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:   Wed, 21 Jun 2023 16:54:36 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Waiman Long <longman@redhat.com>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Josh Poimboeuf <jpoimboe@kernel.org>,
        Pawan Gupta <pawan.kumar.gupta@linux.intel.com>,
        Jacob Pan <jacob.jun.pan@linux.intel.com>,
        Len Brown <lenb@kernel.org>, linux-kernel@vger.kernel.org,
        x86@kernel.org, linux-pm@vger.kernel.org,
        Robin Jarry <rjarry@redhat.com>, Joe Mario <jmario@redhat.com>
Subject: Re: [PATCH v2 2/5] x86/idle: Disable IBRS when cpu is offline
Message-ID: <20230621145436.GK2053369@hirez.programming.kicks-ass.net>
References: <20230620140625.1001886-1-longman@redhat.com>
 <20230620140625.1001886-3-longman@redhat.com>
 <20230621072313.GA2046280@hirez.programming.kicks-ass.net>
 <7f2424df-1846-6c38-e446-b3d5aa693ecd@redhat.com>
 <20230621143602.GI2053369@hirez.programming.kicks-ass.net>
 <5cb81f3b-45a0-d566-3d63-569b5706e9fe@redhat.com>
 <20230621144848.GJ2053369@hirez.programming.kicks-ass.net>
 <309d15f5-0dd8-aee8-14a6-621a071bc363@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <309d15f5-0dd8-aee8-14a6-621a071bc363@redhat.com>
Precedence: bulk

On Wed, Jun 21, 2023 at 10:51:33AM -0400, Waiman Long wrote:
> 
> On 6/21/23 10:48, Peter Zijlstra wrote:
> > On Wed, Jun 21, 2023 at 10:44:23AM -0400, Waiman Long wrote:
> > 
> > > Well, hlt_play_dead() is only called if cpuidle_play_dead() returns an error
> > > which is not the typical case. My testing does confirm that this patch is
> > > able to keep the IBRS bit off when a CPU is offline via its online sysfs
> > > file.
> > The point is; your re-enable IBRS hunk at the end is dead-code. It
> > should never ever run and having it is confusing.
> 
> What I meant is that hlt_play_dead() should never be called unless there is
> some serious problem with the system and native_play_dead() does return in
> normal usage.

This is all through arch_cpu_idle_dead() which is __noreturn. And no,
none of this must ever return.

If you want an offline CPU to come back, you re-init.