Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.35 as permitted sender) client-ip=23.128.96.35;
MIME-Version: 1.0
References: <20231010032833.398033-1-robinlai@tencent.com> <ZSTLxuiJssT9aYb0@casper.infradead.org>
In-Reply-To: <ZSTLxuiJssT9aYb0@casper.infradead.org>
From:   lai bin <sclaibin@gmail.com>
Date:   Tue, 10 Oct 2023 15:04:04 +0800
Message-ID: <CAE82Ze2HSJcTp4aNzMne877L+kyCwG3WXTLMdSwJ+HqsfbnHSw@mail.gmail.com>
Subject: Re: [PATCH] sched/wait: introduce endmark in __wake_up_common
To:     Matthew Wilcox <willy@infradead.org>
Cc:     mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, akpm@linux-foundation.org,
        dietmar.eggemann@arm.com, rostedt@goodmis.org, bsegall@google.com,
        mgorman@suse.de, bristot@redhat.com, vschneid@redhat.com,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-mm@kvack.org, Bin Lai <robinlai@tencent.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Yes,  you are right. Now that bookmarks weren't actually useful any
more, removing  bookmarks  is a better way to address the livelocks.

Matthew Wilcox <willy@infradead.org> =E4=BA=8E2023=E5=B9=B410=E6=9C=8810=E6=
=97=A5=E5=91=A8=E4=BA=8C 11:58=E5=86=99=E9=81=93=EF=BC=9A
>
> On Tue, Oct 10, 2023 at 11:28:33AM +0800, Bin Lai wrote:
> > Without this patch applied, it can cause the waker to fall into an
> > infinite loop in some cases. The commit 2554db916586 ("sched/wait: Brea=
k
> > up long wake list walk") introduces WQ_FLAG_BOOKMARK to break up long
> > wake list walk. When the number of walked entries reach 64, the waker
> > will record scan position and release the queue lock, which reduces
> > interrupts and rescheduling latency.
>
> Maybe we could try removing bookmarks instead?  There was a thread a
> while ago where we agreed they weren't actually useful any more.
>
> Patches to follow ...