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>
19 #include <netinet/ip.h>
21 #include "byte-order.h"
24 #include "dynamic-string.h"
32 #include "ovs-thread.h"
33 #include "openvswitch/types.h"
35 #include "poll-loop.h"
43 VLOG_DEFINE_THIS_MODULE(bfd);
47 * The goal of this module is to replace CFM with something both more flexible
48 * and standards compliant. In service of this goal, the following needs to be
52 * * Implement Demand mode.
53 * * Go through the RFC line by line and verify we comply.
54 * * Test against a hardware implementation. Preferably a popular one.
55 * * Delete BFD packets with nw_ttl != 255 in the datapath to prevent DOS
60 * - Set TOS/PCP on the outer tunnel header when encapped.
62 * - Sending BFD messages should be in its own thread/process.
64 * - Scale testing. How does it operate when there are large number of bfd
65 * sessions? Do we ever have random flaps? What's the CPU utilization?
67 * - Rely on data traffic for liveness by using BFD demand mode.
68 * If we're receiving traffic on a port, we can safely assume it's up (modulo
69 * unidrectional failures). BFD has a demand mode in which it can stay quiet
70 * unless it feels the need to check the status of the port. Using this, we
71 * can implement a strategy in which BFD only sends control messages on dark
74 * - Depending on how one interprets the spec, it appears that a BFD session
75 * can never change bfd.LocalDiag to "No Diagnostic". We should verify that
76 * this is what hardware implementations actually do. Seems like "No
77 * Diagnostic" should be set once a BFD session state goes UP. */
82 FLAG_MULTIPOINT = 1 << 0,
91 STATE_ADMIN_DOWN = 0 << 6,
98 DIAG_NONE = 0, /* No Diagnostic. */
99 DIAG_EXPIRED = 1, /* Control Detection Time Expired. */
100 DIAG_ECHO_FAILED = 2, /* Echo Function Failed. */
101 DIAG_RMT_DOWN = 3, /* Neighbor Signaled Session Down. */
102 DIAG_FWD_RESET = 4, /* Forwarding Plane Reset. */
103 DIAG_PATH_DOWN = 5, /* Path Down. */
104 DIAG_CPATH_DOWN = 6, /* Concatenated Path Down. */
105 DIAG_ADMIN_DOWN = 7, /* Administratively Down. */
106 DIAG_RCPATH_DOWN = 8 /* Reverse Concatenated Path Down. */
109 /* RFC 5880 Section 4.1
111 * 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
112 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
113 * |Vers | Diag |Sta|P|F|C|A|D|M| Detect Mult | Length |
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * | My Discriminator |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 * | Your Discriminator |
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * | Desired Min TX Interval |
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 * | Required Min RX Interval |
122 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
123 * | Required Min Echo RX Interval |
124 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ */
126 uint8_t vers_diag; /* Version and diagnostic. */
127 uint8_t flags; /* 2bit State field followed by flags. */
128 uint8_t mult; /* Fault detection multiplier. */
129 uint8_t length; /* Length of this BFD message. */
130 ovs_be32 my_disc; /* My discriminator. */
131 ovs_be32 your_disc; /* Your discriminator. */
132 ovs_be32 min_tx; /* Desired minimum tx interval. */
133 ovs_be32 min_rx; /* Required minimum rx interval. */
134 ovs_be32 min_rx_echo; /* Required minimum echo rx interval. */
136 BUILD_ASSERT_DECL(BFD_PACKET_LEN == sizeof(struct msg));
138 #define DIAG_MASK 0x1f
140 #define STATE_MASK 0xC0
141 #define FLAGS_MASK 0x3f
144 struct hmap_node node; /* In 'all_bfds'. */
145 uint32_t disc; /* bfd.LocalDiscr. Key in 'all_bfds' hmap. */
147 char *name; /* Name used for logging. */
149 bool cpath_down; /* Concatenated Path Down. */
150 uint8_t mult; /* bfd.DetectMult. */
152 enum state state; /* bfd.SessionState. */
153 enum state rmt_state; /* bfd.RemoteSessionState. */
155 enum diag diag; /* bfd.LocalDiag. */
156 enum diag rmt_diag; /* Remote diagnostic. */
158 enum flags flags; /* Flags sent on messages. */
159 enum flags rmt_flags; /* Flags last received. */
161 uint32_t rmt_disc; /* bfd.RemoteDiscr. */
163 uint8_t eth_dst[ETH_ADDR_LEN];/* Ethernet destination address. */
164 bool eth_dst_set; /* 'eth_dst' set through database. */
166 uint16_t udp_src; /* UDP source port. */
168 /* All timers in milliseconds. */
169 long long int rmt_min_rx; /* bfd.RemoteMinRxInterval. */
170 long long int rmt_min_tx; /* Remote minimum TX interval. */
172 long long int cfg_min_tx; /* Configured minimum TX rate. */
173 long long int cfg_min_rx; /* Configured required minimum RX rate. */
174 long long int poll_min_tx; /* Min TX negotating in a poll sequence. */
175 long long int poll_min_rx; /* Min RX negotating in a poll sequence. */
176 long long int min_tx; /* bfd.DesiredMinTxInterval. */
177 long long int min_rx; /* bfd.RequiredMinRxInterval. */
179 long long int last_tx; /* Last TX time. */
180 long long int next_tx; /* Next TX time. */
181 long long int detect_time; /* RFC 5880 6.8.4 Detection time. */
183 int forwarding_override; /* Manual override of 'forwarding' status. */
185 atomic_bool check_tnl_key; /* Verify tunnel key of inbound packets? */
189 static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
190 static struct hmap all_bfds__ = HMAP_INITIALIZER(&all_bfds__);
191 static struct hmap *const all_bfds OVS_GUARDED_BY(mutex) = &all_bfds__;
193 static bool bfd_forwarding__(const struct bfd *) OVS_REQ_WRLOCK(mutex);
194 static bool bfd_in_poll(const struct bfd *) OVS_REQ_WRLOCK(&mutex);
195 static void bfd_poll(struct bfd *bfd) OVS_REQ_WRLOCK(&mutex);
196 static const char *bfd_diag_str(enum diag) OVS_REQ_WRLOCK(&mutex);
197 static const char *bfd_state_str(enum state) OVS_REQ_WRLOCK(&mutex);
198 static long long int bfd_min_tx(const struct bfd *) OVS_REQ_WRLOCK(&mutex);
199 static long long int bfd_tx_interval(const struct bfd *)
200 OVS_REQ_WRLOCK(&mutex);
201 static long long int bfd_rx_interval(const struct bfd *)
202 OVS_REQ_WRLOCK(&mutex);
203 static void bfd_set_next_tx(struct bfd *) OVS_REQ_WRLOCK(&mutex);
204 static void bfd_set_state(struct bfd *, enum state, enum diag)
205 OVS_REQ_WRLOCK(&mutex);
206 static uint32_t generate_discriminator(void) OVS_REQ_WRLOCK(&mutex);
207 static void bfd_put_details(struct ds *, const struct bfd *)
208 OVS_REQ_WRLOCK(&mutex);
209 static void bfd_unixctl_show(struct unixctl_conn *, int argc,
210 const char *argv[], void *aux OVS_UNUSED);
211 static void bfd_unixctl_set_forwarding_override(struct unixctl_conn *,
212 int argc, const char *argv[],
213 void *aux OVS_UNUSED);
214 static void log_msg(enum vlog_level, const struct msg *, const char *message,
215 const struct bfd *) OVS_REQ_WRLOCK(&mutex);
217 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(20, 20);
219 /* Returns true if the interface on which 'bfd' is running may be used to
220 * forward traffic according to the BFD session state. */
222 bfd_forwarding(const struct bfd *bfd) OVS_EXCLUDED(mutex)
226 ovs_mutex_lock(&mutex);
227 ret = bfd_forwarding__(bfd);
228 ovs_mutex_unlock(&mutex);
232 /* Returns a 'smap' of key value pairs representing the status of 'bfd'
233 * intended for the OVS database. */
235 bfd_get_status(const struct bfd *bfd, struct smap *smap)
238 ovs_mutex_lock(&mutex);
239 smap_add(smap, "forwarding", bfd_forwarding__(bfd)? "true" : "false");
240 smap_add(smap, "state", bfd_state_str(bfd->state));
241 smap_add(smap, "diagnostic", bfd_diag_str(bfd->diag));
243 if (bfd->state != STATE_DOWN) {
244 smap_add(smap, "remote_state", bfd_state_str(bfd->rmt_state));
245 smap_add(smap, "remote_diagnostic", bfd_diag_str(bfd->rmt_diag));
247 ovs_mutex_unlock(&mutex);
250 /* Initializes, destroys, or reconfigures the BFD session 'bfd' (named 'name'),
251 * according to the database configuration contained in 'cfg'. Takes ownership
252 * of 'bfd', which may be NULL. Returns a BFD object which may be used as a
253 * handle for the session, or NULL if BFD is not enabled according to 'cfg'.
254 * Also returns NULL if cfg is NULL. */
256 bfd_configure(struct bfd *bfd, const char *name, const struct smap *cfg)
259 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
260 static atomic_uint16_t udp_src = ATOMIC_VAR_INIT(0);
262 long long int min_tx, min_rx;
265 uint8_t ea[ETH_ADDR_LEN];
267 if (ovsthread_once_start(&once)) {
268 unixctl_command_register("bfd/show", "[interface]", 0, 1,
269 bfd_unixctl_show, NULL);
270 unixctl_command_register("bfd/set-forwarding",
271 "[interface] normal|false|true", 1, 2,
272 bfd_unixctl_set_forwarding_override, NULL);
273 ovsthread_once_done(&once);
276 if (!cfg || !smap_get_bool(cfg, "enable", false)) {
281 ovs_mutex_lock(&mutex);
283 bfd = xzalloc(sizeof *bfd);
284 bfd->name = xstrdup(name);
285 bfd->forwarding_override = -1;
286 bfd->disc = generate_discriminator();
287 hmap_insert(all_bfds, &bfd->node, bfd->disc);
289 bfd->diag = DIAG_NONE;
292 atomic_init(&bfd->ref_cnt, 1);
294 /* RFC 5881 section 4
295 * The source port MUST be in the range 49152 through 65535. The same
296 * UDP source port number MUST be used for all BFD Control packets
297 * associated with a particular session. The source port number SHOULD
298 * be unique among all BFD sessions on the system. */
299 atomic_add(&udp_src, 1, &bfd->udp_src);
300 bfd->udp_src = (bfd->udp_src % 16384) + 49152;
302 bfd_set_state(bfd, STATE_DOWN, DIAG_NONE);
304 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
307 atomic_store(&bfd->check_tnl_key,
308 smap_get_bool(cfg, "check_tnl_key", false));
309 min_tx = smap_get_int(cfg, "min_tx", 100);
310 min_tx = MAX(min_tx, 100);
311 if (bfd->cfg_min_tx != min_tx) {
312 bfd->cfg_min_tx = min_tx;
313 if (bfd->state != STATE_UP
314 || (!bfd_in_poll(bfd) && bfd->cfg_min_tx < bfd->min_tx)) {
315 bfd->min_tx = bfd->cfg_min_tx;
320 min_rx = smap_get_int(cfg, "min_rx", 1000);
321 min_rx = MAX(min_rx, 100);
322 if (bfd->cfg_min_rx != min_rx) {
323 bfd->cfg_min_rx = min_rx;
324 if (bfd->state != STATE_UP
325 || (!bfd_in_poll(bfd) && bfd->cfg_min_rx > bfd->min_rx)) {
326 bfd->min_rx = bfd->cfg_min_rx;
331 cpath_down = smap_get_bool(cfg, "cpath_down", false);
332 if (bfd->cpath_down != cpath_down) {
333 bfd->cpath_down = cpath_down;
334 if (bfd->diag == DIAG_NONE || bfd->diag == DIAG_CPATH_DOWN) {
335 bfd_set_state(bfd, bfd->state, DIAG_NONE);
340 hwaddr = smap_get(cfg, "bfd_dst_mac");
341 if (hwaddr && eth_addr_from_string(hwaddr, ea) && !eth_addr_is_zero(ea)) {
342 memcpy(bfd->eth_dst, ea, ETH_ADDR_LEN);
343 bfd->eth_dst_set = true;
344 } else if (bfd->eth_dst_set) {
345 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
346 bfd->eth_dst_set = false;
349 ovs_mutex_unlock(&mutex);
354 bfd_ref(const struct bfd *bfd_)
356 struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
359 atomic_add(&bfd->ref_cnt, 1, &orig);
360 ovs_assert(orig > 0);
366 bfd_unref(struct bfd *bfd) OVS_EXCLUDED(mutex)
371 atomic_sub(&bfd->ref_cnt, 1, &orig);
372 ovs_assert(orig > 0);
374 ovs_mutex_lock(&mutex);
375 hmap_remove(all_bfds, &bfd->node);
378 ovs_mutex_unlock(&mutex);
384 bfd_wait(const struct bfd *bfd) OVS_EXCLUDED(mutex)
386 ovs_mutex_lock(&mutex);
387 if (bfd->flags & FLAG_FINAL) {
388 poll_immediate_wake();
391 poll_timer_wait_until(bfd->next_tx);
392 if (bfd->state > STATE_DOWN) {
393 poll_timer_wait_until(bfd->detect_time);
395 ovs_mutex_unlock(&mutex);
399 bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
401 ovs_mutex_lock(&mutex);
402 if (bfd->state > STATE_DOWN && time_msec() >= bfd->detect_time) {
403 bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
406 if (bfd->min_tx != bfd->cfg_min_tx || bfd->min_rx != bfd->cfg_min_rx) {
409 ovs_mutex_unlock(&mutex);
413 bfd_should_send_packet(const struct bfd *bfd) OVS_EXCLUDED(mutex)
416 ovs_mutex_lock(&mutex);
417 ret = bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
418 ovs_mutex_unlock(&mutex);
423 bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
424 uint8_t eth_src[ETH_ADDR_LEN]) OVS_EXCLUDED(mutex)
426 long long int min_tx, min_rx;
427 struct udp_header *udp;
428 struct eth_header *eth;
429 struct ip_header *ip;
432 ovs_mutex_lock(&mutex);
434 long long int delay = time_msec() - bfd->next_tx;
435 long long int interval = bfd_tx_interval(bfd);
436 if (delay > interval * 3 / 2) {
437 VLOG_INFO("%s: long delay of %lldms (expected %lldms) sending BFD"
438 " control message", bfd->name, delay, interval);
442 /* RFC 5880 Section 6.5
443 * A BFD Control packet MUST NOT have both the Poll (P) and Final (F) bits
445 ovs_assert(!(bfd->flags & FLAG_POLL) || !(bfd->flags & FLAG_FINAL));
447 ofpbuf_reserve(p, 2); /* Properly align after the ethernet header. */
448 eth = ofpbuf_put_uninit(p, sizeof *eth);
449 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
450 memcpy(eth->eth_dst, bfd->eth_dst, ETH_ADDR_LEN);
451 eth->eth_type = htons(ETH_TYPE_IP);
453 ip = ofpbuf_put_zeros(p, sizeof *ip);
454 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
455 ip->ip_tot_len = htons(sizeof *ip + sizeof *udp + sizeof *msg);
457 ip->ip_tos = IPTOS_LOWDELAY | IPTOS_THROUGHPUT;
458 ip->ip_proto = IPPROTO_UDP;
459 ip->ip_src = htonl(0xA9FE0100); /* 169.254.1.0 Link Local. */
460 ip->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1 Link Local. */
461 ip->ip_csum = csum(ip, sizeof *ip);
463 udp = ofpbuf_put_zeros(p, sizeof *udp);
464 udp->udp_src = htons(bfd->udp_src);
465 udp->udp_dst = htons(BFD_DEST_PORT);
466 udp->udp_len = htons(sizeof *udp + sizeof *msg);
468 msg = ofpbuf_put_uninit(p, sizeof *msg);
469 msg->vers_diag = (BFD_VERSION << 5) | bfd->diag;
470 msg->flags = (bfd->state & STATE_MASK) | bfd->flags;
472 msg->mult = bfd->mult;
473 msg->length = BFD_PACKET_LEN;
474 msg->my_disc = htonl(bfd->disc);
475 msg->your_disc = htonl(bfd->rmt_disc);
476 msg->min_rx_echo = htonl(0);
478 if (bfd_in_poll(bfd)) {
479 min_tx = bfd->poll_min_tx;
480 min_rx = bfd->poll_min_rx;
482 min_tx = bfd_min_tx(bfd);
483 min_rx = bfd->min_rx;
486 msg->min_tx = htonl(min_tx * 1000);
487 msg->min_rx = htonl(min_rx * 1000);
489 bfd->flags &= ~FLAG_FINAL;
491 log_msg(VLL_DBG, msg, "Sending BFD Message", bfd);
493 bfd->last_tx = time_msec();
494 bfd_set_next_tx(bfd);
495 ovs_mutex_unlock(&mutex);
499 bfd_should_process_flow(const struct bfd *bfd, const struct flow *flow,
500 struct flow_wildcards *wc)
503 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
504 if (bfd->eth_dst_set && memcmp(bfd->eth_dst, flow->dl_dst, ETH_ADDR_LEN)) {
508 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
509 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
511 atomic_read(&bfd->check_tnl_key, &check_tnl_key);
513 memset(&wc->masks.tunnel.tun_id, 0xff, sizeof wc->masks.tunnel.tun_id);
515 return (flow->dl_type == htons(ETH_TYPE_IP)
516 && flow->nw_proto == IPPROTO_UDP
517 && flow->tp_dst == htons(BFD_DEST_PORT)
518 && (check_tnl_key || flow->tunnel.tun_id == htonll(0)));
522 bfd_process_packet(struct bfd *bfd, const struct flow *flow,
523 const struct ofpbuf *p) OVS_EXCLUDED(mutex)
525 uint32_t rmt_min_rx, pkt_your_disc;
526 enum state rmt_state;
531 /* This function is designed to follow section RFC 5880 6.8.6 closely. */
533 ovs_mutex_lock(&mutex);
534 if (flow->nw_ttl != 255) {
535 /* XXX Should drop in the kernel to prevent DOS. */
539 msg = ofpbuf_at(p, (uint8_t *)p->l7 - (uint8_t *)p->data, BFD_PACKET_LEN);
541 VLOG_INFO_RL(&rl, "%s: Received unparseable BFD control message.",
546 /* RFC 5880 Section 6.8.6
547 * If the Length field is greater than the payload of the encapsulating
548 * protocol, the packet MUST be discarded.
550 * Note that we make this check implicity. Above we use ofpbuf_at() to
551 * ensure that there are at least BFD_PACKET_LEN bytes in the payload of
552 * the encapsulating protocol. Below we require msg->length to be exactly
553 * BFD_PACKET_LEN bytes. */
555 flags = msg->flags & FLAGS_MASK;
556 rmt_state = msg->flags & STATE_MASK;
557 version = msg->vers_diag >> VERS_SHIFT;
559 log_msg(VLL_DBG, msg, "Received BFD control message", bfd);
561 if (version != BFD_VERSION) {
562 log_msg(VLL_WARN, msg, "Incorrect version", bfd);
566 /* Technically this should happen after the length check. We don't support
567 * authentication however, so it's simpler to do the check first. */
568 if (flags & FLAG_AUTH) {
569 log_msg(VLL_WARN, msg, "Authenticated control message with"
570 " authentication disabled", bfd);
574 if (msg->length != BFD_PACKET_LEN) {
575 log_msg(VLL_WARN, msg, "Unexpected length", bfd);
576 if (msg->length < BFD_PACKET_LEN) {
582 log_msg(VLL_WARN, msg, "Zero multiplier", bfd);
586 if (flags & FLAG_MULTIPOINT) {
587 log_msg(VLL_WARN, msg, "Unsupported multipoint flag", bfd);
592 log_msg(VLL_WARN, msg, "NULL my_disc", bfd);
596 pkt_your_disc = ntohl(msg->your_disc);
598 /* Technically, we should use the your discriminator field to figure
599 * out which 'struct bfd' this packet is destined towards. That way a
600 * bfd session could migrate from one interface to another
601 * transparently. This doesn't fit in with the OVS structure very
602 * well, so in this respect, we are not compliant. */
603 if (pkt_your_disc != bfd->disc) {
604 log_msg(VLL_WARN, msg, "Incorrect your_disc", bfd);
607 } else if (rmt_state > STATE_DOWN) {
608 log_msg(VLL_WARN, msg, "Null your_disc", bfd);
612 bfd->rmt_disc = ntohl(msg->my_disc);
613 bfd->rmt_state = rmt_state;
614 bfd->rmt_flags = flags;
615 bfd->rmt_diag = msg->vers_diag & DIAG_MASK;
617 if (flags & FLAG_FINAL && bfd_in_poll(bfd)) {
618 bfd->min_tx = bfd->poll_min_tx;
619 bfd->min_rx = bfd->poll_min_rx;
620 bfd->flags &= ~FLAG_POLL;
621 log_msg(VLL_INFO, msg, "Poll sequence terminated", bfd);
624 if (flags & FLAG_POLL) {
625 /* RFC 5880 Section 6.5
626 * When the other system receives a Poll, it immediately transmits a
627 * BFD Control packet with the Final (F) bit set, independent of any
628 * periodic BFD Control packets it may be sending
629 * (see section 6.8.7). */
630 bfd->flags &= ~FLAG_POLL;
631 bfd->flags |= FLAG_FINAL;
634 rmt_min_rx = MAX(ntohl(msg->min_rx) / 1000, 1);
635 if (bfd->rmt_min_rx != rmt_min_rx) {
636 bfd->rmt_min_rx = rmt_min_rx;
637 bfd_set_next_tx(bfd);
638 log_msg(VLL_INFO, msg, "New remote min_rx", bfd);
641 bfd->rmt_min_tx = MAX(ntohl(msg->min_tx) / 1000, 1);
642 bfd->detect_time = bfd_rx_interval(bfd) * bfd->mult + time_msec();
644 if (bfd->state == STATE_ADMIN_DOWN) {
645 VLOG_DBG_RL(&rl, "Administratively down, dropping control message.");
649 if (rmt_state == STATE_ADMIN_DOWN) {
650 if (bfd->state != STATE_DOWN) {
651 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
654 switch (bfd->state) {
656 if (rmt_state == STATE_DOWN) {
657 bfd_set_state(bfd, STATE_INIT, bfd->diag);
658 } else if (rmt_state == STATE_INIT) {
659 bfd_set_state(bfd, STATE_UP, bfd->diag);
663 if (rmt_state > STATE_DOWN) {
664 bfd_set_state(bfd, STATE_UP, bfd->diag);
668 if (rmt_state <= STATE_DOWN) {
669 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
670 log_msg(VLL_INFO, msg, "Remote signaled STATE_DOWN", bfd);
673 case STATE_ADMIN_DOWN:
678 /* XXX: RFC 5880 Section 6.8.6 Demand mode related calculations here. */
681 ovs_mutex_unlock(&mutex);
685 bfd_forwarding__(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
687 if (bfd->forwarding_override != -1) {
688 return bfd->forwarding_override == 1;
691 return bfd->state == STATE_UP
692 && bfd->rmt_diag != DIAG_PATH_DOWN
693 && bfd->rmt_diag != DIAG_CPATH_DOWN
694 && bfd->rmt_diag != DIAG_RCPATH_DOWN;
699 bfd_in_poll(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
701 return (bfd->flags & FLAG_POLL) != 0;
705 bfd_poll(struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
707 if (bfd->state > STATE_DOWN && !bfd_in_poll(bfd)
708 && !(bfd->flags & FLAG_FINAL)) {
709 bfd->poll_min_tx = bfd->cfg_min_tx;
710 bfd->poll_min_rx = bfd->cfg_min_rx;
711 bfd->flags |= FLAG_POLL;
713 VLOG_INFO_RL(&rl, "%s: Initiating poll sequence", bfd->name);
718 bfd_min_tx(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
720 /* RFC 5880 Section 6.8.3
721 * When bfd.SessionState is not Up, the system MUST set
722 * bfd.DesiredMinTxInterval to a value of not less than one second
723 * (1,000,000 microseconds). This is intended to ensure that the
724 * bandwidth consumed by BFD sessions that are not Up is negligible,
725 * particularly in the case where a neighbor may not be running BFD. */
726 return (bfd->state == STATE_UP ? bfd->min_tx : MAX(bfd->min_tx, 1000));
730 bfd_tx_interval(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
732 long long int interval = bfd_min_tx(bfd);
733 return MAX(interval, bfd->rmt_min_rx);
737 bfd_rx_interval(const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
739 return MAX(bfd->min_rx, bfd->rmt_min_tx);
743 bfd_set_next_tx(struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
745 long long int interval = bfd_tx_interval(bfd);
746 interval -= interval * random_range(26) / 100;
747 bfd->next_tx = bfd->last_tx + interval;
751 bfd_flag_str(enum flags flags)
753 struct ds ds = DS_EMPTY_INITIALIZER;
754 static char flag_str[128];
760 if (flags & FLAG_MULTIPOINT) {
761 ds_put_cstr(&ds, "multipoint ");
764 if (flags & FLAG_DEMAND) {
765 ds_put_cstr(&ds, "demand ");
768 if (flags & FLAG_AUTH) {
769 ds_put_cstr(&ds, "auth ");
772 if (flags & FLAG_CTL) {
773 ds_put_cstr(&ds, "ctl ");
776 if (flags & FLAG_FINAL) {
777 ds_put_cstr(&ds, "final ");
780 if (flags & FLAG_POLL) {
781 ds_put_cstr(&ds, "poll ");
784 ovs_strlcpy(flag_str, ds_cstr(&ds), sizeof flag_str);
790 bfd_state_str(enum state state)
793 case STATE_ADMIN_DOWN: return "admin_down";
794 case STATE_DOWN: return "down";
795 case STATE_INIT: return "init";
796 case STATE_UP: return "up";
797 default: return "invalid";
802 bfd_diag_str(enum diag diag) {
804 case DIAG_NONE: return "No Diagnostic";
805 case DIAG_EXPIRED: return "Control Detection Time Expired";
806 case DIAG_ECHO_FAILED: return "Echo Function Failed";
807 case DIAG_RMT_DOWN: return "Neighbor Signaled Session Down";
808 case DIAG_FWD_RESET: return "Forwarding Plane Reset";
809 case DIAG_PATH_DOWN: return "Path Down";
810 case DIAG_CPATH_DOWN: return "Concatenated Path Down";
811 case DIAG_ADMIN_DOWN: return "Administratively Down";
812 case DIAG_RCPATH_DOWN: return "Reverse Concatenated Path Down";
813 default: return "Invalid Diagnostic";
818 log_msg(enum vlog_level level, const struct msg *p, const char *message,
819 const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
821 struct ds ds = DS_EMPTY_INITIALIZER;
823 if (vlog_should_drop(THIS_MODULE, level, &rl)) {
829 "\n\tvers:%"PRIu8" diag:\"%s\" state:%s mult:%"PRIu8
832 "\n\tmy_disc:0x%"PRIx32" your_disc:0x%"PRIx32
833 "\n\tmin_tx:%"PRIu32"us (%"PRIu32"ms)"
834 "\n\tmin_rx:%"PRIu32"us (%"PRIu32"ms)"
835 "\n\tmin_rx_echo:%"PRIu32"us (%"PRIu32"ms)",
836 bfd->name, message, p->vers_diag >> VERS_SHIFT,
837 bfd_diag_str(p->vers_diag & DIAG_MASK),
838 bfd_state_str(p->flags & STATE_MASK),
839 p->mult, p->length, bfd_flag_str(p->flags & FLAGS_MASK),
840 ntohl(p->my_disc), ntohl(p->your_disc),
841 ntohl(p->min_tx), ntohl(p->min_tx) / 1000,
842 ntohl(p->min_rx), ntohl(p->min_rx) / 1000,
843 ntohl(p->min_rx_echo), ntohl(p->min_rx_echo) / 1000);
844 bfd_put_details(&ds, bfd);
845 VLOG(level, "%s", ds_cstr(&ds));
850 bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
851 OVS_REQ_WRLOCK(mutex)
853 if (diag == DIAG_NONE && bfd->cpath_down) {
854 diag = DIAG_CPATH_DOWN;
857 if (bfd->state != state || bfd->diag != diag) {
858 if (!VLOG_DROP_INFO(&rl)) {
859 struct ds ds = DS_EMPTY_INITIALIZER;
861 ds_put_format(&ds, "%s: BFD state change: %s->%s"
862 " \"%s\"->\"%s\".\n",
863 bfd->name, bfd_state_str(bfd->state),
864 bfd_state_str(state), bfd_diag_str(bfd->diag),
866 bfd_put_details(&ds, bfd);
867 VLOG_INFO("%s", ds_cstr(&ds));
874 if (bfd->state <= STATE_DOWN) {
875 bfd->rmt_state = STATE_DOWN;
876 bfd->rmt_diag = DIAG_NONE;
886 generate_discriminator(void)
890 /* RFC 5880 Section 6.8.1
891 * It SHOULD be set to a random (but still unique) value to improve
892 * security. The value is otherwise outside the scope of this
898 /* 'disc' is by definition random, so there's no reason to waste time
900 disc = random_uint32();
901 HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, all_bfds) {
902 if (bfd->disc == disc) {
913 bfd_find_by_name(const char *name) OVS_REQ_WRLOCK(mutex)
917 HMAP_FOR_EACH (bfd, node, all_bfds) {
918 if (!strcmp(bfd->name, name)) {
926 bfd_put_details(struct ds *ds, const struct bfd *bfd) OVS_REQ_WRLOCK(mutex)
928 ds_put_format(ds, "\tForwarding: %s\n",
929 bfd_forwarding__(bfd) ? "true" : "false");
930 ds_put_format(ds, "\tDetect Multiplier: %d\n", bfd->mult);
931 ds_put_format(ds, "\tConcatenated Path Down: %s\n",
932 bfd->cpath_down ? "true" : "false");
933 ds_put_format(ds, "\tTX Interval: Approx %lldms\n", bfd_tx_interval(bfd));
934 ds_put_format(ds, "\tRX Interval: Approx %lldms\n", bfd_rx_interval(bfd));
935 ds_put_format(ds, "\tDetect Time: now %+lldms\n",
936 time_msec() - bfd->detect_time);
937 ds_put_format(ds, "\tNext TX Time: now %+lldms\n",
938 time_msec() - bfd->next_tx);
939 ds_put_format(ds, "\tLast TX Time: now %+lldms\n",
940 time_msec() - bfd->last_tx);
942 ds_put_cstr(ds, "\n");
944 ds_put_format(ds, "\tLocal Flags: %s\n", bfd_flag_str(bfd->flags));
945 ds_put_format(ds, "\tLocal Session State: %s\n",
946 bfd_state_str(bfd->state));
947 ds_put_format(ds, "\tLocal Diagnostic: %s\n", bfd_diag_str(bfd->diag));
948 ds_put_format(ds, "\tLocal Discriminator: 0x%"PRIx32"\n", bfd->disc);
949 ds_put_format(ds, "\tLocal Minimum TX Interval: %lldms\n",
951 ds_put_format(ds, "\tLocal Minimum RX Interval: %lldms\n", bfd->min_rx);
953 ds_put_cstr(ds, "\n");
955 ds_put_format(ds, "\tRemote Flags: %s\n", bfd_flag_str(bfd->rmt_flags));
956 ds_put_format(ds, "\tRemote Session State: %s\n",
957 bfd_state_str(bfd->rmt_state));
958 ds_put_format(ds, "\tRemote Diagnostic: %s\n",
959 bfd_diag_str(bfd->rmt_diag));
960 ds_put_format(ds, "\tRemote Discriminator: 0x%"PRIx32"\n", bfd->rmt_disc);
961 ds_put_format(ds, "\tRemote Minimum TX Interval: %lldms\n",
963 ds_put_format(ds, "\tRemote Minimum RX Interval: %lldms\n",
968 bfd_unixctl_show(struct unixctl_conn *conn, int argc, const char *argv[],
969 void *aux OVS_UNUSED) OVS_EXCLUDED(mutex)
971 struct ds ds = DS_EMPTY_INITIALIZER;
974 ovs_mutex_lock(&mutex);
976 bfd = bfd_find_by_name(argv[1]);
978 unixctl_command_reply_error(conn, "no such bfd object");
981 bfd_put_details(&ds, bfd);
983 HMAP_FOR_EACH (bfd, node, all_bfds) {
984 ds_put_format(&ds, "---- %s ----\n", bfd->name);
985 bfd_put_details(&ds, bfd);
988 unixctl_command_reply(conn, ds_cstr(&ds));
992 ovs_mutex_unlock(&mutex);
997 bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
998 const char *argv[], void *aux OVS_UNUSED)
1001 const char *forward_str = argv[argc - 1];
1002 int forwarding_override;
1005 ovs_mutex_lock(&mutex);
1006 if (!strcasecmp("true", forward_str)) {
1007 forwarding_override = 1;
1008 } else if (!strcasecmp("false", forward_str)) {
1009 forwarding_override = 0;
1010 } else if (!strcasecmp("normal", forward_str)) {
1011 forwarding_override = -1;
1013 unixctl_command_reply_error(conn, "unknown fault string");
1018 bfd = bfd_find_by_name(argv[1]);
1020 unixctl_command_reply_error(conn, "no such BFD object");
1023 bfd->forwarding_override = forwarding_override;
1025 HMAP_FOR_EACH (bfd, node, all_bfds) {
1026 bfd->forwarding_override = forwarding_override;
1030 unixctl_command_reply(conn, "OK");
1033 ovs_mutex_unlock(&mutex);