From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-arch-owner@vger.kernel.org>
Received: from gate.crashing.org ([63.228.1.57]:45977 "EHLO gate.crashing.org"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1752857AbXGWGGo (ORCPT <rfc822;linux-arch@vger.kernel.org>);
	Mon, 23 Jul 2007 02:06:44 -0400
Subject: Re: [patch] bitops: lock bitops
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In-Reply-To: <20070712031419.GF32414@wotan.suse.de>
References: <20070712031419.GF32414@wotan.suse.de>
Content-Type: text/plain
Date: Mon, 23 Jul 2007 16:06:31 +1000
Message-Id: <1185170792.5439.105.camel@localhost.localdomain>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-arch-owner@vger.kernel.org
To: Nick Piggin <npiggin@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>, Paul Mackerras <paulus@samba.org>, tony.luck@intel.com, linux-arch@vger.kernel.org
List-ID: <linux-arch.vger.kernel.org>

On Thu, 2007-07-12 at 05:14 +0200, Nick Piggin wrote:
> Here is a set of patches that aims to mitigate some of the lock_page
> overhead on powerpc introduced in the fault path by another set.
> Fortunately it also improves various other things too :)

 Some nits...

> +static __inline__ void clear_bit_unlock(int nr, volatile unsigned long *addr)
> +{
> +	unsigned long old;
> +	unsigned long mask = BITOP_MASK(nr);
> +	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
> +
> +	__asm__ __volatile__(
> +	LWSYNC_ON_SMP
> +"1:"	PPC_LLARX "%0,0,%3	# clear_bit_unlock\n"
> +	"andc	%0,%0,%2\n"
> +	PPC405_ERR77(0,%3)
> +	PPC_STLCX "%0,0,%3\n"
> +	"bne-	1b"
> +	: "=&r" (old), "+m" (*p)
> +	: "r" (mask), "r" (p)
> +	: "cc" );
> +}

You also want a "memory" clobber on clear_bit_unlock() to tell gcc to
consider it as a barrier for memory accesses.

In addition, it's worth documenting that while spin_unlock() provides
synchronization with outstanding MMIOs (look at the io sync stuff we
have in spinlock.h), these bitops don't (and I don't want to add that
here anyway), so drivers abusing bitops for lock might want to use
mmiowb() explicitely.

> +static __inline__ void __clear_bit_unlock(int nr, volatile unsigned long *addr)
> +{
> +	__asm__ __volatile__(LWSYNC_ON_SMP);
> +	__clear_bit(nr, addr);
> +}

The above needs a "memory" clobber too.

Cheers,
Ben.