Re: [Qemu-devel] [patch 2/2] QEMU BOCHS bios patches to use maxcpus value.

[Filip Navara <filip.navara@gmail.com>]

On Tue, Jul 14, 2009 at 12:16 PM, Alexander Graf<agraf@suse.de> wrote:
>
> On 14.07.2009, at 11:32, Gleb Natapov wrote:
>
>> On Tue, Jul 14, 2009 at 11:21:53AM +0200, Filip Navara wrote:
>>>
>>> On Tue, Jul 14, 2009 at 10:38 AM, Jes Sorensen<jes@sgi.com> wrote:
>>>>
>>>> On 07/09/2009 11:57 PM, Anthony Liguori wrote:
>>>>>
>>>>> These changes make my Ubuntu server guest very unhappy.  I get a bunch
>>>>> of messages about "Not responding." on startup.
>>>>>
>>>>> If nothing else, maxcpus ==smp_cpus under QEMU because we don't do CPU
>>>>> hotplug (and I don't think we should).
>>>>
>>>> Anthony,
>>>>
>>>> Sorry haven't gotten back to you earlier as I was on vacation. Are you
>>>> saying the Ubuntu kernel doesn't like having more CPU entries in the
>>>> ACPI table than it actually boots on?
>>>>
>>>> Does the same guest boot using an older KVM setup? Curious since it does
>>>> have the larger CPU table in the DSDT.
>>>>
>>>> Cheers,
>>>> Jes
>>>>
>>>
>>> BTW, many other guests complain when ACPI describes more processors
>>> than actually present in machine. That's why I implemented the dynamic
>>> DSDT generation in Bochs BIOS in the first place. One that comes to
>>> mind is MacOS X, or the Darwin kernel respectively.
>>>
>> There is nothing wrong in describing more processors than actually
>> present. The disable flag is defined by ACPI for a reason. My real HW
>> IBM server does this.
>
> Yeah, last time I tried MacOS X was happy with more CPU descriptions than
> actual CPUs too as long as they were in disabled state. Has anything changed
> there I should know about?
>
> Alex

Not really... it just spits warnings that are not seen unless you
explicitly ask for complete debug/io logs (not sure which one it is
in). In any case the disabled flag was not set in QEMU when I was
debugging the code.

Best regards,
Filip Navara