Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 31 Oct 2000 01:11:27 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 31 Oct 2000 01:11:18 -0500 Received: from ha1.rdc2.mi.home.com ([24.2.68.68]:26076 "EHLO mail.rdc2.mi.home.com") by vger.kernel.org with ESMTP id ; Tue, 31 Oct 2000 01:11:02 -0500 Message-ID: <39FE6291.FA8162A7@didntduck.org> Date: Tue, 31 Oct 2000 01:11:29 -0500 From: Brian Gerst X-Mailer: Mozilla 4.75 [en] (X11; U; Linux 2.2.16-3 i586) X-Accept-Language: en MIME-Version: 1.0 To: "H. Peter Anvin" CC: linux-kernel@vger.kernel.org Subject: Re: kmalloc() allocation. In-Reply-To: <8tll94$hc9$1@cesium.transmeta.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org "H. Peter Anvin" wrote: > > Followup to: > By author: "Richard B. Johnson" > In newsgroup: linux.dev.kernel > > > > > 64K probably less. kmalloc allocates physically linear spaces. vmalloc will > > > happily grab you 2Mb of space but it will not be physically linear > > > > > > > Okay. Thanks. > > > > FWIW, vmalloc()-allocated pages are definitely pinned-down and > available to interrupts. However, you should keep in mind that the > vmalloc() call *itself* is quite expensive on SMP machines (have to > interrupt all CPUs and flush their TLBs!!) so if you're using > vmalloc(), be careful with the number of calls you make. Of course, > this is usually not a problem. This was just changed in 2.4 so that vmalloced pages are faulted in on demand. -- Brian Gerst - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/