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: <0994ea3a-e197-e938-6eab-72433d6547fd@yandex.ru>
Date:   Wed, 29 Mar 2023 22:43:38 +0500
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [PATCH 6.0.y / 6.1.y] x86/split_lock: Add sysctl to control the
 misery mode
Content-Language: en-US
To:     Sean Christopherson <seanjc@google.com>,
        "Guilherme G. Piccoli" <gpiccoli@igalia.com>
Cc:     Peter Zijlstra <peterz@infradead.org>, Greg KH <greg@kroah.com>,
        stable@vger.kernel.org, x86@kernel.org,
        linux-kernel@vger.kernel.org, corbet@lwn.net,
        kernel-dev@igalia.com, kernel@gpiccoli.net, fenghua.yu@intel.com,
        joshua@froggi.es, pgofman@codeweavers.com, pavel@denx.de,
        pgriffais@valvesoftware.com, zfigura@codeweavers.com,
        cristian.ciocaltea@collabora.com,
        Thomas Gleixner <tglx@linutronix.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Tony Luck <tony.luck@intel.com>,
        Andre Almeida <andrealmeid@igalia.com>
References: <20221218234400.795055-1-gpiccoli@igalia.com>
 <Y6A6Q57/qz7w7cxM@kroah.com>
 <Y6BD9W7hk4CjhSdh@hirez.programming.kicks-ass.net>
 <3ff9f56c-479b-2dbd-9ee6-c7d00c7bd285@igalia.com>
 <ZCRx8YaAt7ybDlLM@google.com>
From:   stsp <stsp2@yandex.ru>
In-Reply-To: <ZCRx8YaAt7ybDlLM@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

Hi Sean, thanks for a head-up!

29.03.2023 22:14, Sean Christopherson пишет:
> Resurrecting this thread with a concrete use case.
> dosemu2, which uses KVM to accelerate DOS emulation (stating the obvious), ran
> into a problem where the hardcoded (prior to this patch) behavior will effectively
> hang userspace if the 10ms sleep is interrupted, e.g. by a periodic SIGALRM[*].

Yes, and we also created a ready-to-use
host test-case with no dosemu2 involved.

> Alternatively, we could try to figure out a way to ensure forward progress without
> letting userpace run an end-around on the enforced misery, but backporting this
> patch to stable kernels seems easier.
>
> Stas, do you happen to know what the oldest stable kernel your users use, i.e.
> how far back this backport would need to go?
It seems, the "broken" code was added by
the patch b041b525dab95 which is "described"
as v5.18-rc4-1-gb041b525dab9.
So I guess the answer would be "down to 5.18".

Of course I don't believe the patch you
mention, is a real solution. Its good for
-stable, but something else needs to be
developed in the future.