This bug has been copied from bug #542954 and has been proposed
to be backported to 5.5 z-stream (EUS).
Comment from Kevin, QE please take note and make sure the suggestions made by
Kevin are well covered.
Anything that uses lots of synchronous reads/writes (i.e. metadata
operations). Long snapshots chains where a lot of COW happens seems to
be a good candidate.
It's probably enough to test intensively with one backing file format,
preferably qcow2 which may issue synchronous metadata I/O again and
therefore makes the scenario more complex.
For verification of the fix, you need to use virtio-blk (multiple
requests running at once are required to even trigger this bug). On the
other hand, only IDE can directly call synchronous bdrv_read/write which
is touched by this patch, so in order to avoid regressions some tests on
IDE should be run, too.
Tested with the steps in bug #542954, and that issue does not happen on kvm-83-164.el5_5.12. Besides, we will do more test for this bug, and will upload the test result after finished.
Test runners for verifying this bug:
https://tcms.engineering.redhat.com/run/10369/ (block device)
kvm version: kvm-83-164.el5_5.12
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.