Re: [PATCH] perf/core: Allow list_del during perf_event_overflow()
From: Peter Zijlstra
Date: Tue Sep 08 2026 - 06:46:57 EST
On Tue, Sep 08, 2026 at 08:47:59AM +0200, Thomas Richter wrote:
> A PMU might use perf_sched_cb_inc() and perf_sched_cb_dec()
> interface to get the PMU call back function pmu::sched_task
> invoked at schedule in and schedule out. This is achieved
> by walking along the list anchored by sched_cb_list.
>
> The following scenario might lead to a list corruption.
>
> perf_pmu_sched_task()
> for_each_list_entry(..., &sched_cb_list)
> +--> __perf_pmu_sched_task()
> +--> event->pmu->sched_task())
> +--> PMU_push_sample()
> +--> perf_event_overflow()
> +--> __perf_event_overflow()
> +--> pmu->stop()
> +--> perf_sched_cb_dec()
> remove entry from sched_cb_list
> while list node in use.
>
> This happens when ioctl(fd, PERF_EVENT_IOC_REFRESH, xxx) has been
> invoked and perf_event::event_limit hits zero.
>
> Prevent the list corruption and convert for_each_list_entry()
> to for_each_list_entry_safe().
>
This reads like it wants a Fixes tag.
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Signed-off-by: Thomas Richter <tmricht@xxxxxxxxxxxxx>
> ---
> kernel/events/core.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index a6c8e38a3110..cdd9518ac046 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -3925,13 +3925,13 @@ static void perf_pmu_sched_task(struct task_struct *prev,
> bool sched_in)
> {
> struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context);
> - struct perf_cpu_pmu_context *cpc;
> + struct perf_cpu_pmu_context *cpc, *cpc2;
>
> /* cpuctx->task_ctx will be handled in perf_event_context_sched_in/out */
> if (prev == next || cpuctx->task_ctx)
> return;
>
> - list_for_each_entry(cpc, this_cpu_ptr(&sched_cb_list), sched_cb_entry)
> + list_for_each_entry_safe(cpc, cpc2, this_cpu_ptr(&sched_cb_list), sched_cb_entry)
> __perf_pmu_sched_task(cpc, sched_in ? next : prev, sched_in);
> }
>
> --
> 2.55.0
>