From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from fed1rmmtao06.cox.net (fed1rmmtao06.cox.net [68.230.241.33]) by ozlabs.org (Postfix) with ESMTP id C770F2BD46 for ; Wed, 20 Oct 2004 00:56:27 +1000 (EST) Date: Tue, 19 Oct 2004 07:27:54 -0700 From: Matt Porter To: Conor McLoughlin Message-ID: <20041019072754.A29957@home.com> References: <1098072098.751.32.camel@nighteyes.localdomain> <41738D27.4060007@eircom.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <41738D27.4060007@eircom.net>; from cml3227@eircom.net on Mon, Oct 18, 2004 at 10:30:15AM +0100 Cc: linuxppc-embedded Subject: Re: Does kmalloc on MPC82xx work correctly with GFP_DMA? List-Id: Linux on Embedded PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Mon, Oct 18, 2004 at 10:30:15AM +0100, Conor McLoughlin wrote: > I have been looking at the ethernet device driver (fcc_enet) for the > mpc82xx platform. This allocates buffer descriptors using kmalloc with > the GFP_DMA flag. As far as I can see on my platform, this allocates GFP_DMA has no meaning on PPC. All memory is DMAable. GFP_DMA is for PeeCees with a limited ISA DMA space. Drivers should be using GFP_KERNEL so they don't confuse people. > from the regular kernel memory (0xCxxxxxxx). As the attributes of this > block of memory are controlled by the block address translation > registers, this cannot be DMA safe, can it? Sure it can. It works with hardware snooping. > Is there something I am missing here? Yes, it's a 603e core and hardware snooping support manages cache coherency between system memory and devices. -Matt