Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230328221644.803272-5-yosryahmed@google.com>
 <ZCQfZJFufkJ10o01@dhcp22.suse.cz> <CAJD7tkb-UpKm2QbjYzB=B=oGk6Hyj9cbUviZUPC+7VsvBecH7g@mail.gmail.com>
 <20230329192059.2nlme5ubshzdbpg6@google.com> <ZCU1Bp+5bKNJzWIu@dhcp22.suse.cz>
 <CAJD7tka0CmRvcvB0k8DZuid1vC9OK_mFriHHbXNTUkVE7OjaTA@mail.gmail.com>
 <ZCU+8lSi+e4WgT3F@dhcp22.suse.cz> <CAJD7tkaKd9Bcb2-e83Q-kzF7G+crr1U+7uqUPBARXWq-LpyKvw@mail.gmail.com>
 <ZCVFA78lDj2/Uy0C@dhcp22.suse.cz> <CAJD7tkbjmBaXghQ+14Hy28r2LoWSim+LEjOPxaamYeA_kr2uVw@mail.gmail.com>
 <ZCVKqN2nDkkQFvO0@dhcp22.suse.cz>
In-Reply-To: <ZCVKqN2nDkkQFvO0@dhcp22.suse.cz>
From:   Yosry Ahmed <yosryahmed@google.com>
Date:   Thu, 30 Mar 2023 01:53:38 -0700
Message-ID: <CAJD7tkYEOVRcXs-Ag3mWn69EwE4rjFt9j5MAcTGCNE8BuhTd+A@mail.gmail.com>
Subject: Re: [PATCH v2 4/9] cgroup: rstat: add WARN_ON_ONCE() if flushing
 outside task context
To:     Michal Hocko <mhocko@suse.com>
Cc:     Johannes Weiner <hannes@cmpxchg.org>,
        Shakeel Butt <shakeelb@google.com>, Tejun Heo <tj@kernel.org>,
        Josef Bacik <josef@toxicpanda.com>,
        Jens Axboe <axboe@kernel.dk>,
        Zefan Li <lizefan.x@bytedance.com>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Muchun Song <muchun.song@linux.dev>,
        Andrew Morton <akpm@linux-foundation.org>,
        =?UTF-8?Q?Michal_Koutn=C3=BD?= <mkoutny@suse.com>,
        Vasily Averin <vasily.averin@linux.dev>,
        cgroups@vger.kernel.org, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        bpf@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Mar 30, 2023 at 1:39=E2=80=AFAM Michal Hocko <mhocko@suse.com> wrot=
e:
>
> On Thu 30-03-23 01:19:29, Yosry Ahmed wrote:
> > On Thu, Mar 30, 2023 at 1:15=E2=80=AFAM Michal Hocko <mhocko@suse.com> =
wrote:
> > >
> > > On Thu 30-03-23 01:06:26, Yosry Ahmed wrote:
> > > [...]
> > > > If we achieve that, do you think it makes sense to add
> > > > WARN_ON_ONCE(irqs_disabled()) instead to prevent future users from
> > > > flushing while disabling irqs or in irq context?
> > >
> > > WARN_ON (similar to BUG_ON) will not prevent anybody from doing bad
> > > things. We already have means to shout about sleepable code being
> > > invoked from an atomic context and there is no reason to duplicate th=
at.
> > > As I've said earlier WARN_ON might panic the system in some
> > > configurations (and yes they are used also in production systems - do
> > > not ask me why...). So please be careful about that and use that only
> > > when something really bad (yet recoverable) is going on.
> >
> > Thanks for the information (I was about to ask why about production
> > systems, but okay..). I will avoid WARN_ON completely. For the
> > purposes of this series I will drop this patch anyway.
>
> Thanks! People do strange things sometimes...
>
> > Any idea how to shout about "hey this may take too long, why are you
> > doing it with irqs disabled?!"?
>
> Well we have a hard lockup detector. It hits at a much higher stall by
> default but if you care about IRQ latencies in general then you likely
> want to lower. Another thing would be IRQ tracing. In any case this code
> path shouldn't be any special. Sure it can take long on large systems
> but I bet there are more of those.

We did see hard lockups in extreme cases, and we do have setups with
"nmi_watchdog=3Dpanic" that will panic when this happens, so we would
rather catch the code paths that can lead to that before it actually
happens.

Maybe we can add a primitive like might_sleep() for this, just food for tho=
ught.

> --
> Michal Hocko
> SUSE Labs