Re: [PATCH v5] KVM: VMX: Execute WBINVD to keep data consistency with assigned devices

[Avi Kivity <avi@redhat.com>]

On 06/29/2010 04:35 PM, Marcelo Tosatti wrote:
>>>
>>> work_on_cpu() loop instead of smp_call_function_many(), to avoid executing
>>> wbinvd with interrupts disabled.
>>>        
>> Why?  wbinvd is not interruptible.
>>      
> Right. But still, smp_call_function_many() is going to busy-spin until
> the target CPUs finish their work, while work_on_cpu() will schedule.
>
>    

Good point.  So in the worst case we double the hit.

> Also the IPI request has to handled immediately, bypassing the
> scheduler.
>    

We're screwed on that point in any case, once wbinvd starts the 
scheduler is bypassed.

But work_on_cpu() has its own problems.  It creates a kthread (perhaps 
not too bad).  Maybe queue_work_on()?

btw, smp_call_function_many() appears not to run the function on the 
current cpu.

-- 
error compiling committee.c: too many arguments to function