Date: Wed, 13 Aug 2008 19:31:15 +0200
From: Ingo Molnar <mingo@elte.hu>
To: Mark Langsdorf <mark.langsdorf@amd.com>
Cc: linux-kernel@vger.kernel.org,
       Linus Torvalds <torvalds@linux-foundation.org>,
       Thomas Gleixner <tglx@linutronix.de>, "H. Peter Anvin" <hpa@zytor.com>
Subject: Re: [PATCH][retry 1] 2.6.27-rc2: invalidate caches before going
	into suspend
Message-ID: <20080813173115.GA28891@elte.hu>
References: <200808131228.50415.mark.langsdorf@amd.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <200808131228.50415.mark.langsdorf@amd.com>
User-Agent: Mutt/1.5.18 (2008-05-17)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1267
Lines: 39


(please keep Cc:s intact)

* Mark Langsdorf <mark.langsdorf@amd.com> wrote:

> @@ -104,6 +103,7 @@ static inline void play_dead(void)
>  	 * With physical CPU hotplug, we should halt the cpu
>  	 */
>  	local_irq_disable();
> +	wbinvd();
>  	while (1)
>  		halt();

hm, why not do what i suggested in my first mail:

        if (cpu >= i486)
                asm("cli; wbinvd; cli; 1: hlt; jmp 1b")
        else
		halt();

perhaps turn it into a wbivd_halt() primitive, to make it clean and even 
more obvious.

This sequence does matter to reliable suspend/resume, and in theory gcc 
could insert something before a halt() as well.

[ i only have a pretty far-fetched example that in all likelyhood wont
  happen in practice: for example halt could be decided to be uninlined
  by a braindead compiler, ftrace could hook in there, and dirty some
  state. But still - the point is that we had a difficult bug in this 
  code for a long time and in such situations we should just over-do 
  robustness by default. ]

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/