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;
Date:   Thu, 10 Feb 2022 13:33:39 +0100
From:   Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To:     Christoph Hellwig <hch@infradead.org>
Cc:     linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH REPOST] irq_poll: Add local_bh_disable() in cpu_dead
 notifier
Message-ID: <YgUGI9qAKUh4AOUY@linutronix.de>
References: <YgJ/XWVxxWDVBBVA@linutronix.de>
 <YgNzsnIE9bwQZ1Zg@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
In-Reply-To: <YgNzsnIE9bwQZ1Zg@infradead.org>
Precedence: bulk

On 2022-02-08 23:56:34 [-0800], Christoph Hellwig wrote:
> On Tue, Feb 08, 2022 at 03:34:05PM +0100, Sebastian Andrzej Siewior wrote:
> > __raise_softirq_irqoff() adds a bit to the pending sofirq mask and this
> > is it. The softirq won't be handled in a deterministic way but randomly
> > when an interrupt fires and handles the softirq in its irq_exit() routi=
ne or
> > if something randomly checks and handles pending softirqs in the call
> > chain before the CPU goes idle.
> >=20
> > Add a local_bh_disable/enable() around the IRQ-off section which will
> > handle pending softirqs.
>=20
> And I still haven't seen any good explanation of why this is useful.

You need to handle the pending softirqs. If you don't handle them
immediately or in a deterministic say (like on IRQ exit) then they will
be handled at a random point. If you don't handle them at all, the CPU
will go idle and at least the NO_HZ will complain about pending softirqs
(can_stop_idle_tick()).

You could still argue that the CPU will go down and the there are
latencies involved but=E2=80=A6
I want to avoid waking ksoftirqd for that since there is no need to wake
it and the pending work can be done in-context, right away.

Sebastian