Re: xfstest failure with xfs, dax and v4.4-rc3

[Ross Zwisler <ross.zwisler@linux.intel.com>]

On Thu, Dec 03, 2015 at 05:51:27PM +1100, Dave Chinner wrote:
> As it is, that test does not fail on my DAX testing on RAM disks.
> ISTR it failing up until recently, though. Yeah:
> 
> Last login: Tue Nov 17 08:45:55 2015 from 192.168.1.103
> $ cd ~/src/xfstests-dev; sudo mkfs.xfs -f /dev/ram0 ; sudo xfs_admin -U generate /dev/ram0 ; sudo MOUNT_OPTIONS="-o dax" ./check -g auto
> SECTION       -- xfs
> FSTYP         -- xfs (debug)
> PLATFORM      -- Linux/x86_64 test4 4.3.0-dgc+
> MKFS_OPTIONS  -- -f -bsize=4096 /dev/ram1
> MOUNT_OPTIONS -- -o dax /dev/ram1 /mnt/scratch
> ....
> generic/083 2s ... 1s
> _check_xfs_filesystem: filesystem on /dev/ram1 is inconsistent (c) (see /home/dave/src/xfstests-dev/results//xfs/generic/083.full)
> _check_xfs_filesystem: filesystem on /dev/ram1 is inconsistent (r) (see /home/dave/src/xfstests-dev/results//xfs/generic/083.full)
> ....
> 
> But as I ran earlier today when testing the ENOSPC fix:
> 
> PLATFORM      -- Linux/x86_64 test4 4.4.0-rc2-dgc+
> MKFS_OPTIONS  -- -f -bsize=4096 /dev/ram1
> MOUNT_OPTIONS -- -o dax /dev/ram1 /mnt/scratch
> ....
> generic/083 1s ... 2s
> ....
> 
> So I'm seeing that the current 4.4-rc2 + my local dev tree patches
> appear to have fixed whatever was causing generic/083 to fail
> here...

Yea, interesting, the patch you sent out yesterday:

xfs: Don't use reserved blocks for data blocks with DAX

Makes this issue disappear in my system as well.  Testing with v4.4-rc3
generic/083 fails 50% of the time or so in my setup, but with just that patch
added to v4.4-rc3 I can't make it fail.

> What version of xfsprogs are you using, Ross?

I'm currently using xfsprogs v4.3.0, but the failure occurred when I was using
v4.2.0 as well.