From: Horms Subject: Re: [PATCH 01/04] Kexec / Kdump: Generic code Date: Wed, 18 Oct 2006 17:13:03 +0900 Message-ID: <20061018081302.GC2958@verge.net.au> References: <20061018010003.GC11422@verge.net.au> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Keir Fraser Cc: Ian Pratt , Kazuo Moriwaka , xen-devel@lists.xensource.com, Akio Takebe , magnus.damm@gmail.com, Isaku Yamahata , Magnus Damm List-Id: xen-devel@lists.xenproject.org On Wed, Oct 18, 2006 at 08:15:45AM +0100, Keir Fraser wrote: > On 18/10/06 2:00 am, "Horms" wrote: > > >> Yes please. This isn't performance critical. Folding will make the interface > >> better match other hypercalls. > > > > The main thing that I am concerned about is, that in the > > case that dom0 panics, and we are asking the hypercall to take a dump > > for us, do we really want a copy_from guest in that path? > > It seems to me that its best to have that path as simple as possible. > > But I do argree that it comes at the expense of making the interface > > a bit unclean. > > We need to assume that kernel address space hasn't been too badly > compromised, or we couldn't execute as far as making the hypercall. A failed > copy_from_guest() in the crash hypercall could simply cause you to use the > guest_cpu_user_regs() instead. So you still crash, but the crash site is now > the hypercall call site. Indeed, that is true. I do think there is some value in avoiding guest_cpu_user_regs(), but its theoretical at best. Magnus and I will fold the arguments as you requested. -- Horms H: http://www.vergenet.net/~horms/ W: http://www.valinux.co.jp/en/