Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:4 as permitted sender) client-ip=2620:137:e000::3:4;
Date:   Wed, 13 Sep 2023 18:17:49 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Rik van Riel <riel@surriel.com>
Cc:     linux-kernel@vger.kernel.org, kernel-team@meta.com,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Valentin Schneider <vschneid@redhat.com>,
        Juergen Gross <jgross@suse.com>
Subject: Re: [PATCH,RFC] smp,csd: throw an error if a CSD lock is stuck for
 too long
Message-ID: <20230913161749.GK692@noisy.programming.kicks-ass.net>
References: <20230821160409.663b8ba9@imladris.surriel.com>
 <20230913132251.GE22758@noisy.programming.kicks-ass.net>
 <2189326aaca37487b17eb1103830156ff1684c27.camel@surriel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <2189326aaca37487b17eb1103830156ff1684c27.camel@surriel.com>
Precedence: bulk

On Wed, Sep 13, 2023 at 10:33:51AM -0400, Rik van Riel wrote:
> On Wed, 2023-09-13 at 15:22 +0200, Peter Zijlstra wrote:
> > On Mon, Aug 21, 2023 at 04:04:09PM -0400, Rik van Riel wrote:
> > > 
> > > +???????/* How long since this CSD lock was stuck. */
> > > +???????ts_delta = ts2 - ts0;
> > > 
> > > +???????/*
> > > +??????? * If the CSD lock is still stuck after 5 minutes, it is
> > > unlikely
> > > +??????? * to become unstuck. Use a signed comparison to avoid
> > > triggering
> > > +??????? * on underflows when the TSC is out of sync between
> > > sockets.
> > > +??????? */
> > > +???????BUG_ON((s64)ts_delta > 300000000000LL);
> > > ????????if (cpu_cur_csd && csd != cpu_cur_csd) {
> > > ????????????????pr_alert("\tcsd: CSD lock (#%d) handling prior
> > > %pS(%ps) request.\n",
> > > ???????????????????????? *bug_id, READ_ONCE(per_cpu(cur_csd_func,
> > > cpux)),
> > 
> > How are you guys still seeing this? I thought the KVM APIC thing was
> > fixed a while ago?
> > 
> It's more fun than that. We're seeing this on bare metal.

Oh, 'fun' indeed, *groan*.

> Unfortunately, when a system gets wedged that way currently,
> it ends up being power cycled automatically, and we aren't
> getting crash dumps with clues on what causes the issue.
> 
> Doing a BUG_ON() + panic, followed by a kexec into the kdump
> kernel will hopefully give us some clues on what might be
> causing the issue.

I'm conflicted on the need to push such a debug patch upstream, otoh.
given the amount of debug code already in csd, why not.

But yeah, curious hear what comes out of this.