Subject: Re: [PATCH] Define wc_wmb, a write barrier for PCI write combining
From: "Bryan O'Sullivan" <bos@pathscale.com>
To: Benjamin LaHaise <bcrl@kvack.org>, Andi Kleen <ak@suse.de>
Cc: Benjamin LaHaise <bcrl@kvack.org>, Andrew Morton <akpm@osdl.org>,
       linux-kernel <linux-kernel@vger.kernel.org>
In-Reply-To: <200602282033.48570.ak@suse.de>
References: <1140841250.2587.33.camel@localhost.localdomain>
	 <20060228190354.GE24306@kvack.org>
	 <1141154424.20227.11.camel@serpentine.pathscale.com>
	 <200602282033.48570.ak@suse.de>
Content-Type: text/plain
Organization: PathScale, Inc.
Date: Tue, 28 Feb 2006 11:44:51 -0800
Message-Id: <1141155892.20227.26.camel@serpentine.pathscale.com>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 736
Lines: 19

On Tue, 2006-02-28 at 20:33 +0100, Andi Kleen wrote:

> Are you sure you used the right instruction? Normally CLFLUSH is used
> for such things, not a write barrier which really only changes ordering.

Hmm.  It's possible we're just getting lucky because another write
happens somewhere else soon after the last write we perform as part of a
packet send.  Perhaps, for complete correctness, we should be using
CLFLUSH instead of the last barrier.  I'll have to look into this.

Thanks,

	<b

-
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/