clogd + dm-log-clustered: Improve communication error handling (K <-> U)
Connector (netlink) is an unreliable protocol. From time to time, I get
a connector header in userspace that has all the right fields (seq #, size
of data, etc), but the data is all zeros. When this happens, I use the seq
number in the header to request the kernel to resend the request. This
should avoid those nasty 'Invalid request' messages... which are then
followed by 'Request timed out' (because the invalid request was ignored).