net: fix lock_sock_bh/unlock_sock_bh
authorEric Dumazet <eric.dumazet@gmail.com>
Wed, 26 May 2010 19:20:18 +0000 (19:20 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 27 May 2010 07:30:53 +0000 (00:30 -0700)
commit8a74ad60a546b13bd1096b2a61a7a5c6fd9ae17c
tree3110e7e59883597b5d0f617e8507e15b8f965f3f
parenta56635a56f2afb3d22d9ce07e8f8d69537416b2d
net: fix lock_sock_bh/unlock_sock_bh

This new sock lock primitive was introduced to speedup some user context
socket manipulation. But it is unsafe to protect two threads, one using
regular lock_sock/release_sock, one using lock_sock_bh/unlock_sock_bh

This patch changes lock_sock_bh to be careful against 'owned' state.
If owned is found to be set, we must take the slow path.
lock_sock_bh() now returns a boolean to say if the slow path was taken,
and this boolean is used at unlock_sock_bh time to call the appropriate
unlock function.

After this change, BH are either disabled or enabled during the
lock_sock_bh/unlock_sock_bh protected section. This might be misleading,
so we rename these functions to lock_sock_fast()/unlock_sock_fast().

Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/sock.h
net/core/datagram.c
net/core/sock.c
net/ipv4/udp.c
net/ipv6/udp.c