Re: [PATCH] x86/resctrl: avoid compiler optimization in __resctrl_sched_in

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>,
	Stephane Eranian <eranian@google.com>,
	linux-kernel@vger.kernel.org, tony.luck@intel.com,
	reinette.chatre@intel.com, fenghua.yu@intel.com,
	peternewman@google.com, james.morse@arm.com, babu.moger@amd.com,
	ananth.narayan@amd.com, vschneid@redhat.com,
	Nathan Chancellor <nathan@kernel.org>,
	clang-built-linux <llvm@lists.linux.dev>,
	Borislav Petkov <bp@alien8.de>, "H. Peter Anvin" <hpa@zytor.com>,
	linux-toolchains@vger.kernel.org
Subject: Re: [PATCH] x86/resctrl: avoid compiler optimization in __resctrl_sched_in
Date: Tue, 7 Mar 2023 09:48:33 -0800	[thread overview]
Message-ID: <CAHk-=whX6g3QPA_g_ovsig8CsSHRS-vTWkdn9GCUCAZY=XPtCw@mail.gmail.com> (raw)
In-Reply-To: <20230307113545.GB2017917@hirez.programming.kicks-ass.net>

On Tue, Mar 7, 2023 at 3:36 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> > I don't have a satisfactory answer yet, but am looking into this.
>
> Oh, geez, what a twisty tale that... So Linus knew back in '09 that "p"
> was icky, but it sorta was the only thing and it 'worked' -- until now
> :/

Yeah, so 'p' definitely is about the pointer, and I do worry that it
is only a dependency on exactly that - not the memory behind it.

I have this dim memory about us having talked about this with some gcc
person, and coming to the conclusion that it was all fine, but I
suspect it was in some very specific case where it might have been
fine for other reasons.

> Is there a way to explicitly order these things? barrier() obviously
> isn't going to help here.

So one "asm volatile" should always be ordered wrt another "asm volatile".

I have this other dim memory of it not even being clear whether "asm"
and "asm volatile" are ordered. I don't think they necessarily are
(with the obvious caveat that an asm without any arguments - a
so-called "basic asm" - is always volatile whether the "volatile" is
there or not).

I have a lot of dim memories, in other words. Should that worry me?

And then there's the "memory" clobber, of course.

But both of those are also going to disable CSE.

I do think that percpu_stable_op can use "p", but only when the value
is *truly* stable and there is no question about it being moved around
things that might modify it.

                         Linus