From: Hollis Blanchard Subject: Re: [RFC PATCH 07/35] Make LOAD_OFFSET defined by subarch Date: Tue, 28 Mar 2006 11:00:56 -0600 Message-ID: <200603281100.57586.hollisb@us.ibm.com> References: <20060322063040.960068000@sorel.sous-sol.org> <200603271256.44698.hollisb@us.ibm.com> <4428F8A1.9030909@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4428F8A1.9030909@suse.de> Content-Disposition: inline List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Gerd Hoffmann Cc: xen-devel@lists.xensource.com, xen-ppc-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org On Tuesday 28 March 2006 02:49, Gerd Hoffmann wrote: > > > This makes me a little nervous, since the ELF loading code is working fine for > > PPC at the moment. What exactly are you planning on changing? > > Basically what the suggestion outlined: use physical address + VIRT_BASE > instead of placing virtual addresses into the physical address fields. > Some discussions on that went over the list some weeks ago. Patch > attached for reference. > + * bug comparibility alert: old linux kernels used to have > + * virtual addresses in the paddr headers, whereas newer ones > + * (since kexec merge, around 2.6.14) correctly use physical > + * addresses. ("comparibility" -> "compatibility") I can't speak for IA64, but PPC kernels still have virtual addresses in the paddr fields. Accordingly, I would reword it like this: Some Linux kernels have virtual addresses in the paddr headers, and some correctly use physical addresses. Please CC xen-ppc-devel (and probably xen-ia64-devel) when you submit the patch so we can make sure this logic works for PPC. Thanks! -- Hollis Blanchard IBM Linux Technology Center