IO hang with kernel above 4.4.21, heavy IO, btrfs, xen

From: Cheyenne Wills
Date: Wed Mar 22 2017 - 15:44:18 EST


I have a reproducible system hang that appears to be the result of heavy IO.

I originally posted the problem to the linux-block mailing list back
in January, but haven't heard back.

Not sure if I reported to the correct area.

There is a bugzill.kernel.org issue open that contains the details of
the problem https://bugzilla.kernel.org/show_bug.cgi?id=193331

The problem occurs if there is a heavy IO load (the system in question
is a FTP server that locked up during a heavy load). I am able to
reproduce the problem using a btrfs scrub to the file system (the
filesystem spans multiple devices). I have a "test system" where I
can reproduce the problem at will (it takes only a few minutes to
cause the IO to hang up).

With a 4.0.5 kernel, I'm unable to reproduce the hang (and the system
in question runs just fine). A kernel above 4.4.21 (maybe earlier?)
results in the hang described in the above bugzilla ticket.

I would appreciate any suggestions on performing further further
diagnostics, or if there is additional information needed to dig
deeper into the problem.

I'm not subscribed to the lkml, so a cc to cheyenne.wills@xxxxxxxxx
would be appreciated.

Thanks in advance.

Cheyenne Wills
cheyenne.wills@xxxxxxxxx