Re: [patch] allow dynamic sizing the kvm vmm area on ia64

[Jes Sorensen <jes@sgi.com>]

Zhang, Xiantao wrote:
> Hi, Jes
>      I am in holidays these days, sorry for late response.  I think
 > you are saying vm area not vmm area.  In addition,  do you have plan
 > to add the part to dynamicly set the value of kvm_vm_data_shift ?
 > You know, it has no difference with before until this part is ready.
> Maybe we can dynamicly cacluate the value in upspace according to the
> amount of vcups, and transfer it to kernel at the stage of vm
> creation.  Another thing we need to consider is its value should
> follow tlb's page size, otherwise, it may result in other issues. 
> Thanks for your effort! 

Hi Xiantao,

Thanks for taking a look at this. Yes I do plan to add some logic for
setting this from QEMU. Basically what I talked to Hollis about was to
create some KVM_CONTEXT_CREATE or similar ioctl, where one requests a
maximum size of the KVM instance, like maximum possible number of vcpus,
memory etc.

Cheers,
Jes


> Xiantao
> 
> Jes Sorensen wrote:
>> Hi,
>>
>> This patch allows dynamically sizing the vmm area on ia64. I have left
>> it with the same size as we used to have with the constants, but it
>> seems to work going higher. I put in an artificial limit of 64 vcpus
>> for now anyway as scaling wise it doesn't work well going higher
>> anyway..... we'll get there :-)
>>
>> Cheers,
>> Jes