Re: Crash Consistency xfstests

[Josef Bacik <josef@toxicpanda.com>]

On Tue, Aug 22, 2017 at 04:47:32PM +0300, Amir Goldstein wrote:
> On Mon, Aug 21, 2017 at 9:49 PM, Amir Goldstein <amir73il@gmail.com> wrote:
> > On Mon, Aug 21, 2017 at 6:48 PM, Josef Bacik <josef@toxicpanda.com> wrote:
> ...
> >>
> >>> I noticed that if SCRATCH_DEV is a dm target itself (linear), then
> >>> log-writes target creation fails. Is that by design? Can be fixed?
> >>> If not, the test would have to require_scratch_not_dm_target or so.
> >>>
> >>> Please let me know if have any other tip or pointers for me.
> >>
> >> Huh that's weird, I was using it with dm-snapshot and it worked fine.  Maybe I
> >> was doing something else and it's never worked, but it's definitely not by
> >> design.  I'll look into this when I get some time.  Thanks,
> >>
> >
> > Well, if it's not by design, I'll try to figure out the root cause of the error
> > with my logical volume SCRATCH_DEV setup.
> >
> 
> I could not reproduce the logical volume SCRATCH_DEV issue with upstream
> kernel, so chalking it up to setup error or whatever.
> 
> BTW, is there a reason I am missing why you needed to compose FSX_OPTS
> on the fly for the test using the new helper _test_falloc_support?
> I see that fsx already detects if any of the falloc ops are not supported
> via test_fallocate() and auto-disabled the unsupported ops.
> I started fixing _test_falloc_support() which was a bit broken before I realized
> it's not needed. Am I missing something?

Nope I missed that fsx did that, so that part can be dropped.

> 
> I am still seeing test failures from time to time (checksum errors),
> but they do not always reproduce on same system with same random seed
> and when they reproduce its not always the same checksum that fails,
> so I'm still gathering test results.
> 

Yeah timing is everything, when I saw issues things weren't failing
consistently, so I spent a lot of time trying to figure out if it was log-writes
or the fs.  I have definitely found issues with replay log and dm-snapshot, even
recently I added an fsync() before running the fsck from replay-log because
those changes weren't getting picked up by dm-snapshot which resulted in false
positives.  I'm pretty sure I've gotten all the kinks worked out, but I wouldn't
be surprised if there was some other dark corner I missed.  Thanks,

Josef