On Thu, 1 Feb 2024 14:18:45 +0800
Andy Chiu wrote:
> smp_processor_id() should be called with migration disabled. This mean
> we may safely call smp_processor_id() in percpu thread. However, this is
> not the case if the work is (re-)queued into unbound workqueue, during
> cpu-hotplog. So, detect and return early if this work happens to run on
> an unbound wq.
>
Yeah I triggered this too, but never made it a priority to fix.
> Signed-off-by: Andy Chiu
> ---
> kernel/trace/trace_osnoise.c | 7 ++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/trace/trace_osnoise.c b/kernel/trace/trace_osnoise.c
> index bd0d01d00fb9..cf7f716d3f35 100644
> --- a/kernel/trace/trace_osnoise.c
> +++ b/kernel/trace/trace_osnoise.c
> @@ -2068,7 +2068,12 @@ static int start_per_cpu_kthreads(void)
> #ifdef CONFIG_HOTPLUG_CPU
> static void osnoise_hotplug_workfn(struct work_struct *dummy)
> {
> - unsigned int cpu = smp_processor_id();
> + unsigned int cpu;
> +
> + if (!is_percpu_thread())
> + return;
But doesn't this then fail to register the CPU thread?
I wonder if this has some race with schedule_work_on() and hotplug?
-- Steve
> +
> + cpu = smp_processor_id();
>
> mutex_lock(&trace_types_lock);
>
