Il 30/03/2013 14:20, Kevin O'Connor ha scritto:
On Fri, Mar 29, 2013 at 02:49:12PM +0100, Paolo Bonzini wrote:
Il 29/03/2013 14:33, Kevin O'Connor ha scritto:
On Fri, Mar 29, 2013 at 04:18:44PM +0800, Hu Tao wrote:
pvpanic device is used to notify host(qemu) when guest panic happens.
Thanks. However, we're planning a move of ACPI tables from SeaBIOS to QEMU. I think this should wait until after the move.
The device should be in QEMU 1.5, and the SSDT probably will still be in SeaBIOS by then (and might even be the last to move, since it's quite complex and dynamic). I don't think it is fair to block this patch on those grounds...
What is the user visible impact of not having a panic device?
Management cannot restart or dump the guest after a panic, and cannot log that a panic happened (as opposed to a random watchdog event).
My main concern is that the patch creates a new fw_cfg channel between qemu and seabios thats sole purpose is to alter the OS visible ACPI tables. These types of QEMU->SeaBIOS interfaces are fragile and are (in sum) quite complex.
Yes, this is true.
Paolo