From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1LsXqm-00085J-Pq for qemu-devel@nongnu.org; Sat, 11 Apr 2009 03:42:04 -0400 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1LsXqi-00084u-7b for qemu-devel@nongnu.org; Sat, 11 Apr 2009 03:42:04 -0400 Received: from [199.232.76.173] (port=52897 helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1LsXqh-00084m-V7 for qemu-devel@nongnu.org; Sat, 11 Apr 2009 03:42:00 -0400 Received: from mx2.redhat.com ([66.187.237.31]:44230) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1LsXqh-0004Lj-9h for qemu-devel@nongnu.org; Sat, 11 Apr 2009 03:41:59 -0400 Message-ID: <49E049AE.2090508@redhat.com> Date: Fri, 10 Apr 2009 21:41:34 -1000 From: Zachary Amsden MIME-Version: 1.0 Subject: Re: [Qemu-devel] Re: [PATCH] register a single area for vga bios and option roms References: <1239292380-10032-1-git-send-email-glommer@redhat.com> <49DE34D2.3000803@us.ibm.com> In-Reply-To: <49DE34D2.3000803@us.ibm.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Reply-To: qemu-devel@nongnu.org List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: jan.kiszka@siemens.com, Glauber Costa Anthony Liguori wrote: > Glauber Costa wrote: >> Those guys are not different in nature. They're all roms, >> not blessed with the graces of being written to. So there's >> not need to issue multiple requests to memory registration areas: >> just treat them as brothers, and put them all in the same >> region. >> >> It also has the nice side effect of improving the loading code >> a little bit. >> >> Signed-off-by: Glauber Costa >> > > The vga bios is just an option rom, right? So why can't we treat the > vga bios as an option rom and eliminate the specific code for dealing > with VGA bios. > > The real problem here is that we should be registering the option rom > space once and doing that in page boundaries. Um... are we sure about this? First off, option roms can be written to, it's part of the PnP option ROM spec.. they are to be backed in RAM and the init code will actually write to the ROM to resize it (allowing it to jettison startup / init code); they get remapped read-only again later. Will registering the roms as one physical memory region prevent this possibility? Sorry I am still naive and coming up to speed on qemu if this is a naive question. Second thing, option roms are not necessarily on page boundaries, technically they are mapped in chunks, which are 512 byte blocks, at the discretion of the BIOS / chipset. This is mostly a non-issue as all modern BIOS will map them on page boundaries but worth noting. Zach