Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <CAHk-=whewL14RgwLZTXcNAnrDPt0H+sRJS6iDq0oGb6zwaBMxg@mail.gmail.com>
 <CAHk-=whb0=rjc1WR+F_r_syw5Ld4=ebuNJmmpaPEzfjZRD5Y-w@mail.gmail.com>
 <alpine.LSU.2.11.2007221359450.1017@eggly.anvils> <CAHk-=wi=vuc6sdu0m9nYd3gb8x5Xgnc6=TH=DTOy7qU96rZ9nw@mail.gmail.com>
 <CAHk-=whEjnsANEhTA3aqpNLZ3vv7huP7QAmcAEd-GUxm2YMo-Q@mail.gmail.com>
 <20200723124749.GA7428@redhat.com> <CAHk-=wgyc7en4=HddEYiz_RKJXfqe1JYv3BzHc=+_wYq9ti+LQ@mail.gmail.com>
 <CAHk-=whQK3OGwExTzCrwwvuuVaQAgs8KsR-Yv8m1BmXoNZZ=jQ@mail.gmail.com>
 <alpine.LSU.2.11.2007231549540.1016@eggly.anvils> <CAHk-=wgvGOnMF0ePU4xS236bOsP8jouj3rps+ysCaGXvCjh2Dg@mail.gmail.com>
 <20200724152424.GC17209@redhat.com> <CAHk-=whuG+5pUeUqdiW4gk0prvqu7GZSMo-6oWv5PdDC5dBr=A@mail.gmail.com>
 <CAHk-=wjYHvbOs9i39EnUsC6VEJiuJ2e_5gZB5-J5CRKxq80B_Q@mail.gmail.com> <alpine.LSU.2.11.2007241848300.8192@eggly.anvils>
In-Reply-To: <alpine.LSU.2.11.2007241848300.8192@eggly.anvils>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri, 24 Jul 2020 19:46:21 -0700
Message-ID: <CAHk-=wj1qMgNLMcSuA7EFffaqFVDUzNooMSoaon2094qbwCakQ@mail.gmail.com>
Subject: Re: [RFC PATCH] mm: silence soft lockups from unlock_page
To:     Hugh Dickins <hughd@google.com>
Cc:     Oleg Nesterov <oleg@redhat.com>, Michal Hocko <mhocko@kernel.org>,
        Linux-MM <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Tim Chen <tim.c.chen@linux.intel.com>,
        Michal Hocko <mhocko@suse.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Jul 24, 2020 at 7:08 PM Hugh Dickins <hughd@google.com> wrote:
>
> But whatever, what happens on the next run, with these latest patches,
> will be more important; and I'll follow this next run with a run on
> the baseline without them, to compare results.

So the loads you are running are known to have sensitivity to this
particular area, and are why you've done your patches to the page wait
bit code?

Because yes, I think that last version in particular might make a big
difference with the "let people continue even if they only saw the
wakeup event, and never actually tested and saw the bit clear".

Of course, there's a possibility that "big difference" ends up being a
negative one. I think it will make the page wait queues shorter (which
is good for that latency and lockup thing), but waking things up more
aggressively _may_ also end up adding more CPU load, if they then all
decide to retry the operation for whatever reason.

And hey, it's also possible that it makes no difference at all,
because your load mainly tests the exclusive "lock_page()" case, which
won't have changed.

And that's all assuming they don't show some instability, of course.
But the code actually seems fairly simple now, and the basic
synchronization hasn't changed, just a behavioral optimization.

Famous last words.

               Linus