Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <32423b7c0e3a490093ceaca750e8669ac67902c6.camel@sipsolutions.net>
Subject: Re: UML time-travel warning from __run_timers
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Thomas Gleixner <tglx@linutronix.de>,
        Vincent Whitchurch <vincent.whitchurch@axis.com>
Cc:     linux-um@lists.infradead.org, linux-kernel@vger.kernel.org,
        Anna-Maria Gleixner <anna-maria@linutronix.de>,
        Frederic Weisbecker <frederic@kernel.org>
Date:   Sun, 03 Apr 2022 19:13:19 +0200
In-Reply-To: <877d86m978.ffs@tglx>
References: <20220330110156.GA9250@axis.com>
         <84f9d627092660c38400b607198c3b83f795be7f.camel@sipsolutions.net>
         <877d86m978.ffs@tglx>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.42.4 (3.42.4-1.fc35) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Sun, 2022-04-03 at 18:18 +0200, Thomas Gleixner wrote:
> On Sat, Apr 02 2022 at 16:09, Johannes Berg wrote:
> > At init, we get
> > 
> > init_timer_cpu(0) base 0 clk=0xffff8ad0, next_expiry=0x13fff8acf
> > init_timer_cpu(0) base 1 clk=0xffff8ad0, next_expiry=0x13fff8acf
> > 
> > which makes sense, jiffies is set up to wrap very quickly after boot.
> > 
> > The warning triggers when we have jiffies=0x13fff9600, so it's just
> > after the "next_expiry", so in this code:
> 
> which does not make sense.

If you say so, I have no idea :)

> If next_expiry is 0x13fff8acf and jiffies
> advanced to 0x13fff9600 when the warning triggered, then either it
> missed to expire the timer at 0x13fff8acf or it failed to recalculate
> next_expiry.

There was no timer. If there's ever a timer on this base (BASE_DEF) then
this doesn't happen.

So it has to be the latter, but I'm trying to understand in the code
where it would*ever* recalculate next_expiry if it in fact never
expires?

> Could you enable all [hr]timer tracepoints on the kernel command line,
> set panic on warn and ftrace_dump_on_oops which should spill out the
> tracebuffer on the console and provide the output. That should at least
> give us an hint what's going on.
> 

Sure, but since we simulate ~50 days of uptime, that's a massive number
of events and lots of them are not there:

https://p.sipsolutions.net/fb491cbbde82c600.txt

johannes