[bug report] locking/qrwlock, arm64: Move rwlock implementation over to qrwlocks causes CPU crashes/stalls when killing java processes

[Tai Tri Nguyen]

Hi Will,

Recently I have observed the CPU crashes/stalls when rebooting after I
ran cassandra benchmark.
The issue happens randomly.

I could reproduce the issue easier by keep starting/killing cassandra
tests. The issue would occur within about 50 iterations.

I could also reproduce the issue by keep running/killing specjvm2008.
However it takes me much longer time to see the CPU crash/stall issue.

I bisected kernel and found that this changelist below is causing the issue.

commit 087133ac90763cd339b6b67f2998f87dcc136c52
Author: Will Deacon <will.deacon at arm.com>
Date:   Thu Oct 12 13:20:50 2017 +0100

    locking/qrwlock, arm64: Move rwlock implementation over to qrwlocks

    Now that the qrwlock can make use of WFE, remove our homebrewed rwlock
    code in favour of the generic queued implementation.

    Tested-by: Waiman Long <longman at redhat.com>
    Tested-by: Jeremy Linton <jeremy.linton at arm.com>
    Tested-by: Adam Wallis <awallis at codeaurora.org>
    Tested-by: Jan Glauber <jglauber at cavium.com>
    Signed-off-by: Will Deacon <will.deacon at arm.com>
    Acked-by: Peter Zijlstra <peterz at infradead.org>
    Cc: Jeremy.Linton at arm.com
    Cc: Linus Torvalds <torvalds at linux-foundation.org>
    Cc: Thomas Gleixner <tglx at linutronix.de>
    Cc: boqun.feng at gmail.com
    Cc: linux-arm-kernel at lists.infradead.org
    Cc: paulmck at linux.vnet.ibm.com
    Link: http://lkml.kernel.org/r/1507810851-306-5-git-send-email-will.deacon@arm.com
    Signed-off-by: Ingo Molnar <mingo at kernel.org>

Do you have any suggestion or thought how to debug it?

Thanks
-- 
Tai