Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
From:   Qian Cai <cai@lca.pw>
Mime-Version: 1.0 (1.0)
Subject: Re: [PATCH -next] mm: mark a intentional data race in page_zonenum()
Date:   Thu, 6 Feb 2020 06:14:55 -0500
Message-Id: <3E1FB0B8-F499-45F7-9B03-7B80395FFF19@lca.pw>
References: <20200206090436.GF14001@quack2.suse.cz>
Cc:     John Hubbard <jhubbard@nvidia.com>, akpm@linux-foundation.org,
        ira.weiny@intel.com, dan.j.williams@intel.com, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
In-Reply-To: <20200206090436.GF14001@quack2.suse.cz>
To:     Jan Kara <jack@suse.cz>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Feb 6, 2020, at 4:04 AM, Jan Kara <jack@suse.cz> wrote:
>=20
> So in this particular case, store tearing is non-issue because we use
> atomic operation to store the value in page_cpupid_xchg_last(). I think it=

> would make some sense to use READ_ONCE(page->flags) here to prevent
> compiler from loading page->flags several times - I have hard time finding=

> a reason why a compiler would want to do that but conceptually that
> protection makes sense, it is for free performance wise, and will still
> allow KCSAN to find a race in case we ever grow a place that modifies
> page's zone non-atomically (which might be a real problem). And it should
> also silence the KCSAN warning AFAIU.

Ah, read up to 3 bits might be an issue then. I=E2=80=99ll post an alternati=
ve version which uses READ_ONCE() just for the old page ( because the new pa=
ge has not been published yet) in wp_page_copy() then.=