Re: [BUG] WARNING in perf_cgroup_switch: EVENT_FROZEN assertion during kthread sched-out

[Christopher Lusk]

The WARNING was observed on 6.12.0 (RHEL 10 build 124.38.1.el10_1).

I'm also running a parallel syzkaller campaign against 6.19.9
(mainline, built from stable) which has been running for ~6 days with
no hits on this WARNING. That campaign uses the same syzkaller
revision and similar syscall coverage but does not have KASAN enabled.

The 6.12.0 campaign that triggered this does have KASAN enabled, which
may affect timing enough to widen the race window.

Happy to test against a specific tree or commit if that would help.

On Wed, Apr 1, 2026 at 12:05 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Wed, Apr 01, 2026 at 10:15:56AM -0400, Christopher Lusk wrote:
> > Hello,
> >
> > I'm hitting a WARNING in perf_cgroup_switch() on kernel 6.12.0 during
> > continuous syzkaller fuzzing. The assertion WARN_ON_ONCE(ctx->is_active &
> > EVENT_FROZEN) in __perf_ctx_lock() fires when the rcu_preempt kthread is
> > being scheduled out.
> >
> > Kernel: 6.12.0 (RHEL 10 build, no out-of-tree modules)
>
> Does this reproduce on anything recent?