MIME-Version: 1.0
References: <20211124151915.GA6163@haolee.io> <YZ5hBtWPBpHDWzE4@casper.infradead.org>
In-Reply-To: <YZ5hBtWPBpHDWzE4@casper.infradead.org>
From:   Hao Lee <haolee.swjtu@gmail.com>
Date:   Thu, 25 Nov 2021 11:13:54 +0800
Message-ID: <CA+PpKPm4nfFM1XNifQoQT6Q_f3hQ6gSNnCgUumLDnruaLaUgvQ@mail.gmail.com>
Subject: Re: [PATCH] mm: reduce spinlock contention in release_pages()
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Linux MM <linux-mm@kvack.org>,
        Johannes Weiner <hannes@cmpxchg.org>, mhocko@kernel.org,
        vdavydov.dev@gmail.com, Shakeel Butt <shakeelb@google.com>,
        cgroups@vger.kernel.org, LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Nov 24, 2021 at 11:58 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Wed, Nov 24, 2021 at 03:19:15PM +0000, Hao Lee wrote:
> > When several tasks are terminated simultaneously, lots of pages will be
> > released, which can cause severe spinlock contention. Other tasks which
> > are running on the same core will be seriously affected. We can yield
> > cpu to fix this problem.
>
> The realtime people will eat you alive for this suggestion.

Thanks for pointing out this.

>
> > +++ b/mm/swap.c
> > @@ -960,8 +960,14 @@ void release_pages(struct page **pages, int nr)
> >               if (PageLRU(page)) {
> >                       struct lruvec *prev_lruvec = lruvec;
> >
> > -                     lruvec = folio_lruvec_relock_irqsave(folio, lruvec,
> > +retry:
> > +                     lruvec = folio_lruvec_tryrelock_irqsave(folio, lruvec,
> >                                                                       &flags);
> > +                     if (!lruvec) {
> > +                             cond_resched();
> > +                             goto retry;
> > +                     }
> > +
> >                       if (prev_lruvec != lruvec)