Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754514AbYGCSAR (ORCPT ); Thu, 3 Jul 2008 14:00:17 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752985AbYGCSAF (ORCPT ); Thu, 3 Jul 2008 14:00:05 -0400 Received: from outbound-sin.frontbridge.com ([207.46.51.80]:34730 "EHLO SG2EHSOBE001.bigfish.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752604AbYGCSAD convert rfc822-to-8bit (ORCPT ); Thu, 3 Jul 2008 14:00:03 -0400 X-BigFish: VPS-23(zz98dR7efVzzzzz32i6bh61h) X-Spam-TCS-SCL: 0:0 X-WSS-ID: 0K3FZBH-02-RQ3-01 x-mimeole: Produced By Microsoft Exchange V6.5 Content-Class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8BIT Subject: RE: [PATCH 3/6] AMD IOMMU: flush domain TLB when there is more than onepage to flush Date: Thu, 3 Jul 2008 13:01:21 -0500 Message-ID: <6453C3CB8E2B3646B0D020C112613273013D91CF@sausexmb4.amd.com> In-Reply-To: <1215106511-12915-4-git-send-email-joerg.roedel@amd.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PATCH 3/6] AMD IOMMU: flush domain TLB when there is more than onepage to flush Thread-Index: AcjdM43RxO4320PASnuxIq1jBk1D0QAAh92A References: <1215106511-12915-1-git-send-email-joerg.roedel@amd.com> <1215106511-12915-4-git-send-email-joerg.roedel@amd.com> From: "Duran, Leo" To: "Roedel, Joerg" , , CC: "Richter, Robert" , , , "Sarathy, Bhavna" X-OriginalArrivalTime: 03 Jul 2008 17:59:51.0268 (UTC) FILETIME=[9713AA40:01C8DD36] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 928 Lines: 27 On Thursday, July 03, 2008 12:35 PM, Joerg Roedel wrote: > This patch changes the domain TLB flushing behavior of the driver. When > there > is more than one page to flush it flushes the whole domain TLB instead > of every > single page. So we send only a single command to the IOMMU in every > case which > is faster to execute. > Joerg, Instead of flushing the complete TLB, how about just setting the page-size encoding bits so as to cover all of the pages (rounded up to a power of 2... which in some cases may flush a few more pages, but not the complete TLB). This way other devices don't pay the 'flushing' penalty every time you unmap a buffer for a given device. Leo. -- 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/