From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [PATCH] NUMA aware allocation of transmit and receive buffers for e1000 Date: 30 May 2005 11:41:07 +0200 Message-ID: <20050530094107.GJ86087@muc.de> References: <428AF378.5020109@cosmosbay.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Eric Dumazet , netdev@oss.sgi.com, akpm@osdl.org, shai@scalex86.org Return-path: Date: Mon, 30 May 2005 11:41:07 +0200 To: Christoph Lameter Content-Disposition: inline In-Reply-To: Sender: netdev-bounce@oss.sgi.com Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Fri, May 27, 2005 at 02:15:55PM -0700, Christoph Lameter wrote: > On Wed, 18 May 2005, Eric Dumazet wrote: > > > Related question, is pci_alloc_consistent() already NUMA aware ? > > Nope. It also ultimately acquires memory via vmalloc. x86-64 vmalloc has been NUMA aware forever. I believe the HP BigTux patchkit for IA64 added it there too, although it might not have hit mainline. > > pci_alloc_consistent will also need some rework to allocate memory > consistent (well the name is coherent right?) with the node that the > device is on. Second case of a need for a node aware vmalloc? You can already have it today by temporarily changing the process mempolicy. However for pci_alloc_consistent just using alloc_pages_node directly seems to work fine. -Andi