From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner+w=401wt.eu-S1759659AbZE0MSU@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1759659AbZE0MSU (ORCPT <rfc822;w@1wt.eu>);
	Wed, 27 May 2009 08:18:20 -0400
Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752115AbZE0MSM
	(ORCPT <rfc822;linux-kernel-outgoing>);
	Wed, 27 May 2009 08:18:12 -0400
Received: from mx-out2.daemonmail.net ([216.104.160.39]:51618 "EHLO
	mx-out2.daemonmail.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751795AbZE0MSL convert rfc822-to-8bit (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 27 May 2009 08:18:11 -0400
From: "Michael S. Zick" <lkml@morethan.org>
Reply-To: lkml@morethan.org
To: Harald Welte <HaraldWelte@viatech.com>
Subject: Re:[VIA Support] was: [BUG FIX] Make x86_32 uni-processor Atomic ops, Atomic
Date: Wed, 27 May 2009 07:18:08 -0500
User-Agent: KMail/1.9.9
Cc: "H. Peter Anvin" <hpa@zytor.com>, Ingo Molnar <mingo@elte.hu>,
       Thomas Gleixner <tglx@linutronix.de>, linux-kernel@vger.kernel.org,
       Alan Cox <alan@lxorguk.ukuu.org.uk>
References: <200905221139.26941.lkml@morethan.org> <4A188A48.5000208@zytor.com> <20090524064905.GH20452@prithivi.gnumonks.org>
In-Reply-To: <20090524064905.GH20452@prithivi.gnumonks.org>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 8BIT
Content-Disposition: inline
Message-Id: <200905270718.11093.lkml@morethan.org>
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Sun May 24 2009, Harald Welte wrote:
> 
> Once I understand it in full detail, I can talk to the right people inside
> CentaurLabs (VIA's CPU division).  
> 
> If somebody (optionally) can phrase a precise technical question that I can
> directly forward to somebody with low-level x86 knowledge but no Linux background,
> it would definitely help speeding up the process.
> 

What is the PCI Cache Line size in the CX700?  In the CN896?

Ref:
arch/x86/pci/common.c

As in:
        /*
         * Assume PCI cacheline size of 32 bytes for all x86s except K7/K8
         * and P4. It's also good for 386/486s (which actually have 16)
         * as quite a few PCI devices do not support smaller values.
         */

        pci_cache_line_size = 32 >> 2;
        if (c->x86 >= 6 && c->x86_vendor == X86_VENDOR_AMD)
                pci_cache_line_size = 64 >> 2;  /* K7 & K8 */
        else if (c->x86 > 6 && c->x86_vendor == X86_VENDOR_INTEL)
                pci_cache_line_size = 128 >> 2; /* P4 */

A problem with cache coherency, alignment, or consistency would explain
the problems I am seeing - and the differences in the test cases.

Mike