Re: [PATCH] kprobes: Make optprobe optimizer multi-generational and asynchronous

[Google]

On Thu, 24 Sep 2026 16:17:38 +0000 (UTC)
bot+bpf-ci@xxxxxxxxxx wrote:

> > diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> > index 6337da5cab9e..1e25980303c4 100644
> > --- a/kernel/kprobes.c
> > +++ b/kernel/kprobes.c
> 
> [ ... ]
> 
> > @@ -530,50 +553,78 @@ static DECLARE_COMPLETION(optimizer_completion);
> >
> >  #define OPTIMIZE_DELAY 5
> >
> > +static bool optprobe_has_queued_probes(void)
> > +{
> > +	struct optprobe_generation *gen = &optprobe_gens[optprobe_cur_gen];
> > +
> > +	return !list_empty(&gen->optimizing_list) ||
> > +	       !list_empty(&gen->unoptimizing_list);
> > +}
> 
> Does this logic need to check gen->freeing_list as well?

Ah, indeed. It should check the freeing_list too.

[ ... ]
> 
> > @@ -622,67 +675,119 @@ static void do_free_cleaned_kprobes(void)
> >
> >  static void kick_kprobe_optimizer(void);
> >
> > +static void optprobe_generation_rcu_cb(struct rcu_head *rcu)
> > +{
> > +	struct optprobe_generation *gen;
> > +
> > +	gen = container_of(rcu, struct optprobe_generation, rcu);
> > +	WRITE_ONCE(gen->ready, true);
> > +	wake_up(&kprobe_optimizer_wait);
> > +}
> > +
> > +static void optprobe_finalize_generation(struct optprobe_generation *gen)
> > +{
> > +	lockdep_assert_held(&kprobe_mutex);
> > +
> > +	scoped_guard(cpus_read_lock) {
> > +		guard(mutex)(&text_mutex);
> > +
> > +		/* Optimization never be done when disarmed */
> > +		if (!kprobes_all_disarmed && kprobes_allow_optimization &&
> > +		    !list_empty(&gen->optimizing_list))
> > +			arch_optimize_kprobes(&gen->optimizing_list);
> > +	}
> > +
> > +	/* Free cleaned kprobes after quiescence period */
> > +	do_free_cleaned_kprobes(&gen->freeing_list);
> > +
> > +	gen->in_flight = false;
> > +	WRITE_ONCE(gen->ready, false);
> > +}
> > +
> > +static void optprobe_dispatch_generation(void)
> > +{
> > +	struct optprobe_generation *gen;
> > +
> > +	lockdep_assert_held(&kprobe_mutex);
> > +
> > +	if (!optprobe_can_fire() || !optprobe_has_queued_probes())
> > +		return;
> > +
> > +	gen = &optprobe_gens[optprobe_cur_gen];
> > +
> > +	scoped_guard(cpus_read_lock) {
> > +		guard(mutex)(&text_mutex);
> > +
> > +		/*
> > +		 * Unoptimize kprobes and collect cleaned (unused and disarmed)
> > +		 * kprobes before waiting for quiescence period.
> > +		 */
> > +		do_unoptimize_kprobes(&gen->unoptimizing_list, &gen->freeing_list);
> > +	}
> > +
> > +	/* Advance cur_gen to the next generation slot */
> > +	optprobe_cur_gen = (optprobe_cur_gen + 1) % OPTPROBE_GEN_MAX;
> > +
> > +	gen->in_flight = true;
> > +	WRITE_ONCE(gen->ready, false);
> > +
> > +	call_rcu_tasks(&gen->rcu, optprobe_generation_rcu_cb);
> > +}
> 
> Now that kprobe_mutex is dropped during the Tasks RCU grace period, can
> an aggregator probe that was unoptimized but is still in use be freed
> while a task is still preempted inside its detour buffer?

Good catch! Yes, we need to keep tracking the unoptimized but not
freed kprobes until finalize it.

[ ... ]
> 
> > @@ -707,12 +812,11 @@ static void wait_for_kprobe_optimizer_locked(void)
> >  {
> >  	lockdep_assert_held(&kprobe_mutex);
> >
> > -	while (!list_empty(&optimizing_list) || !list_empty(&unoptimizing_list)) {
> > +	while (optprobe_optimizer_busy()) {
> >  		init_completion(&optimizer_completion);
> > -		/*
> > -		 * Set state to OPTIMIZER_ST_FLUSHING and wake up the thread if it's
> > -		 * idle. If it's already kicked, it will see the state change.
> > -		 */
> > +		optprobe_flush_requested = true;
> > +
> > +		/* Wake up optimizer thread */
> >  		if (atomic_xchg_acquire(&optimizer_state,
> >  			OPTIMIZER_ST_FLUSHING) != OPTIMIZER_ST_FLUSHING)
> >  			wake_up(&kprobe_optimizer_wait);
> 
> Andrea Parri raised a concern on 2026-09-24 about a race condition in
> this code (https://lore.kernel.org/all/arTuBodyGqQ0mErI@andreayoga.localdomain/).

Yes, so I pick his patch at first and rebase this on it.

Thanks,

-- 
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>