Re: [PATCH 0/9 v3] ext4: Punch hole and DAX fixes

[Dave Chinner <david@fromorbit.com>]

On Fri, Nov 06, 2015 at 05:57:04PM +0000, Boylston, Brian wrote:
> Hi,
> 
> I've written a test tool (included below) that exercises page faults on
> hole-y portions of an mmapped file.  The file is created, sized using
> various methods, mmapped, and then two threads race to write a marker to
> different offsets within each mapped page.  Once the threads have
> finished marking each page, the pages are checked for the presence of
> the markers.
> 
> With vanilla 4.2 and 4.3 kernels, this test easily exposes corruption on
> pmem-backed, DAX-mounted xfs and ext4 file systems.

4.2/4.3 kernels do not have a fix for this race problem in them, so
it's no surprise that they fail the test. The fixes for this problem
in XFS are currently in the for-next branch, and will be merged into
4.4 during the merge window.

And FWIW, with XFS on a ramdisk on a 4.3+for-next branch kernel,
this test passes just fine:

$ ./dax-hole-test -f /mnt/scratch/holetest 1000
holetest r0

INFO: zero-filled test...
INFO: sz = 3e800000, npages = 256000
INFO: vastart = 00007fc5a90bd000
INFO: thread 0 is 7fc5a90bc700
INFO: thread 1 is 7fc5a88bb700
INFO: 0 error(s) detected

INFO: posix_fallocate test...
INFO: sz = 3e800000, npages = 256000
INFO: vastart = 00007fc5a90bd000
INFO: thread 0 is 7fc5a88bb700
INFO: thread 1 is 7fc5a90bc700
INFO: 0 error(s) detected

INFO: fallocate test...
INFO: sz = 3e800000, npages = 256000
INFO: vastart = 00007fc5a90bd000
INFO: thread 0 is 7fc5a90bc700
INFO: thread 1 is 7fc5a88bb700
INFO: 0 error(s) detected

INFO: ftruncate test...
INFO: sz = 3e800000, npages = 256000
INFO: vastart = 00007fc5a90bd000
INFO: thread 0 is 7fc5a88bb700
INFO: thread 1 is 7fc5a90bc700
INFO: 0 error(s) detected

That tends to indicate a bug in the ext4 patchset, not that the
method we've taken to solve the problem is fundamentally wrong.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com