On 13/05/2015 17:17, Fam Zheng wrote: >> > >> > It can be the topic of a separate series. But this patch brings a >> > false sense of security (either the blocker is unnecessary, or it >> > needs to last after bdrv_drain returns), so I think it should be >> > dropped. > Doesn't this let bdrv_drain_all return when virtio-blk-dataplane is having > high > workload, in places where you say "the blocker is unnecessary"?
Yes, you're right. Please document it in the commit message and the code, it's tricky. Paolo