1 /* Copyright (C) 2007 Board of Trustees, Leland Stanford Jr. University.
3 * Permission is hereby granted, free of charge, to any person obtaining a copy
4 * of this software and associated documentation files (the "Software"), to
5 * deal in the Software without restriction, including without limitation the
6 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
7 * sell copies of the Software, and to permit persons to whom the Software is
8 * furnished to do so, subject to the following conditions:
10 * The above copyright notice and this permission notice shall be included in
11 * all copies or substantial portions of the Software.
13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
18 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
26 #include <netinet/in.h>
33 #include "command-line.h"
40 #include "ofp-print.h"
42 #include "poll-loop.h"
46 #include "vconn-ssl.h"
48 #include "vlog-socket.h"
52 #define THIS_MODULE VLM_controller
54 #define MAX_SWITCHES 16
62 time_t last_control_hello;
67 /* -H, --hub: Use dumb hub instead of learning switch? */
68 static bool hub = false;
70 /* -n, --noflow: Pass traffic, but don't setup flows in switch */
71 static bool noflow = false;
73 static void parse_options(int argc, char *argv[]);
74 static void usage(void) NO_RETURN;
76 static struct switch_ *connect_switch(const char *name);
77 static struct switch_ *new_switch(const char *name, struct vconn *);
78 static void close_switch(struct switch_ *);
80 static void queue_tx(struct switch_ *, struct buffer *);
82 static void send_control_hello(struct switch_ *);
84 static int do_switch_recv(struct switch_ *this);
85 static int do_switch_send(struct switch_ *this);
87 static void process_packet(struct switch_ *, struct buffer *);
88 static void process_hub(struct switch_ *, struct ofp_packet_in *);
89 static void process_noflow(struct switch_ *, struct ofp_packet_in *);
91 static void switch_init(void);
92 static void process_switch(struct switch_ *, struct ofp_packet_in *);
95 main(int argc, char *argv[])
97 struct switch_ *switches[MAX_SWITCHES];
102 set_program_name(argv[0]);
103 register_fault_handlers();
105 parse_options(argc, argv);
107 if (!hub && !noflow) {
111 if (argc - optind < 1) {
112 fatal(0, "at least one vconn argument required; use --help for usage");
115 retval = vlog_server_listen(NULL, NULL);
117 fatal(retval, "Could not listen for vlog connections");
121 for (i = 0; i < argc - optind; i++) {
122 struct switch_ *this = connect_switch(argv[optind + i]);
124 if (n_switches >= MAX_SWITCHES) {
125 fatal(0, "max %d switch connections", n_switches);
127 switches[n_switches++] = this;
130 if (n_switches == 0) {
131 fatal(0, "could not connect to any switches");
134 while (n_switches > 0) {
135 /* Do some work. Limit the number of iterations so that callbacks
136 * registered with the poll loop don't starve. */
139 for (iteration = 0; iteration < 50; iteration++) {
140 bool progress = false;
141 for (i = 0; i < n_switches; ) {
142 struct switch_ *this = switches[i];
145 if (vconn_is_passive(this->vconn)) {
147 while (n_switches < MAX_SWITCHES) {
148 struct vconn *new_vconn;
149 retval = vconn_accept(this->vconn, &new_vconn);
153 switches[n_switches++] = new_switch("tcp", new_vconn);
156 retval = do_switch_recv(this);
157 if (!retval || retval == EAGAIN) {
159 retval = do_switch_send(this);
167 if (retval && retval != EAGAIN) {
169 switches[i] = switches[--n_switches];
179 /* Wait for something to happen. */
180 for (i = 0; i < n_switches; i++) {
181 struct switch_ *this = switches[i];
182 if (vconn_is_passive(this->vconn)) {
183 if (n_switches < MAX_SWITCHES) {
184 vconn_accept_wait(this->vconn);
187 vconn_recv_wait(this->vconn);
189 vconn_send_wait(this->vconn);
200 do_switch_recv(struct switch_ *this)
205 retval = vconn_recv(this->vconn, &msg);
207 process_packet(this, msg);
214 do_switch_send(struct switch_ *this)
218 struct buffer *next = this->txq.head->next;
219 retval = vconn_send(this->vconn, this->txq.head);
223 queue_advance_head(&this->txq, next);
230 connect_switch(const char *name)
235 retval = vconn_open(name, &vconn);
237 VLOG_ERR("%s: connect: %s", name, strerror(retval));
241 return new_switch(name, vconn);
244 static struct switch_ *
245 new_switch(const char *name, struct vconn *vconn)
247 struct switch_ *this = xmalloc(sizeof *this);
248 memset(this, 0, sizeof *this);
249 this->name = xstrdup(name);
251 queue_init(&this->txq);
252 this->last_control_hello = 0;
253 if (!vconn_is_passive(vconn)) {
254 send_control_hello(this);
260 close_switch(struct switch_ *this)
264 vconn_close(this->vconn);
265 queue_destroy(&this->txq);
271 send_control_hello(struct switch_ *this)
273 time_t now = time(0);
274 if (now >= this->last_control_hello + 1) {
276 struct ofp_control_hello *och;
279 och = buffer_put_uninit(b, sizeof *och);
280 memset(och, 0, sizeof *och);
281 och->header.version = OFP_VERSION;
282 och->header.length = htons(sizeof *och);
284 och->version = htonl(OFP_VERSION);
285 och->flags = htons(OFP_CHELLO_SEND_FLOW_EXP);
286 och->miss_send_len = htons(OFP_DEFAULT_MISS_SEND_LEN);
289 this->last_control_hello = now;
294 queue_tx(struct switch_ *this, struct buffer *b)
296 queue_push_tail(&this->txq, b);
300 process_packet(struct switch_ *sw, struct buffer *msg)
302 static const size_t min_size[UINT8_MAX + 1] = {
303 [0 ... UINT8_MAX] = SIZE_MAX,
304 [OFPT_CONTROL_HELLO] = sizeof (struct ofp_control_hello),
305 [OFPT_DATA_HELLO] = sizeof (struct ofp_data_hello),
306 [OFPT_PACKET_IN] = offsetof (struct ofp_packet_in, data),
307 [OFPT_PACKET_OUT] = sizeof (struct ofp_packet_out),
308 [OFPT_FLOW_MOD] = sizeof (struct ofp_flow_mod),
309 [OFPT_FLOW_EXPIRED] = sizeof (struct ofp_flow_expired),
310 [OFPT_TABLE] = sizeof (struct ofp_table),
311 [OFPT_PORT_MOD] = sizeof (struct ofp_port_mod),
312 [OFPT_PORT_STATUS] = sizeof (struct ofp_port_status),
313 [OFPT_FLOW_STAT_REQUEST] = sizeof (struct ofp_flow_stat_request),
314 [OFPT_FLOW_STAT_REPLY] = sizeof (struct ofp_flow_stat_reply),
316 struct ofp_header *oh;
319 if (msg->size < min_size[oh->type]) {
320 VLOG_WARN("%s: too short (%zu bytes) for type %"PRIu8" (min %zu)",
321 sw->name, msg->size, oh->type, min_size[oh->type]);
325 if (oh->type == OFPT_DATA_HELLO) {
326 struct ofp_data_hello *odh = msg->data;
327 sw->datapath_id = odh->datapath_id;
328 } else if (sw->datapath_id == 0) {
329 send_control_hello(sw);
333 if (oh->type == OFPT_PACKET_IN) {
334 if (sw->txq.n >= MAX_TXQ) {
335 /* FIXME: ratelimit. */
336 VLOG_WARN("%s: tx queue overflow", sw->name);
338 process_noflow(sw, msg->data);
340 process_hub(sw, msg->data);
342 process_switch(sw, msg->data);
347 ofp_print(stdout, msg->data, msg->size, 2);
351 process_hub(struct switch_ *sw, struct ofp_packet_in *opi)
353 size_t pkt_ofs, pkt_len;
357 /* Extract flow data from 'opi' into 'flow'. */
358 pkt_ofs = offsetof(struct ofp_packet_in, data);
359 pkt_len = ntohs(opi->header.length) - pkt_ofs;
360 pkt.data = opi->data;
362 flow_extract(&pkt, ntohs(opi->in_port), &flow);
365 queue_tx(sw, make_add_simple_flow(&flow, ntohl(opi->buffer_id),
368 /* If the switch didn't buffer the packet, we need to send a copy. */
369 if (ntohl(opi->buffer_id) == UINT32_MAX) {
370 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(flow.in_port),
376 process_noflow(struct switch_ *sw, struct ofp_packet_in *opi)
378 /* If the switch didn't buffer the packet, we need to send a copy. */
379 if (ntohl(opi->buffer_id) == UINT32_MAX) {
380 size_t pkt_ofs, pkt_len;
383 /* Extract flow data from 'opi' into 'flow'. */
384 pkt_ofs = offsetof(struct ofp_packet_in, data);
385 pkt_len = ntohs(opi->header.length) - pkt_ofs;
386 pkt.data = opi->data;
389 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(opi->in_port),
392 queue_tx(sw, make_buffered_packet_out(ntohl(opi->buffer_id),
393 ntohs(opi->in_port), OFPP_FLOOD));
398 #define MAC_HASH_BITS 10
399 #define MAC_HASH_MASK (MAC_HASH_SIZE - 1)
400 #define MAC_HASH_SIZE (1u << MAC_HASH_BITS)
405 struct list hash_list;
406 struct list lru_list;
407 uint64_t datapath_id;
408 uint8_t mac[ETH_ADDR_LEN];
412 static struct list mac_table[MAC_HASH_SIZE];
413 static struct list lrus;
414 static size_t mac_count;
422 for (i = 0; i < MAC_HASH_SIZE; i++) {
423 list_init(&mac_table[i]);
428 mac_table_bucket(uint64_t datapath_id, const uint8_t mac[ETH_ADDR_LEN])
431 hash = hash_fnv(&datapath_id, sizeof datapath_id, HASH_FNV_BASIS);
432 hash = hash_fnv(mac, ETH_ADDR_LEN, hash);
433 return &mac_table[hash & MAC_HASH_BITS];
437 process_switch(struct switch_ *sw, struct ofp_packet_in *opi)
439 size_t pkt_ofs, pkt_len;
445 /* Extract flow data from 'opi' into 'flow'. */
446 pkt_ofs = offsetof(struct ofp_packet_in, data);
447 pkt_len = ntohs(opi->header.length) - pkt_ofs;
448 pkt.data = opi->data;
450 flow_extract(&pkt, ntohs(opi->in_port), &flow);
452 /* Learn the source. */
453 if (!mac_is_multicast(flow.dl_src)) {
454 struct mac_source *src;
458 bucket = mac_table_bucket(sw->datapath_id, flow.dl_src);
460 LIST_FOR_EACH (src, struct mac_source, hash_list, bucket) {
461 if (src->datapath_id == sw->datapath_id
462 && mac_equals(src->mac, flow.dl_src)) {
469 /* Learn a new address. */
471 if (mac_count >= MAC_MAX) {
472 /* Drop the least recently used mac source. */
473 struct mac_source *lru;
474 lru = CONTAINER_OF(lrus.next, struct mac_source, lru_list);
475 list_remove(&lru->hash_list);
476 list_remove(&lru->lru_list);
482 /* Create new mac source */
483 src = xmalloc(sizeof *src);
484 src->datapath_id = sw->datapath_id;
485 memcpy(src->mac, flow.dl_src, ETH_ADDR_LEN);
487 list_push_front(bucket, &src->hash_list);
488 list_push_back(&lrus, &src->lru_list);
490 /* Make 'src' most-recently-used. */
491 list_remove(&src->lru_list);
492 list_push_back(&lrus, &src->lru_list);
495 if (ntohs(flow.in_port) != src->port) {
496 src->port = ntohs(flow.in_port);
497 VLOG_DBG("learned that "MAC_FMT" is on datapath %"PRIx64" port %d",
498 MAC_ARGS(src->mac), ntohll(src->datapath_id),
502 VLOG_DBG("multicast packet source "MAC_FMT, MAC_ARGS(flow.dl_src));
505 /* Figure out the destination. */
506 out_port = OFPP_FLOOD;
507 if (!mac_is_multicast(flow.dl_dst)) {
508 struct mac_source *dst;
511 bucket = mac_table_bucket(sw->datapath_id, flow.dl_dst);
512 LIST_FOR_EACH (dst, struct mac_source, hash_list, bucket) {
513 if (dst->datapath_id == sw->datapath_id
514 && mac_equals(dst->mac, flow.dl_dst)) {
515 out_port = dst->port;
521 if (out_port != OFPP_FLOOD) {
522 /* The output port is known, so add a new flow. */
523 queue_tx(sw, make_add_simple_flow(&flow, ntohl(opi->buffer_id),
526 /* If the switch didn't buffer the packet, we need to send a copy. */
527 if (ntohl(opi->buffer_id) == UINT32_MAX) {
528 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(flow.in_port),
532 /* We don't know that MAC. Flood the packet. */
534 if (ntohl(opi->buffer_id) == UINT32_MAX) {
535 b = make_unbuffered_packet_out(&pkt, ntohs(flow.in_port), out_port);
537 b = make_buffered_packet_out(ntohl(opi->buffer_id),
538 ntohs(flow.in_port), out_port);
545 parse_options(int argc, char *argv[])
547 static struct option long_options[] = {
548 {"hub", no_argument, 0, 'H'},
549 {"noflow", no_argument, 0, 'n'},
550 {"verbose", optional_argument, 0, 'v'},
551 {"help", no_argument, 0, 'h'},
552 {"version", no_argument, 0, 'V'},
554 {"private-key", required_argument, 0, 'p'},
555 {"certificate", required_argument, 0, 'c'},
556 {"ca-cert", required_argument, 0, 'C'},
560 char *short_options = long_options_to_short_options(long_options);
566 c = getopt_long(argc, argv, short_options, long_options, &indexptr);
584 printf("%s "VERSION" compiled "__DATE__" "__TIME__"\n", argv[0]);
588 vlog_set_verbosity(optarg);
593 vconn_ssl_set_private_key_file(optarg);
597 vconn_ssl_set_certificate_file(optarg);
601 vconn_ssl_set_ca_cert_file(optarg);
618 printf("%s: OpenFlow controller\n"
619 "usage: %s [OPTIONS] VCONN\n"
620 "where VCONN is one of the following:\n"
621 " ptcp:[PORT] listen to TCP PORT (default: %d)\n",
622 program_name, program_name, OFP_TCP_PORT);
624 printf(" nl:DP_IDX via netlink to local datapath DP_IDX\n");
627 printf(" pssl:[PORT] listen for SSL on PORT (default: %d)\n"
628 "\nPKI configuration (required to use SSL):\n"
629 " -p, --private-key=FILE file with private key\n"
630 " -c, --certificate=FILE file with certificate for private key\n"
631 " -C, --ca-cert=FILE file with peer CA certificate\n",
634 printf("\nOther options:\n"
635 " -H, --hub act as hub instead of learning switch\n"
636 " -n, --noflow pass traffic, but don't add flows\n"
637 " -v, --verbose set maximum verbosity level\n"
638 " -h, --help display this help message\n"
639 " -V, --version display version information\n");