Re: [RFC PATCH] nmi,printk: fix ABBA deadlock between nmi_backtrace and dump_stack_lvl
From: Petr Mladek
Date: Mon Sep 16 2024 - 10:34:06 EST
On Fri 2024-09-13 13:25:33, Rik van Riel wrote:
> On Wed, 2024-07-24 at 17:08 +0200, Petr Mladek wrote:
> > On Wed 2024-07-24 16:51:46, John Ogness wrote:
> >
> >
> > I see. I agree that deferring printk on that CPU [0] is the right
> > solution.
> >
> > > [0]
> > > https://lore.kernel.org/lkml/87plrcqyii.fsf@xxxxxxxxxxxxxxxxxxxxx
> >
> Hi Petr,
>
> we have been running with John Ogness's fix (from the mail linked
> above) for about a month now, and it has reduced the number of
> deadlocks in the Linux kernel in our fleet by maybe 90%, comparing
> a kernel with the fix, to a kernel without the fix.
>
> Would it be possible to get John Ogness's fix queued up for merging
> in the printk tree, and then later to Linus?
Sure. Thanks for reminder.
John, could you please send the fix as a proper patch with some
reasonable commit message?
Best Regards,
Petr