2 * net/tipc/node.c: TIPC node management routines
4 * Copyright (c) 2000-2006, 2012-2015, Ericsson AB
5 * Copyright (c) 2005-2006, 2010-2014, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
40 #include "name_distr.h"
45 /* Node FSM states and events:
48 SELF_DOWN_PEER_DOWN
= 0xdd,
49 SELF_UP_PEER_UP
= 0xaa,
50 SELF_DOWN_PEER_LEAVING
= 0xd1,
51 SELF_UP_PEER_COMING
= 0xac,
52 SELF_COMING_PEER_UP
= 0xca,
53 SELF_LEAVING_PEER_DOWN
= 0x1d,
54 NODE_FAILINGOVER
= 0xf0,
59 SELF_ESTABL_CONTACT_EVT
= 0xece,
60 SELF_LOST_CONTACT_EVT
= 0x1ce,
61 PEER_ESTABL_CONTACT_EVT
= 0x9ece,
62 PEER_LOST_CONTACT_EVT
= 0x91ce,
63 NODE_FAILOVER_BEGIN_EVT
= 0xfbe,
64 NODE_FAILOVER_END_EVT
= 0xfee,
65 NODE_SYNCH_BEGIN_EVT
= 0xcbe,
66 NODE_SYNCH_END_EVT
= 0xcee
69 static void __tipc_node_link_down(struct tipc_node
*n
, int *bearer_id
,
70 struct sk_buff_head
*xmitq
,
71 struct tipc_media_addr
**maddr
);
72 static void tipc_node_link_down(struct tipc_node
*n
, int bearer_id
,
74 static void node_lost_contact(struct tipc_node
*n
, struct sk_buff_head
*inputq
);
75 static void tipc_node_delete(struct tipc_node
*node
);
76 static void tipc_node_timeout(unsigned long data
);
77 static void tipc_node_fsm_evt(struct tipc_node
*n
, int evt
);
79 struct tipc_sock_conn
{
83 struct list_head list
;
86 static const struct nla_policy tipc_nl_node_policy
[TIPC_NLA_NODE_MAX
+ 1] = {
87 [TIPC_NLA_NODE_UNSPEC
] = { .type
= NLA_UNSPEC
},
88 [TIPC_NLA_NODE_ADDR
] = { .type
= NLA_U32
},
89 [TIPC_NLA_NODE_UP
] = { .type
= NLA_FLAG
}
93 * A trivial power-of-two bitmask technique is used for speed, since this
94 * operation is done for every incoming TIPC packet. The number of hash table
95 * entries has been chosen so that no hash chain exceeds 8 nodes and will
96 * usually be much smaller (typically only a single node).
98 static unsigned int tipc_hashfn(u32 addr
)
100 return addr
& (NODE_HTABLE_SIZE
- 1);
103 static void tipc_node_kref_release(struct kref
*kref
)
105 struct tipc_node
*node
= container_of(kref
, struct tipc_node
, kref
);
107 tipc_node_delete(node
);
110 void tipc_node_put(struct tipc_node
*node
)
112 kref_put(&node
->kref
, tipc_node_kref_release
);
115 static void tipc_node_get(struct tipc_node
*node
)
117 kref_get(&node
->kref
);
121 * tipc_node_find - locate specified node object, if it exists
123 struct tipc_node
*tipc_node_find(struct net
*net
, u32 addr
)
125 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
126 struct tipc_node
*node
;
128 if (unlikely(!in_own_cluster_exact(net
, addr
)))
132 hlist_for_each_entry_rcu(node
, &tn
->node_htable
[tipc_hashfn(addr
)],
134 if (node
->addr
== addr
) {
144 struct tipc_node
*tipc_node_create(struct net
*net
, u32 addr
, u16 capabilities
)
146 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
147 struct tipc_node
*n_ptr
, *temp_node
;
149 spin_lock_bh(&tn
->node_list_lock
);
150 n_ptr
= tipc_node_find(net
, addr
);
153 n_ptr
= kzalloc(sizeof(*n_ptr
), GFP_ATOMIC
);
155 pr_warn("Node creation failed, no memory\n");
160 n_ptr
->capabilities
= capabilities
;
161 kref_init(&n_ptr
->kref
);
162 spin_lock_init(&n_ptr
->lock
);
163 INIT_HLIST_NODE(&n_ptr
->hash
);
164 INIT_LIST_HEAD(&n_ptr
->list
);
165 INIT_LIST_HEAD(&n_ptr
->publ_list
);
166 INIT_LIST_HEAD(&n_ptr
->conn_sks
);
167 skb_queue_head_init(&n_ptr
->bc_entry
.namedq
);
168 skb_queue_head_init(&n_ptr
->bc_entry
.inputq1
);
169 __skb_queue_head_init(&n_ptr
->bc_entry
.arrvq
);
170 skb_queue_head_init(&n_ptr
->bc_entry
.inputq2
);
171 hlist_add_head_rcu(&n_ptr
->hash
, &tn
->node_htable
[tipc_hashfn(addr
)]);
172 list_for_each_entry_rcu(temp_node
, &tn
->node_list
, list
) {
173 if (n_ptr
->addr
< temp_node
->addr
)
176 list_add_tail_rcu(&n_ptr
->list
, &temp_node
->list
);
177 n_ptr
->state
= SELF_DOWN_PEER_LEAVING
;
178 n_ptr
->signature
= INVALID_NODE_SIG
;
179 n_ptr
->active_links
[0] = INVALID_BEARER_ID
;
180 n_ptr
->active_links
[1] = INVALID_BEARER_ID
;
181 if (!tipc_link_bc_create(net
, tipc_own_addr(net
), n_ptr
->addr
,
182 U16_MAX
, tipc_bc_sndlink(net
)->window
,
184 &n_ptr
->bc_entry
.inputq1
,
185 &n_ptr
->bc_entry
.namedq
,
186 tipc_bc_sndlink(net
),
187 &n_ptr
->bc_entry
.link
)) {
188 pr_warn("Broadcast rcv link creation failed, no memory\n");
193 tipc_node_get(n_ptr
);
194 setup_timer(&n_ptr
->timer
, tipc_node_timeout
, (unsigned long)n_ptr
);
195 n_ptr
->keepalive_intv
= U32_MAX
;
197 spin_unlock_bh(&tn
->node_list_lock
);
201 static void tipc_node_calculate_timer(struct tipc_node
*n
, struct tipc_link
*l
)
203 unsigned long tol
= l
->tolerance
;
204 unsigned long intv
= ((tol
/ 4) > 500) ? 500 : tol
/ 4;
205 unsigned long keepalive_intv
= msecs_to_jiffies(intv
);
207 /* Link with lowest tolerance determines timer interval */
208 if (keepalive_intv
< n
->keepalive_intv
)
209 n
->keepalive_intv
= keepalive_intv
;
211 /* Ensure link's abort limit corresponds to current interval */
212 l
->abort_limit
= l
->tolerance
/ jiffies_to_msecs(n
->keepalive_intv
);
215 static void tipc_node_delete(struct tipc_node
*node
)
217 list_del_rcu(&node
->list
);
218 hlist_del_rcu(&node
->hash
);
219 kfree(node
->bc_entry
.link
);
220 kfree_rcu(node
, rcu
);
223 void tipc_node_stop(struct net
*net
)
225 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
226 struct tipc_node
*node
, *t_node
;
228 spin_lock_bh(&tn
->node_list_lock
);
229 list_for_each_entry_safe(node
, t_node
, &tn
->node_list
, list
) {
230 if (del_timer(&node
->timer
))
234 spin_unlock_bh(&tn
->node_list_lock
);
237 void tipc_node_subscribe(struct net
*net
, struct list_head
*subscr
, u32 addr
)
241 if (in_own_node(net
, addr
))
244 n
= tipc_node_find(net
, addr
);
246 pr_warn("Node subscribe rejected, unknown node 0x%x\n", addr
);
250 list_add_tail(subscr
, &n
->publ_list
);
255 void tipc_node_unsubscribe(struct net
*net
, struct list_head
*subscr
, u32 addr
)
259 if (in_own_node(net
, addr
))
262 n
= tipc_node_find(net
, addr
);
264 pr_warn("Node unsubscribe rejected, unknown node 0x%x\n", addr
);
268 list_del_init(subscr
);
273 int tipc_node_add_conn(struct net
*net
, u32 dnode
, u32 port
, u32 peer_port
)
275 struct tipc_node
*node
;
276 struct tipc_sock_conn
*conn
;
279 if (in_own_node(net
, dnode
))
282 node
= tipc_node_find(net
, dnode
);
284 pr_warn("Connecting sock to node 0x%x failed\n", dnode
);
285 return -EHOSTUNREACH
;
287 conn
= kmalloc(sizeof(*conn
), GFP_ATOMIC
);
292 conn
->peer_node
= dnode
;
294 conn
->peer_port
= peer_port
;
296 tipc_node_lock(node
);
297 list_add_tail(&conn
->list
, &node
->conn_sks
);
298 tipc_node_unlock(node
);
304 void tipc_node_remove_conn(struct net
*net
, u32 dnode
, u32 port
)
306 struct tipc_node
*node
;
307 struct tipc_sock_conn
*conn
, *safe
;
309 if (in_own_node(net
, dnode
))
312 node
= tipc_node_find(net
, dnode
);
316 tipc_node_lock(node
);
317 list_for_each_entry_safe(conn
, safe
, &node
->conn_sks
, list
) {
318 if (port
!= conn
->port
)
320 list_del(&conn
->list
);
323 tipc_node_unlock(node
);
327 /* tipc_node_timeout - handle expiration of node timer
329 static void tipc_node_timeout(unsigned long data
)
331 struct tipc_node
*n
= (struct tipc_node
*)data
;
332 struct tipc_link_entry
*le
;
333 struct sk_buff_head xmitq
;
337 __skb_queue_head_init(&xmitq
);
339 for (bearer_id
= 0; bearer_id
< MAX_BEARERS
; bearer_id
++) {
341 le
= &n
->links
[bearer_id
];
343 /* Link tolerance may change asynchronously: */
344 tipc_node_calculate_timer(n
, le
->link
);
345 rc
= tipc_link_timeout(le
->link
, &xmitq
);
348 tipc_bearer_xmit(n
->net
, bearer_id
, &xmitq
, &le
->maddr
);
349 if (rc
& TIPC_LINK_DOWN_EVT
)
350 tipc_node_link_down(n
, bearer_id
, false);
352 if (!mod_timer(&n
->timer
, jiffies
+ n
->keepalive_intv
))
358 * __tipc_node_link_up - handle addition of link
359 * Node lock must be held by caller
360 * Link becomes active (alone or shared) or standby, depending on its priority.
362 static void __tipc_node_link_up(struct tipc_node
*n
, int bearer_id
,
363 struct sk_buff_head
*xmitq
)
365 int *slot0
= &n
->active_links
[0];
366 int *slot1
= &n
->active_links
[1];
367 struct tipc_link
*ol
= node_active_link(n
, 0);
368 struct tipc_link
*nl
= n
->links
[bearer_id
].link
;
373 tipc_link_fsm_evt(nl
, LINK_ESTABLISH_EVT
);
374 if (!tipc_link_is_up(nl
))
378 n
->action_flags
|= TIPC_NOTIFY_LINK_UP
;
379 n
->link_id
= nl
->peer_bearer_id
<< 16 | bearer_id
;
381 /* Leave room for tunnel header when returning 'mtu' to users: */
382 n
->links
[bearer_id
].mtu
= nl
->mtu
- INT_H_SIZE
;
384 tipc_bearer_add_dest(n
->net
, bearer_id
, n
->addr
);
385 tipc_bcast_inc_bearer_dst_cnt(n
->net
, bearer_id
);
387 pr_debug("Established link <%s> on network plane %c\n",
388 nl
->name
, nl
->net_plane
);
390 /* First link? => give it both slots */
394 tipc_node_fsm_evt(n
, SELF_ESTABL_CONTACT_EVT
);
395 n
->action_flags
|= TIPC_NOTIFY_NODE_UP
;
396 tipc_bcast_add_peer(n
->net
, nl
, xmitq
);
400 /* Second link => redistribute slots */
401 if (nl
->priority
> ol
->priority
) {
402 pr_debug("Old link <%s> becomes standby\n", ol
->name
);
405 tipc_link_set_active(nl
, true);
406 tipc_link_set_active(ol
, false);
407 } else if (nl
->priority
== ol
->priority
) {
408 tipc_link_set_active(nl
, true);
411 pr_debug("New link <%s> is standby\n", nl
->name
);
414 /* Prepare synchronization with first link */
415 tipc_link_tnl_prepare(ol
, nl
, SYNCH_MSG
, xmitq
);
419 * tipc_node_link_up - handle addition of link
421 * Link becomes active (alone or shared) or standby, depending on its priority.
423 static void tipc_node_link_up(struct tipc_node
*n
, int bearer_id
,
424 struct sk_buff_head
*xmitq
)
427 __tipc_node_link_up(n
, bearer_id
, xmitq
);
432 * __tipc_node_link_down - handle loss of link
434 static void __tipc_node_link_down(struct tipc_node
*n
, int *bearer_id
,
435 struct sk_buff_head
*xmitq
,
436 struct tipc_media_addr
**maddr
)
438 struct tipc_link_entry
*le
= &n
->links
[*bearer_id
];
439 int *slot0
= &n
->active_links
[0];
440 int *slot1
= &n
->active_links
[1];
442 struct tipc_link
*l
, *_l
, *tnl
;
444 l
= n
->links
[*bearer_id
].link
;
445 if (!l
|| tipc_link_is_reset(l
))
449 n
->action_flags
|= TIPC_NOTIFY_LINK_DOWN
;
450 n
->link_id
= l
->peer_bearer_id
<< 16 | *bearer_id
;
452 tipc_bearer_remove_dest(n
->net
, *bearer_id
, n
->addr
);
454 pr_debug("Lost link <%s> on network plane %c\n",
455 l
->name
, l
->net_plane
);
457 /* Select new active link if any available */
458 *slot0
= INVALID_BEARER_ID
;
459 *slot1
= INVALID_BEARER_ID
;
460 for (i
= 0; i
< MAX_BEARERS
; i
++) {
461 _l
= n
->links
[i
].link
;
462 if (!_l
|| !tipc_link_is_up(_l
))
466 if (_l
->priority
< highest
)
468 if (_l
->priority
> highest
) {
469 highest
= _l
->priority
;
477 if (!tipc_node_is_up(n
)) {
478 if (tipc_link_peer_is_down(l
))
479 tipc_node_fsm_evt(n
, PEER_LOST_CONTACT_EVT
);
480 tipc_node_fsm_evt(n
, SELF_LOST_CONTACT_EVT
);
481 tipc_link_fsm_evt(l
, LINK_RESET_EVT
);
483 tipc_link_build_reset_msg(l
, xmitq
);
484 *maddr
= &n
->links
[*bearer_id
].maddr
;
485 node_lost_contact(n
, &le
->inputq
);
486 tipc_bcast_dec_bearer_dst_cnt(n
->net
, *bearer_id
);
489 tipc_bcast_dec_bearer_dst_cnt(n
->net
, *bearer_id
);
491 /* There is still a working link => initiate failover */
492 tnl
= node_active_link(n
, 0);
493 tipc_link_fsm_evt(tnl
, LINK_SYNCH_END_EVT
);
494 tipc_node_fsm_evt(n
, NODE_SYNCH_END_EVT
);
495 n
->sync_point
= tnl
->rcv_nxt
+ (U16_MAX
/ 2 - 1);
496 tipc_link_tnl_prepare(l
, tnl
, FAILOVER_MSG
, xmitq
);
498 tipc_link_fsm_evt(l
, LINK_RESET_EVT
);
499 tipc_link_fsm_evt(l
, LINK_FAILOVER_BEGIN_EVT
);
500 tipc_node_fsm_evt(n
, NODE_FAILOVER_BEGIN_EVT
);
501 *maddr
= &n
->links
[tnl
->bearer_id
].maddr
;
502 *bearer_id
= tnl
->bearer_id
;
505 static void tipc_node_link_down(struct tipc_node
*n
, int bearer_id
, bool delete)
507 struct tipc_link_entry
*le
= &n
->links
[bearer_id
];
508 struct tipc_link
*l
= le
->link
;
509 struct tipc_media_addr
*maddr
;
510 struct sk_buff_head xmitq
;
515 __skb_queue_head_init(&xmitq
);
518 if (!tipc_link_is_establishing(l
)) {
519 __tipc_node_link_down(n
, &bearer_id
, &xmitq
, &maddr
);
526 /* Defuse pending tipc_node_link_up() */
527 tipc_link_fsm_evt(l
, LINK_RESET_EVT
);
530 tipc_bearer_xmit(n
->net
, bearer_id
, &xmitq
, maddr
);
531 tipc_sk_rcv(n
->net
, &le
->inputq
);
534 bool tipc_node_is_up(struct tipc_node
*n
)
536 return n
->active_links
[0] != INVALID_BEARER_ID
;
539 void tipc_node_check_dest(struct net
*net
, u32 onode
,
540 struct tipc_bearer
*b
,
541 u16 capabilities
, u32 signature
,
542 struct tipc_media_addr
*maddr
,
543 bool *respond
, bool *dupl_addr
)
547 struct tipc_link_entry
*le
;
548 bool addr_match
= false;
549 bool sign_match
= false;
550 bool link_up
= false;
551 bool accept_addr
= false;
558 n
= tipc_node_create(net
, onode
, capabilities
);
564 le
= &n
->links
[b
->identity
];
566 /* Prepare to validate requesting node's signature and media address */
568 link_up
= l
&& tipc_link_is_up(l
);
569 addr_match
= l
&& !memcmp(&le
->maddr
, maddr
, sizeof(*maddr
));
570 sign_match
= (signature
== n
->signature
);
572 /* These three flags give us eight permutations: */
574 if (sign_match
&& addr_match
&& link_up
) {
575 /* All is fine. Do nothing. */
577 } else if (sign_match
&& addr_match
&& !link_up
) {
578 /* Respond. The link will come up in due time */
580 } else if (sign_match
&& !addr_match
&& link_up
) {
581 /* Peer has changed i/f address without rebooting.
582 * If so, the link will reset soon, and the next
583 * discovery will be accepted. So we can ignore it.
584 * It may also be an cloned or malicious peer having
585 * chosen the same node address and signature as an
587 * Ignore requests until the link goes down, if ever.
590 } else if (sign_match
&& !addr_match
&& !link_up
) {
591 /* Peer link has changed i/f address without rebooting.
592 * It may also be a cloned or malicious peer; we can't
593 * distinguish between the two.
594 * The signature is correct, so we must accept.
598 } else if (!sign_match
&& addr_match
&& link_up
) {
599 /* Peer node rebooted. Two possibilities:
600 * - Delayed re-discovery; this link endpoint has already
601 * reset and re-established contact with the peer, before
602 * receiving a discovery message from that node.
603 * (The peer happened to receive one from this node first).
604 * - The peer came back so fast that our side has not
605 * discovered it yet. Probing from this side will soon
606 * reset the link, since there can be no working link
607 * endpoint at the peer end, and the link will re-establish.
608 * Accept the signature, since it comes from a known peer.
610 n
->signature
= signature
;
611 } else if (!sign_match
&& addr_match
&& !link_up
) {
612 /* The peer node has rebooted.
613 * Accept signature, since it is a known peer.
615 n
->signature
= signature
;
617 } else if (!sign_match
&& !addr_match
&& link_up
) {
618 /* Peer rebooted with new address, or a new/duplicate peer.
619 * Ignore until the link goes down, if ever.
622 } else if (!sign_match
&& !addr_match
&& !link_up
) {
623 /* Peer rebooted with new address, or it is a new peer.
624 * Accept signature and address.
626 n
->signature
= signature
;
634 /* Now create new link if not already existing */
636 if (n
->link_cnt
== 2) {
637 pr_warn("Cannot establish 3rd link to %x\n", n
->addr
);
640 if_name
= strchr(b
->name
, ':') + 1;
641 if (!tipc_link_create(net
, if_name
, b
->identity
, b
->tolerance
,
642 b
->net_plane
, b
->mtu
, b
->priority
,
643 b
->window
, mod(tipc_net(net
)->random
),
644 tipc_own_addr(net
), onode
,
646 tipc_bc_sndlink(n
->net
), n
->bc_entry
.link
,
648 &n
->bc_entry
.namedq
, &l
)) {
653 tipc_link_fsm_evt(l
, LINK_RESET_EVT
);
654 if (n
->state
== NODE_FAILINGOVER
)
655 tipc_link_fsm_evt(l
, LINK_FAILOVER_BEGIN_EVT
);
658 tipc_node_calculate_timer(n
, l
);
659 if (n
->link_cnt
== 1)
660 if (!mod_timer(&n
->timer
, jiffies
+ n
->keepalive_intv
))
663 memcpy(&le
->maddr
, maddr
, sizeof(*maddr
));
666 if (reset
&& !tipc_link_is_reset(l
))
667 tipc_node_link_down(n
, b
->identity
, false);
671 void tipc_node_delete_links(struct net
*net
, int bearer_id
)
673 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
677 list_for_each_entry_rcu(n
, &tn
->node_list
, list
) {
678 tipc_node_link_down(n
, bearer_id
, true);
683 static void tipc_node_reset_links(struct tipc_node
*n
)
685 char addr_string
[16];
688 pr_warn("Resetting all links to %s\n",
689 tipc_addr_string_fill(addr_string
, n
->addr
));
691 for (i
= 0; i
< MAX_BEARERS
; i
++) {
692 tipc_node_link_down(n
, i
, false);
696 /* tipc_node_fsm_evt - node finite state machine
697 * Determines when contact is allowed with peer node
699 static void tipc_node_fsm_evt(struct tipc_node
*n
, int evt
)
701 int state
= n
->state
;
704 case SELF_DOWN_PEER_DOWN
:
706 case SELF_ESTABL_CONTACT_EVT
:
707 state
= SELF_UP_PEER_COMING
;
709 case PEER_ESTABL_CONTACT_EVT
:
710 state
= SELF_COMING_PEER_UP
;
712 case SELF_LOST_CONTACT_EVT
:
713 case PEER_LOST_CONTACT_EVT
:
715 case NODE_SYNCH_END_EVT
:
716 case NODE_SYNCH_BEGIN_EVT
:
717 case NODE_FAILOVER_BEGIN_EVT
:
718 case NODE_FAILOVER_END_EVT
:
723 case SELF_UP_PEER_UP
:
725 case SELF_LOST_CONTACT_EVT
:
726 state
= SELF_DOWN_PEER_LEAVING
;
728 case PEER_LOST_CONTACT_EVT
:
729 state
= SELF_LEAVING_PEER_DOWN
;
731 case NODE_SYNCH_BEGIN_EVT
:
732 state
= NODE_SYNCHING
;
734 case NODE_FAILOVER_BEGIN_EVT
:
735 state
= NODE_FAILINGOVER
;
737 case SELF_ESTABL_CONTACT_EVT
:
738 case PEER_ESTABL_CONTACT_EVT
:
739 case NODE_SYNCH_END_EVT
:
740 case NODE_FAILOVER_END_EVT
:
746 case SELF_DOWN_PEER_LEAVING
:
748 case PEER_LOST_CONTACT_EVT
:
749 state
= SELF_DOWN_PEER_DOWN
;
751 case SELF_ESTABL_CONTACT_EVT
:
752 case PEER_ESTABL_CONTACT_EVT
:
753 case SELF_LOST_CONTACT_EVT
:
755 case NODE_SYNCH_END_EVT
:
756 case NODE_SYNCH_BEGIN_EVT
:
757 case NODE_FAILOVER_BEGIN_EVT
:
758 case NODE_FAILOVER_END_EVT
:
763 case SELF_UP_PEER_COMING
:
765 case PEER_ESTABL_CONTACT_EVT
:
766 state
= SELF_UP_PEER_UP
;
768 case SELF_LOST_CONTACT_EVT
:
769 state
= SELF_DOWN_PEER_LEAVING
;
771 case SELF_ESTABL_CONTACT_EVT
:
772 case PEER_LOST_CONTACT_EVT
:
773 case NODE_SYNCH_END_EVT
:
774 case NODE_FAILOVER_BEGIN_EVT
:
776 case NODE_SYNCH_BEGIN_EVT
:
777 case NODE_FAILOVER_END_EVT
:
782 case SELF_COMING_PEER_UP
:
784 case SELF_ESTABL_CONTACT_EVT
:
785 state
= SELF_UP_PEER_UP
;
787 case PEER_LOST_CONTACT_EVT
:
788 state
= SELF_LEAVING_PEER_DOWN
;
790 case SELF_LOST_CONTACT_EVT
:
791 case PEER_ESTABL_CONTACT_EVT
:
793 case NODE_SYNCH_END_EVT
:
794 case NODE_SYNCH_BEGIN_EVT
:
795 case NODE_FAILOVER_BEGIN_EVT
:
796 case NODE_FAILOVER_END_EVT
:
801 case SELF_LEAVING_PEER_DOWN
:
803 case SELF_LOST_CONTACT_EVT
:
804 state
= SELF_DOWN_PEER_DOWN
;
806 case SELF_ESTABL_CONTACT_EVT
:
807 case PEER_ESTABL_CONTACT_EVT
:
808 case PEER_LOST_CONTACT_EVT
:
810 case NODE_SYNCH_END_EVT
:
811 case NODE_SYNCH_BEGIN_EVT
:
812 case NODE_FAILOVER_BEGIN_EVT
:
813 case NODE_FAILOVER_END_EVT
:
818 case NODE_FAILINGOVER
:
820 case SELF_LOST_CONTACT_EVT
:
821 state
= SELF_DOWN_PEER_LEAVING
;
823 case PEER_LOST_CONTACT_EVT
:
824 state
= SELF_LEAVING_PEER_DOWN
;
826 case NODE_FAILOVER_END_EVT
:
827 state
= SELF_UP_PEER_UP
;
829 case NODE_FAILOVER_BEGIN_EVT
:
830 case SELF_ESTABL_CONTACT_EVT
:
831 case PEER_ESTABL_CONTACT_EVT
:
833 case NODE_SYNCH_BEGIN_EVT
:
834 case NODE_SYNCH_END_EVT
:
841 case SELF_LOST_CONTACT_EVT
:
842 state
= SELF_DOWN_PEER_LEAVING
;
844 case PEER_LOST_CONTACT_EVT
:
845 state
= SELF_LEAVING_PEER_DOWN
;
847 case NODE_SYNCH_END_EVT
:
848 state
= SELF_UP_PEER_UP
;
850 case NODE_FAILOVER_BEGIN_EVT
:
851 state
= NODE_FAILINGOVER
;
853 case NODE_SYNCH_BEGIN_EVT
:
854 case SELF_ESTABL_CONTACT_EVT
:
855 case PEER_ESTABL_CONTACT_EVT
:
857 case NODE_FAILOVER_END_EVT
:
863 pr_err("Unknown node fsm state %x\n", state
);
870 pr_err("Illegal node fsm evt %x in state %x\n", evt
, state
);
873 bool tipc_node_filter_pkt(struct tipc_node
*n
, struct tipc_msg
*hdr
)
875 int state
= n
->state
;
877 if (likely(state
== SELF_UP_PEER_UP
))
880 if (state
== SELF_LEAVING_PEER_DOWN
)
883 if (state
== SELF_DOWN_PEER_LEAVING
) {
884 if (msg_peer_node_is_up(hdr
))
891 static void node_lost_contact(struct tipc_node
*n
,
892 struct sk_buff_head
*inputq
)
894 char addr_string
[16];
895 struct tipc_sock_conn
*conn
, *safe
;
897 struct list_head
*conns
= &n
->conn_sks
;
901 pr_debug("Lost contact with %s\n",
902 tipc_addr_string_fill(addr_string
, n
->addr
));
904 /* Clean up broadcast state */
905 tipc_bcast_remove_peer(n
->net
, n
->bc_entry
.link
);
907 /* Abort any ongoing link failover */
908 for (i
= 0; i
< MAX_BEARERS
; i
++) {
909 l
= n
->links
[i
].link
;
911 tipc_link_fsm_evt(l
, LINK_FAILOVER_END_EVT
);
914 /* Notify publications from this node */
915 n
->action_flags
|= TIPC_NOTIFY_NODE_DOWN
;
917 /* Notify sockets connected to node */
918 list_for_each_entry_safe(conn
, safe
, conns
, list
) {
919 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
920 SHORT_H_SIZE
, 0, tipc_own_addr(n
->net
),
921 conn
->peer_node
, conn
->port
,
922 conn
->peer_port
, TIPC_ERR_NO_NODE
);
924 skb_queue_tail(inputq
, skb
);
925 list_del(&conn
->list
);
931 * tipc_node_get_linkname - get the name of a link
933 * @bearer_id: id of the bearer
934 * @node: peer node address
935 * @linkname: link name output buffer
937 * Returns 0 on success
939 int tipc_node_get_linkname(struct net
*net
, u32 bearer_id
, u32 addr
,
940 char *linkname
, size_t len
)
942 struct tipc_link
*link
;
944 struct tipc_node
*node
= tipc_node_find(net
, addr
);
949 if (bearer_id
>= MAX_BEARERS
)
952 tipc_node_lock(node
);
953 link
= node
->links
[bearer_id
].link
;
955 strncpy(linkname
, link
->name
, len
);
959 tipc_node_unlock(node
);
964 void tipc_node_unlock(struct tipc_node
*node
)
966 struct net
*net
= node
->net
;
968 u32 flags
= node
->action_flags
;
970 struct list_head
*publ_list
;
972 if (likely(!flags
)) {
973 spin_unlock_bh(&node
->lock
);
978 link_id
= node
->link_id
;
979 publ_list
= &node
->publ_list
;
981 node
->action_flags
&= ~(TIPC_NOTIFY_NODE_DOWN
| TIPC_NOTIFY_NODE_UP
|
982 TIPC_NOTIFY_LINK_DOWN
| TIPC_NOTIFY_LINK_UP
);
984 spin_unlock_bh(&node
->lock
);
986 if (flags
& TIPC_NOTIFY_NODE_DOWN
)
987 tipc_publ_notify(net
, publ_list
, addr
);
989 if (flags
& TIPC_NOTIFY_NODE_UP
)
990 tipc_named_node_up(net
, addr
);
992 if (flags
& TIPC_NOTIFY_LINK_UP
)
993 tipc_nametbl_publish(net
, TIPC_LINK_STATE
, addr
, addr
,
994 TIPC_NODE_SCOPE
, link_id
, addr
);
996 if (flags
& TIPC_NOTIFY_LINK_DOWN
)
997 tipc_nametbl_withdraw(net
, TIPC_LINK_STATE
, addr
,
1002 /* Caller should hold node lock for the passed node */
1003 static int __tipc_nl_add_node(struct tipc_nl_msg
*msg
, struct tipc_node
*node
)
1006 struct nlattr
*attrs
;
1008 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
1009 NLM_F_MULTI
, TIPC_NL_NODE_GET
);
1013 attrs
= nla_nest_start(msg
->skb
, TIPC_NLA_NODE
);
1017 if (nla_put_u32(msg
->skb
, TIPC_NLA_NODE_ADDR
, node
->addr
))
1019 if (tipc_node_is_up(node
))
1020 if (nla_put_flag(msg
->skb
, TIPC_NLA_NODE_UP
))
1023 nla_nest_end(msg
->skb
, attrs
);
1024 genlmsg_end(msg
->skb
, hdr
);
1029 nla_nest_cancel(msg
->skb
, attrs
);
1031 genlmsg_cancel(msg
->skb
, hdr
);
1036 static struct tipc_link
*tipc_node_select_link(struct tipc_node
*n
, int sel
,
1038 struct tipc_media_addr
**maddr
)
1040 int id
= n
->active_links
[sel
& 1];
1042 if (unlikely(id
< 0))
1046 *maddr
= &n
->links
[id
].maddr
;
1047 return n
->links
[id
].link
;
1051 * tipc_node_xmit() is the general link level function for message sending
1052 * @net: the applicable net namespace
1053 * @list: chain of buffers containing message
1054 * @dnode: address of destination node
1055 * @selector: a number used for deterministic link selection
1056 * Consumes the buffer chain, except when returning -ELINKCONG
1057 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1059 int tipc_node_xmit(struct net
*net
, struct sk_buff_head
*list
,
1060 u32 dnode
, int selector
)
1062 struct tipc_link
*l
= NULL
;
1063 struct tipc_node
*n
;
1064 struct sk_buff_head xmitq
;
1065 struct tipc_media_addr
*maddr
;
1067 int rc
= -EHOSTUNREACH
;
1069 __skb_queue_head_init(&xmitq
);
1070 n
= tipc_node_find(net
, dnode
);
1073 l
= tipc_node_select_link(n
, selector
, &bearer_id
, &maddr
);
1075 rc
= tipc_link_xmit(l
, list
, &xmitq
);
1076 tipc_node_unlock(n
);
1077 if (unlikely(rc
== -ENOBUFS
))
1078 tipc_node_link_down(n
, bearer_id
, false);
1082 tipc_bearer_xmit(net
, bearer_id
, &xmitq
, maddr
);
1085 if (likely(in_own_node(net
, dnode
))) {
1086 tipc_sk_rcv(net
, list
);
1092 /* tipc_node_xmit_skb(): send single buffer to destination
1093 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1094 * messages, which will not be rejected
1095 * The only exception is datagram messages rerouted after secondary
1096 * lookup, which are rare and safe to dispose of anyway.
1097 * TODO: Return real return value, and let callers use
1098 * tipc_wait_for_sendpkt() where applicable
1100 int tipc_node_xmit_skb(struct net
*net
, struct sk_buff
*skb
, u32 dnode
,
1103 struct sk_buff_head head
;
1106 skb_queue_head_init(&head
);
1107 __skb_queue_tail(&head
, skb
);
1108 rc
= tipc_node_xmit(net
, &head
, dnode
, selector
);
1109 if (rc
== -ELINKCONG
)
1114 void tipc_node_broadcast(struct net
*net
, struct sk_buff
*skb
)
1116 struct sk_buff
*txskb
;
1117 struct tipc_node
*n
;
1121 list_for_each_entry_rcu(n
, tipc_nodes(net
), list
) {
1123 if (in_own_node(net
, dst
))
1125 if (!tipc_node_is_up(n
))
1127 txskb
= pskb_copy(skb
, GFP_ATOMIC
);
1130 msg_set_destnode(buf_msg(txskb
), dst
);
1131 tipc_node_xmit_skb(net
, txskb
, dst
, 0);
1139 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1140 * @net: the applicable net namespace
1142 * @bearer_id: id of bearer message arrived on
1144 * Invoked with no locks held.
1146 static void tipc_node_bc_rcv(struct net
*net
, struct sk_buff
*skb
, int bearer_id
)
1149 struct sk_buff_head xmitq
;
1150 struct tipc_bclink_entry
*be
;
1151 struct tipc_link_entry
*le
;
1152 struct tipc_msg
*hdr
= buf_msg(skb
);
1153 int usr
= msg_user(hdr
);
1154 u32 dnode
= msg_destnode(hdr
);
1155 struct tipc_node
*n
;
1157 __skb_queue_head_init(&xmitq
);
1159 /* If NACK for other node, let rcv link for that node peek into it */
1160 if ((usr
== BCAST_PROTOCOL
) && (dnode
!= tipc_own_addr(net
)))
1161 n
= tipc_node_find(net
, dnode
);
1163 n
= tipc_node_find(net
, msg_prevnode(hdr
));
1169 le
= &n
->links
[bearer_id
];
1171 rc
= tipc_bcast_rcv(net
, be
->link
, skb
);
1173 /* Broadcast link reset may happen at reassembly failure */
1174 if (rc
& TIPC_LINK_DOWN_EVT
)
1175 tipc_node_reset_links(n
);
1177 /* Broadcast ACKs are sent on a unicast link */
1178 if (rc
& TIPC_LINK_SND_BC_ACK
) {
1180 tipc_link_build_ack_msg(le
->link
, &xmitq
);
1181 tipc_node_unlock(n
);
1184 if (!skb_queue_empty(&xmitq
))
1185 tipc_bearer_xmit(net
, bearer_id
, &xmitq
, &le
->maddr
);
1187 /* Deliver. 'arrvq' is under inputq2's lock protection */
1188 if (!skb_queue_empty(&be
->inputq1
)) {
1189 spin_lock_bh(&be
->inputq2
.lock
);
1190 spin_lock_bh(&be
->inputq1
.lock
);
1191 skb_queue_splice_tail_init(&be
->inputq1
, &be
->arrvq
);
1192 spin_unlock_bh(&be
->inputq1
.lock
);
1193 spin_unlock_bh(&be
->inputq2
.lock
);
1194 tipc_sk_mcast_rcv(net
, &be
->arrvq
, &be
->inputq2
);
1200 * tipc_node_check_state - check and if necessary update node state
1202 * @bearer_id: identity of bearer delivering the packet
1203 * Returns true if state is ok, otherwise consumes buffer and returns false
1205 static bool tipc_node_check_state(struct tipc_node
*n
, struct sk_buff
*skb
,
1206 int bearer_id
, struct sk_buff_head
*xmitq
)
1208 struct tipc_msg
*hdr
= buf_msg(skb
);
1209 int usr
= msg_user(hdr
);
1210 int mtyp
= msg_type(hdr
);
1211 u16 oseqno
= msg_seqno(hdr
);
1212 u16 iseqno
= msg_seqno(msg_get_wrapped(hdr
));
1213 u16 exp_pkts
= msg_msgcnt(hdr
);
1214 u16 rcv_nxt
, syncpt
, dlv_nxt
;
1215 int state
= n
->state
;
1216 struct tipc_link
*l
, *tnl
, *pl
= NULL
;
1217 struct tipc_media_addr
*maddr
;
1220 l
= n
->links
[bearer_id
].link
;
1223 rcv_nxt
= l
->rcv_nxt
;
1226 if (likely((state
== SELF_UP_PEER_UP
) && (usr
!= TUNNEL_PROTOCOL
)))
1229 /* Find parallel link, if any */
1230 for (i
= 0; i
< MAX_BEARERS
; i
++) {
1231 if ((i
!= bearer_id
) && n
->links
[i
].link
) {
1232 pl
= n
->links
[i
].link
;
1237 /* Update node accesibility if applicable */
1238 if (state
== SELF_UP_PEER_COMING
) {
1239 if (!tipc_link_is_up(l
))
1241 if (!msg_peer_link_is_up(hdr
))
1243 tipc_node_fsm_evt(n
, PEER_ESTABL_CONTACT_EVT
);
1246 if (state
== SELF_DOWN_PEER_LEAVING
) {
1247 if (msg_peer_node_is_up(hdr
))
1249 tipc_node_fsm_evt(n
, PEER_LOST_CONTACT_EVT
);
1253 /* Ignore duplicate packets */
1254 if ((usr
!= LINK_PROTOCOL
) && less(oseqno
, rcv_nxt
))
1257 /* Initiate or update failover mode if applicable */
1258 if ((usr
== TUNNEL_PROTOCOL
) && (mtyp
== FAILOVER_MSG
)) {
1259 syncpt
= oseqno
+ exp_pkts
- 1;
1260 if (pl
&& tipc_link_is_up(pl
)) {
1261 pb_id
= pl
->bearer_id
;
1262 __tipc_node_link_down(n
, &pb_id
, xmitq
, &maddr
);
1263 tipc_skb_queue_splice_tail_init(pl
->inputq
, l
->inputq
);
1265 /* If pkts arrive out of order, use lowest calculated syncpt */
1266 if (less(syncpt
, n
->sync_point
))
1267 n
->sync_point
= syncpt
;
1270 /* Open parallel link when tunnel link reaches synch point */
1271 if ((n
->state
== NODE_FAILINGOVER
) && tipc_link_is_up(l
)) {
1272 if (!more(rcv_nxt
, n
->sync_point
))
1274 tipc_node_fsm_evt(n
, NODE_FAILOVER_END_EVT
);
1276 tipc_link_fsm_evt(pl
, LINK_FAILOVER_END_EVT
);
1280 /* No synching needed if only one link */
1281 if (!pl
|| !tipc_link_is_up(pl
))
1284 /* Initiate synch mode if applicable */
1285 if ((usr
== TUNNEL_PROTOCOL
) && (mtyp
== SYNCH_MSG
) && (oseqno
== 1)) {
1286 syncpt
= iseqno
+ exp_pkts
- 1;
1287 if (!tipc_link_is_up(l
)) {
1288 tipc_link_fsm_evt(l
, LINK_ESTABLISH_EVT
);
1289 __tipc_node_link_up(n
, bearer_id
, xmitq
);
1291 if (n
->state
== SELF_UP_PEER_UP
) {
1292 n
->sync_point
= syncpt
;
1293 tipc_link_fsm_evt(l
, LINK_SYNCH_BEGIN_EVT
);
1294 tipc_node_fsm_evt(n
, NODE_SYNCH_BEGIN_EVT
);
1298 /* Open tunnel link when parallel link reaches synch point */
1299 if (n
->state
== NODE_SYNCHING
) {
1300 if (tipc_link_is_synching(l
)) {
1306 dlv_nxt
= pl
->rcv_nxt
- mod(skb_queue_len(pl
->inputq
));
1307 if (more(dlv_nxt
, n
->sync_point
)) {
1308 tipc_link_fsm_evt(tnl
, LINK_SYNCH_END_EVT
);
1309 tipc_node_fsm_evt(n
, NODE_SYNCH_END_EVT
);
1314 if ((usr
== TUNNEL_PROTOCOL
) && (mtyp
== SYNCH_MSG
))
1316 if (usr
== LINK_PROTOCOL
)
1324 * tipc_rcv - process TIPC packets/messages arriving from off-node
1325 * @net: the applicable net namespace
1327 * @bearer: pointer to bearer message arrived on
1329 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1330 * structure (i.e. cannot be NULL), but bearer can be inactive.
1332 void tipc_rcv(struct net
*net
, struct sk_buff
*skb
, struct tipc_bearer
*b
)
1334 struct sk_buff_head xmitq
;
1335 struct tipc_node
*n
;
1336 struct tipc_msg
*hdr
= buf_msg(skb
);
1337 int usr
= msg_user(hdr
);
1338 int bearer_id
= b
->identity
;
1339 struct tipc_link_entry
*le
;
1340 u16 bc_ack
= msg_bcast_ack(hdr
);
1343 __skb_queue_head_init(&xmitq
);
1345 /* Ensure message is well-formed */
1346 if (unlikely(!tipc_msg_validate(skb
)))
1349 /* Handle arrival of discovery or broadcast packet */
1350 if (unlikely(msg_non_seq(hdr
))) {
1351 if (unlikely(usr
== LINK_CONFIG
))
1352 return tipc_disc_rcv(net
, skb
, b
);
1354 return tipc_node_bc_rcv(net
, skb
, bearer_id
);
1357 /* Locate neighboring node that sent packet */
1358 n
= tipc_node_find(net
, msg_prevnode(hdr
));
1361 le
= &n
->links
[bearer_id
];
1363 /* Ensure broadcast reception is in synch with peer's send state */
1364 if (unlikely(usr
== LINK_PROTOCOL
))
1365 tipc_bcast_sync_rcv(net
, n
->bc_entry
.link
, hdr
);
1366 else if (unlikely(n
->bc_entry
.link
->acked
!= bc_ack
))
1367 tipc_bcast_ack_rcv(net
, n
->bc_entry
.link
, bc_ack
);
1371 /* Is reception permitted at the moment ? */
1372 if (!tipc_node_filter_pkt(n
, hdr
))
1375 /* Check and if necessary update node state */
1376 if (likely(tipc_node_check_state(n
, skb
, bearer_id
, &xmitq
))) {
1377 rc
= tipc_link_rcv(le
->link
, skb
, &xmitq
);
1381 tipc_node_unlock(n
);
1383 if (unlikely(rc
& TIPC_LINK_UP_EVT
))
1384 tipc_node_link_up(n
, bearer_id
, &xmitq
);
1386 if (unlikely(rc
& TIPC_LINK_DOWN_EVT
))
1387 tipc_node_link_down(n
, bearer_id
, false);
1389 if (unlikely(!skb_queue_empty(&n
->bc_entry
.namedq
)))
1390 tipc_named_rcv(net
, &n
->bc_entry
.namedq
);
1392 if (!skb_queue_empty(&le
->inputq
))
1393 tipc_sk_rcv(net
, &le
->inputq
);
1395 if (!skb_queue_empty(&xmitq
))
1396 tipc_bearer_xmit(net
, bearer_id
, &xmitq
, &le
->maddr
);
1403 int tipc_nl_node_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1406 struct net
*net
= sock_net(skb
->sk
);
1407 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
1408 int done
= cb
->args
[0];
1409 int last_addr
= cb
->args
[1];
1410 struct tipc_node
*node
;
1411 struct tipc_nl_msg msg
;
1417 msg
.portid
= NETLINK_CB(cb
->skb
).portid
;
1418 msg
.seq
= cb
->nlh
->nlmsg_seq
;
1422 node
= tipc_node_find(net
, last_addr
);
1425 /* We never set seq or call nl_dump_check_consistent()
1426 * this means that setting prev_seq here will cause the
1427 * consistence check to fail in the netlink callback
1428 * handler. Resulting in the NLMSG_DONE message having
1429 * the NLM_F_DUMP_INTR flag set if the node state
1430 * changed while we released the lock.
1435 tipc_node_put(node
);
1438 list_for_each_entry_rcu(node
, &tn
->node_list
, list
) {
1440 if (node
->addr
== last_addr
)
1446 tipc_node_lock(node
);
1447 err
= __tipc_nl_add_node(&msg
, node
);
1449 last_addr
= node
->addr
;
1450 tipc_node_unlock(node
);
1454 tipc_node_unlock(node
);
1459 cb
->args
[1] = last_addr
;