Re: [PATCH v2] genirq: add support for warning on long-running IRQ handlers

[Wladislav Wiebe]

On 22/07/2025 10:21, Jiri Slaby wrote:

>
> On 14. 07. 25, 10:41, Wladislav Wiebe wrote:
>> This patch adds a mechanism to detect and warn about long-running IRQ
>> handlers exceeding a user-defined duration threshold in microseconds.
>>
>> The feature is enabled via the kernel boot parameter:
>> "irqhandler.duration_warn_us=<threshold_in_us>"
>>
>> For example, passing irqhandler.duration_warn_us=1000 will warn if an
>> IRQ handler takes more than 1000 microseconds.
>>
>> Implementation uses local_clock() to measure the execution duration of
>> IRQ handlers. When the threshold is exceeded, a ratelimited warning is
>> printed:
>>
>> "[CPU14] long duration on IRQ[159:bad_irq_handler [long_irq]], took: 1330 us"
>>
>> Signed-off-by: Wladislav Wiebe <wladislav.wiebe@xxxxxxxxx>
>> ---
>> V1 -> V2: refactor to use local_clock() instead of jiffies and replace
>>         Kconfig knobs by a new command-line parameter.
>> V1 link:  https://lore.kernel.org/lkml/20250630124721.18232-1-wladislav.wiebe@xxxxxxxxx/
>> ---
>>   .../admin-guide/kernel-parameters.txt         |  5 ++
>>   kernel/irq/handle.c                           | 48 ++++++++++++++++++-
>>   2 files changed, 52 insertions(+), 1 deletion(-)
>>
>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>> index f1f2c0874da9..fa89f21ea1e6 100644
>> --- a/Documentation/admin-guide/kernel-parameters.txt
>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>> @@ -2543,6 +2543,11 @@
>>                       for it. Intended to get systems with badly broken
>>                       firmware running.
>>
>> +     irqhandler.duration_warn_us= [KNL,EARLY]
>> +                     Warn if an IRQ handler exceeds the specified duration
>> +                     threshold in microseconds. Useful for identifying
>> +                     long-running IRQs in the system.
>> +
>>       irqpoll         [HW]
>>                       When an interrupt is not handled search all handlers
>>                       for it. Also check all handlers each timer
>> diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
>> index 9489f93b3db3..eab8fdfab8d8 100644
>> --- a/kernel/irq/handle.c
>> +++ b/kernel/irq/handle.c
>> @@ -136,6 +136,44 @@ void __irq_wake_thread(struct irq_desc *desc, struct irqaction *action)
>>       wake_up_process(action->thread);
>>   }
>>
>> +static DEFINE_STATIC_KEY_FALSE(irqhandler_duration_check_enabled);
>> +static u64 irqhandler_duration_threshold_us __ro_after_init;
>> +
>> +static int __init irqhandler_duration_check_setup(char *arg)
>> +{
>> +     unsigned long val;
>> +     int ret;
>> +
>> +     if (!arg)
>> +             return 0;
>> +
>> +     ret = kstrtoul(arg, 0, &val);
>> +     if (ret)
>> +             return ret;
>> +
>> +     if (val > 0) {
>> +             irqhandler_duration_threshold_us = val;
>> +             static_branch_enable(&irqhandler_duration_check_enabled);
>> +     } else {
>> +             pr_err("Invalid irqhandler.duration_warn_us setting (%lu)\n", val);
>> +             return -EINVAL;
>
> Perhaps invert the condition and drop the "else {}"?
>
>> +     }
>> +
>> +     return 0;
>> +}
>> +early_param("irqhandler.duration_warn_us", irqhandler_duration_check_setup);
>> +
>> +static inline void irqhandler_duration_check(u64 ts_start, unsigned int irq,
>> +                                           struct irqaction *action)
>
> Can be const.
>
>> +{
>> +     u64 delta_us = (local_clock() - ts_start) >> 10;
>> +
>> +     if (unlikely(delta_us > irqhandler_duration_threshold_us)) {
>> +             pr_warn_ratelimited("[CPU%d] long duration on IRQ[%u:%ps], took: %llu us\n",
>
> s/%d/%u/.
> Do you mean "of IRQ[...]"?
>
>> +                     smp_processor_id(), irq, action->handler, delta_us);
>> +     }
>> +}
>> +
>>   irqreturn_t __handle_irq_event_percpu(struct irq_desc *desc)
>>   {
>>       irqreturn_t retval = IRQ_NONE;

Thanks for the comments, I've addressed them in v3:
https://lore.kernel.org/lkml/20250723182836.1177-1-wladislav.wiebe@xxxxxxxxx/

- W.W.