On Wed, May 17, 2023 at 11:26:51AM -0400, Mathieu Desnoyers 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@xxxxxxxxxxxx>
> Cc: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> Cc: Jonathan Corbet <corbet@xxxxxxx>
> Cc: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx>
> Cc: Carlos O'Donell <carlos@xxxxxxxxxx>
> Cc: Florian Weimer <fweimer@xxxxxxxxxx>
> Cc: libc-alpha@xxxxxxxxxxxxxx
> ---
> 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);
> }
>
> /* rseq_migrate() requires preemption to be disabled. */
> @@ -2405,6 +2414,9 @@ static inline void rseq_signal_deliver(struct ksignal *ksig,
> struct pt_regs *regs)
> {
> }
> +static inline void rseq_set_sched_state(struct task_struct *t, unsigned int state)
> +{
> +}
> static inline void rseq_preempt(struct task_struct *t)
> {
> }
> diff --git a/include/uapi/linux/rseq.h b/include/uapi/linux/rseq.h
> index c233aae5eac9..c6d8537e23ca 100644
> --- a/include/uapi/linux/rseq.h
> +++ b/include/uapi/linux/rseq.h
> @@ -37,6 +37,13 @@ enum rseq_cs_flags {
> (1U << RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT),
> };
>
> +enum rseq_sched_state {
> + /*
> + * Task is currently running on a CPU if bit is set.
> + */
> + RSEQ_SCHED_STATE_ON_CPU = (1U << 0),
> +};
> +
> /*
> * struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always
> * contained within a single cache-line. It is usually declared as
> @@ -148,6 +155,16 @@ struct rseq {
> */
> __u32 mm_cid;
>
> + /*
> + * Restartable sequences sched_state field. Updated by the kernel. Read
> + * by user-space with single-copy atomicity semantics. This fields can
> + * be read by any userspace thread. Aligned on 32-bit. Contains a
Maybe this is a premature optimization, but since most of the time the
bit would be read by another thread, does it make sense putting the
"sched_state" into a different cache line to avoid false sharing?
Also We could have a "sched_state_local" and "sched_state_remote" for
different usages (local reads vs remote reads).
Regards,
Boqun
> + * bitmask of enum rseq_sched_state. This field is provided as a hint
> + * by the scheduler, and requires that the page holding struct rseq is
> + * faulted-in for the state update to be performed by the scheduler.
> + */
> + __u32 sched_state;
> +
> /*
> * Flexible array member at end of structure, after last feature field.
> */
> diff --git a/kernel/rseq.c b/kernel/rseq.c
> index 9de6e35fe679..b2eb3bbaa9ef 100644
> --- a/kernel/rseq.c
> +++ b/kernel/rseq.c
> @@ -91,6 +91,7 @@ static int rseq_update_cpu_node_id(struct task_struct *t)
> u32 cpu_id = raw_smp_processor_id();
> u32 node_id = cpu_to_node(cpu_id);
> u32 mm_cid = task_mm_cid(t);
> + u32 sched_state = RSEQ_SCHED_STATE_ON_CPU;
>
> WARN_ON_ONCE((int) mm_cid < 0);
> if (!user_write_access_begin(rseq, t->rseq_len))
> @@ -99,6 +100,7 @@ static int rseq_update_cpu_node_id(struct task_struct *t)
> unsafe_put_user(cpu_id, &rseq->cpu_id, efault_end);
> unsafe_put_user(node_id, &rseq->node_id, efault_end);
> unsafe_put_user(mm_cid, &rseq->mm_cid, efault_end);
> + unsafe_put_user(sched_state, &rseq->sched_state, efault_end);
> /*
> * Additional feature fields added after ORIG_RSEQ_SIZE
> * need to be conditionally updated only if
> @@ -339,6 +341,18 @@ void __rseq_handle_notify_resume(struct ksignal *ksig, struct pt_regs *regs)
> force_sigsegv(sig);
> }
>
> +/*
> + * Attempt to update rseq scheduler state.
> + */
> +void __rseq_set_sched_state(struct task_struct *t, unsigned int state)
> +{
> + if (unlikely(t->flags & PF_EXITING))
> + return;
> + pagefault_disable();
> + (void) put_user(state, &t->rseq->sched_state);
> + pagefault_enable();
> +}
> +
> #ifdef CONFIG_DEBUG_RSEQ
>
> /*
> --
> 2.25.1
>
