linux-kernel - Re: [PATCH] acpi_idle: use raw_safe_halt() from acpi_idle_play

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 27 Oct 2023 21:14:35 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     David Woodhouse <dwmw2@...radead.org>
Cc:     linux-acpi <linux-acpi@...r.kernel.org>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Len Brown <lenb@...nel.org>, Juergen Gross <jgross@...e.com>,
        xen-devel <xen-devel@...ts.xenproject.org>,
        Ingo Molnar <mingo@...hat.com>, Will Deacon <will@...nel.org>,
        Waiman Long <longman@...hat.com>,
        Boqun Feng <boqun.feng@...il.com>
Subject: Re: [PATCH] acpi_idle: use raw_safe_halt() from acpi_idle_play_dead()

On Fri, Oct 27, 2023 at 07:36:51PM +0100, David Woodhouse wrote:
> From: David Woodhouse <dwmw@...zon.co.uk>
> 
> Xen HVM guests were observed taking triple-faults when attempting to
> online a previously offlined vCPU.
> 
> Investigation showed that the fault was coming from a failing call
> to lockdep_assert_irqs_disabled(), in load_current_idt() which was
> too early in the CPU bringup to actually catch the exception and
> report the failure cleanly.
> 
> This was a false positive, caused by acpi_idle_play_dead() setting
> the per-cpu hardirqs_enabled flag by calling safe_halt(). Switch it
> to use raw_safe_halt() instead, which doesn't do so.
> 
> Signed-off-by: David Woodhouse <dwmw@...zon.co.uk>
> ---
> We might {also,instead} explicitly set the hardirqs_enabled flag to
> zero when bringing up an AP?

So I fixed up the idle paths the other day (see all that __cpuidle
stuff) but I've not yet gone through the whole hotplug thing :/

This seems right, at this point everything, including RCU is very much
gone, any instrumentation is undesired.

Acked-by: Peter Zijlstra (Intel) <peterz@...radead.org>

> 
>  drivers/acpi/processor_idle.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
> index 3a34a8c425fe..55437f5e0c3a 100644
> --- a/drivers/acpi/processor_idle.c
> +++ b/drivers/acpi/processor_idle.c
> @@ -592,7 +592,7 @@ static int acpi_idle_play_dead(struct cpuidle_device *dev, int index)
>  	while (1) {
>  
>  		if (cx->entry_method == ACPI_CSTATE_HALT)
> -			safe_halt();
> +			raw_safe_halt();
>  		else if (cx->entry_method == ACPI_CSTATE_SYSTEMIO) {
>  			io_idle(cx->address);
>  		} else
> -- 
> 2.41.0
> 
>