From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate.crashing.org (gate.crashing.org [63.228.1.57]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id E4B4467B89 for ; Tue, 24 Oct 2006 12:24:14 +1000 (EST) Subject: Re: [RFC]: map 4K iommu pages even on 64K largepage systems. From: Benjamin Herrenschmidt To: Geoff Levand In-Reply-To: <453D635B.20405@am.sony.com> References: <20061024002540.GA6360@austin.ibm.com> <453D635B.20405@am.sony.com> Content-Type: text/plain Date: Tue, 24 Oct 2006 12:23:59 +1000 Message-Id: <1161656640.10524.527.camel@localhost.localdomain> Mime-Version: 1.0 Cc: Olof Johansson , linuxppc-dev@ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , > This is a step in the right direction. Cell allows each device to > have its own io pagesize, from among those that the io controller > supports (4k, 64k, 1m,16m). This limitation of the current iommu > code that you try to address here has caused me to use platform > specific dma alloc routines. Are devices -actually- using different page sizes ? AFAIK, for your platform, you are using a 4k base page size, thus there is very little point (in fact, it's more like a problem) to have an IO mapping using larger page sizes. In any case, read my reply to Linas, I'm proposing a scheme where each iommu_table instance could have its own page size, in which case you can just create a table instance per device. Ben.