in the 2.6.22.4 distribution). > > > Shortlog of the current set of patches follows: > > Mark Fasheh (1): > ocfs2: fix write() performance regression > > Roel Kluin (1): > Fix priority mistakes in fs/ocfs2/{alloc.c, dlmglue.c} > > Srinivas Eeda (1): > ocfs2: fix rename vs unlink race > > > Finally, any feedback about this process (positive or negative) would be > greatly appreciated. > > Thanks, > --Mark > > -- > Mark Fasheh > Senior Software Developer, Oracle > [EMAIL PROTECTED] > > _______________________________________________ > Ocfs2-users mailing list > [email protected] > http://oss.oracle.com/mailman/listinfo/ocfs2-users -- Mark Fasheh Senior Software Developer, Oracle [EMAIL PROTECTED] _______________________________________________ Ocfs2-users mailing list [email protected] http://oss.oracle.com/mailman/listinfo/ocfs2-users
