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:   Tue, 20 Dec 2022 20:38:39 +0800
From:   Baoquan He <bhe@redhat.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-kernel@vger.kernel.org, Thomas Gleixner <tglx@linutronix.de>,
        x86@kernel.org, kexec@lists.infradead.org, dgilbert@redhat.com
Subject: Re: [PATCH] Revert "x86/apic/x2apic: Implement IPI shorthands
 support"
Message-ID: <Y6Gsz4mqnZCtC9rn@MiWiFi-R3L-srv>
References: <Y6FJgh3fOYoF0wSV@MiWiFi-R3L-srv>
 <Y6GextAcO4OZig3Y@hirez.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y6GextAcO4OZig3Y@hirez.programming.kicks-ass.net>
Precedence: bulk

On 12/20/22 at 12:38pm, Peter Zijlstra wrote:
> On Tue, Dec 20, 2022 at 01:34:58PM +0800, Baoquan He wrote:
> > This reverts commit 43931d350f30c6cd8c2f498d54ef7d65750abc92.
> > 
> > On kvm guest with 4 cpus deployed, when adding 'nr_cpus=2' to normal
> > kernel's cmdline, and triggering crash to jump to kdump kernel, kdump
> > kernel will stably hang. Reverting commit 43931d350f30 ("x86/apic/x2apic:
> > Implement IPI shorthands support") can fix it.
> > 
> > The problem will disappear if removing 'nr_cpus=2' from normal kerne's
> > cmdline.
> 
> And the root cause for this is... ? Does the kvm x2apic emulation
> somehow get upset when we shorthand CPUs that haven't been initialized?

Thanks for checking.

I haven't figure out the root cause. I haven't read the apic code for
long time, and not familiar with the kvm code. So raise the issue to
upstream.

I can do testing if any suggestion.

Add our virt dev Dr. David Alan Gilbert to CC.