Re: [PATCH 9/9] watchdog: skip checks when panic is in progress

[Jinchao Wang <wangjinchao600@gmail.com>]

On 8/20/25 23:18, Yury Norov wrote:
> On Wed, Aug 20, 2025 at 05:14:54PM +0800, Jinchao Wang wrote:
>> Both watchdog_buddy_check_hardlockup() and
>> watchdog_overflow_callback() may trigger
>> during a panic. This can lead to recursive
>> panic handling.
>>
>> Add panic_in_progress() checks so watchdog
>> activity is skipped once a panic has begun.
>>
>> This prevents recursive panic and keeps the
>> panic path more reliable.
>>
>> Signed-off-by: Jinchao Wang <wangjinchao600@gmail.com>
>> ---
>>   kernel/watchdog_buddy.c | 5 +++++
>>   kernel/watchdog_perf.c  | 3 +++
>>   2 files changed, 8 insertions(+)
>>
>> diff --git a/kernel/watchdog_buddy.c b/kernel/watchdog_buddy.c
>> index ee754d767c21..79a85623028c 100644
>> --- a/kernel/watchdog_buddy.c
>> +++ b/kernel/watchdog_buddy.c
>> @@ -93,6 +93,11 @@ void watchdog_buddy_check_hardlockup(int hrtimer_interrupts)
>>   	 */
>>   	if (hrtimer_interrupts % 3 != 0)
>>   		return;
>> +	/*
>> +	 * pass the buddy check if a panic is in process
>> +	 */
>> +	if (panic_in_progress())
>> +		return;
>>   
>>   	/* check for a hardlockup on the next CPU */
>>   	next_cpu = watchdog_next_cpu(smp_processor_id());
>> diff --git a/kernel/watchdog_perf.c b/kernel/watchdog_perf.c
>> index 9c58f5b4381d..7641de750ca5 100644
>> --- a/kernel/watchdog_perf.c
>> +++ b/kernel/watchdog_perf.c
>> @@ -12,6 +12,7 @@
>>   
>>   #define pr_fmt(fmt) "NMI watchdog: " fmt
>>   
>> +#include <linux/panic.h>
>>   #include <linux/nmi.h>
>>   #include <linux/atomic.h>
>>   #include <linux/module.h>
>> @@ -110,6 +111,8 @@ static void watchdog_overflow_callback(struct perf_event *event,
>>   
>>   	if (!watchdog_check_timestamp())
>>   		return;
>> +	if (panic_in_progress())
>> +		return;
> 
> It looks like watchdog_check_timestamp() does some real work, like
> updates last_timestamp and so on. Under the panic condition all this
> may be unreliable, right?
> 
> Maybe it's worth to make panic_in_progress() the first check in the
> chain?
> 
That's a good point. Thank you.
> With that,
> 
> Reviewed-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> 
>>   
>>   	watchdog_hardlockup_check(smp_processor_id(), regs);
>>   }
>> -- 
>> 2.43.0


-- 
Best regards,
Jinchao