Re: [PATCH v10 6/6] migration/dirtyrate: implement dirty-ring dirtyrate calculation

[Peter Xu]

On Tue, Jun 29, 2021 at 09:04:07PM +0800, Hyman wrote:
> 
> 
> 在 2021/6/29 7:05, Peter Xu 写道:
> > On Mon, Jun 28, 2021 at 12:27:46AM +0800, huangy81@chinatelecom.cn wrote:
> > > +static void dirtyrate_global_dirty_log_stop(void)
> > > +{
> > > +    qemu_mutex_lock_iothread();
> > 
> > One thing I missed: maybe we want to sync dirty before stop, so the value 
> > will
> > be more accurate.  Perhaps a call to memory_global_dirty_log_sync() here?
> seems good, if the dirty ring reaper is sleeping, we can sync log by
> ourselves, making dirtyrate more accurate.

Btw, I commented here because there has the bql so it's easier; please still
remember to record the pages after sync is done!  Otherwise it'll be an effort
of no real use (which I think that's missing in your latest repost..).

-- 
Peter Xu