From: Eric Dumazet <edumazet@google.com>
Date: Tue, 17 May 2016 17:44:05 -0700
Four layers using TCP stack were assuming sk_callback_lock could
be locked using read_lock() in their handlers because TCP stack
was running with BH disabled.
This is no longer the case. Since presumably the rest could
also depend on BH being disabled, just use read_lock_bh().
Then each layer might consider switching to RCU protection
and no longer depend on BH.
Series applied, thanks Eric.