On Mon, Mar 03, 2025 at 04:17:53PM +0530, Prasad Pandit wrote: > * I think we (you, me, Peter) are all looking at things differently. > - In my view Patch-2 is the minimal change _required_ to enable > multifd & postcopy. In your view we are _revamping_ channel discovery > parts while _sneaking_ in a feature of enabling multifd & postcopy > together. > - In my view Patch-5 in this series is an isolated change because > it adds a new migration command to allow multifd threads sync from > source side. But Peter thinks without that 'flush and sync' Patch-2 is > incomplete, so we should merge it back there.
Just to mention, my suggestion does not conflict with splitting patch 2, as long as you keep every patch complete on its own. Patch 5 needs to be squashed to either patch 2 or a split patch out of patch 2, because current patch 2 (or any possible way to split it into smaller ones, then one of them which enables the feature) is buggy. Thanks, -- Peter Xu
