[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20251008101408.dj46r66gcfo26sgl@lcpd911>
Date: Wed, 8 Oct 2025 15:44:08 +0530
From: Dhruva Gole <d-gole@...com>
To: Sergey Senozhatsky <senozhatsky@...omium.org>
CC: "Rafael J. Wysocki" <rafael@...nel.org>, Pavel Machek <pavel@...nel.org>,
Tomasz Figa <tfiga@...omium.org>, <linux-pm@...r.kernel.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCHv2] PM: dpm: add module param to backtrace all CPUs
On Oct 07, 2025 at 15:35:40 +0900, Sergey Senozhatsky wrote:
> Add dpm_watchdog_all_cpu_backtrace module parameter which
> controls all CPU backtrace dump before DPM panics the system.
> This is expected to help understanding what might have caused
> device timeout.
This will indeed be really helpful for debugging some nasty bugs!
>
> Signed-off-by: Sergey Senozhatsky <senozhatsky@...omium.org>
> ---
> drivers/base/power/main.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> index e83503bdc1fd..7a8807ec9a5d 100644
> --- a/drivers/base/power/main.c
> +++ b/drivers/base/power/main.c
> @@ -34,6 +34,7 @@
> #include <linux/cpufreq.h>
> #include <linux/devfreq.h>
> #include <linux/timer.h>
> +#include <linux/nmi.h>
>
> #include "../base.h"
> #include "power.h"
> @@ -515,6 +516,11 @@ struct dpm_watchdog {
> #define DECLARE_DPM_WATCHDOG_ON_STACK(wd) \
> struct dpm_watchdog wd
>
> +static bool __read_mostly dpm_watchdog_all_cpu_backtrace;
> +module_param(dpm_watchdog_all_cpu_backtrace, bool, 0644);
> +MODULE_PARM_DESC(dpm_watchdog_all_cpu_backtrace,
> + "Backtrace all CPUs on DPM watchdog timeout");
> +
Have you considered runtime configurability instead of a module param?
> /**
> * dpm_watchdog_handler - Driver suspend / resume watchdog handler.
> * @t: The timer that PM watchdog depends on.
> @@ -530,8 +536,12 @@ static void dpm_watchdog_handler(struct timer_list *t)
> unsigned int time_left;
>
> if (wd->fatal) {
> + unsigned int this_cpu = smp_processor_id();
> +
> dev_emerg(wd->dev, "**** DPM device timeout ****\n");
> show_stack(wd->tsk, NULL, KERN_EMERG);
> + if (dpm_watchdog_all_cpu_backtrace)
> + trigger_allbutcpu_cpu_backtrace(this_cpu);
IMO it would be useful to check the ret val of this as well, I mean just
incase this silently returns false it maybe confusing to figure out what
hapenned in the system inspite of setting the mod param.
--
Best regards,
Dhruva Gole
Texas Instruments Incorporated
Powered by blists - more mailing lists