aboutsummaryrefslogtreecommitdiff
path: root/net/rds
diff options
context:
space:
mode:
Diffstat (limited to 'net/rds')
-rw-r--r--net/rds/af_rds.c20
-rw-r--r--net/rds/bind.c1
-rw-r--r--net/rds/cong.c10
-rw-r--r--net/rds/connection.c32
-rw-r--r--net/rds/ib.c26
-rw-r--r--net/rds/ib_cm.c1
-rw-r--r--net/rds/rds.h17
-rw-r--r--net/rds/send.c37
-rw-r--r--net/rds/tcp.c113
-rw-r--r--net/rds/tcp.h1
-rw-r--r--net/rds/tcp_connect.c2
-rw-r--r--net/rds/tcp_recv.c8
-rw-r--r--net/rds/tcp_send.c5
-rw-r--r--net/rds/threads.c20
14 files changed, 196 insertions, 97 deletions
diff --git a/net/rds/af_rds.c b/net/rds/af_rds.c
index b405f77d664c..744c637c86b0 100644
--- a/net/rds/af_rds.c
+++ b/net/rds/af_rds.c
@@ -137,27 +137,27 @@ static int rds_getname(struct socket *sock, struct sockaddr *uaddr,
/*
* RDS' poll is without a doubt the least intuitive part of the interface,
- * as POLLIN and POLLOUT do not behave entirely as you would expect from
+ * as EPOLLIN and EPOLLOUT do not behave entirely as you would expect from
* a network protocol.
*
- * POLLIN is asserted if
+ * EPOLLIN is asserted if
* - there is data on the receive queue.
* - to signal that a previously congested destination may have become
* uncongested
* - A notification has been queued to the socket (this can be a congestion
* update, or a RDMA completion).
*
- * POLLOUT is asserted if there is room on the send queue. This does not mean
+ * EPOLLOUT is asserted if there is room on the send queue. This does not mean
* however, that the next sendmsg() call will succeed. If the application tries
* to send to a congested destination, the system call may still fail (and
* return ENOBUFS).
*/
-static unsigned int rds_poll(struct file *file, struct socket *sock,
+static __poll_t rds_poll(struct file *file, struct socket *sock,
poll_table *wait)
{
struct sock *sk = sock->sk;
struct rds_sock *rs = rds_sk_to_rs(sk);
- unsigned int mask = 0;
+ __poll_t mask = 0;
unsigned long flags;
poll_wait(file, sk_sleep(sk), wait);
@@ -167,22 +167,22 @@ static unsigned int rds_poll(struct file *file, struct socket *sock,
read_lock_irqsave(&rs->rs_recv_lock, flags);
if (!rs->rs_cong_monitor) {
- /* When a congestion map was updated, we signal POLLIN for
+ /* When a congestion map was updated, we signal EPOLLIN for
* "historical" reasons. Applications can also poll for
* WRBAND instead. */
if (rds_cong_updated_since(&rs->rs_cong_track))
- mask |= (POLLIN | POLLRDNORM | POLLWRBAND);
+ mask |= (EPOLLIN | EPOLLRDNORM | EPOLLWRBAND);
} else {
spin_lock(&rs->rs_lock);
if (rs->rs_cong_notify)
- mask |= (POLLIN | POLLRDNORM);
+ mask |= (EPOLLIN | EPOLLRDNORM);
spin_unlock(&rs->rs_lock);
}
if (!list_empty(&rs->rs_recv_queue) ||
!list_empty(&rs->rs_notify_queue))
- mask |= (POLLIN | POLLRDNORM);
+ mask |= (EPOLLIN | EPOLLRDNORM);
if (rs->rs_snd_bytes < rds_sk_sndbuf(rs))
- mask |= (POLLOUT | POLLWRNORM);
+ mask |= (EPOLLOUT | EPOLLWRNORM);
read_unlock_irqrestore(&rs->rs_recv_lock, flags);
/* clear state any time we wake a seen-congested socket */
diff --git a/net/rds/bind.c b/net/rds/bind.c
index 75d43dc8e96b..5aa3a64aa4f0 100644
--- a/net/rds/bind.c
+++ b/net/rds/bind.c
@@ -114,6 +114,7 @@ static int rds_add_bound(struct rds_sock *rs, __be32 addr, __be16 *port)
rs, &addr, (int)ntohs(*port));
break;
} else {
+ rs->rs_bound_addr = 0;
rds_sock_put(rs);
ret = -ENOMEM;
break;
diff --git a/net/rds/cong.c b/net/rds/cong.c
index 8398fee7c866..63da9d2f142d 100644
--- a/net/rds/cong.c
+++ b/net/rds/cong.c
@@ -219,7 +219,11 @@ void rds_cong_queue_updates(struct rds_cong_map *map)
spin_lock_irqsave(&rds_cong_lock, flags);
list_for_each_entry(conn, &map->m_conn_list, c_map_item) {
- if (!test_and_set_bit(0, &conn->c_map_queued)) {
+ struct rds_conn_path *cp = &conn->c_path[0];
+
+ rcu_read_lock();
+ if (!test_and_set_bit(0, &conn->c_map_queued) &&
+ !rds_destroy_pending(cp->cp_conn)) {
rds_stats_inc(s_cong_update_queued);
/* We cannot inline the call to rds_send_xmit() here
* for two reasons (both pertaining to a TCP transport):
@@ -235,9 +239,9 @@ void rds_cong_queue_updates(struct rds_cong_map *map)
* therefore trigger warnings.
* Defer the xmit to rds_send_worker() instead.
*/
- queue_delayed_work(rds_wq,
- &conn->c_path[0].cp_send_w, 0);
+ queue_delayed_work(rds_wq, &cp->cp_send_w, 0);
}
+ rcu_read_unlock();
}
spin_unlock_irqrestore(&rds_cong_lock, flags);
diff --git a/net/rds/connection.c b/net/rds/connection.c
index 7ee2d5d68b78..2da3176bf792 100644
--- a/net/rds/connection.c
+++ b/net/rds/connection.c
@@ -220,8 +220,13 @@ static struct rds_connection *__rds_conn_create(struct net *net,
is_outgoing);
conn->c_path[i].cp_index = i;
}
- ret = trans->conn_alloc(conn, gfp);
+ rcu_read_lock();
+ if (rds_destroy_pending(conn))
+ ret = -ENETDOWN;
+ else
+ ret = trans->conn_alloc(conn, GFP_ATOMIC);
if (ret) {
+ rcu_read_unlock();
kfree(conn->c_path);
kmem_cache_free(rds_conn_slab, conn);
conn = ERR_PTR(ret);
@@ -230,8 +235,8 @@ static struct rds_connection *__rds_conn_create(struct net *net,
rdsdebug("allocated conn %p for %pI4 -> %pI4 over %s %s\n",
conn, &laddr, &faddr,
- trans->t_name ? trans->t_name : "[unknown]",
- is_outgoing ? "(outgoing)" : "");
+ strnlen(trans->t_name, sizeof(trans->t_name)) ? trans->t_name :
+ "[unknown]", is_outgoing ? "(outgoing)" : "");
/*
* Since we ran without holding the conn lock, someone could
@@ -283,6 +288,7 @@ static struct rds_connection *__rds_conn_create(struct net *net,
}
}
spin_unlock_irqrestore(&rds_conn_lock, flags);
+ rcu_read_unlock();
out:
return conn;
@@ -403,6 +409,11 @@ static void rds_conn_path_destroy(struct rds_conn_path *cp)
if (cp->cp_xmit_rm)
rds_message_put(cp->cp_xmit_rm);
+ WARN_ON(delayed_work_pending(&cp->cp_send_w));
+ WARN_ON(delayed_work_pending(&cp->cp_recv_w));
+ WARN_ON(delayed_work_pending(&cp->cp_conn_w));
+ WARN_ON(work_pending(&cp->cp_down_w));
+
cp->cp_conn->c_trans->conn_free(cp->cp_transport_data);
}
@@ -424,7 +435,6 @@ void rds_conn_destroy(struct rds_connection *conn)
"%pI4\n", conn, &conn->c_laddr,
&conn->c_faddr);
- conn->c_destroy_in_prog = 1;
/* Ensure conn will not be scheduled for reconnect */
spin_lock_irq(&rds_conn_lock);
hlist_del_init_rcu(&conn->c_hash_node);
@@ -445,7 +455,6 @@ void rds_conn_destroy(struct rds_connection *conn)
*/
rds_cong_remove_conn(conn);
- put_net(conn->c_net);
kfree(conn->c_path);
kmem_cache_free(rds_conn_slab, conn);
@@ -684,10 +693,13 @@ void rds_conn_path_drop(struct rds_conn_path *cp, bool destroy)
{
atomic_set(&cp->cp_state, RDS_CONN_ERROR);
- if (!destroy && cp->cp_conn->c_destroy_in_prog)
+ rcu_read_lock();
+ if (!destroy && rds_destroy_pending(cp->cp_conn)) {
+ rcu_read_unlock();
return;
-
+ }
queue_work(rds_wq, &cp->cp_down_w);
+ rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(rds_conn_path_drop);
@@ -704,9 +716,15 @@ EXPORT_SYMBOL_GPL(rds_conn_drop);
*/
void rds_conn_path_connect_if_down(struct rds_conn_path *cp)
{
+ rcu_read_lock();
+ if (rds_destroy_pending(cp->cp_conn)) {
+ rcu_read_unlock();
+ return;
+ }
if (rds_conn_path_state(cp) == RDS_CONN_DOWN &&
!test_and_set_bit(RDS_RECONNECT_PENDING, &cp->cp_flags))
queue_delayed_work(rds_wq, &cp->cp_conn_w, 0);
+ rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(rds_conn_path_connect_if_down);
diff --git a/net/rds/ib.c b/net/rds/ib.c
index 36dd2099048a..50a88f3e7e39 100644
--- a/net/rds/ib.c
+++ b/net/rds/ib.c
@@ -48,6 +48,7 @@
static unsigned int rds_ib_mr_1m_pool_size = RDS_MR_1M_POOL_SIZE;
static unsigned int rds_ib_mr_8k_pool_size = RDS_MR_8K_POOL_SIZE;
unsigned int rds_ib_retry_count = RDS_IB_DEFAULT_RETRY_COUNT;
+static atomic_t rds_ib_unloading;
module_param(rds_ib_mr_1m_pool_size, int, 0444);
MODULE_PARM_DESC(rds_ib_mr_1m_pool_size, " Max number of 1M mr per HCA");
@@ -301,13 +302,11 @@ static int rds_ib_conn_info_visitor(struct rds_connection *conn,
memset(&iinfo->dst_gid, 0, sizeof(iinfo->dst_gid));
if (rds_conn_state(conn) == RDS_CONN_UP) {
struct rds_ib_device *rds_ibdev;
- struct rdma_dev_addr *dev_addr;
ic = conn->c_transport_data;
- dev_addr = &ic->i_cm_id->route.addr.dev_addr;
- rdma_addr_get_sgid(dev_addr, (union ib_gid *) &iinfo->src_gid);
- rdma_addr_get_dgid(dev_addr, (union ib_gid *) &iinfo->dst_gid);
+ rdma_read_gids(ic->i_cm_id, (union ib_gid *)&iinfo->src_gid,
+ (union ib_gid *)&iinfo->dst_gid);
rds_ibdev = ic->rds_ibdev;
iinfo->max_send_wr = ic->i_send_ring.w_nr;
@@ -347,7 +346,8 @@ static int rds_ib_laddr_check(struct net *net, __be32 addr)
/* Create a CMA ID and try to bind it. This catches both
* IB and iWARP capable NICs.
*/
- cm_id = rdma_create_id(&init_net, NULL, NULL, RDMA_PS_TCP, IB_QPT_RC);
+ cm_id = rdma_create_id(&init_net, rds_rdma_cm_event_handler,
+ NULL, RDMA_PS_TCP, IB_QPT_RC);
if (IS_ERR(cm_id))
return PTR_ERR(cm_id);
@@ -379,8 +379,23 @@ static void rds_ib_unregister_client(void)
flush_workqueue(rds_wq);
}
+static void rds_ib_set_unloading(void)
+{
+ atomic_set(&rds_ib_unloading, 1);
+}
+
+static bool rds_ib_is_unloading(struct rds_connection *conn)
+{
+ struct rds_conn_path *cp = &conn->c_path[0];
+
+ return (test_bit(RDS_DESTROY_PENDING, &cp->cp_flags) ||
+ atomic_read(&rds_ib_unloading) != 0);
+}
+
void rds_ib_exit(void)
{
+ rds_ib_set_unloading();
+ synchronize_rcu();
rds_info_deregister_func(RDS_INFO_IB_CONNECTIONS, rds_ib_ic_info);
rds_ib_unregister_client();
rds_ib_destroy_nodev_conns();
@@ -414,6 +429,7 @@ struct rds_transport rds_ib_transport = {
.flush_mrs = rds_ib_flush_mrs,
.t_owner = THIS_MODULE,
.t_name = "infiniband",
+ .t_unloading = rds_ib_is_unloading,
.t_type = RDS_TRANS_IB
};
diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c
index 80fb6f63e768..eea1d8611b20 100644
--- a/net/rds/ib_cm.c
+++ b/net/rds/ib_cm.c
@@ -117,6 +117,7 @@ void rds_ib_cm_connect_complete(struct rds_connection *conn, struct rdma_cm_even
&conn->c_laddr, &conn->c_faddr,
RDS_PROTOCOL_MAJOR(conn->c_version),
RDS_PROTOCOL_MINOR(conn->c_version));
+ set_bit(RDS_DESTROY_PENDING, &conn->c_path[0].cp_flags);
rds_conn_destroy(conn);
return;
} else {
diff --git a/net/rds/rds.h b/net/rds/rds.h
index c349c71babff..7301b9b01890 100644
--- a/net/rds/rds.h
+++ b/net/rds/rds.h
@@ -88,6 +88,7 @@ enum {
#define RDS_RECONNECT_PENDING 1
#define RDS_IN_XMIT 2
#define RDS_RECV_REFILL 3
+#define RDS_DESTROY_PENDING 4
/* Max number of multipaths per RDS connection. Must be a power of 2 */
#define RDS_MPATH_WORKERS 8
@@ -139,8 +140,7 @@ struct rds_connection {
__be32 c_faddr;
unsigned int c_loopback:1,
c_ping_triggered:1,
- c_destroy_in_prog:1,
- c_pad_to_32:29;
+ c_pad_to_32:30;
int c_npaths;
struct rds_connection *c_passive;
struct rds_transport *c_trans;
@@ -150,7 +150,7 @@ struct rds_connection {
/* Protocol version */
unsigned int c_version;
- struct net *c_net;
+ possible_net_t c_net;
struct list_head c_map_item;
unsigned long c_map_queued;
@@ -165,13 +165,13 @@ struct rds_connection {
static inline
struct net *rds_conn_net(struct rds_connection *conn)
{
- return conn->c_net;
+ return read_pnet(&conn->c_net);
}
static inline
void rds_conn_net_set(struct rds_connection *conn, struct net *net)
{
- conn->c_net = get_net(net);
+ write_pnet(&conn->c_net, net);
}
#define RDS_FLAG_CONG_BITMAP 0x01
@@ -518,6 +518,7 @@ struct rds_transport {
void (*sync_mr)(void *trans_private, int direction);
void (*free_mr)(void *trans_private, int invalidate);
void (*flush_mrs)(void);
+ bool (*t_unloading)(struct rds_connection *conn);
};
struct rds_sock {
@@ -862,6 +863,12 @@ static inline void rds_mr_put(struct rds_mr *mr)
__rds_put_mr_final(mr);
}
+static inline bool rds_destroy_pending(struct rds_connection *conn)
+{
+ return !check_net(rds_conn_net(conn)) ||
+ (conn->c_trans->t_unloading && conn->c_trans->t_unloading(conn));
+}
+
/* stats.c */
DECLARE_PER_CPU_SHARED_ALIGNED(struct rds_statistics, rds_stats);
#define rds_stats_inc_which(which, member) do { \
diff --git a/net/rds/send.c b/net/rds/send.c
index f72466c63f0c..b1b0022b8370 100644
--- a/net/rds/send.c
+++ b/net/rds/send.c
@@ -162,6 +162,12 @@ restart:
goto out;
}
+ if (rds_destroy_pending(cp->cp_conn)) {
+ release_in_xmit(cp);
+ ret = -ENETUNREACH; /* dont requeue send work */
+ goto out;
+ }
+
/*
* we record the send generation after doing the xmit acquire.
* if someone else manages to jump in and do some work, we'll use
@@ -437,7 +443,12 @@ over_batch:
!list_empty(&cp->cp_send_queue)) && !raced) {
if (batch_count < send_batch_count)
goto restart;
- queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
+ rcu_read_lock();
+ if (rds_destroy_pending(cp->cp_conn))
+ ret = -ENETUNREACH;
+ else
+ queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
+ rcu_read_unlock();
} else if (raced) {
rds_stats_inc(s_send_lock_queue_raced);
}
@@ -1151,6 +1162,11 @@ int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
else
cpath = &conn->c_path[0];
+ if (rds_destroy_pending(conn)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+
rds_conn_path_connect_if_down(cpath);
ret = rds_cong_wait(conn->c_fcong, dport, nonblock, rs);
@@ -1190,9 +1206,17 @@ int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
rds_stats_inc(s_send_queued);
ret = rds_send_xmit(cpath);
- if (ret == -ENOMEM || ret == -EAGAIN)
- queue_delayed_work(rds_wq, &cpath->cp_send_w, 1);
-
+ if (ret == -ENOMEM || ret == -EAGAIN) {
+ ret = 0;
+ rcu_read_lock();
+ if (rds_destroy_pending(cpath->cp_conn))
+ ret = -ENETUNREACH;
+ else
+ queue_delayed_work(rds_wq, &cpath->cp_send_w, 1);
+ rcu_read_unlock();
+ }
+ if (ret)
+ goto out;
rds_message_put(rm);
return payload_len;
@@ -1270,7 +1294,10 @@ rds_send_probe(struct rds_conn_path *cp, __be16 sport,
rds_stats_inc(s_send_pong);
/* schedule the send work on rds_wq */
- queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
+ rcu_read_lock();
+ if (!rds_destroy_pending(cp->cp_conn))
+ queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
+ rcu_read_unlock();
rds_message_put(rm);
return 0;
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index ab7356e0ba83..44c4652721af 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -49,6 +49,7 @@ static unsigned int rds_tcp_tc_count;
/* Track rds_tcp_connection structs so they can be cleaned up */
static DEFINE_SPINLOCK(rds_tcp_conn_lock);
static LIST_HEAD(rds_tcp_conn_list);
+static atomic_t rds_tcp_unloading = ATOMIC_INIT(0);
static struct kmem_cache *rds_tcp_conn_slab;
@@ -271,16 +272,32 @@ static int rds_tcp_laddr_check(struct net *net, __be32 addr)
return -EADDRNOTAVAIL;
}
+static void rds_tcp_conn_free(void *arg)
+{
+ struct rds_tcp_connection *tc = arg;
+
+ rdsdebug("freeing tc %p\n", tc);
+
+ spin_lock_bh(&rds_tcp_conn_lock);
+ if (!tc->t_tcp_node_detached)
+ list_del(&tc->t_tcp_node);
+ spin_unlock_bh(&rds_tcp_conn_lock);
+
+ kmem_cache_free(rds_tcp_conn_slab, tc);
+}
+
static int rds_tcp_conn_alloc(struct rds_connection *conn, gfp_t gfp)
{
struct rds_tcp_connection *tc;
- int i;
+ int i, j;
+ int ret = 0;
for (i = 0; i < RDS_MPATH_WORKERS; i++) {
tc = kmem_cache_alloc(rds_tcp_conn_slab, gfp);
- if (!tc)
- return -ENOMEM;
-
+ if (!tc) {
+ ret = -ENOMEM;
+ goto fail;
+ }
mutex_init(&tc->t_conn_path_lock);
tc->t_sock = NULL;
tc->t_tinc = NULL;
@@ -289,28 +306,24 @@ static int rds_tcp_conn_alloc(struct rds_connection *conn, gfp_t gfp)
conn->c_path[i].cp_transport_data = tc;
tc->t_cpath = &conn->c_path[i];
+ tc->t_tcp_node_detached = true;
- spin_lock_irq(&rds_tcp_conn_lock);
- list_add_tail(&tc->t_tcp_node, &rds_tcp_conn_list);
- spin_unlock_irq(&rds_tcp_conn_lock);
rdsdebug("rds_conn_path [%d] tc %p\n", i,
conn->c_path[i].cp_transport_data);
}
-
- return 0;
-}
-
-static void rds_tcp_conn_free(void *arg)
-{
- struct rds_tcp_connection *tc = arg;
- unsigned long flags;
- rdsdebug("freeing tc %p\n", tc);
-
- spin_lock_irqsave(&rds_tcp_conn_lock, flags);
- list_del(&tc->t_tcp_node);
- spin_unlock_irqrestore(&rds_tcp_conn_lock, flags);
-
- kmem_cache_free(rds_tcp_conn_slab, tc);
+ spin_lock_bh(&rds_tcp_conn_lock);
+ for (i = 0; i < RDS_MPATH_WORKERS; i++) {
+ tc = conn->c_path[i].cp_transport_data;
+ tc->t_tcp_node_detached = false;
+ list_add_tail(&tc->t_tcp_node, &rds_tcp_conn_list);
+ }
+ spin_unlock_bh(&rds_tcp_conn_lock);
+fail:
+ if (ret) {
+ for (j = 0; j < i; j++)
+ rds_tcp_conn_free(conn->c_path[j].cp_transport_data);
+ }
+ return ret;
}
static bool list_has_conn(struct list_head *list, struct rds_connection *conn)
@@ -324,6 +337,16 @@ static bool list_has_conn(struct list_head *list, struct rds_connection *conn)
return false;
}
+static void rds_tcp_set_unloading(void)
+{
+ atomic_set(&rds_tcp_unloading, 1);
+}
+
+static bool rds_tcp_is_unloading(struct rds_connection *conn)
+{
+ return atomic_read(&rds_tcp_unloading) != 0;
+}
+
static void rds_tcp_destroy_conns(void)
{
struct rds_tcp_connection *tc, *_tc;
@@ -362,6 +385,7 @@ struct rds_transport rds_tcp_transport = {
.t_type = RDS_TRANS_TCP,
.t_prefer_loopback = 1,
.t_mp_capable = 1,
+ .t_unloading = rds_tcp_is_unloading,
};
static unsigned int rds_tcp_netid;
@@ -496,27 +520,6 @@ static struct pernet_operations rds_tcp_net_ops = {
.size = sizeof(struct rds_tcp_net),
};
-/* explicitly send a RST on each socket, thereby releasing any socket refcnts
- * that may otherwise hold up netns deletion.
- */
-static void rds_tcp_conn_paths_destroy(struct rds_connection *conn)
-{
- struct rds_conn_path *cp;
- struct rds_tcp_connection *tc;
- int i;
- struct sock *sk;
-
- for (i = 0; i < RDS_MPATH_WORKERS; i++) {
- cp = &conn->c_path[i];
- tc = cp->cp_transport_data;
- if (!tc->t_sock)
- continue;
- sk = tc->t_sock->sk;
- sk->sk_prot->disconnect(sk, 0);
- tcp_done(sk);
- }
-}
-
static void rds_tcp_kill_sock(struct net *net)
{
struct rds_tcp_connection *tc, *_tc;
@@ -526,20 +529,22 @@ static void rds_tcp_kill_sock(struct net *net)
rtn->rds_tcp_listen_sock = NULL;
rds_tcp_listen_stop(lsock, &rtn->rds_tcp_accept_w);
- spin_lock_irq(&rds_tcp_conn_lock);
+ spin_lock_bh(&rds_tcp_conn_lock);
list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) {
- struct net *c_net = tc->t_cpath->cp_conn->c_net;
+ struct net *c_net = read_pnet(&tc->t_cpath->cp_conn->c_net);
if (net != c_net || !tc->t_sock)
continue;
- if (!list_has_conn(&tmp_list, tc->t_cpath->cp_conn))
+ if (!list_has_conn(&tmp_list, tc->t_cpath->cp_conn)) {
list_move_tail(&tc->t_tcp_node, &tmp_list);
+ } else {
+ list_del(&tc->t_tcp_node);
+ tc->t_tcp_node_detached = true;
+ }
}
- spin_unlock_irq(&rds_tcp_conn_lock);
- list_for_each_entry_safe(tc, _tc, &tmp_list, t_tcp_node) {
- rds_tcp_conn_paths_destroy(tc->t_cpath->cp_conn);
+ spin_unlock_bh(&rds_tcp_conn_lock);
+ list_for_each_entry_safe(tc, _tc, &tmp_list, t_tcp_node)
rds_conn_destroy(tc->t_cpath->cp_conn);
- }
}
void *rds_tcp_listen_sock_def_readable(struct net *net)
@@ -585,9 +590,9 @@ static void rds_tcp_sysctl_reset(struct net *net)
{
struct rds_tcp_connection *tc, *_tc;
- spin_lock_irq(&rds_tcp_conn_lock);
+ spin_lock_bh(&rds_tcp_conn_lock);
list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) {
- struct net *c_net = tc->t_cpath->cp_conn->c_net;
+ struct net *c_net = read_pnet(&tc->t_cpath->cp_conn->c_net);
if (net != c_net || !tc->t_sock)
continue;
@@ -595,7 +600,7 @@ static void rds_tcp_sysctl_reset(struct net *net)
/* reconnect with new parameters */
rds_conn_path_drop(tc->t_cpath, false);
}
- spin_unlock_irq(&rds_tcp_conn_lock);
+ spin_unlock_bh(&rds_tcp_conn_lock);
}
static int rds_tcp_skbuf_handler(struct ctl_table *ctl, int write,
@@ -618,6 +623,8 @@ static int rds_tcp_skbuf_handler(struct ctl_table *ctl, int write,
static void rds_tcp_exit(void)
{
+ rds_tcp_set_unloading();
+ synchronize_rcu();
rds_info_deregister_func(RDS_INFO_TCP_SOCKETS, rds_tcp_tc_info);
unregister_pernet_subsys(&rds_tcp_net_ops);
if (unregister_netdevice_notifier(&rds_tcp_dev_notifier))
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 864ca7d8f019..c6fa080e9b6d 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -12,6 +12,7 @@ struct rds_tcp_incoming {
struct rds_tcp_connection {
struct list_head t_tcp_node;
+ bool t_tcp_node_detached;
struct rds_conn_path *t_cpath;
/* t_conn_path_lock synchronizes the connection establishment between
* rds_tcp_accept_one and rds_tcp_conn_path_connect
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 46f74dad0e16..d999e7075645 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -170,7 +170,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
cp->cp_conn, tc, sock);
if (sock) {
- if (cp->cp_conn->c_destroy_in_prog)
+ if (rds_destroy_pending(cp->cp_conn))
rds_tcp_set_linger(sock);
sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
lock_sock(sock->sk);
diff --git a/net/rds/tcp_recv.c b/net/rds/tcp_recv.c
index e006ef8e6d40..b9fbd2ee74ef 100644
--- a/net/rds/tcp_recv.c
+++ b/net/rds/tcp_recv.c
@@ -321,8 +321,12 @@ void rds_tcp_data_ready(struct sock *sk)
ready = tc->t_orig_data_ready;
rds_tcp_stats_inc(s_tcp_data_ready_calls);
- if (rds_tcp_read_sock(cp, GFP_ATOMIC) == -ENOMEM)
- queue_delayed_work(rds_wq, &cp->cp_recv_w, 0);
+ if (rds_tcp_read_sock(cp, GFP_ATOMIC) == -ENOMEM) {
+ rcu_read_lock();
+ if (!rds_destroy_pending(cp->cp_conn))
+ queue_delayed_work(rds_wq, &cp->cp_recv_w, 0);
+ rcu_read_unlock();
+ }
out:
read_unlock_bh(&sk->sk_callback_lock);
ready(sk);
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 9b76e0fa1722..7df869d37afd 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -202,8 +202,11 @@ void rds_tcp_write_space(struct sock *sk)
tc->t_last_seen_una = rds_tcp_snd_una(tc);
rds_send_path_drop_acked(cp, rds_tcp_snd_una(tc), rds_tcp_is_acked);
- if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
+ rcu_read_lock();
+ if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf &&
+ !rds_destroy_pending(cp->cp_conn))
queue_delayed_work(rds_wq, &cp->cp_send_w, 0);
+ rcu_read_unlock();
out:
read_unlock_bh(&sk->sk_callback_lock);
diff --git a/net/rds/threads.c b/net/rds/threads.c
index f121daa402c8..c52861d77a59 100644
--- a/net/rds/threads.c
+++ b/net/rds/threads.c
@@ -87,8 +87,12 @@ void rds_connect_path_complete(struct rds_conn_path *cp, int curr)
cp->cp_reconnect_jiffies = 0;
set_bit(0, &cp->cp_conn->c_map_queued);
- queue_delayed_work(rds_wq, &cp->cp_send_w, 0);
- queue_delayed_work(rds_wq, &cp->cp_recv_w, 0);
+ rcu_read_lock();
+ if (!rds_destroy_pending(cp->cp_conn)) {
+ queue_delayed_work(rds_wq, &cp->cp_send_w, 0);
+ queue_delayed_work(rds_wq, &cp->cp_recv_w, 0);
+ }
+ rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(rds_connect_path_complete);
@@ -133,7 +137,10 @@ void rds_queue_reconnect(struct rds_conn_path *cp)
set_bit(RDS_RECONNECT_PENDING, &cp->cp_flags);
if (cp->cp_reconnect_jiffies == 0) {
cp->cp_reconnect_jiffies = rds_sysctl_reconnect_min_jiffies;
- queue_delayed_work(rds_wq, &cp->cp_conn_w, 0);
+ rcu_read_lock();
+ if (!rds_destroy_pending(cp->cp_conn))
+ queue_delayed_work(rds_wq, &cp->cp_conn_w, 0);
+ rcu_read_unlock();
return;
}
@@ -141,8 +148,11 @@ void rds_queue_reconnect(struct rds_conn_path *cp)
rdsdebug("%lu delay %lu ceil conn %p for %pI4 -> %pI4\n",
rand % cp->cp_reconnect_jiffies, cp->cp_reconnect_jiffies,
conn, &conn->c_laddr, &conn->c_faddr);
- queue_delayed_work(rds_wq, &cp->cp_conn_w,
- rand % cp->cp_reconnect_jiffies);
+ rcu_read_lock();
+ if (!rds_destroy_pending(cp->cp_conn))
+ queue_delayed_work(rds_wq, &cp->cp_conn_w,
+ rand % cp->cp_reconnect_jiffies);
+ rcu_read_unlock();
cp->cp_reconnect_jiffies = min(cp->cp_reconnect_jiffies * 2,
rds_sysctl_reconnect_max_jiffies);
Powered by cgit, Gruvboxed by Blaster4385.