From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by smtp.lore.kernel.org (Postfix) with ESMTP id BD0EEC61DA3 for ; Tue, 21 Feb 2023 23:58:24 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id 3E35C6B0072; Tue, 21 Feb 2023 18:58:24 -0500 (EST) Received: by kanga.kvack.org (Postfix, from userid 40) id 393186B0073; Tue, 21 Feb 2023 18:58:24 -0500 (EST) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 25B286B0074; Tue, 21 Feb 2023 18:58:24 -0500 (EST) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0016.hostedemail.com [216.40.44.16]) by kanga.kvack.org (Postfix) with ESMTP id 15A176B0072 for ; Tue, 21 Feb 2023 18:58:24 -0500 (EST) Received: from smtpin20.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay09.hostedemail.com (Postfix) with ESMTP id DF88B80200 for ; Tue, 21 Feb 2023 23:58:23 +0000 (UTC) X-FDA: 80492965686.20.E93D74E Received: from out-59.mta1.migadu.com (out-59.mta1.migadu.com [95.215.58.59]) by imf12.hostedemail.com (Postfix) with ESMTP id E4A2040009 for ; Tue, 21 Feb 2023 23:58:21 +0000 (UTC) Authentication-Results: imf12.hostedemail.com; dkim=pass header.d=linux.dev header.s=key1 header.b=e3iOG7gV; dmarc=pass (policy=none) header.from=linux.dev; spf=pass (imf12.hostedemail.com: domain of roman.gushchin@linux.dev designates 95.215.58.59 as permitted sender) smtp.mailfrom=roman.gushchin@linux.dev ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1677023902; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references:dkim-signature; bh=0w2q1+fgkXM8VqGxzCmpXcrJt8lRH7j7myHhNUQUFhc=; b=CZada8997thVg6IhuE/EcmaOsIEf+pAkodL+JV+E7AMv5mucCK5o26KfkQHCi74r2FBqVB 2HooWYtbTNs3888+Y9EAj0Y+ZNDDyU6Xy7Dm4PqcfFDEWUlN+3UIp4VYR9jxTuC9VRP7T1 XBRpJfWPk4pBQ/QyRFLVZwfi6aGBH7c= ARC-Authentication-Results: i=1; imf12.hostedemail.com; dkim=pass header.d=linux.dev header.s=key1 header.b=e3iOG7gV; dmarc=pass (policy=none) header.from=linux.dev; spf=pass (imf12.hostedemail.com: domain of roman.gushchin@linux.dev designates 95.215.58.59 as permitted sender) smtp.mailfrom=roman.gushchin@linux.dev ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1677023902; a=rsa-sha256; cv=none; b=ZyQXB+FwYQdrr+0hb8vqPwxPpXaxGa3dVrdBcwH+aJDOhV1+eKYZfu/pFmgrkBKlX8rEaS si+LEnj3osBGAIekTvOLXMoinURE7L32wDBWWCr47M8Vg+JpUVTwclIZ7mHIouPqACfVxR xkjNZjSCpACT8bKVXbwirodfpu0OcdI= Date: Tue, 21 Feb 2023 15:57:58 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linux.dev; s=key1; t=1677023899; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=0w2q1+fgkXM8VqGxzCmpXcrJt8lRH7j7myHhNUQUFhc=; b=e3iOG7gVb1rYOmpMnUAF4daiLb4LdLdCSXjubiHfiSkJkv33MuwWdFg80gcmHfFGQtv+4f 7ffeHmGTYooclzjURTnuwUVAhKJY5hQrPEd1/oishU5tNFMrWRwiLWqtmzuO0nxawSHfS8 3NCSH8KSbZylrUo1nUmXqgTxAweW634= X-Report-Abuse: Please report any abuse attempt to abuse@migadu.com and include these headers. From: Roman Gushchin To: "Paul E. McKenney" Cc: Shakeel Butt , Matthew Wilcox , Marco Elver , Yue Zhao , linux-mm@kvack.org, akpm@linux-foundation.org, hannes@cmpxchg.org, mhocko@kernel.org, muchun.song@linux.dev, cgroups@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH] mm: change memcg->oom_group access with atomic operations Message-ID: References: <20230220230624.lkobqeagycx7bi7p@google.com> <6563189C-7765-4FFA-A8F2-A5CC4860A1EF@linux.dev> <20230221182359.GJ2948950@paulmck-ThinkPad-P17-Gen-1> <20230221223811.GK2948950@paulmck-ThinkPad-P17-Gen-1> <20230221233824.GM2948950@paulmck-ThinkPad-P17-Gen-1> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20230221233824.GM2948950@paulmck-ThinkPad-P17-Gen-1> X-Migadu-Flow: FLOW_OUT X-Rspamd-Queue-Id: E4A2040009 X-Rspamd-Server: rspam09 X-Rspam-User: X-Stat-Signature: ujddqqgwmu7m3cpke1c5d7z1pd1g6cfq X-HE-Tag: 1677023901-405506 X-HE-Meta: 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 vAN8Jjux QUF+JtgGmAVNn92SHnE35aUlAAf5qZaFF6NC63bOCBpXg4BF890RE26fdZI9P6lwOqXtrCMOutMt2oLtsn7dwSe+Axacs5yJEW437Arfga5NSG83D9NZsrfxMsGQoYcLbdo7uQDsothwywhsLzmRndssDH33bfD7fP0c7LXQ0KBEHbQu3nenzTQgBqeWwgMZO/nYs51PyY7Y2igMuYDAQqdd9inlJDt4nHc9r33O3f3q7qkn5Hgeya84r4hHzlda1xpl3v899jkSQNygeH8TUAW7XKvd41NFr7H5Ol+d0lf4C25aZRMITWcS8q1+O07hxEhd6VusgndRB65vWt24uiwRVbClHrK2khTzt0vKQ0p2Bcj6Yjb0tUF+j3g== X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On Tue, Feb 21, 2023 at 03:38:24PM -0800, Paul E. McKenney wrote: > On Tue, Feb 21, 2023 at 03:13:36PM -0800, Shakeel Butt wrote: > > On Tue, Feb 21, 2023 at 2:38 PM Paul E. McKenney wrote: > > > > > > On Tue, Feb 21, 2023 at 02:23:31PM -0800, Roman Gushchin wrote: > > > > On Tue, Feb 21, 2023 at 10:23:59AM -0800, Paul E. McKenney wrote: > > > > > On Tue, Feb 21, 2023 at 08:56:59AM -0800, Shakeel Butt wrote: > > > > > > +Paul & Marco > > > > > > > > > > > > On Tue, Feb 21, 2023 at 5:51 AM Matthew Wilcox wrote: > > > > > > > > > > > > > > On Mon, Feb 20, 2023 at 10:52:10PM -0800, Shakeel Butt wrote: > > > > > > > > On Mon, Feb 20, 2023 at 9:17 PM Roman Gushchin wrote: > > > > > > > > > > On Feb 20, 2023, at 3:06 PM, Shakeel Butt wrote: > > > > > > > > > > > > > > > > > > > > On Mon, Feb 20, 2023 at 01:09:44PM -0800, Roman Gushchin wrote: > > > > > > > > > >>> On Mon, Feb 20, 2023 at 11:16:38PM +0800, Yue Zhao wrote: > > > > > > > > > >>> The knob for cgroup v2 memory controller: memory.oom.group > > > > > > > > > >>> will be read and written simultaneously by user space > > > > > > > > > >>> programs, thus we'd better change memcg->oom_group access > > > > > > > > > >>> with atomic operations to avoid concurrency problems. > > > > > > > > > >>> > > > > > > > > > >>> Signed-off-by: Yue Zhao > > > > > > > > > >> > > > > > > > > > >> Hi Yue! > > > > > > > > > >> > > > > > > > > > >> I'm curious, have any seen any real issues which your patch is solving? > > > > > > > > > >> Can you, please, provide a bit more details. > > > > > > > > > >> > > > > > > > > > > > > > > > > > > > > IMHO such details are not needed. oom_group is being accessed > > > > > > > > > > concurrently and one of them can be a write access. At least > > > > > > > > > > READ_ONCE/WRITE_ONCE is needed here. > > > > > > > > > > > > > > > > > > Needed for what? > > > > > > > > > > > > > > > > For this particular case, documenting such an access. Though I don't > > > > > > > > think there are any architectures which may tear a one byte read/write > > > > > > > > and merging/refetching is not an issue for this. > > > > > > > > > > > > > > Wouldn't a compiler be within its rights to implement a one byte store as: > > > > > > > > > > > > > > load-word > > > > > > > modify-byte-in-word > > > > > > > store-word > > > > > > > > > > > > > > and if this is a lockless store to a word which has an adjacent byte also > > > > > > > being modified by another CPU, one of those CPUs can lose its store? > > > > > > > And WRITE_ONCE would prevent the compiler from implementing the store > > > > > > > in that way. > > > > > > > > > > > > Thanks Willy for pointing this out. If the compiler can really do this > > > > > > then [READ|WRITE]_ONCE are required here. I always have big bad > > > > > > compiler lwn article open in a tab. I couldn't map this transformation > > > > > > to ones mentioned in that article. Do we have name of this one? > > > > > > > > > > No, recent compilers are absolutely forbidden from doing this sort of > > > > > thing except under very special circumstances. > > > > > > > > > > Before C11, compilers could and in fact did do things like this. This is > > > > > after all a great way to keep the CPU's vector unit from getting bored. > > > > > Unfortunately for those who prize optimization above all else, doing > > > > > this can introduce data races, for example: > > > > > > > > > > char a; > > > > > char b; > > > > > spin_lock la; > > > > > spin_lock lb; > > > > > > > > > > void change_a(char new_a) > > > > > { > > > > > spin_lock(&la); > > > > > a = new_a; > > > > > spin_unlock(&la); > > > > > } > > > > > > > > > > void change_b(char new_b) > > > > > { > > > > > spin_lock(&lb); > > > > > b = new_b; > > > > > spin_unlock(&lb); > > > > > } > > > > > > > > > > If the compiler "optimized" that "a = new_a" so as to produce a non-atomic > > > > > read-modify-write sequence, it would be introducing a data race. > > > > > And since C11, the compiler is absolutely forbidden from introducing > > > > > data races. So, again, no, the compiler cannot invent writes to > > > > > variables. > > > > > > > > > > What are those very special circumstances? > > > > > > > > > > 1. The other variables were going to be written to anyway, and > > > > > none of the writes was non-volatile and there was no ordering > > > > > directive between any of those writes. > > > > > > > > > > 2. The other variables are dead, as in there are no subsequent > > > > > reads from them anywhere in the program. Of course in that case, > > > > > there is no need to read the prior values of those variables. > > > > > > > > > > 3. All accesses to all of the variables are visible to the compiler, > > > > > and the compiler can prove that there are no concurrent accesses > > > > > to any of them. For example, all of the variables are on-stack > > > > > variables whose addresses are never taken. > > > > > > > > > > Does that help, or am I misunderstanding the question? > > > > > > > > Thank you, Paul! > > > > > > > > So it seems like READ_ONCE()/WRITE_ONCE() are totally useless here. > > > > Or I still miss something? > > > > > > Yes, given that the compiler will already avoid inventing data-race-prone > > > C-language accesses to shared variables, so if that was the only reason > > > that you were using READ_ONCE() or WRITE_ONCE(), then READ_ONCE() and > > > WRITE_ONCE() won't be helping you. > > > > > > Or perhaps better to put it a different way... The fact that the compiler > > > is not permitted to invent data-racy reads and writes is exactly why > > > you do not normally need READ_ONCE() and WRITE_ONCE() for accesses in > > > lock-based critical sections. Instead, you only need READ_ONCE() and > > > WRITE_ONCE() when you have lockless accesses to the same shared variables. > > > > This is lockless access to memcg->oom_group potentially from multiple > > CPUs, so, READ_ONCE() and WRITE_ONCE() are needed, right? > > Agreed, lockless concurrent accesses should use READ_ONCE() and WRITE_ONCE(). > And if either conflicting access is lockless, it is lockless. ;-) Now I'm confused, why we should use it here? Writing is happening from a separate syscall (a single write from a syscall), reading is happening from a oom context. The variable is boolean, it's either 0 or 1. What difference READ_ONCE()/WRITE_ONCE() will make here? Thanks!