Re: [Qemu-devel] Re: [Qemu-commits] [COMMIT 3086844] Instead of writing a zero page, madvise it away

[Anthony Liguori <aliguori@us.ibm.com>]

Avi Kivity wrote:
> You mean, a NONZERO bit which is cleared by ballooning and set on any 
> write.  This will work naturally with the qemu dirty bytemap.

Yes.

>>
>> For KVM, we would have to enable dirty tracking always to keep 
>> ZERO_DIRTY up to date.  Since write faults are going to happen anyway 
>> at start up, perhaps the cost of doing this wouldn't be so bad?
>
> You need to do this on the source node.  Unfortunately, there's no way 
> to initialize the values racelessly when you start live migration 
> without introducing a new ioctl.  I'd like a more general solution 
> rather than something that targets this specific problem.

I'm saying, always enable dirty tracking from start-of-day with KVM.  
Then the QEMU dirty bitmap is always accurate.  The trick is to never 
start resetting it until you need to do live migration.

The idea being that once the dirty bits have been set, the overhead 
(hopefully) should be zero.

-- 
Regards,

Anthony Liguori