MIME-Version: 1.0
References: <1638780405-38026-1-git-send-email-quic_pragalla@quicinc.com>
 <CAJfpegvDfc9JUo6VASRyYXzj1=j3t6oU9W3QGWO08vhfWHf-UA@mail.gmail.com>
 <Ya8ycLODlcvLx4xB@hirez.programming.kicks-ass.net> <CAJfpegsVg2K0CvrPvXGSu=Jz_R3VZZOy49Jw51rThQUJ1_9e6g@mail.gmail.com>
 <Ya86coKm4RuQDmVS@hirez.programming.kicks-ass.net> <CAJfpegumZ1RQLBCtbrOiOAT9ygDtDThpySwb8yCpWGBu1fRQmw@mail.gmail.com>
 <Ya9ljdrOkhBhhnJX@hirez.programming.kicks-ass.net> <Ya9m0ME1pom49b+D@hirez.programming.kicks-ass.net>
In-Reply-To: <Ya9m0ME1pom49b+D@hirez.programming.kicks-ass.net>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Tue, 7 Dec 2021 15:03:01 +0100
Message-ID: <CAJfpegt2x1ztuzh0niY7fgx1UKxDGsAkJbS0wVPp5awxwyhRpA@mail.gmail.com>
Subject: Re: [PATCH V1] fuse: give wakeup hints to the scheduler
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Ingo Molnar <mingo@redhat.com>, quic_stummala@quicinc.com,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        quic_pkondeti@quicinc.com, quic_sayalil@quicinc.com,
        quic_aiquny@quicinc.com, quic_zljing@quicinc.com,
        quic_blong@quicinc.com, quic_richardp@quicinc.com,
        quic_cdevired@quicinc.com,
        Pradeep P V K <quic_pragalla@quicinc.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, 7 Dec 2021 at 14:51, Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Tue, Dec 07, 2021 at 02:45:49PM +0100, Peter Zijlstra wrote:
>
> > > What would be much nicer, is to look at all the threads on the waitq
> > > and pick one that previously ran on the current CPU if there's one.
> > > Could this be implemented?
> >
> > It would violate the FIFO semantics of _exclusive.
>
> That said, look at
> kernel/locking/percpu-rwsem.c:percpu_rwsem_wake_function() for how to do
> really terrible things with waitqueues, possibly including what you
> suggest.

Okay, so it looks doable, but rather more involved than just sticking
that _sync onto the wake helper.

FIFO is used so that we always wake the most recently used thread, right?

That makes sense if it doesn't involve migration, but if the hot
thread is going to be moved to another CPU then we'd lost most of the
advantages.  Am I missing something?

Thanks,
Miklos