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/in_systm.h>
20 #include <netinet/ip.h>
22 #include "byte-order.h"
25 #include "dynamic-string.h"
33 #include "ovs-thread.h"
34 #include "openvswitch/types.h"
36 #include "poll-loop.h"
44 VLOG_DEFINE_THIS_MODULE(bfd);
48 * The goal of this module is to replace CFM with something both more flexible
49 * and standards compliant. In service of this goal, the following needs to be
53 * * Implement Demand mode.
54 * * Go through the RFC line by line and verify we comply.
55 * * Test against a hardware implementation. Preferably a popular one.
56 * * Delete BFD packets with nw_ttl != 255 in the datapath to prevent DOS
61 * - Set TOS/PCP on the 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 uint8_t eth_dst[ETH_ADDR_LEN];/* Ethernet destination address. */
165 bool eth_dst_set; /* 'eth_dst' set through database. */
167 uint16_t udp_src; /* UDP source port. */
169 /* All timers in milliseconds. */
170 long long int rmt_min_rx; /* bfd.RemoteMinRxInterval. */
171 long long int rmt_min_tx; /* Remote minimum TX interval. */
173 long long int cfg_min_tx; /* Configured minimum TX rate. */
174 long long int cfg_min_rx; /* Configured required minimum RX rate. */
175 long long int poll_min_tx; /* Min TX negotating in a poll sequence. */
176 long long int poll_min_rx; /* Min RX negotating in a poll sequence. */
177 long long int min_tx; /* bfd.DesiredMinTxInterval. */
178 long long int min_rx; /* bfd.RequiredMinRxInterval. */
180 long long int last_tx; /* Last TX time. */
181 long long int next_tx; /* Next TX time. */
182 long long int detect_time; /* RFC 5880 6.8.4 Detection time. */
184 int forwarding_override; /* Manual override of 'forwarding' status. */
186 atomic_bool check_tnl_key; /* Verify tunnel key of inbound packets? */
190 static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
191 static struct hmap all_bfds__ = HMAP_INITIALIZER(&all_bfds__);
192 static struct hmap *const all_bfds OVS_GUARDED_BY(mutex) = &all_bfds__;
194 static bool bfd_forwarding__(const struct bfd *) OVS_REQUIRES(mutex);
195 static bool bfd_in_poll(const struct bfd *) OVS_REQUIRES(mutex);
196 static void bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex);
197 static const char *bfd_diag_str(enum diag) OVS_REQUIRES(mutex);
198 static const char *bfd_state_str(enum state) OVS_REQUIRES(mutex);
199 static long long int bfd_min_tx(const struct bfd *) OVS_REQUIRES(mutex);
200 static long long int bfd_tx_interval(const struct bfd *)
202 static long long int bfd_rx_interval(const struct bfd *)
204 static void bfd_set_next_tx(struct bfd *) OVS_REQUIRES(mutex);
205 static void bfd_set_state(struct bfd *, enum state, enum diag)
207 static uint32_t generate_discriminator(void) OVS_REQUIRES(mutex);
208 static void bfd_put_details(struct ds *, const struct bfd *)
210 static void bfd_unixctl_show(struct unixctl_conn *, int argc,
211 const char *argv[], void *aux OVS_UNUSED);
212 static void bfd_unixctl_set_forwarding_override(struct unixctl_conn *,
213 int argc, const char *argv[],
214 void *aux OVS_UNUSED);
215 static void log_msg(enum vlog_level, const struct msg *, const char *message,
216 const struct bfd *) OVS_REQUIRES(mutex);
218 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(20, 20);
220 /* Returns true if the interface on which 'bfd' is running may be used to
221 * forward traffic according to the BFD session state. */
223 bfd_forwarding(const struct bfd *bfd) OVS_EXCLUDED(mutex)
227 ovs_mutex_lock(&mutex);
228 ret = bfd_forwarding__(bfd);
229 ovs_mutex_unlock(&mutex);
233 /* Returns a 'smap' of key value pairs representing the status of 'bfd'
234 * intended for the OVS database. */
236 bfd_get_status(const struct bfd *bfd, struct smap *smap)
239 ovs_mutex_lock(&mutex);
240 smap_add(smap, "forwarding", bfd_forwarding__(bfd)? "true" : "false");
241 smap_add(smap, "state", bfd_state_str(bfd->state));
242 smap_add(smap, "diagnostic", bfd_diag_str(bfd->diag));
244 if (bfd->state != STATE_DOWN) {
245 smap_add(smap, "remote_state", bfd_state_str(bfd->rmt_state));
246 smap_add(smap, "remote_diagnostic", bfd_diag_str(bfd->rmt_diag));
248 ovs_mutex_unlock(&mutex);
251 /* Initializes, destroys, or reconfigures the BFD session 'bfd' (named 'name'),
252 * according to the database configuration contained in 'cfg'. Takes ownership
253 * of 'bfd', which may be NULL. Returns a BFD object which may be used as a
254 * handle for the session, or NULL if BFD is not enabled according to 'cfg'.
255 * Also returns NULL if cfg is NULL. */
257 bfd_configure(struct bfd *bfd, const char *name, const struct smap *cfg)
260 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
261 static atomic_uint16_t udp_src = ATOMIC_VAR_INIT(0);
263 long long int min_tx, min_rx;
266 uint8_t ea[ETH_ADDR_LEN];
268 if (ovsthread_once_start(&once)) {
269 unixctl_command_register("bfd/show", "[interface]", 0, 1,
270 bfd_unixctl_show, NULL);
271 unixctl_command_register("bfd/set-forwarding",
272 "[interface] normal|false|true", 1, 2,
273 bfd_unixctl_set_forwarding_override, NULL);
274 ovsthread_once_done(&once);
277 if (!cfg || !smap_get_bool(cfg, "enable", false)) {
282 ovs_mutex_lock(&mutex);
284 bfd = xzalloc(sizeof *bfd);
285 bfd->name = xstrdup(name);
286 bfd->forwarding_override = -1;
287 bfd->disc = generate_discriminator();
288 hmap_insert(all_bfds, &bfd->node, bfd->disc);
290 bfd->diag = DIAG_NONE;
293 atomic_init(&bfd->ref_cnt, 1);
295 /* RFC 5881 section 4
296 * The source port MUST be in the range 49152 through 65535. The same
297 * UDP source port number MUST be used for all BFD Control packets
298 * associated with a particular session. The source port number SHOULD
299 * be unique among all BFD sessions on the system. */
300 atomic_add(&udp_src, 1, &bfd->udp_src);
301 bfd->udp_src = (bfd->udp_src % 16384) + 49152;
303 bfd_set_state(bfd, STATE_DOWN, DIAG_NONE);
305 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
308 atomic_store(&bfd->check_tnl_key,
309 smap_get_bool(cfg, "check_tnl_key", false));
310 min_tx = smap_get_int(cfg, "min_tx", 100);
311 min_tx = MAX(min_tx, 100);
312 if (bfd->cfg_min_tx != min_tx) {
313 bfd->cfg_min_tx = min_tx;
314 if (bfd->state != STATE_UP
315 || (!bfd_in_poll(bfd) && bfd->cfg_min_tx < bfd->min_tx)) {
316 bfd->min_tx = bfd->cfg_min_tx;
321 min_rx = smap_get_int(cfg, "min_rx", 1000);
322 min_rx = MAX(min_rx, 100);
323 if (bfd->cfg_min_rx != min_rx) {
324 bfd->cfg_min_rx = min_rx;
325 if (bfd->state != STATE_UP
326 || (!bfd_in_poll(bfd) && bfd->cfg_min_rx > bfd->min_rx)) {
327 bfd->min_rx = bfd->cfg_min_rx;
332 cpath_down = smap_get_bool(cfg, "cpath_down", false);
333 if (bfd->cpath_down != cpath_down) {
334 bfd->cpath_down = cpath_down;
335 if (bfd->diag == DIAG_NONE || bfd->diag == DIAG_CPATH_DOWN) {
336 bfd_set_state(bfd, bfd->state, DIAG_NONE);
341 hwaddr = smap_get(cfg, "bfd_dst_mac");
342 if (hwaddr && eth_addr_from_string(hwaddr, ea) && !eth_addr_is_zero(ea)) {
343 memcpy(bfd->eth_dst, ea, ETH_ADDR_LEN);
344 bfd->eth_dst_set = true;
345 } else if (bfd->eth_dst_set) {
346 memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
347 bfd->eth_dst_set = false;
350 ovs_mutex_unlock(&mutex);
355 bfd_ref(const struct bfd *bfd_)
357 struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
360 atomic_add(&bfd->ref_cnt, 1, &orig);
361 ovs_assert(orig > 0);
367 bfd_unref(struct bfd *bfd) OVS_EXCLUDED(mutex)
372 atomic_sub(&bfd->ref_cnt, 1, &orig);
373 ovs_assert(orig > 0);
375 ovs_mutex_lock(&mutex);
376 hmap_remove(all_bfds, &bfd->node);
379 ovs_mutex_unlock(&mutex);
385 bfd_wait(const struct bfd *bfd) OVS_EXCLUDED(mutex)
387 ovs_mutex_lock(&mutex);
388 if (bfd->flags & FLAG_FINAL) {
389 poll_immediate_wake();
392 poll_timer_wait_until(bfd->next_tx);
393 if (bfd->state > STATE_DOWN) {
394 poll_timer_wait_until(bfd->detect_time);
396 ovs_mutex_unlock(&mutex);
400 bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
402 ovs_mutex_lock(&mutex);
403 if (bfd->state > STATE_DOWN && time_msec() >= bfd->detect_time) {
404 bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
407 if (bfd->min_tx != bfd->cfg_min_tx || bfd->min_rx != bfd->cfg_min_rx) {
410 ovs_mutex_unlock(&mutex);
414 bfd_should_send_packet(const struct bfd *bfd) OVS_EXCLUDED(mutex)
417 ovs_mutex_lock(&mutex);
418 ret = bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
419 ovs_mutex_unlock(&mutex);
424 bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
425 uint8_t eth_src[ETH_ADDR_LEN]) OVS_EXCLUDED(mutex)
427 long long int min_tx, min_rx;
428 struct udp_header *udp;
429 struct eth_header *eth;
430 struct ip_header *ip;
433 ovs_mutex_lock(&mutex);
435 long long int delay = time_msec() - bfd->next_tx;
436 long long int interval = bfd_tx_interval(bfd);
437 if (delay > interval * 3 / 2) {
438 VLOG_INFO("%s: long delay of %lldms (expected %lldms) sending BFD"
439 " control message", bfd->name, delay, interval);
443 /* RFC 5880 Section 6.5
444 * A BFD Control packet MUST NOT have both the Poll (P) and Final (F) bits
446 ovs_assert(!(bfd->flags & FLAG_POLL) || !(bfd->flags & FLAG_FINAL));
448 ofpbuf_reserve(p, 2); /* Properly align after the ethernet header. */
449 eth = ofpbuf_put_uninit(p, sizeof *eth);
450 memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
451 memcpy(eth->eth_dst, bfd->eth_dst, ETH_ADDR_LEN);
452 eth->eth_type = htons(ETH_TYPE_IP);
454 ip = ofpbuf_put_zeros(p, sizeof *ip);
455 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
456 ip->ip_tot_len = htons(sizeof *ip + sizeof *udp + sizeof *msg);
458 ip->ip_tos = IPTOS_LOWDELAY | IPTOS_THROUGHPUT;
459 ip->ip_proto = IPPROTO_UDP;
460 ip->ip_src = htonl(0xA9FE0100); /* 169.254.1.0 Link Local. */
461 ip->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1 Link Local. */
462 ip->ip_csum = csum(ip, sizeof *ip);
464 udp = ofpbuf_put_zeros(p, sizeof *udp);
465 udp->udp_src = htons(bfd->udp_src);
466 udp->udp_dst = htons(BFD_DEST_PORT);
467 udp->udp_len = htons(sizeof *udp + sizeof *msg);
469 msg = ofpbuf_put_uninit(p, sizeof *msg);
470 msg->vers_diag = (BFD_VERSION << 5) | bfd->diag;
471 msg->flags = (bfd->state & STATE_MASK) | bfd->flags;
473 msg->mult = bfd->mult;
474 msg->length = BFD_PACKET_LEN;
475 msg->my_disc = htonl(bfd->disc);
476 msg->your_disc = htonl(bfd->rmt_disc);
477 msg->min_rx_echo = htonl(0);
479 if (bfd_in_poll(bfd)) {
480 min_tx = bfd->poll_min_tx;
481 min_rx = bfd->poll_min_rx;
483 min_tx = bfd_min_tx(bfd);
484 min_rx = bfd->min_rx;
487 msg->min_tx = htonl(min_tx * 1000);
488 msg->min_rx = htonl(min_rx * 1000);
490 bfd->flags &= ~FLAG_FINAL;
492 log_msg(VLL_DBG, msg, "Sending BFD Message", bfd);
494 bfd->last_tx = time_msec();
495 bfd_set_next_tx(bfd);
496 ovs_mutex_unlock(&mutex);
500 bfd_should_process_flow(const struct bfd *bfd, const struct flow *flow,
501 struct flow_wildcards *wc)
504 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
505 if (bfd->eth_dst_set && memcmp(bfd->eth_dst, flow->dl_dst, ETH_ADDR_LEN)) {
509 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
510 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
512 atomic_read(&bfd->check_tnl_key, &check_tnl_key);
514 memset(&wc->masks.tunnel.tun_id, 0xff, sizeof wc->masks.tunnel.tun_id);
516 return (flow->dl_type == htons(ETH_TYPE_IP)
517 && flow->nw_proto == IPPROTO_UDP
518 && flow->tp_dst == htons(BFD_DEST_PORT)
519 && (check_tnl_key || flow->tunnel.tun_id == htonll(0)));
523 bfd_process_packet(struct bfd *bfd, const struct flow *flow,
524 const struct ofpbuf *p) OVS_EXCLUDED(mutex)
526 uint32_t rmt_min_rx, pkt_your_disc;
527 enum state rmt_state;
532 /* This function is designed to follow section RFC 5880 6.8.6 closely. */
534 ovs_mutex_lock(&mutex);
535 if (flow->nw_ttl != 255) {
536 /* XXX Should drop in the kernel to prevent DOS. */
540 msg = ofpbuf_at(p, (uint8_t *)p->l7 - (uint8_t *)p->data, BFD_PACKET_LEN);
542 VLOG_INFO_RL(&rl, "%s: Received unparseable BFD control message.",
547 /* RFC 5880 Section 6.8.6
548 * If the Length field is greater than the payload of the encapsulating
549 * protocol, the packet MUST be discarded.
551 * Note that we make this check implicity. Above we use ofpbuf_at() to
552 * ensure that there are at least BFD_PACKET_LEN bytes in the payload of
553 * the encapsulating protocol. Below we require msg->length to be exactly
554 * BFD_PACKET_LEN bytes. */
556 flags = msg->flags & FLAGS_MASK;
557 rmt_state = msg->flags & STATE_MASK;
558 version = msg->vers_diag >> VERS_SHIFT;
560 log_msg(VLL_DBG, msg, "Received BFD control message", bfd);
562 if (version != BFD_VERSION) {
563 log_msg(VLL_WARN, msg, "Incorrect version", bfd);
567 /* Technically this should happen after the length check. We don't support
568 * authentication however, so it's simpler to do the check first. */
569 if (flags & FLAG_AUTH) {
570 log_msg(VLL_WARN, msg, "Authenticated control message with"
571 " authentication disabled", bfd);
575 if (msg->length != BFD_PACKET_LEN) {
576 log_msg(VLL_WARN, msg, "Unexpected length", bfd);
577 if (msg->length < BFD_PACKET_LEN) {
583 log_msg(VLL_WARN, msg, "Zero multiplier", bfd);
587 if (flags & FLAG_MULTIPOINT) {
588 log_msg(VLL_WARN, msg, "Unsupported multipoint flag", bfd);
593 log_msg(VLL_WARN, msg, "NULL my_disc", bfd);
597 pkt_your_disc = ntohl(msg->your_disc);
599 /* Technically, we should use the your discriminator field to figure
600 * out which 'struct bfd' this packet is destined towards. That way a
601 * bfd session could migrate from one interface to another
602 * transparently. This doesn't fit in with the OVS structure very
603 * well, so in this respect, we are not compliant. */
604 if (pkt_your_disc != bfd->disc) {
605 log_msg(VLL_WARN, msg, "Incorrect your_disc", bfd);
608 } else if (rmt_state > STATE_DOWN) {
609 log_msg(VLL_WARN, msg, "Null your_disc", bfd);
613 bfd->rmt_disc = ntohl(msg->my_disc);
614 bfd->rmt_state = rmt_state;
615 bfd->rmt_flags = flags;
616 bfd->rmt_diag = msg->vers_diag & DIAG_MASK;
618 if (flags & FLAG_FINAL && bfd_in_poll(bfd)) {
619 bfd->min_tx = bfd->poll_min_tx;
620 bfd->min_rx = bfd->poll_min_rx;
621 bfd->flags &= ~FLAG_POLL;
622 log_msg(VLL_INFO, msg, "Poll sequence terminated", bfd);
625 if (flags & FLAG_POLL) {
626 /* RFC 5880 Section 6.5
627 * When the other system receives a Poll, it immediately transmits a
628 * BFD Control packet with the Final (F) bit set, independent of any
629 * periodic BFD Control packets it may be sending
630 * (see section 6.8.7). */
631 bfd->flags &= ~FLAG_POLL;
632 bfd->flags |= FLAG_FINAL;
635 rmt_min_rx = MAX(ntohl(msg->min_rx) / 1000, 1);
636 if (bfd->rmt_min_rx != rmt_min_rx) {
637 bfd->rmt_min_rx = rmt_min_rx;
638 bfd_set_next_tx(bfd);
639 log_msg(VLL_INFO, msg, "New remote min_rx", bfd);
642 bfd->rmt_min_tx = MAX(ntohl(msg->min_tx) / 1000, 1);
643 bfd->detect_time = bfd_rx_interval(bfd) * bfd->mult + time_msec();
645 if (bfd->state == STATE_ADMIN_DOWN) {
646 VLOG_DBG_RL(&rl, "Administratively down, dropping control message.");
650 if (rmt_state == STATE_ADMIN_DOWN) {
651 if (bfd->state != STATE_DOWN) {
652 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
655 switch (bfd->state) {
657 if (rmt_state == STATE_DOWN) {
658 bfd_set_state(bfd, STATE_INIT, bfd->diag);
659 } else if (rmt_state == STATE_INIT) {
660 bfd_set_state(bfd, STATE_UP, bfd->diag);
664 if (rmt_state > STATE_DOWN) {
665 bfd_set_state(bfd, STATE_UP, bfd->diag);
669 if (rmt_state <= STATE_DOWN) {
670 bfd_set_state(bfd, STATE_DOWN, DIAG_RMT_DOWN);
671 log_msg(VLL_INFO, msg, "Remote signaled STATE_DOWN", bfd);
674 case STATE_ADMIN_DOWN:
679 /* XXX: RFC 5880 Section 6.8.6 Demand mode related calculations here. */
682 ovs_mutex_unlock(&mutex);
686 bfd_forwarding__(const struct bfd *bfd) OVS_REQUIRES(mutex)
688 if (bfd->forwarding_override != -1) {
689 return bfd->forwarding_override == 1;
692 return bfd->state == STATE_UP
693 && bfd->rmt_diag != DIAG_PATH_DOWN
694 && bfd->rmt_diag != DIAG_CPATH_DOWN
695 && bfd->rmt_diag != DIAG_RCPATH_DOWN;
700 bfd_in_poll(const struct bfd *bfd) OVS_REQUIRES(mutex)
702 return (bfd->flags & FLAG_POLL) != 0;
706 bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex)
708 if (bfd->state > STATE_DOWN && !bfd_in_poll(bfd)
709 && !(bfd->flags & FLAG_FINAL)) {
710 bfd->poll_min_tx = bfd->cfg_min_tx;
711 bfd->poll_min_rx = bfd->cfg_min_rx;
712 bfd->flags |= FLAG_POLL;
714 VLOG_INFO_RL(&rl, "%s: Initiating poll sequence", bfd->name);
719 bfd_min_tx(const struct bfd *bfd) OVS_REQUIRES(mutex)
721 /* RFC 5880 Section 6.8.3
722 * When bfd.SessionState is not Up, the system MUST set
723 * bfd.DesiredMinTxInterval to a value of not less than one second
724 * (1,000,000 microseconds). This is intended to ensure that the
725 * bandwidth consumed by BFD sessions that are not Up is negligible,
726 * particularly in the case where a neighbor may not be running BFD. */
727 return (bfd->state == STATE_UP ? bfd->min_tx : MAX(bfd->min_tx, 1000));
731 bfd_tx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
733 long long int interval = bfd_min_tx(bfd);
734 return MAX(interval, bfd->rmt_min_rx);
738 bfd_rx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
740 return MAX(bfd->min_rx, bfd->rmt_min_tx);
744 bfd_set_next_tx(struct bfd *bfd) OVS_REQUIRES(mutex)
746 long long int interval = bfd_tx_interval(bfd);
747 interval -= interval * random_range(26) / 100;
748 bfd->next_tx = bfd->last_tx + interval;
752 bfd_flag_str(enum flags flags)
754 struct ds ds = DS_EMPTY_INITIALIZER;
755 static char flag_str[128];
761 if (flags & FLAG_MULTIPOINT) {
762 ds_put_cstr(&ds, "multipoint ");
765 if (flags & FLAG_DEMAND) {
766 ds_put_cstr(&ds, "demand ");
769 if (flags & FLAG_AUTH) {
770 ds_put_cstr(&ds, "auth ");
773 if (flags & FLAG_CTL) {
774 ds_put_cstr(&ds, "ctl ");
777 if (flags & FLAG_FINAL) {
778 ds_put_cstr(&ds, "final ");
781 if (flags & FLAG_POLL) {
782 ds_put_cstr(&ds, "poll ");
785 ovs_strlcpy(flag_str, ds_cstr(&ds), sizeof flag_str);
791 bfd_state_str(enum state state)
794 case STATE_ADMIN_DOWN: return "admin_down";
795 case STATE_DOWN: return "down";
796 case STATE_INIT: return "init";
797 case STATE_UP: return "up";
798 default: return "invalid";
803 bfd_diag_str(enum diag diag) {
805 case DIAG_NONE: return "No Diagnostic";
806 case DIAG_EXPIRED: return "Control Detection Time Expired";
807 case DIAG_ECHO_FAILED: return "Echo Function Failed";
808 case DIAG_RMT_DOWN: return "Neighbor Signaled Session Down";
809 case DIAG_FWD_RESET: return "Forwarding Plane Reset";
810 case DIAG_PATH_DOWN: return "Path Down";
811 case DIAG_CPATH_DOWN: return "Concatenated Path Down";
812 case DIAG_ADMIN_DOWN: return "Administratively Down";
813 case DIAG_RCPATH_DOWN: return "Reverse Concatenated Path Down";
814 default: return "Invalid Diagnostic";
819 log_msg(enum vlog_level level, const struct msg *p, const char *message,
820 const struct bfd *bfd) OVS_REQUIRES(mutex)
822 struct ds ds = DS_EMPTY_INITIALIZER;
824 if (vlog_should_drop(THIS_MODULE, level, &rl)) {
830 "\n\tvers:%"PRIu8" diag:\"%s\" state:%s mult:%"PRIu8
833 "\n\tmy_disc:0x%"PRIx32" your_disc:0x%"PRIx32
834 "\n\tmin_tx:%"PRIu32"us (%"PRIu32"ms)"
835 "\n\tmin_rx:%"PRIu32"us (%"PRIu32"ms)"
836 "\n\tmin_rx_echo:%"PRIu32"us (%"PRIu32"ms)",
837 bfd->name, message, p->vers_diag >> VERS_SHIFT,
838 bfd_diag_str(p->vers_diag & DIAG_MASK),
839 bfd_state_str(p->flags & STATE_MASK),
840 p->mult, p->length, bfd_flag_str(p->flags & FLAGS_MASK),
841 ntohl(p->my_disc), ntohl(p->your_disc),
842 ntohl(p->min_tx), ntohl(p->min_tx) / 1000,
843 ntohl(p->min_rx), ntohl(p->min_rx) / 1000,
844 ntohl(p->min_rx_echo), ntohl(p->min_rx_echo) / 1000);
845 bfd_put_details(&ds, bfd);
846 VLOG(level, "%s", ds_cstr(&ds));
851 bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
854 if (diag == DIAG_NONE && bfd->cpath_down) {
855 diag = DIAG_CPATH_DOWN;
858 if (bfd->state != state || bfd->diag != diag) {
859 if (!VLOG_DROP_INFO(&rl)) {
860 struct ds ds = DS_EMPTY_INITIALIZER;
862 ds_put_format(&ds, "%s: BFD state change: %s->%s"
863 " \"%s\"->\"%s\".\n",
864 bfd->name, bfd_state_str(bfd->state),
865 bfd_state_str(state), bfd_diag_str(bfd->diag),
867 bfd_put_details(&ds, bfd);
868 VLOG_INFO("%s", ds_cstr(&ds));
875 if (bfd->state <= STATE_DOWN) {
876 bfd->rmt_state = STATE_DOWN;
877 bfd->rmt_diag = DIAG_NONE;
887 generate_discriminator(void)
891 /* RFC 5880 Section 6.8.1
892 * It SHOULD be set to a random (but still unique) value to improve
893 * security. The value is otherwise outside the scope of this
899 /* 'disc' is by definition random, so there's no reason to waste time
901 disc = random_uint32();
902 HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, all_bfds) {
903 if (bfd->disc == disc) {
914 bfd_find_by_name(const char *name) OVS_REQUIRES(mutex)
918 HMAP_FOR_EACH (bfd, node, all_bfds) {
919 if (!strcmp(bfd->name, name)) {
927 bfd_put_details(struct ds *ds, const struct bfd *bfd) OVS_REQUIRES(mutex)
929 ds_put_format(ds, "\tForwarding: %s\n",
930 bfd_forwarding__(bfd) ? "true" : "false");
931 ds_put_format(ds, "\tDetect Multiplier: %d\n", bfd->mult);
932 ds_put_format(ds, "\tConcatenated Path Down: %s\n",
933 bfd->cpath_down ? "true" : "false");
934 ds_put_format(ds, "\tTX Interval: Approx %lldms\n", bfd_tx_interval(bfd));
935 ds_put_format(ds, "\tRX Interval: Approx %lldms\n", bfd_rx_interval(bfd));
936 ds_put_format(ds, "\tDetect Time: now %+lldms\n",
937 time_msec() - bfd->detect_time);
938 ds_put_format(ds, "\tNext TX Time: now %+lldms\n",
939 time_msec() - bfd->next_tx);
940 ds_put_format(ds, "\tLast TX Time: now %+lldms\n",
941 time_msec() - bfd->last_tx);
943 ds_put_cstr(ds, "\n");
945 ds_put_format(ds, "\tLocal Flags: %s\n", bfd_flag_str(bfd->flags));
946 ds_put_format(ds, "\tLocal Session State: %s\n",
947 bfd_state_str(bfd->state));
948 ds_put_format(ds, "\tLocal Diagnostic: %s\n", bfd_diag_str(bfd->diag));
949 ds_put_format(ds, "\tLocal Discriminator: 0x%"PRIx32"\n", bfd->disc);
950 ds_put_format(ds, "\tLocal Minimum TX Interval: %lldms\n",
952 ds_put_format(ds, "\tLocal Minimum RX Interval: %lldms\n", bfd->min_rx);
954 ds_put_cstr(ds, "\n");
956 ds_put_format(ds, "\tRemote Flags: %s\n", bfd_flag_str(bfd->rmt_flags));
957 ds_put_format(ds, "\tRemote Session State: %s\n",
958 bfd_state_str(bfd->rmt_state));
959 ds_put_format(ds, "\tRemote Diagnostic: %s\n",
960 bfd_diag_str(bfd->rmt_diag));
961 ds_put_format(ds, "\tRemote Discriminator: 0x%"PRIx32"\n", bfd->rmt_disc);
962 ds_put_format(ds, "\tRemote Minimum TX Interval: %lldms\n",
964 ds_put_format(ds, "\tRemote Minimum RX Interval: %lldms\n",
969 bfd_unixctl_show(struct unixctl_conn *conn, int argc, const char *argv[],
970 void *aux OVS_UNUSED) OVS_EXCLUDED(mutex)
972 struct ds ds = DS_EMPTY_INITIALIZER;
975 ovs_mutex_lock(&mutex);
977 bfd = bfd_find_by_name(argv[1]);
979 unixctl_command_reply_error(conn, "no such bfd object");
982 bfd_put_details(&ds, bfd);
984 HMAP_FOR_EACH (bfd, node, all_bfds) {
985 ds_put_format(&ds, "---- %s ----\n", bfd->name);
986 bfd_put_details(&ds, bfd);
989 unixctl_command_reply(conn, ds_cstr(&ds));
993 ovs_mutex_unlock(&mutex);
998 bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
999 const char *argv[], void *aux OVS_UNUSED)
1002 const char *forward_str = argv[argc - 1];
1003 int forwarding_override;
1006 ovs_mutex_lock(&mutex);
1007 if (!strcasecmp("true", forward_str)) {
1008 forwarding_override = 1;
1009 } else if (!strcasecmp("false", forward_str)) {
1010 forwarding_override = 0;
1011 } else if (!strcasecmp("normal", forward_str)) {
1012 forwarding_override = -1;
1014 unixctl_command_reply_error(conn, "unknown fault string");
1019 bfd = bfd_find_by_name(argv[1]);
1021 unixctl_command_reply_error(conn, "no such BFD object");
1024 bfd->forwarding_override = forwarding_override;
1026 HMAP_FOR_EACH (bfd, node, all_bfds) {
1027 bfd->forwarding_override = forwarding_override;
1031 unixctl_command_reply(conn, "OK");
1034 ovs_mutex_unlock(&mutex);