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;
Message-ID: <e08a08dd-6cfe-1550-5d13-b76eef38bb27@igalia.com>
Date:   Fri, 14 Oct 2022 21:19:11 -0300
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.3.2
Subject: Re: [PATCH V2] x86/split_lock: Add sysctl to control the misery mode
Content-Language: en-US
To:     "Luck, Tony" <tony.luck@intel.com>,
        "x86@kernel.org" <x86@kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Cc:     "tglx@linutronix.de" <tglx@linutronix.de>,
        "mingo@redhat.com" <mingo@redhat.com>,
        "bp@alien8.de" <bp@alien8.de>,
        "dave.hansen@linux.intel.com" <dave.hansen@linux.intel.com>,
        "hpa@zytor.com" <hpa@zytor.com>,
        "Lutomirski, Andy" <luto@kernel.org>,
        "corbet@lwn.net" <corbet@lwn.net>,
        "linux-doc@vger.kernel.org" <linux-doc@vger.kernel.org>,
        "kernel-dev@igalia.com" <kernel-dev@igalia.com>,
        "kernel@gpiccoli.net" <kernel@gpiccoli.net>,
        Andre Almeida <andrealmeid@igalia.com>,
        "Yu, Fenghua" <fenghua.yu@intel.com>,
        Joshua Ashton <joshua@froggi.es>,
        Melissa Wen <mwen@igalia.com>,
        Paul Gofman <pgofman@codeweavers.com>,
        Pavel Machek <pavel@denx.de>,
        Pierre-Loup Griffais <pgriffais@valvesoftware.com>,
        Zebediah Figura <zfigura@codeweavers.com>
References: <20221014180506.211592-1-gpiccoli@igalia.com>
 <SJ1PR11MB60832858BA290E844F6C13E9FC249@SJ1PR11MB6083.namprd11.prod.outlook.com>
From:   "Guilherme G. Piccoli" <gpiccoli@igalia.com>
In-Reply-To: <SJ1PR11MB60832858BA290E844F6C13E9FC249@SJ1PR11MB6083.namprd11.prod.outlook.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 14/10/2022 15:26, Luck, Tony wrote:
> Looks reasonable.
> 
> Are these games multi-threaded with split locks happening on multiple CPUs in parallel?
> If they are, then skipping both the 10ms delay and the serialization is needed.
> 
> But if split locks are only from one CPU at a time, then possibly it would have
> been enough to just have this mitigation skip the:
> 
> +		if (msleep_interruptible(10) > 0)
> +			return;
> 
> Maybe best not to second guess. You have left the default as "mitigation on",
> so I'm happy.
> 
> -Tony

Hi Tony, thanks for your review!

Some games are indeed multi-threaded, so as you said, I think it's
better if we skip the whole thing when the sysctl is off - a kind of
fallback to the old "warn" mode before the misery was added heh

Cheers,


Guilherme