Message-ID: <CAGsJ_4ysvUj1OWobTpWhgJ1TAwRvVU+X0S8qgeBPE8=z2SbumQ@mail.gmail.com>
Date: Sat, 5 Oct 2024 00:03:17 +0800
From: Barry Song <21cnbao@...il.com>
To: Chris Li <chrisl@...nel.org>
Cc: "Huang, Ying" <ying.huang@...el.com>, akpm@...ux-foundation.org, david@...hat.com, 
	hannes@...xchg.org, hughd@...gle.com, kaleshsingh@...gle.com, 
	kasong@...cent.com, linux-kernel@...r.kernel.org, linux-mm@...ck.org, 
	liyangouwen1@...o.com, mhocko@...e.com, minchan@...nel.org, sj@...nel.org, 
	stable@...r.kernel.org, surenb@...gle.com, v-songbaohua@...o.com, 
	willy@...radead.org, yosryahmed@...gle.com, yuzhao@...gle.com
Subject: Re: [PATCH] mm: avoid unconditional one-tick sleep when
 swapcache_prepare fails

On Fri, Oct 4, 2024 at 7:03 AM Chris Li <chrisl@...nel.org> wrote:
>
> On Wed, Oct 2, 2024 at 5:35 PM Huang, Ying <ying.huang@...el.com> wrote:
> >
> > Barry Song <21cnbao@...il.com> writes:
> >
> > > On Wed, Oct 2, 2024 at 8:43 AM Huang, Ying <ying.huang@...el.com> wrote:
> > >>
> > >> Barry Song <21cnbao@...il.com> writes:
> > >>
> > >> > On Tue, Oct 1, 2024 at 7:43 AM Huang, Ying <ying.huang@...el.com> wrote:
> > >> >>
> > >> >> Barry Song <21cnbao@...il.com> writes:
> > >> >>
> > >> >> > On Sun, Sep 29, 2024 at 3:43 PM Huang, Ying <ying.huang@...el.com> wrote:
> > >> >> >>
> > >> >> >> Hi, Barry,
> > >> >> >>
> > >> >> >> Barry Song <21cnbao@...il.com> writes:
> > >> >> >>
> > >> >> >> > From: Barry Song <v-songbaohua@...o.com>
> > >> >> >> >
> > >> >> >> > Commit 13ddaf26be32 ("mm/swap: fix race when skipping swapcache")
> > >> >> >> > introduced an unconditional one-tick sleep when `swapcache_prepare()`
> > >> >> >> > fails, which has led to reports of UI stuttering on latency-sensitive
> > >> >> >> > Android devices. To address this, we can use a waitqueue to wake up
> > >> >> >> > tasks that fail `swapcache_prepare()` sooner, instead of always
> > >> >> >> > sleeping for a full tick. While tasks may occasionally be woken by an
> > >> >> >> > unrelated `do_swap_page()`, this method is preferable to two scenarios:
> > >> >> >> > rapid re-entry into page faults, which can cause livelocks, and
> > >> >> >> > multiple millisecond sleeps, which visibly degrade user experience.
> > >> >> >>
> > >> >> >> In general, I think that this works.  Why not extend the solution to
> > >> >> >> cover schedule_timeout_uninterruptible() in __read_swap_cache_async()
> > >> >> >> too?  We can call wake_up() when we clear SWAP_HAS_CACHE.  To avoid
> > >> >> >
> > >> >> > Hi Ying,
> > >> >> > Thanks for your comments.
> > >> >> > I feel extending the solution to __read_swap_cache_async() should be done
> > >> >> > in a separate patch. On phones, I've never encountered any issues reported
> > >> >> > on that path, so it might be better suited for an optimization rather than a
> > >> >> > hotfix?
> > >> >>
> > >> >> Yes.  It's fine to do that in another patch as optimization.
> > >> >
> > >> > Ok. I'll prepare a separate patch for optimizing that path.
> > >>
> > >> Thanks!
> > >>
> > >> >>
> > >> >> >> overhead to call wake_up() when there's no task waiting, we can use an
> > >> >> >> atomic to count waiting tasks.
> > >> >> >
> > >> >> > I'm not sure it's worth adding the complexity, as wake_up() on an empty
> > >> >> > waitqueue should have a very low cost on its own?
> > >> >>
> > >> >> wake_up() needs to call spin_lock_irqsave() unconditionally on a global
> > >> >> shared lock.  On systems with many CPUs (such servers), this may cause
> > >> >> severe lock contention.  Even the cache ping-pong may hurt performance
> > >> >> much.
> > >> >
> > >> > I understand that cache synchronization was a significant issue before
> > >> > qspinlock, but it seems to be less of a concern after its implementation.
> > >>
> > >> Unfortunately, qspinlock cannot eliminate cache ping-pong issue, as
> > >> discussed in the following thread.
> > >>
> > >> https://lore.kernel.org/lkml/20220510192708.GQ76023@worktop.programming.kicks-ass.net/
> > >>
> > >> > However, using a global atomic variable would still trigger cache broadcasts,
> > >> > correct?
> > >>
> > >> We can only change the atomic variable to non-zero when
> > >> swapcache_prepare() returns non-zero, and call wake_up() when the atomic
> > >> variable is non-zero.  Because swapcache_prepare() returns 0 most times,
> > >> the atomic variable is 0 most times.  If we don't change the value of
> > >> atomic variable, cache ping-pong will not be triggered.
> > >
> > > yes. this can be implemented by adding another atomic variable.
> >
> > Just realized that we don't need another atomic variable for this, just
> > use waitqueue_active() before wake_up() should be enough.
> >
> > >>
> > >> Hi, Kairui,
> > >>
> > >> Do you have some test cases to test parallel zram swap-in?  If so, that
> > >> can be used to verify whether cache ping-pong is an issue and whether it
> > >> can be fixed via a global atomic variable.
> > >>
> > >
> > > Yes, Kairui please run a test on your machine with lots of cores before
> > > and after adding a global atomic variable as suggested by Ying. I am
> > > sorry I don't have a server machine.
> > >
> > > if it turns out you find cache ping-pong can be an issue, another
> > > approach would be a waitqueue hash:
> >
> > Yes.  waitqueue hash may help reduce lock contention.  And, we can have
> > both waitqueue_active() and waitqueue hash if necessary.  As the first
> > step, waitqueue_active() appears simpler.
>
> Interesting. Just take a look at the waitqueue_active(), it requires
> smp_mb() if using without holding the lock.
> Quote from the comment of waitqueue_active():
> * Also note that this 'optimization' trades a spin_lock() for an smp_mb(),
>  * which (when the lock is uncontended) are of roughly equal cost.
>

probably not a problem in our case. two reasons:
1. we don't have a condition here
2. false postive/negative wake_up() won't cause a problem here.

We used to always sleep at least 4ms for an embedded system, if we can
kill 99% of the possibilities, it is all good.

Ideally, we could combine wait queue hash and wakeup_active(), but
Kairui's test shows even if we did neither of the above, it is still acceptable
in performance. so probably we can make things simple by just
adding a if(waitqueue_active()) before wake_up().

> Chris
>
> >
> > > diff --git a/mm/memory.c b/mm/memory.c
> > > index 2366578015ad..aae0e532d8b6 100644
> > > --- a/mm/memory.c
> > > +++ b/mm/memory.c
> > > @@ -4192,6 +4192,23 @@ static struct folio *alloc_swap_folio(struct vm_fault *vmf)
> > >  }
> > >  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> > >
> > > +/*
> > > + * Alleviating the 'thundering herd' phenomenon using a waitqueue hash
> > > + * when multiple do_swap_page() operations occur simultaneously.
> > > + */
> > > +#define SWAPCACHE_WAIT_TABLE_BITS 5
> > > +#define SWAPCACHE_WAIT_TABLE_SIZE (1 << SWAPCACHE_WAIT_TABLE_BITS)
> > > +static wait_queue_head_t swapcache_wqs[SWAPCACHE_WAIT_TABLE_SIZE];
> > > +
> > > +static int __init swapcache_wqs_init(void)
> > > +{
> > > +     for (int i = 0; i < SWAPCACHE_WAIT_TABLE_SIZE; i++)
> > > +             init_waitqueue_head(&swapcache_wqs[i]);
> > > +
> > > +        return 0;
> > > +}
> > > +late_initcall(swapcache_wqs_init);
> > > +
> > >  /*
> > >   * We enter with non-exclusive mmap_lock (to exclude vma changes,
> > >   * but allow concurrent faults), and pte mapped but not yet locked.
> > > @@ -4204,6 +4221,8 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > >  {
> > >       struct vm_area_struct *vma = vmf->vma;
> > >       struct folio *swapcache, *folio = NULL;
> > > +     DECLARE_WAITQUEUE(wait, current);
> > > +     wait_queue_head_t *swapcache_wq;
> > >       struct page *page;
> > >       struct swap_info_struct *si = NULL;
> > >       rmap_t rmap_flags = RMAP_NONE;
> > > @@ -4297,12 +4316,16 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > >                                * undetectable as pte_same() returns true due
> > >                                * to entry reuse.
> > >                                */
> > > +                             swapcache_wq = &swapcache_wqs[hash_long(vmf->address & PMD_MASK,
> > > +                                                     SWAPCACHE_WAIT_TABLE_BITS)];
> > >                               if (swapcache_prepare(entry, nr_pages)) {
> > >                                       /*
> > >                                        * Relax a bit to prevent rapid
> > >                                        * repeated page faults.
> > >                                        */
> > > +                                     add_wait_queue(swapcache_wq, &wait);
> > >                                       schedule_timeout_uninterruptible(1);
> > > +                                     remove_wait_queue(swapcache_wq, &wait);
> > >                                       goto out_page;
> > >                               }
> > >                               need_clear_cache = true;
> > > @@ -4609,8 +4632,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > >               pte_unmap_unlock(vmf->pte, vmf->ptl);
> > >  out:
> > >       /* Clear the swap cache pin for direct swapin after PTL unlock */
> > > -     if (need_clear_cache)
> > > +     if (need_clear_cache) {
> > >               swapcache_clear(si, entry, nr_pages);
> > > +             wake_up(swapcache_wq);
> > > +     }
> > >       if (si)
> > >               put_swap_device(si);
> > >       return ret;
> > > @@ -4625,8 +4650,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> > >               folio_unlock(swapcache);
> > >               folio_put(swapcache);
> > >       }
> > > -     if (need_clear_cache)
> > > +     if (need_clear_cache) {
> > >               swapcache_clear(si, entry, nr_pages);
> > > +             wake_up(swapcache_wq);
> > > +     }
> > >       if (si)
> > >               put_swap_device(si);
> > >       return ret;
> >
> > --
> > Best Regards,
> > Huang, Ying

Thanks
Barry

Powered by blists - more mailing lists