]>
Commit | Line | Data |
---|---|---|
b0ec0f27 | 1 | /* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks |
c93b1d6a | 2 | * |
a14bc59f BP |
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: | |
064af421 | 6 | * |
a14bc59f | 7 | * http://www.apache.org/licenses/LICENSE-2.0 |
064af421 | 8 | * |
a14bc59f BP |
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. | |
064af421 BP |
14 | */ |
15 | ||
16 | #include <config.h> | |
17 | #include "bridge.h" | |
10a24935 | 18 | #include "byte-order.h" |
064af421 BP |
19 | #include <assert.h> |
20 | #include <errno.h> | |
21 | #include <arpa/inet.h> | |
22 | #include <ctype.h> | |
23 | #include <inttypes.h> | |
9d82ec47 | 24 | #include <sys/socket.h> |
064af421 BP |
25 | #include <net/if.h> |
26 | #include <openflow/openflow.h> | |
27 | #include <signal.h> | |
28 | #include <stdlib.h> | |
29 | #include <strings.h> | |
30 | #include <sys/stat.h> | |
31 | #include <sys/socket.h> | |
7e40e21d | 32 | #include <sys/types.h> |
064af421 BP |
33 | #include <unistd.h> |
34 | #include "bitmap.h" | |
b31bcf60 | 35 | #include "cfm.h" |
cf3fad8a | 36 | #include "classifier.h" |
064af421 BP |
37 | #include "coverage.h" |
38 | #include "dirs.h" | |
39 | #include "dpif.h" | |
40 | #include "dynamic-string.h" | |
41 | #include "flow.h" | |
42 | #include "hash.h" | |
d9a8717a | 43 | #include "hmap.h" |
cd11000b | 44 | #include "jsonrpc.h" |
064af421 BP |
45 | #include "list.h" |
46 | #include "mac-learning.h" | |
47 | #include "netdev.h" | |
cdee00fd | 48 | #include "netlink.h" |
064af421 BP |
49 | #include "odp-util.h" |
50 | #include "ofp-print.h" | |
51 | #include "ofpbuf.h" | |
d65349ea | 52 | #include "ofproto/netflow.h" |
8cd4882f | 53 | #include "ofproto/ofproto.h" |
dd0d105c | 54 | #include "ovsdb-data.h" |
da285df4 | 55 | #include "packets.h" |
064af421 | 56 | #include "poll-loop.h" |
064af421 BP |
57 | #include "proc-net-compat.h" |
58 | #include "process.h" | |
76343538 | 59 | #include "sha1.h" |
6c88d577 | 60 | #include "shash.h" |
064af421 | 61 | #include "socket-util.h" |
fe55ad15 | 62 | #include "stream-ssl.h" |
064af421 | 63 | #include "svec.h" |
ce887677 | 64 | #include "system-stats.h" |
064af421 BP |
65 | #include "timeval.h" |
66 | #include "util.h" | |
da285df4 | 67 | #include "unixctl.h" |
064af421 | 68 | #include "vconn.h" |
b36682d8 | 69 | #include "vswitchd/vswitch-idl.h" |
064af421 | 70 | #include "xenserver.h" |
5136ce49 | 71 | #include "vlog.h" |
72b06300 | 72 | #include "sflow_api.h" |
064af421 | 73 | |
d98e6007 | 74 | VLOG_DEFINE_THIS_MODULE(bridge); |
064af421 | 75 | |
d76f09ea BP |
76 | COVERAGE_DEFINE(bridge_flush); |
77 | COVERAGE_DEFINE(bridge_process_flow); | |
78 | COVERAGE_DEFINE(bridge_reconfigure); | |
79 | ||
064af421 BP |
80 | struct dst { |
81 | uint16_t vlan; | |
82 | uint16_t dp_ifidx; | |
83 | }; | |
84 | ||
0c62a7ad BP |
85 | struct dst_set { |
86 | struct dst builtin[32]; | |
87 | struct dst *dsts; | |
88 | size_t n, allocated; | |
89 | }; | |
90 | ||
91 | static void dst_set_init(struct dst_set *); | |
92 | static void dst_set_add(struct dst_set *, const struct dst *); | |
93 | static void dst_set_free(struct dst_set *); | |
94 | ||
064af421 | 95 | struct iface { |
0c6aea3f | 96 | /* These members are always valid. */ |
064af421 BP |
97 | struct port *port; /* Containing port. */ |
98 | size_t port_ifidx; /* Index within containing port. */ | |
064af421 | 99 | char *name; /* Host network device name. */ |
064af421 | 100 | tag_type tag; /* Tag associated with this interface. */ |
064af421 | 101 | long long delay_expires; /* Time after which 'enabled' may change. */ |
064af421 | 102 | |
0c6aea3f | 103 | /* These members are valid only after bridge_reconfigure() causes them to |
1e0b752d | 104 | * be initialized. */ |
d9a8717a | 105 | struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */ |
0c6aea3f BP |
106 | int dp_ifidx; /* Index within kernel datapath. */ |
107 | struct netdev *netdev; /* Network device. */ | |
064af421 | 108 | bool enabled; /* May be chosen for flows? */ |
6cefe1da | 109 | const char *type; /* Usually same as cfg->type. */ |
b31bcf60 | 110 | struct cfm *cfm; /* Connectivity Fault Management */ |
76343538 | 111 | const struct ovsrec_interface *cfg; |
064af421 BP |
112 | }; |
113 | ||
114 | #define BOND_MASK 0xff | |
115 | struct bond_entry { | |
116 | int iface_idx; /* Index of assigned iface, or -1 if none. */ | |
117 | uint64_t tx_bytes; /* Count of bytes recently transmitted. */ | |
118 | tag_type iface_tag; /* Tag associated with iface_idx. */ | |
119 | }; | |
120 | ||
27dcaa1a EJ |
121 | enum bond_mode { |
122 | BM_SLB, /* Source Load Balance (Default). */ | |
123 | BM_AB /* Active Backup. */ | |
be02e7c3 EJ |
124 | }; |
125 | ||
064af421 BP |
126 | #define MAX_MIRRORS 32 |
127 | typedef uint32_t mirror_mask_t; | |
128 | #define MIRROR_MASK_C(X) UINT32_C(X) | |
129 | BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS); | |
130 | struct mirror { | |
131 | struct bridge *bridge; | |
132 | size_t idx; | |
133 | char *name; | |
dd0d105c | 134 | struct uuid uuid; /* UUID of this "mirror" record in database. */ |
064af421 BP |
135 | |
136 | /* Selection criteria. */ | |
37e7f427 BP |
137 | struct shash src_ports; /* Name is port name; data is always NULL. */ |
138 | struct shash dst_ports; /* Name is port name; data is always NULL. */ | |
064af421 BP |
139 | int *vlans; |
140 | size_t n_vlans; | |
141 | ||
142 | /* Output. */ | |
143 | struct port *out_port; | |
144 | int out_vlan; | |
145 | }; | |
146 | ||
147 | #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */ | |
148 | struct port { | |
149 | struct bridge *bridge; | |
150 | size_t port_idx; | |
151 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ | |
3e9c481c BP |
152 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. |
153 | * NULL if all VLANs are trunked. */ | |
1e0b752d | 154 | const struct ovsrec_port *cfg; |
064af421 BP |
155 | char *name; |
156 | ||
157 | /* An ordinary bridge port has 1 interface. | |
158 | * A bridge port for bonding has at least 2 interfaces. */ | |
159 | struct iface **ifaces; | |
160 | size_t n_ifaces, allocated_ifaces; | |
161 | ||
162 | /* Bonding info. */ | |
27dcaa1a | 163 | enum bond_mode bond_mode; /* Type of the bond. BM_SLB is the default. */ |
064af421 BP |
164 | int active_iface; /* Ifidx on which bcasts accepted, or -1. */ |
165 | tag_type active_iface_tag; /* Tag for bcast flows. */ | |
166 | tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */ | |
167 | int updelay, downdelay; /* Delay before iface goes up/down, in ms. */ | |
85c74638 | 168 | bool bond_compat_is_stale; /* Need to call port_update_bond_compat()? */ |
76343538 | 169 | bool bond_fake_iface; /* Fake a bond interface for legacy compat? */ |
63331829 EJ |
170 | bool miimon; /* Use miimon instead of carrier? */ |
171 | long long int bond_miimon_interval; /* Miimon status refresh interval. */ | |
172 | long long int bond_miimon_next_update; /* Time of next miimon update. */ | |
e5ae7df8 | 173 | long long int bond_next_fake_iface_update; /* Time of next update. */ |
be02e7c3 EJ |
174 | struct netdev_monitor *monitor; /* Tracks carrier up/down status. */ |
175 | ||
176 | /* SLB specific bonding info. */ | |
177 | struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */ | |
c8143c88 BP |
178 | int bond_rebalance_interval; /* Interval between rebalances, in ms. */ |
179 | long long int bond_next_rebalance; /* Next rebalancing time. */ | |
064af421 BP |
180 | |
181 | /* Port mirroring info. */ | |
182 | mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */ | |
183 | mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */ | |
184 | bool is_mirror_output_port; /* Does port mirroring send frames here? */ | |
064af421 BP |
185 | }; |
186 | ||
064af421 BP |
187 | struct bridge { |
188 | struct list node; /* Node in global list of bridges. */ | |
189 | char *name; /* User-specified arbitrary name. */ | |
93dfc067 | 190 | struct mac_learning *ml; /* MAC learning table. */ |
064af421 | 191 | uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */ |
1e0b752d | 192 | const struct ovsrec_bridge *cfg; |
064af421 | 193 | |
064af421 BP |
194 | /* OpenFlow switch processing. */ |
195 | struct ofproto *ofproto; /* OpenFlow switch. */ | |
196 | ||
197 | /* Kernel datapath information. */ | |
c228a364 | 198 | struct dpif *dpif; /* Datapath. */ |
d9a8717a | 199 | struct hmap ifaces; /* Contains "struct iface"s. */ |
064af421 BP |
200 | |
201 | /* Bridge ports. */ | |
202 | struct port **ports; | |
203 | size_t n_ports, allocated_ports; | |
4a1ee6ae | 204 | struct shash iface_by_name; /* "struct iface"s indexed by name. */ |
836fad5e | 205 | struct shash port_by_name; /* "struct port"s indexed by name. */ |
064af421 BP |
206 | |
207 | /* Bonding. */ | |
208 | bool has_bonded_ports; | |
064af421 BP |
209 | |
210 | /* Flow tracking. */ | |
211 | bool flush; | |
212 | ||
064af421 BP |
213 | /* Port mirroring. */ |
214 | struct mirror *mirrors[MAX_MIRRORS]; | |
064af421 BP |
215 | }; |
216 | ||
217 | /* List of all bridges. */ | |
218 | static struct list all_bridges = LIST_INITIALIZER(&all_bridges); | |
219 | ||
c5187f17 BP |
220 | /* OVSDB IDL used to obtain configuration. */ |
221 | static struct ovsdb_idl *idl; | |
222 | ||
cd0cd65f BP |
223 | /* Each time this timer expires, the bridge fetches systems and interface |
224 | * statistics and pushes them into the database. */ | |
225 | #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */ | |
226 | static long long int stats_timer = LLONG_MIN; | |
018f1525 | 227 | |
1a6f1e2a | 228 | static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg); |
064af421 BP |
229 | static void bridge_destroy(struct bridge *); |
230 | static struct bridge *bridge_lookup(const char *name); | |
8ca79daa | 231 | static unixctl_cb_func bridge_unixctl_dump_flows; |
fa05809b | 232 | static unixctl_cb_func bridge_unixctl_reconnect; |
064af421 | 233 | static int bridge_run_one(struct bridge *); |
1a048029 | 234 | static size_t bridge_get_controllers(const struct bridge *br, |
76ce9432 | 235 | struct ovsrec_controller ***controllersp); |
1a048029 JP |
236 | static void bridge_reconfigure_one(struct bridge *); |
237 | static void bridge_reconfigure_remotes(struct bridge *, | |
11bbff1b BP |
238 | const struct sockaddr_in *managers, |
239 | size_t n_managers); | |
76343538 | 240 | static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces); |
064af421 BP |
241 | static void bridge_fetch_dp_ifaces(struct bridge *); |
242 | static void bridge_flush(struct bridge *); | |
243 | static void bridge_pick_local_hw_addr(struct bridge *, | |
244 | uint8_t ea[ETH_ADDR_LEN], | |
07c318f4 | 245 | struct iface **hw_addr_iface); |
064af421 BP |
246 | static uint64_t bridge_pick_datapath_id(struct bridge *, |
247 | const uint8_t bridge_ea[ETH_ADDR_LEN], | |
07c318f4 BP |
248 | struct iface *hw_addr_iface); |
249 | static struct iface *bridge_get_local_iface(struct bridge *); | |
064af421 BP |
250 | static uint64_t dpid_from_hash(const void *, size_t nbytes); |
251 | ||
8ca79daa | 252 | static unixctl_cb_func bridge_unixctl_fdb_show; |
8c4c1387 | 253 | |
da285df4 | 254 | static void bond_init(void); |
064af421 BP |
255 | static void bond_run(struct bridge *); |
256 | static void bond_wait(struct bridge *); | |
257 | static void bond_rebalance_port(struct port *); | |
2303f3b2 | 258 | static void bond_send_learning_packets(struct port *); |
8b2a2f4a | 259 | static void bond_enable_slave(struct iface *iface, bool enable); |
064af421 | 260 | |
76343538 BP |
261 | static struct port *port_create(struct bridge *, const char *name); |
262 | static void port_reconfigure(struct port *, const struct ovsrec_port *); | |
4a1ee6ae | 263 | static void port_del_ifaces(struct port *, const struct ovsrec_port *); |
064af421 BP |
264 | static void port_destroy(struct port *); |
265 | static struct port *port_lookup(const struct bridge *, const char *name); | |
da285df4 | 266 | static struct iface *port_lookup_iface(const struct port *, const char *name); |
064af421 BP |
267 | static struct port *port_from_dp_ifidx(const struct bridge *, |
268 | uint16_t dp_ifidx); | |
269 | static void port_update_bond_compat(struct port *); | |
270 | static void port_update_vlan_compat(struct port *); | |
0c6aea3f | 271 | static void port_update_bonding(struct port *); |
064af421 | 272 | |
dd0d105c | 273 | static void mirror_create(struct bridge *, struct ovsrec_mirror *); |
064af421 BP |
274 | static void mirror_destroy(struct mirror *); |
275 | static void mirror_reconfigure(struct bridge *); | |
37e7f427 | 276 | static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *); |
064af421 BP |
277 | static bool vlan_is_mirrored(const struct mirror *, int vlan); |
278 | ||
d295e8e9 | 279 | static struct iface *iface_create(struct port *port, |
a740f0de | 280 | const struct ovsrec_interface *if_cfg); |
064af421 BP |
281 | static void iface_destroy(struct iface *); |
282 | static struct iface *iface_lookup(const struct bridge *, const char *name); | |
283 | static struct iface *iface_from_dp_ifidx(const struct bridge *, | |
284 | uint16_t dp_ifidx); | |
52df17e7 | 285 | static void iface_set_mac(struct iface *); |
bcd49a45 | 286 | static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport); |
c1c9c9c4 | 287 | static void iface_update_qos(struct iface *, const struct ovsrec_qos *); |
b31bcf60 EJ |
288 | static void iface_update_cfm(struct iface *); |
289 | static void iface_refresh_cfm_stats(struct iface *iface); | |
290 | static void iface_send_packet(struct iface *, struct ofpbuf *packet); | |
064af421 | 291 | |
43776b8f BP |
292 | static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, |
293 | struct shash *); | |
ea763e0e EJ |
294 | static void shash_to_ovs_idl_map(struct shash *, |
295 | char ***keys, char ***values, size_t *n); | |
296 | ||
43776b8f | 297 | |
064af421 BP |
298 | /* Hooks into ofproto processing. */ |
299 | static struct ofhooks bridge_ofhooks; | |
300 | \f | |
301 | /* Public functions. */ | |
302 | ||
c5187f17 BP |
303 | /* Initializes the bridge module, configuring it to obtain its configuration |
304 | * from an OVSDB server accessed over 'remote', which should be a string in a | |
305 | * form acceptable to ovsdb_idl_create(). */ | |
064af421 | 306 | void |
c5187f17 BP |
307 | bridge_init(const char *remote) |
308 | { | |
309 | /* Create connection to database. */ | |
ef73f86c | 310 | idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true); |
c5187f17 | 311 | |
ef73f86c BP |
312 | ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg); |
313 | ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics); | |
e85bbd75 BP |
314 | ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids); |
315 | ||
316 | ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); | |
317 | ||
318 | ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); | |
319 | ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); | |
320 | ||
ef73f86c BP |
321 | ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport); |
322 | ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); | |
e85bbd75 BP |
323 | ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); |
324 | ||
c5187f17 BP |
325 | /* Register unixctl commands. */ |
326 | unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL); | |
327 | unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows, | |
328 | NULL); | |
fa05809b BP |
329 | unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect, |
330 | NULL); | |
c5187f17 BP |
331 | bond_init(); |
332 | } | |
333 | ||
ee45ad81 BP |
334 | void |
335 | bridge_exit(void) | |
336 | { | |
337 | struct bridge *br, *next_br; | |
338 | ||
339 | LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) { | |
340 | bridge_destroy(br); | |
341 | } | |
342 | ovsdb_idl_destroy(idl); | |
343 | } | |
344 | ||
c5187f17 BP |
345 | /* Performs configuration that is only necessary once at ovs-vswitchd startup, |
346 | * but for which the ovs-vswitchd configuration 'cfg' is required. */ | |
347 | static void | |
348 | bridge_configure_once(const struct ovsrec_open_vswitch *cfg) | |
064af421 | 349 | { |
c5187f17 | 350 | static bool already_configured_once; |
76343538 | 351 | struct svec bridge_names; |
1a6f1e2a | 352 | struct svec dpif_names, dpif_types; |
d3d22744 | 353 | size_t i; |
da285df4 | 354 | |
c5187f17 BP |
355 | /* Only do this once per ovs-vswitchd run. */ |
356 | if (already_configured_once) { | |
357 | return; | |
358 | } | |
359 | already_configured_once = true; | |
8c4c1387 | 360 | |
cd0cd65f | 361 | stats_timer = time_msec() + STATS_INTERVAL; |
018f1525 | 362 | |
c5187f17 | 363 | /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */ |
76343538 BP |
364 | svec_init(&bridge_names); |
365 | for (i = 0; i < cfg->n_bridges; i++) { | |
366 | svec_add(&bridge_names, cfg->bridges[i]->name); | |
367 | } | |
368 | svec_sort(&bridge_names); | |
369 | ||
c5187f17 BP |
370 | /* Iterate over all system dpifs and delete any of them that do not appear |
371 | * in 'cfg'. */ | |
d2a34569 | 372 | svec_init(&dpif_names); |
1a6f1e2a JG |
373 | svec_init(&dpif_types); |
374 | dp_enumerate_types(&dpif_types); | |
375 | for (i = 0; i < dpif_types.n; i++) { | |
1a6f1e2a | 376 | size_t j; |
064af421 | 377 | |
1a6f1e2a | 378 | dp_enumerate_names(dpif_types.names[i], &dpif_names); |
d3d22744 | 379 | |
3d8c9535 | 380 | /* Delete each dpif whose name is not in 'bridge_names'. */ |
1a6f1e2a | 381 | for (j = 0; j < dpif_names.n; j++) { |
3d8c9535 BP |
382 | if (!svec_contains(&bridge_names, dpif_names.names[j])) { |
383 | struct dpif *dpif; | |
384 | int retval; | |
385 | ||
386 | retval = dpif_open(dpif_names.names[j], dpif_types.names[i], | |
387 | &dpif); | |
388 | if (!retval) { | |
389 | dpif_delete(dpif); | |
390 | dpif_close(dpif); | |
d3d22744 | 391 | } |
064af421 | 392 | } |
064af421 BP |
393 | } |
394 | } | |
7128cd6b | 395 | svec_destroy(&bridge_names); |
d0c9d304 | 396 | svec_destroy(&dpif_names); |
1a6f1e2a | 397 | svec_destroy(&dpif_types); |
064af421 BP |
398 | } |
399 | ||
64d64dd7 | 400 | /* Callback for iterate_and_prune_ifaces(). */ |
0c6aea3f | 401 | static bool |
64d64dd7 | 402 | check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED) |
0c6aea3f | 403 | { |
149f577a | 404 | if (!iface->netdev) { |
64d64dd7 BP |
405 | /* We already reported a related error, don't bother duplicating it. */ |
406 | return false; | |
0c6aea3f | 407 | } |
149f577a | 408 | |
64d64dd7 | 409 | if (iface->dp_ifidx < 0) { |
6ae39834 BP |
410 | VLOG_ERR("%s interface not in %s, dropping", |
411 | iface->name, dpif_name(br->dpif)); | |
412 | return false; | |
413 | } | |
64d64dd7 BP |
414 | |
415 | VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif), | |
416 | iface->name, iface->dp_ifidx); | |
417 | return true; | |
6ae39834 BP |
418 | } |
419 | ||
64d64dd7 | 420 | /* Callback for iterate_and_prune_ifaces(). */ |
795fe1fb | 421 | static bool |
67a4917b BP |
422 | set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface, |
423 | void *aux OVS_UNUSED) | |
795fe1fb | 424 | { |
4d678233 | 425 | /* Set policing attributes. */ |
76343538 BP |
426 | netdev_set_policing(iface->netdev, |
427 | iface->cfg->ingress_policing_rate, | |
428 | iface->cfg->ingress_policing_burst); | |
4d678233 BP |
429 | |
430 | /* Set MAC address of internal interfaces other than the local | |
431 | * interface. */ | |
6cefe1da | 432 | if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) { |
4d678233 BP |
433 | iface_set_mac(iface); |
434 | } | |
435 | ||
795fe1fb BP |
436 | return true; |
437 | } | |
438 | ||
6ae39834 BP |
439 | /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument. |
440 | * Deletes from 'br' all the interfaces for which 'cb' returns false, and then | |
441 | * deletes from 'br' any ports that no longer have any interfaces. */ | |
442 | static void | |
443 | iterate_and_prune_ifaces(struct bridge *br, | |
444 | bool (*cb)(struct bridge *, struct iface *, | |
445 | void *aux), | |
446 | void *aux) | |
447 | { | |
448 | size_t i, j; | |
449 | ||
450 | for (i = 0; i < br->n_ports; ) { | |
451 | struct port *port = br->ports[i]; | |
452 | for (j = 0; j < port->n_ifaces; ) { | |
453 | struct iface *iface = port->ifaces[j]; | |
454 | if (cb(br, iface, aux)) { | |
455 | j++; | |
456 | } else { | |
bcd49a45 | 457 | iface_set_ofport(iface->cfg, -1); |
6ae39834 BP |
458 | iface_destroy(iface); |
459 | } | |
460 | } | |
461 | ||
462 | if (port->n_ifaces) { | |
463 | i++; | |
464 | } else { | |
465 | VLOG_ERR("%s port has no interfaces, dropping", port->name); | |
466 | port_destroy(port); | |
467 | } | |
468 | } | |
469 | } | |
470 | ||
cd11000b BP |
471 | /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP |
472 | * addresses and ports into '*managersp' and '*n_managersp'. The caller is | |
473 | * responsible for freeing '*managersp' (with free()). | |
474 | * | |
475 | * You may be asking yourself "why does ovs-vswitchd care?", because | |
476 | * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd | |
477 | * should not be and in fact is not directly involved in that. But | |
478 | * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so | |
479 | * it has to tell in-band control where the managers are to enable that. | |
94db5407 | 480 | * (Thus, only managers connected in-band are collected.) |
cd11000b BP |
481 | */ |
482 | static void | |
94db5407 BP |
483 | collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, |
484 | struct sockaddr_in **managersp, size_t *n_managersp) | |
cd11000b BP |
485 | { |
486 | struct sockaddr_in *managers = NULL; | |
487 | size_t n_managers = 0; | |
94db5407 BP |
488 | struct shash targets; |
489 | size_t i; | |
490 | ||
491 | /* Collect all of the potential targets, as the union of the "managers" | |
492 | * column and the "targets" columns of the rows pointed to by | |
493 | * "manager_options", excluding any that are out-of-band. */ | |
494 | shash_init(&targets); | |
495 | for (i = 0; i < ovs_cfg->n_managers; i++) { | |
496 | shash_add_once(&targets, ovs_cfg->managers[i], NULL); | |
497 | } | |
498 | for (i = 0; i < ovs_cfg->n_manager_options; i++) { | |
499 | struct ovsrec_manager *m = ovs_cfg->manager_options[i]; | |
500 | ||
501 | if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) { | |
502 | shash_find_and_delete(&targets, m->target); | |
503 | } else { | |
504 | shash_add_once(&targets, m->target, NULL); | |
505 | } | |
506 | } | |
cd11000b | 507 | |
94db5407 BP |
508 | /* Now extract the targets' IP addresses. */ |
509 | if (!shash_is_empty(&targets)) { | |
510 | struct shash_node *node; | |
cd11000b | 511 | |
94db5407 BP |
512 | managers = xmalloc(shash_count(&targets) * sizeof *managers); |
513 | SHASH_FOR_EACH (node, &targets) { | |
514 | const char *target = node->name; | |
515 | struct sockaddr_in *sin = &managers[n_managers]; | |
cd11000b | 516 | |
94db5407 BP |
517 | if ((!strncmp(target, "tcp:", 4) |
518 | && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) || | |
519 | (!strncmp(target, "ssl:", 4) | |
520 | && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) { | |
cd11000b BP |
521 | n_managers++; |
522 | } | |
523 | } | |
524 | } | |
94db5407 | 525 | shash_destroy(&targets); |
cd11000b BP |
526 | |
527 | *managersp = managers; | |
528 | *n_managersp = n_managers; | |
529 | } | |
530 | ||
c5187f17 | 531 | static void |
76343538 | 532 | bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) |
064af421 | 533 | { |
76343538 BP |
534 | struct shash old_br, new_br; |
535 | struct shash_node *node; | |
064af421 | 536 | struct bridge *br, *next; |
cd11000b BP |
537 | struct sockaddr_in *managers; |
538 | size_t n_managers; | |
6ae39834 | 539 | size_t i; |
72b06300 | 540 | int sflow_bridge_number; |
064af421 BP |
541 | |
542 | COVERAGE_INC(bridge_reconfigure); | |
543 | ||
94db5407 | 544 | collect_in_band_managers(ovs_cfg, &managers, &n_managers); |
cd11000b | 545 | |
632d136c | 546 | /* Collect old and new bridges. */ |
76343538 BP |
547 | shash_init(&old_br); |
548 | shash_init(&new_br); | |
4e8e4213 | 549 | LIST_FOR_EACH (br, node, &all_bridges) { |
76343538 BP |
550 | shash_add(&old_br, br->name, br); |
551 | } | |
552 | for (i = 0; i < ovs_cfg->n_bridges; i++) { | |
553 | const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i]; | |
554 | if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) { | |
555 | VLOG_WARN("more than one bridge named %s", br_cfg->name); | |
556 | } | |
064af421 | 557 | } |
064af421 BP |
558 | |
559 | /* Get rid of deleted bridges and add new bridges. */ | |
4e8e4213 | 560 | LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) { |
76343538 BP |
561 | struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name); |
562 | if (br_cfg) { | |
563 | br->cfg = br_cfg; | |
564 | } else { | |
064af421 BP |
565 | bridge_destroy(br); |
566 | } | |
567 | } | |
76343538 BP |
568 | SHASH_FOR_EACH (node, &new_br) { |
569 | const char *br_name = node->name; | |
570 | const struct ovsrec_bridge *br_cfg = node->data; | |
1a6f1e2a JG |
571 | br = shash_find_data(&old_br, br_name); |
572 | if (br) { | |
573 | /* If the bridge datapath type has changed, we need to tear it | |
574 | * down and recreate. */ | |
575 | if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) { | |
576 | bridge_destroy(br); | |
577 | bridge_create(br_cfg); | |
fa33d64a | 578 | } |
1a6f1e2a JG |
579 | } else { |
580 | bridge_create(br_cfg); | |
064af421 BP |
581 | } |
582 | } | |
76343538 BP |
583 | shash_destroy(&old_br); |
584 | shash_destroy(&new_br); | |
064af421 | 585 | |
064af421 | 586 | /* Reconfigure all bridges. */ |
4e8e4213 | 587 | LIST_FOR_EACH (br, node, &all_bridges) { |
1a048029 | 588 | bridge_reconfigure_one(br); |
064af421 BP |
589 | } |
590 | ||
591 | /* Add and delete ports on all datapaths. | |
592 | * | |
593 | * The kernel will reject any attempt to add a given port to a datapath if | |
594 | * that port already belongs to a different datapath, so we must do all | |
595 | * port deletions before any port additions. */ | |
4e8e4213 | 596 | LIST_FOR_EACH (br, node, &all_bridges) { |
b0ec0f27 | 597 | struct dpif_port_dump dump; |
76343538 | 598 | struct shash want_ifaces; |
4c738a8d | 599 | struct dpif_port dpif_port; |
064af421 | 600 | |
064af421 | 601 | bridge_get_all_ifaces(br, &want_ifaces); |
4c738a8d BP |
602 | DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { |
603 | if (!shash_find(&want_ifaces, dpif_port.name) | |
604 | && strcmp(dpif_port.name, br->name)) { | |
605 | int retval = dpif_port_del(br->dpif, dpif_port.port_no); | |
064af421 | 606 | if (retval) { |
b29ba128 | 607 | VLOG_ERR("failed to remove %s interface from %s: %s", |
4c738a8d | 608 | dpif_port.name, dpif_name(br->dpif), |
b29ba128 | 609 | strerror(retval)); |
064af421 BP |
610 | } |
611 | } | |
612 | } | |
76343538 | 613 | shash_destroy(&want_ifaces); |
064af421 | 614 | } |
4e8e4213 | 615 | LIST_FOR_EACH (br, node, &all_bridges) { |
76343538 | 616 | struct shash cur_ifaces, want_ifaces; |
b0ec0f27 | 617 | struct dpif_port_dump dump; |
4c738a8d | 618 | struct dpif_port dpif_port; |
064af421 | 619 | |
76343538 | 620 | /* Get the set of interfaces currently in this datapath. */ |
76343538 | 621 | shash_init(&cur_ifaces); |
4c738a8d | 622 | DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { |
b0ec0f27 | 623 | struct dpif_port *port_info = xmalloc(sizeof *port_info); |
4c738a8d BP |
624 | dpif_port_clone(port_info, &dpif_port); |
625 | shash_add(&cur_ifaces, dpif_port.name, port_info); | |
064af421 | 626 | } |
064af421 | 627 | |
76343538 BP |
628 | /* Get the set of interfaces we want on this datapath. */ |
629 | bridge_get_all_ifaces(br, &want_ifaces); | |
6c88d577 | 630 | |
3a6ccc8c | 631 | hmap_clear(&br->ifaces); |
76343538 BP |
632 | SHASH_FOR_EACH (node, &want_ifaces) { |
633 | const char *if_name = node->name; | |
634 | struct iface *iface = node->data; | |
b0ec0f27 BP |
635 | struct dpif_port *dpif_port; |
636 | const char *type; | |
3a6ccc8c BP |
637 | int error; |
638 | ||
b0ec0f27 BP |
639 | type = iface ? iface->type : "internal"; |
640 | dpif_port = shash_find_data(&cur_ifaces, if_name); | |
641 | ||
3a6ccc8c BP |
642 | /* If we have a port or a netdev already, and it's not the type we |
643 | * want, then delete the port (if any) and close the netdev (if | |
644 | * any). */ | |
c3827f61 BP |
645 | if ((dpif_port && strcmp(dpif_port->type, type)) |
646 | || (iface && iface->netdev | |
647 | && strcmp(type, netdev_get_type(iface->netdev)))) { | |
3a6ccc8c | 648 | if (dpif_port) { |
b0ec0f27 | 649 | error = ofproto_port_del(br->ofproto, dpif_port->port_no); |
3a6ccc8c BP |
650 | if (error) { |
651 | continue; | |
652 | } | |
653 | dpif_port = NULL; | |
76343538 | 654 | } |
3a6ccc8c BP |
655 | if (iface) { |
656 | netdev_close(iface->netdev); | |
657 | iface->netdev = NULL; | |
658 | } | |
659 | } | |
76343538 | 660 | |
c3827f61 BP |
661 | /* If the port doesn't exist or we don't have the netdev open, |
662 | * we need to do more work. */ | |
663 | if (!dpif_port || (iface && !iface->netdev)) { | |
664 | struct netdev_options options; | |
665 | struct netdev *netdev; | |
666 | struct shash args; | |
667 | ||
668 | /* First open the network device. */ | |
669 | options.name = if_name; | |
670 | options.type = type; | |
671 | options.args = &args; | |
672 | options.ethertype = NETDEV_ETH_TYPE_NONE; | |
673 | ||
674 | shash_init(&args); | |
675 | if (iface) { | |
6b4186af | 676 | shash_from_ovs_idl_map(iface->cfg->key_options, |
fe4838bc JP |
677 | iface->cfg->value_options, |
678 | iface->cfg->n_options, &args); | |
82057f51 | 679 | } |
c3827f61 BP |
680 | error = netdev_open(&options, &netdev); |
681 | shash_destroy(&args); | |
82057f51 | 682 | |
c3827f61 BP |
683 | if (error) { |
684 | VLOG_WARN("could not open network device %s (%s)", | |
685 | if_name, strerror(error)); | |
3a6ccc8c | 686 | continue; |
76343538 | 687 | } |
3a6ccc8c | 688 | |
c3827f61 BP |
689 | /* Then add the port if we haven't already. */ |
690 | if (!dpif_port) { | |
691 | error = dpif_port_add(br->dpif, netdev, NULL); | |
3a6ccc8c | 692 | if (error) { |
c3827f61 BP |
693 | netdev_close(netdev); |
694 | if (error == EFBIG) { | |
695 | VLOG_ERR("ran out of valid port numbers on %s", | |
696 | dpif_name(br->dpif)); | |
697 | break; | |
698 | } else { | |
699 | VLOG_ERR("failed to add %s interface to %s: %s", | |
700 | if_name, dpif_name(br->dpif), | |
701 | strerror(error)); | |
702 | continue; | |
703 | } | |
3a6ccc8c BP |
704 | } |
705 | } | |
c3827f61 BP |
706 | |
707 | /* Update 'iface'. */ | |
708 | if (iface) { | |
709 | iface->netdev = netdev; | |
710 | iface->enabled = netdev_get_carrier(iface->netdev); | |
711 | } | |
712 | } else if (iface && iface->netdev) { | |
713 | struct shash args; | |
714 | ||
715 | shash_init(&args); | |
6b4186af | 716 | shash_from_ovs_idl_map(iface->cfg->key_options, |
fe4838bc JP |
717 | iface->cfg->value_options, |
718 | iface->cfg->n_options, &args); | |
6d9e6eb4 | 719 | netdev_set_config(iface->netdev, &args); |
c3827f61 | 720 | shash_destroy(&args); |
3a6ccc8c | 721 | } |
064af421 | 722 | } |
76343538 | 723 | shash_destroy(&want_ifaces); |
b0ec0f27 BP |
724 | |
725 | SHASH_FOR_EACH (node, &cur_ifaces) { | |
726 | struct dpif_port *port_info = node->data; | |
4c738a8d | 727 | dpif_port_destroy(port_info); |
b0ec0f27 BP |
728 | free(port_info); |
729 | } | |
730 | shash_destroy(&cur_ifaces); | |
064af421 | 731 | } |
72b06300 | 732 | sflow_bridge_number = 0; |
4e8e4213 | 733 | LIST_FOR_EACH (br, node, &all_bridges) { |
064af421 BP |
734 | uint8_t ea[8]; |
735 | uint64_t dpid; | |
07c318f4 BP |
736 | struct iface *local_iface; |
737 | struct iface *hw_addr_iface; | |
093e47f4 | 738 | char *dpid_string; |
064af421 | 739 | |
064af421 | 740 | bridge_fetch_dp_ifaces(br); |
064af421 | 741 | |
64d64dd7 | 742 | iterate_and_prune_ifaces(br, check_iface, NULL); |
064af421 BP |
743 | |
744 | /* Pick local port hardware address, datapath ID. */ | |
07c318f4 BP |
745 | bridge_pick_local_hw_addr(br, ea, &hw_addr_iface); |
746 | local_iface = bridge_get_local_iface(br); | |
064af421 | 747 | if (local_iface) { |
07c318f4 | 748 | int error = netdev_set_etheraddr(local_iface->netdev, ea); |
064af421 BP |
749 | if (error) { |
750 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
751 | VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge " | |
752 | "Ethernet address: %s", | |
753 | br->name, strerror(error)); | |
754 | } | |
755 | } | |
756 | ||
07c318f4 | 757 | dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface); |
064af421 BP |
758 | ofproto_set_datapath_id(br->ofproto, dpid); |
759 | ||
fc09119a | 760 | dpid_string = xasprintf("%016"PRIx64, dpid); |
093e47f4 BP |
761 | ovsrec_bridge_set_datapath_id(br->cfg, dpid_string); |
762 | free(dpid_string); | |
763 | ||
064af421 | 764 | /* Set NetFlow configuration on this bridge. */ |
76343538 BP |
765 | if (br->cfg->netflow) { |
766 | struct ovsrec_netflow *nf_cfg = br->cfg->netflow; | |
767 | struct netflow_options opts; | |
768 | ||
769 | memset(&opts, 0, sizeof opts); | |
770 | ||
771 | dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id); | |
772 | if (nf_cfg->engine_type) { | |
cd746526 | 773 | opts.engine_type = *nf_cfg->engine_type; |
76343538 BP |
774 | } |
775 | if (nf_cfg->engine_id) { | |
cd746526 | 776 | opts.engine_id = *nf_cfg->engine_id; |
76343538 BP |
777 | } |
778 | ||
779 | opts.active_timeout = nf_cfg->active_timeout; | |
780 | if (!opts.active_timeout) { | |
781 | opts.active_timeout = -1; | |
782 | } else if (opts.active_timeout < 0) { | |
e9e2856e JG |
783 | VLOG_WARN("bridge %s: active timeout interval set to negative " |
784 | "value, using default instead (%d seconds)", br->name, | |
785 | NF_ACTIVE_TIMEOUT_DEFAULT); | |
786 | opts.active_timeout = -1; | |
76343538 BP |
787 | } |
788 | ||
789 | opts.add_id_to_iface = nf_cfg->add_id_to_interface; | |
790 | if (opts.add_id_to_iface) { | |
791 | if (opts.engine_id > 0x7f) { | |
792 | VLOG_WARN("bridge %s: netflow port mangling may conflict " | |
793 | "with another vswitch, choose an engine id less " | |
794 | "than 128", br->name); | |
795 | } | |
796 | if (br->n_ports > 508) { | |
797 | VLOG_WARN("bridge %s: netflow port mangling will conflict " | |
798 | "with another port when more than 508 ports are " | |
799 | "used", br->name); | |
800 | } | |
801 | } | |
802 | ||
803 | opts.collectors.n = nf_cfg->n_targets; | |
804 | opts.collectors.names = nf_cfg->targets; | |
805 | if (ofproto_set_netflow(br->ofproto, &opts)) { | |
d295e8e9 | 806 | VLOG_ERR("bridge %s: problem setting netflow collectors", |
76343538 BP |
807 | br->name); |
808 | } | |
809 | } else { | |
810 | ofproto_set_netflow(br->ofproto, NULL); | |
811 | } | |
064af421 | 812 | |
a4af0040 JP |
813 | /* Set sFlow configuration on this bridge. */ |
814 | if (br->cfg->sflow) { | |
d49d354b | 815 | const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow; |
76ce9432 | 816 | struct ovsrec_controller **controllers; |
72b06300 | 817 | struct ofproto_sflow_options oso; |
76ce9432 | 818 | size_t n_controllers; |
72b06300 | 819 | |
a4af0040 JP |
820 | memset(&oso, 0, sizeof oso); |
821 | ||
822 | oso.targets.n = sflow_cfg->n_targets; | |
823 | oso.targets.names = sflow_cfg->targets; | |
72b06300 BP |
824 | |
825 | oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE; | |
a4af0040 JP |
826 | if (sflow_cfg->sampling) { |
827 | oso.sampling_rate = *sflow_cfg->sampling; | |
72b06300 BP |
828 | } |
829 | ||
830 | oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL; | |
a4af0040 JP |
831 | if (sflow_cfg->polling) { |
832 | oso.polling_interval = *sflow_cfg->polling; | |
72b06300 BP |
833 | } |
834 | ||
835 | oso.header_len = SFL_DEFAULT_HEADER_SIZE; | |
a4af0040 JP |
836 | if (sflow_cfg->header) { |
837 | oso.header_len = *sflow_cfg->header; | |
72b06300 BP |
838 | } |
839 | ||
840 | oso.sub_id = sflow_bridge_number++; | |
a4af0040 JP |
841 | oso.agent_device = sflow_cfg->agent; |
842 | ||
76ce9432 | 843 | oso.control_ip = NULL; |
1a048029 | 844 | n_controllers = bridge_get_controllers(br, &controllers); |
76ce9432 BP |
845 | for (i = 0; i < n_controllers; i++) { |
846 | if (controllers[i]->local_ip) { | |
847 | oso.control_ip = controllers[i]->local_ip; | |
848 | break; | |
849 | } | |
850 | } | |
72b06300 BP |
851 | ofproto_set_sflow(br->ofproto, &oso); |
852 | ||
7b99db05 | 853 | /* Do not destroy oso.targets because it is owned by sflow_cfg. */ |
72b06300 BP |
854 | } else { |
855 | ofproto_set_sflow(br->ofproto, NULL); | |
856 | } | |
857 | ||
064af421 BP |
858 | /* Update the controller and related settings. It would be more |
859 | * straightforward to call this from bridge_reconfigure_one(), but we | |
860 | * can't do it there for two reasons. First, and most importantly, at | |
861 | * that point we don't know the dp_ifidx of any interfaces that have | |
862 | * been added to the bridge (because we haven't actually added them to | |
863 | * the datapath). Second, at that point we haven't set the datapath ID | |
864 | * yet; when a controller is configured, resetting the datapath ID will | |
865 | * immediately disconnect from the controller, so it's better to set | |
866 | * the datapath ID before the controller. */ | |
1a048029 | 867 | bridge_reconfigure_remotes(br, managers, n_managers); |
064af421 | 868 | } |
4e8e4213 | 869 | LIST_FOR_EACH (br, node, &all_bridges) { |
064af421 BP |
870 | for (i = 0; i < br->n_ports; i++) { |
871 | struct port *port = br->ports[i]; | |
c1c9c9c4 | 872 | int j; |
52df17e7 | 873 | |
064af421 | 874 | port_update_vlan_compat(port); |
0c6aea3f | 875 | port_update_bonding(port); |
c1c9c9c4 BP |
876 | |
877 | for (j = 0; j < port->n_ifaces; j++) { | |
878 | iface_update_qos(port->ifaces[j], port->cfg->qos); | |
879 | } | |
064af421 BP |
880 | } |
881 | } | |
4e8e4213 | 882 | LIST_FOR_EACH (br, node, &all_bridges) { |
4d678233 | 883 | iterate_and_prune_ifaces(br, set_iface_properties, NULL); |
064af421 | 884 | } |
093e47f4 | 885 | |
b31bcf60 EJ |
886 | LIST_FOR_EACH (br, node, &all_bridges) { |
887 | struct iface *iface; | |
888 | HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { | |
889 | iface_update_cfm(iface); | |
890 | } | |
891 | } | |
892 | ||
cd11000b | 893 | free(managers); |
093e47f4 BP |
894 | } |
895 | ||
896 | static const char * | |
af6278e1 BP |
897 | get_ovsrec_key_value(const struct ovsdb_idl_row *row, |
898 | const struct ovsdb_idl_column *column, | |
899 | const char *key) | |
093e47f4 | 900 | { |
af6278e1 BP |
901 | const struct ovsdb_datum *datum; |
902 | union ovsdb_atom atom; | |
903 | unsigned int idx; | |
093e47f4 | 904 | |
af6278e1 BP |
905 | datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING); |
906 | atom.string = (char *) key; | |
907 | idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING); | |
908 | return idx == UINT_MAX ? NULL : datum->values[idx].string; | |
064af421 BP |
909 | } |
910 | ||
c8143c88 BP |
911 | static const char * |
912 | bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key) | |
913 | { | |
af6278e1 BP |
914 | return get_ovsrec_key_value(&br_cfg->header_, |
915 | &ovsrec_bridge_col_other_config, key); | |
c8143c88 BP |
916 | } |
917 | ||
064af421 BP |
918 | static void |
919 | bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], | |
07c318f4 | 920 | struct iface **hw_addr_iface) |
064af421 | 921 | { |
093e47f4 | 922 | const char *hwaddr; |
064af421 BP |
923 | size_t i, j; |
924 | int error; | |
925 | ||
07c318f4 | 926 | *hw_addr_iface = NULL; |
064af421 BP |
927 | |
928 | /* Did the user request a particular MAC? */ | |
093e47f4 BP |
929 | hwaddr = bridge_get_other_config(br->cfg, "hwaddr"); |
930 | if (hwaddr && eth_addr_from_string(hwaddr, ea)) { | |
064af421 BP |
931 | if (eth_addr_is_multicast(ea)) { |
932 | VLOG_ERR("bridge %s: cannot set MAC address to multicast " | |
933 | "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); | |
934 | } else if (eth_addr_is_zero(ea)) { | |
935 | VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name); | |
936 | } else { | |
937 | return; | |
938 | } | |
939 | } | |
940 | ||
141f4942 BP |
941 | /* Otherwise choose the minimum non-local MAC address among all of the |
942 | * interfaces. */ | |
064af421 BP |
943 | memset(ea, 0xff, sizeof ea); |
944 | for (i = 0; i < br->n_ports; i++) { | |
945 | struct port *port = br->ports[i]; | |
58b7527e | 946 | uint8_t iface_ea[ETH_ADDR_LEN]; |
58b7527e BP |
947 | struct iface *iface; |
948 | ||
949 | /* Mirror output ports don't participate. */ | |
064af421 BP |
950 | if (port->is_mirror_output_port) { |
951 | continue; | |
952 | } | |
58b7527e BP |
953 | |
954 | /* Choose the MAC address to represent the port. */ | |
76343538 | 955 | if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) { |
ba09980a BP |
956 | /* Find the interface with this Ethernet address (if any) so that |
957 | * we can provide the correct devname to the caller. */ | |
958 | iface = NULL; | |
959 | for (j = 0; j < port->n_ifaces; j++) { | |
960 | struct iface *candidate = port->ifaces[j]; | |
961 | uint8_t candidate_ea[ETH_ADDR_LEN]; | |
8fef8c71 | 962 | if (!netdev_get_etheraddr(candidate->netdev, candidate_ea) |
ba09980a BP |
963 | && eth_addr_equals(iface_ea, candidate_ea)) { |
964 | iface = candidate; | |
965 | } | |
966 | } | |
58b7527e BP |
967 | } else { |
968 | /* Choose the interface whose MAC address will represent the port. | |
969 | * The Linux kernel bonding code always chooses the MAC address of | |
970 | * the first slave added to a bond, and the Fedora networking | |
971 | * scripts always add slaves to a bond in alphabetical order, so | |
972 | * for compatibility we choose the interface with the name that is | |
973 | * first in alphabetical order. */ | |
974 | iface = port->ifaces[0]; | |
975 | for (j = 1; j < port->n_ifaces; j++) { | |
976 | struct iface *candidate = port->ifaces[j]; | |
977 | if (strcmp(candidate->name, iface->name) < 0) { | |
978 | iface = candidate; | |
979 | } | |
980 | } | |
981 | ||
982 | /* The local port doesn't count (since we're trying to choose its | |
141f4942 BP |
983 | * MAC address anyway). */ |
984 | if (iface->dp_ifidx == ODPP_LOCAL) { | |
064af421 BP |
985 | continue; |
986 | } | |
58b7527e BP |
987 | |
988 | /* Grab MAC. */ | |
07c318f4 | 989 | error = netdev_get_etheraddr(iface->netdev, iface_ea); |
58b7527e | 990 | if (error) { |
064af421 BP |
991 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); |
992 | VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s", | |
993 | iface->name, strerror(error)); | |
58b7527e | 994 | continue; |
064af421 BP |
995 | } |
996 | } | |
58b7527e BP |
997 | |
998 | /* Compare against our current choice. */ | |
999 | if (!eth_addr_is_multicast(iface_ea) && | |
141f4942 | 1000 | !eth_addr_is_local(iface_ea) && |
58b7527e BP |
1001 | !eth_addr_is_reserved(iface_ea) && |
1002 | !eth_addr_is_zero(iface_ea) && | |
1003 | memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0) | |
1004 | { | |
0babc06f | 1005 | memcpy(ea, iface_ea, ETH_ADDR_LEN); |
8fef8c71 | 1006 | *hw_addr_iface = iface; |
58b7527e | 1007 | } |
064af421 | 1008 | } |
141f4942 | 1009 | if (eth_addr_is_multicast(ea)) { |
064af421 | 1010 | memcpy(ea, br->default_ea, ETH_ADDR_LEN); |
07c318f4 | 1011 | *hw_addr_iface = NULL; |
064af421 BP |
1012 | VLOG_WARN("bridge %s: using default bridge Ethernet " |
1013 | "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); | |
1014 | } else { | |
1015 | VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT, | |
1016 | br->name, ETH_ADDR_ARGS(ea)); | |
1017 | } | |
1018 | } | |
1019 | ||
1020 | /* Choose and returns the datapath ID for bridge 'br' given that the bridge | |
1021 | * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of | |
07c318f4 BP |
1022 | * an interface on 'br', then that interface must be passed in as |
1023 | * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then | |
1024 | * 'hw_addr_iface' must be passed in as a null pointer. */ | |
064af421 BP |
1025 | static uint64_t |
1026 | bridge_pick_datapath_id(struct bridge *br, | |
1027 | const uint8_t bridge_ea[ETH_ADDR_LEN], | |
07c318f4 | 1028 | struct iface *hw_addr_iface) |
064af421 BP |
1029 | { |
1030 | /* | |
1031 | * The procedure for choosing a bridge MAC address will, in the most | |
1032 | * ordinary case, also choose a unique MAC that we can use as a datapath | |
1033 | * ID. In some special cases, though, multiple bridges will end up with | |
1034 | * the same MAC address. This is OK for the bridges, but it will confuse | |
1035 | * the OpenFlow controller, because each datapath needs a unique datapath | |
1036 | * ID. | |
1037 | * | |
1038 | * Datapath IDs must be unique. It is also very desirable that they be | |
1039 | * stable from one run to the next, so that policy set on a datapath | |
1040 | * "sticks". | |
1041 | */ | |
093e47f4 | 1042 | const char *datapath_id; |
064af421 BP |
1043 | uint64_t dpid; |
1044 | ||
093e47f4 BP |
1045 | datapath_id = bridge_get_other_config(br->cfg, "datapath-id"); |
1046 | if (datapath_id && dpid_from_string(datapath_id, &dpid)) { | |
064af421 BP |
1047 | return dpid; |
1048 | } | |
1049 | ||
07c318f4 | 1050 | if (hw_addr_iface) { |
064af421 | 1051 | int vlan; |
b2f17b15 | 1052 | if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) { |
064af421 BP |
1053 | /* |
1054 | * A bridge whose MAC address is taken from a VLAN network device | |
1055 | * (that is, a network device created with vconfig(8) or similar | |
1056 | * tool) will have the same MAC address as a bridge on the VLAN | |
1057 | * device's physical network device. | |
1058 | * | |
1059 | * Handle this case by hashing the physical network device MAC | |
1060 | * along with the VLAN identifier. | |
1061 | */ | |
1062 | uint8_t buf[ETH_ADDR_LEN + 2]; | |
1063 | memcpy(buf, bridge_ea, ETH_ADDR_LEN); | |
1064 | buf[ETH_ADDR_LEN] = vlan >> 8; | |
1065 | buf[ETH_ADDR_LEN + 1] = vlan; | |
1066 | return dpid_from_hash(buf, sizeof buf); | |
1067 | } else { | |
1068 | /* | |
1069 | * Assume that this bridge's MAC address is unique, since it | |
1070 | * doesn't fit any of the cases we handle specially. | |
1071 | */ | |
1072 | } | |
1073 | } else { | |
1074 | /* | |
1075 | * A purely internal bridge, that is, one that has no non-virtual | |
1076 | * network devices on it at all, is more difficult because it has no | |
1077 | * natural unique identifier at all. | |
1078 | * | |
1079 | * When the host is a XenServer, we handle this case by hashing the | |
1080 | * host's UUID with the name of the bridge. Names of bridges are | |
1081 | * persistent across XenServer reboots, although they can be reused if | |
1082 | * an internal network is destroyed and then a new one is later | |
1083 | * created, so this is fairly effective. | |
1084 | * | |
1085 | * When the host is not a XenServer, we punt by using a random MAC | |
1086 | * address on each run. | |
1087 | */ | |
1088 | const char *host_uuid = xenserver_get_host_uuid(); | |
1089 | if (host_uuid) { | |
1090 | char *combined = xasprintf("%s,%s", host_uuid, br->name); | |
1091 | dpid = dpid_from_hash(combined, strlen(combined)); | |
1092 | free(combined); | |
1093 | return dpid; | |
1094 | } | |
1095 | } | |
1096 | ||
1097 | return eth_addr_to_uint64(bridge_ea); | |
1098 | } | |
1099 | ||
1100 | static uint64_t | |
1101 | dpid_from_hash(const void *data, size_t n) | |
1102 | { | |
5eccf359 | 1103 | uint8_t hash[SHA1_DIGEST_SIZE]; |
064af421 BP |
1104 | |
1105 | BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN); | |
5eccf359 | 1106 | sha1_bytes(data, n, hash); |
064af421 BP |
1107 | eth_addr_mark_random(hash); |
1108 | return eth_addr_to_uint64(hash); | |
1109 | } | |
1110 | ||
ea83a2fc | 1111 | static void |
ea763e0e | 1112 | iface_refresh_status(struct iface *iface) |
ea83a2fc | 1113 | { |
ea763e0e EJ |
1114 | struct shash sh; |
1115 | ||
e210037e AE |
1116 | enum netdev_flags flags; |
1117 | uint32_t current; | |
1118 | int64_t bps; | |
1119 | int mtu; | |
1120 | int64_t mtu_64; | |
1121 | int error; | |
1122 | ||
ea763e0e EJ |
1123 | shash_init(&sh); |
1124 | ||
1125 | if (!netdev_get_status(iface->netdev, &sh)) { | |
1126 | size_t n; | |
1127 | char **keys, **values; | |
ea83a2fc | 1128 | |
ea763e0e EJ |
1129 | shash_to_ovs_idl_map(&sh, &keys, &values, &n); |
1130 | ovsrec_interface_set_status(iface->cfg, keys, values, n); | |
1131 | ||
1132 | free(keys); | |
1133 | free(values); | |
1134 | } else { | |
1135 | ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0); | |
1136 | } | |
1137 | ||
1138 | shash_destroy_free_data(&sh); | |
e210037e AE |
1139 | |
1140 | error = netdev_get_flags(iface->netdev, &flags); | |
1141 | if (!error) { | |
1142 | ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down"); | |
1143 | } | |
1144 | else { | |
1145 | ovsrec_interface_set_admin_state(iface->cfg, NULL); | |
1146 | } | |
1147 | ||
1148 | error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL); | |
1149 | if (!error) { | |
1150 | ovsrec_interface_set_duplex(iface->cfg, | |
1151 | netdev_features_is_full_duplex(current) | |
1152 | ? "full" : "half"); | |
1153 | /* warning: uint64_t -> int64_t conversion */ | |
1154 | bps = netdev_features_to_bps(current); | |
1155 | ovsrec_interface_set_link_speed(iface->cfg, &bps, 1); | |
1156 | } | |
1157 | else { | |
1158 | ovsrec_interface_set_duplex(iface->cfg, NULL); | |
1159 | ovsrec_interface_set_link_speed(iface->cfg, NULL, 0); | |
1160 | } | |
1161 | ||
1162 | ||
1163 | ovsrec_interface_set_link_state(iface->cfg, | |
1164 | netdev_get_carrier(iface->netdev) | |
1165 | ? "up" : "down"); | |
1166 | ||
1167 | error = netdev_get_mtu(iface->netdev, &mtu); | |
1168 | if (!error) { | |
1169 | mtu_64 = mtu; | |
1170 | ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1); | |
1171 | } | |
1172 | else { | |
1173 | ovsrec_interface_set_mtu(iface->cfg, NULL, 0); | |
1174 | } | |
ea83a2fc EJ |
1175 | } |
1176 | ||
b31bcf60 EJ |
1177 | static void |
1178 | iface_refresh_cfm_stats(struct iface *iface) | |
1179 | { | |
1180 | size_t i; | |
1181 | struct cfm *cfm; | |
1182 | const struct ovsrec_monitor *mon; | |
1183 | ||
1184 | mon = iface->cfg->monitor; | |
1185 | cfm = iface->cfm; | |
1186 | ||
1187 | if (!cfm || !mon) { | |
1188 | return; | |
1189 | } | |
1190 | ||
1191 | for (i = 0; i < mon->n_remote_mps; i++) { | |
1192 | const struct ovsrec_maintenance_point *mp; | |
1193 | const struct remote_mp *rmp; | |
1194 | ||
1195 | mp = mon->remote_mps[i]; | |
1196 | rmp = cfm_get_remote_mp(cfm, mp->mpid); | |
1197 | ||
1198 | ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1); | |
1199 | } | |
1200 | ||
1201 | if (hmap_is_empty(&cfm->x_remote_mps)) { | |
1202 | ovsrec_monitor_set_unexpected_remote_mpids(mon, NULL, 0); | |
1203 | } else { | |
1204 | size_t length; | |
1205 | struct remote_mp *rmp; | |
1206 | int64_t *x_remote_mps; | |
1207 | ||
1208 | length = hmap_count(&cfm->x_remote_mps); | |
1209 | x_remote_mps = xzalloc(length * sizeof *x_remote_mps); | |
1210 | ||
1211 | i = 0; | |
1212 | HMAP_FOR_EACH (rmp, node, &cfm->x_remote_mps) { | |
1213 | x_remote_mps[i++] = rmp->mpid; | |
1214 | } | |
1215 | ||
1216 | ovsrec_monitor_set_unexpected_remote_mpids(mon, x_remote_mps, length); | |
1217 | free(x_remote_mps); | |
1218 | } | |
1219 | ||
1220 | if (hmap_is_empty(&cfm->x_remote_maids)) { | |
1221 | ovsrec_monitor_set_unexpected_remote_maids(mon, NULL, 0); | |
1222 | } else { | |
1223 | size_t length; | |
1224 | char **x_remote_maids; | |
1225 | struct remote_maid *rmaid; | |
1226 | ||
1227 | length = hmap_count(&cfm->x_remote_maids); | |
1228 | x_remote_maids = xzalloc(length * sizeof *x_remote_maids); | |
1229 | ||
1230 | i = 0; | |
1231 | HMAP_FOR_EACH (rmaid, node, &cfm->x_remote_maids) { | |
1232 | size_t j; | |
1233 | ||
1234 | x_remote_maids[i] = xzalloc(CCM_MAID_LEN * 2 + 1); | |
1235 | ||
1236 | for (j = 0; j < CCM_MAID_LEN; j++) { | |
1237 | snprintf(&x_remote_maids[i][j * 2], 3, "%02hhx", | |
1238 | rmaid->maid[j]); | |
1239 | } | |
1240 | i++; | |
1241 | } | |
1242 | ovsrec_monitor_set_unexpected_remote_maids(mon, x_remote_maids, length); | |
1243 | ||
1244 | for (i = 0; i < length; i++) { | |
1245 | free(x_remote_maids[i]); | |
1246 | } | |
1247 | free(x_remote_maids); | |
1248 | } | |
1249 | ||
1250 | ovsrec_monitor_set_fault(mon, &cfm->fault, 1); | |
1251 | } | |
1252 | ||
018f1525 BP |
1253 | static void |
1254 | iface_refresh_stats(struct iface *iface) | |
1255 | { | |
1256 | struct iface_stat { | |
1257 | char *name; | |
1258 | int offset; | |
1259 | }; | |
1260 | static const struct iface_stat iface_stats[] = { | |
1261 | { "rx_packets", offsetof(struct netdev_stats, rx_packets) }, | |
1262 | { "tx_packets", offsetof(struct netdev_stats, tx_packets) }, | |
1263 | { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) }, | |
1264 | { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) }, | |
1265 | { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) }, | |
1266 | { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) }, | |
1267 | { "rx_errors", offsetof(struct netdev_stats, rx_errors) }, | |
1268 | { "tx_errors", offsetof(struct netdev_stats, tx_errors) }, | |
1269 | { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) }, | |
1270 | { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) }, | |
1271 | { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) }, | |
1272 | { "collisions", offsetof(struct netdev_stats, collisions) }, | |
1273 | }; | |
1274 | enum { N_STATS = ARRAY_SIZE(iface_stats) }; | |
1275 | const struct iface_stat *s; | |
1276 | ||
1277 | char *keys[N_STATS]; | |
1278 | int64_t values[N_STATS]; | |
1279 | int n; | |
1280 | ||
1281 | struct netdev_stats stats; | |
1282 | ||
1283 | /* Intentionally ignore return value, since errors will set 'stats' to | |
1284 | * all-1s, and we will deal with that correctly below. */ | |
1285 | netdev_get_stats(iface->netdev, &stats); | |
1286 | ||
1287 | n = 0; | |
1288 | for (s = iface_stats; s < &iface_stats[N_STATS]; s++) { | |
1289 | uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset); | |
1290 | if (value != UINT64_MAX) { | |
1291 | keys[n] = s->name; | |
1292 | values[n] = value; | |
1293 | n++; | |
1294 | } | |
1295 | } | |
1296 | ||
1297 | ovsrec_interface_set_statistics(iface->cfg, keys, values, n); | |
1298 | } | |
1299 | ||
ce887677 BP |
1300 | static void |
1301 | refresh_system_stats(const struct ovsrec_open_vswitch *cfg) | |
1302 | { | |
1303 | struct ovsdb_datum datum; | |
1304 | struct shash stats; | |
1305 | ||
1306 | shash_init(&stats); | |
1307 | get_system_stats(&stats); | |
1308 | ||
1309 | ovsdb_datum_from_shash(&datum, &stats); | |
1310 | ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, | |
1311 | &datum); | |
1312 | } | |
1313 | ||
bffc0589 AE |
1314 | static inline const char * |
1315 | nx_role_to_str(enum nx_role role) | |
1316 | { | |
1317 | switch (role) { | |
1318 | case NX_ROLE_OTHER: | |
1319 | return "other"; | |
1320 | case NX_ROLE_MASTER: | |
1321 | return "master"; | |
1322 | case NX_ROLE_SLAVE: | |
1323 | return "slave"; | |
1324 | default: | |
1325 | return "*** INVALID ROLE ***"; | |
1326 | } | |
1327 | } | |
1328 | ||
1329 | static void | |
1330 | bridge_refresh_controller_status(const struct bridge *br) | |
1331 | { | |
1332 | struct shash info; | |
1333 | const struct ovsrec_controller *cfg; | |
1334 | ||
1335 | ofproto_get_ofproto_controller_info(br->ofproto, &info); | |
1336 | ||
1337 | OVSREC_CONTROLLER_FOR_EACH(cfg, idl) { | |
eb9b8307 AE |
1338 | struct ofproto_controller_info *cinfo = |
1339 | shash_find_data(&info, cfg->target); | |
1340 | ||
1341 | if (cinfo) { | |
1342 | ovsrec_controller_set_is_connected(cfg, cinfo->is_connected); | |
1343 | ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role)); | |
1344 | ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys, | |
1345 | (char **) cinfo->pairs.values, | |
1346 | cinfo->pairs.n); | |
1347 | } else { | |
1348 | ovsrec_controller_set_is_connected(cfg, false); | |
1349 | ovsrec_controller_set_role(cfg, NULL); | |
1350 | ovsrec_controller_set_status(cfg, NULL, NULL, 0); | |
1351 | } | |
bffc0589 AE |
1352 | } |
1353 | ||
1354 | ofproto_free_ofproto_controller_info(&info); | |
1355 | } | |
1356 | ||
c5187f17 | 1357 | void |
064af421 BP |
1358 | bridge_run(void) |
1359 | { | |
d54ff998 BP |
1360 | const struct ovsrec_open_vswitch *cfg; |
1361 | ||
c5187f17 | 1362 | bool datapath_destroyed; |
d54ff998 | 1363 | bool database_changed; |
c5187f17 | 1364 | struct bridge *br; |
064af421 | 1365 | |
c5187f17 BP |
1366 | /* Let each bridge do the work that it needs to do. */ |
1367 | datapath_destroyed = false; | |
4e8e4213 | 1368 | LIST_FOR_EACH (br, node, &all_bridges) { |
064af421 BP |
1369 | int error = bridge_run_one(br); |
1370 | if (error) { | |
1371 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1372 | VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, " | |
1373 | "forcing reconfiguration", br->name); | |
c5187f17 BP |
1374 | datapath_destroyed = true; |
1375 | } | |
1376 | } | |
1377 | ||
1378 | /* (Re)configure if necessary. */ | |
d54ff998 BP |
1379 | database_changed = ovsdb_idl_run(idl); |
1380 | cfg = ovsrec_open_vswitch_first(idl); | |
d6da96ce BP |
1381 | #ifdef HAVE_OPENSSL |
1382 | /* Re-configure SSL. We do this on every trip through the main loop, | |
1383 | * instead of just when the database changes, because the contents of the | |
1384 | * key and certificate files can change without the database changing. | |
1385 | * | |
1386 | * We do this before bridge_reconfigure() because that function might | |
1387 | * initiate SSL connections and thus requires SSL to be configured. */ | |
1388 | if (cfg && cfg->ssl) { | |
1389 | const struct ovsrec_ssl *ssl = cfg->ssl; | |
1390 | ||
1391 | stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate); | |
1392 | stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); | |
1393 | } | |
1394 | #endif | |
d54ff998 | 1395 | if (database_changed || datapath_destroyed) { |
c5187f17 BP |
1396 | if (cfg) { |
1397 | struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); | |
1398 | ||
1399 | bridge_configure_once(cfg); | |
1400 | bridge_reconfigure(cfg); | |
1401 | ||
1402 | ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg); | |
1403 | ovsdb_idl_txn_commit(txn); | |
1404 | ovsdb_idl_txn_destroy(txn); /* XXX */ | |
1e0b752d BP |
1405 | } else { |
1406 | /* We still need to reconfigure to avoid dangling pointers to | |
1407 | * now-destroyed ovsrec structures inside bridge data. */ | |
1408 | static const struct ovsrec_open_vswitch null_cfg; | |
1409 | ||
1410 | bridge_reconfigure(&null_cfg); | |
064af421 BP |
1411 | } |
1412 | } | |
018f1525 | 1413 | |
cd0cd65f BP |
1414 | /* Refresh system and interface stats if necessary. */ |
1415 | if (time_msec() >= stats_timer) { | |
ce887677 BP |
1416 | if (cfg) { |
1417 | struct ovsdb_idl_txn *txn; | |
018f1525 | 1418 | |
ce887677 | 1419 | txn = ovsdb_idl_txn_create(idl); |
4e8e4213 | 1420 | LIST_FOR_EACH (br, node, &all_bridges) { |
ce887677 | 1421 | size_t i; |
018f1525 | 1422 | |
ce887677 BP |
1423 | for (i = 0; i < br->n_ports; i++) { |
1424 | struct port *port = br->ports[i]; | |
1425 | size_t j; | |
018f1525 | 1426 | |
ce887677 BP |
1427 | for (j = 0; j < port->n_ifaces; j++) { |
1428 | struct iface *iface = port->ifaces[j]; | |
1429 | iface_refresh_stats(iface); | |
b31bcf60 | 1430 | iface_refresh_cfm_stats(iface); |
ea763e0e | 1431 | iface_refresh_status(iface); |
ce887677 | 1432 | } |
018f1525 | 1433 | } |
bffc0589 | 1434 | bridge_refresh_controller_status(br); |
018f1525 | 1435 | } |
ce887677 BP |
1436 | refresh_system_stats(cfg); |
1437 | ovsdb_idl_txn_commit(txn); | |
1438 | ovsdb_idl_txn_destroy(txn); /* XXX */ | |
018f1525 | 1439 | } |
018f1525 | 1440 | |
cd0cd65f | 1441 | stats_timer = time_msec() + STATS_INTERVAL; |
018f1525 | 1442 | } |
064af421 BP |
1443 | } |
1444 | ||
1445 | void | |
1446 | bridge_wait(void) | |
1447 | { | |
1448 | struct bridge *br; | |
b31bcf60 | 1449 | struct iface *iface; |
064af421 | 1450 | |
4e8e4213 | 1451 | LIST_FOR_EACH (br, node, &all_bridges) { |
064af421 | 1452 | ofproto_wait(br->ofproto); |
7d674866 | 1453 | if (ofproto_has_primary_controller(br->ofproto)) { |
064af421 BP |
1454 | continue; |
1455 | } | |
1456 | ||
93dfc067 | 1457 | mac_learning_wait(br->ml); |
064af421 | 1458 | bond_wait(br); |
b31bcf60 EJ |
1459 | |
1460 | HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { | |
1461 | if (iface->cfm) { | |
1462 | cfm_wait(iface->cfm); | |
1463 | } | |
1464 | } | |
064af421 | 1465 | } |
c5187f17 | 1466 | ovsdb_idl_wait(idl); |
cd0cd65f | 1467 | poll_timer_wait_until(stats_timer); |
064af421 BP |
1468 | } |
1469 | ||
1470 | /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s | |
1471 | * configuration changes. */ | |
1472 | static void | |
1473 | bridge_flush(struct bridge *br) | |
1474 | { | |
1475 | COVERAGE_INC(bridge_flush); | |
1476 | br->flush = true; | |
93dfc067 | 1477 | mac_learning_flush(br->ml); |
064af421 | 1478 | } |
07c318f4 BP |
1479 | |
1480 | /* Returns the 'br' interface for the ODPP_LOCAL port, or null if 'br' has no | |
1481 | * such interface. */ | |
1482 | static struct iface * | |
1483 | bridge_get_local_iface(struct bridge *br) | |
1484 | { | |
1485 | size_t i, j; | |
1486 | ||
1487 | for (i = 0; i < br->n_ports; i++) { | |
1488 | struct port *port = br->ports[i]; | |
1489 | for (j = 0; j < port->n_ifaces; j++) { | |
1490 | struct iface *iface = port->ifaces[j]; | |
1491 | if (iface->dp_ifidx == ODPP_LOCAL) { | |
1492 | return iface; | |
1493 | } | |
1494 | } | |
1495 | } | |
1496 | ||
1497 | return NULL; | |
1498 | } | |
064af421 | 1499 | \f |
8c4c1387 BP |
1500 | /* Bridge unixctl user interface functions. */ |
1501 | static void | |
8ca79daa | 1502 | bridge_unixctl_fdb_show(struct unixctl_conn *conn, |
c69ee87c | 1503 | const char *args, void *aux OVS_UNUSED) |
8c4c1387 BP |
1504 | { |
1505 | struct ds ds = DS_EMPTY_INITIALIZER; | |
1506 | const struct bridge *br; | |
93dfc067 | 1507 | const struct mac_entry *e; |
8c4c1387 BP |
1508 | |
1509 | br = bridge_lookup(args); | |
1510 | if (!br) { | |
1511 | unixctl_command_reply(conn, 501, "no such bridge"); | |
1512 | return; | |
1513 | } | |
1514 | ||
1515 | ds_put_cstr(&ds, " port VLAN MAC Age\n"); | |
4e8e4213 | 1516 | LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { |
93dfc067 JG |
1517 | if (e->port < 0 || e->port >= br->n_ports) { |
1518 | continue; | |
8c4c1387 | 1519 | } |
93dfc067 JG |
1520 | ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n", |
1521 | br->ports[e->port]->ifaces[0]->dp_ifidx, | |
1522 | e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e)); | |
8c4c1387 BP |
1523 | } |
1524 | unixctl_command_reply(conn, 200, ds_cstr(&ds)); | |
1525 | ds_destroy(&ds); | |
1526 | } | |
1527 | \f | |
064af421 | 1528 | /* Bridge reconfiguration functions. */ |
064af421 | 1529 | static struct bridge * |
1a6f1e2a | 1530 | bridge_create(const struct ovsrec_bridge *br_cfg) |
064af421 BP |
1531 | { |
1532 | struct bridge *br; | |
1533 | int error; | |
1534 | ||
1a6f1e2a | 1535 | assert(!bridge_lookup(br_cfg->name)); |
ec6fde61 | 1536 | br = xzalloc(sizeof *br); |
064af421 | 1537 | |
1a6f1e2a JG |
1538 | error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type, |
1539 | &br->dpif); | |
efacbce6 | 1540 | if (error) { |
064af421 BP |
1541 | free(br); |
1542 | return NULL; | |
1543 | } | |
efacbce6 | 1544 | dpif_flow_flush(br->dpif); |
064af421 | 1545 | |
1a6f1e2a JG |
1546 | error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks, |
1547 | br, &br->ofproto); | |
064af421 | 1548 | if (error) { |
1a6f1e2a JG |
1549 | VLOG_ERR("failed to create switch %s: %s", br_cfg->name, |
1550 | strerror(error)); | |
c228a364 BP |
1551 | dpif_delete(br->dpif); |
1552 | dpif_close(br->dpif); | |
064af421 BP |
1553 | free(br); |
1554 | return NULL; | |
1555 | } | |
1556 | ||
1a6f1e2a JG |
1557 | br->name = xstrdup(br_cfg->name); |
1558 | br->cfg = br_cfg; | |
064af421 | 1559 | br->ml = mac_learning_create(); |
70150daf | 1560 | eth_addr_nicira_random(br->default_ea); |
064af421 | 1561 | |
d9a8717a | 1562 | hmap_init(&br->ifaces); |
064af421 | 1563 | |
836fad5e | 1564 | shash_init(&br->port_by_name); |
4a1ee6ae BP |
1565 | shash_init(&br->iface_by_name); |
1566 | ||
064af421 | 1567 | br->flush = false; |
064af421 BP |
1568 | |
1569 | list_push_back(&all_bridges, &br->node); | |
1570 | ||
c228a364 | 1571 | VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif)); |
064af421 BP |
1572 | |
1573 | return br; | |
1574 | } | |
1575 | ||
1576 | static void | |
1577 | bridge_destroy(struct bridge *br) | |
1578 | { | |
1579 | if (br) { | |
1580 | int error; | |
1581 | ||
1582 | while (br->n_ports > 0) { | |
1583 | port_destroy(br->ports[br->n_ports - 1]); | |
1584 | } | |
1585 | list_remove(&br->node); | |
c228a364 | 1586 | error = dpif_delete(br->dpif); |
064af421 | 1587 | if (error && error != ENOENT) { |
b29ba128 | 1588 | VLOG_ERR("failed to delete %s: %s", |
c228a364 | 1589 | dpif_name(br->dpif), strerror(error)); |
064af421 | 1590 | } |
c228a364 | 1591 | dpif_close(br->dpif); |
064af421 | 1592 | ofproto_destroy(br->ofproto); |
064af421 | 1593 | mac_learning_destroy(br->ml); |
d9a8717a | 1594 | hmap_destroy(&br->ifaces); |
836fad5e | 1595 | shash_destroy(&br->port_by_name); |
4a1ee6ae | 1596 | shash_destroy(&br->iface_by_name); |
064af421 BP |
1597 | free(br->ports); |
1598 | free(br->name); | |
1599 | free(br); | |
1600 | } | |
1601 | } | |
1602 | ||
1603 | static struct bridge * | |
1604 | bridge_lookup(const char *name) | |
1605 | { | |
1606 | struct bridge *br; | |
1607 | ||
4e8e4213 | 1608 | LIST_FOR_EACH (br, node, &all_bridges) { |
064af421 BP |
1609 | if (!strcmp(br->name, name)) { |
1610 | return br; | |
1611 | } | |
1612 | } | |
1613 | return NULL; | |
1614 | } | |
1615 | ||
4f2cad2c JP |
1616 | /* Handle requests for a listing of all flows known by the OpenFlow |
1617 | * stack, including those normally hidden. */ | |
1618 | static void | |
8ca79daa | 1619 | bridge_unixctl_dump_flows(struct unixctl_conn *conn, |
c69ee87c | 1620 | const char *args, void *aux OVS_UNUSED) |
4f2cad2c JP |
1621 | { |
1622 | struct bridge *br; | |
1623 | struct ds results; | |
d295e8e9 | 1624 | |
4f2cad2c JP |
1625 | br = bridge_lookup(args); |
1626 | if (!br) { | |
1627 | unixctl_command_reply(conn, 501, "Unknown bridge"); | |
1628 | return; | |
1629 | } | |
1630 | ||
1631 | ds_init(&results); | |
1632 | ofproto_get_all_flows(br->ofproto, &results); | |
1633 | ||
1634 | unixctl_command_reply(conn, 200, ds_cstr(&results)); | |
1635 | ds_destroy(&results); | |
1636 | } | |
1637 | ||
fa05809b BP |
1638 | /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller |
1639 | * connections and reconnect. If BRIDGE is not specified, then all bridges | |
1640 | * drop their controller connections and reconnect. */ | |
1641 | static void | |
1642 | bridge_unixctl_reconnect(struct unixctl_conn *conn, | |
1643 | const char *args, void *aux OVS_UNUSED) | |
1644 | { | |
1645 | struct bridge *br; | |
1646 | if (args[0] != '\0') { | |
1647 | br = bridge_lookup(args); | |
1648 | if (!br) { | |
1649 | unixctl_command_reply(conn, 501, "Unknown bridge"); | |
1650 | return; | |
1651 | } | |
1652 | ofproto_reconnect_controllers(br->ofproto); | |
1653 | } else { | |
4e8e4213 | 1654 | LIST_FOR_EACH (br, node, &all_bridges) { |
fa05809b BP |
1655 | ofproto_reconnect_controllers(br->ofproto); |
1656 | } | |
1657 | } | |
1658 | unixctl_command_reply(conn, 200, NULL); | |
1659 | } | |
1660 | ||
064af421 BP |
1661 | static int |
1662 | bridge_run_one(struct bridge *br) | |
1663 | { | |
1664 | int error; | |
b31bcf60 | 1665 | struct iface *iface; |
064af421 BP |
1666 | |
1667 | error = ofproto_run1(br->ofproto); | |
1668 | if (error) { | |
1669 | return error; | |
1670 | } | |
1671 | ||
93dfc067 | 1672 | mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto)); |
064af421 | 1673 | bond_run(br); |
064af421 BP |
1674 | |
1675 | error = ofproto_run2(br->ofproto, br->flush); | |
1676 | br->flush = false; | |
1677 | ||
b31bcf60 EJ |
1678 | HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { |
1679 | struct ofpbuf *packet; | |
1680 | ||
1681 | if (!iface->cfm) { | |
1682 | continue; | |
1683 | } | |
1684 | ||
1685 | packet = cfm_run(iface->cfm); | |
1686 | if (packet) { | |
1687 | iface_send_packet(iface, packet); | |
1688 | ofpbuf_uninit(packet); | |
1689 | free(packet); | |
1690 | } | |
1691 | } | |
1692 | ||
064af421 BP |
1693 | return error; |
1694 | } | |
1695 | ||
76ce9432 | 1696 | static size_t |
1a048029 | 1697 | bridge_get_controllers(const struct bridge *br, |
76ce9432 | 1698 | struct ovsrec_controller ***controllersp) |
064af421 | 1699 | { |
76ce9432 BP |
1700 | struct ovsrec_controller **controllers; |
1701 | size_t n_controllers; | |
064af421 | 1702 | |
1a048029 JP |
1703 | controllers = br->cfg->controller; |
1704 | n_controllers = br->cfg->n_controller; | |
76343538 | 1705 | |
76ce9432 BP |
1706 | if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) { |
1707 | controllers = NULL; | |
1708 | n_controllers = 0; | |
064af421 | 1709 | } |
76343538 | 1710 | |
76ce9432 BP |
1711 | if (controllersp) { |
1712 | *controllersp = controllers; | |
1713 | } | |
1714 | return n_controllers; | |
064af421 BP |
1715 | } |
1716 | ||
1717 | static void | |
1a048029 | 1718 | bridge_reconfigure_one(struct bridge *br) |
064af421 | 1719 | { |
76343538 | 1720 | struct shash old_ports, new_ports; |
064af421 | 1721 | struct svec snoops, old_snoops; |
76343538 | 1722 | struct shash_node *node; |
31681a5d | 1723 | enum ofproto_fail_mode fail_mode; |
6ae39834 | 1724 | size_t i; |
064af421 BP |
1725 | |
1726 | /* Collect old ports. */ | |
76343538 | 1727 | shash_init(&old_ports); |
064af421 | 1728 | for (i = 0; i < br->n_ports; i++) { |
76343538 | 1729 | shash_add(&old_ports, br->ports[i]->name, br->ports[i]); |
064af421 | 1730 | } |
064af421 BP |
1731 | |
1732 | /* Collect new ports. */ | |
76343538 BP |
1733 | shash_init(&new_ports); |
1734 | for (i = 0; i < br->cfg->n_ports; i++) { | |
1735 | const char *name = br->cfg->ports[i]->name; | |
1736 | if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) { | |
1737 | VLOG_WARN("bridge %s: %s specified twice as bridge port", | |
1738 | br->name, name); | |
1739 | } | |
1740 | } | |
e073f944 BP |
1741 | |
1742 | /* If we have a controller, then we need a local port. Complain if the | |
1743 | * user didn't specify one. | |
1744 | * | |
1745 | * XXX perhaps we should synthesize a port ourselves in this case. */ | |
1a048029 | 1746 | if (bridge_get_controllers(br, NULL)) { |
72865317 BP |
1747 | char local_name[IF_NAMESIZE]; |
1748 | int error; | |
1749 | ||
1750 | error = dpif_port_get_name(br->dpif, ODPP_LOCAL, | |
1751 | local_name, sizeof local_name); | |
e073f944 BP |
1752 | if (!error && !shash_find(&new_ports, local_name)) { |
1753 | VLOG_WARN("bridge %s: controller specified but no local port " | |
1754 | "(port named %s) defined", | |
1755 | br->name, local_name); | |
72865317 | 1756 | } |
064af421 | 1757 | } |
064af421 | 1758 | |
4a1ee6ae BP |
1759 | /* Get rid of deleted ports. |
1760 | * Get rid of deleted interfaces on ports that still exist. */ | |
76343538 | 1761 | SHASH_FOR_EACH (node, &old_ports) { |
4a1ee6ae BP |
1762 | struct port *port = node->data; |
1763 | const struct ovsrec_port *port_cfg; | |
1764 | ||
1765 | port_cfg = shash_find_data(&new_ports, node->name); | |
1766 | if (!port_cfg) { | |
1767 | port_destroy(port); | |
1768 | } else { | |
1769 | port_del_ifaces(port, port_cfg); | |
064af421 BP |
1770 | } |
1771 | } | |
4a1ee6ae BP |
1772 | |
1773 | /* Create new ports. | |
1774 | * Add new interfaces to existing ports. | |
1775 | * Reconfigure existing ports. */ | |
76343538 BP |
1776 | SHASH_FOR_EACH (node, &new_ports) { |
1777 | struct port *port = shash_find_data(&old_ports, node->name); | |
1778 | if (!port) { | |
1779 | port = port_create(br, node->name); | |
064af421 | 1780 | } |
ceb4559f | 1781 | |
76343538 | 1782 | port_reconfigure(port, node->data); |
ceb4559f JG |
1783 | if (!port->n_ifaces) { |
1784 | VLOG_WARN("bridge %s: port %s has no interfaces, dropping", | |
1785 | br->name, port->name); | |
1786 | port_destroy(port); | |
1787 | } | |
064af421 | 1788 | } |
76343538 BP |
1789 | shash_destroy(&old_ports); |
1790 | shash_destroy(&new_ports); | |
064af421 | 1791 | |
31681a5d JP |
1792 | /* Set the fail-mode */ |
1793 | fail_mode = !br->cfg->fail_mode | |
1794 | || !strcmp(br->cfg->fail_mode, "standalone") | |
1795 | ? OFPROTO_FAIL_STANDALONE | |
1796 | : OFPROTO_FAIL_SECURE; | |
7d674866 BP |
1797 | if (ofproto_get_fail_mode(br->ofproto) != fail_mode |
1798 | && !ofproto_has_primary_controller(br->ofproto)) { | |
abdfe474 JP |
1799 | ofproto_flush_flows(br->ofproto); |
1800 | } | |
31681a5d JP |
1801 | ofproto_set_fail_mode(br->ofproto, fail_mode); |
1802 | ||
064af421 BP |
1803 | /* Delete all flows if we're switching from connected to standalone or vice |
1804 | * versa. (XXX Should we delete all flows if we are switching from one | |
1805 | * controller to another?) */ | |
1806 | ||
5a243150 | 1807 | /* Configure OpenFlow controller connection snooping. */ |
76343538 BP |
1808 | svec_init(&snoops); |
1809 | svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop", | |
b43c6fe2 | 1810 | ovs_rundir(), br->name)); |
76343538 BP |
1811 | svec_init(&old_snoops); |
1812 | ofproto_get_snoops(br->ofproto, &old_snoops); | |
1813 | if (!svec_equal(&snoops, &old_snoops)) { | |
1814 | ofproto_set_snoops(br->ofproto, &snoops); | |
1815 | } | |
1816 | svec_destroy(&snoops); | |
1817 | svec_destroy(&old_snoops); | |
76343538 | 1818 | |
064af421 BP |
1819 | mirror_reconfigure(br); |
1820 | } | |
1821 | ||
7d674866 BP |
1822 | /* Initializes 'oc' appropriately as a management service controller for |
1823 | * 'br'. | |
1824 | * | |
1825 | * The caller must free oc->target when it is no longer needed. */ | |
1826 | static void | |
1827 | bridge_ofproto_controller_for_mgmt(const struct bridge *br, | |
1828 | struct ofproto_controller *oc) | |
1829 | { | |
b43c6fe2 | 1830 | oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name); |
7d674866 BP |
1831 | oc->max_backoff = 0; |
1832 | oc->probe_interval = 60; | |
1833 | oc->band = OFPROTO_OUT_OF_BAND; | |
1834 | oc->accept_re = NULL; | |
1835 | oc->update_resolv_conf = false; | |
1836 | oc->rate_limit = 0; | |
1837 | oc->burst_limit = 0; | |
1838 | } | |
1839 | ||
1840 | /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */ | |
1841 | static void | |
1842 | bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, | |
1843 | struct ofproto_controller *oc) | |
1844 | { | |
1845 | oc->target = c->target; | |
1846 | oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; | |
1847 | oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5; | |
1848 | oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band") | |
1849 | ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND); | |
1850 | oc->accept_re = c->discover_accept_regex; | |
1851 | oc->update_resolv_conf = c->discover_update_resolv_conf; | |
1852 | oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0; | |
1853 | oc->burst_limit = (c->controller_burst_limit | |
1854 | ? *c->controller_burst_limit : 0); | |
1855 | } | |
1856 | ||
1857 | /* Configures the IP stack for 'br''s local interface properly according to the | |
1858 | * configuration in 'c'. */ | |
1859 | static void | |
1860 | bridge_configure_local_iface_netdev(struct bridge *br, | |
1861 | struct ovsrec_controller *c) | |
1862 | { | |
1863 | struct netdev *netdev; | |
1864 | struct in_addr mask, gateway; | |
1865 | ||
1866 | struct iface *local_iface; | |
1867 | struct in_addr ip; | |
1868 | ||
1869 | /* Controller discovery does its own TCP/IP configuration later. */ | |
1870 | if (strcmp(c->target, "discover")) { | |
1871 | return; | |
1872 | } | |
1873 | ||
1874 | /* If there's no local interface or no IP address, give up. */ | |
1875 | local_iface = bridge_get_local_iface(br); | |
1876 | if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) { | |
1877 | return; | |
1878 | } | |
1879 | ||
1880 | /* Bring up the local interface. */ | |
1881 | netdev = local_iface->netdev; | |
1882 | netdev_turn_flags_on(netdev, NETDEV_UP, true); | |
1883 | ||
1884 | /* Configure the IP address and netmask. */ | |
1885 | if (!c->local_netmask | |
1886 | || !inet_aton(c->local_netmask, &mask) | |
1887 | || !mask.s_addr) { | |
1888 | mask.s_addr = guess_netmask(ip.s_addr); | |
1889 | } | |
1890 | if (!netdev_set_in4(netdev, ip, mask)) { | |
1891 | VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT, | |
1892 | br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr)); | |
1893 | } | |
1894 | ||
1895 | /* Configure the default gateway. */ | |
1896 | if (c->local_gateway | |
1897 | && inet_aton(c->local_gateway, &gateway) | |
1898 | && gateway.s_addr) { | |
1899 | if (!netdev_add_router(netdev, gateway)) { | |
1900 | VLOG_INFO("bridge %s: configured gateway "IP_FMT, | |
1901 | br->name, IP_ARGS(&gateway.s_addr)); | |
1902 | } | |
1903 | } | |
1904 | } | |
1905 | ||
064af421 | 1906 | static void |
1a048029 | 1907 | bridge_reconfigure_remotes(struct bridge *br, |
11bbff1b BP |
1908 | const struct sockaddr_in *managers, |
1909 | size_t n_managers) | |
064af421 | 1910 | { |
b1da6250 BP |
1911 | const char *disable_ib_str, *queue_id_str; |
1912 | bool disable_in_band = false; | |
1913 | int queue_id; | |
1914 | ||
76ce9432 BP |
1915 | struct ovsrec_controller **controllers; |
1916 | size_t n_controllers; | |
7d674866 BP |
1917 | bool had_primary; |
1918 | ||
1919 | struct ofproto_controller *ocs; | |
1920 | size_t n_ocs; | |
1921 | size_t i; | |
064af421 | 1922 | |
8731b2b6 JP |
1923 | /* Check if we should disable in-band control on this bridge. */ |
1924 | disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band"); | |
1925 | if (disable_ib_str && !strcmp(disable_ib_str, "true")) { | |
1926 | disable_in_band = true; | |
1927 | } | |
1928 | ||
b1da6250 BP |
1929 | /* Set OpenFlow queue ID for in-band control. */ |
1930 | queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue"); | |
1931 | queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1; | |
1932 | ofproto_set_in_band_queue(br->ofproto, queue_id); | |
1933 | ||
8731b2b6 JP |
1934 | if (disable_in_band) { |
1935 | ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0); | |
1936 | } else { | |
1937 | ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); | |
1938 | } | |
7d674866 | 1939 | had_primary = ofproto_has_primary_controller(br->ofproto); |
cd11000b | 1940 | |
1a048029 | 1941 | n_controllers = bridge_get_controllers(br, &controllers); |
064af421 | 1942 | |
7d674866 BP |
1943 | ocs = xmalloc((n_controllers + 1) * sizeof *ocs); |
1944 | n_ocs = 0; | |
064af421 | 1945 | |
7d674866 BP |
1946 | bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]); |
1947 | for (i = 0; i < n_controllers; i++) { | |
1948 | struct ovsrec_controller *c = controllers[i]; | |
76ce9432 | 1949 | |
7d674866 BP |
1950 | if (!strncmp(c->target, "punix:", 6) |
1951 | || !strncmp(c->target, "unix:", 5)) { | |
1952 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
76ce9432 | 1953 | |
7d674866 BP |
1954 | /* Prevent remote ovsdb-server users from accessing arbitrary Unix |
1955 | * domain sockets and overwriting arbitrary local files. */ | |
1956 | VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller " | |
1957 | "\"%s\" due to possibility for remote exploit", | |
1958 | dpif_name(br->dpif), c->target); | |
1959 | continue; | |
1960 | } | |
76ce9432 | 1961 | |
7d674866 | 1962 | bridge_configure_local_iface_netdev(br, c); |
8731b2b6 JP |
1963 | bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]); |
1964 | if (disable_in_band) { | |
1965 | ocs[n_ocs].band = OFPROTO_OUT_OF_BAND; | |
1966 | } | |
1967 | n_ocs++; | |
7d674866 | 1968 | } |
76ce9432 | 1969 | |
7d674866 BP |
1970 | ofproto_set_controllers(br->ofproto, ocs, n_ocs); |
1971 | free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */ | |
1972 | free(ocs); | |
064af421 | 1973 | |
7d674866 BP |
1974 | if (had_primary != ofproto_has_primary_controller(br->ofproto)) { |
1975 | ofproto_flush_flows(br->ofproto); | |
1976 | } | |
76ce9432 | 1977 | |
7d674866 BP |
1978 | /* If there are no controllers and the bridge is in standalone |
1979 | * mode, set up a flow that matches every packet and directs | |
1980 | * them to OFPP_NORMAL (which goes to us). Otherwise, the | |
1981 | * switch is in secure mode and we won't pass any traffic until | |
1982 | * a controller has been defined and it tells us to do so. */ | |
1983 | if (!n_controllers | |
1984 | && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) { | |
1985 | union ofp_action action; | |
cf3fad8a | 1986 | struct cls_rule rule; |
76ce9432 | 1987 | |
7d674866 BP |
1988 | memset(&action, 0, sizeof action); |
1989 | action.type = htons(OFPAT_OUTPUT); | |
1990 | action.output.len = htons(sizeof action); | |
1991 | action.output.port = htons(OFPP_NORMAL); | |
cf3fad8a | 1992 | cls_rule_init_catchall(&rule, 0); |
fa8b054f | 1993 | ofproto_add_flow(br->ofproto, &rule, &action, 1); |
064af421 | 1994 | } |
064af421 BP |
1995 | } |
1996 | ||
1997 | static void | |
76343538 | 1998 | bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces) |
064af421 BP |
1999 | { |
2000 | size_t i, j; | |
2001 | ||
76343538 | 2002 | shash_init(ifaces); |
064af421 BP |
2003 | for (i = 0; i < br->n_ports; i++) { |
2004 | struct port *port = br->ports[i]; | |
2005 | for (j = 0; j < port->n_ifaces; j++) { | |
2006 | struct iface *iface = port->ifaces[j]; | |
76343538 | 2007 | shash_add_once(ifaces, iface->name, iface); |
064af421 | 2008 | } |
76343538 BP |
2009 | if (port->n_ifaces > 1 && port->cfg->bond_fake_iface) { |
2010 | shash_add_once(ifaces, port->name, NULL); | |
35c979bf | 2011 | } |
064af421 | 2012 | } |
064af421 BP |
2013 | } |
2014 | ||
2015 | /* For robustness, in case the administrator moves around datapath ports behind | |
2016 | * our back, we re-check all the datapath port numbers here. | |
2017 | * | |
2018 | * This function will set the 'dp_ifidx' members of interfaces that have | |
2019 | * disappeared to -1, so only call this function from a context where those | |
2020 | * 'struct iface's will be removed from the bridge. Otherwise, the -1 | |
2021 | * 'dp_ifidx'es will cause trouble later when we try to send them to the | |
2022 | * datapath, which doesn't support UINT16_MAX+1 ports. */ | |
2023 | static void | |
2024 | bridge_fetch_dp_ifaces(struct bridge *br) | |
2025 | { | |
b0ec0f27 | 2026 | struct dpif_port_dump dump; |
4c738a8d | 2027 | struct dpif_port dpif_port; |
064af421 BP |
2028 | size_t i, j; |
2029 | ||
2030 | /* Reset all interface numbers. */ | |
2031 | for (i = 0; i < br->n_ports; i++) { | |
2032 | struct port *port = br->ports[i]; | |
2033 | for (j = 0; j < port->n_ifaces; j++) { | |
2034 | struct iface *iface = port->ifaces[j]; | |
2035 | iface->dp_ifidx = -1; | |
2036 | } | |
2037 | } | |
d9a8717a | 2038 | hmap_clear(&br->ifaces); |
064af421 | 2039 | |
4c738a8d BP |
2040 | DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { |
2041 | struct iface *iface = iface_lookup(br, dpif_port.name); | |
064af421 BP |
2042 | if (iface) { |
2043 | if (iface->dp_ifidx >= 0) { | |
b29ba128 | 2044 | VLOG_WARN("%s reported interface %s twice", |
4c738a8d BP |
2045 | dpif_name(br->dpif), dpif_port.name); |
2046 | } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) { | |
b29ba128 | 2047 | VLOG_WARN("%s reported interface %"PRIu16" twice", |
4c738a8d | 2048 | dpif_name(br->dpif), dpif_port.port_no); |
064af421 | 2049 | } else { |
4c738a8d | 2050 | iface->dp_ifidx = dpif_port.port_no; |
d9a8717a BP |
2051 | hmap_insert(&br->ifaces, &iface->dp_ifidx_node, |
2052 | hash_int(iface->dp_ifidx, 0)); | |
064af421 | 2053 | } |
093e47f4 | 2054 | |
bcd49a45 BP |
2055 | iface_set_ofport(iface->cfg, |
2056 | (iface->dp_ifidx >= 0 | |
2057 | ? odp_port_to_ofp_port(iface->dp_ifidx) | |
2058 | : -1)); | |
064af421 BP |
2059 | } |
2060 | } | |
064af421 BP |
2061 | } |
2062 | \f | |
2063 | /* Bridge packet processing functions. */ | |
2064 | ||
da285df4 | 2065 | static int |
e58de0e3 | 2066 | bond_hash(const uint8_t mac[ETH_ADDR_LEN], uint16_t vlan) |
da285df4 | 2067 | { |
e58de0e3 | 2068 | return hash_bytes(mac, ETH_ADDR_LEN, vlan) & BOND_MASK; |
da285df4 BP |
2069 | } |
2070 | ||
064af421 | 2071 | static struct bond_entry * |
e58de0e3 EJ |
2072 | lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN], |
2073 | uint16_t vlan) | |
064af421 | 2074 | { |
27dcaa1a | 2075 | assert(port->bond_mode == BM_SLB); |
e58de0e3 | 2076 | return &port->bond_hash[bond_hash(mac, vlan)]; |
064af421 BP |
2077 | } |
2078 | ||
2079 | static int | |
2080 | bond_choose_iface(const struct port *port) | |
2081 | { | |
8b2a2f4a JG |
2082 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); |
2083 | size_t i, best_down_slave = -1; | |
2084 | long long next_delay_expiration = LLONG_MAX; | |
2085 | ||
064af421 | 2086 | for (i = 0; i < port->n_ifaces; i++) { |
8b2a2f4a JG |
2087 | struct iface *iface = port->ifaces[i]; |
2088 | ||
2089 | if (iface->enabled) { | |
064af421 | 2090 | return i; |
8b2a2f4a JG |
2091 | } else if (iface->delay_expires < next_delay_expiration) { |
2092 | best_down_slave = i; | |
2093 | next_delay_expiration = iface->delay_expires; | |
064af421 BP |
2094 | } |
2095 | } | |
8b2a2f4a JG |
2096 | |
2097 | if (best_down_slave != -1) { | |
2098 | struct iface *iface = port->ifaces[best_down_slave]; | |
2099 | ||
2100 | VLOG_INFO_RL(&rl, "interface %s: skipping remaining %lli ms updelay " | |
2101 | "since no other interface is up", iface->name, | |
2102 | iface->delay_expires - time_msec()); | |
2103 | bond_enable_slave(iface, true); | |
2104 | } | |
2105 | ||
2106 | return best_down_slave; | |
064af421 BP |
2107 | } |
2108 | ||
2109 | static bool | |
2303f3b2 | 2110 | choose_output_iface(const struct port *port, const uint8_t *dl_src, |
e58de0e3 | 2111 | uint16_t vlan, uint16_t *dp_ifidx, tag_type *tags) |
064af421 BP |
2112 | { |
2113 | struct iface *iface; | |
2114 | ||
2115 | assert(port->n_ifaces); | |
2116 | if (port->n_ifaces == 1) { | |
2117 | iface = port->ifaces[0]; | |
27dcaa1a | 2118 | } else if (port->bond_mode == BM_AB) { |
be02e7c3 EJ |
2119 | if (port->active_iface < 0) { |
2120 | *tags |= port->no_ifaces_tag; | |
2121 | return false; | |
2122 | } | |
2123 | iface = port->ifaces[port->active_iface]; | |
27dcaa1a | 2124 | } else if (port->bond_mode == BM_SLB){ |
e58de0e3 | 2125 | struct bond_entry *e = lookup_bond_entry(port, dl_src, vlan); |
064af421 BP |
2126 | if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces |
2127 | || !port->ifaces[e->iface_idx]->enabled) { | |
2128 | /* XXX select interface properly. The current interface selection | |
2129 | * is only good for testing the rebalancing code. */ | |
2130 | e->iface_idx = bond_choose_iface(port); | |
2131 | if (e->iface_idx < 0) { | |
2132 | *tags |= port->no_ifaces_tag; | |
2133 | return false; | |
2134 | } | |
2135 | e->iface_tag = tag_create_random(); | |
85c74638 | 2136 | ((struct port *) port)->bond_compat_is_stale = true; |
064af421 BP |
2137 | } |
2138 | *tags |= e->iface_tag; | |
2139 | iface = port->ifaces[e->iface_idx]; | |
be02e7c3 EJ |
2140 | } else { |
2141 | NOT_REACHED(); | |
064af421 BP |
2142 | } |
2143 | *dp_ifidx = iface->dp_ifidx; | |
2144 | *tags |= iface->tag; /* Currently only used for bonding. */ | |
2145 | return true; | |
2146 | } | |
2147 | ||
2148 | static void | |
2149 | bond_link_status_update(struct iface *iface, bool carrier) | |
2150 | { | |
2151 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); | |
2152 | struct port *port = iface->port; | |
2153 | ||
2154 | if ((carrier == iface->enabled) == (iface->delay_expires == LLONG_MAX)) { | |
2155 | /* Nothing to do. */ | |
2156 | return; | |
2157 | } | |
63331829 EJ |
2158 | VLOG_INFO_RL(&rl, "interface %s: link state %s", |
2159 | iface->name, carrier ? "up" : "down"); | |
064af421 BP |
2160 | if (carrier == iface->enabled) { |
2161 | iface->delay_expires = LLONG_MAX; | |
2162 | VLOG_INFO_RL(&rl, "interface %s: will not be %s", | |
2163 | iface->name, carrier ? "disabled" : "enabled"); | |
8b2a2f4a JG |
2164 | } else if (carrier && port->active_iface < 0) { |
2165 | bond_enable_slave(iface, true); | |
2166 | if (port->updelay) { | |
2167 | VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no " | |
2168 | "other interface is up", iface->name, port->updelay); | |
2169 | } | |
064af421 BP |
2170 | } else { |
2171 | int delay = carrier ? port->updelay : port->downdelay; | |
2172 | iface->delay_expires = time_msec() + delay; | |
2173 | if (delay) { | |
2174 | VLOG_INFO_RL(&rl, | |
2175 | "interface %s: will be %s if it stays %s for %d ms", | |
2176 | iface->name, | |
2177 | carrier ? "enabled" : "disabled", | |
2178 | carrier ? "up" : "down", | |
2179 | delay); | |
2180 | } | |
2181 | } | |
2182 | } | |
2183 | ||
2184 | static void | |
2185 | bond_choose_active_iface(struct port *port) | |
2186 | { | |
2187 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); | |
2188 | ||
2189 | port->active_iface = bond_choose_iface(port); | |
2190 | port->active_iface_tag = tag_create_random(); | |
2191 | if (port->active_iface >= 0) { | |
2192 | VLOG_INFO_RL(&rl, "port %s: active interface is now %s", | |
2193 | port->name, port->ifaces[port->active_iface]->name); | |
2194 | } else { | |
2195 | VLOG_WARN_RL(&rl, "port %s: all ports disabled, no active interface", | |
2196 | port->name); | |
2197 | } | |
2198 | } | |
2199 | ||
da285df4 BP |
2200 | static void |
2201 | bond_enable_slave(struct iface *iface, bool enable) | |
2202 | { | |
2203 | struct port *port = iface->port; | |
2204 | struct bridge *br = port->bridge; | |
2205 | ||
8b2a2f4a JG |
2206 | /* This acts as a recursion check. If the act of disabling a slave |
2207 | * causes a different slave to be enabled, the flag will allow us to | |
2208 | * skip redundant work when we reenter this function. It must be | |
2209 | * cleared on exit to keep things safe with multiple bonds. */ | |
2210 | static bool moving_active_iface = false; | |
2211 | ||
da285df4 BP |
2212 | iface->delay_expires = LLONG_MAX; |
2213 | if (enable == iface->enabled) { | |
2214 | return; | |
2215 | } | |
2216 | ||
2217 | iface->enabled = enable; | |
2218 | if (!iface->enabled) { | |
17ea75b2 | 2219 | VLOG_WARN("interface %s: disabled", iface->name); |
da285df4 BP |
2220 | ofproto_revalidate(br->ofproto, iface->tag); |
2221 | if (iface->port_ifidx == port->active_iface) { | |
2222 | ofproto_revalidate(br->ofproto, | |
2223 | port->active_iface_tag); | |
8b2a2f4a JG |
2224 | |
2225 | /* Disabling a slave can lead to another slave being immediately | |
2226 | * enabled if there will be no active slaves but one is waiting | |
2227 | * on an updelay. In this case we do not need to run most of the | |
2228 | * code for the newly enabled slave since there was no period | |
2229 | * without an active slave and it is redundant with the disabling | |
2230 | * path. */ | |
2231 | moving_active_iface = true; | |
da285df4 BP |
2232 | bond_choose_active_iface(port); |
2233 | } | |
2234 | bond_send_learning_packets(port); | |
2235 | } else { | |
17ea75b2 | 2236 | VLOG_WARN("interface %s: enabled", iface->name); |
8b2a2f4a | 2237 | if (port->active_iface < 0 && !moving_active_iface) { |
da285df4 BP |
2238 | ofproto_revalidate(br->ofproto, port->no_ifaces_tag); |
2239 | bond_choose_active_iface(port); | |
2240 | bond_send_learning_packets(port); | |
2241 | } | |
2242 | iface->tag = tag_create_random(); | |
2243 | } | |
8b2a2f4a JG |
2244 | |
2245 | moving_active_iface = false; | |
2246 | port->bond_compat_is_stale = true; | |
da285df4 BP |
2247 | } |
2248 | ||
8722022c BP |
2249 | /* Attempts to make the sum of the bond slaves' statistics appear on the fake |
2250 | * bond interface. */ | |
2251 | static void | |
2252 | bond_update_fake_iface_stats(struct port *port) | |
2253 | { | |
2254 | struct netdev_stats bond_stats; | |
2255 | struct netdev *bond_dev; | |
2256 | size_t i; | |
2257 | ||
2258 | memset(&bond_stats, 0, sizeof bond_stats); | |
2259 | ||
2260 | for (i = 0; i < port->n_ifaces; i++) { | |
2261 | struct netdev_stats slave_stats; | |
2262 | ||
2263 | if (!netdev_get_stats(port->ifaces[i]->netdev, &slave_stats)) { | |
f4b6076a JG |
2264 | /* XXX: We swap the stats here because they are swapped back when |
2265 | * reported by the internal device. The reason for this is | |
2266 | * internal devices normally represent packets going into the system | |
2267 | * but when used as fake bond device they represent packets leaving | |
2268 | * the system. We really should do this in the internal device | |
2269 | * itself because changing it here reverses the counts from the | |
2270 | * perspective of the switch. However, the internal device doesn't | |
2271 | * know what type of device it represents so we have to do it here | |
2272 | * for now. */ | |
2273 | bond_stats.tx_packets += slave_stats.rx_packets; | |
2274 | bond_stats.tx_bytes += slave_stats.rx_bytes; | |
2275 | bond_stats.rx_packets += slave_stats.tx_packets; | |
2276 | bond_stats.rx_bytes += slave_stats.tx_bytes; | |
8722022c BP |
2277 | } |
2278 | } | |
2279 | ||
2280 | if (!netdev_open_default(port->name, &bond_dev)) { | |
2281 | netdev_set_stats(bond_dev, &bond_stats); | |
2282 | netdev_close(bond_dev); | |
2283 | } | |
2284 | } | |
2285 | ||
064af421 BP |
2286 | static void |
2287 | bond_run(struct bridge *br) | |
2288 | { | |
2289 | size_t i, j; | |
2290 | ||
2291 | for (i = 0; i < br->n_ports; i++) { | |
2292 | struct port *port = br->ports[i]; | |
85c74638 | 2293 | |
8b2a2f4a | 2294 | if (port->n_ifaces >= 2) { |
0bb5c3ec BP |
2295 | char *devname; |
2296 | ||
63331829 EJ |
2297 | if (port->monitor) { |
2298 | assert(!port->miimon); | |
0bb5c3ec | 2299 | |
63331829 EJ |
2300 | /* Track carrier going up and down on interfaces. */ |
2301 | while (!netdev_monitor_poll(port->monitor, &devname)) { | |
2302 | struct iface *iface; | |
2303 | ||
2304 | iface = port_lookup_iface(port, devname); | |
2305 | if (iface) { | |
2306 | bool up = netdev_get_carrier(iface->netdev); | |
2307 | ||
2308 | bond_link_status_update(iface, up); | |
2309 | port_update_bond_compat(port); | |
2310 | } | |
2311 | free(devname); | |
2312 | } | |
2313 | } else { | |
2314 | assert(port->miimon); | |
85da620e | 2315 | |
63331829 EJ |
2316 | if (time_msec() >= port->bond_miimon_next_update) { |
2317 | for (j = 0; j < port->n_ifaces; j++) { | |
2318 | struct iface *iface = port->ifaces[j]; | |
2319 | bool up = netdev_get_miimon(iface->netdev); | |
2320 | ||
2321 | bond_link_status_update(iface, up); | |
2322 | port_update_bond_compat(port); | |
2323 | } | |
2324 | port->bond_miimon_next_update = time_msec() + | |
2325 | port->bond_miimon_interval; | |
0bb5c3ec | 2326 | } |
0bb5c3ec BP |
2327 | } |
2328 | ||
8b2a2f4a JG |
2329 | for (j = 0; j < port->n_ifaces; j++) { |
2330 | struct iface *iface = port->ifaces[j]; | |
2331 | if (time_msec() >= iface->delay_expires) { | |
2332 | bond_enable_slave(iface, !iface->enabled); | |
2333 | } | |
2334 | } | |
8722022c BP |
2335 | |
2336 | if (port->bond_fake_iface | |
2337 | && time_msec() >= port->bond_next_fake_iface_update) { | |
2338 | bond_update_fake_iface_stats(port); | |
2339 | port->bond_next_fake_iface_update = time_msec() + 1000; | |
2340 | } | |
8b2a2f4a JG |
2341 | } |
2342 | ||
85c74638 BP |
2343 | if (port->bond_compat_is_stale) { |
2344 | port->bond_compat_is_stale = false; | |
2345 | port_update_bond_compat(port); | |
2346 | } | |
064af421 BP |
2347 | } |
2348 | } | |
2349 | ||
2350 | static void | |
2351 | bond_wait(struct bridge *br) | |
2352 | { | |
2353 | size_t i, j; | |
2354 | ||
2355 | for (i = 0; i < br->n_ports; i++) { | |
2356 | struct port *port = br->ports[i]; | |
2357 | if (port->n_ifaces < 2) { | |
2358 | continue; | |
2359 | } | |
63331829 EJ |
2360 | |
2361 | if (port->monitor) { | |
2362 | netdev_monitor_poll_wait(port->monitor); | |
2363 | } | |
2364 | ||
2365 | if (port->miimon) { | |
2366 | poll_timer_wait_until(port->bond_miimon_next_update); | |
2367 | } | |
2368 | ||
064af421 BP |
2369 | for (j = 0; j < port->n_ifaces; j++) { |
2370 | struct iface *iface = port->ifaces[j]; | |
2371 | if (iface->delay_expires != LLONG_MAX) { | |
7cf8b266 | 2372 | poll_timer_wait_until(iface->delay_expires); |
064af421 BP |
2373 | } |
2374 | } | |
8722022c | 2375 | if (port->bond_fake_iface) { |
7cf8b266 | 2376 | poll_timer_wait_until(port->bond_next_fake_iface_update); |
8722022c | 2377 | } |
064af421 BP |
2378 | } |
2379 | } | |
2380 | ||
2381 | static bool | |
0c62a7ad | 2382 | set_dst(struct dst *dst, const struct flow *flow, |
064af421 BP |
2383 | const struct port *in_port, const struct port *out_port, |
2384 | tag_type *tags) | |
2385 | { | |
0c62a7ad | 2386 | dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE |
064af421 | 2387 | : in_port->vlan >= 0 ? in_port->vlan |
66642cb4 BP |
2388 | : flow->vlan_tci == 0 ? OFP_VLAN_NONE |
2389 | : vlan_tci_to_vid(flow->vlan_tci)); | |
0c62a7ad BP |
2390 | return choose_output_iface(out_port, flow->dl_src, dst->vlan, |
2391 | &dst->dp_ifidx, tags); | |
064af421 BP |
2392 | } |
2393 | ||
2394 | static void | |
2395 | swap_dst(struct dst *p, struct dst *q) | |
2396 | { | |
2397 | struct dst tmp = *p; | |
2398 | *p = *q; | |
2399 | *q = tmp; | |
2400 | } | |
2401 | ||
2402 | /* Moves all the dsts with vlan == 'vlan' to the front of the 'n_dsts' in | |
2403 | * 'dsts'. (This may help performance by reducing the number of VLAN changes | |
2404 | * that we push to the datapath. We could in fact fully sort the array by | |
2405 | * vlan, but in most cases there are at most two different vlan tags so that's | |
2406 | * possibly overkill.) */ | |
2407 | static void | |
0c62a7ad | 2408 | partition_dsts(struct dst_set *set, int vlan) |
064af421 | 2409 | { |
0c62a7ad BP |
2410 | struct dst *first = set->dsts; |
2411 | struct dst *last = set->dsts + set->n; | |
064af421 BP |
2412 | |
2413 | while (first != last) { | |
2414 | /* Invariants: | |
2415 | * - All dsts < first have vlan == 'vlan'. | |
2416 | * - All dsts >= last have vlan != 'vlan'. | |
2417 | * - first < last. */ | |
2418 | while (first->vlan == vlan) { | |
2419 | if (++first == last) { | |
2420 | return; | |
2421 | } | |
2422 | } | |
2423 | ||
2424 | /* Same invariants, plus one additional: | |
2425 | * - first->vlan != vlan. | |
2426 | */ | |
2427 | while (last[-1].vlan != vlan) { | |
2428 | if (--last == first) { | |
2429 | return; | |
2430 | } | |
2431 | } | |
2432 | ||
2433 | /* Same invariants, plus one additional: | |
2434 | * - last[-1].vlan == vlan.*/ | |
2435 | swap_dst(first++, --last); | |
2436 | } | |
2437 | } | |
2438 | ||
2439 | static int | |
2440 | mirror_mask_ffs(mirror_mask_t mask) | |
2441 | { | |
2442 | BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask)); | |
2443 | return ffs(mask); | |
2444 | } | |
2445 | ||
0c62a7ad BP |
2446 | static void |
2447 | dst_set_init(struct dst_set *set) | |
2448 | { | |
2449 | set->dsts = set->builtin; | |
2450 | set->n = 0; | |
2451 | set->allocated = ARRAY_SIZE(set->builtin); | |
2452 | } | |
2453 | ||
2454 | static void | |
2455 | dst_set_add(struct dst_set *set, const struct dst *dst) | |
2456 | { | |
2457 | if (set->n >= set->allocated) { | |
2458 | size_t new_allocated; | |
2459 | struct dst *new_dsts; | |
2460 | ||
2461 | new_allocated = set->allocated * 2; | |
2462 | new_dsts = xmalloc(new_allocated * sizeof *new_dsts); | |
2463 | memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts); | |
2464 | ||
2465 | dst_set_free(set); | |
2466 | ||
2467 | set->dsts = new_dsts; | |
2468 | set->allocated = new_allocated; | |
2469 | } | |
2470 | set->dsts[set->n++] = *dst; | |
2471 | } | |
2472 | ||
2473 | static void | |
2474 | dst_set_free(struct dst_set *set) | |
2475 | { | |
2476 | if (set->dsts != set->builtin) { | |
2477 | free(set->dsts); | |
2478 | } | |
2479 | } | |
2480 | ||
064af421 | 2481 | static bool |
0c62a7ad | 2482 | dst_is_duplicate(const struct dst_set *set, const struct dst *test) |
064af421 BP |
2483 | { |
2484 | size_t i; | |
0c62a7ad BP |
2485 | for (i = 0; i < set->n; i++) { |
2486 | if (set->dsts[i].vlan == test->vlan | |
2487 | && set->dsts[i].dp_ifidx == test->dp_ifidx) { | |
064af421 BP |
2488 | return true; |
2489 | } | |
2490 | } | |
2491 | return false; | |
2492 | } | |
2493 | ||
2494 | static bool | |
2495 | port_trunks_vlan(const struct port *port, uint16_t vlan) | |
2496 | { | |
3e9c481c BP |
2497 | return (port->vlan < 0 |
2498 | && (!port->trunks || bitmap_is_set(port->trunks, vlan))); | |
064af421 BP |
2499 | } |
2500 | ||
2501 | static bool | |
2502 | port_includes_vlan(const struct port *port, uint16_t vlan) | |
2503 | { | |
2504 | return vlan == port->vlan || port_trunks_vlan(port, vlan); | |
2505 | } | |
2506 | ||
a4e2e1f2 EJ |
2507 | static bool |
2508 | port_is_floodable(const struct port *port) | |
2509 | { | |
2510 | int i; | |
2511 | ||
2512 | for (i = 0; i < port->n_ifaces; i++) { | |
2513 | if (!ofproto_port_is_floodable(port->bridge->ofproto, | |
2514 | port->ifaces[i]->dp_ifidx)) { | |
2515 | return false; | |
2516 | } | |
2517 | } | |
2518 | return true; | |
2519 | } | |
2520 | ||
0c62a7ad | 2521 | static void |
ae412e7d | 2522 | compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, |
064af421 | 2523 | const struct port *in_port, const struct port *out_port, |
0c62a7ad | 2524 | struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface) |
064af421 BP |
2525 | { |
2526 | mirror_mask_t mirrors = in_port->src_mirrors; | |
0c62a7ad | 2527 | struct dst dst; |
66642cb4 | 2528 | int flow_vlan; |
064af421 BP |
2529 | size_t i; |
2530 | ||
66642cb4 BP |
2531 | flow_vlan = vlan_tci_to_vid(flow->vlan_tci); |
2532 | if (flow_vlan == 0) { | |
2533 | flow_vlan = OFP_VLAN_NONE; | |
2534 | } | |
2535 | ||
064af421 | 2536 | if (out_port == FLOOD_PORT) { |
064af421 BP |
2537 | for (i = 0; i < br->n_ports; i++) { |
2538 | struct port *port = br->ports[i]; | |
a4e2e1f2 EJ |
2539 | if (port != in_port |
2540 | && port_is_floodable(port) | |
2541 | && port_includes_vlan(port, vlan) | |
064af421 | 2542 | && !port->is_mirror_output_port |
0c62a7ad | 2543 | && set_dst(&dst, flow, in_port, port, tags)) { |
064af421 | 2544 | mirrors |= port->dst_mirrors; |
0c62a7ad | 2545 | dst_set_add(set, &dst); |
064af421 BP |
2546 | } |
2547 | } | |
6a07af36 | 2548 | *nf_output_iface = NF_OUT_FLOOD; |
0c62a7ad BP |
2549 | } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) { |
2550 | dst_set_add(set, &dst); | |
2551 | *nf_output_iface = dst.dp_ifidx; | |
064af421 | 2552 | mirrors |= out_port->dst_mirrors; |
064af421 BP |
2553 | } |
2554 | ||
2555 | while (mirrors) { | |
2556 | struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1]; | |
2557 | if (!m->n_vlans || vlan_is_mirrored(m, vlan)) { | |
2558 | if (m->out_port) { | |
0c62a7ad BP |
2559 | if (set_dst(&dst, flow, in_port, m->out_port, tags) |
2560 | && !dst_is_duplicate(set, &dst)) { | |
2561 | dst_set_add(set, &dst); | |
064af421 BP |
2562 | } |
2563 | } else { | |
2564 | for (i = 0; i < br->n_ports; i++) { | |
2565 | struct port *port = br->ports[i]; | |
2566 | if (port_includes_vlan(port, m->out_vlan) | |
0c62a7ad | 2567 | && set_dst(&dst, flow, in_port, port, tags)) |
064af421 BP |
2568 | { |
2569 | if (port->vlan < 0) { | |
0c62a7ad | 2570 | dst.vlan = m->out_vlan; |
064af421 | 2571 | } |
0c62a7ad | 2572 | if (dst_is_duplicate(set, &dst)) { |
274de4d2 BP |
2573 | continue; |
2574 | } | |
43aa5f47 JG |
2575 | |
2576 | /* Use the vlan tag on the original flow instead of | |
2577 | * the one passed in the vlan parameter. This ensures | |
2578 | * that we compare the vlan from before any implicit | |
2579 | * tagging tags place. This is necessary because | |
2580 | * dst->vlan is the final vlan, after removing implicit | |
2581 | * tags. */ | |
0c62a7ad | 2582 | if (port == in_port && dst.vlan == flow_vlan) { |
064af421 BP |
2583 | /* Don't send out input port on same VLAN. */ |
2584 | continue; | |
2585 | } | |
0c62a7ad | 2586 | dst_set_add(set, &dst); |
064af421 BP |
2587 | } |
2588 | } | |
2589 | } | |
2590 | } | |
2591 | mirrors &= mirrors - 1; | |
2592 | } | |
2593 | ||
0c62a7ad | 2594 | partition_dsts(set, flow_vlan); |
064af421 BP |
2595 | } |
2596 | ||
67a4917b | 2597 | static void OVS_UNUSED |
0c62a7ad | 2598 | print_dsts(const struct dst_set *set) |
064af421 | 2599 | { |
0c62a7ad BP |
2600 | size_t i; |
2601 | ||
2602 | for (i = 0; i < set->n; i++) { | |
2603 | const struct dst *dst = &set->dsts[i]; | |
2604 | ||
2605 | printf(">p%"PRIu16, dst->dp_ifidx); | |
2606 | if (dst->vlan != OFP_VLAN_NONE) { | |
2607 | printf("v%"PRIu16, dst->vlan); | |
064af421 BP |
2608 | } |
2609 | } | |
2610 | } | |
2611 | ||
2612 | static void | |
ae412e7d | 2613 | compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan, |
064af421 | 2614 | const struct port *in_port, const struct port *out_port, |
cdee00fd | 2615 | tag_type *tags, struct ofpbuf *actions, |
6a07af36 | 2616 | uint16_t *nf_output_iface) |
064af421 | 2617 | { |
0c62a7ad | 2618 | struct dst_set set; |
064af421 | 2619 | uint16_t cur_vlan; |
0c62a7ad | 2620 | size_t i; |
064af421 | 2621 | |
0c62a7ad BP |
2622 | dst_set_init(&set); |
2623 | compose_dsts(br, flow, vlan, in_port, out_port, &set, tags, | |
2624 | nf_output_iface); | |
064af421 | 2625 | |
66642cb4 BP |
2626 | cur_vlan = vlan_tci_to_vid(flow->vlan_tci); |
2627 | if (cur_vlan == 0) { | |
2628 | cur_vlan = OFP_VLAN_NONE; | |
2629 | } | |
0c62a7ad BP |
2630 | for (i = 0; i < set.n; i++) { |
2631 | const struct dst *dst = &set.dsts[i]; | |
2632 | if (dst->vlan != cur_vlan) { | |
2633 | if (dst->vlan == OFP_VLAN_NONE) { | |
cdee00fd | 2634 | nl_msg_put_flag(actions, ODPAT_STRIP_VLAN); |
064af421 | 2635 | } else { |
cdee00fd | 2636 | ovs_be16 tci; |
0c62a7ad | 2637 | tci = htons(dst->vlan & VLAN_VID_MASK); |
cdee00fd BP |
2638 | tci |= flow->vlan_tci & htons(VLAN_PCP_MASK); |
2639 | nl_msg_put_be16(actions, ODPAT_SET_DL_TCI, tci); | |
064af421 | 2640 | } |
0c62a7ad | 2641 | cur_vlan = dst->vlan; |
064af421 | 2642 | } |
0c62a7ad | 2643 | nl_msg_put_u32(actions, ODPAT_OUTPUT, dst->dp_ifidx); |
064af421 | 2644 | } |
0c62a7ad | 2645 | dst_set_free(&set); |
064af421 BP |
2646 | } |
2647 | ||
e96a4d80 JG |
2648 | /* Returns the effective vlan of a packet, taking into account both the |
2649 | * 802.1Q header and implicitly tagged ports. A value of 0 indicates that | |
2650 | * the packet is untagged and -1 indicates it has an invalid header and | |
2651 | * should be dropped. */ | |
ae412e7d | 2652 | static int flow_get_vlan(struct bridge *br, const struct flow *flow, |
e96a4d80 JG |
2653 | struct port *in_port, bool have_packet) |
2654 | { | |
66642cb4 | 2655 | int vlan = vlan_tci_to_vid(flow->vlan_tci); |
e96a4d80 JG |
2656 | if (in_port->vlan >= 0) { |
2657 | if (vlan) { | |
2658 | /* XXX support double tagging? */ | |
2659 | if (have_packet) { | |
2660 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
66642cb4 | 2661 | VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged " |
e96a4d80 JG |
2662 | "packet received on port %s configured with " |
2663 | "implicit VLAN %"PRIu16, | |
66642cb4 | 2664 | br->name, vlan, in_port->name, in_port->vlan); |
e96a4d80 JG |
2665 | } |
2666 | return -1; | |
2667 | } | |
2668 | vlan = in_port->vlan; | |
2669 | } else { | |
2670 | if (!port_includes_vlan(in_port, vlan)) { | |
2671 | if (have_packet) { | |
2672 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2673 | VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged " | |
2674 | "packet received on port %s not configured for " | |
2675 | "trunking VLAN %d", | |
2676 | br->name, vlan, in_port->name, vlan); | |
2677 | } | |
2678 | return -1; | |
2679 | } | |
2680 | } | |
2681 | ||
2682 | return vlan; | |
2683 | } | |
2684 | ||
7febb910 JG |
2685 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after |
2686 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
2687 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
2688 | static bool | |
ae412e7d | 2689 | is_gratuitous_arp(const struct flow *flow) |
7febb910 JG |
2690 | { |
2691 | return (flow->dl_type == htons(ETH_TYPE_ARP) | |
2692 | && eth_addr_is_broadcast(flow->dl_dst) | |
2693 | && (flow->nw_proto == ARP_OP_REPLY | |
2694 | || (flow->nw_proto == ARP_OP_REQUEST | |
2695 | && flow->nw_src == flow->nw_dst))); | |
2696 | } | |
2697 | ||
e96a4d80 | 2698 | static void |
ae412e7d | 2699 | update_learning_table(struct bridge *br, const struct flow *flow, int vlan, |
e96a4d80 JG |
2700 | struct port *in_port) |
2701 | { | |
7febb910 JG |
2702 | enum grat_arp_lock_type lock_type; |
2703 | tag_type rev_tag; | |
2704 | ||
2705 | /* We don't want to learn from gratuitous ARP packets that are reflected | |
2706 | * back over bond slaves so we lock the learning table. */ | |
2707 | lock_type = !is_gratuitous_arp(flow) ? GRAT_ARP_LOCK_NONE : | |
2708 | (in_port->n_ifaces == 1) ? GRAT_ARP_LOCK_SET : | |
2709 | GRAT_ARP_LOCK_CHECK; | |
2710 | ||
2711 | rev_tag = mac_learning_learn(br->ml, flow->dl_src, vlan, in_port->port_idx, | |
2712 | lock_type); | |
e96a4d80 JG |
2713 | if (rev_tag) { |
2714 | /* The log messages here could actually be useful in debugging, | |
2715 | * so keep the rate limit relatively high. */ | |
2716 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, | |
2717 | 300); | |
2718 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " | |
2719 | "on port %s in VLAN %d", | |
2720 | br->name, ETH_ADDR_ARGS(flow->dl_src), | |
2721 | in_port->name, vlan); | |
2722 | ofproto_revalidate(br->ofproto, rev_tag); | |
2723 | } | |
2724 | } | |
2725 | ||
14a34d00 BP |
2726 | /* Determines whether packets in 'flow' within 'br' should be forwarded or |
2727 | * dropped. Returns true if they may be forwarded, false if they should be | |
2728 | * dropped. | |
2729 | * | |
2730 | * If 'have_packet' is true, it indicates that the caller is processing a | |
2731 | * received packet. If 'have_packet' is false, then the caller is just | |
2732 | * revalidating an existing flow because configuration has changed. Either | |
2733 | * way, 'have_packet' only affects logging (there is no point in logging errors | |
2734 | * during revalidation). | |
2735 | * | |
2736 | * Sets '*in_portp' to the input port. This will be a null pointer if | |
2737 | * flow->in_port does not designate a known input port (in which case | |
2738 | * is_admissible() returns false). | |
2739 | * | |
2740 | * When returning true, sets '*vlanp' to the effective VLAN of the input | |
2741 | * packet, as returned by flow_get_vlan(). | |
2742 | * | |
2743 | * May also add tags to '*tags', although the current implementation only does | |
2744 | * so in one special case. | |
2745 | */ | |
064af421 | 2746 | static bool |
ae412e7d | 2747 | is_admissible(struct bridge *br, const struct flow *flow, bool have_packet, |
14a34d00 | 2748 | tag_type *tags, int *vlanp, struct port **in_portp) |
064af421 BP |
2749 | { |
2750 | struct iface *in_iface; | |
2751 | struct port *in_port; | |
064af421 BP |
2752 | int vlan; |
2753 | ||
2754 | /* Find the interface and port structure for the received packet. */ | |
2755 | in_iface = iface_from_dp_ifidx(br, flow->in_port); | |
2756 | if (!in_iface) { | |
2757 | /* No interface? Something fishy... */ | |
14a34d00 | 2758 | if (have_packet) { |
064af421 BP |
2759 | /* Odd. A few possible reasons here: |
2760 | * | |
2761 | * - We deleted an interface but there are still a few packets | |
2762 | * queued up from it. | |
2763 | * | |
2764 | * - Someone externally added an interface (e.g. with "ovs-dpctl | |
2765 | * add-if") that we don't know about. | |
2766 | * | |
2767 | * - Packet arrived on the local port but the local port is not | |
2768 | * one of our bridge ports. | |
2769 | */ | |
2770 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2771 | ||
2772 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
d295e8e9 | 2773 | "interface %"PRIu16, br->name, flow->in_port); |
064af421 BP |
2774 | } |
2775 | ||
14a34d00 BP |
2776 | *in_portp = NULL; |
2777 | return false; | |
064af421 | 2778 | } |
14a34d00 BP |
2779 | *in_portp = in_port = in_iface->port; |
2780 | *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet); | |
e96a4d80 | 2781 | if (vlan < 0) { |
14a34d00 | 2782 | return false; |
064af421 BP |
2783 | } |
2784 | ||
064af421 BP |
2785 | /* Drop frames for reserved multicast addresses. */ |
2786 | if (eth_addr_is_reserved(flow->dl_dst)) { | |
14a34d00 | 2787 | return false; |
064af421 BP |
2788 | } |
2789 | ||
2790 | /* Drop frames on ports reserved for mirroring. */ | |
2791 | if (in_port->is_mirror_output_port) { | |
14a34d00 | 2792 | if (have_packet) { |
f925807d BP |
2793 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); |
2794 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2795 | "%s, which is reserved exclusively for mirroring", | |
2796 | br->name, in_port->name); | |
2797 | } | |
14a34d00 | 2798 | return false; |
064af421 BP |
2799 | } |
2800 | ||
3a55ef14 JG |
2801 | /* Packets received on bonds need special attention to avoid duplicates. */ |
2802 | if (in_port->n_ifaces > 1) { | |
2803 | int src_idx; | |
7febb910 | 2804 | bool is_grat_arp_locked; |
3a55ef14 JG |
2805 | |
2806 | if (eth_addr_is_multicast(flow->dl_dst)) { | |
2807 | *tags |= in_port->active_iface_tag; | |
2808 | if (in_port->active_iface != in_iface->port_ifidx) { | |
2809 | /* Drop all multicast packets on inactive slaves. */ | |
14a34d00 | 2810 | return false; |
c93b1d6a | 2811 | } |
064af421 | 2812 | } |
3a55ef14 JG |
2813 | |
2814 | /* Drop all packets for which we have learned a different input | |
2815 | * port, because we probably sent the packet on one slave and got | |
5d0ae138 | 2816 | * it back on the other. Gratuitous ARP packets are an exception |
7febb910 JG |
2817 | * to this rule: the host has moved to another switch. The exception |
2818 | * to the exception is if we locked the learning table to avoid | |
2819 | * reflections on bond slaves. If this is the case, just drop the | |
2820 | * packet now. */ | |
2821 | src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan, | |
2822 | &is_grat_arp_locked); | |
3a55ef14 | 2823 | if (src_idx != -1 && src_idx != in_port->port_idx && |
7febb910 | 2824 | (!is_gratuitous_arp(flow) || is_grat_arp_locked)) { |
14a34d00 | 2825 | return false; |
3a55ef14 | 2826 | } |
064af421 BP |
2827 | } |
2828 | ||
14a34d00 BP |
2829 | return true; |
2830 | } | |
2831 | ||
2832 | /* If the composed actions may be applied to any packet in the given 'flow', | |
2833 | * returns true. Otherwise, the actions should only be applied to 'packet', or | |
2834 | * not at all, if 'packet' was NULL. */ | |
2835 | static bool | |
ae412e7d | 2836 | process_flow(struct bridge *br, const struct flow *flow, |
cdee00fd | 2837 | const struct ofpbuf *packet, struct ofpbuf *actions, |
14a34d00 BP |
2838 | tag_type *tags, uint16_t *nf_output_iface) |
2839 | { | |
2840 | struct port *in_port; | |
2841 | struct port *out_port; | |
2842 | int vlan; | |
2843 | int out_port_idx; | |
2844 | ||
2845 | /* Check whether we should drop packets in this flow. */ | |
2846 | if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) { | |
2847 | out_port = NULL; | |
2848 | goto done; | |
2849 | } | |
2850 | ||
93dfc067 JG |
2851 | /* Learn source MAC (but don't try to learn from revalidation). */ |
2852 | if (packet) { | |
e96a4d80 | 2853 | update_learning_table(br, flow, vlan, in_port); |
93dfc067 JG |
2854 | } |
2855 | ||
2856 | /* Determine output port. */ | |
7febb910 JG |
2857 | out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan, tags, |
2858 | NULL); | |
93dfc067 JG |
2859 | if (out_port_idx >= 0 && out_port_idx < br->n_ports) { |
2860 | out_port = br->ports[out_port_idx]; | |
e96a4d80 | 2861 | } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) { |
93dfc067 | 2862 | /* If we are revalidating but don't have a learning entry then |
e96a4d80 JG |
2863 | * eject the flow. Installing a flow that floods packets opens |
2864 | * up a window of time where we could learn from a packet reflected | |
2865 | * on a bond and blackhole packets before the learning table is | |
2866 | * updated to reflect the correct port. */ | |
93dfc067 | 2867 | return false; |
1609aa03 BP |
2868 | } else { |
2869 | out_port = FLOOD_PORT; | |
064af421 BP |
2870 | } |
2871 | ||
ba54bf4f BP |
2872 | /* Don't send packets out their input ports. */ |
2873 | if (in_port == out_port) { | |
064af421 BP |
2874 | out_port = NULL; |
2875 | } | |
2876 | ||
2877 | done: | |
14a34d00 BP |
2878 | if (in_port) { |
2879 | compose_actions(br, flow, vlan, in_port, out_port, tags, actions, | |
2880 | nf_output_iface); | |
2881 | } | |
064af421 | 2882 | |
69d60f9f | 2883 | return true; |
064af421 BP |
2884 | } |
2885 | ||
064af421 | 2886 | static bool |
ae412e7d | 2887 | bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, |
cdee00fd | 2888 | struct ofpbuf *actions, tag_type *tags, |
6a07af36 | 2889 | uint16_t *nf_output_iface, void *br_) |
064af421 | 2890 | { |
b31bcf60 | 2891 | struct iface *iface; |
064af421 BP |
2892 | struct bridge *br = br_; |
2893 | ||
064af421 | 2894 | COVERAGE_INC(bridge_process_flow); |
26d79bf2 | 2895 | |
b31bcf60 EJ |
2896 | iface = iface_from_dp_ifidx(br, flow->in_port); |
2897 | ||
2898 | if (cfm_should_process_flow(flow)) { | |
2899 | if (packet && iface->cfm) { | |
2900 | cfm_process_heartbeat(iface->cfm, packet); | |
2901 | } | |
2902 | return false; | |
2903 | } | |
2904 | ||
6a07af36 | 2905 | return process_flow(br, flow, packet, actions, tags, nf_output_iface); |
064af421 BP |
2906 | } |
2907 | ||
2908 | static void | |
ae412e7d | 2909 | bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, |
cdee00fd | 2910 | const struct nlattr *actions, |
cf22f8cb | 2911 | size_t actions_len, |
cdee00fd | 2912 | unsigned long long int n_bytes, void *br_) |
064af421 BP |
2913 | { |
2914 | struct bridge *br = br_; | |
cdee00fd | 2915 | const struct nlattr *a; |
db0e2ad1 | 2916 | struct port *in_port; |
26d79bf2 | 2917 | tag_type dummy = 0; |
cdee00fd | 2918 | unsigned int left; |
db0e2ad1 | 2919 | int vlan; |
064af421 | 2920 | |
26d79bf2 BP |
2921 | /* Feed information from the active flows back into the learning table to |
2922 | * ensure that table is always in sync with what is actually flowing | |
2923 | * through the datapath. | |
2924 | * | |
2925 | * We test that 'tags' is nonzero to ensure that only flows that include an | |
2926 | * OFPP_NORMAL action are used for learning. This works because | |
2927 | * bridge_normal_ofhook_cb() always sets a nonzero tag value. */ | |
2928 | if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) { | |
db0e2ad1 | 2929 | update_learning_table(br, flow, vlan, in_port); |
e96a4d80 JG |
2930 | } |
2931 | ||
26d79bf2 | 2932 | /* Account for bond slave utilization. */ |
064af421 BP |
2933 | if (!br->has_bonded_ports) { |
2934 | return; | |
2935 | } | |
cdee00fd BP |
2936 | NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) { |
2937 | if (nl_attr_type(a) == ODPAT_OUTPUT) { | |
2938 | struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a)); | |
be02e7c3 | 2939 | if (out_port && out_port->n_ifaces >= 2 && |
27dcaa1a | 2940 | out_port->bond_mode == BM_SLB) { |
e58de0e3 EJ |
2941 | uint16_t vlan = (flow->vlan_tci |
2942 | ? vlan_tci_to_vid(flow->vlan_tci) | |
2943 | : OFP_VLAN_NONE); | |
e96a4d80 | 2944 | struct bond_entry *e = lookup_bond_entry(out_port, |
e58de0e3 | 2945 | flow->dl_src, vlan); |
064af421 BP |
2946 | e->tx_bytes += n_bytes; |
2947 | } | |
2948 | } | |
2949 | } | |
2950 | } | |
2951 | ||
2952 | static void | |
2953 | bridge_account_checkpoint_ofhook_cb(void *br_) | |
2954 | { | |
2955 | struct bridge *br = br_; | |
c8143c88 | 2956 | long long int now; |
064af421 BP |
2957 | size_t i; |
2958 | ||
2959 | if (!br->has_bonded_ports) { | |
2960 | return; | |
2961 | } | |
2962 | ||
c8143c88 | 2963 | now = time_msec(); |
064af421 BP |
2964 | for (i = 0; i < br->n_ports; i++) { |
2965 | struct port *port = br->ports[i]; | |
27dcaa1a | 2966 | if (port->n_ifaces > 1 && port->bond_mode == BM_SLB |
be02e7c3 | 2967 | && now >= port->bond_next_rebalance) { |
c8143c88 | 2968 | port->bond_next_rebalance = now + port->bond_rebalance_interval; |
064af421 BP |
2969 | bond_rebalance_port(port); |
2970 | } | |
2971 | } | |
2972 | } | |
2973 | ||
2974 | static struct ofhooks bridge_ofhooks = { | |
064af421 BP |
2975 | bridge_normal_ofhook_cb, |
2976 | bridge_account_flow_ofhook_cb, | |
2977 | bridge_account_checkpoint_ofhook_cb, | |
2978 | }; | |
2979 | \f | |
2303f3b2 BP |
2980 | /* Bonding functions. */ |
2981 | ||
064af421 BP |
2982 | /* Statistics for a single interface on a bonded port, used for load-based |
2983 | * bond rebalancing. */ | |
2984 | struct slave_balance { | |
2985 | struct iface *iface; /* The interface. */ | |
2986 | uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */ | |
2987 | ||
2988 | /* All the "bond_entry"s that are assigned to this interface, in order of | |
2989 | * increasing tx_bytes. */ | |
2990 | struct bond_entry **hashes; | |
2991 | size_t n_hashes; | |
2992 | }; | |
2993 | ||
be02e7c3 | 2994 | static const char * |
27dcaa1a EJ |
2995 | bond_mode_to_string(enum bond_mode bm) { |
2996 | static char *bm_slb = "balance-slb"; | |
2997 | static char *bm_ab = "active-backup"; | |
be02e7c3 | 2998 | |
27dcaa1a EJ |
2999 | switch (bm) { |
3000 | case BM_SLB: return bm_slb; | |
3001 | case BM_AB: return bm_ab; | |
be02e7c3 EJ |
3002 | } |
3003 | ||
3004 | NOT_REACHED(); | |
3005 | return NULL; | |
3006 | } | |
3007 | ||
064af421 BP |
3008 | /* Sorts pointers to pointers to bond_entries in ascending order by the |
3009 | * interface to which they are assigned, and within a single interface in | |
3010 | * ascending order of bytes transmitted. */ | |
3011 | static int | |
3012 | compare_bond_entries(const void *a_, const void *b_) | |
3013 | { | |
3014 | const struct bond_entry *const *ap = a_; | |
3015 | const struct bond_entry *const *bp = b_; | |
3016 | const struct bond_entry *a = *ap; | |
3017 | const struct bond_entry *b = *bp; | |
3018 | if (a->iface_idx != b->iface_idx) { | |
3019 | return a->iface_idx > b->iface_idx ? 1 : -1; | |
3020 | } else if (a->tx_bytes != b->tx_bytes) { | |
3021 | return a->tx_bytes > b->tx_bytes ? 1 : -1; | |
3022 | } else { | |
3023 | return 0; | |
3024 | } | |
3025 | } | |
3026 | ||
3027 | /* Sorts slave_balances so that enabled ports come first, and otherwise in | |
3028 | * *descending* order by number of bytes transmitted. */ | |
3029 | static int | |
3030 | compare_slave_balance(const void *a_, const void *b_) | |
3031 | { | |
3032 | const struct slave_balance *a = a_; | |
3033 | const struct slave_balance *b = b_; | |
3034 | if (a->iface->enabled != b->iface->enabled) { | |
3035 | return a->iface->enabled ? -1 : 1; | |
3036 | } else if (a->tx_bytes != b->tx_bytes) { | |
3037 | return a->tx_bytes > b->tx_bytes ? -1 : 1; | |
3038 | } else { | |
3039 | return 0; | |
3040 | } | |
3041 | } | |
3042 | ||
3043 | static void | |
3044 | swap_bals(struct slave_balance *a, struct slave_balance *b) | |
3045 | { | |
3046 | struct slave_balance tmp = *a; | |
3047 | *a = *b; | |
3048 | *b = tmp; | |
3049 | } | |
3050 | ||
3051 | /* Restores the 'n_bals' slave_balance structures in 'bals' to sorted order | |
3052 | * given that 'p' (and only 'p') might be in the wrong location. | |
3053 | * | |
3054 | * This function invalidates 'p', since it might now be in a different memory | |
3055 | * location. */ | |
3056 | static void | |
3057 | resort_bals(struct slave_balance *p, | |
3058 | struct slave_balance bals[], size_t n_bals) | |
3059 | { | |
3060 | if (n_bals > 1) { | |
3061 | for (; p > bals && p->tx_bytes > p[-1].tx_bytes; p--) { | |
3062 | swap_bals(p, p - 1); | |
3063 | } | |
3064 | for (; p < &bals[n_bals - 1] && p->tx_bytes < p[1].tx_bytes; p++) { | |
3065 | swap_bals(p, p + 1); | |
3066 | } | |
3067 | } | |
3068 | } | |
3069 | ||
3070 | static void | |
3071 | log_bals(const struct slave_balance *bals, size_t n_bals, struct port *port) | |
3072 | { | |
3073 | if (VLOG_IS_DBG_ENABLED()) { | |
3074 | struct ds ds = DS_EMPTY_INITIALIZER; | |
3075 | const struct slave_balance *b; | |
3076 | ||
3077 | for (b = bals; b < bals + n_bals; b++) { | |
3078 | size_t i; | |
3079 | ||
3080 | if (b > bals) { | |
3081 | ds_put_char(&ds, ','); | |
3082 | } | |
3083 | ds_put_format(&ds, " %s %"PRIu64"kB", | |
3084 | b->iface->name, b->tx_bytes / 1024); | |
3085 | ||
3086 | if (!b->iface->enabled) { | |
3087 | ds_put_cstr(&ds, " (disabled)"); | |
3088 | } | |
3089 | if (b->n_hashes > 0) { | |
3090 | ds_put_cstr(&ds, " ("); | |
3091 | for (i = 0; i < b->n_hashes; i++) { | |
3092 | const struct bond_entry *e = b->hashes[i]; | |
3093 | if (i > 0) { | |
3094 | ds_put_cstr(&ds, " + "); | |
3095 | } | |
3096 | ds_put_format(&ds, "h%td: %"PRIu64"kB", | |
3097 | e - port->bond_hash, e->tx_bytes / 1024); | |
3098 | } | |
3099 | ds_put_cstr(&ds, ")"); | |
3100 | } | |
3101 | } | |
3102 | VLOG_DBG("bond %s:%s", port->name, ds_cstr(&ds)); | |
3103 | ds_destroy(&ds); | |
3104 | } | |
3105 | } | |
3106 | ||
3107 | /* Shifts 'hash' from 'from' to 'to' within 'port'. */ | |
3108 | static void | |
3109 | bond_shift_load(struct slave_balance *from, struct slave_balance *to, | |
5422a9e1 | 3110 | int hash_idx) |
064af421 | 3111 | { |
5422a9e1 | 3112 | struct bond_entry *hash = from->hashes[hash_idx]; |
064af421 BP |
3113 | struct port *port = from->iface->port; |
3114 | uint64_t delta = hash->tx_bytes; | |
3115 | ||
27dcaa1a | 3116 | assert(port->bond_mode == BM_SLB); |
be02e7c3 | 3117 | |
064af421 BP |
3118 | VLOG_INFO("bond %s: shift %"PRIu64"kB of load (with hash %td) " |
3119 | "from %s to %s (now carrying %"PRIu64"kB and " | |
3120 | "%"PRIu64"kB load, respectively)", | |
3121 | port->name, delta / 1024, hash - port->bond_hash, | |
3122 | from->iface->name, to->iface->name, | |
3123 | (from->tx_bytes - delta) / 1024, | |
3124 | (to->tx_bytes + delta) / 1024); | |
3125 | ||
3126 | /* Delete element from from->hashes. | |
3127 | * | |
3128 | * We don't bother to add the element to to->hashes because not only would | |
3129 | * it require more work, the only purpose it would be to allow that hash to | |
3130 | * be migrated to another slave in this rebalancing run, and there is no | |
3131 | * point in doing that. */ | |
5422a9e1 | 3132 | if (hash_idx == 0) { |
064af421 BP |
3133 | from->hashes++; |
3134 | } else { | |
5422a9e1 JG |
3135 | memmove(from->hashes + hash_idx, from->hashes + hash_idx + 1, |
3136 | (from->n_hashes - (hash_idx + 1)) * sizeof *from->hashes); | |
064af421 BP |
3137 | } |
3138 | from->n_hashes--; | |
3139 | ||
3140 | /* Shift load away from 'from' to 'to'. */ | |
3141 | from->tx_bytes -= delta; | |
3142 | to->tx_bytes += delta; | |
3143 | ||
3144 | /* Arrange for flows to be revalidated. */ | |
3145 | ofproto_revalidate(port->bridge->ofproto, hash->iface_tag); | |
3146 | hash->iface_idx = to->iface->port_ifidx; | |
3147 | hash->iface_tag = tag_create_random(); | |
064af421 BP |
3148 | } |
3149 | ||
3150 | static void | |
3151 | bond_rebalance_port(struct port *port) | |
3152 | { | |
c47cd5f0 | 3153 | struct slave_balance *bals; |
064af421 BP |
3154 | size_t n_bals; |
3155 | struct bond_entry *hashes[BOND_MASK + 1]; | |
3156 | struct slave_balance *b, *from, *to; | |
3157 | struct bond_entry *e; | |
3158 | size_t i; | |
3159 | ||
27dcaa1a | 3160 | assert(port->bond_mode == BM_SLB); |
be02e7c3 | 3161 | |
064af421 BP |
3162 | /* Sets up 'bals' to describe each of the port's interfaces, sorted in |
3163 | * descending order of tx_bytes, so that bals[0] represents the most | |
3164 | * heavily loaded slave and bals[n_bals - 1] represents the least heavily | |
3165 | * loaded slave. | |
3166 | * | |
3167 | * The code is a bit tricky: to avoid dynamically allocating a 'hashes' | |
3168 | * array for each slave_balance structure, we sort our local array of | |
3169 | * hashes in order by slave, so that all of the hashes for a given slave | |
3170 | * become contiguous in memory, and then we point each 'hashes' members of | |
3171 | * a slave_balance structure to the start of a contiguous group. */ | |
3172 | n_bals = port->n_ifaces; | |
c47cd5f0 | 3173 | bals = xmalloc(n_bals * sizeof *bals); |
064af421 BP |
3174 | for (b = bals; b < &bals[n_bals]; b++) { |
3175 | b->iface = port->ifaces[b - bals]; | |
3176 | b->tx_bytes = 0; | |
3177 | b->hashes = NULL; | |
3178 | b->n_hashes = 0; | |
3179 | } | |
3180 | for (i = 0; i <= BOND_MASK; i++) { | |
3181 | hashes[i] = &port->bond_hash[i]; | |
3182 | } | |
3183 | qsort(hashes, BOND_MASK + 1, sizeof *hashes, compare_bond_entries); | |
3184 | for (i = 0; i <= BOND_MASK; i++) { | |
3185 | e = hashes[i]; | |
3186 | if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) { | |
3187 | b = &bals[e->iface_idx]; | |
3188 | b->tx_bytes += e->tx_bytes; | |
3189 | if (!b->hashes) { | |
3190 | b->hashes = &hashes[i]; | |
3191 | } | |
3192 | b->n_hashes++; | |
3193 | } | |
3194 | } | |
3195 | qsort(bals, n_bals, sizeof *bals, compare_slave_balance); | |
3196 | log_bals(bals, n_bals, port); | |
3197 | ||
3198 | /* Discard slaves that aren't enabled (which were sorted to the back of the | |
3199 | * array earlier). */ | |
3200 | while (!bals[n_bals - 1].iface->enabled) { | |
3201 | n_bals--; | |
3202 | if (!n_bals) { | |
c47cd5f0 | 3203 | goto exit; |
064af421 BP |
3204 | } |
3205 | } | |
3206 | ||
3207 | /* Shift load from the most-loaded slaves to the least-loaded slaves. */ | |
3208 | to = &bals[n_bals - 1]; | |
3209 | for (from = bals; from < to; ) { | |
3210 | uint64_t overload = from->tx_bytes - to->tx_bytes; | |
3211 | if (overload < to->tx_bytes >> 5 || overload < 100000) { | |
3212 | /* The extra load on 'from' (and all less-loaded slaves), compared | |
3213 | * to that of 'to' (the least-loaded slave), is less than ~3%, or | |
3214 | * it is less than ~1Mbps. No point in rebalancing. */ | |
3215 | break; | |
3216 | } else if (from->n_hashes == 1) { | |
3217 | /* 'from' only carries a single MAC hash, so we can't shift any | |
3218 | * load away from it, even though we want to. */ | |
3219 | from++; | |
3220 | } else { | |
3221 | /* 'from' is carrying significantly more load than 'to', and that | |
3222 | * load is split across at least two different hashes. Pick a hash | |
3223 | * to migrate to 'to' (the least-loaded slave), given that doing so | |
5422a9e1 JG |
3224 | * must decrease the ratio of the load on the two slaves by at |
3225 | * least 0.1. | |
064af421 BP |
3226 | * |
3227 | * The sort order we use means that we prefer to shift away the | |
3228 | * smallest hashes instead of the biggest ones. There is little | |
3229 | * reason behind this decision; we could use the opposite sort | |
3230 | * order to shift away big hashes ahead of small ones. */ | |
5422a9e1 | 3231 | bool order_swapped; |
064af421 BP |
3232 | |
3233 | for (i = 0; i < from->n_hashes; i++) { | |
5422a9e1 | 3234 | double old_ratio, new_ratio; |
064af421 | 3235 | uint64_t delta = from->hashes[i]->tx_bytes; |
5422a9e1 JG |
3236 | |
3237 | if (delta == 0 || from->tx_bytes - delta == 0) { | |
3238 | /* Pointless move. */ | |
3239 | continue; | |
3240 | } | |
3241 | ||
3242 | order_swapped = from->tx_bytes - delta < to->tx_bytes + delta; | |
3243 | ||
3244 | if (to->tx_bytes == 0) { | |
3245 | /* Nothing on the new slave, move it. */ | |
3246 | break; | |
3247 | } | |
3248 | ||
3249 | old_ratio = (double)from->tx_bytes / to->tx_bytes; | |
3250 | new_ratio = (double)(from->tx_bytes - delta) / | |
3251 | (to->tx_bytes + delta); | |
3252 | ||
3253 | if (new_ratio == 0) { | |
3254 | /* Should already be covered but check to prevent division | |
3255 | * by zero. */ | |
3256 | continue; | |
3257 | } | |
3258 | ||
3259 | if (new_ratio < 1) { | |
3260 | new_ratio = 1 / new_ratio; | |
3261 | } | |
3262 | ||
3263 | if (old_ratio - new_ratio > 0.1) { | |
3264 | /* Would decrease the ratio, move it. */ | |
064af421 BP |
3265 | break; |
3266 | } | |
3267 | } | |
3268 | if (i < from->n_hashes) { | |
5422a9e1 JG |
3269 | bond_shift_load(from, to, i); |
3270 | port->bond_compat_is_stale = true; | |
3271 | ||
3272 | /* If the result of the migration changed the relative order of | |
3273 | * 'from' and 'to' swap them back to maintain invariants. */ | |
3274 | if (order_swapped) { | |
3275 | swap_bals(from, to); | |
3276 | } | |
064af421 BP |
3277 | |
3278 | /* Re-sort 'bals'. Note that this may make 'from' and 'to' | |
3279 | * point to different slave_balance structures. It is only | |
3280 | * valid to do these two operations in a row at all because we | |
3281 | * know that 'from' will not move past 'to' and vice versa. */ | |
3282 | resort_bals(from, bals, n_bals); | |
3283 | resort_bals(to, bals, n_bals); | |
3284 | } else { | |
3285 | from++; | |
3286 | } | |
3287 | } | |
3288 | } | |
3289 | ||
3290 | /* Implement exponentially weighted moving average. A weight of 1/2 causes | |
3291 | * historical data to decay to <1% in 7 rebalancing runs. */ | |
3292 | for (e = &port->bond_hash[0]; e <= &port->bond_hash[BOND_MASK]; e++) { | |
3293 | e->tx_bytes /= 2; | |
3294 | } | |
c47cd5f0 BP |
3295 | |
3296 | exit: | |
3297 | free(bals); | |
064af421 | 3298 | } |
2303f3b2 BP |
3299 | |
3300 | static void | |
3301 | bond_send_learning_packets(struct port *port) | |
3302 | { | |
3303 | struct bridge *br = port->bridge; | |
3304 | struct mac_entry *e; | |
3305 | struct ofpbuf packet; | |
3306 | int error, n_packets, n_errors; | |
3307 | ||
93dfc067 | 3308 | if (!port->n_ifaces || port->active_iface < 0) { |
2303f3b2 BP |
3309 | return; |
3310 | } | |
3311 | ||
3312 | ofpbuf_init(&packet, 128); | |
3313 | error = n_packets = n_errors = 0; | |
4e8e4213 | 3314 | LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { |
2303f3b2 | 3315 | union ofp_action actions[2], *a; |
2303f3b2 BP |
3316 | uint16_t dp_ifidx; |
3317 | tag_type tags = 0; | |
ae412e7d | 3318 | struct flow flow; |
2303f3b2 BP |
3319 | int retval; |
3320 | ||
3321 | if (e->port == port->port_idx | |
e58de0e3 | 3322 | || !choose_output_iface(port, e->mac, e->vlan, &dp_ifidx, &tags)) { |
2303f3b2 BP |
3323 | continue; |
3324 | } | |
3325 | ||
2303f3b2 BP |
3326 | /* Compose actions. */ |
3327 | memset(actions, 0, sizeof actions); | |
3328 | a = actions; | |
3329 | if (e->vlan) { | |
3330 | a->vlan_vid.type = htons(OFPAT_SET_VLAN_VID); | |
3331 | a->vlan_vid.len = htons(sizeof *a); | |
3332 | a->vlan_vid.vlan_vid = htons(e->vlan); | |
3333 | a++; | |
3334 | } | |
3335 | a->output.type = htons(OFPAT_OUTPUT); | |
3336 | a->output.len = htons(sizeof *a); | |
3337 | a->output.port = htons(odp_port_to_ofp_port(dp_ifidx)); | |
3338 | a++; | |
3339 | ||
3340 | /* Send packet. */ | |
3341 | n_packets++; | |
b9e8b45a BP |
3342 | compose_benign_packet(&packet, "Open vSwitch Bond Failover", 0xf177, |
3343 | e->mac); | |
659586ef | 3344 | flow_extract(&packet, 0, ODPP_NONE, &flow); |
2303f3b2 BP |
3345 | retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions, |
3346 | &packet); | |
3347 | if (retval) { | |
3348 | error = retval; | |
3349 | n_errors++; | |
3350 | } | |
3351 | } | |
3352 | ofpbuf_uninit(&packet); | |
3353 | ||
3354 | if (n_errors) { | |
3355 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3356 | VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning " | |
3357 | "packets, last error was: %s", | |
3358 | port->name, n_errors, n_packets, strerror(error)); | |
3359 | } else { | |
3360 | VLOG_DBG("bond %s: sent %d gratuitous learning packets", | |
3361 | port->name, n_packets); | |
3362 | } | |
3363 | } | |
064af421 | 3364 | \f |
da285df4 BP |
3365 | /* Bonding unixctl user interface functions. */ |
3366 | ||
3367 | static void | |
8ca79daa | 3368 | bond_unixctl_list(struct unixctl_conn *conn, |
c69ee87c | 3369 | const char *args OVS_UNUSED, void *aux OVS_UNUSED) |
da285df4 BP |
3370 | { |
3371 | struct ds ds = DS_EMPTY_INITIALIZER; | |
3372 | const struct bridge *br; | |
3373 | ||
be02e7c3 | 3374 | ds_put_cstr(&ds, "bridge\tbond\ttype\tslaves\n"); |
da285df4 | 3375 | |
4e8e4213 | 3376 | LIST_FOR_EACH (br, node, &all_bridges) { |
da285df4 BP |
3377 | size_t i; |
3378 | ||
3379 | for (i = 0; i < br->n_ports; i++) { | |
3380 | const struct port *port = br->ports[i]; | |
3381 | if (port->n_ifaces > 1) { | |
3382 | size_t j; | |
3383 | ||
be02e7c3 | 3384 | ds_put_format(&ds, "%s\t%s\t%s\t", br->name, port->name, |
27dcaa1a | 3385 | bond_mode_to_string(port->bond_mode)); |
da285df4 BP |
3386 | for (j = 0; j < port->n_ifaces; j++) { |
3387 | const struct iface *iface = port->ifaces[j]; | |
3388 | if (j) { | |
3389 | ds_put_cstr(&ds, ", "); | |
3390 | } | |
3391 | ds_put_cstr(&ds, iface->name); | |
3392 | } | |
3393 | ds_put_char(&ds, '\n'); | |
3394 | } | |
3395 | } | |
3396 | } | |
3397 | unixctl_command_reply(conn, 200, ds_cstr(&ds)); | |
3398 | ds_destroy(&ds); | |
3399 | } | |
3400 | ||
3401 | static struct port * | |
3402 | bond_find(const char *name) | |
3403 | { | |
3404 | const struct bridge *br; | |
3405 | ||
4e8e4213 | 3406 | LIST_FOR_EACH (br, node, &all_bridges) { |
da285df4 BP |
3407 | size_t i; |
3408 | ||
3409 | for (i = 0; i < br->n_ports; i++) { | |
3410 | struct port *port = br->ports[i]; | |
3411 | if (!strcmp(port->name, name) && port->n_ifaces > 1) { | |
3412 | return port; | |
3413 | } | |
3414 | } | |
3415 | } | |
3416 | return NULL; | |
3417 | } | |
3418 | ||
3419 | static void | |
8ca79daa | 3420 | bond_unixctl_show(struct unixctl_conn *conn, |
c69ee87c | 3421 | const char *args, void *aux OVS_UNUSED) |
da285df4 BP |
3422 | { |
3423 | struct ds ds = DS_EMPTY_INITIALIZER; | |
3424 | const struct port *port; | |
3425 | size_t j; | |
3426 | ||
3427 | port = bond_find(args); | |
3428 | if (!port) { | |
3429 | unixctl_command_reply(conn, 501, "no such bond"); | |
3430 | return; | |
3431 | } | |
3432 | ||
27dcaa1a EJ |
3433 | ds_put_format(&ds, "bond_mode: %s\n", |
3434 | bond_mode_to_string(port->bond_mode)); | |
63331829 EJ |
3435 | ds_put_format(&ds, "bond-detect-mode: %s\n", |
3436 | port->miimon ? "miimon" : "carrier"); | |
b42c5c24 EJ |
3437 | |
3438 | if (port->miimon) { | |
3439 | ds_put_format(&ds, "bond-miimon-interval: %lld\n", | |
3440 | port->bond_miimon_interval); | |
3441 | } | |
3442 | ||
da285df4 BP |
3443 | ds_put_format(&ds, "updelay: %d ms\n", port->updelay); |
3444 | ds_put_format(&ds, "downdelay: %d ms\n", port->downdelay); | |
be02e7c3 | 3445 | |
27dcaa1a | 3446 | if (port->bond_mode == BM_SLB) { |
be02e7c3 EJ |
3447 | ds_put_format(&ds, "next rebalance: %lld ms\n", |
3448 | port->bond_next_rebalance - time_msec()); | |
3449 | } | |
3450 | ||
da285df4 BP |
3451 | for (j = 0; j < port->n_ifaces; j++) { |
3452 | const struct iface *iface = port->ifaces[j]; | |
3453 | struct bond_entry *be; | |
3454 | ||
3455 | /* Basic info. */ | |
3456 | ds_put_format(&ds, "slave %s: %s\n", | |
3457 | iface->name, iface->enabled ? "enabled" : "disabled"); | |
3458 | if (j == port->active_iface) { | |
3459 | ds_put_cstr(&ds, "\tactive slave\n"); | |
3460 | } | |
3461 | if (iface->delay_expires != LLONG_MAX) { | |
3462 | ds_put_format(&ds, "\t%s expires in %lld ms\n", | |
3463 | iface->enabled ? "downdelay" : "updelay", | |
3464 | iface->delay_expires - time_msec()); | |
3465 | } | |
3466 | ||
27dcaa1a | 3467 | if (port->bond_mode != BM_SLB) { |
be02e7c3 EJ |
3468 | continue; |
3469 | } | |
3470 | ||
da285df4 BP |
3471 | /* Hashes. */ |
3472 | for (be = port->bond_hash; be <= &port->bond_hash[BOND_MASK]; be++) { | |
3473 | int hash = be - port->bond_hash; | |
3474 | struct mac_entry *me; | |
3475 | ||
3476 | if (be->iface_idx != j) { | |
3477 | continue; | |
3478 | } | |
3479 | ||
2886875a | 3480 | ds_put_format(&ds, "\thash %d: %"PRIu64" kB load\n", |
da285df4 BP |
3481 | hash, be->tx_bytes / 1024); |
3482 | ||
3483 | /* MACs. */ | |
4e8e4213 | 3484 | LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) { |
da285df4 BP |
3485 | uint16_t dp_ifidx; |
3486 | tag_type tags = 0; | |
e58de0e3 | 3487 | if (bond_hash(me->mac, me->vlan) == hash |
da285df4 | 3488 | && me->port != port->port_idx |
e58de0e3 EJ |
3489 | && choose_output_iface(port, me->mac, me->vlan, |
3490 | &dp_ifidx, &tags) | |
da285df4 BP |
3491 | && dp_ifidx == iface->dp_ifidx) |
3492 | { | |
3493 | ds_put_format(&ds, "\t\t"ETH_ADDR_FMT"\n", | |
3494 | ETH_ADDR_ARGS(me->mac)); | |
3495 | } | |
3496 | } | |
3497 | } | |
3498 | } | |
3499 | unixctl_command_reply(conn, 200, ds_cstr(&ds)); | |
3500 | ds_destroy(&ds); | |
3501 | } | |
3502 | ||
3503 | static void | |
8ca79daa | 3504 | bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, |
c69ee87c | 3505 | void *aux OVS_UNUSED) |
da285df4 BP |
3506 | { |
3507 | char *args = (char *) args_; | |
3508 | char *save_ptr = NULL; | |
3509 | char *bond_s, *hash_s, *slave_s; | |
da285df4 BP |
3510 | struct port *port; |
3511 | struct iface *iface; | |
3512 | struct bond_entry *entry; | |
3513 | int hash; | |
3514 | ||
3515 | bond_s = strtok_r(args, " ", &save_ptr); | |
3516 | hash_s = strtok_r(NULL, " ", &save_ptr); | |
3517 | slave_s = strtok_r(NULL, " ", &save_ptr); | |
3518 | if (!slave_s) { | |
3519 | unixctl_command_reply(conn, 501, | |
3520 | "usage: bond/migrate BOND HASH SLAVE"); | |
3521 | return; | |
3522 | } | |
3523 | ||
3524 | port = bond_find(bond_s); | |
3525 | if (!port) { | |
3526 | unixctl_command_reply(conn, 501, "no such bond"); | |
3527 | return; | |
3528 | } | |
3529 | ||
27dcaa1a | 3530 | if (port->bond_mode != BM_SLB) { |
be02e7c3 EJ |
3531 | unixctl_command_reply(conn, 501, "not an SLB bond"); |
3532 | return; | |
3533 | } | |
3534 | ||
557c178b | 3535 | if (strspn(hash_s, "0123456789") == strlen(hash_s)) { |
da285df4 BP |
3536 | hash = atoi(hash_s) & BOND_MASK; |
3537 | } else { | |
3538 | unixctl_command_reply(conn, 501, "bad hash"); | |
3539 | return; | |
3540 | } | |
3541 | ||
3542 | iface = port_lookup_iface(port, slave_s); | |
3543 | if (!iface) { | |
3544 | unixctl_command_reply(conn, 501, "no such slave"); | |
3545 | return; | |
3546 | } | |
3547 | ||
3548 | if (!iface->enabled) { | |
3549 | unixctl_command_reply(conn, 501, "cannot migrate to disabled slave"); | |
3550 | return; | |
3551 | } | |
3552 | ||
3553 | entry = &port->bond_hash[hash]; | |
3554 | ofproto_revalidate(port->bridge->ofproto, entry->iface_tag); | |
3555 | entry->iface_idx = iface->port_ifidx; | |
3556 | entry->iface_tag = tag_create_random(); | |
85c74638 | 3557 | port->bond_compat_is_stale = true; |
da285df4 BP |
3558 | unixctl_command_reply(conn, 200, "migrated"); |
3559 | } | |
3560 | ||
3561 | static void | |
8ca79daa | 3562 | bond_unixctl_set_active_slave(struct unixctl_conn *conn, const char *args_, |
c69ee87c | 3563 | void *aux OVS_UNUSED) |
da285df4 BP |
3564 | { |
3565 | char *args = (char *) args_; | |
3566 | char *save_ptr = NULL; | |
3567 | char *bond_s, *slave_s; | |
3568 | struct port *port; | |
3569 | struct iface *iface; | |
3570 | ||
3571 | bond_s = strtok_r(args, " ", &save_ptr); | |
3572 | slave_s = strtok_r(NULL, " ", &save_ptr); | |
3573 | if (!slave_s) { | |
3574 | unixctl_command_reply(conn, 501, | |
3575 | "usage: bond/set-active-slave BOND SLAVE"); | |
3576 | return; | |
3577 | } | |
3578 | ||
3579 | port = bond_find(bond_s); | |
3580 | if (!port) { | |
3581 | unixctl_command_reply(conn, 501, "no such bond"); | |
3582 | return; | |
3583 | } | |
3584 | ||
3585 | iface = port_lookup_iface(port, slave_s); | |
3586 | if (!iface) { | |
3587 | unixctl_command_reply(conn, 501, "no such slave"); | |
3588 | return; | |
3589 | } | |
3590 | ||
3591 | if (!iface->enabled) { | |
3592 | unixctl_command_reply(conn, 501, "cannot make disabled slave active"); | |
3593 | return; | |
3594 | } | |
3595 | ||
3596 | if (port->active_iface != iface->port_ifidx) { | |
3597 | ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag); | |
3598 | port->active_iface = iface->port_ifidx; | |
3599 | port->active_iface_tag = tag_create_random(); | |
3600 | VLOG_INFO("port %s: active interface is now %s", | |
3601 | port->name, iface->name); | |
3602 | bond_send_learning_packets(port); | |
3603 | unixctl_command_reply(conn, 200, "done"); | |
3604 | } else { | |
3605 | unixctl_command_reply(conn, 200, "no change"); | |
3606 | } | |
3607 | } | |
3608 | ||
3609 | static void | |
3610 | enable_slave(struct unixctl_conn *conn, const char *args_, bool enable) | |
3611 | { | |
3612 | char *args = (char *) args_; | |
3613 | char *save_ptr = NULL; | |
3614 | char *bond_s, *slave_s; | |
3615 | struct port *port; | |
3616 | struct iface *iface; | |
3617 | ||
3618 | bond_s = strtok_r(args, " ", &save_ptr); | |
3619 | slave_s = strtok_r(NULL, " ", &save_ptr); | |
3620 | if (!slave_s) { | |
3621 | unixctl_command_reply(conn, 501, | |
3622 | "usage: bond/enable/disable-slave BOND SLAVE"); | |
3623 | return; | |
3624 | } | |
3625 | ||
3626 | port = bond_find(bond_s); | |
3627 | if (!port) { | |
3628 | unixctl_command_reply(conn, 501, "no such bond"); | |
3629 | return; | |
3630 | } | |
3631 | ||
3632 | iface = port_lookup_iface(port, slave_s); | |
3633 | if (!iface) { | |
3634 | unixctl_command_reply(conn, 501, "no such slave"); | |
3635 | return; | |
3636 | } | |
3637 | ||
3638 | bond_enable_slave(iface, enable); | |
3639 | unixctl_command_reply(conn, 501, enable ? "enabled" : "disabled"); | |
3640 | } | |
3641 | ||
3642 | static void | |
8ca79daa | 3643 | bond_unixctl_enable_slave(struct unixctl_conn *conn, const char *args, |
c69ee87c | 3644 | void *aux OVS_UNUSED) |
da285df4 BP |
3645 | { |
3646 | enable_slave(conn, args, true); | |
3647 | } | |
3648 | ||
3649 | static void | |
8ca79daa | 3650 | bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args, |
c69ee87c | 3651 | void *aux OVS_UNUSED) |
da285df4 BP |
3652 | { |
3653 | enable_slave(conn, args, false); | |
3654 | } | |
3655 | ||
e0644b61 | 3656 | static void |
e58de0e3 | 3657 | bond_unixctl_hash(struct unixctl_conn *conn, const char *args_, |
c69ee87c | 3658 | void *aux OVS_UNUSED) |
e0644b61 | 3659 | { |
e58de0e3 | 3660 | char *args = (char *) args_; |
a27598cd EJ |
3661 | uint8_t mac[ETH_ADDR_LEN]; |
3662 | uint8_t hash; | |
3663 | char *hash_cstr; | |
e58de0e3 EJ |
3664 | unsigned int vlan; |
3665 | char *mac_s, *vlan_s; | |
3666 | char *save_ptr = NULL; | |
3667 | ||
3668 | mac_s = strtok_r(args, " ", &save_ptr); | |
3669 | vlan_s = strtok_r(NULL, " ", &save_ptr); | |
3670 | ||
3671 | if (vlan_s) { | |
3672 | if (sscanf(vlan_s, "%u", &vlan) != 1) { | |
3673 | unixctl_command_reply(conn, 501, "invalid vlan"); | |
3674 | return; | |
3675 | } | |
3676 | } else { | |
3677 | vlan = OFP_VLAN_NONE; | |
3678 | } | |
a27598cd | 3679 | |
e58de0e3 | 3680 | if (sscanf(mac_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac)) |
a27598cd | 3681 | == ETH_ADDR_SCAN_COUNT) { |
e58de0e3 | 3682 | hash = bond_hash(mac, vlan); |
a27598cd EJ |
3683 | |
3684 | hash_cstr = xasprintf("%u", hash); | |
3685 | unixctl_command_reply(conn, 200, hash_cstr); | |
3686 | free(hash_cstr); | |
3687 | } else { | |
3688 | unixctl_command_reply(conn, 501, "invalid mac"); | |
3689 | } | |
e0644b61 IC |
3690 | } |
3691 | ||
da285df4 BP |
3692 | static void |
3693 | bond_init(void) | |
3694 | { | |
8ca79daa BP |
3695 | unixctl_command_register("bond/list", bond_unixctl_list, NULL); |
3696 | unixctl_command_register("bond/show", bond_unixctl_show, NULL); | |
3697 | unixctl_command_register("bond/migrate", bond_unixctl_migrate, NULL); | |
da285df4 | 3698 | unixctl_command_register("bond/set-active-slave", |
8ca79daa BP |
3699 | bond_unixctl_set_active_slave, NULL); |
3700 | unixctl_command_register("bond/enable-slave", bond_unixctl_enable_slave, | |
3701 | NULL); | |
3702 | unixctl_command_register("bond/disable-slave", bond_unixctl_disable_slave, | |
3703 | NULL); | |
3704 | unixctl_command_register("bond/hash", bond_unixctl_hash, NULL); | |
da285df4 BP |
3705 | } |
3706 | \f | |
064af421 BP |
3707 | /* Port functions. */ |
3708 | ||
76343538 | 3709 | static struct port * |
064af421 BP |
3710 | port_create(struct bridge *br, const char *name) |
3711 | { | |
3712 | struct port *port; | |
3713 | ||
ec6fde61 | 3714 | port = xzalloc(sizeof *port); |
064af421 BP |
3715 | port->bridge = br; |
3716 | port->port_idx = br->n_ports; | |
3717 | port->vlan = -1; | |
3718 | port->trunks = NULL; | |
3719 | port->name = xstrdup(name); | |
3720 | port->active_iface = -1; | |
064af421 BP |
3721 | |
3722 | if (br->n_ports >= br->allocated_ports) { | |
3723 | br->ports = x2nrealloc(br->ports, &br->allocated_ports, | |
3724 | sizeof *br->ports); | |
3725 | } | |
3726 | br->ports[br->n_ports++] = port; | |
836fad5e | 3727 | shash_add_assert(&br->port_by_name, port->name, port); |
064af421 BP |
3728 | |
3729 | VLOG_INFO("created port %s on bridge %s", port->name, br->name); | |
3730 | bridge_flush(br); | |
76343538 BP |
3731 | |
3732 | return port; | |
064af421 BP |
3733 | } |
3734 | ||
c8143c88 BP |
3735 | static const char * |
3736 | get_port_other_config(const struct ovsrec_port *port, const char *key, | |
3737 | const char *default_value) | |
3738 | { | |
af6278e1 BP |
3739 | const char *value; |
3740 | ||
3741 | value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config, | |
3742 | key); | |
c8143c88 BP |
3743 | return value ? value : default_value; |
3744 | } | |
3745 | ||
4a1ee6ae BP |
3746 | static void |
3747 | port_del_ifaces(struct port *port, const struct ovsrec_port *cfg) | |
3748 | { | |
3749 | struct shash new_ifaces; | |
3750 | size_t i; | |
3751 | ||
3752 | /* Collect list of new interfaces. */ | |
3753 | shash_init(&new_ifaces); | |
3754 | for (i = 0; i < cfg->n_interfaces; i++) { | |
3755 | const char *name = cfg->interfaces[i]->name; | |
3756 | shash_add_once(&new_ifaces, name, NULL); | |
3757 | } | |
3758 | ||
3759 | /* Get rid of deleted interfaces. */ | |
3760 | for (i = 0; i < port->n_ifaces; ) { | |
3761 | if (!shash_find(&new_ifaces, cfg->interfaces[i]->name)) { | |
3762 | iface_destroy(port->ifaces[i]); | |
3763 | } else { | |
3764 | i++; | |
3765 | } | |
3766 | } | |
3767 | ||
3768 | shash_destroy(&new_ifaces); | |
3769 | } | |
3770 | ||
064af421 | 3771 | static void |
76343538 | 3772 | port_reconfigure(struct port *port, const struct ovsrec_port *cfg) |
064af421 | 3773 | { |
63331829 | 3774 | const char *detect_mode; |
4a1ee6ae | 3775 | struct shash new_ifaces; |
63331829 | 3776 | long long int next_rebalance, miimon_next_update; |
064af421 BP |
3777 | unsigned long *trunks; |
3778 | int vlan; | |
3779 | size_t i; | |
3780 | ||
76343538 BP |
3781 | port->cfg = cfg; |
3782 | ||
4a1ee6ae | 3783 | /* Update settings. */ |
76343538 BP |
3784 | port->updelay = cfg->bond_updelay; |
3785 | if (port->updelay < 0) { | |
3786 | port->updelay = 0; | |
3787 | } | |
7ef6b685 | 3788 | port->downdelay = cfg->bond_downdelay; |
76343538 BP |
3789 | if (port->downdelay < 0) { |
3790 | port->downdelay = 0; | |
064af421 | 3791 | } |
c8143c88 BP |
3792 | port->bond_rebalance_interval = atoi( |
3793 | get_port_other_config(cfg, "bond-rebalance-interval", "10000")); | |
3794 | if (port->bond_rebalance_interval < 1000) { | |
3795 | port->bond_rebalance_interval = 1000; | |
3796 | } | |
3797 | next_rebalance = time_msec() + port->bond_rebalance_interval; | |
3798 | if (port->bond_next_rebalance > next_rebalance) { | |
3799 | port->bond_next_rebalance = next_rebalance; | |
3800 | } | |
064af421 | 3801 | |
63331829 EJ |
3802 | detect_mode = get_port_other_config(cfg, "bond-detect-mode", |
3803 | "carrier"); | |
3804 | ||
3805 | if (!strcmp(detect_mode, "carrier")) { | |
3806 | port->miimon = false; | |
3807 | } else if (!strcmp(detect_mode, "miimon")) { | |
3808 | port->miimon = true; | |
3809 | } else { | |
3810 | port->miimon = false; | |
3811 | VLOG_WARN("port %s: unsupported bond-detect-mode %s, defaulting to " | |
3812 | "carrier", port->name, detect_mode); | |
3813 | } | |
3814 | ||
3815 | port->bond_miimon_interval = atoi( | |
3816 | get_port_other_config(cfg, "bond-miimon-interval", "200")); | |
3817 | if (port->bond_miimon_interval < 100) { | |
3818 | port->bond_miimon_interval = 100; | |
3819 | } | |
3820 | miimon_next_update = time_msec() + port->bond_miimon_interval; | |
3821 | if (port->bond_miimon_next_update > miimon_next_update) { | |
3822 | port->bond_miimon_next_update = miimon_next_update; | |
3823 | } | |
3824 | ||
27dcaa1a EJ |
3825 | if (!port->cfg->bond_mode || |
3826 | !strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_SLB))) { | |
3827 | port->bond_mode = BM_SLB; | |
3828 | } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_AB))) { | |
3829 | port->bond_mode = BM_AB; | |
be02e7c3 | 3830 | } else { |
27dcaa1a EJ |
3831 | port->bond_mode = BM_SLB; |
3832 | VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", | |
3833 | port->name, port->cfg->bond_mode, | |
3834 | bond_mode_to_string(port->bond_mode)); | |
be02e7c3 EJ |
3835 | } |
3836 | ||
4a1ee6ae BP |
3837 | /* Add new interfaces and update 'cfg' member of existing ones. */ |
3838 | shash_init(&new_ifaces); | |
3839 | for (i = 0; i < cfg->n_interfaces; i++) { | |
3840 | const struct ovsrec_interface *if_cfg = cfg->interfaces[i]; | |
76343538 BP |
3841 | struct iface *iface; |
3842 | ||
4a1ee6ae BP |
3843 | if (!shash_add_once(&new_ifaces, if_cfg->name, NULL)) { |
3844 | VLOG_WARN("port %s: %s specified twice as port interface", | |
3845 | port->name, if_cfg->name); | |
bcd49a45 | 3846 | iface_set_ofport(if_cfg, -1); |
4a1ee6ae BP |
3847 | continue; |
3848 | } | |
3849 | ||
3850 | iface = iface_lookup(port->bridge, if_cfg->name); | |
3851 | if (iface) { | |
3852 | if (iface->port != port) { | |
3853 | VLOG_ERR("bridge %s: %s interface is on multiple ports, " | |
3854 | "removing from %s", | |
3855 | port->bridge->name, if_cfg->name, iface->port->name); | |
3856 | continue; | |
3857 | } | |
149f577a | 3858 | iface->cfg = if_cfg; |
4a1ee6ae | 3859 | } else { |
6cefe1da | 3860 | iface = iface_create(port, if_cfg); |
064af421 | 3861 | } |
6cefe1da BP |
3862 | |
3863 | /* Determine interface type. The local port always has type | |
3864 | * "internal". Other ports take their type from the database and | |
3865 | * default to "system" if none is specified. */ | |
3866 | iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal" | |
3867 | : if_cfg->type[0] ? if_cfg->type | |
3868 | : "system"); | |
064af421 | 3869 | } |
4a1ee6ae | 3870 | shash_destroy(&new_ifaces); |
064af421 BP |
3871 | |
3872 | /* Get VLAN tag. */ | |
3873 | vlan = -1; | |
76343538 BP |
3874 | if (cfg->tag) { |
3875 | if (port->n_ifaces < 2) { | |
3876 | vlan = *cfg->tag; | |
064af421 BP |
3877 | if (vlan >= 0 && vlan <= 4095) { |
3878 | VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan); | |
76343538 BP |
3879 | } else { |
3880 | vlan = -1; | |
064af421 BP |
3881 | } |
3882 | } else { | |
3883 | /* It's possible that bonded, VLAN-tagged ports make sense. Maybe | |
3884 | * they even work as-is. But they have not been tested. */ | |
3885 | VLOG_WARN("port %s: VLAN tags not supported on bonded ports", | |
3886 | port->name); | |
3887 | } | |
3888 | } | |
3889 | if (port->vlan != vlan) { | |
3890 | port->vlan = vlan; | |
3891 | bridge_flush(port->bridge); | |
3892 | } | |
3893 | ||
3894 | /* Get trunked VLANs. */ | |
3895 | trunks = NULL; | |
3e9c481c | 3896 | if (vlan < 0 && cfg->n_trunks) { |
76343538 | 3897 | size_t n_errors; |
064af421 BP |
3898 | |
3899 | trunks = bitmap_allocate(4096); | |
064af421 | 3900 | n_errors = 0; |
76343538 BP |
3901 | for (i = 0; i < cfg->n_trunks; i++) { |
3902 | int trunk = cfg->trunks[i]; | |
064af421 BP |
3903 | if (trunk >= 0) { |
3904 | bitmap_set1(trunks, trunk); | |
3905 | } else { | |
3906 | n_errors++; | |
3907 | } | |
3908 | } | |
3909 | if (n_errors) { | |
3910 | VLOG_ERR("port %s: invalid values for %zu trunk VLANs", | |
76343538 | 3911 | port->name, cfg->n_trunks); |
064af421 | 3912 | } |
76343538 | 3913 | if (n_errors == cfg->n_trunks) { |
3e9c481c | 3914 | VLOG_ERR("port %s: no valid trunks, trunking all VLANs", |
76343538 | 3915 | port->name); |
3e9c481c BP |
3916 | bitmap_free(trunks); |
3917 | trunks = NULL; | |
064af421 | 3918 | } |
3e9c481c BP |
3919 | } else if (vlan >= 0 && cfg->n_trunks) { |
3920 | VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan", | |
3921 | port->name); | |
064af421 BP |
3922 | } |
3923 | if (trunks == NULL | |
3924 | ? port->trunks != NULL | |
3925 | : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) { | |
3926 | bridge_flush(port->bridge); | |
3927 | } | |
3928 | bitmap_free(port->trunks); | |
3929 | port->trunks = trunks; | |
064af421 BP |
3930 | } |
3931 | ||
3932 | static void | |
3933 | port_destroy(struct port *port) | |
3934 | { | |
3935 | if (port) { | |
3936 | struct bridge *br = port->bridge; | |
3937 | struct port *del; | |
37e7f427 | 3938 | int i; |
064af421 BP |
3939 | |
3940 | proc_net_compat_update_vlan(port->name, NULL, 0); | |
85c74638 | 3941 | proc_net_compat_update_bond(port->name, NULL); |
064af421 BP |
3942 | |
3943 | for (i = 0; i < MAX_MIRRORS; i++) { | |
3944 | struct mirror *m = br->mirrors[i]; | |
3945 | if (m && m->out_port == port) { | |
3946 | mirror_destroy(m); | |
3947 | } | |
3948 | } | |
3949 | ||
3950 | while (port->n_ifaces > 0) { | |
3951 | iface_destroy(port->ifaces[port->n_ifaces - 1]); | |
3952 | } | |
3953 | ||
836fad5e BP |
3954 | shash_find_and_delete_assert(&br->port_by_name, port->name); |
3955 | ||
064af421 BP |
3956 | del = br->ports[port->port_idx] = br->ports[--br->n_ports]; |
3957 | del->port_idx = port->port_idx; | |
3958 | ||
99707a7a JP |
3959 | VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name); |
3960 | ||
0bb5c3ec | 3961 | netdev_monitor_destroy(port->monitor); |
064af421 BP |
3962 | free(port->ifaces); |
3963 | bitmap_free(port->trunks); | |
3964 | free(port->name); | |
3965 | free(port); | |
3966 | bridge_flush(br); | |
3967 | } | |
3968 | } | |
3969 | ||
3970 | static struct port * | |
3971 | port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx) | |
3972 | { | |
3973 | struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx); | |
3974 | return iface ? iface->port : NULL; | |
3975 | } | |
3976 | ||
3977 | static struct port * | |
3978 | port_lookup(const struct bridge *br, const char *name) | |
3979 | { | |
836fad5e | 3980 | return shash_find_data(&br->port_by_name, name); |
064af421 BP |
3981 | } |
3982 | ||
da285df4 BP |
3983 | static struct iface * |
3984 | port_lookup_iface(const struct port *port, const char *name) | |
3985 | { | |
4a1ee6ae BP |
3986 | struct iface *iface = iface_lookup(port->bridge, name); |
3987 | return iface && iface->port == port ? iface : NULL; | |
da285df4 BP |
3988 | } |
3989 | ||
064af421 BP |
3990 | static void |
3991 | port_update_bonding(struct port *port) | |
3992 | { | |
0bb5c3ec BP |
3993 | if (port->monitor) { |
3994 | netdev_monitor_destroy(port->monitor); | |
3995 | port->monitor = NULL; | |
3996 | } | |
064af421 BP |
3997 | if (port->n_ifaces < 2) { |
3998 | /* Not a bonded port. */ | |
3999 | if (port->bond_hash) { | |
4000 | free(port->bond_hash); | |
4001 | port->bond_hash = NULL; | |
85c74638 | 4002 | port->bond_compat_is_stale = true; |
064af421 | 4003 | } |
be02e7c3 EJ |
4004 | |
4005 | port->bond_fake_iface = false; | |
064af421 | 4006 | } else { |
0bb5c3ec | 4007 | size_t i; |
064af421 | 4008 | |
27dcaa1a | 4009 | if (port->bond_mode == BM_SLB && !port->bond_hash) { |
064af421 BP |
4010 | port->bond_hash = xcalloc(BOND_MASK + 1, sizeof *port->bond_hash); |
4011 | for (i = 0; i <= BOND_MASK; i++) { | |
4012 | struct bond_entry *e = &port->bond_hash[i]; | |
4013 | e->iface_idx = -1; | |
4014 | e->tx_bytes = 0; | |
4015 | } | |
4016 | port->no_ifaces_tag = tag_create_random(); | |
4017 | bond_choose_active_iface(port); | |
c8143c88 BP |
4018 | port->bond_next_rebalance |
4019 | = time_msec() + port->bond_rebalance_interval; | |
8722022c BP |
4020 | |
4021 | if (port->cfg->bond_fake_iface) { | |
4022 | port->bond_next_fake_iface_update = time_msec(); | |
4023 | } | |
27dcaa1a | 4024 | } else if (port->bond_mode != BM_SLB) { |
be02e7c3 EJ |
4025 | free(port->bond_hash); |
4026 | port->bond_hash = NULL; | |
064af421 | 4027 | } |
85c74638 | 4028 | port->bond_compat_is_stale = true; |
76343538 | 4029 | port->bond_fake_iface = port->cfg->bond_fake_iface; |
0bb5c3ec | 4030 | |
63331829 EJ |
4031 | if (!port->miimon) { |
4032 | port->monitor = netdev_monitor_create(); | |
4033 | for (i = 0; i < port->n_ifaces; i++) { | |
4034 | netdev_monitor_add(port->monitor, port->ifaces[i]->netdev); | |
4035 | } | |
0bb5c3ec | 4036 | } |
064af421 BP |
4037 | } |
4038 | } | |
4039 | ||
4040 | static void | |
4041 | port_update_bond_compat(struct port *port) | |
4042 | { | |
2aebae83 | 4043 | struct compat_bond_hash compat_hashes[BOND_MASK + 1]; |
064af421 BP |
4044 | struct compat_bond bond; |
4045 | size_t i; | |
4046 | ||
27dcaa1a | 4047 | if (port->n_ifaces < 2 || port->bond_mode != BM_SLB) { |
85c74638 | 4048 | proc_net_compat_update_bond(port->name, NULL); |
064af421 BP |
4049 | return; |
4050 | } | |
4051 | ||
4052 | bond.up = false; | |
4053 | bond.updelay = port->updelay; | |
4054 | bond.downdelay = port->downdelay; | |
2aebae83 BP |
4055 | |
4056 | bond.n_hashes = 0; | |
4057 | bond.hashes = compat_hashes; | |
4058 | if (port->bond_hash) { | |
4059 | const struct bond_entry *e; | |
4060 | for (e = port->bond_hash; e <= &port->bond_hash[BOND_MASK]; e++) { | |
4061 | if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) { | |
4062 | struct compat_bond_hash *cbh = &bond.hashes[bond.n_hashes++]; | |
4063 | cbh->hash = e - port->bond_hash; | |
4064 | cbh->netdev_name = port->ifaces[e->iface_idx]->name; | |
4065 | } | |
4066 | } | |
4067 | } | |
4068 | ||
064af421 BP |
4069 | bond.n_slaves = port->n_ifaces; |
4070 | bond.slaves = xmalloc(port->n_ifaces * sizeof *bond.slaves); | |
4071 | for (i = 0; i < port->n_ifaces; i++) { | |
4072 | struct iface *iface = port->ifaces[i]; | |
4073 | struct compat_bond_slave *slave = &bond.slaves[i]; | |
4074 | slave->name = iface->name; | |
6643c03b JP |
4075 | |
4076 | /* We need to make the same determination as the Linux bonding | |
4077 | * code to determine whether a slave should be consider "up". | |
d295e8e9 | 4078 | * The Linux function bond_miimon_inspect() supports four |
6643c03b | 4079 | * BOND_LINK_* states: |
d295e8e9 | 4080 | * |
6643c03b JP |
4081 | * - BOND_LINK_UP: carrier detected, updelay has passed. |
4082 | * - BOND_LINK_FAIL: carrier lost, downdelay in progress. | |
4083 | * - BOND_LINK_DOWN: carrier lost, downdelay has passed. | |
4084 | * - BOND_LINK_BACK: carrier detected, updelay in progress. | |
4085 | * | |
d295e8e9 | 4086 | * The function bond_info_show_slave() only considers BOND_LINK_UP |
6643c03b JP |
4087 | * to be "up" and anything else to be "down". |
4088 | */ | |
4089 | slave->up = iface->enabled && iface->delay_expires == LLONG_MAX; | |
064af421 BP |
4090 | if (slave->up) { |
4091 | bond.up = true; | |
4092 | } | |
96b9b7a9 | 4093 | netdev_get_etheraddr(iface->netdev, slave->mac); |
064af421 | 4094 | } |
2aebae83 | 4095 | |
76343538 | 4096 | if (port->bond_fake_iface) { |
c6303ea1 JG |
4097 | struct netdev *bond_netdev; |
4098 | ||
149f577a | 4099 | if (!netdev_open_default(port->name, &bond_netdev)) { |
c6303ea1 JG |
4100 | if (bond.up) { |
4101 | netdev_turn_flags_on(bond_netdev, NETDEV_UP, true); | |
4102 | } else { | |
4103 | netdev_turn_flags_off(bond_netdev, NETDEV_UP, true); | |
4104 | } | |
4105 | netdev_close(bond_netdev); | |
4106 | } | |
4107 | } | |
4108 | ||
064af421 BP |
4109 | proc_net_compat_update_bond(port->name, &bond); |
4110 | free(bond.slaves); | |
4111 | } | |
4112 | ||
4113 | static void | |
4114 | port_update_vlan_compat(struct port *port) | |
4115 | { | |
4116 | struct bridge *br = port->bridge; | |
4117 | char *vlandev_name = NULL; | |
4118 | ||
4119 | if (port->vlan > 0) { | |
4120 | /* Figure out the name that the VLAN device should actually have, if it | |
4121 | * existed. This takes some work because the VLAN device would not | |
4122 | * have port->name in its name; rather, it would have the trunk port's | |
4123 | * name, and 'port' would be attached to a bridge that also had the | |
4124 | * VLAN device one of its ports. So we need to find a trunk port that | |
4125 | * includes port->vlan. | |
4126 | * | |
4127 | * There might be more than one candidate. This doesn't happen on | |
4128 | * XenServer, so if it happens we just pick the first choice in | |
4129 | * alphabetical order instead of creating multiple VLAN devices. */ | |
4130 | size_t i; | |
4131 | for (i = 0; i < br->n_ports; i++) { | |
4132 | struct port *p = br->ports[i]; | |
4133 | if (port_trunks_vlan(p, port->vlan) | |
4134 | && p->n_ifaces | |
4135 | && (!vlandev_name || strcmp(p->name, vlandev_name) <= 0)) | |
4136 | { | |
96b9b7a9 BP |
4137 | uint8_t ea[ETH_ADDR_LEN]; |
4138 | netdev_get_etheraddr(p->ifaces[0]->netdev, ea); | |
064af421 BP |
4139 | if (!eth_addr_is_multicast(ea) && |
4140 | !eth_addr_is_reserved(ea) && | |
4141 | !eth_addr_is_zero(ea)) { | |
4142 | vlandev_name = p->name; | |
4143 | } | |
4144 | } | |
4145 | } | |
4146 | } | |
4147 | proc_net_compat_update_vlan(port->name, vlandev_name, port->vlan); | |
4148 | } | |
4149 | \f | |
4150 | /* Interface functions. */ | |
4151 | ||
b31bcf60 EJ |
4152 | static void |
4153 | iface_send_packet(struct iface *iface, struct ofpbuf *packet) | |
4154 | { | |
4155 | struct flow flow; | |
4156 | union ofp_action action; | |
4157 | ||
4158 | memset(&action, 0, sizeof action); | |
4159 | action.output.type = htons(OFPAT_OUTPUT); | |
4160 | action.output.len = htons(sizeof action); | |
4161 | action.output.port = htons(odp_port_to_ofp_port(iface->dp_ifidx)); | |
4162 | ||
4163 | flow_extract(packet, 0, ODPP_NONE, &flow); | |
4164 | ||
4165 | if (ofproto_send_packet(iface->port->bridge->ofproto, &flow, &action, 1, | |
4166 | packet)) { | |
4167 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
4168 | VLOG_WARN_RL(&rl, "interface %s: Failed to send packet.", iface->name); | |
4169 | } | |
4170 | } | |
4171 | ||
76343538 | 4172 | static struct iface * |
a740f0de | 4173 | iface_create(struct port *port, const struct ovsrec_interface *if_cfg) |
064af421 | 4174 | { |
4a1ee6ae | 4175 | struct bridge *br = port->bridge; |
064af421 | 4176 | struct iface *iface; |
a740f0de | 4177 | char *name = if_cfg->name; |
064af421 | 4178 | |
ec6fde61 | 4179 | iface = xzalloc(sizeof *iface); |
064af421 BP |
4180 | iface->port = port; |
4181 | iface->port_ifidx = port->n_ifaces; | |
4182 | iface->name = xstrdup(name); | |
4183 | iface->dp_ifidx = -1; | |
4184 | iface->tag = tag_create_random(); | |
064af421 | 4185 | iface->delay_expires = LLONG_MAX; |
0c6aea3f | 4186 | iface->netdev = NULL; |
149f577a | 4187 | iface->cfg = if_cfg; |
064af421 | 4188 | |
2457b24f JG |
4189 | shash_add_assert(&br->iface_by_name, iface->name, iface); |
4190 | ||
064af421 BP |
4191 | if (port->n_ifaces >= port->allocated_ifaces) { |
4192 | port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces, | |
4193 | sizeof *port->ifaces); | |
4194 | } | |
4195 | port->ifaces[port->n_ifaces++] = iface; | |
4196 | if (port->n_ifaces > 1) { | |
4a1ee6ae | 4197 | br->has_bonded_ports = true; |
064af421 BP |
4198 | } |
4199 | ||
4200 | VLOG_DBG("attached network device %s to port %s", iface->name, port->name); | |
4201 | ||
4a1ee6ae | 4202 | bridge_flush(br); |
76343538 BP |
4203 | |
4204 | return iface; | |
064af421 BP |
4205 | } |
4206 | ||
4207 | static void | |
4208 | iface_destroy(struct iface *iface) | |
4209 | { | |
4210 | if (iface) { | |
4211 | struct port *port = iface->port; | |
4212 | struct bridge *br = port->bridge; | |
4213 | bool del_active = port->active_iface == iface->port_ifidx; | |
4214 | struct iface *del; | |
4215 | ||
7ca5f243 EJ |
4216 | if (port->monitor) { |
4217 | netdev_monitor_remove(port->monitor, iface->netdev); | |
4218 | } | |
4219 | ||
4a1ee6ae BP |
4220 | shash_find_and_delete_assert(&br->iface_by_name, iface->name); |
4221 | ||
064af421 | 4222 | if (iface->dp_ifidx >= 0) { |
d9a8717a | 4223 | hmap_remove(&br->ifaces, &iface->dp_ifidx_node); |
064af421 BP |
4224 | } |
4225 | ||
4226 | del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces]; | |
4227 | del->port_ifidx = iface->port_ifidx; | |
4228 | ||
0c6aea3f | 4229 | netdev_close(iface->netdev); |
064af421 BP |
4230 | |
4231 | if (del_active) { | |
4232 | ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag); | |
4233 | bond_choose_active_iface(port); | |
2303f3b2 | 4234 | bond_send_learning_packets(port); |
064af421 BP |
4235 | } |
4236 | ||
b31bcf60 EJ |
4237 | cfm_destroy(iface->cfm); |
4238 | ||
a740f0de JG |
4239 | free(iface->name); |
4240 | free(iface); | |
4241 | ||
064af421 BP |
4242 | bridge_flush(port->bridge); |
4243 | } | |
4244 | } | |
4245 | ||
4246 | static struct iface * | |
4247 | iface_lookup(const struct bridge *br, const char *name) | |
4248 | { | |
4a1ee6ae | 4249 | return shash_find_data(&br->iface_by_name, name); |
064af421 BP |
4250 | } |
4251 | ||
4252 | static struct iface * | |
4253 | iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx) | |
4254 | { | |
d9a8717a BP |
4255 | struct iface *iface; |
4256 | ||
4e8e4213 | 4257 | HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node, |
d9a8717a BP |
4258 | hash_int(dp_ifidx, 0), &br->ifaces) { |
4259 | if (iface->dp_ifidx == dp_ifidx) { | |
4260 | return iface; | |
4261 | } | |
4262 | } | |
4263 | return NULL; | |
064af421 | 4264 | } |
557d8e6c | 4265 | |
52df17e7 BP |
4266 | /* Set Ethernet address of 'iface', if one is specified in the configuration |
4267 | * file. */ | |
4268 | static void | |
4269 | iface_set_mac(struct iface *iface) | |
4270 | { | |
76343538 | 4271 | uint8_t ea[ETH_ADDR_LEN]; |
52df17e7 | 4272 | |
76343538 | 4273 | if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) { |
52df17e7 BP |
4274 | if (eth_addr_is_multicast(ea)) { |
4275 | VLOG_ERR("interface %s: cannot set MAC to multicast address", | |
4276 | iface->name); | |
4277 | } else if (iface->dp_ifidx == ODPP_LOCAL) { | |
4278 | VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead", | |
4279 | iface->name, iface->name); | |
4280 | } else { | |
4d678233 | 4281 | int error = netdev_set_etheraddr(iface->netdev, ea); |
52df17e7 BP |
4282 | if (error) { |
4283 | VLOG_ERR("interface %s: setting MAC failed (%s)", | |
4284 | iface->name, strerror(error)); | |
4285 | } | |
4286 | } | |
4287 | } | |
4288 | } | |
c1c9c9c4 | 4289 | |
bcd49a45 BP |
4290 | /* Sets the ofport column of 'if_cfg' to 'ofport'. */ |
4291 | static void | |
4292 | iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport) | |
4293 | { | |
4294 | if (if_cfg) { | |
4295 | ovsrec_interface_set_ofport(if_cfg, &ofport, 1); | |
4296 | } | |
4297 | } | |
4298 | ||
43776b8f BP |
4299 | /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'. |
4300 | * | |
4301 | * The value strings in '*shash' are taken directly from values[], not copied, | |
4302 | * so the caller should not modify or free them. */ | |
c1c9c9c4 BP |
4303 | static void |
4304 | shash_from_ovs_idl_map(char **keys, char **values, size_t n, | |
4305 | struct shash *shash) | |
4306 | { | |
4307 | size_t i; | |
4308 | ||
4309 | shash_init(shash); | |
4310 | for (i = 0; i < n; i++) { | |
4311 | shash_add(shash, keys[i], values[i]); | |
4312 | } | |
4313 | } | |
4314 | ||
ea763e0e EJ |
4315 | /* Creates 'keys' and 'values' arrays from 'shash'. |
4316 | * | |
4317 | * Sets 'keys' and 'values' to heap allocated arrays representing the key-value | |
4318 | * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They | |
4319 | * are populated with with strings taken directly from 'shash' and thus have | |
4320 | * the same ownership of the key-value pairs in shash. | |
4321 | */ | |
4322 | static void | |
4323 | shash_to_ovs_idl_map(struct shash *shash, | |
4324 | char ***keys, char ***values, size_t *n) | |
4325 | { | |
4326 | size_t i, count; | |
4327 | char **k, **v; | |
4328 | struct shash_node *sn; | |
4329 | ||
4330 | count = shash_count(shash); | |
4331 | ||
4332 | k = xmalloc(count * sizeof *k); | |
4333 | v = xmalloc(count * sizeof *v); | |
4334 | ||
4335 | i = 0; | |
4336 | SHASH_FOR_EACH(sn, shash) { | |
4337 | k[i] = sn->name; | |
4338 | v[i] = sn->data; | |
4339 | i++; | |
4340 | } | |
4341 | ||
4342 | *n = count; | |
4343 | *keys = k; | |
4344 | *values = v; | |
4345 | } | |
4346 | ||
c1c9c9c4 BP |
4347 | struct iface_delete_queues_cbdata { |
4348 | struct netdev *netdev; | |
44fca7f9 | 4349 | const struct ovsdb_datum *queues; |
c1c9c9c4 BP |
4350 | }; |
4351 | ||
4352 | static bool | |
44fca7f9 | 4353 | queue_ids_include(const struct ovsdb_datum *queues, int64_t target) |
c1c9c9c4 | 4354 | { |
44fca7f9 | 4355 | union ovsdb_atom atom; |
c1c9c9c4 | 4356 | |
44fca7f9 BP |
4357 | atom.integer = target; |
4358 | return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX; | |
c1c9c9c4 BP |
4359 | } |
4360 | ||
4361 | static void | |
4362 | iface_delete_queues(unsigned int queue_id, | |
4363 | const struct shash *details OVS_UNUSED, void *cbdata_) | |
4364 | { | |
4365 | struct iface_delete_queues_cbdata *cbdata = cbdata_; | |
4366 | ||
44fca7f9 | 4367 | if (!queue_ids_include(cbdata->queues, queue_id)) { |
c1c9c9c4 BP |
4368 | netdev_delete_queue(cbdata->netdev, queue_id); |
4369 | } | |
4370 | } | |
4371 | ||
4372 | static void | |
4373 | iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos) | |
4374 | { | |
4375 | if (!qos || qos->type[0] == '\0') { | |
4376 | netdev_set_qos(iface->netdev, NULL, NULL); | |
4377 | } else { | |
4378 | struct iface_delete_queues_cbdata cbdata; | |
4379 | struct shash details; | |
4380 | size_t i; | |
4381 | ||
4382 | /* Configure top-level Qos for 'iface'. */ | |
4383 | shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config, | |
4384 | qos->n_other_config, &details); | |
4385 | netdev_set_qos(iface->netdev, qos->type, &details); | |
4386 | shash_destroy(&details); | |
4387 | ||
4388 | /* Deconfigure queues that were deleted. */ | |
4389 | cbdata.netdev = iface->netdev; | |
44fca7f9 BP |
4390 | cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER, |
4391 | OVSDB_TYPE_UUID); | |
c1c9c9c4 BP |
4392 | netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata); |
4393 | ||
4394 | /* Configure queues for 'iface'. */ | |
4395 | for (i = 0; i < qos->n_queues; i++) { | |
4396 | const struct ovsrec_queue *queue = qos->value_queues[i]; | |
4397 | unsigned int queue_id = qos->key_queues[i]; | |
4398 | ||
4399 | shash_from_ovs_idl_map(queue->key_other_config, | |
4400 | queue->value_other_config, | |
4401 | queue->n_other_config, &details); | |
4402 | netdev_set_queue(iface->netdev, queue_id, &details); | |
4403 | shash_destroy(&details); | |
4404 | } | |
4405 | } | |
4406 | } | |
b31bcf60 EJ |
4407 | |
4408 | static void | |
4409 | iface_update_cfm(struct iface *iface) | |
4410 | { | |
4411 | size_t i; | |
4412 | struct cfm *cfm; | |
4413 | uint16_t *remote_mps; | |
4414 | struct ovsrec_monitor *mon; | |
4415 | uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN]; | |
4416 | ||
4417 | mon = iface->cfg->monitor; | |
4418 | ||
4419 | if (!mon) { | |
4420 | return; | |
4421 | } | |
4422 | ||
4423 | if (netdev_get_etheraddr(iface->netdev, ea)) { | |
4424 | VLOG_WARN("interface %s: Failed to get ethernet address. " | |
4425 | "Skipping Monitor.", iface->name); | |
4426 | return; | |
4427 | } | |
4428 | ||
4429 | if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) { | |
4430 | VLOG_WARN("interface %s: Failed to generate MAID.", iface->name); | |
4431 | return; | |
4432 | } | |
4433 | ||
4434 | if (!iface->cfm) { | |
4435 | iface->cfm = cfm_create(); | |
4436 | } | |
4437 | ||
4438 | cfm = iface->cfm; | |
4439 | cfm->mpid = mon->mpid; | |
4440 | cfm->interval = mon->interval ? *mon->interval : 1000; | |
4441 | ||
4442 | memcpy(cfm->eth_src, ea, sizeof cfm->eth_src); | |
4443 | memcpy(cfm->maid, maid, sizeof cfm->maid); | |
4444 | ||
4445 | remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps); | |
4446 | for(i = 0; i < mon->n_remote_mps; i++) { | |
4447 | remote_mps[i] = mon->remote_mps[i]->mpid; | |
4448 | } | |
4449 | cfm_update_remote_mps(cfm, remote_mps, mon->n_remote_mps); | |
4450 | free(remote_mps); | |
4451 | ||
4452 | if (!cfm_configure(iface->cfm)) { | |
4453 | cfm_destroy(iface->cfm); | |
4454 | iface->cfm = NULL; | |
4455 | } | |
4456 | } | |
064af421 BP |
4457 | \f |
4458 | /* Port mirroring. */ | |
4459 | ||
dd0d105c BP |
4460 | static struct mirror * |
4461 | mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid) | |
4462 | { | |
4463 | int i; | |
4464 | ||
4465 | for (i = 0; i < MAX_MIRRORS; i++) { | |
4466 | struct mirror *m = br->mirrors[i]; | |
4467 | if (m && uuid_equals(uuid, &m->uuid)) { | |
4468 | return m; | |
4469 | } | |
4470 | } | |
4471 | return NULL; | |
4472 | } | |
4473 | ||
064af421 | 4474 | static void |
37e7f427 | 4475 | mirror_reconfigure(struct bridge *br) |
064af421 | 4476 | { |
f2d7fd66 | 4477 | unsigned long *rspan_vlans; |
37e7f427 | 4478 | int i; |
064af421 | 4479 | |
dd0d105c | 4480 | /* Get rid of deleted mirrors. */ |
064af421 | 4481 | for (i = 0; i < MAX_MIRRORS; i++) { |
dd0d105c BP |
4482 | struct mirror *m = br->mirrors[i]; |
4483 | if (m) { | |
4484 | const struct ovsdb_datum *mc; | |
4485 | union ovsdb_atom atom; | |
4486 | ||
4487 | mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID); | |
4488 | atom.uuid = br->mirrors[i]->uuid; | |
4489 | if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) { | |
4490 | mirror_destroy(m); | |
4491 | } | |
064af421 BP |
4492 | } |
4493 | } | |
4494 | ||
dd0d105c | 4495 | /* Add new mirrors and reconfigure existing ones. */ |
37e7f427 BP |
4496 | for (i = 0; i < br->cfg->n_mirrors; i++) { |
4497 | struct ovsrec_mirror *cfg = br->cfg->mirrors[i]; | |
dd0d105c BP |
4498 | struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid); |
4499 | if (m) { | |
4500 | mirror_reconfigure_one(m, cfg); | |
4501 | } else { | |
4502 | mirror_create(br, cfg); | |
064af421 BP |
4503 | } |
4504 | } | |
4505 | ||
4506 | /* Update port reserved status. */ | |
4507 | for (i = 0; i < br->n_ports; i++) { | |
4508 | br->ports[i]->is_mirror_output_port = false; | |
4509 | } | |
4510 | for (i = 0; i < MAX_MIRRORS; i++) { | |
4511 | struct mirror *m = br->mirrors[i]; | |
4512 | if (m && m->out_port) { | |
4513 | m->out_port->is_mirror_output_port = true; | |
4514 | } | |
4515 | } | |
f2d7fd66 | 4516 | |
8f30d09a | 4517 | /* Update flooded vlans (for RSPAN). */ |
f2d7fd66 | 4518 | rspan_vlans = NULL; |
37e7f427 | 4519 | if (br->cfg->n_flood_vlans) { |
f2d7fd66 JG |
4520 | rspan_vlans = bitmap_allocate(4096); |
4521 | ||
37e7f427 BP |
4522 | for (i = 0; i < br->cfg->n_flood_vlans; i++) { |
4523 | int64_t vlan = br->cfg->flood_vlans[i]; | |
4524 | if (vlan >= 0 && vlan < 4096) { | |
f2d7fd66 | 4525 | bitmap_set1(rspan_vlans, vlan); |
37e7f427 | 4526 | VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64, |
42061b2a | 4527 | br->name, vlan); |
f2d7fd66 | 4528 | } else { |
37e7f427 BP |
4529 | VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN", |
4530 | br->name, vlan); | |
f2d7fd66 JG |
4531 | } |
4532 | } | |
4533 | } | |
8f30d09a | 4534 | if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) { |
f2d7fd66 JG |
4535 | bridge_flush(br); |
4536 | } | |
064af421 BP |
4537 | } |
4538 | ||
dd0d105c BP |
4539 | static void |
4540 | mirror_create(struct bridge *br, struct ovsrec_mirror *cfg) | |
064af421 BP |
4541 | { |
4542 | struct mirror *m; | |
4543 | size_t i; | |
4544 | ||
4545 | for (i = 0; ; i++) { | |
4546 | if (i >= MAX_MIRRORS) { | |
4547 | VLOG_WARN("bridge %s: maximum of %d port mirrors reached, " | |
dd0d105c BP |
4548 | "cannot create %s", br->name, MAX_MIRRORS, cfg->name); |
4549 | return; | |
064af421 BP |
4550 | } |
4551 | if (!br->mirrors[i]) { | |
4552 | break; | |
4553 | } | |
4554 | } | |
4555 | ||
dd0d105c | 4556 | VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name); |
064af421 BP |
4557 | bridge_flush(br); |
4558 | ||
ec6fde61 | 4559 | br->mirrors[i] = m = xzalloc(sizeof *m); |
064af421 BP |
4560 | m->bridge = br; |
4561 | m->idx = i; | |
dd0d105c | 4562 | m->name = xstrdup(cfg->name); |
37e7f427 BP |
4563 | shash_init(&m->src_ports); |
4564 | shash_init(&m->dst_ports); | |
064af421 BP |
4565 | m->vlans = NULL; |
4566 | m->n_vlans = 0; | |
4567 | m->out_vlan = -1; | |
4568 | m->out_port = NULL; | |
37e7f427 | 4569 | |
dd0d105c | 4570 | mirror_reconfigure_one(m, cfg); |
064af421 BP |
4571 | } |
4572 | ||
4573 | static void | |
4574 | mirror_destroy(struct mirror *m) | |
4575 | { | |
4576 | if (m) { | |
4577 | struct bridge *br = m->bridge; | |
4578 | size_t i; | |
4579 | ||
4580 | for (i = 0; i < br->n_ports; i++) { | |
4581 | br->ports[i]->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx); | |
4582 | br->ports[i]->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx); | |
4583 | } | |
4584 | ||
37e7f427 BP |
4585 | shash_destroy(&m->src_ports); |
4586 | shash_destroy(&m->dst_ports); | |
064af421 BP |
4587 | free(m->vlans); |
4588 | ||
4589 | m->bridge->mirrors[m->idx] = NULL; | |
786880a5 | 4590 | free(m->name); |
064af421 BP |
4591 | free(m); |
4592 | ||
4593 | bridge_flush(br); | |
4594 | } | |
4595 | } | |
4596 | ||
4597 | static void | |
37e7f427 BP |
4598 | mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports, |
4599 | struct shash *names) | |
064af421 | 4600 | { |
064af421 BP |
4601 | size_t i; |
4602 | ||
37e7f427 BP |
4603 | for (i = 0; i < n_ports; i++) { |
4604 | const char *name = ports[i]->name; | |
064af421 | 4605 | if (port_lookup(m->bridge, name)) { |
37e7f427 | 4606 | shash_add_once(names, name, NULL); |
064af421 | 4607 | } else { |
37e7f427 BP |
4608 | VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent " |
4609 | "port %s", m->bridge->name, m->name, name); | |
064af421 BP |
4610 | } |
4611 | } | |
064af421 BP |
4612 | } |
4613 | ||
4614 | static size_t | |
37e7f427 BP |
4615 | mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg, |
4616 | int **vlans) | |
064af421 | 4617 | { |
37e7f427 BP |
4618 | size_t n_vlans; |
4619 | size_t i; | |
064af421 | 4620 | |
0b523ea7 | 4621 | *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan); |
064af421 | 4622 | n_vlans = 0; |
37e7f427 BP |
4623 | for (i = 0; i < cfg->n_select_vlan; i++) { |
4624 | int64_t vlan = cfg->select_vlan[i]; | |
4625 | if (vlan < 0 || vlan > 4095) { | |
4626 | VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64, | |
4627 | m->bridge->name, m->name, vlan); | |
064af421 BP |
4628 | } else { |
4629 | (*vlans)[n_vlans++] = vlan; | |
4630 | } | |
4631 | } | |
4632 | return n_vlans; | |
4633 | } | |
4634 | ||
4635 | static bool | |
4636 | vlan_is_mirrored(const struct mirror *m, int vlan) | |
4637 | { | |
4638 | size_t i; | |
4639 | ||
4640 | for (i = 0; i < m->n_vlans; i++) { | |
4641 | if (m->vlans[i] == vlan) { | |
4642 | return true; | |
4643 | } | |
4644 | } | |
4645 | return false; | |
4646 | } | |
4647 | ||
4648 | static bool | |
4649 | port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p) | |
4650 | { | |
4651 | size_t i; | |
4652 | ||
4653 | for (i = 0; i < m->n_vlans; i++) { | |
4654 | if (port_trunks_vlan(p, m->vlans[i])) { | |
4655 | return true; | |
4656 | } | |
4657 | } | |
4658 | return false; | |
4659 | } | |
4660 | ||
4661 | static void | |
37e7f427 | 4662 | mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg) |
064af421 | 4663 | { |
37e7f427 | 4664 | struct shash src_ports, dst_ports; |
064af421 | 4665 | mirror_mask_t mirror_bit; |
064af421 BP |
4666 | struct port *out_port; |
4667 | int out_vlan; | |
4668 | size_t n_vlans; | |
4669 | int *vlans; | |
4670 | size_t i; | |
064af421 | 4671 | |
dd0d105c BP |
4672 | /* Set name. */ |
4673 | if (strcmp(cfg->name, m->name)) { | |
4674 | free(m->name); | |
4675 | m->name = xstrdup(cfg->name); | |
4676 | } | |
4677 | ||
064af421 | 4678 | /* Get output port. */ |
37e7f427 BP |
4679 | if (cfg->output_port) { |
4680 | out_port = port_lookup(m->bridge, cfg->output_port->name); | |
064af421 | 4681 | if (!out_port) { |
37e7f427 BP |
4682 | VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge", |
4683 | m->bridge->name, m->name); | |
064af421 | 4684 | mirror_destroy(m); |
064af421 BP |
4685 | return; |
4686 | } | |
4687 | out_vlan = -1; | |
4688 | ||
37e7f427 BP |
4689 | if (cfg->output_vlan) { |
4690 | VLOG_ERR("bridge %s: mirror %s specifies both output port and " | |
4691 | "output vlan; ignoring output vlan", | |
4692 | m->bridge->name, m->name); | |
064af421 | 4693 | } |
37e7f427 | 4694 | } else if (cfg->output_vlan) { |
064af421 | 4695 | out_port = NULL; |
37e7f427 | 4696 | out_vlan = *cfg->output_vlan; |
064af421 | 4697 | } else { |
37e7f427 BP |
4698 | VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring", |
4699 | m->bridge->name, m->name); | |
064af421 | 4700 | mirror_destroy(m); |
064af421 BP |
4701 | return; |
4702 | } | |
4703 | ||
37e7f427 BP |
4704 | shash_init(&src_ports); |
4705 | shash_init(&dst_ports); | |
939ff267 BP |
4706 | if (cfg->select_all) { |
4707 | for (i = 0; i < m->bridge->n_ports; i++) { | |
4708 | const char *name = m->bridge->ports[i]->name; | |
4709 | shash_add_once(&src_ports, name, NULL); | |
4710 | shash_add_once(&dst_ports, name, NULL); | |
4711 | } | |
4712 | vlans = NULL; | |
4713 | n_vlans = 0; | |
4714 | } else { | |
4715 | /* Get ports, and drop duplicates and ports that don't exist. */ | |
4716 | mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port, | |
4717 | &src_ports); | |
4718 | mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port, | |
4719 | &dst_ports); | |
064af421 | 4720 | |
939ff267 BP |
4721 | /* Get all the vlans, and drop duplicate and invalid vlans. */ |
4722 | n_vlans = mirror_collect_vlans(m, cfg, &vlans); | |
37e7f427 | 4723 | } |
064af421 BP |
4724 | |
4725 | /* Update mirror data. */ | |
37e7f427 BP |
4726 | if (!shash_equal_keys(&m->src_ports, &src_ports) |
4727 | || !shash_equal_keys(&m->dst_ports, &dst_ports) | |
064af421 BP |
4728 | || m->n_vlans != n_vlans |
4729 | || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans) | |
4730 | || m->out_port != out_port | |
4731 | || m->out_vlan != out_vlan) { | |
4732 | bridge_flush(m->bridge); | |
4733 | } | |
37e7f427 BP |
4734 | shash_swap(&m->src_ports, &src_ports); |
4735 | shash_swap(&m->dst_ports, &dst_ports); | |
064af421 BP |
4736 | free(m->vlans); |
4737 | m->vlans = vlans; | |
4738 | m->n_vlans = n_vlans; | |
4739 | m->out_port = out_port; | |
4740 | m->out_vlan = out_vlan; | |
4741 | ||
064af421 BP |
4742 | /* Update ports. */ |
4743 | mirror_bit = MIRROR_MASK_C(1) << m->idx; | |
4744 | for (i = 0; i < m->bridge->n_ports; i++) { | |
4745 | struct port *port = m->bridge->ports[i]; | |
4746 | ||
939ff267 | 4747 | if (shash_find(&m->src_ports, port->name) |
064af421 BP |
4748 | || (m->n_vlans |
4749 | && (!port->vlan | |
4750 | ? port_trunks_any_mirrored_vlan(m, port) | |
4751 | : vlan_is_mirrored(m, port->vlan)))) { | |
4752 | port->src_mirrors |= mirror_bit; | |
4753 | } else { | |
4754 | port->src_mirrors &= ~mirror_bit; | |
4755 | } | |
4756 | ||
939ff267 | 4757 | if (shash_find(&m->dst_ports, port->name)) { |
064af421 BP |
4758 | port->dst_mirrors |= mirror_bit; |
4759 | } else { | |
4760 | port->dst_mirrors &= ~mirror_bit; | |
4761 | } | |
4762 | } | |
4763 | ||
4764 | /* Clean up. */ | |
37e7f427 BP |
4765 | shash_destroy(&src_ports); |
4766 | shash_destroy(&dst_ports); | |
064af421 | 4767 | } |