Re: [PATCH 4/4] locking: Introduce smp_cond_acquire()

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>,
	Boqun Feng <boqun.feng@gmail.com>, Ingo Molnar <mingo@kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Paul McKenney <paulmck@linux.vnet.ibm.com>,
	Jonathan Corbet <corbet@lwn.net>,
	Michal Hocko <mhocko@kernel.org>,
	David Howells <dhowells@redhat.com>,
	Will Deacon <will.deacon@arm.com>
Subject: Re: [PATCH 4/4] locking: Introduce smp_cond_acquire()
Date: Wed, 11 Nov 2015 13:23:14 -0800	[thread overview]
Message-ID: <CA+55aFz6SV6ycQT0fopmDw6ktvpCwr5sbhUoy8w=KYMwbiSLUQ@mail.gmail.com> (raw)
In-Reply-To: <20151111193953.GA23515@redhat.com>

On Wed, Nov 11, 2015 at 11:39 AM, Oleg Nesterov <oleg@redhat.com> wrote:
> On 11/11, Peter Zijlstra wrote:
>
> Sometimes I even think it should have both ACQUIRE + RELEASE semantics.
> IOW, it should be "equivalent" to spin_lock() + spin_unlock().

That's insane.

"Release" semantics are - by definition - about stuff that *predeces* it.

It is simply not sensible to have a "wait_for_unlock()" that then
synchronizes loads or stores that happened *before* the wait. That's
some crazy voodoo programming.

And if you do need to have model where you do "store something, then
make sure that the unlock we wait for happens after the store", then
you need to just add a "smp_mb()" in between that store and the
waiting for unlock.

Or just take the damn lock, and don't play any games at all.

                   Linus