MySQL-5.6/5.7 innodb assertion !lock->recursive assert in sync0rw.cc during high load on PPC64

[bugproxy]

---Problem Description---

per upstream bug https://bugs.mysql.com/bug.php?id=80436 MySQL-5.7 can assert on ppc64/ppc64le (and probably ARM64) because load order between "recursive" and "writer_thread" wasn't properly enforced.

Upstream bug report has a patch (originally from MariaDB) that corrects the problem.

---uname output---
Linux p87 4.4.0-62-generic #83-Ubuntu SMP Wed Jan 18 14:09:19 UTC 2017 ppc64le ppc64le ppc64le GNU/Linux

Machine Type = Any P8

---Steps to Reproduce---
Install MySQL-5.6 or 5.7

Perform lots of concurrent innodb operations on Power hardware.

Contact Information = Daniel Black (<email address hidden>), Daniel Axtens (Canonical)

[Daniel Black (daniel-black)]

Previous "Clarification from upstream" is a different bug.

[David Britton (dpb)]

Hello --

1) Where are you observing this failure? 16.04?

2) It would be a bit premature to commit anything into Ubuntu without acknowledgement from upstream on this patch (against MariaDB) and if it is an acceptable change in Mysql. After that, we would want it committed to the dev release, then an SRU could be requested against Xenial (16.04).

[Daniel Black (daniel-black)]

apologies for the delay. still trying to get a test case. I'll continue looking next week.

[Manoj Iyer (manjo)]

Daniel, any updates on this bug? I believe this is blocking bug #1706291

[Daniel Black (daniel-black)]

I haven't managed to reproduce however I see no reason why the bug would of gone away from the code changes made since the mysql versions Daniel Axtens used when https://bugs.mysql.com/bug.php?id=80436 was written. I'm just going to keep trying.

I see [~racb] was trying to bundle this patch with bug #1706291 however from a logical point of view this is an independent bug.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2018-02-08 09:55 EDT-------
Closing as unreproducible. Please reopen if problem resurfaces.