linux-kernel - Re: [PATCH v3 3/4] kernel/smp: add more data to CSD lock debugging

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20220816155841.GD2125313@paulmck-ThinkPad-P17-Gen-1>
Date:   Tue, 16 Aug 2022 08:58:41 -0700
From:   "Paul E. McKenney" <paulmck@...nel.org>
To:     Chang-Ho Cho <changho.cho11@...il.com>
Cc:     jgross@...e.com, linux-kernel@...r.kernel.org, mhocko@...e.com,
        peterz@...radead.org
Subject: Re: [PATCH v3 3/4] kernel/smp: add more data to CSD lock debugging

On Tue, Aug 16, 2022 at 07:53:49PM +0900, Chang-Ho Cho wrote:
> Hello, 
> 
> Google search brought me into this thread.  May I have update on the progress of this issue? Has anyone find root cause, solution or workaround?
> How can I get more about the problem?
> Windows 2008 R2 VM experiencing hang on QEMU-kvm with IPI handling issue. 

Here is a blog entry describing the bug that was causing me trouble:

https://paulmck.livejournal.com/62071.html

But although this appeared to be an IPI-loss problem, it was in fact
an interrupt storm.

Juergen, who you CCed, located a race-condition issue with a non-KVM
hypervisor, if I recall correctly.

Of course, much depends on the exact versions of your various OSes
and firmware.  One useful trick is to look for fixes since whichever
version of Linux you are using, and to backport those.  Also, for issues
with Windows, you are likely better served asking elsewhere.  Since you
are asking here, I am assuming that you have somehow proven that the
interrupt is being lost in KVM or QEMU rather than in the Windows guest.

Please note that observing the hang on QEMU/KVM but not on bare metal
is insufficient, given that the difference in timing can dramatically
change the probability of bugs occuring.

							Thanx, Paul