[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250526055008.GA619414@IPBU-BLR-SERVER1.marvell.com>
Date: Mon, 26 May 2025 11:20:08 +0530
From: George Cherian <george.cherian@...vell.com>
To: Guixiong Wei <weiguixiong@...edance.com>
CC: <wim@...ux-watchdog.org>, <linux@...ck-us.net>,
<linux-watchdog@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [RFC] watchdog: Introduce panic notifier for watchdog device on
kdump
Hi Guixiong,
On 2025-05-25 at 17:49:39, Guixiong Wei (weiguixiong@...edance.com) wrote:
> The watchdog device is not stop after kenrel crash. But the kexec
> kernel may not enable watchdog device. This will interrupt the
> long-time kdump process(e.g., TDX VMs with large-memory).
>
> So introduce panic notifer for watchdog device. When kernel crashes,
> the handler invokes the watchdog stop callback, then the kdump
> process will not be interrupted by the watchdog device.
>
> Signed-off-by: Guixiong Wei <weiguixiong@...edance.com>
A similar attempt is made in the following series here
Patch 1: https://patchwork.kernel.org/project/linux-watchdog/patch/20250522055715.3533356-2-george.cherian@marvell.com/
Patch 2: https://patchwork.kernel.org/project/linux-watchdog/patch/20250522055715.3533356-3-george.cherian@marvell.com/
For your particular use-case you need to set reset_on_panic to zero via
sysfs. This would disable the watchdog in panic notifier.
It would be helpful if you could test the series above and give your
comments/Tested-by.
Regards,
-George
Powered by blists - more mailing lists