Re: [RFC PATCH] Flipped jump labels

From: Borislav Petkov <bp@alien8.de>
To: Ingo Molnar <mingo@kernel.org>
Cc: x86-ml <x86@kernel.org>, lkml <linux-kernel@vger.kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Steven Rostedt <rostedt@goodmis.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Jason Baron <jbaron@akamai.com>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [RFC PATCH] Flipped jump labels
Date: Sun, 10 Aug 2014 17:35:39 +0200	[thread overview]
Message-ID: <20140810153539.GA9490@pd.tnic> (raw)
In-Reply-To: <20140810061303.GA14206@gmail.com>

On Sun, Aug 10, 2014 at 08:13:03AM +0200, Ingo Molnar wrote:
> Wouldn't using STATIC_KEY_INIT_TRUE and static_key_true() [instead of
> !static_key_false()] result in the same good code placement effects?

Nope, not really. static_key_true() is !static_key_false() and we're not
changing anything, logically. ASM looks non-optimal here, in any case,
with the "true" version.

Also, the problem in this particular use case in native_sched_clock is
that we want to do some initialization first before we enable the RDTSC,
i.e., the likely branch.

So we want to have the following static layout after build:

JMP
likely
unlikely

and the initialization code turns the JMP into a NOP at runtime.
Basically something like alternatives. :-)

Current jump labels do not have the possibility to get generated as
jump-to-unlikely labels which get changed at init time - they issue
NOPs only.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--