]>
git.proxmox.com Git - mirror_ovs.git/blob - ofproto/in-band.c
2 * Copyright (c) 2008, 2009, 2010 Nicira Networks.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include <arpa/inet.h>
22 #include <sys/socket.h>
33 #include "openflow/openflow.h"
35 #include "poll-loop.h"
40 VLOG_DEFINE_THIS_MODULE(in_band
)
42 /* In-band control allows a single network to be used for OpenFlow
43 * traffic and other data traffic. Refer to ovs-vswitchd.conf(5) and
44 * secchan(8) for a description of configuring in-band control.
46 * This comment is an attempt to describe how in-band control works at a
47 * wire- and implementation-level. Correctly implementing in-band
48 * control has proven difficult due to its many subtleties, and has thus
49 * gone through many iterations. Please read through and understand the
50 * reasoning behind the chosen rules before making modifications.
52 * In Open vSwitch, in-band control is implemented as "hidden" flows (in that
53 * they are not visible through OpenFlow) and at a higher priority than
54 * wildcarded flows can be set up by through OpenFlow. This is done so that
55 * the OpenFlow controller cannot interfere with them and possibly break
56 * connectivity with its switches. It is possible to see all flows, including
57 * in-band ones, with the ovs-appctl "bridge/dump-flows" command.
59 * The Open vSwitch implementation of in-band control can hide traffic to
60 * arbitrary "remotes", where each remote is one TCP port on one IP address.
61 * Currently the remotes are automatically configured as the in-band OpenFlow
62 * controllers plus the OVSDB managers, if any. (The latter is a requirement
63 * because OVSDB managers are responsible for configuring OpenFlow controllers,
64 * so if the manager cannot be reached then OpenFlow cannot be reconfigured.)
66 * The following rules (with the OFPP_NORMAL action) are set up on any bridge
67 * that has any remotes:
69 * (a) DHCP requests sent from the local port.
70 * (b) ARP replies to the local port's MAC address.
71 * (c) ARP requests from the local port's MAC address.
73 * In-band also sets up the following rules for each unique next-hop MAC
74 * address for the remotes' IPs (the "next hop" is either the remote
75 * itself, if it is on a local subnet, or the gateway to reach the remote):
77 * (d) ARP replies to the next hop's MAC address.
78 * (e) ARP requests from the next hop's MAC address.
80 * In-band also sets up the following rules for each unique remote IP address:
82 * (f) ARP replies containing the remote's IP address as a target.
83 * (g) ARP requests containing the remote's IP address as a source.
85 * In-band also sets up the following rules for each unique remote (IP,port)
88 * (h) TCP traffic to the remote's IP and port.
89 * (i) TCP traffic from the remote's IP and port.
91 * The goal of these rules is to be as narrow as possible to allow a
92 * switch to join a network and be able to communicate with the
93 * remotes. As mentioned earlier, these rules have higher priority
94 * than the controller's rules, so if they are too broad, they may
95 * prevent the controller from implementing its policy. As such,
96 * in-band actively monitors some aspects of flow and packet processing
97 * so that the rules can be made more precise.
99 * In-band control monitors attempts to add flows into the datapath that
100 * could interfere with its duties. The datapath only allows exact
101 * match entries, so in-band control is able to be very precise about
102 * the flows it prevents. Flows that miss in the datapath are sent to
103 * userspace to be processed, so preventing these flows from being
104 * cached in the "fast path" does not affect correctness. The only type
105 * of flow that is currently prevented is one that would prevent DHCP
106 * replies from being seen by the local port. For example, a rule that
107 * forwarded all DHCP traffic to the controller would not be allowed,
108 * but one that forwarded to all ports (including the local port) would.
110 * As mentioned earlier, packets that miss in the datapath are sent to
111 * the userspace for processing. The userspace has its own flow table,
112 * the "classifier", so in-band checks whether any special processing
113 * is needed before the classifier is consulted. If a packet is a DHCP
114 * response to a request from the local port, the packet is forwarded to
115 * the local port, regardless of the flow table. Note that this requires
116 * L7 processing of DHCP replies to determine whether the 'chaddr' field
117 * matches the MAC address of the local port.
119 * It is interesting to note that for an L3-based in-band control
120 * mechanism, the majority of rules are devoted to ARP traffic. At first
121 * glance, some of these rules appear redundant. However, each serves an
122 * important role. First, in order to determine the MAC address of the
123 * remote side (controller or gateway) for other ARP rules, we must allow
124 * ARP traffic for our local port with rules (b) and (c). If we are
125 * between a switch and its connection to the remote, we have to
126 * allow the other switch's ARP traffic to through. This is done with
127 * rules (d) and (e), since we do not know the addresses of the other
128 * switches a priori, but do know the remote's or gateway's. Finally,
129 * if the remote is running in a local guest VM that is not reached
130 * through the local port, the switch that is connected to the VM must
131 * allow ARP traffic based on the remote's IP address, since it will
132 * not know the MAC address of the local port that is sending the traffic
133 * or the MAC address of the remote in the guest VM.
135 * With a few notable exceptions below, in-band should work in most
136 * network setups. The following are considered "supported' in the
137 * current implementation:
139 * - Locally Connected. The switch and remote are on the same
140 * subnet. This uses rules (a), (b), (c), (h), and (i).
142 * - Reached through Gateway. The switch and remote are on
143 * different subnets and must go through a gateway. This uses
144 * rules (a), (b), (c), (h), and (i).
146 * - Between Switch and Remote. This switch is between another
147 * switch and the remote, and we want to allow the other
148 * switch's traffic through. This uses rules (d), (e), (h), and
149 * (i). It uses (b) and (c) indirectly in order to know the MAC
150 * address for rules (d) and (e). Note that DHCP for the other
151 * switch will not work unless an OpenFlow controller explicitly lets this
152 * switch pass the traffic.
154 * - Between Switch and Gateway. This switch is between another
155 * switch and the gateway, and we want to allow the other switch's
156 * traffic through. This uses the same rules and logic as the
157 * "Between Switch and Remote" configuration described earlier.
159 * - Remote on Local VM. The remote is a guest VM on the
160 * system running in-band control. This uses rules (a), (b), (c),
163 * - Remote on Local VM with Different Networks. The remote
164 * is a guest VM on the system running in-band control, but the
165 * local port is not used to connect to the remote. For
166 * example, an IP address is configured on eth0 of the switch. The
167 * remote's VM is connected through eth1 of the switch, but an
168 * IP address has not been configured for that port on the switch.
169 * As such, the switch will use eth0 to connect to the remote,
170 * and eth1's rules about the local port will not work. In the
171 * example, the switch attached to eth0 would use rules (a), (b),
172 * (c), (h), and (i) on eth0. The switch attached to eth1 would use
173 * rules (f), (g), (h), and (i).
175 * The following are explicitly *not* supported by in-band control:
177 * - Specify Remote by Name. Currently, the remote must be
178 * identified by IP address. A naive approach would be to permit
179 * all DNS traffic. Unfortunately, this would prevent the
180 * controller from defining any policy over DNS. Since switches
181 * that are located behind us need to connect to the remote,
182 * in-band cannot simply add a rule that allows DNS traffic from
183 * the local port. The "correct" way to support this is to parse
184 * DNS requests to allow all traffic related to a request for the
185 * remote's name through. Due to the potential security
186 * problems and amount of processing, we decided to hold off for
189 * - Differing Remotes for Switches. All switches must know
190 * the L3 addresses for all the remotes that other switches
191 * may use, since rules need to be set up to allow traffic related
192 * to those remotes through. See rules (f), (g), (h), and (i).
194 * - Differing Routes for Switches. In order for the switch to
195 * allow other switches to connect to a remote through a
196 * gateway, it allows the gateway's traffic through with rules (d)
197 * and (e). If the routes to the remote differ for the two
198 * switches, we will not know the MAC address of the alternate
202 /* Priorities used in classifier for in-band rules. These values are higher
203 * than any that may be set with OpenFlow, and "18" kind of looks like "IB".
204 * The ordering of priorities is not important because all of the rules set up
205 * by in-band control have the same action. The only reason to use more than
206 * one priority is to make the kind of flow easier to see during debugging. */
208 /* One set per bridge. */
209 IBR_FROM_LOCAL_DHCP
= 180000, /* (a) From local port, DHCP. */
210 IBR_TO_LOCAL_ARP
, /* (b) To local port, ARP. */
211 IBR_FROM_LOCAL_ARP
, /* (c) From local port, ARP. */
213 /* One set per unique next-hop MAC. */
214 IBR_TO_NEXT_HOP_ARP
, /* (d) To remote MAC, ARP. */
215 IBR_FROM_NEXT_HOP_ARP
, /* (e) From remote MAC, ARP. */
217 /* One set per unique remote IP address. */
218 IBR_TO_REMOTE_ARP
, /* (f) To remote IP, ARP. */
219 IBR_FROM_REMOTE_ARP
, /* (g) From remote IP, ARP. */
221 /* One set per unique remote (IP,port) pair. */
222 IBR_TO_REMOTE_TCP
, /* (h) To remote IP, TCP port. */
223 IBR_FROM_REMOTE_TCP
/* (i) From remote IP, TCP port. */
226 struct in_band_rule
{
229 unsigned int priority
;
232 /* Track one remote IP and next hop information. */
233 struct in_band_remote
{
234 struct sockaddr_in remote_addr
; /* IP address, in network byte order. */
235 uint8_t remote_mac
[ETH_ADDR_LEN
]; /* Next-hop MAC, all-zeros if unknown. */
236 uint8_t last_remote_mac
[ETH_ADDR_LEN
]; /* Previous nonzero next-hop MAC. */
237 struct netdev
*remote_netdev
; /* Device to send to next-hop MAC. */
241 struct ofproto
*ofproto
;
242 struct status_category
*ss_cat
;
244 /* Remote information. */
245 time_t next_remote_refresh
; /* Refresh timer. */
246 struct in_band_remote
*remotes
;
249 /* Local information. */
250 time_t next_local_refresh
; /* Refresh timer. */
251 uint8_t local_mac
[ETH_ADDR_LEN
]; /* Current MAC. */
252 struct netdev
*local_netdev
; /* Local port's network device. */
254 /* Local and remote addresses that are installed as flows. */
255 uint8_t installed_local_mac
[ETH_ADDR_LEN
];
256 struct sockaddr_in
*remote_addrs
;
257 size_t n_remote_addrs
;
258 uint8_t *remote_macs
;
259 size_t n_remote_macs
;
262 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(60, 60);
265 refresh_remote(struct in_band
*ib
, struct in_band_remote
*r
)
267 struct in_addr next_hop_inaddr
;
271 /* Find the next-hop IP address. */
272 memset(r
->remote_mac
, 0, sizeof r
->remote_mac
);
273 retval
= netdev_get_next_hop(ib
->local_netdev
, &r
->remote_addr
.sin_addr
,
274 &next_hop_inaddr
, &next_hop_dev
);
276 VLOG_WARN("cannot find route for controller ("IP_FMT
"): %s",
277 IP_ARGS(&r
->remote_addr
.sin_addr
), strerror(retval
));
280 if (!next_hop_inaddr
.s_addr
) {
281 next_hop_inaddr
= r
->remote_addr
.sin_addr
;
284 /* Open the next-hop network device. */
285 if (!r
->remote_netdev
286 || strcmp(netdev_get_name(r
->remote_netdev
), next_hop_dev
))
288 netdev_close(r
->remote_netdev
);
290 retval
= netdev_open_default(next_hop_dev
, &r
->remote_netdev
);
292 VLOG_WARN_RL(&rl
, "cannot open netdev %s (next hop "
293 "to controller "IP_FMT
"): %s",
294 next_hop_dev
, IP_ARGS(&r
->remote_addr
.sin_addr
),
302 /* Look up the MAC address of the next-hop IP address. */
303 retval
= netdev_arp_lookup(r
->remote_netdev
, next_hop_inaddr
.s_addr
,
306 VLOG_DBG_RL(&rl
, "cannot look up remote MAC address ("IP_FMT
"): %s",
307 IP_ARGS(&next_hop_inaddr
.s_addr
), strerror(retval
));
310 /* If we don't have a MAC address, then refresh quickly, since we probably
311 * will get a MAC address soon (via ARP). Otherwise, we can afford to wait
313 return eth_addr_is_zero(r
->remote_mac
) ? 1 : 10;
317 refresh_remotes(struct in_band
*ib
)
319 struct in_band_remote
*r
;
322 if (time_now() < ib
->next_remote_refresh
) {
327 ib
->next_remote_refresh
= TIME_MAX
;
328 for (r
= ib
->remotes
; r
< &ib
->remotes
[ib
->n_remotes
]; r
++) {
329 uint8_t old_remote_mac
[ETH_ADDR_LEN
];
333 memcpy(old_remote_mac
, r
->remote_mac
, ETH_ADDR_LEN
);
335 /* Refresh remote information. */
336 next_refresh
= refresh_remote(ib
, r
) + time_now();
337 ib
->next_remote_refresh
= MIN(ib
->next_remote_refresh
, next_refresh
);
339 /* If the MAC changed, log the changes. */
340 if (!eth_addr_equals(r
->remote_mac
, old_remote_mac
)) {
342 if (!eth_addr_is_zero(r
->remote_mac
)
343 && !eth_addr_equals(r
->last_remote_mac
, r
->remote_mac
)) {
344 VLOG_DBG("remote MAC address changed from "ETH_ADDR_FMT
346 ETH_ADDR_ARGS(r
->last_remote_mac
),
347 ETH_ADDR_ARGS(r
->remote_mac
));
348 memcpy(r
->last_remote_mac
, r
->remote_mac
, ETH_ADDR_LEN
);
356 /* Refreshes the MAC address of the local port into ib->local_mac, if it is due
357 * for a refresh. Returns true if anything changed, otherwise false. */
359 refresh_local(struct in_band
*ib
)
361 uint8_t ea
[ETH_ADDR_LEN
];
365 if (now
< ib
->next_local_refresh
) {
368 ib
->next_local_refresh
= now
+ 1;
370 if (netdev_get_etheraddr(ib
->local_netdev
, ea
)
371 || eth_addr_equals(ea
, ib
->local_mac
)) {
375 memcpy(ib
->local_mac
, ea
, ETH_ADDR_LEN
);
380 in_band_status_cb(struct status_reply
*sr
, void *in_band_
)
382 struct in_band
*in_band
= in_band_
;
384 if (!eth_addr_is_zero(in_band
->local_mac
)) {
385 status_reply_put(sr
, "local-mac="ETH_ADDR_FMT
,
386 ETH_ADDR_ARGS(in_band
->local_mac
));
389 if (in_band
->n_remotes
390 && !eth_addr_is_zero(in_band
->remotes
[0].remote_mac
)) {
391 status_reply_put(sr
, "remote-mac="ETH_ADDR_FMT
,
392 ETH_ADDR_ARGS(in_band
->remotes
[0].remote_mac
));
396 /* Returns true if 'packet' should be sent to the local port regardless
397 * of the flow table. */
399 in_band_msg_in_hook(struct in_band
*in_band
, const flow_t
*flow
,
400 const struct ofpbuf
*packet
)
406 /* Regardless of how the flow table is configured, we want to be
407 * able to see replies to our DHCP requests. */
408 if (flow
->dl_type
== htons(ETH_TYPE_IP
)
409 && flow
->nw_proto
== IP_TYPE_UDP
410 && flow
->tp_src
== htons(DHCP_SERVER_PORT
)
411 && flow
->tp_dst
== htons(DHCP_CLIENT_PORT
)
413 struct dhcp_header
*dhcp
;
415 dhcp
= ofpbuf_at(packet
, (char *)packet
->l7
- (char *)packet
->data
,
421 refresh_local(in_band
);
422 if (!eth_addr_is_zero(in_band
->local_mac
)
423 && eth_addr_equals(dhcp
->chaddr
, in_band
->local_mac
)) {
431 /* Returns true if the rule that would match 'flow' with 'actions' is
432 * allowed to be set up in the datapath. */
434 in_band_rule_check(struct in_band
*in_band
, const flow_t
*flow
,
435 const struct odp_actions
*actions
)
441 /* Don't allow flows that would prevent DHCP replies from being seen
442 * by the local port. */
443 if (flow
->dl_type
== htons(ETH_TYPE_IP
)
444 && flow
->nw_proto
== IP_TYPE_UDP
445 && flow
->tp_src
== htons(DHCP_SERVER_PORT
)
446 && flow
->tp_dst
== htons(DHCP_CLIENT_PORT
)) {
449 for (i
=0; i
<actions
->n_actions
; i
++) {
450 if (actions
->actions
[i
].output
.type
== ODPAT_OUTPUT
451 && actions
->actions
[i
].output
.port
== ODPP_LOCAL
) {
462 init_rule(struct in_band_rule
*rule
, unsigned int priority
)
464 rule
->wildcards
= OVSFW_ALL
;
465 rule
->priority
= priority
;
467 /* Not strictly necessary but seems cleaner. */
468 memset(&rule
->flow
, 0, sizeof rule
->flow
);
472 set_in_port(struct in_band_rule
*rule
, uint16_t odp_port
)
474 rule
->wildcards
&= ~OFPFW_IN_PORT
;
475 rule
->flow
.in_port
= odp_port
;
479 set_dl_type(struct in_band_rule
*rule
, uint16_t dl_type
)
481 rule
->wildcards
&= ~OFPFW_DL_TYPE
;
482 rule
->flow
.dl_type
= dl_type
;
486 set_dl_src(struct in_band_rule
*rule
, const uint8_t dl_src
[ETH_ADDR_LEN
])
488 rule
->wildcards
&= ~OFPFW_DL_SRC
;
489 memcpy(rule
->flow
.dl_src
, dl_src
, ETH_ADDR_LEN
);
493 set_dl_dst(struct in_band_rule
*rule
, const uint8_t dl_dst
[ETH_ADDR_LEN
])
495 rule
->wildcards
&= ~OFPFW_DL_DST
;
496 memcpy(rule
->flow
.dl_dst
, dl_dst
, ETH_ADDR_LEN
);
500 set_tp_src(struct in_band_rule
*rule
, uint16_t tp_src
)
502 rule
->wildcards
&= ~OFPFW_TP_SRC
;
503 rule
->flow
.tp_src
= tp_src
;
507 set_tp_dst(struct in_band_rule
*rule
, uint16_t tp_dst
)
509 rule
->wildcards
&= ~OFPFW_TP_DST
;
510 rule
->flow
.tp_dst
= tp_dst
;
514 set_nw_proto(struct in_band_rule
*rule
, uint8_t nw_proto
)
516 rule
->wildcards
&= ~OFPFW_NW_PROTO
;
517 rule
->flow
.nw_proto
= nw_proto
;
521 set_nw_src(struct in_band_rule
*rule
, const struct in_addr nw_src
)
523 rule
->wildcards
&= ~OFPFW_NW_SRC_MASK
;
524 rule
->flow
.nw_src
= nw_src
.s_addr
;
528 set_nw_dst(struct in_band_rule
*rule
, const struct in_addr nw_dst
)
530 rule
->wildcards
&= ~OFPFW_NW_DST_MASK
;
531 rule
->flow
.nw_dst
= nw_dst
.s_addr
;
535 make_rules(struct in_band
*ib
,
536 void (*cb
)(struct in_band
*, const struct in_band_rule
*))
538 struct in_band_rule rule
;
541 if (!eth_addr_is_zero(ib
->installed_local_mac
)) {
542 /* (a) Allow DHCP requests sent from the local port. */
543 init_rule(&rule
, IBR_FROM_LOCAL_DHCP
);
544 set_in_port(&rule
, ODPP_LOCAL
);
545 set_dl_type(&rule
, htons(ETH_TYPE_IP
));
546 set_dl_src(&rule
, ib
->installed_local_mac
);
547 set_nw_proto(&rule
, IP_TYPE_UDP
);
548 set_tp_src(&rule
, htons(DHCP_CLIENT_PORT
));
549 set_tp_dst(&rule
, htons(DHCP_SERVER_PORT
));
552 /* (b) Allow ARP replies to the local port's MAC address. */
553 init_rule(&rule
, IBR_TO_LOCAL_ARP
);
554 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
555 set_dl_dst(&rule
, ib
->installed_local_mac
);
556 set_nw_proto(&rule
, ARP_OP_REPLY
);
559 /* (c) Allow ARP requests from the local port's MAC address. */
560 init_rule(&rule
, IBR_FROM_LOCAL_ARP
);
561 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
562 set_dl_src(&rule
, ib
->installed_local_mac
);
563 set_nw_proto(&rule
, ARP_OP_REQUEST
);
567 for (i
= 0; i
< ib
->n_remote_macs
; i
++) {
568 const uint8_t *remote_mac
= &ib
->remote_macs
[i
* ETH_ADDR_LEN
];
571 const uint8_t *prev_mac
= &ib
->remote_macs
[(i
- 1) * ETH_ADDR_LEN
];
572 if (eth_addr_equals(remote_mac
, prev_mac
)) {
573 /* Skip duplicates. */
578 /* (d) Allow ARP replies to the next hop's MAC address. */
579 init_rule(&rule
, IBR_TO_NEXT_HOP_ARP
);
580 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
581 set_dl_dst(&rule
, remote_mac
);
582 set_nw_proto(&rule
, ARP_OP_REPLY
);
585 /* (e) Allow ARP requests from the next hop's MAC address. */
586 init_rule(&rule
, IBR_FROM_NEXT_HOP_ARP
);
587 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
588 set_dl_src(&rule
, remote_mac
);
589 set_nw_proto(&rule
, ARP_OP_REQUEST
);
593 for (i
= 0; i
< ib
->n_remote_addrs
; i
++) {
594 const struct sockaddr_in
*a
= &ib
->remote_addrs
[i
];
596 if (!i
|| a
->sin_addr
.s_addr
!= a
[-1].sin_addr
.s_addr
) {
597 /* (f) Allow ARP replies containing the remote's IP address as a
599 init_rule(&rule
, IBR_TO_REMOTE_ARP
);
600 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
601 set_nw_proto(&rule
, ARP_OP_REPLY
);
602 set_nw_dst(&rule
, a
->sin_addr
);
605 /* (g) Allow ARP requests containing the remote's IP address as a
607 init_rule(&rule
, IBR_FROM_REMOTE_ARP
);
608 set_dl_type(&rule
, htons(ETH_TYPE_ARP
));
609 set_nw_proto(&rule
, ARP_OP_REQUEST
);
610 set_nw_src(&rule
, a
->sin_addr
);
615 || a
->sin_addr
.s_addr
!= a
[-1].sin_addr
.s_addr
616 || a
->sin_port
!= a
[-1].sin_port
) {
617 /* (h) Allow TCP traffic to the remote's IP and port. */
618 init_rule(&rule
, IBR_TO_REMOTE_TCP
);
619 set_dl_type(&rule
, htons(ETH_TYPE_IP
));
620 set_nw_proto(&rule
, IP_TYPE_TCP
);
621 set_nw_dst(&rule
, a
->sin_addr
);
622 set_tp_dst(&rule
, a
->sin_port
);
625 /* (i) Allow TCP traffic from the remote's IP and port. */
626 init_rule(&rule
, IBR_FROM_REMOTE_TCP
);
627 set_dl_type(&rule
, htons(ETH_TYPE_IP
));
628 set_nw_proto(&rule
, IP_TYPE_TCP
);
629 set_nw_src(&rule
, a
->sin_addr
);
630 set_tp_src(&rule
, a
->sin_port
);
637 drop_rule(struct in_band
*ib
, const struct in_band_rule
*rule
)
639 ofproto_delete_flow(ib
->ofproto
, &rule
->flow
,
640 rule
->wildcards
, rule
->priority
);
643 /* Drops from the flow table all of the flows set up by 'ib', then clears out
644 * the information about the installed flows so that they can be filled in
645 * again if necessary. */
647 drop_rules(struct in_band
*ib
)
650 make_rules(ib
, drop_rule
);
652 /* Clear out state. */
653 memset(ib
->installed_local_mac
, 0, sizeof ib
->installed_local_mac
);
655 free(ib
->remote_addrs
);
656 ib
->remote_addrs
= NULL
;
657 ib
->n_remote_addrs
= 0;
659 free(ib
->remote_macs
);
660 ib
->remote_macs
= NULL
;
661 ib
->n_remote_macs
= 0;
665 add_rule(struct in_band
*ib
, const struct in_band_rule
*rule
)
667 union ofp_action action
;
669 action
.type
= htons(OFPAT_OUTPUT
);
670 action
.output
.len
= htons(sizeof action
);
671 action
.output
.port
= htons(OFPP_NORMAL
);
672 action
.output
.max_len
= htons(0);
673 ofproto_add_flow(ib
->ofproto
, &rule
->flow
, rule
->wildcards
,
674 rule
->priority
, &action
, 1, 0);
677 /* Inserts flows into the flow table for the current state of 'ib'. */
679 add_rules(struct in_band
*ib
)
681 make_rules(ib
, add_rule
);
685 compare_addrs(const void *a_
, const void *b_
)
687 const struct sockaddr_in
*a
= a_
;
688 const struct sockaddr_in
*b
= b_
;
691 cmp
= memcmp(&a
->sin_addr
.s_addr
,
693 sizeof a
->sin_addr
.s_addr
);
697 return memcmp(&a
->sin_port
, &b
->sin_port
, sizeof a
->sin_port
);
701 compare_macs(const void *a
, const void *b
)
703 return memcmp(a
, b
, ETH_ADDR_LEN
);
707 in_band_run(struct in_band
*ib
)
709 struct in_band_remote
*r
;
710 bool local_change
, remote_change
;
712 local_change
= refresh_local(ib
);
713 remote_change
= refresh_remotes(ib
);
714 if (!local_change
&& !remote_change
) {
715 /* Nothing changed, nothing to do. */
719 /* Drop old rules. */
722 /* Figure out new rules. */
723 memcpy(ib
->installed_local_mac
, ib
->local_mac
, ETH_ADDR_LEN
);
724 ib
->remote_addrs
= xmalloc(ib
->n_remotes
* sizeof *ib
->remote_addrs
);
725 ib
->n_remote_addrs
= 0;
726 ib
->remote_macs
= xmalloc(ib
->n_remotes
* ETH_ADDR_LEN
);
727 ib
->n_remote_macs
= 0;
728 for (r
= ib
->remotes
; r
< &ib
->remotes
[ib
->n_remotes
]; r
++) {
729 ib
->remote_addrs
[ib
->n_remote_addrs
++] = r
->remote_addr
;
730 if (!eth_addr_is_zero(r
->remote_mac
)) {
731 memcpy(&ib
->remote_macs
[ib
->n_remote_macs
* ETH_ADDR_LEN
],
732 r
->remote_mac
, ETH_ADDR_LEN
);
737 /* Sort, to allow make_rules() to easily skip duplicates. */
738 qsort(ib
->remote_addrs
, ib
->n_remote_addrs
, sizeof *ib
->remote_addrs
,
740 qsort(ib
->remote_macs
, ib
->n_remote_macs
, ETH_ADDR_LEN
, compare_macs
);
747 in_band_wait(struct in_band
*in_band
)
750 = MIN(in_band
->next_remote_refresh
, in_band
->next_local_refresh
);
751 poll_timer_wait_until(wakeup
* 1000);
754 /* ofproto has flushed all flows from the flow table and it is calling us back
755 * to allow us to reinstall the ones that are important to us. */
757 in_band_flushed(struct in_band
*in_band
)
763 in_band_create(struct ofproto
*ofproto
, struct dpif
*dpif
,
764 struct switch_status
*ss
, struct in_band
**in_bandp
)
766 struct in_band
*in_band
;
767 char local_name
[IF_NAMESIZE
];
768 struct netdev
*local_netdev
;
771 error
= dpif_port_get_name(dpif
, ODPP_LOCAL
,
772 local_name
, sizeof local_name
);
774 VLOG_ERR("failed to initialize in-band control: cannot get name "
775 "of datapath local port (%s)", strerror(error
));
779 error
= netdev_open_default(local_name
, &local_netdev
);
781 VLOG_ERR("failed to initialize in-band control: cannot open "
782 "datapath local port %s (%s)", local_name
, strerror(error
));
786 in_band
= xzalloc(sizeof *in_band
);
787 in_band
->ofproto
= ofproto
;
788 in_band
->ss_cat
= switch_status_register(ss
, "in-band",
789 in_band_status_cb
, in_band
);
790 in_band
->next_remote_refresh
= TIME_MIN
;
791 in_band
->next_local_refresh
= TIME_MIN
;
792 in_band
->local_netdev
= local_netdev
;
800 in_band_destroy(struct in_band
*ib
)
804 in_band_set_remotes(ib
, NULL
, 0);
805 switch_status_unregister(ib
->ss_cat
);
806 netdev_close(ib
->local_netdev
);
812 any_addresses_changed(struct in_band
*ib
,
813 const struct sockaddr_in
*addresses
, size_t n
)
817 if (n
!= ib
->n_remotes
) {
821 for (i
= 0; i
< n
; i
++) {
822 const struct sockaddr_in
*old
= &ib
->remotes
[i
].remote_addr
;
823 const struct sockaddr_in
*new = &addresses
[i
];
825 if (old
->sin_addr
.s_addr
!= new->sin_addr
.s_addr
||
826 old
->sin_port
!= new->sin_port
) {
835 in_band_set_remotes(struct in_band
*ib
,
836 const struct sockaddr_in
*addresses
, size_t n
)
840 if (!any_addresses_changed(ib
, addresses
, n
)) {
844 /* Clear old remotes. */
845 for (i
= 0; i
< ib
->n_remotes
; i
++) {
846 netdev_close(ib
->remotes
[i
].remote_netdev
);
850 /* Set up new remotes. */
851 ib
->remotes
= n
? xzalloc(n
* sizeof *ib
->remotes
) : NULL
;
853 for (i
= 0; i
< n
; i
++) {
854 ib
->remotes
[i
].remote_addr
= addresses
[i
];
857 /* Force refresh in next call to in_band_run(). */
858 ib
->next_remote_refresh
= TIME_MIN
;