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"
45 #include "vconn-ssl.h"
47 #include "vlog-socket.h"
51 #define THIS_MODULE VLM_controller
53 #define MAX_SWITCHES 16
61 time_t last_control_hello;
64 struct buffer *txq, *tx_tail;
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->next;
220 retval = vconn_send(this->vconn, this->txq);
226 if (this->txq == NULL) {
227 this->tx_tail = NULL;
236 connect_switch(const char *name)
241 retval = vconn_open(name, &vconn);
243 VLOG_ERR("%s: connect: %s", name, strerror(retval));
247 return new_switch(name, vconn);
250 static struct switch_ *
251 new_switch(const char *name, struct vconn *vconn)
253 struct switch_ *this = xmalloc(sizeof *this);
254 memset(this, 0, sizeof *this);
255 this->name = xstrdup(name);
259 this->tx_tail = NULL;
260 this->last_control_hello = 0;
261 if (!vconn_is_passive(vconn)) {
262 send_control_hello(this);
268 close_switch(struct switch_ *this)
271 struct buffer *cur, *next;
274 vconn_close(this->vconn);
275 for (cur = this->txq; cur != NULL; cur = next) {
284 send_control_hello(struct switch_ *this)
286 time_t now = time(0);
287 if (now >= this->last_control_hello + 1) {
289 struct ofp_control_hello *och;
292 och = buffer_put_uninit(b, sizeof *och);
293 memset(och, 0, sizeof *och);
294 och->header.version = OFP_VERSION;
295 och->header.length = htons(sizeof *och);
297 och->version = htonl(OFP_VERSION);
298 och->flags = htons(OFP_CHELLO_SEND_FLOW_EXP);
299 och->miss_send_len = htons(OFP_DEFAULT_MISS_SEND_LEN);
302 this->last_control_hello = now;
307 check_txq(struct switch_ *this UNUSED)
313 assert(this->n_txq == 0
314 ? this->txq == NULL && this->tx_tail == NULL
315 : this->txq != NULL && this->tx_tail != NULL);
318 for (iter = this->txq; iter != NULL; iter = iter->next) {
320 assert((iter->next != NULL) == (iter != this->tx_tail));
322 assert(n == this->n_txq);
327 queue_tx(struct switch_ *this, struct buffer *b)
333 this->tx_tail->next = b;
343 process_packet(struct switch_ *sw, struct buffer *msg)
345 static const size_t min_size[UINT8_MAX + 1] = {
346 [0 ... UINT8_MAX] = SIZE_MAX,
347 [OFPT_CONTROL_HELLO] = sizeof (struct ofp_control_hello),
348 [OFPT_DATA_HELLO] = sizeof (struct ofp_data_hello),
349 [OFPT_PACKET_IN] = offsetof (struct ofp_packet_in, data),
350 [OFPT_PACKET_OUT] = sizeof (struct ofp_packet_out),
351 [OFPT_FLOW_MOD] = sizeof (struct ofp_flow_mod),
352 [OFPT_FLOW_EXPIRED] = sizeof (struct ofp_flow_expired),
353 [OFPT_TABLE] = sizeof (struct ofp_table),
354 [OFPT_PORT_MOD] = sizeof (struct ofp_port_mod),
355 [OFPT_PORT_STATUS] = sizeof (struct ofp_port_status),
356 [OFPT_FLOW_STAT_REQUEST] = sizeof (struct ofp_flow_stat_request),
357 [OFPT_FLOW_STAT_REPLY] = sizeof (struct ofp_flow_stat_reply),
359 struct ofp_header *oh;
362 if (msg->size < min_size[oh->type]) {
363 VLOG_WARN("%s: too short (%zu bytes) for type %"PRIu8" (min %zu)",
364 sw->name, msg->size, oh->type, min_size[oh->type]);
368 if (oh->type == OFPT_DATA_HELLO) {
369 struct ofp_data_hello *odh = msg->data;
370 sw->datapath_id = odh->datapath_id;
371 } else if (sw->datapath_id == 0) {
372 send_control_hello(sw);
376 if (oh->type == OFPT_PACKET_IN) {
377 if (sw->n_txq >= MAX_TXQ) {
378 VLOG_WARN("%s: tx queue overflow", sw->name);
380 process_noflow(sw, msg->data);
382 process_hub(sw, msg->data);
384 process_switch(sw, msg->data);
389 ofp_print(stdout, msg->data, msg->size, 2);
393 process_hub(struct switch_ *sw, struct ofp_packet_in *opi)
395 size_t pkt_ofs, pkt_len;
399 /* Extract flow data from 'opi' into 'flow'. */
400 pkt_ofs = offsetof(struct ofp_packet_in, data);
401 pkt_len = ntohs(opi->header.length) - pkt_ofs;
402 pkt.data = opi->data;
404 flow_extract(&pkt, ntohs(opi->in_port), &flow);
407 queue_tx(sw, make_add_simple_flow(&flow, ntohl(opi->buffer_id),
410 /* If the switch didn't buffer the packet, we need to send a copy. */
411 if (ntohl(opi->buffer_id) == UINT32_MAX) {
412 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(flow.in_port),
418 process_noflow(struct switch_ *sw, struct ofp_packet_in *opi)
420 /* If the switch didn't buffer the packet, we need to send a copy. */
421 if (ntohl(opi->buffer_id) == UINT32_MAX) {
422 size_t pkt_ofs, pkt_len;
425 /* Extract flow data from 'opi' into 'flow'. */
426 pkt_ofs = offsetof(struct ofp_packet_in, data);
427 pkt_len = ntohs(opi->header.length) - pkt_ofs;
428 pkt.data = opi->data;
431 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(opi->in_port),
434 queue_tx(sw, make_buffered_packet_out(ntohl(opi->buffer_id),
435 ntohs(opi->in_port), OFPP_FLOOD));
440 #define MAC_HASH_BITS 10
441 #define MAC_HASH_MASK (MAC_HASH_SIZE - 1)
442 #define MAC_HASH_SIZE (1u << MAC_HASH_BITS)
447 struct list hash_list;
448 struct list lru_list;
449 uint64_t datapath_id;
450 uint8_t mac[ETH_ADDR_LEN];
454 static struct list mac_table[MAC_HASH_SIZE];
455 static struct list lrus;
456 static size_t mac_count;
464 for (i = 0; i < MAC_HASH_SIZE; i++) {
465 list_init(&mac_table[i]);
470 mac_table_bucket(uint64_t datapath_id, const uint8_t mac[ETH_ADDR_LEN])
473 hash = hash_fnv(&datapath_id, sizeof datapath_id, HASH_FNV_BASIS);
474 hash = hash_fnv(mac, ETH_ADDR_LEN, hash);
475 return &mac_table[hash & MAC_HASH_BITS];
479 process_switch(struct switch_ *sw, struct ofp_packet_in *opi)
481 size_t pkt_ofs, pkt_len;
487 /* Extract flow data from 'opi' into 'flow'. */
488 pkt_ofs = offsetof(struct ofp_packet_in, data);
489 pkt_len = ntohs(opi->header.length) - pkt_ofs;
490 pkt.data = opi->data;
492 flow_extract(&pkt, ntohs(opi->in_port), &flow);
494 /* Learn the source. */
495 if (!mac_is_multicast(flow.dl_src)) {
496 struct mac_source *src;
500 bucket = mac_table_bucket(sw->datapath_id, flow.dl_src);
502 LIST_FOR_EACH (src, struct mac_source, hash_list, bucket) {
503 if (src->datapath_id == sw->datapath_id
504 && mac_equals(src->mac, flow.dl_src)) {
511 /* Learn a new address. */
513 if (mac_count >= MAC_MAX) {
514 /* Drop the least recently used mac source. */
515 struct mac_source *lru;
516 lru = CONTAINER_OF(lrus.next, struct mac_source, lru_list);
517 list_remove(&lru->hash_list);
518 list_remove(&lru->lru_list);
524 /* Create new mac source */
525 src = xmalloc(sizeof *src);
526 src->datapath_id = sw->datapath_id;
527 memcpy(src->mac, flow.dl_src, ETH_ADDR_LEN);
529 list_push_front(bucket, &src->hash_list);
530 list_push_back(&lrus, &src->lru_list);
532 /* Make 'src' most-recently-used. */
533 list_remove(&src->lru_list);
534 list_push_back(&lrus, &src->lru_list);
537 if (ntohs(flow.in_port) != src->port) {
538 src->port = ntohs(flow.in_port);
539 VLOG_DBG("learned that "MAC_FMT" is on datapath %"PRIx64" port %d",
540 MAC_ARGS(src->mac), ntohll(src->datapath_id),
544 VLOG_DBG("multicast packet source "MAC_FMT, MAC_ARGS(flow.dl_src));
547 /* Figure out the destination. */
548 out_port = OFPP_FLOOD;
549 if (!mac_is_multicast(flow.dl_dst)) {
550 struct mac_source *dst;
553 bucket = mac_table_bucket(sw->datapath_id, flow.dl_dst);
554 LIST_FOR_EACH (dst, struct mac_source, hash_list, bucket) {
555 if (dst->datapath_id == sw->datapath_id
556 && mac_equals(dst->mac, flow.dl_dst)) {
557 out_port = dst->port;
563 if (out_port != OFPP_FLOOD) {
564 /* The output port is known, so add a new flow. */
565 queue_tx(sw, make_add_simple_flow(&flow, ntohl(opi->buffer_id),
568 /* If the switch didn't buffer the packet, we need to send a copy. */
569 if (ntohl(opi->buffer_id) == UINT32_MAX) {
570 queue_tx(sw, make_unbuffered_packet_out(&pkt, ntohs(flow.in_port),
574 /* We don't know that MAC. Flood the packet. */
576 if (ntohl(opi->buffer_id) == UINT32_MAX) {
577 b = make_unbuffered_packet_out(&pkt, ntohs(flow.in_port), out_port);
579 b = make_buffered_packet_out(ntohl(opi->buffer_id),
580 ntohs(flow.in_port), out_port);
587 parse_options(int argc, char *argv[])
589 static struct option long_options[] = {
590 {"hub", no_argument, 0, 'H'},
591 {"noflow", no_argument, 0, 'n'},
592 {"verbose", optional_argument, 0, 'v'},
593 {"help", no_argument, 0, 'h'},
594 {"version", no_argument, 0, 'V'},
596 {"private-key", required_argument, 0, 'p'},
597 {"certificate", required_argument, 0, 'c'},
598 {"ca-cert", required_argument, 0, 'C'},
602 char *short_options = long_options_to_short_options(long_options);
608 c = getopt_long(argc, argv, short_options, long_options, &indexptr);
626 printf("%s "VERSION" compiled "__DATE__" "__TIME__"\n", argv[0]);
630 vlog_set_verbosity(optarg);
635 vconn_ssl_set_private_key_file(optarg);
639 vconn_ssl_set_certificate_file(optarg);
643 vconn_ssl_set_ca_cert_file(optarg);
660 printf("%s: OpenFlow controller\n"
661 "usage: %s [OPTIONS] VCONN\n"
662 "where VCONN is one of the following:\n"
663 " ptcp:[PORT] listen to TCP PORT (default: %d)\n",
664 program_name, program_name, OFP_TCP_PORT);
666 printf(" nl:DP_IDX via netlink to local datapath DP_IDX\n");
669 printf(" pssl:[PORT] listen for SSL on PORT (default: %d)\n"
670 "\nPKI configuration (required to use SSL):\n"
671 " -p, --private-key=FILE file with private key\n"
672 " -c, --certificate=FILE file with certificate for private key\n"
673 " -C, --ca-cert=FILE file with peer CA certificate\n",
676 printf("\nOther options:\n"
677 " -H, --hub act as hub instead of learning switch\n"
678 " -n, --noflow pass traffic, but don't add flows\n"
679 " -v, --verbose set maximum verbosity level\n"
680 " -h, --help display this help message\n"
681 " -V, --version display version information\n");