Masami Hiramatsu  , Google [off-list ref] wrote:

Good catch! Hmm, previously we guaranteed that the find_first_fprobe_node()
must be called under rcu read locked or fprobe_mutex locked, so that the
node list should not be changed. But according to the comment of
rhltable_lookup(), we need to lock the rcu_read_lock() around that.

Just as is the case for RCU in general, rcu read locks are unnecessary
for rhashtable if you already hold the write-side locks.

Cheers,
-- 
Email: Herbert Xu [off-list ref]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt