From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262441AbTJPG6E (ORCPT ); Thu, 16 Oct 2003 02:58:04 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262613AbTJPG6E (ORCPT ); Thu, 16 Oct 2003 02:58:04 -0400 Received: from ns.virtualhost.dk ([195.184.98.160]:52186 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S262441AbTJPG6C (ORCPT ); Thu, 16 Oct 2003 02:58:02 -0400 Date: Thu, 16 Oct 2003 08:57:43 +0200 From: Jens Axboe To: "David S. Miller" Cc: mika.penttila@kolumbus.fi, herbert@gondor.apana.org.au, torvalds@osdl.org, hch@infradead.org, linux-kernel@vger.kernel.org Subject: Re: [SCSI] Set max_phys_segments to sg_tablesize Message-ID: <20031016065743.GG1128@suse.de> References: <20031015123722.CSSJ7495.fep07-app.kolumbus.fi@mta.imail.kolumbus.fi> <20031015215134.6f5a0479.davem@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20031015215134.6f5a0479.davem@redhat.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Oct 15 2003, David S. Miller wrote: > On Wed, 15 Oct 2003 15:37:22 +0300 > wrote: > > > >It does matter if the driver can't cope with it. > > > > But this fix may hurt performance with well behaving drivers in > > iommu systems. It's better to fix the broken drivers I think. > > That's absolutely correct. > > If you add Herbert's change, you will effectively turn off all > of the IOMMU coalescing benefits we designed into the block layer. > Please don't apply it, fix the broken drivers instead. Just changing it to hw segments should be fine, but I agree phys segments is the wrong check. -- Jens Axboe