From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mustang.oldcity.dca.net (mustang.oldcity.dca.net [216.158.38.3]) by ozlabs.org (Postfix) with SMTP id EEE5367B60 for ; Wed, 2 Aug 2006 05:50:42 +1000 (EST) Subject: Re: DMA buffer synchronisation with ioremap() From: Lee Revell To: Phil.Nitschke@avalon.com.au In-Reply-To: <1154351427.3203.59.camel@toby.int.avalon.com.au> References: <1154351427.3203.59.camel@toby.int.avalon.com.au> Content-Type: text/plain Date: Tue, 01 Aug 2006 15:51:15 -0400 Message-Id: <1154461876.14540.70.camel@mindpipe> Mime-Version: 1.0 Cc: linuxppc-embedded@ozlabs.org List-Id: Linux on Embedded PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Mon, 2006-07-31 at 22:40 +0930, Phil Nitschke wrote: > (I go no replies last week, so I'll try again, with less explanation...) > > If I master a DMA from a PCI device into a main memory buffer allocated > with dma_alloc_noncoherent(), I need to synchronise the destination > buffer using dma_sync_single_range_for_xxx() before and after the DMA. > > But if the buffer is a very large chunk of memory (reserved at boot > time) which has been ioremap()-ed into the virtual address space, do I > need to still synchronise that memory? Maybe it's a better question for the linux-kernel list? Lee