Il 21/03/2013 14:12, Michael S. Tsirkin ha scritto:
On Thu, Mar 21, 2013 at 01:04:35PM +0000, David Woodhouse wrote:
On Thu, 2013-03-21 at 13:56 +0100, Laszlo Ersek wrote:
- for an earlier qemu, the option must be set,
- for a later qemu the option must be clear && (no -acpitable switch must be specified on the qemu cmldine || one -acpitable switch must load a MADT)
Hm, that sounds like it won't be possible to build one version of SeaBIOS that works for *both* old and new qemu. That doesn't seem like a great idea. I'd prefer something like:
- If Qemu provides the 'core' ACPI tables (i.e. not just SSDT) then SeaBIOS uses them.
- Otherwise, it makes its own.
-- dwmw2
It might simplify life for someone bisecting qemu as you don't need to rebuild seabios after each bisect but is this really a common workflow?
It would help me working on RHEL's qemu-kvm, which I often do on a Fedora box.
The bios.bin in the RHEL qemu-kvm tree is obsolete (like Gerd showed for Fedora, RHEL's bios.bin comes from a separate package), and using whatever happens to be in /usr or in my local SeaBIOS clone is the simplest thing to do. I'd like this not to break.
Paolo
Anyway, I am not against such runtime flags.
If we add to this an option to build a minimal BIOS that only works with the new QEMU, do we have a deal?
Then the plan is to make progress and apply patches step by step without deciding on the detection interface first. Before QEMU is switched to the new configuration, we'll add the runtime thing for the benefit of developers that bisect.
Makes sense?