diff options
author | Tariq Saeed <tariq.x.saeed@oracle.com> | 2014-04-03 23:47:11 +0200 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2014-04-04 01:20:56 +0200 |
commit | da8ded405de74de4b189d1128f4c102d06328c29 (patch) | |
tree | f89ec00e4c725703aeba00e741f9e5203b9b7594 /fs/ocfs2/cluster/tcp.c | |
parent | ocfs2: rollback alloc_dinode counts when ocfs2_block_group_set_bits() failed (diff) | |
download | linux-da8ded405de74de4b189d1128f4c102d06328c29.tar.xz linux-da8ded405de74de4b189d1128f4c102d06328c29.zip |
ocfs2/o2net: o2net_listen_data_ready should do nothing if socket state is not TCP_LISTEN
Orabug: 17330860
When accepting an incomming connection o2net_accept_one clones a child
data socket from the parent listening socket. It then proceeds to setup
the child with callback o2net_data_ready() and sk_user_data to NULL. If
data arrives in this window, o2net_listen_data_ready will be called with
some non-deterministic value in sk_user_data (not inherited). We panic
when we page fault on sk_user_data -- in parent it is
sock_def_readable().
The fix is to recognize that this is a data socket being set up by
looking at the socket state and do nothing.
Signed-off-by: Tariq Saseed <tariq.x.saeed@oracle.com>
Signed-off-by: Srinivas Eeda <srinivas.eeda@oracle.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'fs/ocfs2/cluster/tcp.c')
-rw-r--r-- | fs/ocfs2/cluster/tcp.c | 22 |
1 files changed, 17 insertions, 5 deletions
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c index b4ab371d46d9..eb649d23a4de 100644 --- a/fs/ocfs2/cluster/tcp.c +++ b/fs/ocfs2/cluster/tcp.c @@ -1964,18 +1964,30 @@ static void o2net_listen_data_ready(struct sock *sk, int bytes) goto out; } - /* ->sk_data_ready is also called for a newly established child socket - * before it has been accepted and the acceptor has set up their - * data_ready.. we only want to queue listen work for our listening - * socket */ + /* This callback may called twice when a new connection + * is being established as a child socket inherits everything + * from a parent LISTEN socket, including the data_ready cb of + * the parent. This leads to a hazard. In o2net_accept_one() + * we are still initializing the child socket but have not + * changed the inherited data_ready callback yet when + * data starts arriving. + * We avoid this hazard by checking the state. + * For the listening socket, the state will be TCP_LISTEN; for the new + * socket, will be TCP_ESTABLISHED. Also, in this case, + * sk->sk_user_data is not a valid function pointer. + */ + if (sk->sk_state == TCP_LISTEN) { mlog(ML_TCP, "bytes: %d\n", bytes); queue_work(o2net_wq, &o2net_listen_work); + } else { + ready = NULL; } out: read_unlock(&sk->sk_callback_lock); - ready(sk, bytes); + if (ready != NULL) + ready(sk, bytes); } static int o2net_open_listening_sock(__be32 addr, __be16 port) |