Gabriele Monaco <[email protected]> writes:

> From: Wen Yang <[email protected]>
>
> da_monitor_start() set monitoring=1 before calling da_monitor_init_hook(),
> may racing with the sched_switch handler:
>
>   da_monitor_start()               sched_switch handler
>   -------------------------        ---------------------------------
>   da_mon->monitoring = 1;
>                                    if (da_monitoring(da_mon))  /* true  */
>                                        ha_start_timer_ns(...);
>                                        /* hrtimer->base == NULL, crash */
>   da_monitor_init_hook(da_mon);
>   /* hrtimer_setup() sets base */
>
> Fix the ordering and pair with release/acquire semantics:
>
>   da_monitor_init_hook(da_mon);
>   smp_store_release(&da_mon->monitoring, 1);    /* da_monitor_start()  */
>   return smp_load_acquire(&da_mon->monitoring); /* da_monitoring()     */
>
> On ARM64 a plain STR + LDR does not form a release-acquire pair, so
> the load can observe monitoring=1 while hrtimer->base is still NULL.
> The plain accesses are also data races under KCSAN.
>
> Use WRITE_ONCE for the monitoring=0 store in da_monitor_reset() to
> cover the reset path.
>
> Fixes: 792575348ff7 ("rv/include: Add deterministic automata monitor 
> definition via C macros")
> Signed-off-by: Wen Yang <[email protected]>
> Reviewed-by: Gabriele Monaco <[email protected]>
> Reviewed-by: Nam Cao <[email protected]>
> Signed-off-by: Gabriele Monaco <[email protected]>
> ---
>  include/rv/da_monitor.h | 8 +++++---
>  1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/include/rv/da_monitor.h b/include/rv/da_monitor.h
> index a7e103654..60dc39f26 100644
> --- a/include/rv/da_monitor.h
> +++ b/include/rv/da_monitor.h
> @@ -82,7 +82,7 @@ static void react(enum states curr_state, enum events event)
>  static inline void da_monitor_reset(struct da_monitor *da_mon)
>  {
>       da_monitor_reset_hook(da_mon);
> -     da_mon->monitoring = 0;
> +     WRITE_ONCE(da_mon->monitoring, 0);
>       da_mon->curr_state = model_get_initial_state();
>  }

Looking at this again, do you need to change it to

static inline void da_monitor_reset(struct da_monitor *da_mon)
{
        WRITE_ONCE(da_mon->monitoring, 0);
        smp_mb();
        da_monitor_reset_hook(da_mon);
        da_mon->curr_state = model_get_initial_state();
}

To prevent another task from seeing monitoring=1 while the timer is
already cancelled?

Nam

Reply via email to