Re: [PATCH 3/4] KVM: Switch to srcu-less get_dirty_log()

[Takuya Yoshikawa <takuya.yoshikawa@gmail.com>]

Avi Kivity <avi@redhat.com> wrote:

> > We cannot get a complete snapshot without mmu_lock; if the guest faults on
> > the Nth page during xchg'ing i = 1, 2, ... then the i = N alone will
> > become newer.
> 
> Ah, so there is no data corruption (missed dirty bits), just the display
> is updated inconsistently?

Yes, no data corruption and just a matter of ... feeling.

The basic rule I wrote in the comment in this patch should be enough to
not lose dirty bits.

> I don't think we can get a consistent snapshot anyway, since the guest
> can update the framebuffer while userspace is processing it.

Yes, nothing will be broken.  I was just not sure what the API should promise
to the userspace.

To some extent the inconsistency may be felt more than before.

> Understood.  I don't think we can get a consistent vga snapshot without
> stopping the guest, and even then, it depends on how the guest updates
> the framebuffer.

OKay, I will not care about this from now.

> > So I want to see how much improvement the proposed API can achieve.
> >
> > I thought this might be a good GSoC project but ...
> 
> It may be too involved for GSoC, the issues are difficult.

I am now checking QEMU code closely - rather readable than before!

Though I can make experimental KVM patches for the student, just changing
QEMU's live migration code seems to be difficult - but not sure now.

Anyway, we should try this before deciding the new API: I mean if I cannot
find anyone who want to try this, no need to be a student, I may do instead
at some point.

	Takuya