Re: [RFC PATCH v3] locking/atomic: Implement atomic{,64,_long}_{fetch_,}{andnot_or}{,_relaxed,_acquire,_release}()

From: hev <r@hev.cc>
To: Will Deacon <will@kernel.org>
Cc: Rui Wang <wangrui@loongson.cn>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>, Arnd Bergmann <arnd@arndb.de>,
	Waiman Long <longman@redhat.com>,
	Boqun Feng <boqun.feng@gmail.com>, Guo Ren <guoren@kernel.org>,
	linux-arch@vger.kernel.org, linux-kernel@vger.kernel.org,
	Xuefeng Li <lixuefeng@loongson.cn>,
	Huacai Chen <chenhuacai@gmail.com>,
	Jiaxun Yang <jiaxun.yang@flygoat.com>,
	Huacai Chen <chenhuacai@loongson.cn>,
	kernel test robot <lkp@intel.com>
Subject: Re: [RFC PATCH v3] locking/atomic: Implement atomic{,64,_long}_{fetch_,}{andnot_or}{,_relaxed,_acquire,_release}()
Date: Thu, 29 Jul 2021 18:58:59 +0800	[thread overview]
Message-ID: <CAHirt9j+UJiNpgmeSOMnUnYomOLgi1oD44ZCzEWA9OAzrnAMaw@mail.gmail.com> (raw)
In-Reply-To: <20210729095551.GE21151@willie-the-truck>

Hi, Will,

On Thu, Jul 29, 2021 at 5:55 PM Will Deacon <will@kernel.org> wrote:
>
> On Thu, Jul 29, 2021 at 05:30:03PM +0800, Rui Wang wrote:
> > This patch introduce a new atomic primitive andnot_or:
> >
> >  * atomic_andnot_or
> >  * atomic_fetch_andnot_or
> >  * atomic_fetch_andnot_or_relaxed
> >  * atomic_fetch_andnot_or_acquire
> >  * atomic_fetch_andnot_or_release
> >  * atomic64_andnot_or
> >  * atomic64_fetch_andnot_or
> >  * atomic64_fetch_andnot_or_relaxed
> >  * atomic64_fetch_andnot_or_acquire
> >  * atomic64_fetch_andnot_or_release
> >  * atomic_long_andnot_or
> >  * atomic_long_fetch_andnot_or
> >  * atomic_long_fetch_andnot_or_relaxed
> >  * atomic_long_fetch_andnot_or_acquire
> >  * atomic_long_fetch_andnot_or_release
> >
> > Reported-by: kernel test robot <lkp@intel.com>
> > Signed-off-by: Rui Wang <wangrui@loongson.cn>
> > ---
> >  include/asm-generic/atomic-instrumented.h |  72 +++++-
> >  include/asm-generic/atomic-long.h         |  62 ++++-
> >  include/linux/atomic-arch-fallback.h      | 262 +++++++++++++++++++++-
> >  lib/atomic64_test.c                       |  92 ++++----
> >  scripts/atomic/atomics.tbl                |   1 +
> >  scripts/atomic/fallbacks/andnot_or        |  25 +++
> >  6 files changed, 471 insertions(+), 43 deletions(-)
> >  create mode 100755 scripts/atomic/fallbacks/andnot_or
>
> Please see my other comments on the other patches you posted:
>
> https://lore.kernel.org/r/20210729093923.GD21151@willie-the-truck
>
> Overall, I'm not thrilled to bits by extending the atomics API with
> operations that cannot be implemented efficiently on any (?) architectures
> and are only used by the qspinlock slowpath on machines with more than 16K
> CPUs.
>
> I also think we're lacking documentation justifying when you would use this
> new primitive over e.g. a sub-word WRITE_ONCE() on architectures that
> support those, especially for the non-returning variants.
>
> Will

I have tried to explain in another thread. At the beginning, I thought
about implementing xchg_mask for the sub-word xchg, but now I agree
that atomic andnot_or is clearer and more general.

Peter, what do you think?

Regards,
Rui