> I see your point, and it makes sense to keep KMI on stable branches, but > mainline I would like to have the proposed approach since all transports > use it to cycle on their sockets. > > So we could do a series with 2 patches: > - Patch 1 fixes the problem in all transports by checking the transport > in the callback (here we insert the fixes tag so we backport it to the > stable branches) > - Patch 2 moves the check in vsock_for_each_connected_socket() removing > it from callbacks so it is less prone to errors and we merge it only > mainline > > What do you think? > > Thanks, > Stefano > That sounds like a plan. Will submit a new series soon.
