MIME-Version: 1.0
References: <20211210161618.645249719@infradead.org> <20211210162313.857673010@infradead.org>
 <20211213164334.GY16608@worktop.programming.kicks-ass.net>
In-Reply-To: <20211213164334.GY16608@worktop.programming.kicks-ass.net>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon, 13 Dec 2021 10:11:28 -0800
Message-ID: <CAHk-=wjc+mr_Rh++5pPDkNFuceyPwFxCtzp124AppBLgbVVV0A@mail.gmail.com>
Subject: Re: [PATCH v2 8/9] atomic,x86: Alternative atomic_*_overflow() scheme
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Will Deacon <will@kernel.org>, Boqun Feng <boqun.feng@gmail.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Marco Elver <elver@google.com>,
        Kees Cook <keescook@chromium.org>,
        Christoph Hellwig <hch@infradead.org>,
        Jens Axboe <axboe@kernel.dk>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Dec 13, 2021 at 8:43 AM Peter Zijlstra <peterz@infradead.org> wrote=
:
>
> So Marco was expressing doubt about this exact interface for the
> atomic_*_overflow() functions, since it's extremely easy to get the
> whole ATOMIC_OVERFLOW_OFFSET thing wrong.

I missed that discussion (maybe it was on irc? Or maybe I just get too
much email).

Anyway, my preferred solution would simply be to make the ref-counting
atomics use a different type.

Voil=C3=A0, problem solved. You can't really misuse them by mistake,
because you can't access it by mistake.

Sure, it could be a wrapper around 'atomic_t' on architectures that
end up using the generic fallback, so it might be as simple as

   typedef atomic_t atomic_ref_t;

in some asm-generic implementation, although I suspect that you'd want
type safety even there, and do

  typedef struct { atomic_t atomic_val; } atomic_ref_t;

But then on x86 - and other architectures that might prefer to use
that offset trick because they have flags - I'm not sure it even makes
sense to have anything to do with 'atomic_t' at all, since there would
basically be zero overlap with the regular atomic operations (partly
due to the offset, but partly simply because the 'ref' operations are
simply different).

(Wrt naming: I do think this is more about the "ref" part than the
"overflow" part - thus I'd suggest the "atomic_ref_t" rather than your
ofl naming).

            Linus