On Thu, 4 Sept 2025 at 05:18, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> In any case, I doubt that it's a scheduler bug; something exposed by
> the timings change, but that's about it...
Not great, but not entirely unexpected. It would have been nice to see
a smoking gun pointing to an obvious cause, but we've had these kinds
of timing-dependent test failures before, and umount tends to be
special.
And generic/475 clearly ends up exposing various issues, as shown by
git log --grep=generic/475 fs/
which shows a variety of filesystems having had issues with it.
Linus
