Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
Date:   Tue, 10 Oct 2023 16:12:44 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Phil Auld <pauld@redhat.com>
Cc:     Chen Yu <yu.c.chen@intel.com>, Ingo Molnar <mingo@redhat.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Biju Das <biju.das.jz@bp.renesas.com>,
        Marek Szyprowski <m.szyprowski@samsung.com>,
        linux-kernel@vger.kernel.org, Tim Chen <tim.c.chen@intel.com>,
        Chen Yu <yu.chen.surf@gmail.com>
Subject: Re: [PATCH] sched/fair: Use printk_deferred instead of printk in
 pick_eevdf()
Message-ID: <20231010141244.GM377@noisy.programming.kicks-ass.net>
References: <20231010032541.339606-1-yu.c.chen@intel.com>
 <20231010075928.GA377@noisy.programming.kicks-ass.net>
 <20231010122600.GA477540@lorien.usersys.redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231010122600.GA477540@lorien.usersys.redhat.com>
Precedence: bulk

On Tue, Oct 10, 2023 at 08:26:00AM -0400, Phil Auld wrote:

> > No.. I detest printk_deferred with a passion. This is effectively a WARN
> > and we don't do silly buggers for them either.
> >
> 
> Sure, printk_deferred is not ideal, but is getting this message in the right
> order worth locking up people's machines?  Not sure you get the message at
> all when that happens.  I have to dig the code location out of the crash
> dump to find which sched warning fired and took down the (usually virtual)
> machine.

Same thing with WARN(), we don't have a silly bugger version of that
either. Just use a sane printk() / console or whatever.

Virt stuff has perfectly functional serial consoles that works just fine
and don't lock up the machine -- mostly. Use my early_printk hacks if
you want something reliable:

  https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=debug/experimental

Boot with: "earlyprintk=serial,ttyS0,115200 force_early_printk". And all
will be well.

The fact that crashdump is more reliable than printk is a *BIG* problem
and the only solution is fixing printk() (people are sorta working on
that). We should not try and work around this problem.

I fundamentally despise the delayed stuff, I've had countless insteances
where delaying output means you have no output because the machine is
dead.