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;
Date:   Wed, 27 Feb 2019 17:48:06 +0100
From:   Peter Zijlstra <peterz@infradead.org>
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>,
        Thomas Gleixner <tglx@linutronix.de>,
        Andrew Lutomirski <luto@kernel.org>,
        Nadav Amit <namit@vmware.com>,
        Linux List Kernel Mailing <linux-kernel@vger.kernel.org>,
        Nadav Amit <nadav.amit@gmail.com>
Subject: Re: [PATCH 1/5] x86/percpu: Differentiate this_cpu_{}() and
 __this_cpu_{}()
Message-ID: <20190227164806.GS32494@hirez.programming.kicks-ass.net>
References: <20190227101252.413192716@infradead.org>
 <20190227101544.210748832@infradead.org>
 <CAHk-=wiv6dgJDj81wukpJCJBVhAVn-CYJdhj+o8JGN1LfRwb0g@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAHk-=wiv6dgJDj81wukpJCJBVhAVn-CYJdhj+o8JGN1LfRwb0g@mail.gmail.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Feb 27, 2019 at 08:14:09AM -0800, Linus Torvalds wrote:
> On Wed, Feb 27, 2019 at 2:16 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > Nadav Amit reported that commit:
> >
> >   b59167ac7baf ("x86/percpu: Fix this_cpu_read()")
> >
> > added a bunch of constraints to all sorts of code; and while some of
> > that was correct and desired, some of that seems superfluous.
> 
> Hmm.
> 
> I have the strong feeling that we should instead relax this_cpu_read()
> again a bit.
> 
> In particular, making it "asm volatile" really is a big hammer
> approach. It's worth noting that the *other* this_cpu_xyz ops don't
> even do that.

Right, this patch 'fixes' that :-)

> I would suggest that instead of making "this_cpu_read()" be asm
> volatile, we mark it as potentially changing the memory location it is
> touching - the same way the modify/write ops do.
> 
> That still means that the read will be forced (like READ_ONCE()), but
> allows gcc a bit more flexibility in instruction scheduling, I think.

Ah, fair enough, I'll spin a version of this patch with "+m" for
this_cpu and "m" for raw_cpu.

> That said, I didn't actually check how it affects code generation.
> Nadav, would you check the code sequences you originally noticed?

Much of it was the ONCE behaviour defeating CSE I think, but yes, it
would be good to have another look.