[PATCH v2 net-next 0/5] tcp: listener refactor part 16

[Eric Dumazet <edumazet@google.com>]

A CONFIG_PROVE_RCU=y build revealed an RCU splat I had to fix.

I added const qualifiers to various md5 methods, as I expect
to call them on behalf of request sock traffic even if
the listener socket is not locked. This seems ok, but adding
const makes the contract clearer. Note a good reduction
of code size thanks to request/establish sockets convergence.

Eric Dumazet (5):
  tcp: md5: fix rcu lockdep splat
  tcp: use C99 initializers in new_state[]
  tcp: md5: input path is run under rcu protected sections
  tcp: md5: remove request sock argument of calc_md5_hash()
  tcp: md5: get rid of tcp_v[46]_reqsk_md5_lookup()

 include/net/tcp.h     | 37 +++++++++++++++++------------------
 net/ipv4/tcp.c        | 25 ++++++++++++------------
 net/ipv4/tcp_ipv4.c   | 53 +++++++++++++++------------------------------------
 net/ipv4/tcp_output.c | 24 ++++++++++++-----------
 net/ipv6/tcp_ipv6.c   | 50 ++++++++++++++----------------------------------
 5 files changed, 72 insertions(+), 117 deletions(-)

-- 
2.2.0.rc0.207.ga3a616c