From mboxrd@z Thu Jan 1 00:00:00 1970 From: Muli Ben-Yehuda Subject: Re: network devices don't handle pci_dma_mapping_error()'s Date: Thu, 7 Dec 2006 15:04:16 +0200 Message-ID: <20061207130416.GD2926@rhun.haifa.ibm.com> References: <200612011340.kB1DeEhR014690@dut39.unminc.com> <20061206101644.27eeaf14@freekitty> <20061206193328.GB3524@rhun.ibm.com> <200612071148.14996.amitkale@linsyssoft.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Stephen Hemminger , Jeff Garzik , "Amit S. Kale" , netdev@vger.kernel.org, brazilnut@us.ibm.com, netxenproj@linsyssoft.com, rob@netxen.com, romieu@fr.zoreil.com, sanjeev@netxen.com, wendyx@us.ibm.com Return-path: Received: from mtagate6.de.ibm.com ([195.212.29.155]:16249 "EHLO mtagate6.de.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1032193AbWLGNEU (ORCPT ); Thu, 7 Dec 2006 08:04:20 -0500 Received: from d12nrmr1607.megacenter.de.ibm.com (d12nrmr1607.megacenter.de.ibm.com [9.149.167.49]) by mtagate6.de.ibm.com (8.13.8/8.13.8) with ESMTP id kB7D4ID7140766 for ; Thu, 7 Dec 2006 13:04:19 GMT Received: from d12av02.megacenter.de.ibm.com (d12av02.megacenter.de.ibm.com [9.149.165.228]) by d12nrmr1607.megacenter.de.ibm.com (8.13.6/8.13.6/NCO v8.1.1) with ESMTP id kB7D4Ix33027190 for ; Thu, 7 Dec 2006 14:04:19 +0100 Received: from d12av02.megacenter.de.ibm.com (loopback [127.0.0.1]) by d12av02.megacenter.de.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id kB7D4HEp023551 for ; Thu, 7 Dec 2006 14:04:18 +0100 To: "Amit S. Kale" Content-Disposition: inline In-Reply-To: <200612071148.14996.amitkale@linsyssoft.com> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Thu, Dec 07, 2006 at 11:48:14AM +0530, Amit S. Kale wrote: > On the x86_64 boxes that don't feature iommu functionality (because the > motherboard disables it or because Linux can't handle it) Linux bounce buffer > framework automatically comes into picture. Could we have the same framework > take over when IOMMU space is over? I don't think this is possible with > present code, though. We probably can have fallback_dma_ops in addition to > dma_ops. In the general case, no - some platforms (including x86-64 on IBM's high end servers!) have an isolation capable IOMMU, which means all DMA mappings need to go through it, so a general mechanism to cope with DMA mappings running out is still needed. Cheers, Muli