[RFC PATCH 1/4] rseq: Add sched_state field to struct rseq
Mathieu Desnoyers
mathieu.desnoyers@efficios.com
Tue May 23 12:49:33 GMT 2023
On 2023-05-19 16:51, Noah Goldstein wrote:
> On Wed, May 17, 2023 at 10:28 AM Mathieu Desnoyers via Libc-alpha
> <libc-alpha@sourceware.org> wrote:
>>
>> Expose the "on-cpu" state for each thread through struct rseq to allow
>> adaptative mutexes to decide more accurately between busy-waiting and
>> calling sys_futex() to release the CPU, based on the on-cpu state of the
>> mutex owner.
>>
>> It is only provided as an optimization hint, because there is no
>> guarantee that the page containing this field is in the page cache, and
>> therefore the scheduler may very well fail to clear the on-cpu state on
>> preemption. This is expected to be rare though, and is resolved as soon
>> as the task returns to user-space.
>>
>> The goal is to improve use-cases where the duration of the critical
>> sections for a given lock follows a multi-modal distribution, preventing
>> statistical guesses from doing a good job at choosing between busy-wait
>> and futex wait behavior.
>>
>> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
>> Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
>> Cc: Jonathan Corbet <corbet@lwn.net>
>> Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
>> Cc: Carlos O'Donell <carlos@redhat.com>
>> Cc: Florian Weimer <fweimer@redhat.com>
>> Cc: libc-alpha@sourceware.org
>> ---
>> include/linux/sched.h | 12 ++++++++++++
>> include/uapi/linux/rseq.h | 17 +++++++++++++++++
>> kernel/rseq.c | 14 ++++++++++++++
>> 3 files changed, 43 insertions(+)
>>
>> diff --git a/include/linux/sched.h b/include/linux/sched.h
>> index eed5d65b8d1f..c7e9248134c1 100644
>> --- a/include/linux/sched.h
>> +++ b/include/linux/sched.h
>> @@ -2351,11 +2351,20 @@ static inline void rseq_signal_deliver(struct ksignal *ksig,
>> rseq_handle_notify_resume(ksig, regs);
>> }
>>
>> +void __rseq_set_sched_state(struct task_struct *t, unsigned int state);
>> +
>> +static inline void rseq_set_sched_state(struct task_struct *t, unsigned int state)
>> +{
>> + if (t->rseq)
>> + __rseq_set_sched_state(t, state);
>> +}
>> +
>> /* rseq_preempt() requires preemption to be disabled. */
>> static inline void rseq_preempt(struct task_struct *t)
>> {
>> __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
>> rseq_set_notify_resume(t);
>> + rseq_set_sched_state(t, 0);
>
> Should rseq_migrate also be made to update the cpu_id of the new core?
> I imagine the usage of this will be something along the lines of:
>
> if(!on_cpu(mutex->owner_rseq_struct) &&
> cpu(mutex->owner_rseq_struct) == this_threads_cpu)
> // goto futex
>
> So I would think updating on migrate would be useful as well.
I don't think we want to act differently based on the cpu on which the
owner is queued.
If the mutex owner is not on-cpu, and queued on the same cpu as the
current thread, we indeed want to call sys_futex WAIT.
If the mutex owner is not on-cpu, but queued on a different cpu than the
current thread, we *still* want to call sys_futex WAIT, because
busy-waiting for a thread which is queued but not currently running is
wasteful.
Or am I missing something ?
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
More information about the Libc-alpha
mailing list