On Sat, Mar 06, 2010 at 05:43:57AM -0500, Christoph Hellwig wrote:
> On Fri, Mar 05, 2010 at 03:40:49PM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> > If the file being defragmented has attributes, then fsr puts a dummy
> > attribute on the temporary file to try to ensure that the inode
> > attribute fork offset is set correctly. This works perfectly well
> > for the old style of attributes that use a fixed fork offset - the
> > presence of any attribute of any size or shape will result in fsr
> > doing the correct thing.
> > However, for attr2 filesystems, the attribute fork offset is
> > dependent on the size and shape of both the data and attribute
> > forks. Hence setting a small attribute on the file does not
> > guarantee that the two inodes have the same fork offset and
> > therefore compatible for a data fork swap.
> > This patch improves the attribute fork handling of fsr. It checks
> > the filesystem version to see if the old style attributes are in
> > use, and if so uses the current method.
> > If attr2 is in use, fsr uses bulkstat output to determine what the
> > fork offset is. If the attribute fork offsets differ then fsr will
> > try to create attributes that will result in the correct offset. If
> > that fails, or the attribute fork is too large, it will give up and just
> > attempt the swap.
> > This fork offset value in bulkstat new functionality in the kernel,
> > so if there are attributes and a zero fork offset, then the kernel
> > does not support this feature and we simply fall back to the existing,
> > less effective code.
> Looks reasonable. It would be good to have a testcase for this in
> xfsqa to verify this works.
Yeah, so far I've tested by running './check -g auto' and finding
the inodes that failed with the standard fsr, then running the fixed
fsr on them. It's not particularly efficient....
FWIW, there's enough information from fsr and xfs_db to be able to
recreate the failure scenario. e.g. from 'xfs_fsr -d -v ...' on a
ino=149 extents=21 can_save=2 tmp=/mnt/test/.fsr/ag0/tmp18175
set temp attr
DEBUG: fsize=2873344 blsz_dio=2873344 d_min=512 d_max=2147483136
Temporary file has 18 extents (21 in original)
XFS_IOC_SWAPEXT failed: ino=149: Invalid argument
$ sudo xfs_db -r -c "inode 149" -c "p" /dev/sdb1
core.magic = 0x494e
core.mode = 0100666
core.version = 2
core.format = 3 (btree)
core.size = 2873344
core.nblocks = 194
core.extsize = 0
core.nextents = 12
core.naextents = 3
core.forkoff = 15
core.aformat = 3 (btree)
has 12 data extents and 3 attribute extents. I should be able to
create this with xfs_io without too much trouble.
Actaully the patch I posted fails on this inode - the previous ones
that were created were btree data fork and extent attribute fork,
so there's more work on this to be done.
More importantly, I discovered a neat option to xfs_fsr yesterday:
$ export FSRXFSTEST=true
$ xfs_fsr -C 18 ....
which allows you to control the number of extents in the temporary
file, so I should be able to recreate the exact conditions that
triggered a failure as well....
I'll see what I can come up with and do some more rigorous testing
of the patch before moving forward with it.