The sanity checks in hrtimer_update_function() are expensive for high
frequency usage like in the io/uring code due to locking.

Hide the sanity checks behind CONFIG_PROVE_LOCKING, which has a decent
chance to be enabled on a regular basis for testing.

Fixes: 8f02e3563bb5 ("hrtimers: Introduce hrtimer_update_function()")
Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 include/linux/hrtimer.h |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

--- a/include/linux/hrtimer.h
+++ b/include/linux/hrtimer.h
@@ -333,6 +333,7 @@ static inline int hrtimer_callback_runni
 static inline void hrtimer_update_function(struct hrtimer *timer,
 					   enum hrtimer_restart (*function)(struct hrtimer *))
 {
+#ifdef CONFIG_PROVE_LOCKING
 	guard(raw_spinlock_irqsave)(&timer->base->cpu_base->lock);
 
 	if (WARN_ON_ONCE(hrtimer_is_queued(timer)))
@@ -340,7 +341,7 @@ static inline void hrtimer_update_functi
 
 	if (WARN_ON_ONCE(!function))
 		return;
-
+#endif
 	timer->function = function;
 }

On 2/7/25 2:16 PM, Thomas Gleixner wrote:
> The sanity checks in hrtimer_update_function() are expensive for high
> frequency usage like in the io/uring code due to locking.
> 
> Hide the sanity checks behind CONFIG_PROVE_LOCKING, which has a decent
> chance to be enabled on a regular basis for testing.

Looks good to me, thanks Thomas. On my side, I always have a debug run
done with PROVE_LOCKING and KASAN, fwiw.

-- 
Jens Axboe

On Sat, Feb 08 2025 at 08:18, Jens Axboe wrote:

> On 2/7/25 2:16 PM, Thomas Gleixner wrote:
>> The sanity checks in hrtimer_update_function() are expensive for high
>> frequency usage like in the io/uring code due to locking.
>> 
>> Hide the sanity checks behind CONFIG_PROVE_LOCKING, which has a decent
>> chance to be enabled on a regular basis for testing.
>
> Looks good to me, thanks Thomas. On my side, I always have a debug run
> done with PROVE_LOCKING and KASAN, fwiw.

I assume that with that your objections against the conversion of ioring
to hrtimer_update_function() is gone too.

Thanks,

        tglx

On 2/10/25 12:52 PM, Thomas Gleixner wrote:
> On Sat, Feb 08 2025 at 08:18, Jens Axboe wrote:
> 
>> On 2/7/25 2:16 PM, Thomas Gleixner wrote:
>>> The sanity checks in hrtimer_update_function() are expensive for high
>>> frequency usage like in the io/uring code due to locking.
>>>
>>> Hide the sanity checks behind CONFIG_PROVE_LOCKING, which has a decent
>>> chance to be enabled on a regular basis for testing.
>>
>> Looks good to me, thanks Thomas. On my side, I always have a debug run
>> done with PROVE_LOCKING and KASAN, fwiw.
> 
> I assume that with that your objections against the conversion of ioring
> to hrtimer_update_function() is gone too.

Certainly, with the expensive bits under PROVE_LOCKING, I have no
objections to the patch.

-- 
Jens Axboe

The following commit has been merged into the timers/core branch of tip:

Commit-ID:     2ea97b76d6712bfb0408e5b81ffd7bc4551d3153
Gitweb:        https://git.kernel.org/tip/2ea97b76d6712bfb0408e5b81ffd7bc4551d3153
Author:        Thomas Gleixner <tglx@linutronix.de>
AuthorDate:    Fri, 07 Feb 2025 22:16:09 +01:00
Committer:     Thomas Gleixner <tglx@linutronix.de>
CommitterDate: Mon, 10 Feb 2025 20:51:12 +01:00

hrtimers: Make hrtimer_update_function() less expensive

The sanity checks in hrtimer_update_function() are expensive for high
frequency usage like in the io/uring code due to locking.

Hide the sanity checks behind CONFIG_PROVE_LOCKING, which has a decent
chance to be enabled on a regular basis for testing.

Fixes: 8f02e3563bb5 ("hrtimers: Introduce hrtimer_update_function()")
Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/87ikpllali.ffs@tglx

---
 include/linux/hrtimer.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/linux/hrtimer.h b/include/linux/hrtimer.h
index f7bfdcf..bdd55c1 100644
--- a/include/linux/hrtimer.h
+++ b/include/linux/hrtimer.h
@@ -333,6 +333,7 @@ static inline int hrtimer_callback_running(struct hrtimer *timer)
 static inline void hrtimer_update_function(struct hrtimer *timer,
 					   enum hrtimer_restart (*function)(struct hrtimer *))
 {
+#ifdef CONFIG_PROVE_LOCKING
 	guard(raw_spinlock_irqsave)(&timer->base->cpu_base->lock);
 
 	if (WARN_ON_ONCE(hrtimer_is_queued(timer)))
@@ -340,7 +341,7 @@ static inline void hrtimer_update_function(struct hrtimer *timer,
 
 	if (WARN_ON_ONCE(!function))
 		return;
-
+#endif
 	timer->function = function;
 }