Re: [linus:master] [x86] 4817f70c25: stress-ng.mmapaddr.ops_per_sec 63.0% regression
From: Frederic Weisbecker
Date: Wed Jan 29 2025 - 11:55:33 EST
Le Wed, Jan 29, 2025 at 08:36:12AM -0800, Paul E. McKenney a écrit :
> On Wed, Jan 29, 2025 at 11:14:29AM -0500, Rik van Riel wrote:
> > On Wed, 2025-01-29 at 16:12 +0000, Matthew Wilcox wrote:
> > > On Wed, Jan 29, 2025 at 10:59:20AM -0500, Rik van Riel wrote:
> > > > Below is a tentative fix for the issue. It is kind of a big hammer,
> > > > and maybe the RCU people have a better idea on how to solve this
> > > > problem, but it may be worth giving this a try to see if it helps
> > > > with the regression you identified.
> > >
> > > Perhaps better to do:
> > >
> > > +++ b/mm/page_alloc.c
> > > @@ -97,8 +97,8 @@ static DEFINE_MUTEX(pcp_batch_high_lock);
> > > * On SMP, spin_trylock is sufficient protection.
> > > * On PREEMPT_RT, spin_trylock is equivalent on both SMP and UP.
> > > */
> > > -#define pcp_trylock_prepare(flags) do { } while (0)
> > > -#define pcp_trylock_finish(flag) do { } while (0)
> > > +#define pcp_trylock_prepare(flags) rcu_read_lock()
> > > +#define pcp_trylock_finish(flag) rcu_reada_unlock()
> > > #else
> > >
> > > /* UP spin_trylock always succeeds so disable IRQs to prevent re-
> > > entrancy. */
> > >
> > > with appropriate comment changes
> > >
> > Agreed. Assuming this change even works :)
> >
> > Paul, does this look like it could do the trick,
> > or do we need something else to make RCU freeing
> > happy again?
>
> I don't claim to fully understand the issue, but this would prevent
> any RCU grace periods starting subsequently from completing. It would
> not prevent RCU callbacks from being invoked for RCU grace periods that
> started earlier.
>
> So it won't prevent RCU callbacks from being invoked.
>
> It *will* ensure that only a finite number of RCU callbacks get invoked.
> For some perhaps rather large value of "finite".
>
> Does that help, or is more required?
I don't fully understand the issue either but would spin_lock_bh() help?
Thanks.