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;
MIME-Version: 1.0
References: <20230801011554.3950435-1-guoren@kernel.org> <CAHirt9ht8AsE=FC8+222JDZXH3T58uLt+o=_pq+1zBhv1MKRjg@mail.gmail.com>
 <CAJF2gTRa5erHomJzLgUFO4SGqd5zSDwn6r3WN7kM8aWpv1vesg@mail.gmail.com>
 <CAJF2gTQ1hV1vipAo3H4X4WiPO84kVVFZcdGq7u4f0bVTry_akQ@mail.gmail.com>
 <CAHirt9gVqE=9vviJEY=kY=booVRmFPHrnFsKCXPXnXiWTB8bZQ@mail.gmail.com>
 <CAJF2gTR2ON33wc87iV564rkDbNiE56h_t0kzKKXdJtGqgJ1sOQ@mail.gmail.com>
 <CAHirt9i_osW_Dy5jNAnNOKrm-+38qN7SF+8ofNHePMAYhRKHsA@mail.gmail.com> <CAJF2gTSLD3jwzV59dGj_RENCvE0zMOAkLHqmT55bxL-4vA1wkA@mail.gmail.com>
In-Reply-To: <CAJF2gTSLD3jwzV59dGj_RENCvE0zMOAkLHqmT55bxL-4vA1wkA@mail.gmail.com>
From:   WANG Rui <wangrui@loongson.cn>
Date:   Wed, 2 Aug 2023 10:23:18 +0800
Message-ID: <CAHirt9iiA_zdBbmnmMNAEmvKBU-1imRxD7wG-0OmnPTsG2RJMA@mail.gmail.com>
Subject: Re: [PATCH] LoongArch: Fixup cmpxchg sematic for memory barrier
To:     Guo Ren <guoren@kernel.org>
Cc:     chenhuacai@kernel.or, kernel@xen0n.name, arnd@arndb.de,
        andi.shyti@linux.intel.com, andrzej.hajda@intel.com,
        peterz@infradead.org, will@kernel.org, boqun.feng@gmail.com,
        mark.rutland@arm.com, loongarch@lists.linux.dev,
        linux-kernel@vger.kernel.org, Guo Ren <guoren@linux.alibaba.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Precedence: bulk

On Wed, Aug 2, 2023 at 7:17 AM Guo Ren <guoren@kernel.org> wrote:
>
> On Tue, Aug 1, 2023 at 12:37 PM WANG Rui <wangrui@loongson.cn> wrote:
> >
> > On Tue, Aug 1, 2023 at 6:50 PM Guo Ren <guoren@kernel.org> wrote:
> > >
> > > On Tue, Aug 1, 2023 at 5:32 PM WANG Rui <wangrui@loongson.cn> wrote:
> > > > No. LL and LL won't reorder because LL implies a memory barrier(though
> > > > not acquire semantics).
> > > That means we could remove __WEAK_LLSC_MB totally, right?
> >
> > More precisely, __WEAK_LLSC_MB is intended to prevent reordering
> > between LL and normal LD used to fetch the expected value for cmpxchg.
> Oh, that's unnecessary when cmpxchg fails.
>
> Maybe you treat cmpxchg as a CoRR antidote in coincidence. Please
> solve the CoRR problem by READ_ONCE.
>
> See alpha architecture.

Unfortunately, the LL instruction has no acquire semantics. Even if
our kernel team improves READ_ONCE, it cannot prevent reordering
between LL and READ_ONCE after cmpxchg fails.

LL (<memory-barrier> + <load-exclusive>); WEAK_LLSC_MB; READ_ONCE
(<normal-load>); ...

vs

LL (<memory-barrier> + <load-exclusive>); READ_ONCE (<normal-load> +
<memory-barrier>); ...

Improving READ_ONCE is really important.

Regards,
-- 
WANG Rui