1 /* Copyright (c) 2013 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License. */
18 #include <arpa/inet.h>
20 #include "byte-order.h"
23 #include "dynamic-string.h"
31 #include "ovs-thread.h"
32 #include "openvswitch/types.h"
34 #include "poll-loop.h"
42 VLOG_DEFINE_THIS_MODULE(bfd);
46 * The goal of this module is to replace CFM with something both more flexible
47 * and standards compliant. In service of this goal, the following needs to be
51 * * Implement Demand mode.
52 * * Go through the RFC line by line and verify we comply.
53 * * Test against a hardware implementation. Preferably a popular one.
54 * * Delete BFD packets with nw_ttl != 255 in the datapath to prevent DOS
59 * - BFD show into ovs-bugtool.
61 * - Set TOS/PCP on inner BFD frame, and outer tunnel header when encapped.
63 * - Sending BFD messages should be in its own thread/process.
65 * - Scale testing. How does it operate when there are large number of bfd
66 * sessions? Do we ever have random flaps? What's the CPU utilization?
68 * - Rely on data traffic for liveness by using BFD demand mode.
69 * If we're receiving traffic on a port, we can safely assume it's up (modulo
70 * unidrectional failures). BFD has a demand mode in which it can stay quiet
71 * unless it feels the need to check the status of the port. Using this, we
72 * can implement a strategy in which BFD only sends control messages on dark
75 * - Depending on how one interprets the spec, it appears that a BFD session
76 * can never change bfd.LocalDiag to "No Diagnostic". We should verify that
77 * this is what hardware implementations actually do. Seems like "No
78 * Diagnostic" should be set once a BFD session state goes UP. */
83 FLAG_MULTIPOINT = 1 << 0,
92 STATE_ADMIN_DOWN = 0 << 6,
99 DIAG_NONE = 0, /* No Diagnostic. */
100 DIAG_EXPIRED = 1, /* Control Detection Time Expired. */
101 DIAG_ECHO_FAILED = 2, /* Echo Function Failed. */
102 DIAG_RMT_DOWN = 3, /* Neighbor Signaled Session Down. */
103 DIAG_FWD_RESET = 4, /* Forwarding Plane Reset. */
104 DIAG_PATH_DOWN = 5, /* Path Down. */
105 DIAG_CPATH_DOWN = 6, /* Concatenated Path Down. */
106 DIAG_ADMIN_DOWN = 7, /* Administratively Down. */
107 DIAG_RCPATH_DOWN = 8 /* Reverse Concatenated Path Down. */
110 /* RFC 5880 Section 4.1
112 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114 * |Vers | Diag |Sta|P|F|C|A|D|M| Detect Mult | Length |
115 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116 * | My Discriminator |
117 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118 * | Your Discriminator |
119 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
120 * | Desired Min TX Interval |
121 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122 * | Required Min RX Interval |
123 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124 * | Required Min Echo RX Interval |
125 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ */
127 uint8_t vers_diag; /* Version and diagnostic. */
128 uint8_t flags; /* 2bit State field followed by flags. */
129 uint8_t mult; /* Fault detection multiplier. */
130 uint8_t length; /* Length of this BFD message. */
131 ovs_be32 my_disc; /* My discriminator. */
132 ovs_be32 your_disc; /* Your discriminator. */
133 ovs_be32 min_tx; /* Desired minimum tx interval. */
134 ovs_be32 min_rx; /* Required minimum rx interval. */
135 ovs_be32 min_rx_echo; /* Required minimum echo rx interval. */
137 BUILD_ASSERT_DECL(BFD_PACKET_LEN == sizeof(struct msg));
139 #define DIAG_MASK 0x1f
141 #define STATE_MASK 0xC0
142 #define FLAGS_MASK 0x3f
145 struct hmap_node node; /* In 'all_bfds'. */
146 uint32_t disc; /* bfd.LocalDiscr. Key in 'all_bfds' hmap. */
148 char *name; /* Name used for logging. */
150 bool cpath_down; /* Concatenated Path Down. */
151 uint8_t mult; /* bfd.DetectMult. */
153 enum state state; /* bfd.SessionState. */
154 enum state rmt_state; /* bfd.RemoteSessionState. */
156 enum diag diag; /* bfd.LocalDiag. */
157 enum diag rmt_diag; /* Remote diagnostic. */
159 enum flags flags; /* Flags sent on messages. */
160 enum flags rmt_flags; /* Flags last received. */
162 uint32_t rmt_disc; /* bfd.RemoteDiscr. */
164 uint16_t udp_src; /* UDP source port. */
166 /* All timers in milliseconds. */
167 long long int rmt_min_rx; /* bfd.RemoteMinRxInterval. */
168 long long int rmt_min_tx; /* Remote minimum TX interval. */
170 long long int cfg_min_tx; /* Configured minimum TX rate. */
171 long long int cfg_min_rx; /* Configured required minimum RX rate. */
172 long long int poll_min_tx; /* Min TX negotating in a poll sequence. */
173 long long int poll_min_rx; /* Min RX negotating in a poll sequence. */
174 long long int min_tx; /* bfd.DesiredMinTxInterval. */
175 long long int min_rx; /* bfd.RequiredMinRxInterval. */
177 long long int last_tx; /* Last TX time. */
178 long long int next_tx; /* Next TX time. */
179 long long int detect_time; /* RFC 5880 6.8.4 Detection time. */
181 int forwarding_override; /* Manual override of 'forwarding' status. */
183 atomic_bool check_tnl_key; /* Verify tunnel key of inbound packets? */
187 static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
188 static struct hmap all_bfds__ = HMAP_INITIALIZER(&all_bfds__);
189 static struct hmap *const all_bfds OVS_GUARDED_BY(mutex) = &all_bfds__;
191 static bool bfd_forwarding__(const struct bfd *) OVS_REQ_WRLOCK(mutex);
192 static bool bfd_in_poll(const struct bfd *) OVS_REQ_WRLOCK(&mutex);
193 static void bfd_poll(struct bfd *bfd) OVS_REQ_WRLOCK(&mutex);
194 static const char *bfd_diag_str(enum diag) OVS_REQ_WRLOCK(&mutex);
195 static const char *bfd_state_str(enum state) OVS_REQ_WRLOCK(&mutex);
196 static long long int bfd_min_tx(const struct bfd *) OVS_REQ_WRLOCK(&mutex);
197 static long long int bfd_tx_interval(const struct bfd *)
198 OVS_REQ_WRLOCK(&mutex);
199 static long long int bfd_rx_interval(const struct bfd *)
200 OVS_REQ_WRLOCK(&mutex);
201 static void bfd_set_next_tx(struct bfd *) OVS_REQ_WRLOCK(&mutex);
202 static void bfd_set_state(struct bfd *, enum state, enum diag)
203 OVS_REQ_WRLOCK(&mutex);
204 static uint32_t generate_discriminator(void) OVS_REQ_WRLOCK(&mutex);
205 static void bfd_put_details(struct ds *, const struct bfd *)
206 OVS_REQ_WRLOCK(&mutex);
207 static void bfd_unixctl_show(struct unixctl_conn *, int argc,
208 const char *argv[], void *aux OVS_UNUSED);
209 static void bfd_unixctl_set_forwarding_override(struct unixctl_conn *,
210 int argc, const char *argv[],
211 void *aux OVS_UNUSED);
212 static void log_msg(enum vlog_level, const struct msg *, const char *message,
213 const struct bfd *) OVS_REQ_WRLOCK(&mutex);
215 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(20, 20);
217 /* Returns true if the interface on which 'bfd' is running may be used to
218 * forward traffic according to the BFD session state. */
220 bfd_forwarding(const struct bfd *bfd) OVS_EXCLUDED(mutex)
224 ovs_mutex_lock(&mutex);
225 ret = bfd_forwarding__(bfd);
226 ovs_mutex_unlock(&mutex);
230 /* Returns a 'smap' of key value pairs representing the status of 'bfd'
231 * intended for the OVS database. */
233 bfd_get_status(const struct bfd *bfd, struct smap *smap)
236 ovs_mutex_lock(&mutex);
237 smap_add(smap, "forwarding", bfd_forwarding__(bfd)? "true" : "false");
238 smap_add(smap, "state", bfd_state_str(bfd->state));
239 smap_add(smap, "diagnostic", bfd_diag_str(bfd->diag));
241 if (bfd->state != STATE_DOWN) {
242 smap_add(smap, "remote_state", bfd_state_str(bfd->rmt_state));
243 smap_add(smap, "remote_diagnostic", bfd_diag_str(bfd->rmt_diag));
245 ovs_mutex_unlock(&mutex);
248 /* Initializes, destroys, or reconfigures the BFD session 'bfd' (named 'name'),
249 * according to the database configuration contained in 'cfg'. Takes ownership
250 * of 'bfd', which may be NULL. Returns a BFD object which may be used as a
251 * handle for the session, or NULL if BFD is not enabled according to 'cfg'.
252 * Also returns NULL if cfg is NULL. */
254 bfd_configure(struct bfd *bfd, const char *name, const struct smap *cfg)
257 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
258 static atomic_uint16_t udp_src = ATOMIC_VAR_INIT(0);
260 long long int min_tx, min_rx;
263 if (ovsthread_once_start(&once)) {
264 unixctl_command_register("bfd/show", "[interface]", 0, 1,
265 bfd_unixctl_show, NULL);
266 unixctl_command_register("bfd/set-forwarding",
267 "[interface] normal|false|true", 1, 2,
268 bfd_unixctl_set_forwarding_override, NULL);
269 ovsthread_once_done(&once);
272 if (!cfg || !smap_get_bool(cfg, "enable", false)) {
277 ovs_mutex_lock(&mutex);
279 bfd = xzalloc(sizeof *bfd);
280 bfd->name = xstrdup(name);
281 bfd->forwarding_override = -1;
282 bfd->disc = generate_discriminator();
283 hmap_insert(all_bfds, &bfd->node, bfd->disc);
285 bfd->diag = DIAG_NONE;
288 atomic_init(&bfd->ref_cnt, 1);
290 /* RFC 5881 section 4
291 * The source port MUST be in the range 49152 through 65535. The same
292 * UDP source port number MUST be used for all BFD Control packets
293 * associated with a particular session. The source port number SHOULD
294 * be unique among all BFD sessions on the system. */
295 atomic_add(&udp_src, 1, &bfd->udp_src);
296 bfd->udp_src = (bfd->udp_src % 16384) + 49152;
298 bfd_set_state(bfd, STATE_DOWN, DIAG_NONE);
301 atomic_store(&bfd->check_tnl_key,
302 smap_get_bool(cfg, "check_tnl_key", false));
303 min_tx = smap_get_int(cfg, "min_tx", 100);
304 min_tx = MAX(min_tx, 100);
305 if (bfd->cfg_min_tx != min_tx) {
306 bfd->cfg_min_tx = min_tx;
307 if (bfd->state != STATE_UP
308 || (!bfd_in_poll(bfd) && bfd->cfg_min_tx < bfd->min_tx)) {
309 bfd->min_tx = bfd->cfg_min_tx;
314 min_rx = smap_get_int(cfg, "min_rx", 1000);
315 min_rx = MAX(min_rx, 100);
316 if (bfd->cfg_min_rx != min_rx) {
317 bfd->cfg_min_rx = min_rx;
318 if (bfd->state != STATE_UP
319 || (!bfd_in_poll(bfd) && bfd->cfg_min_rx > bfd->min_rx)) {
320 bfd->min_rx = bfd->cfg_min_rx;
325 cpath_down = smap_get_bool(cfg, "cpath_down", false);
326 if (bfd->cpath_down != cpath_down) {
327 bfd->cpath_down = cpath_down;
328 if (bfd->diag == DIAG_NONE || bfd->diag == DIAG_CPATH_DOWN) {
329 bfd_set_state(bfd, bfd->state, DIAG_NONE);
333 ovs_mutex_unlock(&mutex);
338 bfd_ref(const struct bfd *bfd_)
340 struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
343 atomic_add(&bfd->ref_cnt, 1, &orig);
344 ovs_assert(orig > 0);
350 bfd_unref(struct bfd *bfd) OVS_EXCLUDED(mutex)
355 atomic_sub(&bfd->ref_cnt, 1, &orig);
356 ovs_assert(orig > 0);
358 ovs_mutex_lock(&mutex);
359 hmap_remove(all_bfds, &bfd->node);
362 ovs_mutex_unlock(&mutex);
368 bfd_wait(const struct bfd *bfd) OVS_EXCLUDED(mutex)
370 ovs_mutex_lock(&mutex);
371 if (bfd->flags & FLAG_FINAL) {
372 poll_immediate_wake();
375 poll_timer_wait_until(bfd->next_tx);
376 if (bfd->state > STATE_DOWN) {
377 poll_timer_wait_until(bfd->detect_time);
379 ovs_mutex_unlock(&mutex);
383 bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
385 ovs_mutex_lock(&mutex);
386 if (bfd->state > STATE_DOWN && time_msec() >= bfd->detect_time) {
387 bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
390 if (bfd->min_tx != bfd->cfg_min_tx || bfd->min_rx != bfd->cfg_min_rx) {
393 ovs_mutex_unlock(&mutex);
397 bfd_should_send_packet(const struct bfd *bfd) OVS_EXCLUDED(mutex)
400 ovs_mutex_lock(&mutex);
401 ret = bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
402 ovs_mutex_unlock(&mutex);
407 bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
408 uint8_t eth_src[ETH_ADDR_LEN]) OVS_EXCLUDED(mutex)
410 long long int min_tx, min_rx;
411 struct udp_header *udp;
412 struct eth_header *eth;
413 struct ip_header *ip;
416 ovs_mutex_lock(&mutex);
418 long long int delay = time_msec() - bfd->next_tx;
419 long long int interval = bfd_tx_interval(bfd);
420 if (delay > interval * 3 / 2) {
421 VLOG_INFO("%s: long delay of %lldms (expected %lldms) sending BFD"
422 " control message", bfd->name, delay, interval);
426 /* RFC 5880 Section 6.5
427 * A BFD Control packet MUST NOT have both the Poll (P) and Final (F) bits
429 ovs_assert(!(bfd->flags & FLAG_POLL) || !(bfd->flags & FLAG_FINAL));
431 ofpbuf_reserve(p, 2); /* Properly align after the ethernet header. */
432 eth = ofpbuf_put_uninit(p, sizeof *eth);
433 memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
434 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
435 eth->eth_type = htons(ETH_TYPE_IP);
437 ip = ofpbuf_put_zeros(p, sizeof *ip);
438 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
439 ip->ip_tot_len = htons(sizeof *ip + sizeof *udp + sizeof *msg);
441 ip->ip_proto = IPPROTO_UDP;
442 ip->ip_src = htonl(0xA9FE0100); /* 169.254.1.0 Link Local. */
443 ip->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1 Link Local. */
444 ip->ip_csum = csum(ip, sizeof *ip);
446 udp = ofpbuf_put_zeros(p, sizeof *udp);
447 udp->udp_src = htons(bfd->udp_src);
448 udp->udp_dst = htons(BFD_DEST_PORT);
449 udp->udp_len = htons(sizeof *udp + sizeof *msg);
451 msg = ofpbuf_put_uninit(p, sizeof *msg);
452 msg->vers_diag = (BFD_VERSION << 5) | bfd->diag;
453 msg->flags = (bfd->state & STATE_MASK) | bfd->flags;
455 msg->mult = bfd->mult;
456 msg->length = BFD_PACKET_LEN;
457 msg->my_disc = htonl(bfd->disc);
458 msg->your_disc = htonl(bfd->rmt_disc);
459 msg->min_rx_echo = htonl(0);
461 if (bfd_in_poll(bfd)) {
462 min_tx = bfd->poll_min_tx;
463 min_rx = bfd->poll_min_rx;
465 min_tx = bfd_min_tx(bfd);
466 min_rx = bfd->min_rx;
469 msg->min_tx = htonl(min_tx * 1000);
470 msg->min_rx = htonl(min_rx * 1000);
472 bfd->flags &= ~FLAG_FINAL;
474 log_msg(VLL_DBG, msg, "Sending BFD Message", bfd);
476 bfd->last_tx = time_msec();
477 bfd_set_next_tx(bfd);
478 ovs_mutex_unlock(&mutex);
482 bfd_should_process_flow(const struct bfd *bfd, const struct flow *flow,
483 struct flow_wildcards *wc)
487 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
488 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
490 atomic_read(&bfd->check_tnl_key, &check_tnl_key);
492 memset(&wc->masks.tunnel.tun_id, 0xff, sizeof wc->masks.tunnel.tun_id);
494 return (flow->dl_type == htons(ETH_TYPE_IP)
495 && flow->nw_proto == IPPROTO_UDP
496 && flow->tp_dst == htons(3784)
497 && (check_tnl_key || flow->tunnel.tun_id == htonll(0)));
501 bfd_process_packet(struct bfd *bfd, const struct flow *flow,
502 const struct ofpbuf *p) OVS_EXCLUDED(mutex)
504 uint32_t rmt_min_rx, pkt_your_disc;
505 enum state rmt_state;
510 /* This function is designed to follow section RFC 5880 6.8.6 closely. */
512 ovs_mutex_lock(&mutex);
513 if (flow->nw_ttl != 255) {
514 /* XXX Should drop in the kernel to prevent DOS. */
518 msg = ofpbuf_at(p, (uint8_t *)p->l7 - (uint8_t *)p->data, BFD_PACKET_LEN);
520 VLOG_INFO_RL(&rl, "%s: Received unparseable BFD control message.",
525 /* RFC 5880 Section 6.8.6
526 * If the Length field is greater than the payload of the encapsulating
527 * protocol, the packet MUST be discarded.
529 * Note that we make this check implicity. Above we use ofpbuf_at() to
530 * ensure that there are at least BFD_PACKET_LEN bytes in the payload of
531 * the encapsulating protocol. Below we require msg->length to be exactly
532 * BFD_PACKET_LEN bytes. */
534 flags = msg->flags & FLAGS_MASK;
535 rmt_state = msg->flags & STATE_MASK;
536 version = msg->vers_diag >> VERS_SHIFT;
538 log_msg(VLL_DBG, msg, "Received BFD control message", bfd);
540 if (version != BFD_VERSION) {
541 log_msg(VLL_WARN, msg, "Incorrect version", bfd);
545 /* Technically this should happen after the length check. We don't support
546 * authentication however, so it's simpler to do the check first. */
547 if (flags & FLAG_AUTH) {
548 log_msg(VLL_WARN, msg, "Authenticated control message with"
549 " authentication disabled", bfd);
553 if (msg->length != BFD_PACKET_LEN) {
554 log_msg(VLL_WARN, msg, "Unexpected length", bfd);
555 if (msg->length < BFD_PACKET_LEN) {
561 log_msg(VLL_WARN, msg, "Zero multiplier", bfd);
565 if (flags & FLAG_MULTIPOINT) {
566 log_msg(VLL_WARN, msg, "Unsupported multipoint flag", bfd);
571 log_msg(VLL_WARN, msg, "NULL my_disc", bfd);
575 pkt_your_disc = ntohl(msg->your_disc);
577 /* Technically, we should use the your discriminator field to figure
578 * out which 'struct bfd' this packet is destined towards. That way a
579 * bfd session could migrate from one interface to another
580 * transparently. This doesn't fit in with the OVS structure very
581 * well, so in this respect, we are not compliant. */
582 if (pkt_your_disc != bfd->disc) {
583 log_msg(VLL_WARN, msg, "Incorrect your_disc", bfd);
586 } else if (rmt_state > STATE_DOWN) {
587 log_msg(VLL_WARN, msg, "Null your_disc", bfd);
591 bfd->rmt_disc = ntohl(msg->my_disc);
592 bfd->rmt_state = rmt_state;
593 bfd->rmt_flags = flags;
594 bfd->rmt_diag = msg->vers_diag & DIAG_MASK;
596 if (flags & FLAG_FINAL && bfd_in_poll(bfd)) {
597 bfd->min_tx = bfd->poll_min_tx;
598 bfd->min_rx = bfd->poll_min_rx;
599 bfd->flags &= ~FLAG_POLL;
600 log_msg(VLL_INFO, msg, "Poll sequence terminated", bfd);
603 if (flags & FLAG_POLL) {
604 /* RFC 5880 Section 6.5
605 * When the other system receives a Poll, it immediately transmits a
606 * BFD Control packet with the Final (F) bit set, independent of any
607 * periodic BFD Control packets it may be sending
608 * (see section 6.8.7). */
609 bfd->flags &= ~FLAG_POLL;
610 bfd->flags |= FLAG_FINAL;
613 rmt_min_rx = MAX(ntohl(msg->min_rx) / 1000, 1);
614 if (bfd->rmt_min_rx != rmt_min_rx) {
615 bfd->rmt_min_rx = rmt_min_rx;
616 bfd_set_next_tx(bfd);
617 log_msg(VLL_INFO, msg, "New remote min_rx", bfd);
620 bfd->rmt_min_tx = MAX(ntohl(msg->min_tx) / 1000, 1);
621 bfd->detect_time = bfd_rx_interval(bfd) * bfd->mult + time_msec();
623 if (bfd->state == STATE_ADMIN_DOWN) {
624 VLOG_DBG_RL(&rl, "Administratively down, dropping control message.");
628 if (rmt_state == STATE_ADMIN_DOWN) {
629 if (bfd->state != STATE_DOWN) {
630 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
633 switch (bfd->state) {
635 if (rmt_state == STATE_DOWN) {
636 bfd_set_state(bfd, STATE_INIT, bfd->diag);
637 } else if (rmt_state == STATE_INIT) {
638 bfd_set_state(bfd, STATE_UP, bfd->diag);
642 if (rmt_state > STATE_DOWN) {
643 bfd_set_state(bfd, STATE_UP, bfd->diag);
647 if (rmt_state <= STATE_DOWN) {
648 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
649 log_msg(VLL_INFO, msg, "Remote signaled STATE_DOWN", bfd);
652 case STATE_ADMIN_DOWN:
657 /* XXX: RFC 5880 Section 6.8.6 Demand mode related calculations here. */
660 ovs_mutex_unlock(&mutex);
664 bfd_forwarding__(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
666 if (bfd->forwarding_override != -1) {
667 return bfd->forwarding_override == 1;
670 return bfd->state == STATE_UP
671 && bfd->rmt_diag != DIAG_PATH_DOWN
672 && bfd->rmt_diag != DIAG_CPATH_DOWN
673 && bfd->rmt_diag != DIAG_RCPATH_DOWN;
678 bfd_in_poll(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
680 return (bfd->flags & FLAG_POLL) != 0;
684 bfd_poll(struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
686 if (bfd->state > STATE_DOWN && !bfd_in_poll(bfd)
687 && !(bfd->flags & FLAG_FINAL)) {
688 bfd->poll_min_tx = bfd->cfg_min_tx;
689 bfd->poll_min_rx = bfd->cfg_min_rx;
690 bfd->flags |= FLAG_POLL;
692 VLOG_INFO_RL(&rl, "%s: Initiating poll sequence", bfd->name);
697 bfd_min_tx(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
699 /* RFC 5880 Section 6.8.3
700 * When bfd.SessionState is not Up, the system MUST set
701 * bfd.DesiredMinTxInterval to a value of not less than one second
702 * (1,000,000 microseconds). This is intended to ensure that the
703 * bandwidth consumed by BFD sessions that are not Up is negligible,
704 * particularly in the case where a neighbor may not be running BFD. */
705 return (bfd->state == STATE_UP ? bfd->min_tx : MAX(bfd->min_tx, 1000));
709 bfd_tx_interval(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
711 long long int interval = bfd_min_tx(bfd);
712 return MAX(interval, bfd->rmt_min_rx);
716 bfd_rx_interval(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
718 return MAX(bfd->min_rx, bfd->rmt_min_tx);
722 bfd_set_next_tx(struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
724 long long int interval = bfd_tx_interval(bfd);
725 interval -= interval * random_range(26) / 100;
726 bfd->next_tx = bfd->last_tx + interval;
730 bfd_flag_str(enum flags flags)
732 struct ds ds = DS_EMPTY_INITIALIZER;
733 static char flag_str[128];
739 if (flags & FLAG_MULTIPOINT) {
740 ds_put_cstr(&ds, "multipoint ");
743 if (flags & FLAG_DEMAND) {
744 ds_put_cstr(&ds, "demand ");
747 if (flags & FLAG_AUTH) {
748 ds_put_cstr(&ds, "auth ");
751 if (flags & FLAG_CTL) {
752 ds_put_cstr(&ds, "ctl ");
755 if (flags & FLAG_FINAL) {
756 ds_put_cstr(&ds, "final ");
759 if (flags & FLAG_POLL) {
760 ds_put_cstr(&ds, "poll ");
763 ovs_strlcpy(flag_str, ds_cstr(&ds), sizeof flag_str);
769 bfd_state_str(enum state state)
772 case STATE_ADMIN_DOWN: return "admin_down";
773 case STATE_DOWN: return "down";
774 case STATE_INIT: return "init";
775 case STATE_UP: return "up";
776 default: return "invalid";
781 bfd_diag_str(enum diag diag) {
783 case DIAG_NONE: return "No Diagnostic";
784 case DIAG_EXPIRED: return "Control Detection Time Expired";
785 case DIAG_ECHO_FAILED: return "Echo Function Failed";
786 case DIAG_RMT_DOWN: return "Neighbor Signaled Session Down";
787 case DIAG_FWD_RESET: return "Forwarding Plane Reset";
788 case DIAG_PATH_DOWN: return "Path Down";
789 case DIAG_CPATH_DOWN: return "Concatenated Path Down";
790 case DIAG_ADMIN_DOWN: return "Administratively Down";
791 case DIAG_RCPATH_DOWN: return "Reverse Concatenated Path Down";
792 default: return "Invalid Diagnostic";
797 log_msg(enum vlog_level level, const struct msg *p, const char *message,
798 const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
800 struct ds ds = DS_EMPTY_INITIALIZER;
802 if (vlog_should_drop(THIS_MODULE, level, &rl)) {
808 "\n\tvers:%"PRIu8" diag:\"%s\" state:%s mult:%"PRIu8
811 "\n\tmy_disc:0x%"PRIx32" your_disc:0x%"PRIx32
812 "\n\tmin_tx:%"PRIu32"us (%"PRIu32"ms)"
813 "\n\tmin_rx:%"PRIu32"us (%"PRIu32"ms)"
814 "\n\tmin_rx_echo:%"PRIu32"us (%"PRIu32"ms)",
815 bfd->name, message, p->vers_diag >> VERS_SHIFT,
816 bfd_diag_str(p->vers_diag & DIAG_MASK),
817 bfd_state_str(p->flags & STATE_MASK),
818 p->mult, p->length, bfd_flag_str(p->flags & FLAGS_MASK),
819 ntohl(p->my_disc), ntohl(p->your_disc),
820 ntohl(p->min_tx), ntohl(p->min_tx) / 1000,
821 ntohl(p->min_rx), ntohl(p->min_rx) / 1000,
822 ntohl(p->min_rx_echo), ntohl(p->min_rx_echo) / 1000);
823 bfd_put_details(&ds, bfd);
824 VLOG(level, "%s", ds_cstr(&ds));
829 bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
830 OVS_REQ_WRLOCK(mutex)
832 if (diag == DIAG_NONE && bfd->cpath_down) {
833 diag = DIAG_CPATH_DOWN;
836 if (bfd->state != state || bfd->diag != diag) {
837 if (!VLOG_DROP_INFO(&rl)) {
838 struct ds ds = DS_EMPTY_INITIALIZER;
840 ds_put_format(&ds, "%s: BFD state change: %s->%s"
841 " \"%s\"->\"%s\".\n",
842 bfd->name, bfd_state_str(bfd->state),
843 bfd_state_str(state), bfd_diag_str(bfd->diag),
845 bfd_put_details(&ds, bfd);
846 VLOG_INFO("%s", ds_cstr(&ds));
853 if (bfd->state <= STATE_DOWN) {
854 bfd->rmt_state = STATE_DOWN;
855 bfd->rmt_diag = DIAG_NONE;
865 generate_discriminator(void)
869 /* RFC 5880 Section 6.8.1
870 * It SHOULD be set to a random (but still unique) value to improve
871 * security. The value is otherwise outside the scope of this
877 /* 'disc' is by definition random, so there's no reason to waste time
879 disc = random_uint32();
880 HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, all_bfds) {
881 if (bfd->disc == disc) {
892 bfd_find_by_name(const char *name) OVS_REQ_WRLOCK(mutex)
896 HMAP_FOR_EACH (bfd, node, all_bfds) {
897 if (!strcmp(bfd->name, name)) {
905 bfd_put_details(struct ds *ds, const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
907 ds_put_format(ds, "\tForwarding: %s\n",
908 bfd_forwarding__(bfd) ? "true" : "false");
909 ds_put_format(ds, "\tDetect Multiplier: %d\n", bfd->mult);
910 ds_put_format(ds, "\tConcatenated Path Down: %s\n",
911 bfd->cpath_down ? "true" : "false");
912 ds_put_format(ds, "\tTX Interval: Approx %lldms\n", bfd_tx_interval(bfd));
913 ds_put_format(ds, "\tRX Interval: Approx %lldms\n", bfd_rx_interval(bfd));
914 ds_put_format(ds, "\tDetect Time: now %+lldms\n",
915 time_msec() - bfd->detect_time);
916 ds_put_format(ds, "\tNext TX Time: now %+lldms\n",
917 time_msec() - bfd->next_tx);
918 ds_put_format(ds, "\tLast TX Time: now %+lldms\n",
919 time_msec() - bfd->last_tx);
921 ds_put_cstr(ds, "\n");
923 ds_put_format(ds, "\tLocal Flags: %s\n", bfd_flag_str(bfd->flags));
924 ds_put_format(ds, "\tLocal Session State: %s\n",
925 bfd_state_str(bfd->state));
926 ds_put_format(ds, "\tLocal Diagnostic: %s\n", bfd_diag_str(bfd->diag));
927 ds_put_format(ds, "\tLocal Discriminator: 0x%"PRIx32"\n", bfd->disc);
928 ds_put_format(ds, "\tLocal Minimum TX Interval: %lldms\n",
930 ds_put_format(ds, "\tLocal Minimum RX Interval: %lldms\n", bfd->min_rx);
932 ds_put_cstr(ds, "\n");
934 ds_put_format(ds, "\tRemote Flags: %s\n", bfd_flag_str(bfd->rmt_flags));
935 ds_put_format(ds, "\tRemote Session State: %s\n",
936 bfd_state_str(bfd->rmt_state));
937 ds_put_format(ds, "\tRemote Diagnostic: %s\n",
938 bfd_diag_str(bfd->rmt_diag));
939 ds_put_format(ds, "\tRemote Discriminator: 0x%"PRIx32"\n", bfd->rmt_disc);
940 ds_put_format(ds, "\tRemote Minimum TX Interval: %lldms\n",
942 ds_put_format(ds, "\tRemote Minimum RX Interval: %lldms\n",
947 bfd_unixctl_show(struct unixctl_conn *conn, int argc, const char *argv[],
948 void *aux OVS_UNUSED) OVS_EXCLUDED(mutex)
950 struct ds ds = DS_EMPTY_INITIALIZER;
953 ovs_mutex_lock(&mutex);
955 bfd = bfd_find_by_name(argv[1]);
957 unixctl_command_reply_error(conn, "no such bfd object");
960 bfd_put_details(&ds, bfd);
962 HMAP_FOR_EACH (bfd, node, all_bfds) {
963 ds_put_format(&ds, "---- %s ----\n", bfd->name);
964 bfd_put_details(&ds, bfd);
967 unixctl_command_reply(conn, ds_cstr(&ds));
971 ovs_mutex_unlock(&mutex);
976 bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
977 const char *argv[], void *aux OVS_UNUSED)
980 const char *forward_str = argv[argc - 1];
981 int forwarding_override;
984 ovs_mutex_lock(&mutex);
985 if (!strcasecmp("true", forward_str)) {
986 forwarding_override = 1;
987 } else if (!strcasecmp("false", forward_str)) {
988 forwarding_override = 0;
989 } else if (!strcasecmp("normal", forward_str)) {
990 forwarding_override = -1;
992 unixctl_command_reply_error(conn, "unknown fault string");
997 bfd = bfd_find_by_name(argv[1]);
999 unixctl_command_reply_error(conn, "no such BFD object");
1002 bfd->forwarding_override = forwarding_override;
1004 HMAP_FOR_EACH (bfd, node, all_bfds) {
1005 bfd->forwarding_override = forwarding_override;
1009 unixctl_command_reply(conn, "OK");
1012 ovs_mutex_unlock(&mutex);