Re: [PATCH] x86/power: Fix 'nosmt' vs. hibernation triple fault during resume

[Peter Zijlstra]

On Tue, May 28, 2019 at 11:31:45PM +0200, Jiri Kosina wrote:

> diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c
> index a7d966964c6f..bde8ce1f6c6c 100644
> --- a/arch/x86/power/cpu.c
> +++ b/arch/x86/power/cpu.c
> @@ -299,9 +299,20 @@ int hibernate_resume_nonboot_cpu_disable(void)
>  	 * address in its instruction pointer may not be possible to resolve
>  	 * any more at that point (the page tables used by it previously may
>  	 * have been overwritten by hibernate image data).
> +	 *
> +	 * First, make sure that we wake up all the potentially disabled SMT
> +	 * threads which have been initially brought up and then put into
> +	 * mwait/cpuidle sleep.
> +	 * Those will be put to proper (not interfering with hibernation
> +	 * resume) sleep afterwards, and the resumed kernel will decide itself
> +	 * what to do with them.
>  	 */
>  	smp_ops.play_dead = resume_play_dead;

Oooh, teh yuck!, but this explains my confusion from the other thread.

> +	ret = cpuhp_smt_enable();
> +	if (ret)
> +		goto out;
>  	ret = disable_nonboot_cpus();
> +out:
>  	smp_ops.play_dead = play_dead;
>  	return ret;
>  }

I think you can avoid the goto like:

	ret = cpuhp_smt_enable();
	if (ret)
		return ret;

	smp_ops.play_dead = resume_play_dead;
	ret = disable_nonboot_cpus();
	smp_ops.play_dead = play_dead;
	return ret;

We don't need the play dead change to online CPUs.