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, 3 May 2022 11:03:05 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     LKML <linux-kernel@vger.kernel.org>, x86@kernel.org,
        Filipe Manana <fdmanana@suse.com>
Subject: Re: [patch 3/3] x86/fpu: Make FPU protection more robust
Message-ID: <YnDvye07/jRMBiPx@hirez.programming.kicks-ass.net>
References: <20220501192740.203963477@linutronix.de>
 <20220501193102.704267030@linutronix.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220501193102.704267030@linutronix.de>
Precedence: bulk

On Sun, May 01, 2022 at 09:31:47PM +0200, Thomas Gleixner wrote:
> +void fpregs_lock(void)
> +{
> +	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
> +		local_bh_disable();
> +	else
> +		preempt_disable();
> +
> +	WARN_ON_ONCE(this_cpu_read(fpu_in_use));
> +	this_cpu_write(fpu_in_use, true);

	barrier();
> +}
> +EXPORT_SYMBOL_GPL(fpregs_lock);

> +void fpregs_unlock(void)
> +{
	barrier();

> +	WARN_ON_ONCE(!this_cpu_read(fpu_in_use));
> +	this_cpu_write(fpu_in_use, false);
> +
> +	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
> +		local_bh_enable();
> +	else
> +		preempt_enable();
> +}
> +EXPORT_SYMBOL_GPL(fpregs_unlock);

I think this isn't currently a problem because a function call is a C
sequence point, but 'funnily' C doesn't preserve sequence points when
inlining so LTO can actually break this without barrier() on.