Re: [PATCH] perf symbols: Add 'intel_idle_irq' to the list of idle symbols

[Honglei Wang]

On 3/26/26 12:07 AM, Ian Rogers wrote:
> On Wed, Mar 25, 2026 at 12:05 AM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
>>
>> On Mon, Mar 23, 2026 at 11:35:05AM -0700, Ian Rogers wrote:
>>> On Mon, Mar 23, 2026 at 1:53 AM Honglei Wang <jameshongleiwang@xxxxxxx> wrote:
>>>>
>>>> Commit 32d4fd5751ea ("cpuidle,intel_idle: Fix CPUIDLE_FLAG_IRQ_ENABLE")
>>>> introduced intel_idle_irq as entry of those with
>>>> CPUIDLE_FLAG_IRQ_ENABLE, say, skx and icx so far. Let's add a new
>>>> one to the list of idle symbols to avoid noise.
>>>>
>>>> Signed-off-by: Honglei Wang <jameshongleiwang@xxxxxxx>
>>>> ---
>>>>  tools/perf/util/symbol.c | 1 +
>>>>  1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
>>>> index 8662001e1e25..7cff66e7ce06 100644
>>>> --- a/tools/perf/util/symbol.c
>>>> +++ b/tools/perf/util/symbol.c
>>>> @@ -730,6 +730,7 @@ static bool symbol__is_idle(const char *name)
>>>>                 "cpu_startup_entry",
>>>>                 "idle_cpu",
>>>>                 "intel_idle",
>>>> +               "intel_idle_irq",
>>>
>>> It would be nice to make this an Intel specific idle symbol. I sent
>>> out a patch set to do this:
>>> https://lore.kernel.org/lkml/20260302234343.564937-1-irogers@xxxxxxxxxx/
>>> but there are no reviews. Perhaps we can follow up that patch by
>>> adding this symbol. Arnaldo, Namhyung, wdyt? I believe Arnaldo was
>>> supportive in:
>>> https://lore.kernel.org/lkml/aahC-zp56irr5e2k@x1/
>>
>> Yeah, I'm ok in general.  I've left a comment there.
> 
> As that patch adds:
> ```
> if (e_machine == EM_386 || e_machine == EM_X86_64) {
>     if (strstarts(name, "mwait_idle") ||
>          strstarts(name, "intel_idle")) {
>         sym->idle = SYMBOL_IDLE__IDLE;
>     return true;
>     }
> }
> ```
> 
> The strstarts covers intel_idle_irq and that change will incorporate
> this modification.
> 
> Thanks,
> Ian
> 

Sorry, I was a bit busy these two days.. I'm OK to make this general.
But at a fast test on my machine with your v2 patch, the intel_idle_irq
was still there at top of the output. I'll do more debug to see why it
happened.

Thanks,
Honglei

>> Thanks,
>> Namhyung