On Tue, Oct 1, 2013 at 10:28 AM, Paolo Bonzini <pbonz...@redhat.com> wrote:
> Il 01/10/2013 17:12, Peter Maydell ha scritto: > >> > I just need some way to have different default firmware images for > >> > different pc machine type versions. Requests to considering other > >> > firmware loading needs (such as the two images needed for sPAPR) when > >> > doing this is reasonable. Blaming me for how inconistent qemu's > >> > firmware loading is today is not ok. > > You're defining new command line syntax -- it seems reasonable > > to think about what it ought to actually mean rather than just > > saying "it should mean the same collection of random stuff > > that -bios means". If you're going to do that you should just stick > > with -bios... > > What it ought to mean could be "the ROM that bootstraps the machine by > loading stuff from a drive (could be NAND, SD or a hardware component > such as virtio/IDE/SCSI)". But most machines sidestep this by loading > the kernel directly in RAM. > Firmware should not be a special concept in QEMU but it is. On real hardware, the firmware is stored in some sort of memory device that happens to be addressed at a specific location. This should be modelled as we do with -pflash as a block device and a ROM device. But we have a world today where firmware is special so we need to deal with that. And since the vast majority of systems have exactly one firmware blob, we should model the 99% case and let folks figure out how to deal with the remaining 1%. BTW, sPAPR is never a good reason to change how something is modelled. It is not a hardware system. It's a virtual machine. The concepts it has are not going to cleanly relate to other systems and RTAS is a very good example of this. RTAS has more in common with Linux's VDSO than anything on real hardware. Regards, Anthony Liguori > Paolo >
