Il 14/08/2013 13:21, Michael S. Tsirkin ha scritto:
On Wed, Aug 14, 2013 at 01:16:24PM +0200, Laszlo Ersek wrote:
On 08/14/13 12:54, Fabio Fantoni wrote:
Il 14/08/2013 11:56, Laszlo Ersek ha scritto:
On 08/14/13 11:19, Fabio Fantoni wrote:
Tried with qemu 1.4.2 and it works also with 4 gb of ram. This ram regression seems to be introduced with qemu 1.5, and there is another regression more critical with qemu 1.6.
Can you save qemu's stderr for the 1.5->1.6 regression?
Tried to add seabios debug options to domU xl configuration file...
device_model_args=["-chardev stdio,id=seabios","-device","isa-debugcon,iobase=0x402,chardev=seabios"]
But it doesn't boot
/var/log/xen/qemu-dm-W7.log qemu-system-x86_64: -chardev stdio,id=seabios: invalid option
That's probably because you forgot to break up this option into separate python list entries:
device_model_args=["-chardev", "stdio,id=seabios", "-device", "isa-debugcon,iobase=0x402,chardev=seabios"]
Laszlo
Thanks for your reply and sorry for my very stupid error. I retried with the correct parameters and on both cases (2 gb of ram working and 4gb of ram not working) I get only this one more line on qemu log: Start bios (version debian/1.7.3-1-1-ga76c6f1-dirty-20130813_122010-test)
The probable cause is that on xen, seabios doesn't generate some tables but it takes the static ones from hvmloader.
Without having any specifics in mind, maybe this can be attributed to the ACPI table move that affects both qemu and seabios, and hvmloader is the collateral damage. Probably stuff should be updated in qemu/seabios so that hvmloader be happy again. CC'ing Michael...
Of course I might as well be completely off.
Laszlo
Seems unlikely - the ACPI table move code is still in my tree.
I see on 1.6 changelog:
... QEMU can expose the full set of ACPI tables to the guest: See Features/ACPITableGeneration http://wiki.qemu.org/Features/ACPITableGeneration ...
Could be this the regression problem of 1.6?