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;
MIME-Version: 1.0
References: <CAHk-=wgnjMEvqHnu_iJcbr_kdFyBQLhYojwv5T7p9F+CHxA9pg@mail.gmail.com>
 <Pine.LNX.4.44L0.1911121639540.1567-100000@iolanthe.rowland.org>
 <CANn89iKjWH86kChzPiVtCgVpt3GookwGk2x1YCTMeBSPpKU+Ww@mail.gmail.com>
 <20191112224441.2kxmt727qy4l4ncb@ast-mbp.dhcp.thefacebook.com>
 <CANn89iKLy-5rnGmVt-nzf6as4MvXgZzSH+BSReXZKpSTjhoWAw@mail.gmail.com>
 <CAHk-=wj_rFOPF9Sw_-h-6J93tP9qO_UOEvd-e02z9+DEDs2kLQ@mail.gmail.com> <CANpmjNNkbWoqckyPfhq52W4WfJWR2=rt8WXzs+WXEZzv9xxL0g@mail.gmail.com>
In-Reply-To: <CANpmjNNkbWoqckyPfhq52W4WfJWR2=rt8WXzs+WXEZzv9xxL0g@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed, 13 Nov 2019 08:57:49 -0800
Message-ID: <CAHk-=wg5CkOEF8DTez1Qu0XTEFw_oHhxN98bDnFqbY7HL5AB2g@mail.gmail.com>
Subject: Re: KCSAN: data-race in __alloc_file / __alloc_file
To:     Marco Elver <elver@google.com>
Cc:     Eric Dumazet <edumazet@google.com>,
        Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Alan Stern <stern@rowland.harvard.edu>,
        Eric Dumazet <eric.dumazet@gmail.com>,
        syzbot <syzbot+3ef049d50587836c0606@syzkaller.appspotmail.com>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        syzkaller-bugs <syzkaller-bugs@googlegroups.com>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Andrea Parri <parri.andrea@gmail.com>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        LKMM Maintainers -- Akira Yokosawa <akiyks@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Nov 13, 2019 at 7:00 AM Marco Elver <elver@google.com> wrote:
>
> Just to summarize the options we've had so far:
> 1. Add a comment, and let the tool parse it somehow.
> 2. Add attribute to variables.
> 3. Add some new macro to use with expressions, which doesn't do
> anything if the tool is disabled. E.g. "racy__(counter++)",
> "lossy__(counter++);" or any suitable name.

I guess I could live with "data_race(x)" or something simple like
that, assuming we really can just surround a whole expression with it,
and we don't have to make a hundred different versions for the
different cases ("racy plain assignment" vs "racy statistics update"
vs "racy u64 addition" etc etc).

I just want the source code to be very legible, which is one of the
problems with the ugly READ_ONCE() conversions.

Part of that "legible source code" implies no crazy double
underscores. But a plain "data_race(x)" might not look too bad, and
would be easy to grep for, and doesn't seem to exist in the current
kernel as anything else.

One question is if it would be a statement expression or an actual
expression. I think the expression would read much better, IOW you
could do

    val = data_race(p->field);

instead of having to write it as

    data_race(val = p->field);

to really point out the race. But at the same time, maybe you need to
surround several statements, ie

    // intentionally racy xchg because we don't care and it generates
better code
    data_race(a = p->field; p->field = b);

which all would work fine with a non-instrumented macro something like this:

    #define data_race(x) ({ x; })

which would hopefully give the proper syntax rules.

But that might be very very inconvenient for KCSAN, depending on how
you annotate the thing.

So I _suspect_ that what you actually want is to do it as a statement,
not as an expression. What's the actual underlying syntax for "ignore
this code for thread safety checking"?

                    Linus