X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lnet%2Fklnds%2Fsocklnd%2Fsocklnd.c;h=35d12d969a1fb6e338d4c2fc6b9d95ebae744d41;hp=4ed04583c909f5678338876928496f3c9066c05a;hb=34b57a6f8fcd;hpb=45a64335d0e326a411fe6a68cce77e618924f59f diff --git a/lnet/klnds/socklnd/socklnd.c b/lnet/klnds/socklnd/socklnd.c index 4ed0458..35d12d9 100644 --- a/lnet/klnds/socklnd/socklnd.c +++ b/lnet/klnds/socklnd/socklnd.c @@ -27,7 +27,6 @@ */ /* * This file is part of Lustre, http://www.lustre.org/ - * Lustre is a trademark of Sun Microsystems, Inc. * * lnet/klnds/socklnd/socklnd.c * @@ -37,63 +36,117 @@ * Author: Eric Barton */ -#include "socklnd.h" #include +#include "socklnd.h" +#include static const struct lnet_lnd the_ksocklnd; struct ksock_nal_data ksocknal_data; static struct ksock_interface * -ksocknal_ip2iface(struct lnet_ni *ni, __u32 ip) +ksocknal_ip2iface(struct lnet_ni *ni, struct sockaddr *addr) { struct ksock_net *net = ni->ni_data; - int i; struct ksock_interface *iface; - for (i = 0; i < net->ksnn_ninterfaces; i++) { - LASSERT(i < LNET_INTERFACES_NUM); - iface = &net->ksnn_interfaces[i]; + iface = &net->ksnn_interface; - if (iface->ksni_ipaddr == ip) - return iface; - } + if (rpc_cmp_addr((struct sockaddr *)&iface->ksni_addr, addr)) + return iface; return NULL; } -static struct ksock_route * -ksocknal_create_route(__u32 ipaddr, int port) +static struct ksock_interface * +ksocknal_index2iface(struct lnet_ni *ni, int index) +{ + struct ksock_net *net = ni->ni_data; + struct ksock_interface *iface; + + iface = &net->ksnn_interface; + + if (iface->ksni_index == index) + return iface; + + return NULL; +} + +static int ksocknal_ip2index(struct sockaddr *addr, struct lnet_ni *ni) +{ + struct net_device *dev; + int ret = -1; + DECLARE_CONST_IN_IFADDR(ifa); + + if (addr->sa_family != AF_INET) + /* No IPv6 support yet */ + return ret; + + rcu_read_lock(); + for_each_netdev(ni->ni_net_ns, dev) { + int flags = dev_get_flags(dev); + struct in_device *in_dev; + + if (flags & IFF_LOOPBACK) /* skip the loopback IF */ + continue; + + if (!(flags & IFF_UP)) + continue; + + in_dev = __in_dev_get_rcu(dev); + if (!in_dev) + continue; + + in_dev_for_each_ifa_rcu(ifa, in_dev) { + if (ifa->ifa_local == + ((struct sockaddr_in *)addr)->sin_addr.s_addr) + ret = dev->ifindex; + } + endfor_ifa(in_dev); + if (ret >= 0) + break; + } + rcu_read_unlock(); + + return ret; +} + +static struct ksock_conn_cb * +ksocknal_create_conn_cb(struct sockaddr *addr) { - struct ksock_route *route; - - LIBCFS_ALLOC (route, sizeof (*route)); - if (route == NULL) - return (NULL); - - atomic_set (&route->ksnr_refcount, 1); - route->ksnr_peer = NULL; - route->ksnr_retry_interval = 0; /* OK to connect at any time */ - route->ksnr_ipaddr = ipaddr; - route->ksnr_port = port; - route->ksnr_scheduled = 0; - route->ksnr_connecting = 0; - route->ksnr_connected = 0; - route->ksnr_deleted = 0; - route->ksnr_conn_count = 0; - route->ksnr_share_count = 0; - - return (route); + struct ksock_conn_cb *conn_cb; + + LIBCFS_ALLOC(conn_cb, sizeof(*conn_cb)); + if (!conn_cb) + return NULL; + + refcount_set(&conn_cb->ksnr_refcount, 1); + conn_cb->ksnr_peer = NULL; + conn_cb->ksnr_retry_interval = 0; /* OK to connect at any time */ + rpc_copy_addr((struct sockaddr *)&conn_cb->ksnr_addr, addr); + rpc_set_port((struct sockaddr *)&conn_cb->ksnr_addr, + rpc_get_port(addr)); + conn_cb->ksnr_myiface = -1; + conn_cb->ksnr_scheduled = 0; + conn_cb->ksnr_connecting = 0; + conn_cb->ksnr_connected = 0; + conn_cb->ksnr_deleted = 0; + conn_cb->ksnr_conn_count = 0; + conn_cb->ksnr_ctrl_conn_count = 0; + conn_cb->ksnr_blki_conn_count = 0; + conn_cb->ksnr_blko_conn_count = 0; + + return conn_cb; } void -ksocknal_destroy_route(struct ksock_route *route) +ksocknal_destroy_conn_cb(struct ksock_conn_cb *conn_cb) { - LASSERT (atomic_read(&route->ksnr_refcount) == 0); + LASSERT(refcount_read(&conn_cb->ksnr_refcount) == 0); - if (route->ksnr_peer != NULL) - ksocknal_peer_decref(route->ksnr_peer); + if (conn_cb->ksnr_peer) + ksocknal_peer_decref(conn_cb->ksnr_peer); - LIBCFS_FREE (route, sizeof (*route)); + LIBCFS_FREE(conn_cb, sizeof(*conn_cb)); } static struct ksock_peer_ni * @@ -120,15 +173,15 @@ ksocknal_create_peer(struct lnet_ni *ni, struct lnet_process_id id) peer_ni->ksnp_ni = ni; peer_ni->ksnp_id = id; - atomic_set(&peer_ni->ksnp_refcount, 1); /* 1 ref for caller */ + refcount_set(&peer_ni->ksnp_refcount, 1); /* 1 ref for caller */ peer_ni->ksnp_closing = 0; peer_ni->ksnp_accepting = 0; peer_ni->ksnp_proto = NULL; peer_ni->ksnp_last_alive = 0; peer_ni->ksnp_zc_next_cookie = SOCKNAL_KEEPALIVE_PING + 1; + peer_ni->ksnp_conn_cb = NULL; INIT_LIST_HEAD(&peer_ni->ksnp_conns); - INIT_LIST_HEAD(&peer_ni->ksnp_routes); INIT_LIST_HEAD(&peer_ni->ksnp_tx_queue); INIT_LIST_HEAD(&peer_ni->ksnp_zc_req_list); spin_lock_init(&peer_ni->ksnp_lock); @@ -144,21 +197,22 @@ ksocknal_destroy_peer(struct ksock_peer_ni *peer_ni) CDEBUG (D_NET, "peer_ni %s %p deleted\n", libcfs_id2str(peer_ni->ksnp_id), peer_ni); - LASSERT(atomic_read(&peer_ni->ksnp_refcount) == 0); + LASSERT(refcount_read(&peer_ni->ksnp_refcount) == 0); LASSERT(peer_ni->ksnp_accepting == 0); LASSERT(list_empty(&peer_ni->ksnp_conns)); - LASSERT(list_empty(&peer_ni->ksnp_routes)); + LASSERT(peer_ni->ksnp_conn_cb == NULL); LASSERT(list_empty(&peer_ni->ksnp_tx_queue)); LASSERT(list_empty(&peer_ni->ksnp_zc_req_list)); LIBCFS_FREE(peer_ni, sizeof(*peer_ni)); - /* NB a peer_ni's connections and routes keep a reference on their + /* NB a peer_ni's connections and conn_cb keep a reference on their * peer_ni until they are destroyed, so we can be assured that _all_ * state to do with this peer_ni has been cleaned up when its refcount * drops to zero. */ - atomic_dec(&net->ksnn_npeers); + if (atomic_dec_and_test(&net->ksnn_npeers)) + wake_up_var(&net->ksnn_npeers); } struct ksock_peer_ni * @@ -179,7 +233,7 @@ ksocknal_find_peer_locked(struct lnet_ni *ni, struct lnet_process_id id) CDEBUG(D_NET, "got peer_ni [%p] -> %s (%d)\n", peer_ni, libcfs_id2str(id), - atomic_read(&peer_ni->ksnp_refcount)); + refcount_read(&peer_ni->ksnp_refcount)); return peer_ni; } return NULL; @@ -203,14 +257,15 @@ static void ksocknal_unlink_peer_locked(struct ksock_peer_ni *peer_ni) { int i; - __u32 ip; struct ksock_interface *iface; for (i = 0; i < peer_ni->ksnp_n_passive_ips; i++) { + struct sockaddr_in sa = { .sin_family = AF_INET }; LASSERT(i < LNET_INTERFACES_NUM); - ip = peer_ni->ksnp_passive_ips[i]; + sa.sin_addr.s_addr = htonl(peer_ni->ksnp_passive_ips[i]); - iface = ksocknal_ip2iface(peer_ni->ksnp_ni, ip); + iface = ksocknal_ip2iface(peer_ni->ksnp_ni, + (struct sockaddr *)&sa); /* * All IPs in peer_ni->ksnp_passive_ips[] come from the * interface list, therefore the call must succeed. @@ -223,7 +278,7 @@ ksocknal_unlink_peer_locked(struct ksock_peer_ni *peer_ni) } LASSERT(list_empty(&peer_ni->ksnp_conns)); - LASSERT(list_empty(&peer_ni->ksnp_routes)); + LASSERT(peer_ni->ksnp_conn_cb == NULL); LASSERT(!peer_ni->ksnp_closing); peer_ni->ksnp_closing = 1; hlist_del(&peer_ni->ksnp_list); @@ -237,8 +292,7 @@ ksocknal_get_peer_info(struct lnet_ni *ni, int index, int *port, int *conn_count, int *share_count) { struct ksock_peer_ni *peer_ni; - struct ksock_route *route; - struct list_head *rtmp; + struct ksock_conn_cb *conn_cb; int i; int j; int rc = -ENOENT; @@ -251,7 +305,7 @@ ksocknal_get_peer_info(struct lnet_ni *ni, int index, continue; if (peer_ni->ksnp_n_passive_ips == 0 && - list_empty(&peer_ni->ksnp_routes)) { + peer_ni->ksnp_conn_cb == NULL) { if (index-- > 0) continue; @@ -279,20 +333,31 @@ ksocknal_get_peer_info(struct lnet_ni *ni, int index, goto out; } - list_for_each(rtmp, &peer_ni->ksnp_routes) { + if (peer_ni->ksnp_conn_cb) { if (index-- > 0) continue; - route = list_entry(rtmp, struct ksock_route, - ksnr_list); + conn_cb = peer_ni->ksnp_conn_cb; *id = peer_ni->ksnp_id; - *myip = route->ksnr_myipaddr; - *peer_ip = route->ksnr_ipaddr; - *port = route->ksnr_port; - *conn_count = route->ksnr_conn_count; - *share_count = route->ksnr_share_count; - rc = 0; + if (conn_cb->ksnr_addr.ss_family == AF_INET) { + struct sockaddr_in *sa = + (void *)&conn_cb->ksnr_addr; + + rc = choose_ipv4_src(myip, + conn_cb->ksnr_myiface, + ntohl(sa->sin_addr.s_addr), + ni->ni_net_ns); + *peer_ip = ntohl(sa->sin_addr.s_addr); + *port = ntohs(sa->sin_port); + } else { + *myip = 0xFFFFFFFF; + *peer_ip = 0xFFFFFFFF; + *port = 0; + rc = -ENOTSUPP; + } + *conn_count = conn_cb->ksnr_conn_count; + *share_count = 1; goto out; } } @@ -301,159 +366,221 @@ out: return rc; } +static unsigned int +ksocknal_get_conn_count_by_type(struct ksock_conn_cb *conn_cb, + int type) +{ + unsigned int count = 0; + + switch (type) { + case SOCKLND_CONN_CONTROL: + count = conn_cb->ksnr_ctrl_conn_count; + break; + case SOCKLND_CONN_BULK_IN: + count = conn_cb->ksnr_blki_conn_count; + break; + case SOCKLND_CONN_BULK_OUT: + count = conn_cb->ksnr_blko_conn_count; + break; + case SOCKLND_CONN_ANY: + count = conn_cb->ksnr_conn_count; + break; + default: + LBUG(); + break; + } + + return count; +} + +static void +ksocknal_incr_conn_count(struct ksock_conn_cb *conn_cb, + int type) +{ + conn_cb->ksnr_conn_count++; + + /* check if all connections of the given type got created */ + switch (type) { + case SOCKLND_CONN_CONTROL: + conn_cb->ksnr_ctrl_conn_count++; + /* there's a single control connection per peer */ + conn_cb->ksnr_connected |= BIT(type); + break; + case SOCKLND_CONN_BULK_IN: + conn_cb->ksnr_blki_conn_count++; + if (conn_cb->ksnr_blki_conn_count >= + *ksocknal_tunables.ksnd_conns_per_peer) + conn_cb->ksnr_connected |= BIT(type); + break; + case SOCKLND_CONN_BULK_OUT: + conn_cb->ksnr_blko_conn_count++; + if (conn_cb->ksnr_blko_conn_count >= + *ksocknal_tunables.ksnd_conns_per_peer) + conn_cb->ksnr_connected |= BIT(type); + break; + case SOCKLND_CONN_ANY: + if (conn_cb->ksnr_conn_count >= + *ksocknal_tunables.ksnd_conns_per_peer) + conn_cb->ksnr_connected |= BIT(type); + break; + default: + LBUG(); + break; + + } + + CDEBUG(D_NET, "Add conn type %d, ksnr_connected %x conns_per_peer %d\n", + type, conn_cb->ksnr_connected, *ksocknal_tunables.ksnd_conns_per_peer); +} + static void -ksocknal_associate_route_conn_locked(struct ksock_route *route, struct ksock_conn *conn) +ksocknal_associate_cb_conn_locked(struct ksock_conn_cb *conn_cb, + struct ksock_conn *conn) { - struct ksock_peer_ni *peer_ni = route->ksnr_peer; + struct ksock_peer_ni *peer_ni = conn_cb->ksnr_peer; int type = conn->ksnc_type; struct ksock_interface *iface; + int conn_iface; - conn->ksnc_route = route; - ksocknal_route_addref(route); + conn_iface = ksocknal_ip2index((struct sockaddr *)&conn->ksnc_myaddr, + peer_ni->ksnp_ni); + conn->ksnc_conn_cb = conn_cb; + ksocknal_conn_cb_addref(conn_cb); - if (route->ksnr_myipaddr != conn->ksnc_myipaddr) { - if (route->ksnr_myipaddr == 0) { + if (conn_cb->ksnr_myiface != conn_iface) { + if (conn_cb->ksnr_myiface < 0) { /* route wasn't bound locally yet (the initial route) */ - CDEBUG(D_NET, "Binding %s %pI4h to %pI4h\n", + CDEBUG(D_NET, "Binding %s %pIS to interface %d\n", libcfs_id2str(peer_ni->ksnp_id), - &route->ksnr_ipaddr, - &conn->ksnc_myipaddr); + &conn_cb->ksnr_addr, + conn_iface); } else { - CDEBUG(D_NET, "Rebinding %s %pI4h from %pI4h " - "to %pI4h\n", libcfs_id2str(peer_ni->ksnp_id), - &route->ksnr_ipaddr, - &route->ksnr_myipaddr, - &conn->ksnc_myipaddr); - - iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni, - route->ksnr_myipaddr); - if (iface != NULL) - iface->ksni_nroutes--; - } - route->ksnr_myipaddr = conn->ksnc_myipaddr; - iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni, - route->ksnr_myipaddr); - if (iface != NULL) - iface->ksni_nroutes++; - } + CDEBUG(D_NET, + "Rebinding %s %pIS from interface %d to %d\n", + libcfs_id2str(peer_ni->ksnp_id), + &conn_cb->ksnr_addr, + conn_cb->ksnr_myiface, + conn_iface); + + iface = ksocknal_index2iface(peer_ni->ksnp_ni, + conn_cb->ksnr_myiface); + if (iface) + iface->ksni_nroutes--; + } + conn_cb->ksnr_myiface = conn_iface; + iface = ksocknal_index2iface(peer_ni->ksnp_ni, + conn_cb->ksnr_myiface); + if (iface) + iface->ksni_nroutes++; + } - route->ksnr_connected |= (1<ksnr_conn_count++; + ksocknal_incr_conn_count(conn_cb, type); - /* Successful connection => further attempts can - * proceed immediately */ - route->ksnr_retry_interval = 0; + /* Successful connection => further attempts can + * proceed immediately + */ + conn_cb->ksnr_retry_interval = 0; } static void -ksocknal_add_route_locked(struct ksock_peer_ni *peer_ni, struct ksock_route *route) +ksocknal_add_conn_cb_locked(struct ksock_peer_ni *peer_ni, + struct ksock_conn_cb *conn_cb) { struct list_head *tmp; struct ksock_conn *conn; - struct ksock_route *route2; + struct ksock_net *net = peer_ni->ksnp_ni->ni_data; LASSERT(!peer_ni->ksnp_closing); - LASSERT(route->ksnr_peer == NULL); - LASSERT(!route->ksnr_scheduled); - LASSERT(!route->ksnr_connecting); - LASSERT(route->ksnr_connected == 0); + LASSERT(!conn_cb->ksnr_peer); + LASSERT(!conn_cb->ksnr_scheduled); + LASSERT(!conn_cb->ksnr_connecting); + LASSERT(conn_cb->ksnr_connected == 0); - /* LASSERT(unique) */ - list_for_each(tmp, &peer_ni->ksnp_routes) { - route2 = list_entry(tmp, struct ksock_route, ksnr_list); + conn_cb->ksnr_peer = peer_ni; + ksocknal_peer_addref(peer_ni); - if (route2->ksnr_ipaddr == route->ksnr_ipaddr) { - CERROR("Duplicate route %s %pI4h\n", - libcfs_id2str(peer_ni->ksnp_id), - &route->ksnr_ipaddr); - LBUG(); - } - } + /* set the conn_cb's interface to the current net's interface */ + conn_cb->ksnr_myiface = net->ksnn_interface.ksni_index; + net->ksnn_interface.ksni_nroutes++; - route->ksnr_peer = peer_ni; - ksocknal_peer_addref(peer_ni); - /* peer_ni's routelist takes over my ref on 'route' */ - list_add_tail(&route->ksnr_list, &peer_ni->ksnp_routes); + /* peer_ni's route list takes over my ref on 'route' */ + peer_ni->ksnp_conn_cb = conn_cb; list_for_each(tmp, &peer_ni->ksnp_conns) { conn = list_entry(tmp, struct ksock_conn, ksnc_list); - if (conn->ksnc_ipaddr != route->ksnr_ipaddr) + if (!rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr, + (struct sockaddr *)&conn_cb->ksnr_addr)) continue; - ksocknal_associate_route_conn_locked(route, conn); - /* keep going (typed routes) */ + ksocknal_associate_cb_conn_locked(conn_cb, conn); + /* keep going (typed conns) */ } } static void -ksocknal_del_route_locked(struct ksock_route *route) +ksocknal_del_conn_cb_locked(struct ksock_conn_cb *conn_cb) { - struct ksock_peer_ni *peer_ni = route->ksnr_peer; + struct ksock_peer_ni *peer_ni = conn_cb->ksnr_peer; struct ksock_interface *iface; struct ksock_conn *conn; - struct list_head *ctmp; - struct list_head *cnxt; + struct ksock_conn *cnxt; - LASSERT(!route->ksnr_deleted); + LASSERT(!conn_cb->ksnr_deleted); /* Close associated conns */ - list_for_each_safe(ctmp, cnxt, &peer_ni->ksnp_conns) { - conn = list_entry(ctmp, struct ksock_conn, ksnc_list); - - if (conn->ksnc_route != route) + list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns, ksnc_list) { + if (conn->ksnc_conn_cb != conn_cb) continue; ksocknal_close_conn_locked(conn, 0); } - if (route->ksnr_myipaddr != 0) { - iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni, - route->ksnr_myipaddr); - if (iface != NULL) + if (conn_cb->ksnr_myiface >= 0) { + iface = ksocknal_index2iface(peer_ni->ksnp_ni, + conn_cb->ksnr_myiface); + if (iface) iface->ksni_nroutes--; } - route->ksnr_deleted = 1; - list_del(&route->ksnr_list); - ksocknal_route_decref(route); /* drop peer_ni's ref */ + conn_cb->ksnr_deleted = 1; + ksocknal_conn_cb_decref(conn_cb); /* drop peer_ni's ref */ + peer_ni->ksnp_conn_cb = NULL; - if (list_empty(&peer_ni->ksnp_routes) && - list_empty(&peer_ni->ksnp_conns)) { + if (list_empty(&peer_ni->ksnp_conns)) { /* I've just removed the last route to a peer_ni with no active - * connections */ + * connections + */ ksocknal_unlink_peer_locked(peer_ni); } } int -ksocknal_add_peer(struct lnet_ni *ni, struct lnet_process_id id, __u32 ipaddr, - int port) +ksocknal_add_peer(struct lnet_ni *ni, struct lnet_process_id id, + struct sockaddr *addr) { - struct list_head *tmp; struct ksock_peer_ni *peer_ni; struct ksock_peer_ni *peer2; - struct ksock_route *route; - struct ksock_route *route2; + struct ksock_conn_cb *conn_cb; - if (id.nid == LNET_NID_ANY || - id.pid == LNET_PID_ANY) - return (-EINVAL); + if (id.nid == LNET_NID_ANY || + id.pid == LNET_PID_ANY) + return (-EINVAL); /* Have a brand new peer_ni ready... */ peer_ni = ksocknal_create_peer(ni, id); if (IS_ERR(peer_ni)) return PTR_ERR(peer_ni); - route = ksocknal_create_route (ipaddr, port); - if (route == NULL) { - ksocknal_peer_decref(peer_ni); - return (-ENOMEM); - } + conn_cb = ksocknal_create_conn_cb(addr); + if (!conn_cb) { + ksocknal_peer_decref(peer_ni); + return -ENOMEM; + } write_lock_bh(&ksocknal_data.ksnd_global_lock); - /* always called with a ref on ni, so shutdown can't have started */ + /* always called with a ref on ni, so shutdown can't have started */ LASSERT(atomic_read(&((struct ksock_net *)ni->ni_data)->ksnn_npeers) >= 0); @@ -466,22 +593,7 @@ ksocknal_add_peer(struct lnet_ni *ni, struct lnet_process_id id, __u32 ipaddr, hash_add(ksocknal_data.ksnd_peers, &peer_ni->ksnp_list, id.nid); } - route2 = NULL; - list_for_each(tmp, &peer_ni->ksnp_routes) { - route2 = list_entry(tmp, struct ksock_route, ksnr_list); - - if (route2->ksnr_ipaddr == ipaddr) - break; - - route2 = NULL; - } - if (route2 == NULL) { - ksocknal_add_route_locked(peer_ni, route); - route->ksnr_share_count++; - } else { - ksocknal_route_decref(route); - route2->ksnr_share_count++; - } + ksocknal_add_conn_cb_locked(peer_ni, conn_cb); write_unlock_bh(&ksocknal_data.ksnd_global_lock); @@ -492,55 +604,23 @@ static void ksocknal_del_peer_locked(struct ksock_peer_ni *peer_ni, __u32 ip) { struct ksock_conn *conn; - struct ksock_route *route; - struct list_head *tmp; - struct list_head *nxt; - int nshared; + struct ksock_conn *cnxt; + struct ksock_conn_cb *conn_cb; LASSERT(!peer_ni->ksnp_closing); /* Extra ref prevents peer_ni disappearing until I'm done with it */ ksocknal_peer_addref(peer_ni); + conn_cb = peer_ni->ksnp_conn_cb; + if (conn_cb) + ksocknal_del_conn_cb_locked(conn_cb); - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) { - route = list_entry(tmp, struct ksock_route, ksnr_list); - - /* no match */ - if (!(ip == 0 || route->ksnr_ipaddr == ip)) - continue; - - route->ksnr_share_count = 0; - /* This deletes associated conns too */ - ksocknal_del_route_locked(route); - } - - nshared = 0; - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) { - route = list_entry(tmp, struct ksock_route, ksnr_list); - nshared += route->ksnr_share_count; - } - - if (nshared == 0) { - /* remove everything else if there are no explicit entries - * left */ - - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) { - route = list_entry(tmp, struct ksock_route, ksnr_list); - - /* we should only be removing auto-entries */ - LASSERT(route->ksnr_share_count == 0); - ksocknal_del_route_locked(route); - } - - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_conns) { - conn = list_entry(tmp, struct ksock_conn, ksnc_list); - - ksocknal_close_conn_locked(conn, 0); - } - } + list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns, + ksnc_list) + ksocknal_close_conn_locked(conn, 0); ksocknal_peer_decref(peer_ni); - /* NB peer_ni unlinks itself when last conn/route is removed */ + /* NB peer_ni unlinks itself when last conn/conn_cb is removed */ } static int @@ -584,7 +664,7 @@ ksocknal_del_peer(struct lnet_ni *ni, struct lnet_process_id id, __u32 ip) if (peer_ni->ksnp_closing && !list_empty(&peer_ni->ksnp_tx_queue)) { LASSERT(list_empty(&peer_ni->ksnp_conns)); - LASSERT(list_empty(&peer_ni->ksnp_routes)); + LASSERT(peer_ni->ksnp_conn_cb == NULL); list_splice_init(&peer_ni->ksnp_tx_queue, &zombies); @@ -655,308 +735,21 @@ ksocknal_choose_scheduler_locked(unsigned int cpt) return sched; } -static int -ksocknal_local_ipvec(struct lnet_ni *ni, __u32 *ipaddrs) -{ - struct ksock_net *net = ni->ni_data; - int i; - int nip; - - read_lock(&ksocknal_data.ksnd_global_lock); - - nip = net->ksnn_ninterfaces; - LASSERT(nip <= LNET_INTERFACES_NUM); - - /* - * Only offer interfaces for additional connections if I have - * more than one. - */ - if (nip < 2) { - read_unlock(&ksocknal_data.ksnd_global_lock); - return 0; - } - - for (i = 0; i < nip; i++) { - ipaddrs[i] = net->ksnn_interfaces[i].ksni_ipaddr; - LASSERT(ipaddrs[i] != 0); - } - - read_unlock(&ksocknal_data.ksnd_global_lock); - return nip; -} - -static int -ksocknal_match_peerip(struct ksock_interface *iface, __u32 *ips, int nips) -{ - int best_netmatch = 0; - int best_xor = 0; - int best = -1; - int this_xor; - int this_netmatch; - int i; - - for (i = 0; i < nips; i++) { - if (ips[i] == 0) - continue; - - this_xor = (ips[i] ^ iface->ksni_ipaddr); - this_netmatch = ((this_xor & iface->ksni_netmask) == 0) ? 1 : 0; - - if (!(best < 0 || - best_netmatch < this_netmatch || - (best_netmatch == this_netmatch && - best_xor > this_xor))) - continue; - - best = i; - best_netmatch = this_netmatch; - best_xor = this_xor; - } - - LASSERT (best >= 0); - return (best); -} - -static int -ksocknal_select_ips(struct ksock_peer_ni *peer_ni, __u32 *peerips, int n_peerips) -{ - rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock; - struct ksock_net *net = peer_ni->ksnp_ni->ni_data; - struct ksock_interface *iface; - struct ksock_interface *best_iface; - int n_ips; - int i; - int j; - int k; - u32 ip; - u32 xor; - int this_netmatch; - int best_netmatch; - int best_npeers; - - /* CAVEAT EMPTOR: We do all our interface matching with an - * exclusive hold of global lock at IRQ priority. We're only - * expecting to be dealing with small numbers of interfaces, so the - * O(n**3)-ness shouldn't matter */ - - /* Also note that I'm not going to return more than n_peerips - * interfaces, even if I have more myself */ - - write_lock_bh(global_lock); - - LASSERT(n_peerips <= LNET_INTERFACES_NUM); - LASSERT(net->ksnn_ninterfaces <= LNET_INTERFACES_NUM); - - /* Only match interfaces for additional connections - * if I have > 1 interface - */ - n_ips = (net->ksnn_ninterfaces < 2) ? 0 : - min(n_peerips, net->ksnn_ninterfaces); - - for (i = 0; peer_ni->ksnp_n_passive_ips < n_ips; i++) { - /* ^ yes really... */ - - /* If we have any new interfaces, first tick off all the - * peer_ni IPs that match old interfaces, then choose new - * interfaces to match the remaining peer_ni IPS. - * We don't forget interfaces we've stopped using; we might - * start using them again... */ - - if (i < peer_ni->ksnp_n_passive_ips) { - /* Old interface. */ - ip = peer_ni->ksnp_passive_ips[i]; - best_iface = ksocknal_ip2iface(peer_ni->ksnp_ni, ip); - - /* peer_ni passive ips are kept up to date */ - LASSERT(best_iface != NULL); - } else { - /* choose a new interface */ - LASSERT (i == peer_ni->ksnp_n_passive_ips); - - best_iface = NULL; - best_netmatch = 0; - best_npeers = 0; - - for (j = 0; j < net->ksnn_ninterfaces; j++) { - iface = &net->ksnn_interfaces[j]; - ip = iface->ksni_ipaddr; - - for (k = 0; k < peer_ni->ksnp_n_passive_ips; k++) - if (peer_ni->ksnp_passive_ips[k] == ip) - break; - - if (k < peer_ni->ksnp_n_passive_ips) /* using it already */ - continue; - - k = ksocknal_match_peerip(iface, peerips, n_peerips); - xor = (ip ^ peerips[k]); - this_netmatch = ((xor & iface->ksni_netmask) == 0) ? 1 : 0; - - if (!(best_iface == NULL || - best_netmatch < this_netmatch || - (best_netmatch == this_netmatch && - best_npeers > iface->ksni_npeers))) - continue; - - best_iface = iface; - best_netmatch = this_netmatch; - best_npeers = iface->ksni_npeers; - } - - LASSERT(best_iface != NULL); - - best_iface->ksni_npeers++; - ip = best_iface->ksni_ipaddr; - peer_ni->ksnp_passive_ips[i] = ip; - peer_ni->ksnp_n_passive_ips = i+1; - } - - /* mark the best matching peer_ni IP used */ - j = ksocknal_match_peerip(best_iface, peerips, n_peerips); - peerips[j] = 0; - } - - /* Overwrite input peer_ni IP addresses */ - memcpy(peerips, peer_ni->ksnp_passive_ips, n_ips * sizeof(*peerips)); - - write_unlock_bh(global_lock); - - return (n_ips); -} - -static void -ksocknal_create_routes(struct ksock_peer_ni *peer_ni, int port, - __u32 *peer_ipaddrs, int npeer_ipaddrs) -{ - struct ksock_route *newroute = NULL; - rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock; - struct lnet_ni *ni = peer_ni->ksnp_ni; - struct ksock_net *net = ni->ni_data; - struct list_head *rtmp; - struct ksock_route *route; - struct ksock_interface *iface; - struct ksock_interface *best_iface; - int best_netmatch; - int this_netmatch; - int best_nroutes; - int i; - int j; - - /* CAVEAT EMPTOR: We do all our interface matching with an - * exclusive hold of global lock at IRQ priority. We're only - * expecting to be dealing with small numbers of interfaces, so the - * O(n**3)-ness here shouldn't matter */ - - write_lock_bh(global_lock); - - if (net->ksnn_ninterfaces < 2) { - /* Only create additional connections - * if I have > 1 interface */ - write_unlock_bh(global_lock); - return; - } - - LASSERT(npeer_ipaddrs <= LNET_INTERFACES_NUM); - - for (i = 0; i < npeer_ipaddrs; i++) { - if (newroute != NULL) { - newroute->ksnr_ipaddr = peer_ipaddrs[i]; - } else { - write_unlock_bh(global_lock); - - newroute = ksocknal_create_route(peer_ipaddrs[i], port); - if (newroute == NULL) - return; - - write_lock_bh(global_lock); - } - - if (peer_ni->ksnp_closing) { - /* peer_ni got closed under me */ - break; - } - - /* Already got a route? */ - route = NULL; - list_for_each(rtmp, &peer_ni->ksnp_routes) { - route = list_entry(rtmp, struct ksock_route, ksnr_list); - - if (route->ksnr_ipaddr == newroute->ksnr_ipaddr) - break; - - route = NULL; - } - if (route != NULL) - continue; - - best_iface = NULL; - best_nroutes = 0; - best_netmatch = 0; - - LASSERT(net->ksnn_ninterfaces <= LNET_INTERFACES_NUM); - - /* Select interface to connect from */ - for (j = 0; j < net->ksnn_ninterfaces; j++) { - iface = &net->ksnn_interfaces[j]; - - /* Using this interface already? */ - list_for_each(rtmp, &peer_ni->ksnp_routes) { - route = list_entry(rtmp, struct ksock_route, - ksnr_list); - - if (route->ksnr_myipaddr == iface->ksni_ipaddr) - break; - - route = NULL; - } - if (route != NULL) - continue; - - this_netmatch = (((iface->ksni_ipaddr ^ - newroute->ksnr_ipaddr) & - iface->ksni_netmask) == 0) ? 1 : 0; - - if (!(best_iface == NULL || - best_netmatch < this_netmatch || - (best_netmatch == this_netmatch && - best_nroutes > iface->ksni_nroutes))) - continue; - - best_iface = iface; - best_netmatch = this_netmatch; - best_nroutes = iface->ksni_nroutes; - } - - if (best_iface == NULL) - continue; - - newroute->ksnr_myipaddr = best_iface->ksni_ipaddr; - best_iface->ksni_nroutes++; - - ksocknal_add_route_locked(peer_ni, newroute); - newroute = NULL; - } - - write_unlock_bh(global_lock); - if (newroute != NULL) - ksocknal_route_decref(newroute); -} - int ksocknal_accept(struct lnet_ni *ni, struct socket *sock) { struct ksock_connreq *cr; int rc; - u32 peer_ip; - int peer_port; + struct sockaddr_storage peer; - rc = lnet_sock_getaddr(sock, true, &peer_ip, &peer_port); + rc = lnet_sock_getaddr(sock, true, &peer); LASSERT(rc == 0); /* we succeeded before */ LIBCFS_ALLOC(cr, sizeof(*cr)); if (cr == NULL) { - LCONSOLE_ERROR_MSG(0x12f, "Dropping connection request from " - "%pI4h: memory exhausted\n", &peer_ip); + LCONSOLE_ERROR_MSG(0x12f, + "Dropping connection request from %pIS: memory exhausted\n", + &peer); return -ENOMEM; } @@ -974,19 +767,16 @@ ksocknal_accept(struct lnet_ni *ni, struct socket *sock) } static int -ksocknal_connecting(struct ksock_peer_ni *peer_ni, __u32 ipaddr) +ksocknal_connecting(struct ksock_conn_cb *conn_cb, struct sockaddr *sa) { - struct ksock_route *route; - - list_for_each_entry(route, &peer_ni->ksnp_routes, ksnr_list) { - if (route->ksnr_ipaddr == ipaddr) - return route->ksnr_connecting; - } + if (conn_cb && + rpc_cmp_addr((struct sockaddr *)&conn_cb->ksnr_addr, sa)) + return conn_cb->ksnr_connecting; return 0; } int -ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, +ksocknal_create_conn(struct lnet_ni *ni, struct ksock_conn_cb *conn_cb, struct socket *sock, int type) { rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock; @@ -1006,27 +796,28 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, int rc; int rc2; int active; + int num_dup = 0; char *warn = NULL; - active = (route != NULL); + active = (conn_cb != NULL); - LASSERT (active == (type != SOCKLND_CONN_NONE)); + LASSERT(active == (type != SOCKLND_CONN_NONE)); - LIBCFS_ALLOC(conn, sizeof(*conn)); - if (conn == NULL) { - rc = -ENOMEM; - goto failed_0; - } + LIBCFS_ALLOC(conn, sizeof(*conn)); + if (conn == NULL) { + rc = -ENOMEM; + goto failed_0; + } - conn->ksnc_peer = NULL; - conn->ksnc_route = NULL; - conn->ksnc_sock = sock; + conn->ksnc_peer = NULL; + conn->ksnc_conn_cb = NULL; + conn->ksnc_sock = sock; /* 2 ref, 1 for conn, another extra ref prevents socket * being closed before establishment of connection */ - atomic_set (&conn->ksnc_sock_refcount, 2); + refcount_set(&conn->ksnc_sock_refcount, 2); conn->ksnc_type = type; ksocknal_lib_save_callback(sock, conn); - atomic_set (&conn->ksnc_conn_refcount, 1); /* 1 ref for me */ + refcount_set(&conn->ksnc_conn_refcount, 1); /* 1 ref for me */ conn->ksnc_rx_ready = 0; conn->ksnc_rx_scheduled = 0; @@ -1039,61 +830,62 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, LIBCFS_ALLOC(hello, offsetof(struct ksock_hello_msg, kshm_ips[LNET_INTERFACES_NUM])); - if (hello == NULL) { - rc = -ENOMEM; - goto failed_1; - } + if (hello == NULL) { + rc = -ENOMEM; + goto failed_1; + } - /* stash conn's local and remote addrs */ - rc = ksocknal_lib_get_conn_addrs (conn); - if (rc != 0) - goto failed_1; + /* stash conn's local and remote addrs */ + rc = ksocknal_lib_get_conn_addrs(conn); + if (rc != 0) + goto failed_1; - /* Find out/confirm peer_ni's NID and connection type and get the - * vector of interfaces she's willing to let me connect to. - * Passive connections use the listener timeout since the peer_ni sends - * eagerly */ + /* Find out/confirm peer_ni's NID and connection type and get the + * vector of interfaces she's willing to let me connect to. + * Passive connections use the listener timeout since the peer_ni sends + * eagerly + */ - if (active) { - peer_ni = route->ksnr_peer; - LASSERT(ni == peer_ni->ksnp_ni); + if (active) { + peer_ni = conn_cb->ksnr_peer; + LASSERT(ni == peer_ni->ksnp_ni); - /* Active connection sends HELLO eagerly */ - hello->kshm_nips = ksocknal_local_ipvec(ni, hello->kshm_ips); - peerid = peer_ni->ksnp_id; + /* Active connection sends HELLO eagerly */ + hello->kshm_nips = 0; + peerid = peer_ni->ksnp_id; write_lock_bh(global_lock); - conn->ksnc_proto = peer_ni->ksnp_proto; + conn->ksnc_proto = peer_ni->ksnp_proto; write_unlock_bh(global_lock); - if (conn->ksnc_proto == NULL) { - conn->ksnc_proto = &ksocknal_protocol_v3x; + if (conn->ksnc_proto == NULL) { + conn->ksnc_proto = &ksocknal_protocol_v3x; #if SOCKNAL_VERSION_DEBUG - if (*ksocknal_tunables.ksnd_protocol == 2) - conn->ksnc_proto = &ksocknal_protocol_v2x; - else if (*ksocknal_tunables.ksnd_protocol == 1) - conn->ksnc_proto = &ksocknal_protocol_v1x; + if (*ksocknal_tunables.ksnd_protocol == 2) + conn->ksnc_proto = &ksocknal_protocol_v2x; + else if (*ksocknal_tunables.ksnd_protocol == 1) + conn->ksnc_proto = &ksocknal_protocol_v1x; #endif - } + } - rc = ksocknal_send_hello (ni, conn, peerid.nid, hello); - if (rc != 0) - goto failed_1; - } else { - peerid.nid = LNET_NID_ANY; - peerid.pid = LNET_PID_ANY; + rc = ksocknal_send_hello(ni, conn, peerid.nid, hello); + if (rc != 0) + goto failed_1; + } else { + peerid.nid = LNET_NID_ANY; + peerid.pid = LNET_PID_ANY; - /* Passive, get protocol from peer_ni */ - conn->ksnc_proto = NULL; - } + /* Passive, get protocol from peer_ni */ + conn->ksnc_proto = NULL; + } - rc = ksocknal_recv_hello (ni, conn, hello, &peerid, &incarnation); - if (rc < 0) - goto failed_1; + rc = ksocknal_recv_hello(ni, conn, hello, &peerid, &incarnation); + if (rc < 0) + goto failed_1; - LASSERT (rc == 0 || active); - LASSERT (conn->ksnc_proto != NULL); - LASSERT (peerid.nid != LNET_NID_ANY); + LASSERT(rc == 0 || active); + LASSERT(conn->ksnc_proto != NULL); + LASSERT(peerid.nid != LNET_NID_ANY); cpt = lnet_cpt_of_nid(peerid.nid, ni); @@ -1123,26 +915,28 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, peer_ni = peer2; } - /* +1 ref for me */ - ksocknal_peer_addref(peer_ni); - peer_ni->ksnp_accepting++; - - /* Am I already connecting to this guy? Resolve in - * favour of higher NID... */ - if (peerid.nid < ni->ni_nid && - ksocknal_connecting(peer_ni, conn->ksnc_ipaddr)) { - rc = EALREADY; - warn = "connection race resolution"; - goto failed_2; - } - } + /* +1 ref for me */ + ksocknal_peer_addref(peer_ni); + peer_ni->ksnp_accepting++; - if (peer_ni->ksnp_closing || - (active && route->ksnr_deleted)) { - /* peer_ni/route got closed under me */ - rc = -ESTALE; - warn = "peer_ni/route removed"; - goto failed_2; + /* Am I already connecting to this guy? Resolve in + * favour of higher NID... + */ + if (peerid.nid < ni->ni_nid && + ksocknal_connecting(peer_ni->ksnp_conn_cb, + ((struct sockaddr *) &conn->ksnc_peeraddr))) { + rc = EALREADY; + warn = "connection race resolution"; + goto failed_2; + } + } + + if (peer_ni->ksnp_closing || + (active && conn_cb->ksnr_deleted)) { + /* peer_ni/conn_cb got closed under me */ + rc = -ESTALE; + warn = "peer_ni/conn_cb removed"; + goto failed_2; } if (peer_ni->ksnp_proto == NULL) { @@ -1156,78 +950,86 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, peer_ni->ksnp_incarnation = incarnation; } - if (peer_ni->ksnp_proto != conn->ksnc_proto || - peer_ni->ksnp_incarnation != incarnation) { - /* peer_ni rebooted or I've got the wrong protocol version */ - ksocknal_close_peer_conns_locked(peer_ni, 0, 0); - - peer_ni->ksnp_proto = NULL; - rc = ESTALE; - warn = peer_ni->ksnp_incarnation != incarnation ? - "peer_ni rebooted" : - "wrong proto version"; - goto failed_2; - } + if (peer_ni->ksnp_proto != conn->ksnc_proto || + peer_ni->ksnp_incarnation != incarnation) { + /* peer_ni rebooted or I've got the wrong protocol version */ + ksocknal_close_peer_conns_locked(peer_ni, NULL, 0); - switch (rc) { - default: - LBUG(); - case 0: - break; - case EALREADY: - warn = "lost conn race"; - goto failed_2; - case EPROTO: - warn = "retry with different protocol version"; - goto failed_2; - } + peer_ni->ksnp_proto = NULL; + rc = ESTALE; + warn = peer_ni->ksnp_incarnation != incarnation ? + "peer_ni rebooted" : + "wrong proto version"; + goto failed_2; + } + + switch (rc) { + default: + LBUG(); + case 0: + break; + case EALREADY: + warn = "lost conn race"; + goto failed_2; + case EPROTO: + warn = "retry with different protocol version"; + goto failed_2; + } /* Refuse to duplicate an existing connection, unless this is a * loopback connection */ - if (conn->ksnc_ipaddr != conn->ksnc_myipaddr) { + if (!rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr, + (struct sockaddr *)&conn->ksnc_myaddr)) { list_for_each(tmp, &peer_ni->ksnp_conns) { conn2 = list_entry(tmp, struct ksock_conn, ksnc_list); - if (conn2->ksnc_ipaddr != conn->ksnc_ipaddr || - conn2->ksnc_myipaddr != conn->ksnc_myipaddr || - conn2->ksnc_type != conn->ksnc_type) - continue; + if (!rpc_cmp_addr( + (struct sockaddr *)&conn2->ksnc_peeraddr, + (struct sockaddr *)&conn->ksnc_peeraddr) || + !rpc_cmp_addr( + (struct sockaddr *)&conn2->ksnc_myaddr, + (struct sockaddr *)&conn->ksnc_myaddr) || + conn2->ksnc_type != conn->ksnc_type) + continue; - /* Reply on a passive connection attempt so the peer_ni - * realises we're connected. */ - LASSERT (rc == 0); - if (!active) - rc = EALREADY; + num_dup++; + if (num_dup < *ksocknal_tunables.ksnd_conns_per_peer) + continue; - warn = "duplicate"; - goto failed_2; - } - } + /* Reply on a passive connection attempt so the peer_ni + * realises we're connected. + */ + LASSERT(rc == 0); + if (!active) + rc = EALREADY; - /* If the connection created by this route didn't bind to the IP - * address the route connected to, the connection/route matching - * code below probably isn't going to work. */ - if (active && - route->ksnr_ipaddr != conn->ksnc_ipaddr) { - CERROR("Route %s %pI4h connected to %pI4h\n", - libcfs_id2str(peer_ni->ksnp_id), - &route->ksnr_ipaddr, - &conn->ksnc_ipaddr); - } + warn = "duplicate"; + goto failed_2; + } + } + /* If the connection created by this route didn't bind to the IP + * address the route connected to, the connection/route matching + * code below probably isn't going to work. + */ + if (active && + !rpc_cmp_addr((struct sockaddr *)&conn_cb->ksnr_addr, + (struct sockaddr *)&conn->ksnc_peeraddr)) { + CERROR("Route %s %pIS connected to %pIS\n", + libcfs_id2str(peer_ni->ksnp_id), + &conn_cb->ksnr_addr, + &conn->ksnc_peeraddr); + } - /* Search for a route corresponding to the new connection and + /* Search for a conn_cb corresponding to the new connection and * create an association. This allows incoming connections created - * by routes in my peer_ni to match my own route entries so I don't - * continually create duplicate routes. */ - list_for_each(tmp, &peer_ni->ksnp_routes) { - route = list_entry(tmp, struct ksock_route, ksnr_list); - - if (route->ksnr_ipaddr != conn->ksnc_ipaddr) - continue; + * by conn_cbs in my peer_ni to match my own conn_cb entries so I don't + * continually create duplicate conn_cbs. + */ + conn_cb = peer_ni->ksnp_conn_cb; - ksocknal_associate_route_conn_locked(route, conn); - break; - } + if (conn_cb && rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr, + (struct sockaddr *)&conn_cb->ksnr_addr)) + ksocknal_associate_cb_conn_locked(conn_cb, conn); conn->ksnc_peer = peer_ni; /* conn takes my ref on peer_ni */ peer_ni->ksnp_last_alive = ktime_get_seconds(); @@ -1244,14 +1046,14 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, * native scheduler. So use the scheduler's cpt instead. */ cpt = sched->kss_cpt; - sched->kss_nconns++; - conn->ksnc_scheduler = sched; + sched->kss_nconns++; + conn->ksnc_scheduler = sched; conn->ksnc_tx_last_post = ktime_get_seconds(); /* Set the deadline for the outgoing HELLO to drain */ conn->ksnc_tx_bufnob = sock->sk->sk_wmem_queued; conn->ksnc_tx_deadline = ktime_get_seconds() + - lnet_get_lnd_timeout(); + ksocknal_timeout(); smp_mb(); /* order with adding to peer_ni's conn list */ list_add(&conn->ksnc_list, &peer_ni->ksnp_conns); @@ -1259,7 +1061,7 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, ksocknal_new_packet(conn, 0); - conn->ksnc_zc_capable = ksocknal_lib_zc_capable(conn); + conn->ksnc_zc_capable = ksocknal_lib_zc_capable(conn); /* Take packets blocking for this connection. */ list_for_each_entry_safe(tx, txtmp, &peer_ni->ksnp_tx_queue, tx_list) { @@ -1273,101 +1075,99 @@ ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route, write_unlock_bh(global_lock); - /* We've now got a new connection. Any errors from here on are just - * like "normal" comms errors and we close the connection normally. - * NB (a) we still have to send the reply HELLO for passive + /* We've now got a new connection. Any errors from here on are just + * like "normal" comms errors and we close the connection normally. + * NB (a) we still have to send the reply HELLO for passive * connections, - * (b) normal I/O on the conn is blocked until I setup and call the - * socket callbacks. - */ + * (b) normal I/O on the conn is blocked until I setup and call the + * socket callbacks. + */ - CDEBUG(D_NET, "New conn %s p %d.x %pI4h -> %pI4h/%d" + CDEBUG(D_NET, "New conn %s p %d.x %pIS -> %pISp" " incarnation:%lld sched[%d]\n", libcfs_id2str(peerid), conn->ksnc_proto->pro_version, - &conn->ksnc_myipaddr, &conn->ksnc_ipaddr, - conn->ksnc_port, incarnation, cpt); - - if (active) { - /* additional routes after interface exchange? */ - ksocknal_create_routes(peer_ni, conn->ksnc_port, - hello->kshm_ips, hello->kshm_nips); - } else { - hello->kshm_nips = ksocknal_select_ips(peer_ni, hello->kshm_ips, - hello->kshm_nips); - rc = ksocknal_send_hello(ni, conn, peerid.nid, hello); - } + &conn->ksnc_myaddr, &conn->ksnc_peeraddr, + incarnation, cpt); + + if (!active) { + hello->kshm_nips = 0; + rc = ksocknal_send_hello(ni, conn, peerid.nid, hello); + } LIBCFS_FREE(hello, offsetof(struct ksock_hello_msg, kshm_ips[LNET_INTERFACES_NUM])); - /* setup the socket AFTER I've received hello (it disables - * SO_LINGER). I might call back to the acceptor who may want - * to send a protocol version response and then close the - * socket; this ensures the socket only tears down after the - * response has been sent. */ - if (rc == 0) - rc = ksocknal_lib_setup_sock(sock); + /* setup the socket AFTER I've received hello (it disables + * SO_LINGER). I might call back to the acceptor who may want + * to send a protocol version response and then close the + * socket; this ensures the socket only tears down after the + * response has been sent. + */ + if (rc == 0) + rc = ksocknal_lib_setup_sock(sock); write_lock_bh(global_lock); - /* NB my callbacks block while I hold ksnd_global_lock */ - ksocknal_lib_set_callback(sock, conn); + /* NB my callbacks block while I hold ksnd_global_lock */ + ksocknal_lib_set_callback(sock, conn); - if (!active) - peer_ni->ksnp_accepting--; + if (!active) + peer_ni->ksnp_accepting--; write_unlock_bh(global_lock); - if (rc != 0) { + if (rc != 0) { write_lock_bh(global_lock); - if (!conn->ksnc_closing) { - /* could be closed by another thread */ - ksocknal_close_conn_locked(conn, rc); - } + if (!conn->ksnc_closing) { + /* could be closed by another thread */ + ksocknal_close_conn_locked(conn, rc); + } write_unlock_bh(global_lock); - } else if (ksocknal_connsock_addref(conn) == 0) { - /* Allow I/O to proceed. */ - ksocknal_read_callback(conn); - ksocknal_write_callback(conn); - ksocknal_connsock_decref(conn); - } + } else if (ksocknal_connsock_addref(conn) == 0) { + /* Allow I/O to proceed. */ + ksocknal_read_callback(conn); + ksocknal_write_callback(conn); + ksocknal_connsock_decref(conn); + } - ksocknal_connsock_decref(conn); - ksocknal_conn_decref(conn); - return rc; + ksocknal_connsock_decref(conn); + ksocknal_conn_decref(conn); + return rc; failed_2: + if (!peer_ni->ksnp_closing && list_empty(&peer_ni->ksnp_conns) && - list_empty(&peer_ni->ksnp_routes)) { + peer_ni->ksnp_conn_cb == NULL) { list_splice_init(&peer_ni->ksnp_tx_queue, &zombies); ksocknal_unlink_peer_locked(peer_ni); } write_unlock_bh(global_lock); - if (warn != NULL) { - if (rc < 0) - CERROR("Not creating conn %s type %d: %s\n", - libcfs_id2str(peerid), conn->ksnc_type, warn); - else - CDEBUG(D_NET, "Not creating conn %s type %d: %s\n", - libcfs_id2str(peerid), conn->ksnc_type, warn); - } + if (warn != NULL) { + if (rc < 0) + CERROR("Not creating conn %s type %d: %s\n", + libcfs_id2str(peerid), conn->ksnc_type, warn); + else + CDEBUG(D_NET, "Not creating conn %s type %d: %s\n", + libcfs_id2str(peerid), conn->ksnc_type, warn); + } - if (!active) { - if (rc > 0) { + if (!active) { + if (rc > 0) { /* Request retry by replying with CONN_NONE - * ksnc_proto has been set already */ - conn->ksnc_type = SOCKLND_CONN_NONE; - hello->kshm_nips = 0; - ksocknal_send_hello(ni, conn, peerid.nid, hello); - } + * ksnc_proto has been set already + */ + conn->ksnc_type = SOCKLND_CONN_NONE; + hello->kshm_nips = 0; + ksocknal_send_hello(ni, conn, peerid.nid, hello); + } write_lock_bh(global_lock); - peer_ni->ksnp_accepting--; + peer_ni->ksnp_accepting--; write_unlock_bh(global_lock); - } + } /* * If we get here without an error code, just use -EALREADY. @@ -1376,7 +1176,7 @@ failed_2: */ rc2 = (rc == 0 ? -EALREADY : (rc > 0 ? -rc : rc)); ksocknal_txlist_done(ni, &zombies, rc2); - ksocknal_peer_decref(peer_ni); + ksocknal_peer_decref(peer_ni); failed_1: if (hello != NULL) @@ -1387,6 +1187,7 @@ failed_1: failed_0: sock_release(sock); + return rc; } @@ -1397,7 +1198,7 @@ ksocknal_close_conn_locked(struct ksock_conn *conn, int error) * connection for the reaper to terminate. * Caller holds ksnd_global_lock exclusively in irq context */ struct ksock_peer_ni *peer_ni = conn->ksnc_peer; - struct ksock_route *route; + struct ksock_conn_cb *conn_cb; struct ksock_conn *conn2; struct list_head *tmp; @@ -1408,28 +1209,36 @@ ksocknal_close_conn_locked(struct ksock_conn *conn, int error) /* ksnd_deathrow_conns takes over peer_ni's ref */ list_del(&conn->ksnc_list); - route = conn->ksnc_route; - if (route != NULL) { - /* dissociate conn from route... */ - LASSERT(!route->ksnr_deleted); - LASSERT((route->ksnr_connected & (1 << conn->ksnc_type)) != 0); + conn_cb = conn->ksnc_conn_cb; + if (conn_cb != NULL) { + /* dissociate conn from cb... */ + LASSERT(!conn_cb->ksnr_deleted); + + /* connected bit is set only if all connections + * of the given type got created + */ + if (ksocknal_get_conn_count_by_type(conn_cb, conn->ksnc_type) == + *ksocknal_tunables.ksnd_conns_per_peer) + LASSERT((conn_cb->ksnr_connected & + BIT(conn->ksnc_type)) != 0); conn2 = NULL; list_for_each(tmp, &peer_ni->ksnp_conns) { conn2 = list_entry(tmp, struct ksock_conn, ksnc_list); - if (conn2->ksnc_route == route && + if (conn2->ksnc_conn_cb == conn_cb && conn2->ksnc_type == conn->ksnc_type) break; conn2 = NULL; } if (conn2 == NULL) - route->ksnr_connected &= ~(1 << conn->ksnc_type); + conn_cb->ksnr_connected &= ~BIT(conn->ksnc_type); - conn->ksnc_route = NULL; + conn->ksnc_conn_cb = NULL; - ksocknal_route_decref(route); /* drop conn's ref on route */ + /* drop conn's ref on conn_cb */ + ksocknal_conn_cb_decref(conn_cb); } if (list_empty(&peer_ni->ksnp_conns)) { @@ -1457,9 +1266,10 @@ ksocknal_close_conn_locked(struct ksock_conn *conn, int error) /* stash last conn close reason */ peer_ni->ksnp_error = error; - if (list_empty(&peer_ni->ksnp_routes)) { + if (peer_ni->ksnp_conn_cb == NULL) { /* I've just closed last conn belonging to a - * peer_ni with no routes to it */ + * peer_ni with no connections to it + */ ksocknal_unlink_peer_locked(peer_ni); } } @@ -1475,7 +1285,7 @@ ksocknal_close_conn_locked(struct ksock_conn *conn, int error) void ksocknal_peer_failed(struct ksock_peer_ni *peer_ni) { - int notify = 0; + bool notify = false; time64_t last_alive = 0; /* There has been a connection failure or comms error; but I'll only @@ -1487,8 +1297,8 @@ ksocknal_peer_failed(struct ksock_peer_ni *peer_ni) if ((peer_ni->ksnp_id.pid & LNET_PID_USERFLAG) == 0 && list_empty(&peer_ni->ksnp_conns) && peer_ni->ksnp_accepting == 0 && - ksocknal_find_connecting_route_locked(peer_ni) == NULL) { - notify = 1; + !ksocknal_find_connecting_conn_cb_locked(peer_ni)) { + notify = true; last_alive = peer_ni->ksnp_last_alive; } @@ -1526,9 +1336,8 @@ ksocknal_finalize_zcreq(struct ksock_conn *conn) spin_unlock(&peer_ni->ksnp_lock); - while (!list_empty(&zlist)) { - tx = list_entry(zlist.next, struct ksock_tx, tx_zc_list); - + while ((tx = list_first_entry_or_null(&zlist, struct ksock_tx, + tx_zc_list)) != NULL) { list_del(&tx->tx_zc_list); ksocknal_tx_decref(tx); } @@ -1537,31 +1346,32 @@ ksocknal_finalize_zcreq(struct ksock_conn *conn) void ksocknal_terminate_conn(struct ksock_conn *conn) { - /* This gets called by the reaper (guaranteed thread context) to - * disengage the socket from its callbacks and close it. - * ksnc_refcount will eventually hit zero, and then the reaper will - * destroy it. */ + /* This gets called by the reaper (guaranteed thread context) to + * disengage the socket from its callbacks and close it. + * ksnc_refcount will eventually hit zero, and then the reaper will + * destroy it. + */ struct ksock_peer_ni *peer_ni = conn->ksnc_peer; struct ksock_sched *sched = conn->ksnc_scheduler; - int failed = 0; + bool failed = false; - LASSERT(conn->ksnc_closing); + LASSERT(conn->ksnc_closing); - /* wake up the scheduler to "send" all remaining packets to /dev/null */ + /* wake up the scheduler to "send" all remaining packets to /dev/null */ spin_lock_bh(&sched->kss_lock); - /* a closing conn is always ready to tx */ - conn->ksnc_tx_ready = 1; + /* a closing conn is always ready to tx */ + conn->ksnc_tx_ready = 1; - if (!conn->ksnc_tx_scheduled && + if (!conn->ksnc_tx_scheduled && !list_empty(&conn->ksnc_tx_queue)) { list_add_tail(&conn->ksnc_tx_list, - &sched->kss_tx_conns); - conn->ksnc_tx_scheduled = 1; - /* extra ref for scheduler */ + &sched->kss_tx_conns); + conn->ksnc_tx_scheduled = 1; + /* extra ref for scheduler */ ksocknal_conn_addref(conn); - wake_up (&sched->kss_waitq); + wake_up(&sched->kss_waitq); } spin_unlock_bh(&sched->kss_lock); @@ -1569,37 +1379,39 @@ ksocknal_terminate_conn(struct ksock_conn *conn) /* serialise with callbacks */ write_lock_bh(&ksocknal_data.ksnd_global_lock); - ksocknal_lib_reset_callback(conn->ksnc_sock, conn); + ksocknal_lib_reset_callback(conn->ksnc_sock, conn); - /* OK, so this conn may not be completely disengaged from its - * scheduler yet, but it _has_ committed to terminate... */ - conn->ksnc_scheduler->kss_nconns--; + /* OK, so this conn may not be completely disengaged from its + * scheduler yet, but it _has_ committed to terminate... + */ + conn->ksnc_scheduler->kss_nconns--; - if (peer_ni->ksnp_error != 0) { - /* peer_ni's last conn closed in error */ + if (peer_ni->ksnp_error != 0) { + /* peer_ni's last conn closed in error */ LASSERT(list_empty(&peer_ni->ksnp_conns)); - failed = 1; - peer_ni->ksnp_error = 0; /* avoid multiple notifications */ - } + failed = true; + peer_ni->ksnp_error = 0; /* avoid multiple notifications */ + } write_unlock_bh(&ksocknal_data.ksnd_global_lock); - if (failed) - ksocknal_peer_failed(peer_ni); + if (failed) + ksocknal_peer_failed(peer_ni); - /* The socket is closed on the final put; either here, or in - * ksocknal_{send,recv}msg(). Since we set up the linger2 option - * when the connection was established, this will close the socket - * immediately, aborting anything buffered in it. Any hung - * zero-copy transmits will therefore complete in finite time. */ - ksocknal_connsock_decref(conn); + /* The socket is closed on the final put; either here, or in + * ksocknal_{send,recv}msg(). Since we set up the linger2 option + * when the connection was established, this will close the socket + * immediately, aborting anything buffered in it. Any hung + * zero-copy transmits will therefore complete in finite time. + */ + ksocknal_connsock_decref(conn); } void ksocknal_queue_zombie_conn(struct ksock_conn *conn) { /* Queue the conn for the reaper to destroy */ - LASSERT(atomic_read(&conn->ksnc_conn_refcount) == 0); + LASSERT(refcount_read(&conn->ksnc_conn_refcount) == 0); spin_lock_bh(&ksocknal_data.ksnd_reaper_lock); list_add_tail(&conn->ksnc_list, &ksocknal_data.ksnd_zombie_conns); @@ -1614,26 +1426,24 @@ ksocknal_destroy_conn(struct ksock_conn *conn) time64_t last_rcv; /* Final coup-de-grace of the reaper */ - CDEBUG (D_NET, "connection %p\n", conn); - - LASSERT (atomic_read (&conn->ksnc_conn_refcount) == 0); - LASSERT (atomic_read (&conn->ksnc_sock_refcount) == 0); - LASSERT (conn->ksnc_sock == NULL); - LASSERT (conn->ksnc_route == NULL); - LASSERT (!conn->ksnc_tx_scheduled); - LASSERT (!conn->ksnc_rx_scheduled); + CDEBUG(D_NET, "connection %p\n", conn); + + LASSERT(refcount_read(&conn->ksnc_conn_refcount) == 0); + LASSERT(refcount_read(&conn->ksnc_sock_refcount) == 0); + LASSERT(conn->ksnc_sock == NULL); + LASSERT(conn->ksnc_conn_cb == NULL); + LASSERT(!conn->ksnc_tx_scheduled); + LASSERT(!conn->ksnc_rx_scheduled); LASSERT(list_empty(&conn->ksnc_tx_queue)); /* complete current receive if any */ switch (conn->ksnc_rx_state) { case SOCKNAL_RX_LNET_PAYLOAD: last_rcv = conn->ksnc_rx_deadline - - lnet_get_lnd_timeout(); - CERROR("Completing partial receive from %s[%d], " - "ip %pI4h:%d, with error, wanted: %d, left: %d, " - "last alive is %lld secs ago\n", + ksocknal_timeout(); + CERROR("Completing partial receive from %s[%d], ip %pISp, with error, wanted: %d, left: %d, last alive is %lld secs ago\n", libcfs_id2str(conn->ksnc_peer->ksnp_id), conn->ksnc_type, - &conn->ksnc_ipaddr, conn->ksnc_port, + &conn->ksnc_peeraddr, conn->ksnc_rx_nob_wanted, conn->ksnc_rx_nob_left, ktime_get_seconds() - last_rcv); if (conn->ksnc_lnet_msg) @@ -1641,28 +1451,25 @@ ksocknal_destroy_conn(struct ksock_conn *conn) LNET_MSG_STATUS_REMOTE_ERROR; lnet_finalize(conn->ksnc_lnet_msg, -EIO); break; - case SOCKNAL_RX_LNET_HEADER: - if (conn->ksnc_rx_started) - CERROR("Incomplete receive of lnet header from %s, " - "ip %pI4h:%d, with error, protocol: %d.x.\n", - libcfs_id2str(conn->ksnc_peer->ksnp_id), - &conn->ksnc_ipaddr, conn->ksnc_port, - conn->ksnc_proto->pro_version); - break; + case SOCKNAL_RX_LNET_HEADER: + if (conn->ksnc_rx_started) + CERROR("Incomplete receive of lnet header from %s, ip %pISp, with error, protocol: %d.x.\n", + libcfs_id2str(conn->ksnc_peer->ksnp_id), + &conn->ksnc_peeraddr, + conn->ksnc_proto->pro_version); + break; case SOCKNAL_RX_KSM_HEADER: if (conn->ksnc_rx_started) - CERROR("Incomplete receive of ksock message from %s, " - "ip %pI4h:%d, with error, protocol: %d.x.\n", - libcfs_id2str(conn->ksnc_peer->ksnp_id), - &conn->ksnc_ipaddr, conn->ksnc_port, - conn->ksnc_proto->pro_version); + CERROR("Incomplete receive of ksock message from %s, ip %pISp, with error, protocol: %d.x.\n", + libcfs_id2str(conn->ksnc_peer->ksnp_id), + &conn->ksnc_peeraddr, + conn->ksnc_proto->pro_version); break; case SOCKNAL_RX_SLOP: if (conn->ksnc_rx_started) - CERROR("Incomplete receive of slops from %s, " - "ip %pI4h:%d, with error\n", - libcfs_id2str(conn->ksnc_peer->ksnp_id), - &conn->ksnc_ipaddr, conn->ksnc_port); + CERROR("Incomplete receive of slops from %s, ip %pISp, with error\n", + libcfs_id2str(conn->ksnc_peer->ksnp_id), + &conn->ksnc_peeraddr); break; default: LBUG (); @@ -1675,40 +1482,39 @@ ksocknal_destroy_conn(struct ksock_conn *conn) } int -ksocknal_close_peer_conns_locked(struct ksock_peer_ni *peer_ni, __u32 ipaddr, int why) +ksocknal_close_peer_conns_locked(struct ksock_peer_ni *peer_ni, + struct sockaddr *addr, int why) { struct ksock_conn *conn; - struct list_head *ctmp; - struct list_head *cnxt; + struct ksock_conn *cnxt; int count = 0; - list_for_each_safe(ctmp, cnxt, &peer_ni->ksnp_conns) { - conn = list_entry(ctmp, struct ksock_conn, ksnc_list); - - if (ipaddr == 0 || - conn->ksnc_ipaddr == ipaddr) { - count++; - ksocknal_close_conn_locked (conn, why); - } - } + list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns, ksnc_list) { + if (!addr || + rpc_cmp_addr(addr, + (struct sockaddr *)&conn->ksnc_peeraddr)) { + count++; + ksocknal_close_conn_locked(conn, why); + } + } - return (count); + return count; } int ksocknal_close_conn_and_siblings(struct ksock_conn *conn, int why) { struct ksock_peer_ni *peer_ni = conn->ksnc_peer; - u32 ipaddr = conn->ksnc_ipaddr; int count; write_lock_bh(&ksocknal_data.ksnd_global_lock); - count = ksocknal_close_peer_conns_locked (peer_ni, ipaddr, why); + count = ksocknal_close_peer_conns_locked( + peer_ni, (struct sockaddr *)&conn->ksnc_peeraddr, why); write_unlock_bh(&ksocknal_data.ksnd_global_lock); - return (count); + return count; } int @@ -1720,6 +1526,7 @@ ksocknal_close_matching_conns(struct lnet_process_id id, __u32 ipaddr) int hi; int i; int count = 0; + struct sockaddr_in sa = {.sin_family = AF_INET}; write_lock_bh(&ksocknal_data.ksnd_global_lock); @@ -1731,6 +1538,7 @@ ksocknal_close_matching_conns(struct lnet_process_id id, __u32 ipaddr) hi = HASH_SIZE(ksocknal_data.ksnd_peers) - 1; } + sa.sin_addr.s_addr = htonl(ipaddr); for (i = lo; i <= hi; i++) { hlist_for_each_entry_safe(peer_ni, pnxt, &ksocknal_data.ksnd_peers[i], @@ -1742,8 +1550,9 @@ ksocknal_close_matching_conns(struct lnet_process_id id, __u32 ipaddr) id.pid == peer_ni->ksnp_id.pid))) continue; - count += ksocknal_close_peer_conns_locked(peer_ni, - ipaddr, 0); + count += ksocknal_close_peer_conns_locked( + peer_ni, + ipaddr ? (struct sockaddr *)&sa : NULL, 0); } } @@ -1861,186 +1670,41 @@ ksocknal_push(struct lnet_ni *ni, struct lnet_process_id id) return rc; } -static int -ksocknal_add_interface(struct lnet_ni *ni, __u32 ipaddress, __u32 netmask) -{ - struct ksock_net *net = ni->ni_data; - struct ksock_interface *iface; - int rc; - int i; - int j; - struct ksock_peer_ni *peer_ni; - struct list_head *rtmp; - struct ksock_route *route; - - if (ipaddress == 0 || - netmask == 0) - return -EINVAL; - - write_lock_bh(&ksocknal_data.ksnd_global_lock); - - iface = ksocknal_ip2iface(ni, ipaddress); - if (iface != NULL) { - /* silently ignore dups */ - rc = 0; - } else if (net->ksnn_ninterfaces == LNET_INTERFACES_NUM) { - rc = -ENOSPC; - } else { - iface = &net->ksnn_interfaces[net->ksnn_ninterfaces++]; - - iface->ksni_ipaddr = ipaddress; - iface->ksni_netmask = netmask; - iface->ksni_nroutes = 0; - iface->ksni_npeers = 0; - - hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list) { - for (j = 0; j < peer_ni->ksnp_n_passive_ips; j++) - if (peer_ni->ksnp_passive_ips[j] == ipaddress) - iface->ksni_npeers++; - - list_for_each(rtmp, &peer_ni->ksnp_routes) { - route = list_entry(rtmp, - struct ksock_route, - ksnr_list); - - if (route->ksnr_myipaddr == ipaddress) - iface->ksni_nroutes++; - } - } - - rc = 0; - /* NB only new connections will pay attention to the new - * interface! - */ - } - - write_unlock_bh(&ksocknal_data.ksnd_global_lock); - - return rc; -} - -static void -ksocknal_peer_del_interface_locked(struct ksock_peer_ni *peer_ni, __u32 ipaddr) -{ - struct list_head *tmp; - struct list_head *nxt; - struct ksock_route *route; - struct ksock_conn *conn; - int i; - int j; - - for (i = 0; i < peer_ni->ksnp_n_passive_ips; i++) - if (peer_ni->ksnp_passive_ips[i] == ipaddr) { - for (j = i+1; j < peer_ni->ksnp_n_passive_ips; j++) - peer_ni->ksnp_passive_ips[j-1] = - peer_ni->ksnp_passive_ips[j]; - peer_ni->ksnp_n_passive_ips--; - break; - } - - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) { - route = list_entry(tmp, struct ksock_route, ksnr_list); - - if (route->ksnr_myipaddr != ipaddr) - continue; - - if (route->ksnr_share_count != 0) { - /* Manually created; keep, but unbind */ - route->ksnr_myipaddr = 0; - } else { - ksocknal_del_route_locked(route); - } - } - - list_for_each_safe(tmp, nxt, &peer_ni->ksnp_conns) { - conn = list_entry(tmp, struct ksock_conn, ksnc_list); - - if (conn->ksnc_myipaddr == ipaddr) - ksocknal_close_conn_locked (conn, 0); - } -} - -static int -ksocknal_del_interface(struct lnet_ni *ni, __u32 ipaddress) -{ - struct ksock_net *net = ni->ni_data; - int rc = -ENOENT; - struct hlist_node *nxt; - struct ksock_peer_ni *peer_ni; - u32 this_ip; - int i; - int j; - - write_lock_bh(&ksocknal_data.ksnd_global_lock); - - for (i = 0; i < net->ksnn_ninterfaces; i++) { - this_ip = net->ksnn_interfaces[i].ksni_ipaddr; - - if (!(ipaddress == 0 || - ipaddress == this_ip)) - continue; - - rc = 0; - - for (j = i+1; j < net->ksnn_ninterfaces; j++) - net->ksnn_interfaces[j-1] = - net->ksnn_interfaces[j]; - - net->ksnn_ninterfaces--; - - hash_for_each_safe(ksocknal_data.ksnd_peers, j, - nxt, peer_ni, ksnp_list) { - if (peer_ni->ksnp_ni != ni) - continue; - - ksocknal_peer_del_interface_locked(peer_ni, this_ip); - } - } - - write_unlock_bh(&ksocknal_data.ksnd_global_lock); - - return rc; -} - int ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg) { struct lnet_process_id id = {0}; - struct libcfs_ioctl_data *data = arg; - int rc; + struct libcfs_ioctl_data *data = arg; + int rc; - switch(cmd) { - case IOC_LIBCFS_GET_INTERFACE: { + switch(cmd) { + case IOC_LIBCFS_GET_INTERFACE: { struct ksock_net *net = ni->ni_data; struct ksock_interface *iface; + struct sockaddr_in *sa; read_lock(&ksocknal_data.ksnd_global_lock); - if (data->ioc_count >= (__u32)net->ksnn_ninterfaces) { - rc = -ENOENT; - } else { - rc = 0; - iface = &net->ksnn_interfaces[data->ioc_count]; - - data->ioc_u32[0] = iface->ksni_ipaddr; - data->ioc_u32[1] = iface->ksni_netmask; - data->ioc_u32[2] = iface->ksni_npeers; - data->ioc_u32[3] = iface->ksni_nroutes; - } + if (data->ioc_count >= 1) { + rc = -ENOENT; + } else { + rc = 0; + iface = &net->ksnn_interface; + + sa = (void *)&iface->ksni_addr; + if (sa->sin_family == AF_INET) + data->ioc_u32[0] = ntohl(sa->sin_addr.s_addr); + else + data->ioc_u32[0] = 0xFFFFFFFF; + data->ioc_u32[1] = iface->ksni_netmask; + data->ioc_u32[2] = iface->ksni_npeers; + data->ioc_u32[3] = iface->ksni_nroutes; + } read_unlock(&ksocknal_data.ksnd_global_lock); - return rc; + return rc; } - case IOC_LIBCFS_ADD_INTERFACE: - return ksocknal_add_interface(ni, - data->ioc_u32[0], /* IP address */ - data->ioc_u32[1]); /* net mask */ - - case IOC_LIBCFS_DEL_INTERFACE: - return ksocknal_del_interface(ni, - data->ioc_u32[0]); /* IP address */ - case IOC_LIBCFS_GET_PEER: { __u32 myip = 0; __u32 ip = 0; @@ -2064,13 +1728,15 @@ ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg) return 0; } - case IOC_LIBCFS_ADD_PEER: - id.nid = data->ioc_nid; - id.pid = LNET_PID_LUSTRE; - return ksocknal_add_peer (ni, id, - data->ioc_u32[0], /* IP */ - data->ioc_u32[1]); /* port */ + case IOC_LIBCFS_ADD_PEER: { + struct sockaddr_in sa = {.sin_family = AF_INET}; + id.nid = data->ioc_nid; + id.pid = LNET_PID_LUSTRE; + sa.sin_addr.s_addr = htonl(data->ioc_u32[0]); + sa.sin_port = htons(data->ioc_u32[1]); + return ksocknal_add_peer(ni, id, (struct sockaddr *)&sa); + } case IOC_LIBCFS_DEL_PEER: id.nid = data->ioc_nid; id.pid = LNET_PID_ANY; @@ -2082,6 +1748,8 @@ ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg) int rxmem; int nagle; struct ksock_conn *conn = ksocknal_get_conn_by_idx(ni, data->ioc_count); + struct sockaddr_in *psa = (void *)&conn->ksnc_peeraddr; + struct sockaddr_in *mysa = (void *)&conn->ksnc_myaddr; if (conn == NULL) return -ENOENT; @@ -2091,10 +1759,17 @@ ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg) data->ioc_count = txmem; data->ioc_nid = conn->ksnc_peer->ksnp_id.nid; data->ioc_flags = nagle; - data->ioc_u32[0] = conn->ksnc_ipaddr; - data->ioc_u32[1] = conn->ksnc_port; - data->ioc_u32[2] = conn->ksnc_myipaddr; - data->ioc_u32[3] = conn->ksnc_type; + if (psa->sin_family == AF_INET) + data->ioc_u32[0] = ntohl(psa->sin_addr.s_addr); + else + data->ioc_u32[0] = 0xFFFFFFFF; + data->ioc_u32[1] = rpc_get_port((struct sockaddr *) + &conn->ksnc_peeraddr); + if (mysa->sin_family == AF_INET) + data->ioc_u32[2] = ntohl(mysa->sin_addr.s_addr); + else + data->ioc_u32[2] = 0xFFFFFFFF; + data->ioc_u32[3] = conn->ksnc_type; data->ioc_u32[4] = conn->ksnc_scheduler->kss_cpt; data->ioc_u32[5] = rxmem; data->ioc_u32[6] = conn->ksnc_peer->ksnp_id.pid; @@ -2146,8 +1821,8 @@ ksocknal_free_buffers (void) list_splice_init(&ksocknal_data.ksnd_idle_noop_txs, &zlist); spin_unlock(&ksocknal_data.ksnd_tx_lock); - while (!list_empty(&zlist)) { - tx = list_entry(zlist.next, struct ksock_tx, tx_list); + while ((tx = list_first_entry_or_null(&zlist, struct ksock_tx, + tx_list)) != NULL) { list_del(&tx->tx_list); LIBCFS_FREE(tx, tx->tx_desc_size); } @@ -2163,8 +1838,8 @@ ksocknal_base_shutdown(void) struct ksock_peer_ni *peer_ni; int i; - CDEBUG(D_MALLOC, "before NAL cleanup: kmem %d\n", - atomic_read (&libcfs_kmemory)); + CDEBUG(D_MALLOC, "before NAL cleanup: kmem %lld\n", + libcfs_kmem_read()); LASSERT (ksocknal_data.ksnd_nnets == 0); switch (ksocknal_data.ksnd_init) { @@ -2197,7 +1872,7 @@ ksocknal_base_shutdown(void) /* flag threads to terminate; wake and wait for them to die */ ksocknal_data.ksnd_shuttingdown = 1; wake_up_all(&ksocknal_data.ksnd_connd_waitq); - wake_up_all(&ksocknal_data.ksnd_reaper_waitq); + wake_up(&ksocknal_data.ksnd_reaper_waitq); if (ksocknal_data.ksnd_schedulers != NULL) { cfs_percpt_for_each(sched, i, @@ -2205,28 +1880,19 @@ ksocknal_base_shutdown(void) wake_up_all(&sched->kss_waitq); } - i = 4; - read_lock(&ksocknal_data.ksnd_global_lock); - while (ksocknal_data.ksnd_nthreads != 0) { - i++; - /* power of 2? */ - CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, - "waiting for %d threads to terminate\n", - ksocknal_data.ksnd_nthreads); - read_unlock(&ksocknal_data.ksnd_global_lock); - schedule_timeout_uninterruptible(cfs_time_seconds(1)); - read_lock(&ksocknal_data.ksnd_global_lock); - } - read_unlock(&ksocknal_data.ksnd_global_lock); + wait_var_event_warning(&ksocknal_data.ksnd_nthreads, + atomic_read(&ksocknal_data.ksnd_nthreads) == 0, + "waiting for %d threads to terminate\n", + atomic_read(&ksocknal_data.ksnd_nthreads)); - ksocknal_free_buffers(); + ksocknal_free_buffers(); - ksocknal_data.ksnd_init = SOCKNAL_INIT_NOTHING; - break; - } + ksocknal_data.ksnd_init = SOCKNAL_INIT_NOTHING; + break; + } - CDEBUG(D_MALLOC, "after NAL cleanup: kmem %d\n", - atomic_read (&libcfs_kmemory)); + CDEBUG(D_MALLOC, "after NAL cleanup: kmem %lld\n", + libcfs_kmem_read()); module_put(THIS_MODULE); } @@ -2352,7 +2018,7 @@ ksocknal_base_startup(void) return -ENETDOWN; } -static void +static int ksocknal_debug_peerhash(struct lnet_ni *ni) { struct ksock_peer_ni *peer_ni; @@ -2361,7 +2027,7 @@ ksocknal_debug_peerhash(struct lnet_ni *ni) read_lock(&ksocknal_data.ksnd_global_lock); hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list) { - struct ksock_route *route; + struct ksock_conn_cb *conn_cb; struct ksock_conn *conn; if (peer_ni->ksnp_ni != ni) @@ -2370,30 +2036,32 @@ ksocknal_debug_peerhash(struct lnet_ni *ni) CWARN("Active peer_ni on shutdown: %s, ref %d, " "closing %d, accepting %d, err %d, zcookie %llu, " "txq %d, zc_req %d\n", libcfs_id2str(peer_ni->ksnp_id), - atomic_read(&peer_ni->ksnp_refcount), + refcount_read(&peer_ni->ksnp_refcount), peer_ni->ksnp_closing, peer_ni->ksnp_accepting, peer_ni->ksnp_error, peer_ni->ksnp_zc_next_cookie, !list_empty(&peer_ni->ksnp_tx_queue), !list_empty(&peer_ni->ksnp_zc_req_list)); - list_for_each_entry(route, &peer_ni->ksnp_routes, ksnr_list) { - CWARN("Route: ref %d, schd %d, conn %d, cnted %d, " - "del %d\n", atomic_read(&route->ksnr_refcount), - route->ksnr_scheduled, route->ksnr_connecting, - route->ksnr_connected, route->ksnr_deleted); + conn_cb = peer_ni->ksnp_conn_cb; + if (conn_cb) { + CWARN("ConnCB: ref %d, schd %d, conn %d, cnted %d, del %d\n", + refcount_read(&conn_cb->ksnr_refcount), + conn_cb->ksnr_scheduled, conn_cb->ksnr_connecting, + conn_cb->ksnr_connected, conn_cb->ksnr_deleted); } list_for_each_entry(conn, &peer_ni->ksnp_conns, ksnc_list) { CWARN("Conn: ref %d, sref %d, t %d, c %d\n", - atomic_read(&conn->ksnc_conn_refcount), - atomic_read(&conn->ksnc_sock_refcount), + refcount_read(&conn->ksnc_conn_refcount), + refcount_read(&conn->ksnc_sock_refcount), conn->ksnc_type, conn->ksnc_closing); } break; } read_unlock(&ksocknal_data.ksnd_global_lock); + return 0; } void @@ -2404,7 +2072,6 @@ ksocknal_shutdown(struct lnet_ni *ni) .nid = LNET_NID_ANY, .pid = LNET_PID_ANY, }; - int i; LASSERT(ksocknal_data.ksnd_init == SOCKNAL_INIT_ALL); LASSERT(ksocknal_data.ksnd_nnets > 0); @@ -2416,69 +2083,53 @@ ksocknal_shutdown(struct lnet_ni *ni) ksocknal_del_peer(ni, anyid, 0); /* Wait for all peer_ni state to clean up */ - i = 2; - while (atomic_read(&net->ksnn_npeers) > SOCKNAL_SHUTDOWN_BIAS) { - i++; - CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* power of 2? */ - "waiting for %d peers to disconnect\n", - atomic_read(&net->ksnn_npeers) - SOCKNAL_SHUTDOWN_BIAS); - schedule_timeout_uninterruptible(cfs_time_seconds(1)); - - ksocknal_debug_peerhash(ni); - } + wait_var_event_warning(&net->ksnn_npeers, + atomic_read(&net->ksnn_npeers) == + SOCKNAL_SHUTDOWN_BIAS, + "waiting for %d peers to disconnect\n", + ksocknal_debug_peerhash(ni) + + atomic_read(&net->ksnn_npeers) - + SOCKNAL_SHUTDOWN_BIAS); - for (i = 0; i < net->ksnn_ninterfaces; i++) { - LASSERT(net->ksnn_interfaces[i].ksni_npeers == 0); - LASSERT(net->ksnn_interfaces[i].ksni_nroutes == 0); - } + LASSERT(net->ksnn_interface.ksni_npeers == 0); + LASSERT(net->ksnn_interface.ksni_nroutes == 0); list_del(&net->ksnn_list); LIBCFS_FREE(net, sizeof(*net)); - ksocknal_data.ksnd_nnets--; - if (ksocknal_data.ksnd_nnets == 0) - ksocknal_base_shutdown(); + ksocknal_data.ksnd_nnets--; + if (ksocknal_data.ksnd_nnets == 0) + ksocknal_base_shutdown(); } static int ksocknal_search_new_ipif(struct ksock_net *net) { int new_ipif = 0; - int i; + char *ifnam = &net->ksnn_interface.ksni_name[0]; + char *colon = strchr(ifnam, ':'); + bool found = false; + struct ksock_net *tmp; - for (i = 0; i < net->ksnn_ninterfaces; i++) { - char *ifnam = &net->ksnn_interfaces[i].ksni_name[0]; - char *colon = strchr(ifnam, ':'); - int found = 0; - struct ksock_net *tmp; - int j; - - if (colon != NULL) /* ignore alias device */ - *colon = 0; - - list_for_each_entry(tmp, &ksocknal_data.ksnd_nets, - ksnn_list) { - for (j = 0; !found && j < tmp->ksnn_ninterfaces; j++) { - char *ifnam2 = &tmp->ksnn_interfaces[j].\ - ksni_name[0]; - char *colon2 = strchr(ifnam2, ':'); - - if (colon2 != NULL) - *colon2 = 0; - - found = strcmp(ifnam, ifnam2) == 0; - if (colon2 != NULL) - *colon2 = ':'; - } - if (found) - break; - } + if (colon != NULL) + *colon = 0; + + list_for_each_entry(tmp, &ksocknal_data.ksnd_nets, ksnn_list) { + char *ifnam2 = &tmp->ksnn_interface.ksni_name[0]; + char *colon2 = strchr(ifnam2, ':'); + + if (colon2 != NULL) + *colon2 = 0; - new_ipif += !found; - if (colon != NULL) - *colon = ':'; + found = strcmp(ifnam, ifnam2) == 0; + if (colon2 != NULL) + *colon2 = ':'; } + new_ipif += !found; + if (colon != NULL) + *colon = ':'; + return new_ipif; } @@ -2561,25 +2212,22 @@ ksocknal_startup(struct lnet_ni *ni) struct lnet_ioctl_config_lnd_cmn_tunables *net_tunables; struct ksock_interface *ksi = NULL; struct lnet_inetdev *ifaces = NULL; + struct sockaddr_in *sa; int i = 0; int rc; LASSERT (ni->ni_net->net_lnd == &the_ksocklnd); - if (ksocknal_data.ksnd_init == SOCKNAL_INIT_NOTHING) { rc = ksocknal_base_startup(); if (rc != 0) return rc; } - LIBCFS_ALLOC(net, sizeof(*net)); if (net == NULL) goto fail_0; - net->ksnn_incarnation = ktime_get_real_ns(); ni->ni_data = net; net_tunables = &ni->ni_net->net_tunables; - if (net_tunables->lct_peer_timeout == -1) net_tunables->lct_peer_timeout = *ksocknal_tunables.ksnd_peertimeout; @@ -2605,87 +2253,54 @@ ksocknal_startup(struct lnet_ni *ni) if (rc < 0) goto fail_1; - if (!ni->ni_interfaces[0]) { - ksi = &net->ksnn_interfaces[0]; - - /* Use the first discovered interface */ - net->ksnn_ninterfaces = 1; - ni->ni_dev_cpt = ifaces[0].li_cpt; - ksi->ksni_ipaddr = ifaces[0].li_ipaddr; - ksi->ksni_netmask = ifaces[0].li_netmask; - strlcpy(ksi->ksni_name, ifaces[0].li_name, - sizeof(ksi->ksni_name)); - } else { - /* Before Multi-Rail ksocklnd would manage - * multiple interfaces with its own tcp bonding. - * If we encounter an old configuration using - * this tcp bonding approach then we need to - * handle more than one ni_interfaces. - * - * In Multi-Rail configuration only ONE ni_interface - * should exist. Each IP alias should be mapped to - * each 'struct net_ni'. - */ - for (i = 0; i < LNET_INTERFACES_NUM; i++) { - int j; + ksi = &net->ksnn_interface; - if (!ni->ni_interfaces[i]) + /* Use the first discovered interface or look in the list */ + if (ni->ni_interface) { + for (i = 0; i < rc; i++) + if (strcmp(ifaces[i].li_name, ni->ni_interface) == 0) break; - for (j = 0; j < LNET_INTERFACES_NUM; j++) { - if (i != j && ni->ni_interfaces[j] && - strcmp(ni->ni_interfaces[i], - ni->ni_interfaces[j]) == 0) { - rc = -EEXIST; - CERROR("ksocklnd: found duplicate %s at %d and %d, rc = %d\n", - ni->ni_interfaces[i], i, j, rc); - goto fail_1; - } - } - - for (j = 0; j < rc; j++) { - if (strcmp(ifaces[j].li_name, - ni->ni_interfaces[i]) != 0) - continue; - - ksi = &net->ksnn_interfaces[j]; - ni->ni_dev_cpt = ifaces[j].li_cpt; - ksi->ksni_ipaddr = ifaces[j].li_ipaddr; - ksi->ksni_netmask = ifaces[j].li_netmask; - strlcpy(ksi->ksni_name, ifaces[j].li_name, - sizeof(ksi->ksni_name)); - net->ksnn_ninterfaces++; - break; - } - } - /* ni_interfaces don't map to all network interfaces */ - if (!ksi || net->ksnn_ninterfaces != i) { - CERROR("ksocklnd: requested %d but only %d interfaces found\n", - i, net->ksnn_ninterfaces); + /* ni_interfaces doesn't contain the interface we want */ + if (i == rc) { + CERROR("ksocklnd: failed to find interface %s\n", + ni->ni_interface); goto fail_1; } } + ni->ni_dev_cpt = ifaces[i].li_cpt; + sa = (void *)&ksi->ksni_addr; + memset(sa, 0, sizeof(*sa)); + sa->sin_family = AF_INET; + sa->sin_addr.s_addr = htonl(ifaces[i].li_ipaddr); + ksi->ksni_index = ksocknal_ip2index((struct sockaddr *)sa, ni); + ksi->ksni_netmask = ifaces[i].li_netmask; + strlcpy(ksi->ksni_name, ifaces[i].li_name, sizeof(ksi->ksni_name)); + /* call it before add it to ksocknal_data.ksnd_nets */ rc = ksocknal_net_start_threads(net, ni->ni_cpts, ni->ni_ncpts); if (rc != 0) goto fail_1; LASSERT(ksi); - ni->ni_nid = LNET_MKNID(LNET_NIDNET(ni->ni_nid), ksi->ksni_ipaddr); + LASSERT(ksi->ksni_addr.ss_family == AF_INET); + ni->ni_nid = LNET_MKNID( + LNET_NIDNET(ni->ni_nid), + ntohl(((struct sockaddr_in *) + &ksi->ksni_addr)->sin_addr.s_addr)); list_add(&net->ksnn_list, &ksocknal_data.ksnd_nets); + ksocknal_data.ksnd_nnets++; - ksocknal_data.ksnd_nnets++; - - return 0; + return 0; - fail_1: - LIBCFS_FREE(net, sizeof(*net)); - fail_0: - if (ksocknal_data.ksnd_nnets == 0) - ksocknal_base_shutdown(); +fail_1: + LIBCFS_FREE(net, sizeof(*net)); +fail_0: + if (ksocknal_data.ksnd_nnets == 0) + ksocknal_base_shutdown(); - return -ENETDOWN; + return -ENETDOWN; }