[Severity: Medium]
Can a transmitter on another CPU still cache the route to the old remote
after this reset?
Yes, as answered for v2: dst_cache tracks freshness in jiffies, so a
transmit that built its key before the reset can refill the cache in a
later jiffy. vxlan_fdb_replace() has the same window. Closing it means
replacing the remote, or its dst_cache, through RCU instead of updating
it in place, which would also change vxlan_fdb_replace(). I would
rather do that, if wanted, as a separate net-next change. Without this
patch every move uses the stale route; with it only this race does.
[Severity: Medium]
Can concurrent snoops of the same MAC leave remote_ip and remote_ifindex
inconsistent?
Yes. v6 updates the remote under vxlan->hash_lock and checks the entry
again once the lock is held, so concurrent snoops are serialized with
each other and with vxlan_fdb_replace(). v6 also addresses the three
notes on patch 2.
pw-bot: cr