diff options
Diffstat (limited to 'drivers/infiniband/sw/rxe/rxe_net.c')
| -rw-r--r-- | drivers/infiniband/sw/rxe/rxe_net.c | 81 |
1 files changed, 81 insertions, 0 deletions
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c index ac0183a2ff7a..0bd0902b11f7 100644 --- a/drivers/infiniband/sw/rxe/rxe_net.c +++ b/drivers/infiniband/sw/rxe/rxe_net.c @@ -20,6 +20,86 @@ static struct rxe_recv_sockets recv_sockets; +#ifdef CONFIG_DEBUG_LOCK_ALLOC +/* + * lockdep can detect false positive circular dependencies + * when there are user-space socket API users or in kernel + * users switching between a tcp and rdma transport. + * Maybe also switching between siw and rxe may cause + * problems as per default sockets are only classified + * by family and not by ip protocol. And there might + * be different locks used between the application + * and the low level sockets. + * + * Problems were seen with ksmbd.ko and cifs.ko, + * switching transports, use git blame to find + * more details. + */ +static struct lock_class_key rxe_recv_sk_key[2]; +static struct lock_class_key rxe_recv_slock_key[2]; +#endif /* CONFIG_DEBUG_LOCK_ALLOC */ + +static inline void rxe_reclassify_recv_socket(struct socket *sock) +{ +#ifdef CONFIG_DEBUG_LOCK_ALLOC + struct sock *sk = sock->sk; + + if (WARN_ON_ONCE(!sock_allow_reclassification(sk))) + return; + + switch (sk->sk_family) { + case AF_INET: + sock_lock_init_class_and_name(sk, + "slock-AF_INET-RDMA-RXE-RECV", + &rxe_recv_slock_key[0], + "sk_lock-AF_INET-RDMA-RXE-RECV", + &rxe_recv_sk_key[0]); + break; + case AF_INET6: + sock_lock_init_class_and_name(sk, + "slock-AF_INET6-RDMA-RXE-RECV", + &rxe_recv_slock_key[1], + "sk_lock-AF_INET6-RDMA-RXE-RECV", + &rxe_recv_sk_key[1]); + break; + default: + WARN_ON_ONCE(1); + return; + } + /* + * sock_lock_init_class_and_name() calls + * sk_owner_set(sk, THIS_MODULE); in order + * to make sure the referenced global + * variables rxe_recv_slock_key and + * rxe_recv_sk_key are not removed + * before the socket is closed. + * + * However this prevents rxe_net_exit() + * from being called and 'rmmod rdma_rxe' + * is refused because of the references. + * + * For the global sockets in recv_sockets, + * we are sure that rxe_net_exit() will call + * rxe_release_udp_tunnel -> udp_tunnel_sock_release. + * + * So we don't need the additional reference to + * our own (THIS_MODULE). + */ + sk_owner_put(sk); + /* + * We also call sk_owner_clear() otherwise + * sk_owner_put(sk) in sk_prot_free will + * fail, which is called via + * sk_free -> __sk_free -> sk_destruct + * and sk_destruct calls __sk_destruct + * directly or via call_rcu() + * so sk_prot_free() might be called + * after rxe_net_exit(). + */ + sk_owner_clear(sk); +#endif /* CONFIG_DEBUG_LOCK_ALLOC */ +} + static struct dst_entry *rxe_find_route4(struct rxe_qp *qp, struct net_device *ndev, struct in_addr *saddr, @@ -192,6 +272,7 @@ static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port, err = udp_sock_create(net, &udp_cfg, &sock); if (err < 0) return ERR_PTR(err); + rxe_reclassify_recv_socket(sock); tnl_cfg.encap_type = 1; tnl_cfg.encap_rcv = rxe_udp_encap_recv; |