]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
net: sk_dst_cache RCUification
authorEric Dumazet <eric.dumazet@gmail.com>
Thu, 8 Apr 2010 23:03:29 +0000 (23:03 +0000)
committerDavid S. Miller <davem@davemloft.net>
Tue, 13 Apr 2010 08:41:33 +0000 (01:41 -0700)
commitb6c6712a42ca3f9fa7f4a3d7c40e3a9dd1fd9e03
tree42032b4978874e8ffcf6c851d13324b8c8c7c113
parent7a161ea92471087a1579239d7a58dd06eaa5601c
net: sk_dst_cache RCUification

With latest CONFIG_PROVE_RCU stuff, I felt more comfortable to make this
work.

sk->sk_dst_cache is currently protected by a rwlock (sk_dst_lock)

This rwlock is readlocked for a very small amount of time, and dst
entries are already freed after RCU grace period. This calls for RCU
again :)

This patch converts sk_dst_lock to a spinlock, and use RCU for readers.

__sk_dst_get() is supposed to be called with rcu_read_lock() or if
socket locked by user, so use appropriate rcu_dereference_check()
condition (rcu_read_lock_held() || sock_owned_by_user(sk))

This patch avoids two atomic ops per tx packet on UDP connected sockets,
for example, and permits sk_dst_lock to be much less dirtied.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/dst.h
include/net/ip6_route.h
include/net/sock.h
net/core/dev.c
net/core/sock.c
net/dccp/timer.c
net/decnet/af_decnet.c
net/ipv4/af_inet.c
net/ipv4/tcp_input.c
net/ipv4/tcp_timer.c
net/ipv6/ipv6_sockglue.c