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 <sys/types.h>
19 #include <arpa/inet.h>
20 #include <netinet/in_systm.h>
21 #include <netinet/ip.h>
23 #include "byte-order.h"
26 #include "dynamic-string.h"
34 #include "ovs-thread.h"
35 #include "openvswitch/types.h"
37 #include "poll-loop.h"
45 VLOG_DEFINE_THIS_MODULE(bfd);
49 * The goal of this module is to replace CFM with something both more flexible
50 * and standards compliant. In service of this goal, the following needs to be
54 * * Implement Demand mode.
55 * * Go through the RFC line by line and verify we comply.
56 * * Test against a hardware implementation. Preferably a popular one.
57 * * Delete BFD packets with nw_ttl != 255 in the datapath to prevent DOS
62 * - Set TOS/PCP on the outer tunnel header when encapped.
64 * - Sending BFD messages should be in its own thread/process.
66 * - Scale testing. How does it operate when there are large number of bfd
67 * sessions? Do we ever have random flaps? What's the CPU utilization?
69 * - Rely on data traffic for liveness by using BFD demand mode.
70 * If we're receiving traffic on a port, we can safely assume it's up (modulo
71 * unidrectional failures). BFD has a demand mode in which it can stay quiet
72 * unless it feels the need to check the status of the port. Using this, we
73 * can implement a strategy in which BFD only sends control messages on dark
76 * - Depending on how one interprets the spec, it appears that a BFD session
77 * can never change bfd.LocalDiag to "No Diagnostic". We should verify that
78 * this is what hardware implementations actually do. Seems like "No
79 * Diagnostic" should be set once a BFD session state goes UP. */
84 FLAG_MULTIPOINT = 1 << 0,
93 STATE_ADMIN_DOWN = 0 << 6,
100 DIAG_NONE = 0, /* No Diagnostic. */
101 DIAG_EXPIRED = 1, /* Control Detection Time Expired. */
102 DIAG_ECHO_FAILED = 2, /* Echo Function Failed. */
103 DIAG_RMT_DOWN = 3, /* Neighbor Signaled Session Down. */
104 DIAG_FWD_RESET = 4, /* Forwarding Plane Reset. */
105 DIAG_PATH_DOWN = 5, /* Path Down. */
106 DIAG_CPATH_DOWN = 6, /* Concatenated Path Down. */
107 DIAG_ADMIN_DOWN = 7, /* Administratively Down. */
108 DIAG_RCPATH_DOWN = 8 /* Reverse Concatenated Path Down. */
111 /* RFC 5880 Section 4.1
113 * 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
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * |Vers | Diag |Sta|P|F|C|A|D|M| Detect Mult | Length |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * | My Discriminator |
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * | Your Discriminator |
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 * | Desired Min TX Interval |
122 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123 * | Required Min RX Interval |
124 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
125 * | Required Min Echo RX Interval |
126 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ */
128 uint8_t vers_diag; /* Version and diagnostic. */
129 uint8_t flags; /* 2bit State field followed by flags. */
130 uint8_t mult; /* Fault detection multiplier. */
131 uint8_t length; /* Length of this BFD message. */
132 ovs_be32 my_disc; /* My discriminator. */
133 ovs_be32 your_disc; /* Your discriminator. */
134 ovs_be32 min_tx; /* Desired minimum tx interval. */
135 ovs_be32 min_rx; /* Required minimum rx interval. */
136 ovs_be32 min_rx_echo; /* Required minimum echo rx interval. */
138 BUILD_ASSERT_DECL(BFD_PACKET_LEN == sizeof(struct msg));
140 #define DIAG_MASK 0x1f
142 #define STATE_MASK 0xC0
143 #define FLAGS_MASK 0x3f
146 struct hmap_node node; /* In 'all_bfds'. */
147 uint32_t disc; /* bfd.LocalDiscr. Key in 'all_bfds' hmap. */
149 char *name; /* Name used for logging. */
151 bool cpath_down; /* Concatenated Path Down. */
152 uint8_t mult; /* bfd.DetectMult. */
154 enum state state; /* bfd.SessionState. */
155 enum state rmt_state; /* bfd.RemoteSessionState. */
157 enum diag diag; /* bfd.LocalDiag. */
158 enum diag rmt_diag; /* Remote diagnostic. */
160 enum flags flags; /* Flags sent on messages. */
161 enum flags rmt_flags; /* Flags last received. */
163 uint32_t rmt_disc; /* bfd.RemoteDiscr. */
165 uint8_t eth_dst[ETH_ADDR_LEN];/* Ethernet destination address. */
166 bool eth_dst_set; /* 'eth_dst' set through database. */
168 uint16_t udp_src; /* UDP source port. */
170 /* All timers in milliseconds. */
171 long long int rmt_min_rx; /* bfd.RemoteMinRxInterval. */
172 long long int rmt_min_tx; /* Remote minimum TX interval. */
174 long long int cfg_min_tx; /* Configured minimum TX rate. */
175 long long int cfg_min_rx; /* Configured required minimum RX rate. */
176 long long int poll_min_tx; /* Min TX negotating in a poll sequence. */
177 long long int poll_min_rx; /* Min RX negotating in a poll sequence. */
178 long long int min_tx; /* bfd.DesiredMinTxInterval. */
179 long long int min_rx; /* bfd.RequiredMinRxInterval. */
181 long long int last_tx; /* Last TX time. */
182 long long int next_tx; /* Next TX time. */
183 long long int detect_time; /* RFC 5880 6.8.4 Detection time. */
185 int forwarding_override; /* Manual override of 'forwarding' status. */
187 atomic_bool check_tnl_key; /* Verify tunnel key of inbound packets? */
191 static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
192 static struct hmap all_bfds__ = HMAP_INITIALIZER(&all_bfds__);
193 static struct hmap *const all_bfds OVS_GUARDED_BY(mutex) = &all_bfds__;
195 static bool bfd_forwarding__(const struct bfd *) OVS_REQUIRES(mutex);
196 static bool bfd_in_poll(const struct bfd *) OVS_REQUIRES(mutex);
197 static void bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex);
198 static const char *bfd_diag_str(enum diag) OVS_REQUIRES(mutex);
199 static const char *bfd_state_str(enum state) OVS_REQUIRES(mutex);
200 static long long int bfd_min_tx(const struct bfd *) OVS_REQUIRES(mutex);
201 static long long int bfd_tx_interval(const struct bfd *)
203 static long long int bfd_rx_interval(const struct bfd *)
205 static void bfd_set_next_tx(struct bfd *) OVS_REQUIRES(mutex);
206 static void bfd_set_state(struct bfd *, enum state, enum diag)
208 static uint32_t generate_discriminator(void) OVS_REQUIRES(mutex);
209 static void bfd_put_details(struct ds *, const struct bfd *)
211 static void bfd_unixctl_show(struct unixctl_conn *, int argc,
212 const char *argv[], void *aux OVS_UNUSED);
213 static void bfd_unixctl_set_forwarding_override(struct unixctl_conn *,
214 int argc, const char *argv[],
215 void *aux OVS_UNUSED);
216 static void log_msg(enum vlog_level, const struct msg *, const char *message,
217 const struct bfd *) OVS_REQUIRES(mutex);
219 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(20, 20);
221 /* Returns true if the interface on which 'bfd' is running may be used to
222 * forward traffic according to the BFD session state. */
224 bfd_forwarding(const struct bfd *bfd) OVS_EXCLUDED(mutex)
228 ovs_mutex_lock(&mutex);
229 ret = bfd_forwarding__(bfd);
230 ovs_mutex_unlock(&mutex);
234 /* Returns a 'smap' of key value pairs representing the status of 'bfd'
235 * intended for the OVS database. */
237 bfd_get_status(const struct bfd *bfd, struct smap *smap)
240 ovs_mutex_lock(&mutex);
241 smap_add(smap, "forwarding", bfd_forwarding__(bfd)? "true" : "false");
242 smap_add(smap, "state", bfd_state_str(bfd->state));
243 smap_add(smap, "diagnostic", bfd_diag_str(bfd->diag));
245 if (bfd->state != STATE_DOWN) {
246 smap_add(smap, "remote_state", bfd_state_str(bfd->rmt_state));
247 smap_add(smap, "remote_diagnostic", bfd_diag_str(bfd->rmt_diag));
249 ovs_mutex_unlock(&mutex);
252 /* Initializes, destroys, or reconfigures the BFD session 'bfd' (named 'name'),
253 * according to the database configuration contained in 'cfg'. Takes ownership
254 * of 'bfd', which may be NULL. Returns a BFD object which may be used as a
255 * handle for the session, or NULL if BFD is not enabled according to 'cfg'.
256 * Also returns NULL if cfg is NULL. */
258 bfd_configure(struct bfd *bfd, const char *name, const struct smap *cfg)
261 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
262 static atomic_uint16_t udp_src = ATOMIC_VAR_INIT(0);
264 long long int min_tx, min_rx;
265 bool need_poll = false;
268 uint8_t ea[ETH_ADDR_LEN];
270 if (ovsthread_once_start(&once)) {
271 unixctl_command_register("bfd/show", "[interface]", 0, 1,
272 bfd_unixctl_show, NULL);
273 unixctl_command_register("bfd/set-forwarding",
274 "[interface] normal|false|true", 1, 2,
275 bfd_unixctl_set_forwarding_override, NULL);
276 ovsthread_once_done(&once);
279 if (!cfg || !smap_get_bool(cfg, "enable", false)) {
284 ovs_mutex_lock(&mutex);
286 bfd = xzalloc(sizeof *bfd);
287 bfd->name = xstrdup(name);
288 bfd->forwarding_override = -1;
289 bfd->disc = generate_discriminator();
290 hmap_insert(all_bfds, &bfd->node, bfd->disc);
292 bfd->diag = DIAG_NONE;
295 atomic_init(&bfd->ref_cnt, 1);
297 /* RFC 5881 section 4
298 * The source port MUST be in the range 49152 through 65535. The same
299 * UDP source port number MUST be used for all BFD Control packets
300 * associated with a particular session. The source port number SHOULD
301 * be unique among all BFD sessions on the system. */
302 atomic_add(&udp_src, 1, &bfd->udp_src);
303 bfd->udp_src = (bfd->udp_src % 16384) + 49152;
305 bfd_set_state(bfd, STATE_DOWN, DIAG_NONE);
307 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
310 atomic_store(&bfd->check_tnl_key,
311 smap_get_bool(cfg, "check_tnl_key", false));
312 min_tx = smap_get_int(cfg, "min_tx", 100);
313 min_tx = MAX(min_tx, 100);
314 if (bfd->cfg_min_tx != min_tx) {
315 bfd->cfg_min_tx = min_tx;
316 if (bfd->state != STATE_UP
317 || (!bfd_in_poll(bfd) && bfd->cfg_min_tx < bfd->min_tx)) {
318 bfd->min_tx = bfd->cfg_min_tx;
323 min_rx = smap_get_int(cfg, "min_rx", 1000);
324 min_rx = MAX(min_rx, 100);
325 if (bfd->cfg_min_rx != min_rx) {
326 bfd->cfg_min_rx = min_rx;
327 if (bfd->state != STATE_UP
328 || (!bfd_in_poll(bfd) && bfd->cfg_min_rx > bfd->min_rx)) {
329 bfd->min_rx = bfd->cfg_min_rx;
334 cpath_down = smap_get_bool(cfg, "cpath_down", false);
335 if (bfd->cpath_down != cpath_down) {
336 bfd->cpath_down = cpath_down;
337 if (bfd->diag == DIAG_NONE || bfd->diag == DIAG_CPATH_DOWN) {
338 bfd_set_state(bfd, bfd->state, DIAG_NONE);
343 hwaddr = smap_get(cfg, "bfd_dst_mac");
344 if (hwaddr && eth_addr_from_string(hwaddr, ea) && !eth_addr_is_zero(ea)) {
345 memcpy(bfd->eth_dst, ea, ETH_ADDR_LEN);
346 bfd->eth_dst_set = true;
347 } else if (bfd->eth_dst_set) {
348 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
349 bfd->eth_dst_set = false;
355 ovs_mutex_unlock(&mutex);
360 bfd_ref(const struct bfd *bfd_)
362 struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
365 atomic_add(&bfd->ref_cnt, 1, &orig);
366 ovs_assert(orig > 0);
372 bfd_unref(struct bfd *bfd) OVS_EXCLUDED(mutex)
377 atomic_sub(&bfd->ref_cnt, 1, &orig);
378 ovs_assert(orig > 0);
380 ovs_mutex_lock(&mutex);
381 hmap_remove(all_bfds, &bfd->node);
384 ovs_mutex_unlock(&mutex);
390 bfd_wait(const struct bfd *bfd) OVS_EXCLUDED(mutex)
392 ovs_mutex_lock(&mutex);
393 if (bfd->flags & FLAG_FINAL) {
394 poll_immediate_wake();
397 poll_timer_wait_until(bfd->next_tx);
398 if (bfd->state > STATE_DOWN) {
399 poll_timer_wait_until(bfd->detect_time);
401 ovs_mutex_unlock(&mutex);
405 bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
407 ovs_mutex_lock(&mutex);
408 if (bfd->state > STATE_DOWN && time_msec() >= bfd->detect_time) {
409 bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
412 if (bfd->min_tx != bfd->cfg_min_tx || bfd->min_rx != bfd->cfg_min_rx) {
415 ovs_mutex_unlock(&mutex);
419 bfd_should_send_packet(const struct bfd *bfd) OVS_EXCLUDED(mutex)
422 ovs_mutex_lock(&mutex);
423 ret = bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
424 ovs_mutex_unlock(&mutex);
429 bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
430 uint8_t eth_src[ETH_ADDR_LEN]) OVS_EXCLUDED(mutex)
432 long long int min_tx, min_rx;
433 struct udp_header *udp;
434 struct eth_header *eth;
435 struct ip_header *ip;
438 ovs_mutex_lock(&mutex);
440 long long int delay = time_msec() - bfd->next_tx;
441 long long int interval = bfd_tx_interval(bfd);
442 if (delay > interval * 3 / 2) {
443 VLOG_INFO("%s: long delay of %lldms (expected %lldms) sending BFD"
444 " control message", bfd->name, delay, interval);
448 /* RFC 5880 Section 6.5
449 * A BFD Control packet MUST NOT have both the Poll (P) and Final (F) bits
451 ovs_assert(!(bfd->flags & FLAG_POLL) || !(bfd->flags & FLAG_FINAL));
453 ofpbuf_reserve(p, 2); /* Properly align after the ethernet header. */
454 eth = ofpbuf_put_uninit(p, sizeof *eth);
455 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
456 memcpy(eth->eth_dst, bfd->eth_dst, ETH_ADDR_LEN);
457 eth->eth_type = htons(ETH_TYPE_IP);
459 ip = ofpbuf_put_zeros(p, sizeof *ip);
460 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
461 ip->ip_tot_len = htons(sizeof *ip + sizeof *udp + sizeof *msg);
463 ip->ip_tos = IPTOS_LOWDELAY | IPTOS_THROUGHPUT;
464 ip->ip_proto = IPPROTO_UDP;
465 ip->ip_src = htonl(0xA9FE0100); /* 169.254.1.0 Link Local. */
466 ip->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1 Link Local. */
467 ip->ip_csum = csum(ip, sizeof *ip);
469 udp = ofpbuf_put_zeros(p, sizeof *udp);
470 udp->udp_src = htons(bfd->udp_src);
471 udp->udp_dst = htons(BFD_DEST_PORT);
472 udp->udp_len = htons(sizeof *udp + sizeof *msg);
474 msg = ofpbuf_put_uninit(p, sizeof *msg);
475 msg->vers_diag = (BFD_VERSION << 5) | bfd->diag;
476 msg->flags = (bfd->state & STATE_MASK) | bfd->flags;
478 msg->mult = bfd->mult;
479 msg->length = BFD_PACKET_LEN;
480 msg->my_disc = htonl(bfd->disc);
481 msg->your_disc = htonl(bfd->rmt_disc);
482 msg->min_rx_echo = htonl(0);
484 if (bfd_in_poll(bfd)) {
485 min_tx = bfd->poll_min_tx;
486 min_rx = bfd->poll_min_rx;
488 min_tx = bfd_min_tx(bfd);
489 min_rx = bfd->min_rx;
492 msg->min_tx = htonl(min_tx * 1000);
493 msg->min_rx = htonl(min_rx * 1000);
495 bfd->flags &= ~FLAG_FINAL;
497 log_msg(VLL_DBG, msg, "Sending BFD Message", bfd);
499 bfd->last_tx = time_msec();
500 bfd_set_next_tx(bfd);
501 ovs_mutex_unlock(&mutex);
505 bfd_should_process_flow(const struct bfd *bfd, const struct flow *flow,
506 struct flow_wildcards *wc)
509 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
510 if (bfd->eth_dst_set && memcmp(bfd->eth_dst, flow->dl_dst, ETH_ADDR_LEN)) {
514 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
515 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
517 atomic_read(&bfd->check_tnl_key, &check_tnl_key);
519 memset(&wc->masks.tunnel.tun_id, 0xff, sizeof wc->masks.tunnel.tun_id);
521 return (flow->dl_type == htons(ETH_TYPE_IP)
522 && flow->nw_proto == IPPROTO_UDP
523 && flow->tp_dst == htons(BFD_DEST_PORT)
524 && (!check_tnl_key || flow->tunnel.tun_id == htonll(0)));
528 bfd_process_packet(struct bfd *bfd, const struct flow *flow,
529 const struct ofpbuf *p) OVS_EXCLUDED(mutex)
531 uint32_t rmt_min_rx, pkt_your_disc;
532 enum state rmt_state;
537 /* This function is designed to follow section RFC 5880 6.8.6 closely. */
539 ovs_mutex_lock(&mutex);
540 if (flow->nw_ttl != 255) {
541 /* XXX Should drop in the kernel to prevent DOS. */
545 msg = ofpbuf_at(p, (uint8_t *)p->l7 - (uint8_t *)p->data, BFD_PACKET_LEN);
547 VLOG_INFO_RL(&rl, "%s: Received unparseable BFD control message.",
552 /* RFC 5880 Section 6.8.6
553 * If the Length field is greater than the payload of the encapsulating
554 * protocol, the packet MUST be discarded.
556 * Note that we make this check implicity. Above we use ofpbuf_at() to
557 * ensure that there are at least BFD_PACKET_LEN bytes in the payload of
558 * the encapsulating protocol. Below we require msg->length to be exactly
559 * BFD_PACKET_LEN bytes. */
561 flags = msg->flags & FLAGS_MASK;
562 rmt_state = msg->flags & STATE_MASK;
563 version = msg->vers_diag >> VERS_SHIFT;
565 log_msg(VLL_DBG, msg, "Received BFD control message", bfd);
567 if (version != BFD_VERSION) {
568 log_msg(VLL_WARN, msg, "Incorrect version", bfd);
572 /* Technically this should happen after the length check. We don't support
573 * authentication however, so it's simpler to do the check first. */
574 if (flags & FLAG_AUTH) {
575 log_msg(VLL_WARN, msg, "Authenticated control message with"
576 " authentication disabled", bfd);
580 if (msg->length != BFD_PACKET_LEN) {
581 log_msg(VLL_WARN, msg, "Unexpected length", bfd);
582 if (msg->length < BFD_PACKET_LEN) {
588 log_msg(VLL_WARN, msg, "Zero multiplier", bfd);
592 if (flags & FLAG_MULTIPOINT) {
593 log_msg(VLL_WARN, msg, "Unsupported multipoint flag", bfd);
598 log_msg(VLL_WARN, msg, "NULL my_disc", bfd);
602 pkt_your_disc = ntohl(msg->your_disc);
604 /* Technically, we should use the your discriminator field to figure
605 * out which 'struct bfd' this packet is destined towards. That way a
606 * bfd session could migrate from one interface to another
607 * transparently. This doesn't fit in with the OVS structure very
608 * well, so in this respect, we are not compliant. */
609 if (pkt_your_disc != bfd->disc) {
610 log_msg(VLL_WARN, msg, "Incorrect your_disc", bfd);
613 } else if (rmt_state > STATE_DOWN) {
614 log_msg(VLL_WARN, msg, "Null your_disc", bfd);
618 bfd->rmt_disc = ntohl(msg->my_disc);
619 bfd->rmt_state = rmt_state;
620 bfd->rmt_flags = flags;
621 bfd->rmt_diag = msg->vers_diag & DIAG_MASK;
623 if (flags & FLAG_FINAL && bfd_in_poll(bfd)) {
624 bfd->min_tx = bfd->poll_min_tx;
625 bfd->min_rx = bfd->poll_min_rx;
626 bfd->flags &= ~FLAG_POLL;
627 log_msg(VLL_INFO, msg, "Poll sequence terminated", bfd);
630 if (flags & FLAG_POLL) {
631 /* RFC 5880 Section 6.5
632 * When the other system receives a Poll, it immediately transmits a
633 * BFD Control packet with the Final (F) bit set, independent of any
634 * periodic BFD Control packets it may be sending
635 * (see section 6.8.7). */
636 bfd->flags &= ~FLAG_POLL;
637 bfd->flags |= FLAG_FINAL;
640 rmt_min_rx = MAX(ntohl(msg->min_rx) / 1000, 1);
641 if (bfd->rmt_min_rx != rmt_min_rx) {
642 bfd->rmt_min_rx = rmt_min_rx;
643 bfd_set_next_tx(bfd);
644 log_msg(VLL_INFO, msg, "New remote min_rx", bfd);
647 bfd->rmt_min_tx = MAX(ntohl(msg->min_tx) / 1000, 1);
648 bfd->detect_time = bfd_rx_interval(bfd) * bfd->mult + time_msec();
650 if (bfd->state == STATE_ADMIN_DOWN) {
651 VLOG_DBG_RL(&rl, "Administratively down, dropping control message.");
655 if (rmt_state == STATE_ADMIN_DOWN) {
656 if (bfd->state != STATE_DOWN) {
657 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
660 switch (bfd->state) {
662 if (rmt_state == STATE_DOWN) {
663 bfd_set_state(bfd, STATE_INIT, bfd->diag);
664 } else if (rmt_state == STATE_INIT) {
665 bfd_set_state(bfd, STATE_UP, bfd->diag);
669 if (rmt_state > STATE_DOWN) {
670 bfd_set_state(bfd, STATE_UP, bfd->diag);
674 if (rmt_state <= STATE_DOWN) {
675 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
676 log_msg(VLL_INFO, msg, "Remote signaled STATE_DOWN", bfd);
679 case STATE_ADMIN_DOWN:
684 /* XXX: RFC 5880 Section 6.8.6 Demand mode related calculations here. */
687 ovs_mutex_unlock(&mutex);
691 bfd_forwarding__(const struct bfd *bfd) OVS_REQUIRES(mutex)
693 if (bfd->forwarding_override != -1) {
694 return bfd->forwarding_override == 1;
697 return bfd->state == STATE_UP
698 && bfd->rmt_diag != DIAG_PATH_DOWN
699 && bfd->rmt_diag != DIAG_CPATH_DOWN
700 && bfd->rmt_diag != DIAG_RCPATH_DOWN;
705 bfd_in_poll(const struct bfd *bfd) OVS_REQUIRES(mutex)
707 return (bfd->flags & FLAG_POLL) != 0;
711 bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex)
713 if (bfd->state > STATE_DOWN && !bfd_in_poll(bfd)
714 && !(bfd->flags & FLAG_FINAL)) {
715 bfd->poll_min_tx = bfd->cfg_min_tx;
716 bfd->poll_min_rx = bfd->cfg_min_rx;
717 bfd->flags |= FLAG_POLL;
719 VLOG_INFO_RL(&rl, "%s: Initiating poll sequence", bfd->name);
724 bfd_min_tx(const struct bfd *bfd) OVS_REQUIRES(mutex)
726 /* RFC 5880 Section 6.8.3
727 * When bfd.SessionState is not Up, the system MUST set
728 * bfd.DesiredMinTxInterval to a value of not less than one second
729 * (1,000,000 microseconds). This is intended to ensure that the
730 * bandwidth consumed by BFD sessions that are not Up is negligible,
731 * particularly in the case where a neighbor may not be running BFD. */
732 return (bfd->state == STATE_UP ? bfd->min_tx : MAX(bfd->min_tx, 1000));
736 bfd_tx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
738 long long int interval = bfd_min_tx(bfd);
739 return MAX(interval, bfd->rmt_min_rx);
743 bfd_rx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
745 return MAX(bfd->min_rx, bfd->rmt_min_tx);
749 bfd_set_next_tx(struct bfd *bfd) OVS_REQUIRES(mutex)
751 long long int interval = bfd_tx_interval(bfd);
752 interval -= interval * random_range(26) / 100;
753 bfd->next_tx = bfd->last_tx + interval;
757 bfd_flag_str(enum flags flags)
759 struct ds ds = DS_EMPTY_INITIALIZER;
760 static char flag_str[128];
766 if (flags & FLAG_MULTIPOINT) {
767 ds_put_cstr(&ds, "multipoint ");
770 if (flags & FLAG_DEMAND) {
771 ds_put_cstr(&ds, "demand ");
774 if (flags & FLAG_AUTH) {
775 ds_put_cstr(&ds, "auth ");
778 if (flags & FLAG_CTL) {
779 ds_put_cstr(&ds, "ctl ");
782 if (flags & FLAG_FINAL) {
783 ds_put_cstr(&ds, "final ");
786 if (flags & FLAG_POLL) {
787 ds_put_cstr(&ds, "poll ");
790 /* Do not copy the trailing whitespace. */
792 ovs_strlcpy(flag_str, ds_cstr(&ds), sizeof flag_str);
798 bfd_state_str(enum state state)
801 case STATE_ADMIN_DOWN: return "admin_down";
802 case STATE_DOWN: return "down";
803 case STATE_INIT: return "init";
804 case STATE_UP: return "up";
805 default: return "invalid";
810 bfd_diag_str(enum diag diag) {
812 case DIAG_NONE: return "No Diagnostic";
813 case DIAG_EXPIRED: return "Control Detection Time Expired";
814 case DIAG_ECHO_FAILED: return "Echo Function Failed";
815 case DIAG_RMT_DOWN: return "Neighbor Signaled Session Down";
816 case DIAG_FWD_RESET: return "Forwarding Plane Reset";
817 case DIAG_PATH_DOWN: return "Path Down";
818 case DIAG_CPATH_DOWN: return "Concatenated Path Down";
819 case DIAG_ADMIN_DOWN: return "Administratively Down";
820 case DIAG_RCPATH_DOWN: return "Reverse Concatenated Path Down";
821 default: return "Invalid Diagnostic";
826 log_msg(enum vlog_level level, const struct msg *p, const char *message,
827 const struct bfd *bfd) OVS_REQUIRES(mutex)
829 struct ds ds = DS_EMPTY_INITIALIZER;
831 if (vlog_should_drop(THIS_MODULE, level, &rl)) {
837 "\n\tvers:%"PRIu8" diag:\"%s\" state:%s mult:%"PRIu8
840 "\n\tmy_disc:0x%"PRIx32" your_disc:0x%"PRIx32
841 "\n\tmin_tx:%"PRIu32"us (%"PRIu32"ms)"
842 "\n\tmin_rx:%"PRIu32"us (%"PRIu32"ms)"
843 "\n\tmin_rx_echo:%"PRIu32"us (%"PRIu32"ms)",
844 bfd->name, message, p->vers_diag >> VERS_SHIFT,
845 bfd_diag_str(p->vers_diag & DIAG_MASK),
846 bfd_state_str(p->flags & STATE_MASK),
847 p->mult, p->length, bfd_flag_str(p->flags & FLAGS_MASK),
848 ntohl(p->my_disc), ntohl(p->your_disc),
849 ntohl(p->min_tx), ntohl(p->min_tx) / 1000,
850 ntohl(p->min_rx), ntohl(p->min_rx) / 1000,
851 ntohl(p->min_rx_echo), ntohl(p->min_rx_echo) / 1000);
852 bfd_put_details(&ds, bfd);
853 VLOG(level, "%s", ds_cstr(&ds));
858 bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
861 if (diag == DIAG_NONE && bfd->cpath_down) {
862 diag = DIAG_CPATH_DOWN;
865 if (bfd->state != state || bfd->diag != diag) {
866 if (!VLOG_DROP_INFO(&rl)) {
867 struct ds ds = DS_EMPTY_INITIALIZER;
869 ds_put_format(&ds, "%s: BFD state change: %s->%s"
870 " \"%s\"->\"%s\".\n",
871 bfd->name, bfd_state_str(bfd->state),
872 bfd_state_str(state), bfd_diag_str(bfd->diag),
874 bfd_put_details(&ds, bfd);
875 VLOG_INFO("%s", ds_cstr(&ds));
882 if (bfd->state <= STATE_DOWN) {
883 bfd->rmt_state = STATE_DOWN;
884 bfd->rmt_diag = DIAG_NONE;
894 generate_discriminator(void)
898 /* RFC 5880 Section 6.8.1
899 * It SHOULD be set to a random (but still unique) value to improve
900 * security. The value is otherwise outside the scope of this
906 /* 'disc' is by definition random, so there's no reason to waste time
908 disc = random_uint32();
909 HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, all_bfds) {
910 if (bfd->disc == disc) {
921 bfd_find_by_name(const char *name) OVS_REQUIRES(mutex)
925 HMAP_FOR_EACH (bfd, node, all_bfds) {
926 if (!strcmp(bfd->name, name)) {
934 bfd_put_details(struct ds *ds, const struct bfd *bfd) OVS_REQUIRES(mutex)
936 ds_put_format(ds, "\tForwarding: %s\n",
937 bfd_forwarding__(bfd) ? "true" : "false");
938 ds_put_format(ds, "\tDetect Multiplier: %d\n", bfd->mult);
939 ds_put_format(ds, "\tConcatenated Path Down: %s\n",
940 bfd->cpath_down ? "true" : "false");
941 ds_put_format(ds, "\tTX Interval: Approx %lldms\n", bfd_tx_interval(bfd));
942 ds_put_format(ds, "\tRX Interval: Approx %lldms\n", bfd_rx_interval(bfd));
943 ds_put_format(ds, "\tDetect Time: now %+lldms\n",
944 time_msec() - bfd->detect_time);
945 ds_put_format(ds, "\tNext TX Time: now %+lldms\n",
946 time_msec() - bfd->next_tx);
947 ds_put_format(ds, "\tLast TX Time: now %+lldms\n",
948 time_msec() - bfd->last_tx);
950 ds_put_cstr(ds, "\n");
952 ds_put_format(ds, "\tLocal Flags: %s\n", bfd_flag_str(bfd->flags));
953 ds_put_format(ds, "\tLocal Session State: %s\n",
954 bfd_state_str(bfd->state));
955 ds_put_format(ds, "\tLocal Diagnostic: %s\n", bfd_diag_str(bfd->diag));
956 ds_put_format(ds, "\tLocal Discriminator: 0x%"PRIx32"\n", bfd->disc);
957 ds_put_format(ds, "\tLocal Minimum TX Interval: %lldms\n",
959 ds_put_format(ds, "\tLocal Minimum RX Interval: %lldms\n", bfd->min_rx);
961 ds_put_cstr(ds, "\n");
963 ds_put_format(ds, "\tRemote Flags: %s\n", bfd_flag_str(bfd->rmt_flags));
964 ds_put_format(ds, "\tRemote Session State: %s\n",
965 bfd_state_str(bfd->rmt_state));
966 ds_put_format(ds, "\tRemote Diagnostic: %s\n",
967 bfd_diag_str(bfd->rmt_diag));
968 ds_put_format(ds, "\tRemote Discriminator: 0x%"PRIx32"\n", bfd->rmt_disc);
969 ds_put_format(ds, "\tRemote Minimum TX Interval: %lldms\n",
971 ds_put_format(ds, "\tRemote Minimum RX Interval: %lldms\n",
976 bfd_unixctl_show(struct unixctl_conn *conn, int argc, const char *argv[],
977 void *aux OVS_UNUSED) OVS_EXCLUDED(mutex)
979 struct ds ds = DS_EMPTY_INITIALIZER;
982 ovs_mutex_lock(&mutex);
984 bfd = bfd_find_by_name(argv[1]);
986 unixctl_command_reply_error(conn, "no such bfd object");
989 bfd_put_details(&ds, bfd);
991 HMAP_FOR_EACH (bfd, node, all_bfds) {
992 ds_put_format(&ds, "---- %s ----\n", bfd->name);
993 bfd_put_details(&ds, bfd);
996 unixctl_command_reply(conn, ds_cstr(&ds));
1000 ovs_mutex_unlock(&mutex);
1005 bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
1006 const char *argv[], void *aux OVS_UNUSED)
1009 const char *forward_str = argv[argc - 1];
1010 int forwarding_override;
1013 ovs_mutex_lock(&mutex);
1014 if (!strcasecmp("true", forward_str)) {
1015 forwarding_override = 1;
1016 } else if (!strcasecmp("false", forward_str)) {
1017 forwarding_override = 0;
1018 } else if (!strcasecmp("normal", forward_str)) {
1019 forwarding_override = -1;
1021 unixctl_command_reply_error(conn, "unknown fault string");
1026 bfd = bfd_find_by_name(argv[1]);
1028 unixctl_command_reply_error(conn, "no such BFD object");
1031 bfd->forwarding_override = forwarding_override;
1033 HMAP_FOR_EACH (bfd, node, all_bfds) {
1034 bfd->forwarding_override = forwarding_override;
1038 unixctl_command_reply(conn, "OK");
1041 ovs_mutex_unlock(&mutex);