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;
From:   Thomas Gleixner <tglx@linutronix.de>
To:     Paolo Abeni <pabeni@redhat.com>, linux-kernel@vger.kernel.org
Cc:     "Paul E. McKenney" <paulmck@kernel.org>, peterz@infradead.org,
        netdev@vger.kernel.org, Jakub Kicinski <kuba@kernel.org>,
        Jason Xing <kerneljasonxing@gmail.com>,
        Eric Dumazet <edumazet@google.com>
Subject: Re: [PATCH] revert: "softirq: Let ksoftirqd do its job"
In-Reply-To: <57e66b364f1b6f09c9bc0316742c3b14f4ce83bd.1683526542.git.pabeni@redhat.com>
Date:   Mon, 08 May 2023 23:21:14 +0200
Message-ID: <87bkiu34fp.ffs@tglx>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Paolo!

On Mon, May 08 2023 at 08:17, Paolo Abeni wrote:
> Due to the mentioned commit, when the ksoftirqd processes take charge
> of softirq processing, the system can experience high latencies.
>
> In the past a few workarounds have been implemented for specific
> side-effects of the above:
>
> commit 1ff688209e2e ("watchdog: core: make sure the watchdog_worker is not deferred")
> commit 8d5755b3f77b ("watchdog: softdog: fire watchdog even if softirqs do not get to run")
> commit 217f69743681 ("net: busy-poll: allow preemption in sk_busy_loop()")
> commit 3c53776e29f8 ("Mark HI and TASKLET softirq synchronous")
>
> but the latency problem still exists in real-life workloads, see the
> link below.
>
> The reverted commit intended to solve a live-lock scenario that can now
> be addressed with the NAPI threaded mode, introduced with commit
> 29863d41bb6e ("net: implement threaded-able napi poll loop support"),
> and nowadays in a pretty stable status.
>
> While a complete solution to put softirq processing under nice resource
> control would be preferable, that has proven to be a very hard task. In
> the short term, remove the main pain point, and also simplify a bit the
> current softirq implementation.
>
> Note that this change also reverts commit 3c53776e29f8 ("Mark HI and
> TASKLET softirq synchronous") and commit 1342d8080f61 ("softirq: Don't
> skip softirq execution when softirq thread is parking"), which are
> direct follow-ups of the feature commit. A single change is preferred to
> avoid known bad intermediate states introduced by a patch series
> reverting them individually.

I'm fine with this change, but I definitely want that to be
acked/reviewed by the other stakeholders in the networking arena.

Thanks,

        tglx