Re: [PATCH] xfs/444: test log replay after XFS_IOC_SWAPEXT

[Brian Foster <bfoster@redhat.com>]

On Fri, Mar 23, 2018 at 05:47:35PM -0500, Eric Sandeen wrote:
> On 2/26/18 7:02 AM, Brian Foster wrote:
> > On Fri, Feb 23, 2018 at 12:33:41PM -0600, Eric Sandeen wrote:
> >> This is a mashup of xfs/042 and some of the log replay tests;
> >> it checks whether the log can be replayed if we crash immediately
> >> after an xfs_fsr / XFS_IOC_SWAPEXT.
> >>
> >> Hint: it can't.  It fails because the temporary donor inode has
> >> been deleted and has invalid mode 0 when we try to replay its
> >> swapext operation.  Kernel patches to fix it will follow soon.
> >>
> >> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
> >> ---
> >>
> >> diff --git a/tests/xfs/444 b/tests/xfs/444
> >> new file mode 100755
> >> index 0000000..e88438a
> >> --- /dev/null
> >> +++ b/tests/xfs/444
> >> @@ -0,0 +1,144 @@
> > ...
> >> +# Test performs several operations to produce a badly fragmented file, then
> >> +# create enough contiguous free space for xfs_fsr to defragment the fragmented
> >> +# file:
> >> +#
> >> +# - create fs with 3 minimum sized (16Mb) allocation groups
> >> +# - create 16x1MB contiguous files which will become large free space extents
> >> +#   when deleted
> >> +# - put a small "space" between each of the 16 contiuguous files to ensure we
> >> +#   have separated free space extents
> >> +# - fill the remaining free space with a "fill file"
> >> +# - mount/unmount/fill remaining free space with a pad file
> >> +# - punch alternate single block holes in the the "fill file" to create
> >> +#   fragmented free space.
> >> +# - use fill2 to generate a very large fragmented file
> >> +# - delete the 16 large contiguous files created initially
> >> +# - run xfs_fsr on the filesystem
> >> +# - check checksums for remaining files
> >> +
> > 
> > Without having dug into the core issue, I wonder whether this sequence
> > could be simplified a bit by using 'xfs_io -c swapext' followed by a
> > shutdown?
> 
> I haven't been able to beat xfs_io into submission here; it requires unlinking
> as well, and even after patching & adding an unlink into the mix:
> 
> xfs_io -x -c "open -f mnt/file1" -c "pwrite 0 4k" -c close -c "open -f mnt/file2" -c "pwrite 0 4k" -c unlink -c "swapext mnt/file1" -c "shutdown -f"
> 
> I can't seem to make this reproduce.  Not sure why.  I'd rather just stick with the proven reproducer.
> 

swapext owner changes required btree format inodes. I don't think a
single 4k write is going to be enough. Indeed, if I do something like
the following:

	xfs_io -fc "falloc 0 100m" /mnt/file1
	xfs_io -fc "falloc 0 100m" /mnt/file2
	xfstests-dev/src/punch-alternating /mnt/file1
	xfstests-dev/src/punch-alternating /mnt/file2
	xfs_io -c "swapext /mnt/file1" /mnt/file2
	rm -f /mnt/file1
	xfs_io -xc "shutdown -f" /mnt/

I see this on a subsequent mount:

	# umount  /mnt ; mount /dev/test/scratch /mnt 
	mount: /mnt: mount(2) system call failed: Structure needs cleaning.

(which I haven't actually confirmed is the original problem).

Brian

> -Eric
> --
> To unsubscribe from this list: send the line "unsubscribe fstests" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html