Re: [patch 3/5] KVM: hypercall batching (v2)

[Avi Kivity <avi@qumranet.com>]

Marcelo Tosatti wrote:
> On Thu, Feb 21, 2008 at 08:30:04PM +0200, Avi Kivity wrote:
>   
>>> Perhaps you want to move that enforcement to the host.
>>>
>>> This allows batching of future hypercalls (if appropriate) to be easy.
>>>
>>>  
>>>       
>> I'm still uneasy about it, though I have no rational reasons left now.
>>
>> Oh, there is one: with a MMU_OP hypercall you can take the mmu spinlock 
>> once per batch (dropping it once in a while to let another vcpu make 
>> progress or to inject an interrupt).
>>     
>
> emulator_write_phys() needs to do blocking work for each pte
> (mmu_guess_page_from_pte), so that optimization would need quite some
> work (separate mmu_guess_page_from_pte from kvm_mmu_pte_write).
>
> And you can do this optimization even without MMU_OP, just check in
> advance how many operations will take the mmu lock, do the non-blocking
> part of them, and then manipulate the mmu lock protected shadow data.
>
> Also, holding the spinlock for a longer period is not necessarily an
> improvement (with the slots_lock it clearly is because there is no write
> contention).
>
>   

Yes, you are right.

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/