tipc: move link synch and failover to link aggregation level
[cascardo/linux.git] / net / tipc / link.c
index 2ccdb6f..d5f4005 100644 (file)
@@ -48,7 +48,7 @@
 /*
  * Error message prefixes
  */
-static const char *link_co_err = "Link changeover error, ";
+static const char *link_co_err = "Link tunneling error, ";
 static const char *link_rst_msg = "Resetting link ";
 static const char *link_unk_evt = "Unknown link event ";
 
@@ -139,24 +139,6 @@ static void tipc_link_build_bcast_sync_msg(struct tipc_link *l,
 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb);
 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb);
-static int tipc_link_failover_rcv(struct tipc_link *l, struct sk_buff **skb);
-
-/*
- *  Simple link routines
- */
-static unsigned int align(unsigned int i)
-{
-       return (i + 3) & ~3u;
-}
-
-static struct tipc_link *tipc_parallel_link(struct tipc_link *l)
-{
-       struct tipc_node *n = l->owner;
-
-       if (node_active_link(n, 0) != l)
-               return node_active_link(n, 0);
-       return node_active_link(n, 1);
-}
 
 /*
  *  Simple non-static link routines (i.e. referenced outside this file)
@@ -394,12 +376,10 @@ static int tipc_link_fsm_evt(struct tipc_link *l, int evt,
        /* Perform actions as decided by FSM */
        if (actions & LINK_RESET) {
                l->exec_mode = TIPC_LINK_BLOCKED;
-               rc |= TIPC_LINK_DOWN_EVT;
-       }
-       if (actions & LINK_ACTIVATE) {
-               l->exec_mode = TIPC_LINK_OPEN;
-               rc |= TIPC_LINK_UP_EVT;
+               rc = TIPC_LINK_DOWN_EVT;
        }
+       if (actions & LINK_ACTIVATE)
+               rc = TIPC_LINK_UP_EVT;
        if (actions & (SND_STATE | SND_PROBE))
                mtyp = STATE_MSG;
        if (actions & SND_RESET)
@@ -461,6 +441,9 @@ int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
 {
        int rc = 0;
 
+       if (l->exec_mode == TIPC_LINK_BLOCKED)
+               return rc;
+
        link_profile_stats(l);
        if (l->silent_intv_cnt)
                rc = tipc_link_fsm_evt(l, SILENCE_EVT, xmitq);
@@ -563,52 +546,42 @@ void tipc_link_purge_queues(struct tipc_link *l_ptr)
        tipc_link_reset_fragments(l_ptr);
 }
 
-void tipc_link_reset(struct tipc_link *l_ptr)
+void tipc_link_reset(struct tipc_link *l)
 {
-       u32 prev_state = l_ptr->state;
-       struct tipc_node *owner = l_ptr->owner;
-       struct tipc_link *pl = tipc_parallel_link(l_ptr);
+       struct tipc_node *owner = l->owner;
 
-       msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
+       l->state = TIPC_LINK_RESETTING;
 
        /* Link is down, accept any session */
-       l_ptr->peer_session = WILDCARD_SESSION;
+       l->peer_session = WILDCARD_SESSION;
 
-       /* Prepare for renewed mtu size negotiation */
-       l_ptr->mtu = l_ptr->advertised_mtu;
-
-       l_ptr->state = TIPC_LINK_RESETTING;
+       /* If peer is up, it only accepts an incremented session number */
+       msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
 
-       if ((prev_state == TIPC_LINK_RESETTING) ||
-           (prev_state == TIPC_LINK_ESTABLISHING))
-               return;
+       /* Prepare for renewed mtu size negotiation */
+       l->mtu = l->advertised_mtu;
 
-       if (tipc_node_is_up(l_ptr->owner) && (pl != l_ptr)) {
-               l_ptr->exec_mode = TIPC_LINK_BLOCKED;
-               l_ptr->failover_checkpt = l_ptr->rcv_nxt;
-               pl->failover_pkts = FIRST_FAILOVER;
-               pl->failover_checkpt = l_ptr->rcv_nxt;
-               pl->failover_skb = l_ptr->reasm_buf;
-       } else {
-               kfree_skb(l_ptr->reasm_buf);
-       }
        /* Clean up all queues, except inputq: */
-       __skb_queue_purge(&l_ptr->transmq);
-       __skb_queue_purge(&l_ptr->deferdq);
+       __skb_queue_purge(&l->transmq);
+       __skb_queue_purge(&l->deferdq);
        if (!owner->inputq)
-               owner->inputq = l_ptr->inputq;
-       skb_queue_splice_init(&l_ptr->wakeupq, owner->inputq);
+               owner->inputq = l->inputq;
+       skb_queue_splice_init(&l->wakeupq, owner->inputq);
        if (!skb_queue_empty(owner->inputq))
                owner->action_flags |= TIPC_MSG_EVT;
-       tipc_link_purge_backlog(l_ptr);
-       l_ptr->reasm_buf = NULL;
-       l_ptr->rcv_unacked = 0;
-       l_ptr->snd_nxt = 1;
-       l_ptr->rcv_nxt = 1;
-       l_ptr->silent_intv_cnt = 0;
-       l_ptr->stats.recv_info = 0;
-       l_ptr->stale_count = 0;
-       link_reset_statistics(l_ptr);
+
+       tipc_link_purge_backlog(l);
+       kfree_skb(l->reasm_buf);
+       kfree_skb(l->failover_reasm_skb);
+       l->reasm_buf = NULL;
+       l->failover_reasm_skb = NULL;
+       l->rcv_unacked = 0;
+       l->snd_nxt = 1;
+       l->rcv_nxt = 1;
+       l->silent_intv_cnt = 0;
+       l->stats.recv_info = 0;
+       l->stale_count = 0;
+       link_reset_statistics(l);
 }
 
 /**
@@ -751,20 +724,6 @@ int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
        return 0;
 }
 
-static void skb2list(struct sk_buff *skb, struct sk_buff_head *list)
-{
-       skb_queue_head_init(list);
-       __skb_queue_tail(list, skb);
-}
-
-static int __tipc_link_xmit_skb(struct tipc_link *link, struct sk_buff *skb)
-{
-       struct sk_buff_head head;
-
-       skb2list(skb, &head);
-       return __tipc_link_xmit(link->owner->net, link, &head);
-}
-
 /*
  * tipc_link_sync_rcv - synchronize broadcast link endpoints.
  * Receive the sequence number where we should start receiving and
@@ -955,32 +914,6 @@ static int tipc_link_retransm(struct tipc_link *l, int retransm,
        return 0;
 }
 
-/* link_synch(): check if all packets arrived before the synch
- *               point have been consumed
- * Returns true if the parallel links are synched, otherwise false
- */
-static bool link_synch(struct tipc_link *l)
-{
-       unsigned int post_synch;
-       struct tipc_link *pl;
-
-       pl  = tipc_parallel_link(l);
-       if (pl == l)
-               goto synched;
-
-       /* Was last pre-synch packet added to input queue ? */
-       if (less_eq(pl->rcv_nxt, l->synch_point))
-               return false;
-
-       /* Is it still in the input queue ? */
-       post_synch = mod(pl->rcv_nxt - l->synch_point) - 1;
-       if (skb_queue_len(pl->inputq) > post_synch)
-               return false;
-synched:
-       l->exec_mode = TIPC_LINK_OPEN;
-       return true;
-}
-
 /* tipc_data_input - deliver data and name distr msgs to upper layer
  *
  * Consumes buffer if message is of right type
@@ -1025,54 +958,59 @@ static bool tipc_data_input(struct tipc_link *link, struct sk_buff *skb)
 /* tipc_link_input - process packet that has passed link protocol check
  *
  * Consumes buffer
- * Node lock must be held
  */
-static int tipc_link_input(struct tipc_link *link, struct sk_buff *skb)
+static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb)
 {
-       struct tipc_node *node = link->owner;
-       struct tipc_msg *msg = buf_msg(skb);
+       struct tipc_node *node = l->owner;
+       struct tipc_msg *hdr = buf_msg(skb);
+       struct sk_buff **reasm_skb = &l->reasm_buf;
        struct sk_buff *iskb;
-       int pos = 0;
+       int usr = msg_user(hdr);
        int rc = 0;
+       int pos = 0;
+       int ipos = 0;
 
-       switch (msg_user(msg)) {
-       case TUNNEL_PROTOCOL:
-               if (msg_dup(msg)) {
-                       link->exec_mode = TIPC_LINK_TUNNEL;
-                       link->synch_point = msg_seqno(msg_get_wrapped(msg));
-                       kfree_skb(skb);
-                       break;
+       if (unlikely(usr == TUNNEL_PROTOCOL)) {
+               if (msg_type(hdr) == SYNCH_MSG) {
+                       __skb_queue_purge(&l->deferdq);
+                       goto drop;
                }
-               rc |= tipc_link_failover_rcv(link, &skb);
-               if (!skb)
-                       break;
-               if (msg_user(buf_msg(skb)) != MSG_BUNDLER) {
-                       tipc_data_input(link, skb);
-                       break;
-               }
-       case MSG_BUNDLER:
-               link->stats.recv_bundles++;
-               link->stats.recv_bundled += msg_msgcnt(msg);
+               if (!tipc_msg_extract(skb, &iskb, &ipos))
+                       return rc;
+               kfree_skb(skb);
+               skb = iskb;
+               hdr = buf_msg(skb);
+               if (less(msg_seqno(hdr), l->drop_point))
+                       goto drop;
+               if (tipc_data_input(l, skb))
+                       return rc;
+               usr = msg_user(hdr);
+               reasm_skb = &l->failover_reasm_skb;
+       }
 
+       if (usr == MSG_BUNDLER) {
+               l->stats.recv_bundles++;
+               l->stats.recv_bundled += msg_msgcnt(hdr);
                while (tipc_msg_extract(skb, &iskb, &pos))
-                       tipc_data_input(link, iskb);
-               break;
-       case MSG_FRAGMENTER:
-               link->stats.recv_fragments++;
-               if (tipc_buf_append(&link->reasm_buf, &skb)) {
-                       link->stats.recv_fragmented++;
-                       tipc_data_input(link, skb);
-               } else if (!link->reasm_buf) {
-                       link->exec_mode = TIPC_LINK_BLOCKED;
-                       rc |= TIPC_LINK_DOWN_EVT;
+                       tipc_data_input(l, iskb);
+               return rc;
+       } else if (usr == MSG_FRAGMENTER) {
+               l->stats.recv_fragments++;
+               if (tipc_buf_append(reasm_skb, &skb)) {
+                       l->stats.recv_fragmented++;
+                       tipc_data_input(l, skb);
+               } else if (!*reasm_skb) {
+                       l->exec_mode = TIPC_LINK_BLOCKED;
+                       l->state = TIPC_LINK_RESETTING;
+                       rc = TIPC_LINK_DOWN_EVT;
                }
-               break;
-       case BCAST_PROTOCOL:
+               return rc;
+       } else if (usr == BCAST_PROTOCOL) {
                tipc_link_sync_rcv(node, skb);
-               break;
-       default:
-               break;
-       };
+               return rc;
+       }
+drop:
+       kfree_skb(skb);
        return rc;
 }
 
@@ -1100,7 +1038,6 @@ int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
                  struct sk_buff_head *xmitq)
 {
        struct sk_buff_head *arrvq = &l->deferdq;
-       struct sk_buff *tmp;
        struct tipc_msg *hdr;
        u16 seqno, rcv_nxt;
        int rc = 0;
@@ -1112,18 +1049,18 @@ int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
                return rc;
        }
 
-       skb_queue_walk_safe(arrvq, skb, tmp) {
+       while ((skb = skb_peek(arrvq))) {
                hdr = buf_msg(skb);
 
                /* Verify and update link state */
                if (unlikely(msg_user(hdr) == LINK_PROTOCOL)) {
                        __skb_dequeue(arrvq);
-                       rc |= tipc_link_proto_rcv(l, skb, xmitq);
+                       rc = tipc_link_proto_rcv(l, skb, xmitq);
                        continue;
                }
 
                if (unlikely(!link_working(l))) {
-                       rc |= tipc_link_fsm_evt(l, TRAFFIC_EVT, xmitq);
+                       rc = tipc_link_fsm_evt(l, TRAFFIC_EVT, xmitq);
                        if (!link_working(l)) {
                                kfree_skb(__skb_dequeue(arrvq));
                                return rc;
@@ -1156,18 +1093,11 @@ int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
                        return rc;
                }
 
-               /* Synchronize with parallel link if applicable */
-               if (unlikely(l->exec_mode == TIPC_LINK_TUNNEL))
-                       if (!msg_dup(hdr) && !link_synch(l)) {
-                               kfree_skb(skb);
-                               return rc;
-                       }
-
                /* Packet can be delivered */
                l->rcv_nxt++;
                l->stats.recv_info++;
                if (unlikely(!tipc_data_input(l, skb)))
-                       rc |= tipc_link_input(l, skb);
+                       rc = tipc_link_input(l, skb);
 
                /* Ack at regular intervals */
                if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN)) {
@@ -1288,7 +1218,7 @@ static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
        } else {
                /* RESET_MSG or ACTIVATE_MSG */
                msg_set_max_pkt(hdr, l->advertised_mtu);
-               msg_set_ack(hdr, l->failover_checkpt - 1);
+               msg_set_ack(hdr, l->rcv_nxt - 1);
                msg_set_next_sent(hdr, 1);
        }
        skb = tipc_buf_acquire(msg_size(hdr));
@@ -1296,223 +1226,75 @@ static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
                return;
        skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
        skb->priority = TC_PRIO_CONTROL;
-       __skb_queue_head(xmitq, skb);
-}
-
-/* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
- * a different bearer. Owner node is locked.
- */
-static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
-                                 struct tipc_msg *tunnel_hdr,
-                                 struct tipc_msg *msg,
-                                 u32 selector)
-{
-       struct tipc_link *tunnel;
-       struct sk_buff *skb;
-       u32 length = msg_size(msg);
-
-       tunnel = node_active_link(l_ptr->owner, selector & 1);
-       if (!tipc_link_is_up(tunnel)) {
-               pr_warn("%stunnel link no longer available\n", link_co_err);
-               return;
-       }
-       msg_set_size(tunnel_hdr, length + INT_H_SIZE);
-       skb = tipc_buf_acquire(length + INT_H_SIZE);
-       if (!skb) {
-               pr_warn("%sunable to send tunnel msg\n", link_co_err);
-               return;
-       }
-       skb_copy_to_linear_data(skb, tunnel_hdr, INT_H_SIZE);
-       skb_copy_to_linear_data_offset(skb, INT_H_SIZE, msg, length);
-       __tipc_link_xmit_skb(tunnel, skb);
+       __skb_queue_tail(xmitq, skb);
 }
 
-
-/* tipc_link_failover_send_queue(): A link has gone down, but a second
- * link is still active. We can do failover. Tunnel the failing link's
- * whole send queue via the remaining link. This way, we don't lose
- * any packets, and sequence order is preserved for subsequent traffic
- * sent over the remaining link. Owner node is locked.
+/* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
+ * with contents of the link's tranmsit and backlog queues.
  */
-void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
+void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
+                          int mtyp, struct sk_buff_head *xmitq)
 {
-       int msgcount;
-       struct tipc_link *tunnel = node_active_link(l_ptr->owner, 0);
-       struct tipc_msg tunnel_hdr;
-       struct sk_buff *skb;
-       int split_bundles;
+       struct sk_buff *skb, *tnlskb;
+       struct tipc_msg *hdr, tnlhdr;
+       struct sk_buff_head *queue = &l->transmq;
+       struct sk_buff_head tmpxq, tnlq;
+       u16 pktlen, pktcnt, seqno = l->snd_nxt;
 
-       if (!tunnel)
+       if (!tnl)
                return;
 
-       tipc_msg_init(link_own_addr(l_ptr), &tunnel_hdr, TUNNEL_PROTOCOL,
-                     FAILOVER_MSG, INT_H_SIZE, l_ptr->addr);
+       skb_queue_head_init(&tnlq);
+       skb_queue_head_init(&tmpxq);
 
-       skb_queue_walk(&l_ptr->backlogq, skb) {
-               msg_set_seqno(buf_msg(skb), l_ptr->snd_nxt);
-               l_ptr->snd_nxt = mod(l_ptr->snd_nxt + 1);
-       }
-       skb_queue_splice_tail_init(&l_ptr->backlogq, &l_ptr->transmq);
-       tipc_link_purge_backlog(l_ptr);
-       msgcount = skb_queue_len(&l_ptr->transmq);
-       msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
-       msg_set_msgcnt(&tunnel_hdr, msgcount);
-
-       if (skb_queue_empty(&l_ptr->transmq)) {
-               skb = tipc_buf_acquire(INT_H_SIZE);
-               if (skb) {
-                       skb_copy_to_linear_data(skb, &tunnel_hdr, INT_H_SIZE);
-                       msg_set_size(&tunnel_hdr, INT_H_SIZE);
-                       __tipc_link_xmit_skb(tunnel, skb);
-               } else {
-                       pr_warn("%sunable to send changeover msg\n",
-                               link_co_err);
-               }
+       /* At least one packet required for safe algorithm => add dummy */
+       skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
+                             BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
+                             0, 0, TIPC_ERR_NO_PORT);
+       if (!skb) {
+               pr_warn("%sunable to create tunnel packet\n", link_co_err);
                return;
        }
-
-       split_bundles = (node_active_link(l_ptr->owner, 0) !=
-                        node_active_link(l_ptr->owner, 0));
-
-       skb_queue_walk(&l_ptr->transmq, skb) {
-               struct tipc_msg *msg = buf_msg(skb);
-
-               if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
-                       struct tipc_msg *m = msg_get_wrapped(msg);
-                       unchar *pos = (unchar *)m;
-
-                       msgcount = msg_msgcnt(msg);
-                       while (msgcount--) {
-                               msg_set_seqno(m, msg_seqno(msg));
-                               tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
-                                                     msg_link_selector(m));
-                               pos += align(msg_size(m));
-                               m = (struct tipc_msg *)pos;
-                       }
-               } else {
-                       tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
-                                             msg_link_selector(msg));
-               }
-       }
-}
-
-/* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
- * duplicate of the first link's send queue via the new link. This way, we
- * are guaranteed that currently queued packets from a socket are delivered
- * before future traffic from the same socket, even if this is using the
- * new link. The last arriving copy of each duplicate packet is dropped at
- * the receiving end by the regular protocol check, so packet cardinality
- * and sequence order is preserved per sender/receiver socket pair.
- * Owner node is locked.
- */
-void tipc_link_dup_queue_xmit(struct tipc_link *link,
-                             struct tipc_link *tnl)
-{
-       struct sk_buff *skb;
-       struct tipc_msg tnl_hdr;
-       struct sk_buff_head *queue = &link->transmq;
-       int mcnt;
-       u16 seqno;
-
-       tipc_msg_init(link_own_addr(link), &tnl_hdr, TUNNEL_PROTOCOL,
-                     SYNCH_MSG, INT_H_SIZE, link->addr);
-       mcnt = skb_queue_len(&link->transmq) + skb_queue_len(&link->backlogq);
-       msg_set_msgcnt(&tnl_hdr, mcnt);
-       msg_set_bearer_id(&tnl_hdr, link->peer_bearer_id);
-
-tunnel_queue:
+       skb_queue_tail(&tnlq, skb);
+       tipc_link_xmit(l, &tnlq, &tmpxq);
+       __skb_queue_purge(&tmpxq);
+
+       /* Initialize reusable tunnel packet header */
+       tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
+                     mtyp, INT_H_SIZE, l->addr);
+       pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
+       msg_set_msgcnt(&tnlhdr, pktcnt);
+       msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
+tnl:
+       /* Wrap each packet into a tunnel packet */
        skb_queue_walk(queue, skb) {
-               struct sk_buff *outskb;
-               struct tipc_msg *msg = buf_msg(skb);
-               u32 len = msg_size(msg);
-
-               msg_set_ack(msg, mod(link->rcv_nxt - 1));
-               msg_set_bcast_ack(msg, link->owner->bclink.last_in);
-               msg_set_size(&tnl_hdr, len + INT_H_SIZE);
-               outskb = tipc_buf_acquire(len + INT_H_SIZE);
-               if (outskb == NULL) {
-                       pr_warn("%sunable to send duplicate msg\n",
-                               link_co_err);
+               hdr = buf_msg(skb);
+               if (queue == &l->backlogq)
+                       msg_set_seqno(hdr, seqno++);
+               pktlen = msg_size(hdr);
+               msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
+               tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
+               if (!tnlskb) {
+                       pr_warn("%sunable to send packet\n", link_co_err);
                        return;
                }
-               skb_copy_to_linear_data(outskb, &tnl_hdr, INT_H_SIZE);
-               skb_copy_to_linear_data_offset(outskb, INT_H_SIZE,
-                                              skb->data, len);
-               __tipc_link_xmit_skb(tnl, outskb);
-               if (!tipc_link_is_up(link))
-                       return;
+               skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
+               skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
+               __skb_queue_tail(&tnlq, tnlskb);
        }
-       if (queue == &link->backlogq)
-               return;
-       seqno = link->snd_nxt;
-       skb_queue_walk(&link->backlogq, skb) {
-               msg_set_seqno(buf_msg(skb), seqno);
-               seqno = mod(seqno + 1);
-       }
-       queue = &link->backlogq;
-       goto tunnel_queue;
-}
-
-/*  tipc_link_failover_rcv(): Receive a tunnelled FAILOVER_MSG packet
- *  Owner node is locked.
- */
-static int tipc_link_failover_rcv(struct tipc_link *link,
-                                 struct sk_buff **skb)
-{
-       struct tipc_msg *msg = buf_msg(*skb);
-       struct sk_buff *iskb = NULL;
-       struct tipc_link *pl = NULL;
-       int bearer_id = msg_bearer_id(msg);
-       int pos = 0;
-       int rc = 0;
-
-       if (msg_type(msg) != FAILOVER_MSG) {
-               pr_warn("%sunknown tunnel pkt received\n", link_co_err);
-               goto exit;
+       if (queue != &l->backlogq) {
+               queue = &l->backlogq;
+               goto tnl;
        }
-       if (bearer_id >= MAX_BEARERS)
-               goto exit;
-
-       if (bearer_id == link->bearer_id)
-               goto exit;
-
-       pl = link->owner->links[bearer_id].link;
-
-       if (link->failover_pkts == FIRST_FAILOVER)
-               link->failover_pkts = msg_msgcnt(msg);
-
-       /* Should we expect an inner packet? */
-       if (!link->failover_pkts)
-               goto exit;
 
-       if (!tipc_msg_extract(*skb, &iskb, &pos)) {
-               pr_warn("%sno inner failover pkt\n", link_co_err);
-               *skb = NULL;
-               goto exit;
-       }
-       link->failover_pkts--;
-       *skb = NULL;
+       tipc_link_xmit(tnl, &tnlq, xmitq);
 
-       /* Was this packet already delivered? */
-       if (less(buf_seqno(iskb), link->failover_checkpt)) {
-               kfree_skb(iskb);
-               iskb = NULL;
-               goto exit;
-       }
-       if (msg_user(buf_msg(iskb)) == MSG_FRAGMENTER) {
-               link->stats.recv_fragments++;
-               if (!tipc_buf_append(&link->failover_skb, &iskb) &&
-                   !link->failover_skb) {
-                       link->exec_mode = TIPC_LINK_BLOCKED;
-                       rc |= TIPC_LINK_DOWN_EVT;
-               }
+       if (mtyp == FAILOVER_MSG) {
+               tnl->drop_point = l->rcv_nxt;
+               tnl->failover_reasm_skb = l->reasm_buf;
+               l->reasm_buf = NULL;
+               l->exec_mode = TIPC_LINK_BLOCKED;
        }
-exit:
-       if (!link->failover_pkts && pl)
-               pl->exec_mode = TIPC_LINK_OPEN;
-       kfree_skb(*skb);
-       *skb = iskb;
-       return rc;
 }
 
 /* tipc_link_proto_rcv(): receive link level protocol message :
@@ -1593,7 +1375,7 @@ static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
 
                /* If NACK, retransmit will now start at right position */
                if (nacked_gap) {
-                       rc |= tipc_link_retransm(l, nacked_gap, xmitq);
+                       rc = tipc_link_retransm(l, nacked_gap, xmitq);
                        l->stats.recv_nacks++;
                }
                tipc_link_advance_backlog(l, xmitq);