On Thu, Oct 13, 2022 at 07:22:42PM +0200, Frederic Weisbecker wrote:
> Using the NMI-unsafe reader API from within NMIs is very likely to be
> buggy for three reasons:
>
> 1) NMIs aren't strictly re-entrant (a pending nested NMI will execute
> at the end of the current one) so it should be fine to use a
> non-atomic increment here. However breakpoints can still interrupt
> NMIs and if a breakpoint callback has a reader on that same ssp, a
> racy increment can happen.
>
> 2) If the only reader site for a given ssp is in an NMI, RCU is definetly
definitely
> a better choice over SRCU.
Just checking - because NMI are by definition not-preemptibe, so SRCU over
RCU doesn't make much sense right?
Reviewed-by: Joel Fernandes (Google) <joel@xxxxxxxxxxxxxxxxx>
thanks,
- Joel
>
> 3) Because of the previous reason (2), an ssp having an SRCU read side
> critical section in an NMI is likely to have another one from a task
> context.
>
> For all these reasons, warn if an nmi unsafe reader API is used from an
> NMI.
>
> Signed-off-by: Frederic Weisbecker <frederic@xxxxxxxxxx>
> ---
> kernel/rcu/srcutree.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index c54142374793..8b7ef1031d89 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -642,6 +642,8 @@ static void srcu_check_nmi_safety(struct srcu_struct *ssp, bool nmi_safe)
>
> if (!IS_ENABLED(CONFIG_PROVE_RCU))
> return;
> + /* NMI-unsafe use in NMI is a bad sign */
> + WARN_ON_ONCE(!nmi_safe && in_nmi());
> sdp = raw_cpu_ptr(ssp->sda);
> old_nmi_safe_mask = READ_ONCE(sdp->srcu_nmi_safety);
> if (!old_nmi_safe_mask) {
> --
> 2.25.1
>
