X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=fs%2Focfs2%2Fcluster%2Ftcp.c;h=ae4ff4a6636b23759522994898a95c148a4401f1;hb=97bf2856c6014879bd04983a3e9dfcdac1e7fe85;hp=0f60cc0d3985d9e80596e3f63b7059cb594d7d97;hpb=76828883507a47dae78837ab5dec5a5b4513c667;p=linux-2.6.git diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c index 0f60cc0d3..ae4ff4a66 100644 --- a/fs/ocfs2/cluster/tcp.c +++ b/fs/ocfs2/cluster/tcp.c @@ -108,7 +108,7 @@ ##args); \ } while (0) -static rwlock_t o2net_handler_lock = RW_LOCK_UNLOCKED; +static DEFINE_RWLOCK(o2net_handler_lock); static struct rb_root o2net_handler_tree = RB_ROOT; static struct o2net_node o2net_nodes[O2NM_MAX_NODES]; @@ -140,13 +140,35 @@ static int o2net_sys_err_translations[O2NET_ERR_MAX] = [O2NET_ERR_DIED] = -EHOSTDOWN,}; /* can't quite avoid *all* internal declarations :/ */ -static void o2net_sc_connect_completed(void *arg); -static void o2net_rx_until_empty(void *arg); -static void o2net_shutdown_sc(void *arg); +static void o2net_sc_connect_completed(struct work_struct *work); +static void o2net_rx_until_empty(struct work_struct *work); +static void o2net_shutdown_sc(struct work_struct *work); static void o2net_listen_data_ready(struct sock *sk, int bytes); -static void o2net_sc_send_keep_req(void *arg); +static void o2net_sc_send_keep_req(struct work_struct *work); static void o2net_idle_timer(unsigned long data); static void o2net_sc_postpone_idle(struct o2net_sock_container *sc); +static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc); + +/* + * FIXME: These should use to_o2nm_cluster_from_node(), but we end up + * losing our parent link to the cluster during shutdown. This can be + * solved by adding a pre-removal callback to configfs, or passing + * around the cluster with the node. -jeffm + */ +static inline int o2net_reconnect_delay(struct o2nm_node *node) +{ + return o2nm_single_cluster->cl_reconnect_delay_ms; +} + +static inline int o2net_keepalive_delay(struct o2nm_node *node) +{ + return o2nm_single_cluster->cl_keepalive_delay_ms; +} + +static inline int o2net_idle_timeout(struct o2nm_node *node) +{ + return o2nm_single_cluster->cl_idle_timeout_ms; +} static inline int o2net_sys_err_to_errno(enum o2net_system_error err) { @@ -271,6 +293,8 @@ static void sc_kref_release(struct kref *kref) { struct o2net_sock_container *sc = container_of(kref, struct o2net_sock_container, sc_kref); + BUG_ON(timer_pending(&sc->sc_idle_timeout)); + sclog(sc, "releasing\n"); if (sc->sc_sock) { @@ -300,7 +324,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) struct page *page = NULL; page = alloc_page(GFP_NOFS); - sc = kcalloc(1, sizeof(*sc), GFP_NOFS); + sc = kzalloc(sizeof(*sc), GFP_NOFS); if (sc == NULL || page == NULL) goto out; @@ -308,10 +332,10 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) o2nm_node_get(node); sc->sc_node = node; - INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed, sc); - INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty, sc); - INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc, sc); - INIT_WORK(&sc->sc_keepalive_work, o2net_sc_send_keep_req, sc); + INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed); + INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty); + INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc); + INIT_DELAYED_WORK(&sc->sc_keepalive_work, o2net_sc_send_keep_req); init_timer(&sc->sc_idle_timeout); sc->sc_idle_timeout.function = o2net_idle_timer; @@ -342,7 +366,7 @@ static void o2net_sc_queue_work(struct o2net_sock_container *sc, sc_put(sc); } static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc, - struct work_struct *work, + struct delayed_work *work, int delay) { sc_get(sc); @@ -350,12 +374,19 @@ static void o2net_sc_queue_delayed_work(struct o2net_sock_container *sc, sc_put(sc); } static void o2net_sc_cancel_delayed_work(struct o2net_sock_container *sc, - struct work_struct *work) + struct delayed_work *work) { if (cancel_delayed_work(work)) sc_put(sc); } +static atomic_t o2net_connected_peers = ATOMIC_INIT(0); + +int o2net_num_connected_peers(void) +{ + return atomic_read(&o2net_connected_peers); +} + static void o2net_set_nn_state(struct o2net_node *nn, struct o2net_sock_container *sc, unsigned valid, int err) @@ -366,6 +397,11 @@ static void o2net_set_nn_state(struct o2net_node *nn, assert_spin_locked(&nn->nn_lock); + if (old_sc && !sc) + atomic_dec(&o2net_connected_peers); + else if (!old_sc && sc) + atomic_inc(&o2net_connected_peers); + /* the node num comparison and single connect/accept path should stop * an non-null sc from being overwritten with another */ BUG_ON(sc && nn->nn_sc && nn->nn_sc != sc); @@ -396,8 +432,8 @@ static void o2net_set_nn_state(struct o2net_node *nn, } if (was_valid && !valid) { - mlog(ML_NOTICE, "no longer connected to " SC_NODEF_FMT "\n", - SC_NODEF_ARGS(old_sc)); + printk(KERN_INFO "o2net: no longer connected to " + SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc)); o2net_complete_nodes_nsw(nn); } @@ -409,10 +445,10 @@ static void o2net_set_nn_state(struct o2net_node *nn, * the only way to start connecting again is to down * heartbeat and bring it back up. */ cancel_delayed_work(&nn->nn_connect_expired); - mlog(ML_NOTICE, "%s " SC_NODEF_FMT "\n", - o2nm_this_node() > sc->sc_node->nd_num ? - "connected to" : "accepted connection from", - SC_NODEF_ARGS(sc)); + printk(KERN_INFO "o2net: %s " SC_NODEF_FMT "\n", + o2nm_this_node() > sc->sc_node->nd_num ? + "connected to" : "accepted connection from", + SC_NODEF_ARGS(sc)); } /* trigger the connecting worker func as long as we're not valid, @@ -424,9 +460,9 @@ static void o2net_set_nn_state(struct o2net_node *nn, /* delay if we're withing a RECONNECT_DELAY of the * last attempt */ delay = (nn->nn_last_connect_attempt + - msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS)) + msecs_to_jiffies(o2net_reconnect_delay(sc->sc_node))) - jiffies; - if (delay > msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS)) + if (delay > msecs_to_jiffies(o2net_reconnect_delay(sc->sc_node))) delay = 0; mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay); queue_delayed_work(o2net_wq, &nn->nn_connect_work, delay); @@ -564,9 +600,11 @@ static void o2net_ensure_shutdown(struct o2net_node *nn, * ourselves as state_change couldn't get the nn_lock and call set_nn_state * itself. */ -static void o2net_shutdown_sc(void *arg) +static void o2net_shutdown_sc(struct work_struct *work) { - struct o2net_sock_container *sc = arg; + struct o2net_sock_container *sc = + container_of(work, struct o2net_sock_container, + sc_shutdown_work); struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); sclog(sc, "shutting down\n"); @@ -676,7 +714,7 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len, goto out; } - nmh = kcalloc(1, sizeof(struct o2net_msg_handler), GFP_NOFS); + nmh = kzalloc(sizeof(struct o2net_msg_handler), GFP_NOFS); if (nmh == NULL) { ret = -ENOMEM; goto out; @@ -1097,13 +1135,51 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) return -1; } + /* + * Ensure timeouts are consistent with other nodes, otherwise + * we can end up with one node thinking that the other must be down, + * but isn't. This can ultimately cause corruption. + */ + if (be32_to_cpu(hand->o2net_idle_timeout_ms) != + o2net_idle_timeout(sc->sc_node)) { + mlog(ML_NOTICE, SC_NODEF_FMT " uses a network idle timeout of " + "%u ms, but we use %u ms locally. disconnecting\n", + SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_idle_timeout_ms), + o2net_idle_timeout(sc->sc_node)); + o2net_ensure_shutdown(nn, sc, -ENOTCONN); + return -1; + } + + if (be32_to_cpu(hand->o2net_keepalive_delay_ms) != + o2net_keepalive_delay(sc->sc_node)) { + mlog(ML_NOTICE, SC_NODEF_FMT " uses a keepalive delay of " + "%u ms, but we use %u ms locally. disconnecting\n", + SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_keepalive_delay_ms), + o2net_keepalive_delay(sc->sc_node)); + o2net_ensure_shutdown(nn, sc, -ENOTCONN); + return -1; + } + + if (be32_to_cpu(hand->o2hb_heartbeat_timeout_ms) != + O2HB_MAX_WRITE_TIMEOUT_MS) { + mlog(ML_NOTICE, SC_NODEF_FMT " uses a heartbeat timeout of " + "%u ms, but we use %u ms locally. disconnecting\n", + SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2hb_heartbeat_timeout_ms), + O2HB_MAX_WRITE_TIMEOUT_MS); + o2net_ensure_shutdown(nn, sc, -ENOTCONN); + return -1; + } + sc->sc_handshake_ok = 1; spin_lock(&nn->nn_lock); /* set valid and queue the idle timers only if it hasn't been * shut down already */ if (nn->nn_sc == sc) { - o2net_sc_postpone_idle(sc); + o2net_sc_reset_idle_timer(sc); o2net_set_nn_state(nn, sc, 1, 0); } spin_unlock(&nn->nn_lock); @@ -1129,6 +1205,23 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) sclog(sc, "receiving\n"); do_gettimeofday(&sc->sc_tv_advance_start); + if (unlikely(sc->sc_handshake_ok == 0)) { + if(sc->sc_page_off < sizeof(struct o2net_handshake)) { + data = page_address(sc->sc_page) + sc->sc_page_off; + datalen = sizeof(struct o2net_handshake) - sc->sc_page_off; + ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen); + if (ret > 0) + sc->sc_page_off += ret; + } + + if (sc->sc_page_off == sizeof(struct o2net_handshake)) { + o2net_check_handshake(sc); + if (unlikely(sc->sc_handshake_ok == 0)) + ret = -EPROTO; + } + goto out; + } + /* do we need more header? */ if (sc->sc_page_off < sizeof(struct o2net_msg)) { data = page_address(sc->sc_page) + sc->sc_page_off; @@ -1136,15 +1229,6 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) ret = o2net_recv_tcp_msg(sc->sc_sock, data, datalen); if (ret > 0) { sc->sc_page_off += ret; - - /* this working relies on the handshake being - * smaller than the normal message header */ - if (sc->sc_page_off >= sizeof(struct o2net_handshake)&& - !sc->sc_handshake_ok && o2net_check_handshake(sc)) { - ret = -EPROTO; - goto out; - } - /* only swab incoming here.. we can * only get here once as we cross from * being under to over */ @@ -1201,9 +1285,10 @@ out: /* this work func is triggerd by data ready. it reads until it can read no * more. it interprets 0, eof, as fatal. if data_ready hits while we're doing * our work the work struct will be marked and we'll be called again. */ -static void o2net_rx_until_empty(void *arg) +static void o2net_rx_until_empty(struct work_struct *work) { - struct o2net_sock_container *sc = arg; + struct o2net_sock_container *sc = + container_of(work, struct o2net_sock_container, sc_rx_work); int ret; do { @@ -1245,26 +1330,43 @@ static int o2net_set_nodelay(struct socket *sock) return ret; } +static void o2net_initialize_handshake(void) +{ + o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32( + O2HB_MAX_WRITE_TIMEOUT_MS); + o2net_hand->o2net_idle_timeout_ms = cpu_to_be32( + o2net_idle_timeout(NULL)); + o2net_hand->o2net_keepalive_delay_ms = cpu_to_be32( + o2net_keepalive_delay(NULL)); + o2net_hand->o2net_reconnect_delay_ms = cpu_to_be32( + o2net_reconnect_delay(NULL)); +} + /* ------------------------------------------------------------ */ /* called when a connect completes and after a sock is accepted. the * rx path will see the response and mark the sc valid */ -static void o2net_sc_connect_completed(void *arg) +static void o2net_sc_connect_completed(struct work_struct *work) { - struct o2net_sock_container *sc = arg; + struct o2net_sock_container *sc = + container_of(work, struct o2net_sock_container, + sc_connect_work); mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n", (unsigned long long)O2NET_PROTOCOL_VERSION, (unsigned long long)be64_to_cpu(o2net_hand->connector_id)); + o2net_initialize_handshake(); o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand)); sc_put(sc); } /* this is called as a work_struct func. */ -static void o2net_sc_send_keep_req(void *arg) +static void o2net_sc_send_keep_req(struct work_struct *work) { - struct o2net_sock_container *sc = arg; + struct o2net_sock_container *sc = + container_of(work, struct o2net_sock_container, + sc_keepalive_work.work); o2net_sendpage(sc, o2net_keep_req, sizeof(*o2net_keep_req)); sc_put(sc); @@ -1280,8 +1382,10 @@ static void o2net_idle_timer(unsigned long data) do_gettimeofday(&now); - mlog(ML_NOTICE, "connection to " SC_NODEF_FMT " has been idle for 10 " - "seconds, shutting it down.\n", SC_NODEF_ARGS(sc)); + printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u " + "seconds, shutting it down.\n", SC_NODEF_ARGS(sc), + o2net_idle_timeout(sc->sc_node) / 1000, + o2net_idle_timeout(sc->sc_node) % 1000); mlog(ML_NOTICE, "here are some times that might help debug the " "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv " "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n", @@ -1299,14 +1403,21 @@ static void o2net_idle_timer(unsigned long data) o2net_sc_queue_work(sc, &sc->sc_shutdown_work); } -static void o2net_sc_postpone_idle(struct o2net_sock_container *sc) +static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc) { o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work); o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work, - O2NET_KEEPALIVE_DELAY_SECS * HZ); + msecs_to_jiffies(o2net_keepalive_delay(sc->sc_node))); do_gettimeofday(&sc->sc_tv_timer); mod_timer(&sc->sc_idle_timeout, - jiffies + (O2NET_IDLE_TIMEOUT_SECS * HZ)); + jiffies + msecs_to_jiffies(o2net_idle_timeout(sc->sc_node))); +} + +static void o2net_sc_postpone_idle(struct o2net_sock_container *sc) +{ + /* Only push out an existing timer */ + if (timer_pending(&sc->sc_idle_timeout)) + o2net_sc_reset_idle_timer(sc); } /* this work func is kicked whenever a path sets the nn state which doesn't @@ -1314,14 +1425,15 @@ static void o2net_sc_postpone_idle(struct o2net_sock_container *sc) * having a connect attempt fail, etc. This centralizes the logic which decides * if a connect attempt should be made or if we should give up and all future * transmit attempts should fail */ -static void o2net_start_connect(void *arg) +static void o2net_start_connect(struct work_struct *work) { - struct o2net_node *nn = arg; + struct o2net_node *nn = + container_of(work, struct o2net_node, nn_connect_work.work); struct o2net_sock_container *sc = NULL; struct o2nm_node *node = NULL, *mynode = NULL; struct socket *sock = NULL; struct sockaddr_in myaddr = {0, }, remoteaddr = {0, }; - int ret = 0; + int ret = 0, stop; /* if we're greater we initiate tx, otherwise we accept */ if (o2nm_this_node() <= o2net_num_from_nn(nn)) @@ -1342,10 +1454,9 @@ static void o2net_start_connect(void *arg) spin_lock(&nn->nn_lock); /* see if we already have one pending or have given up */ - if (nn->nn_sc || nn->nn_persistent_error) - arg = NULL; + stop = (nn->nn_sc || nn->nn_persistent_error); spin_unlock(&nn->nn_lock); - if (arg == NULL) /* *shrug*, needed some indicator */ + if (stop) goto out; nn->nn_last_connect_attempt = jiffies; @@ -1421,24 +1532,29 @@ out: return; } -static void o2net_connect_expired(void *arg) +static void o2net_connect_expired(struct work_struct *work) { - struct o2net_node *nn = arg; + struct o2net_node *nn = + container_of(work, struct o2net_node, nn_connect_expired.work); spin_lock(&nn->nn_lock); if (!nn->nn_sc_valid) { + struct o2nm_node *node = nn->nn_sc->sc_node; mlog(ML_ERROR, "no connection established with node %u after " - "%u seconds, giving up and returning errors.\n", - o2net_num_from_nn(nn), O2NET_IDLE_TIMEOUT_SECS); + "%u.%u seconds, giving up and returning errors.\n", + o2net_num_from_nn(nn), + o2net_idle_timeout(node) / 1000, + o2net_idle_timeout(node) % 1000); o2net_set_nn_state(nn, NULL, 0, -ENOTCONN); } spin_unlock(&nn->nn_lock); } -static void o2net_still_up(void *arg) +static void o2net_still_up(struct work_struct *work) { - struct o2net_node *nn = arg; + struct o2net_node *nn = + container_of(work, struct o2net_node, nn_still_up.work); o2quo_hb_still_up(o2net_num_from_nn(nn)); } @@ -1469,6 +1585,8 @@ static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num, if (node_num != o2nm_this_node()) o2net_disconnect_node(node); + + BUG_ON(atomic_read(&o2net_connected_peers) < 0); } static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num, @@ -1480,14 +1598,14 @@ static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num, /* ensure an immediate connect attempt */ nn->nn_last_connect_attempt = jiffies - - (msecs_to_jiffies(O2NET_RECONNECT_DELAY_MS) + 1); + (msecs_to_jiffies(o2net_reconnect_delay(node)) + 1); if (node_num != o2nm_this_node()) { /* heartbeat doesn't work unless a local node number is * configured and doing so brings up the o2net_wq, so we can * use it.. */ queue_delayed_work(o2net_wq, &nn->nn_connect_expired, - O2NET_IDLE_TIMEOUT_SECS * HZ); + msecs_to_jiffies(o2net_idle_timeout(node))); /* believe it or not, accept and node hearbeating testing * can succeed for this node before we got here.. so @@ -1632,6 +1750,7 @@ static int o2net_accept_one(struct socket *sock) o2net_register_callbacks(sc->sc_sock->sk, sc); o2net_sc_queue_work(sc, &sc->sc_rx_work); + o2net_initialize_handshake(); o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand)); out: @@ -1644,9 +1763,9 @@ out: return ret; } -static void o2net_accept_many(void *arg) +static void o2net_accept_many(struct work_struct *work) { - struct socket *sock = arg; + struct socket *sock = o2net_listen_sock; while (o2net_accept_one(sock) == 0) cond_resched(); } @@ -1700,7 +1819,7 @@ static int o2net_open_listening_sock(__be16 port) write_unlock_bh(&sock->sk->sk_callback_lock); o2net_listen_sock = sock; - INIT_WORK(&o2net_listen_work, o2net_accept_many, sock); + INIT_WORK(&o2net_listen_work, o2net_accept_many); sock->sk->sk_reuse = 1; ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); @@ -1799,9 +1918,9 @@ int o2net_init(void) o2quo_init(); - o2net_hand = kcalloc(1, sizeof(struct o2net_handshake), GFP_KERNEL); - o2net_keep_req = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL); - o2net_keep_resp = kcalloc(1, sizeof(struct o2net_msg), GFP_KERNEL); + o2net_hand = kzalloc(sizeof(struct o2net_handshake), GFP_KERNEL); + o2net_keep_req = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL); + o2net_keep_resp = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL); if (!o2net_hand || !o2net_keep_req || !o2net_keep_resp) { kfree(o2net_hand); kfree(o2net_keep_req); @@ -1819,9 +1938,10 @@ int o2net_init(void) struct o2net_node *nn = o2net_nn_from_num(i); spin_lock_init(&nn->nn_lock); - INIT_WORK(&nn->nn_connect_work, o2net_start_connect, nn); - INIT_WORK(&nn->nn_connect_expired, o2net_connect_expired, nn); - INIT_WORK(&nn->nn_still_up, o2net_still_up, nn); + INIT_DELAYED_WORK(&nn->nn_connect_work, o2net_start_connect); + INIT_DELAYED_WORK(&nn->nn_connect_expired, + o2net_connect_expired); + INIT_DELAYED_WORK(&nn->nn_still_up, o2net_still_up); /* until we see hb from a node we'll return einval */ nn->nn_persistent_error = -ENOTCONN; init_waitqueue_head(&nn->nn_sc_wq);