]>
Commit | Line | Data |
---|---|---|
e9e28be3 | 1 | /* |
149f577a | 2 | * Copyright (c) 2009, 2010 Nicira Networks. |
e9e28be3 BP |
3 | * |
4 | * Licensed under the Apache License, Version 2.0 (the "License"); | |
5 | * you may not use this file except in compliance with the License. | |
6 | * You may obtain a copy of the License at: | |
7 | * | |
8 | * http://www.apache.org/licenses/LICENSE-2.0 | |
9 | * | |
10 | * Unless required by applicable law or agreed to in writing, software | |
11 | * distributed under the License is distributed on an "AS IS" BASIS, | |
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
13 | * See the License for the specific language governing permissions and | |
14 | * limitations under the License. | |
15 | */ | |
16 | ||
17 | #include <config.h> | |
8b61709d | 18 | #include <assert.h> |
e9e28be3 | 19 | #include <errno.h> |
8b61709d BP |
20 | #include <fcntl.h> |
21 | #include <arpa/inet.h> | |
22 | #include <inttypes.h> | |
23 | #include <linux/if_tun.h> | |
a740f0de | 24 | #include <linux/ip.h> |
8b61709d BP |
25 | #include <linux/types.h> |
26 | #include <linux/ethtool.h> | |
6f42c8ea | 27 | #include <linux/pkt_sched.h> |
e9e28be3 | 28 | #include <linux/rtnetlink.h> |
8b61709d BP |
29 | #include <linux/sockios.h> |
30 | #include <linux/version.h> | |
31 | #include <sys/types.h> | |
32 | #include <sys/ioctl.h> | |
33 | #include <sys/socket.h> | |
34 | #include <netpacket/packet.h> | |
35 | #include <net/ethernet.h> | |
36 | #include <net/if.h> | |
a740f0de | 37 | #include <linux/if_tunnel.h> |
8b61709d BP |
38 | #include <net/if_arp.h> |
39 | #include <net/if_packet.h> | |
40 | #include <net/route.h> | |
41 | #include <netinet/in.h> | |
e9e28be3 | 42 | #include <poll.h> |
8b61709d BP |
43 | #include <stdlib.h> |
44 | #include <string.h> | |
45 | #include <unistd.h> | |
e9e28be3 BP |
46 | |
47 | #include "coverage.h" | |
8b61709d BP |
48 | #include "dynamic-string.h" |
49 | #include "fatal-signal.h" | |
50 | #include "netdev-provider.h" | |
e9e28be3 BP |
51 | #include "netlink.h" |
52 | #include "ofpbuf.h" | |
8b61709d | 53 | #include "openflow/openflow.h" |
8722022c BP |
54 | #include "openvswitch/internal_dev.h" |
55 | #include "openvswitch/gre.h" | |
8b61709d BP |
56 | #include "packets.h" |
57 | #include "poll-loop.h" | |
559843ed | 58 | #include "rtnetlink.h" |
8b61709d BP |
59 | #include "socket-util.h" |
60 | #include "shash.h" | |
61 | #include "svec.h" | |
e9e28be3 BP |
62 | |
63 | #define THIS_MODULE VLM_netdev_linux | |
64 | #include "vlog.h" | |
8b61709d BP |
65 | \f |
66 | /* These were introduced in Linux 2.6.14, so they might be missing if we have | |
67 | * old headers. */ | |
68 | #ifndef ADVERTISED_Pause | |
69 | #define ADVERTISED_Pause (1 << 13) | |
70 | #endif | |
71 | #ifndef ADVERTISED_Asym_Pause | |
72 | #define ADVERTISED_Asym_Pause (1 << 14) | |
73 | #endif | |
74 | ||
149f577a | 75 | static struct rtnetlink_notifier netdev_linux_cache_notifier; |
46415c90 | 76 | static int cache_notifier_refcount; |
8b61709d BP |
77 | |
78 | enum { | |
79 | VALID_IFINDEX = 1 << 0, | |
80 | VALID_ETHERADDR = 1 << 1, | |
81 | VALID_IN4 = 1 << 2, | |
82 | VALID_IN6 = 1 << 3, | |
83 | VALID_MTU = 1 << 4, | |
fe6b0e03 | 84 | VALID_CARRIER = 1 << 5, |
80a86fbe BP |
85 | VALID_IS_PSEUDO = 1 << 6, /* Represents is_internal and is_tap. */ |
86 | VALID_POLICING = 1 << 7 | |
8b61709d BP |
87 | }; |
88 | ||
149f577a JG |
89 | struct tap_state { |
90 | int fd; | |
91 | }; | |
92 | ||
93 | struct netdev_dev_linux { | |
94 | struct netdev_dev netdev_dev; | |
95 | ||
8b61709d | 96 | struct shash_node *shash_node; |
149f577a | 97 | unsigned int cache_valid; |
8b61709d | 98 | |
8722022c BP |
99 | /* The following are figured out "on demand" only. They are only valid |
100 | * when the corresponding VALID_* bit in 'cache_valid' is set. */ | |
8b61709d BP |
101 | int ifindex; |
102 | uint8_t etheraddr[ETH_ADDR_LEN]; | |
f1acd62b | 103 | struct in_addr address, netmask; |
8b61709d BP |
104 | struct in6_addr in6; |
105 | int mtu; | |
106 | int carrier; | |
8722022c BP |
107 | bool is_internal; /* Is this an openvswitch internal device? */ |
108 | bool is_tap; /* Is this a tuntap device? */ | |
80a86fbe BP |
109 | uint32_t kbits_rate; /* Policing data. */ |
110 | uint32_t kbits_burst; | |
149f577a JG |
111 | |
112 | union { | |
113 | struct tap_state tap; | |
114 | } state; | |
8b61709d BP |
115 | }; |
116 | ||
149f577a JG |
117 | struct netdev_linux { |
118 | struct netdev netdev; | |
5b7448ed | 119 | int fd; |
149f577a | 120 | }; |
8b61709d | 121 | |
8b61709d BP |
122 | /* An AF_INET socket (used for ioctl operations). */ |
123 | static int af_inet_sock = -1; | |
124 | ||
125 | struct netdev_linux_notifier { | |
126 | struct netdev_notifier notifier; | |
127 | struct list node; | |
128 | }; | |
129 | ||
130 | static struct shash netdev_linux_notifiers = | |
131 | SHASH_INITIALIZER(&netdev_linux_notifiers); | |
46097491 | 132 | static struct rtnetlink_notifier netdev_linux_poll_notifier; |
8b61709d BP |
133 | |
134 | /* This is set pretty low because we probably won't learn anything from the | |
135 | * additional log messages. */ | |
136 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); | |
137 | ||
15b3596a | 138 | static int netdev_linux_init(void); |
6f643e49 | 139 | |
0b0544d7 | 140 | static int netdev_linux_do_ethtool(const char *name, struct ethtool_cmd *, |
8b61709d | 141 | int cmd, const char *cmd_name); |
149f577a JG |
142 | static int netdev_linux_do_ioctl(const char *name, struct ifreq *, int cmd, |
143 | const char *cmd_name); | |
f1acd62b BP |
144 | static int netdev_linux_get_ipv4(const struct netdev *, struct in_addr *, |
145 | int cmd, const char *cmd_name); | |
8b61709d BP |
146 | static int get_flags(const struct netdev *, int *flagsp); |
147 | static int set_flags(struct netdev *, int flags); | |
148 | static int do_get_ifindex(const char *netdev_name); | |
149 | static int get_ifindex(const struct netdev *, int *ifindexp); | |
150 | static int do_set_addr(struct netdev *netdev, | |
151 | int ioctl_nr, const char *ioctl_name, | |
152 | struct in_addr addr); | |
153 | static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]); | |
154 | static int set_etheraddr(const char *netdev_name, int hwaddr_family, | |
155 | const uint8_t[ETH_ADDR_LEN]); | |
156 | static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats); | |
157 | static int get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats); | |
a5af30fb | 158 | static int get_rtnl_sock(struct nl_sock **); |
8b61709d | 159 | |
15b3596a JG |
160 | static bool |
161 | is_netdev_linux_class(const struct netdev_class *netdev_class) | |
162 | { | |
163 | return netdev_class->init == netdev_linux_init; | |
164 | } | |
165 | ||
149f577a JG |
166 | static struct netdev_dev_linux * |
167 | netdev_dev_linux_cast(const struct netdev_dev *netdev_dev) | |
6c88d577 | 168 | { |
15b3596a JG |
169 | const struct netdev_class *netdev_class = netdev_dev_get_class(netdev_dev); |
170 | assert(is_netdev_linux_class(netdev_class)); | |
171 | ||
149f577a | 172 | return CONTAINER_OF(netdev_dev, struct netdev_dev_linux, netdev_dev); |
6c88d577 JP |
173 | } |
174 | ||
8b61709d BP |
175 | static struct netdev_linux * |
176 | netdev_linux_cast(const struct netdev *netdev) | |
177 | { | |
15b3596a JG |
178 | struct netdev_dev *netdev_dev = netdev_get_dev(netdev); |
179 | const struct netdev_class *netdev_class = netdev_dev_get_class(netdev_dev); | |
180 | assert(is_netdev_linux_class(netdev_class)); | |
181 | ||
8b61709d BP |
182 | return CONTAINER_OF(netdev, struct netdev_linux, netdev); |
183 | } | |
184 | ||
185 | static int | |
186 | netdev_linux_init(void) | |
187 | { | |
188 | static int status = -1; | |
189 | if (status < 0) { | |
190 | af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0); | |
191 | status = af_inet_sock >= 0 ? 0 : errno; | |
192 | if (status) { | |
193 | VLOG_ERR("failed to create inet socket: %s", strerror(status)); | |
194 | } | |
195 | } | |
196 | return status; | |
197 | } | |
198 | ||
199 | static void | |
200 | netdev_linux_run(void) | |
201 | { | |
46097491 | 202 | rtnetlink_notifier_run(); |
8b61709d BP |
203 | } |
204 | ||
205 | static void | |
206 | netdev_linux_wait(void) | |
207 | { | |
46097491 | 208 | rtnetlink_notifier_wait(); |
8b61709d BP |
209 | } |
210 | ||
211 | static void | |
46097491 | 212 | netdev_linux_cache_cb(const struct rtnetlink_change *change, |
67a4917b | 213 | void *aux OVS_UNUSED) |
8b61709d | 214 | { |
149f577a | 215 | struct netdev_dev_linux *dev; |
8b61709d | 216 | if (change) { |
46415c90 JG |
217 | struct netdev_dev *base_dev = netdev_dev_from_name(change->ifname); |
218 | if (base_dev) { | |
15b3596a JG |
219 | const struct netdev_class *netdev_class = |
220 | netdev_dev_get_class(base_dev); | |
221 | ||
222 | if (is_netdev_linux_class(netdev_class)) { | |
223 | dev = netdev_dev_linux_cast(base_dev); | |
224 | dev->cache_valid = 0; | |
225 | } | |
8b61709d BP |
226 | } |
227 | } else { | |
46415c90 | 228 | struct shash device_shash; |
8b61709d | 229 | struct shash_node *node; |
46415c90 JG |
230 | |
231 | shash_init(&device_shash); | |
232 | netdev_dev_get_devices(&netdev_linux_class, &device_shash); | |
233 | SHASH_FOR_EACH (node, &device_shash) { | |
149f577a JG |
234 | dev = node->data; |
235 | dev->cache_valid = 0; | |
8b61709d | 236 | } |
46415c90 | 237 | shash_destroy(&device_shash); |
8b61709d BP |
238 | } |
239 | } | |
240 | ||
149f577a | 241 | /* Creates the netdev device of 'type' with 'name'. */ |
8b61709d | 242 | static int |
c69ee87c | 243 | netdev_linux_create_system(const char *name, const char *type OVS_UNUSED, |
149f577a | 244 | const struct shash *args, struct netdev_dev **netdev_devp) |
6c88d577 | 245 | { |
149f577a JG |
246 | struct netdev_dev_linux *netdev_dev; |
247 | int error; | |
6c88d577 JP |
248 | |
249 | if (!shash_is_empty(args)) { | |
149f577a | 250 | VLOG_WARN("%s: arguments for system devices should be empty", name); |
6c88d577 JP |
251 | } |
252 | ||
46415c90 | 253 | if (!cache_notifier_refcount) { |
149f577a JG |
254 | error = rtnetlink_notifier_register(&netdev_linux_cache_notifier, |
255 | netdev_linux_cache_cb, NULL); | |
256 | if (error) { | |
257 | return error; | |
258 | } | |
259 | } | |
46415c90 | 260 | cache_notifier_refcount++; |
6c88d577 | 261 | |
149f577a | 262 | netdev_dev = xzalloc(sizeof *netdev_dev); |
149f577a | 263 | netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_linux_class); |
46415c90 | 264 | |
149f577a | 265 | *netdev_devp = &netdev_dev->netdev_dev; |
a740f0de JG |
266 | return 0; |
267 | } | |
268 | ||
5b7448ed JG |
269 | /* For most types of netdevs we open the device for each call of |
270 | * netdev_open(). However, this is not the case with tap devices, | |
271 | * since it is only possible to open the device once. In this | |
272 | * situation we share a single file descriptor, and consequently | |
273 | * buffers, across all readers. Therefore once data is read it will | |
274 | * be unavailable to other reads for tap devices. */ | |
a740f0de | 275 | static int |
c69ee87c | 276 | netdev_linux_create_tap(const char *name, const char *type OVS_UNUSED, |
149f577a | 277 | const struct shash *args, struct netdev_dev **netdev_devp) |
a740f0de | 278 | { |
149f577a | 279 | struct netdev_dev_linux *netdev_dev; |
a740f0de JG |
280 | struct tap_state *state; |
281 | static const char tap_dev[] = "/dev/net/tun"; | |
282 | struct ifreq ifr; | |
283 | int error; | |
284 | ||
285 | if (!shash_is_empty(args)) { | |
149f577a | 286 | VLOG_WARN("%s: arguments for TAP devices should be empty", name); |
6c88d577 JP |
287 | } |
288 | ||
149f577a JG |
289 | netdev_dev = xzalloc(sizeof *netdev_dev); |
290 | state = &netdev_dev->state.tap; | |
a740f0de | 291 | |
6c88d577 | 292 | /* Open tap device. */ |
149f577a JG |
293 | state->fd = open(tap_dev, O_RDWR); |
294 | if (state->fd < 0) { | |
6c88d577 JP |
295 | error = errno; |
296 | VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error)); | |
297 | goto error; | |
298 | } | |
299 | ||
300 | /* Create tap device. */ | |
301 | ifr.ifr_flags = IFF_TAP | IFF_NO_PI; | |
302 | strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name); | |
149f577a | 303 | if (ioctl(state->fd, TUNSETIFF, &ifr) == -1) { |
6c88d577 JP |
304 | VLOG_WARN("%s: creating tap device failed: %s", name, |
305 | strerror(errno)); | |
306 | error = errno; | |
307 | goto error; | |
308 | } | |
309 | ||
310 | /* Make non-blocking. */ | |
149f577a | 311 | error = set_nonblocking(state->fd); |
a740f0de JG |
312 | if (error) { |
313 | goto error; | |
314 | } | |
315 | ||
149f577a JG |
316 | netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_tap_class); |
317 | *netdev_devp = &netdev_dev->netdev_dev; | |
a740f0de JG |
318 | return 0; |
319 | ||
320 | error: | |
149f577a | 321 | free(netdev_dev); |
a740f0de JG |
322 | return error; |
323 | } | |
324 | ||
a740f0de | 325 | static void |
149f577a | 326 | destroy_tap(struct netdev_dev_linux *netdev_dev) |
a740f0de | 327 | { |
149f577a JG |
328 | struct tap_state *state = &netdev_dev->state.tap; |
329 | ||
330 | if (state->fd >= 0) { | |
331 | close(state->fd); | |
a740f0de JG |
332 | } |
333 | } | |
334 | ||
149f577a | 335 | /* Destroys the netdev device 'netdev_dev_'. */ |
6c88d577 | 336 | static void |
149f577a | 337 | netdev_linux_destroy(struct netdev_dev *netdev_dev_) |
6c88d577 | 338 | { |
149f577a JG |
339 | struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_dev_); |
340 | const char *type = netdev_dev_get_type(netdev_dev_); | |
6c88d577 | 341 | |
149f577a | 342 | if (!strcmp(type, "system")) { |
46415c90 | 343 | cache_notifier_refcount--; |
149f577a | 344 | |
46415c90 | 345 | if (!cache_notifier_refcount) { |
149f577a JG |
346 | rtnetlink_notifier_unregister(&netdev_linux_cache_notifier); |
347 | } | |
348 | } else if (!strcmp(type, "tap")) { | |
349 | destroy_tap(netdev_dev); | |
6c88d577 | 350 | } |
149f577a | 351 | |
658797c8 | 352 | free(netdev_dev); |
6c88d577 JP |
353 | } |
354 | ||
8b61709d | 355 | static int |
5b7448ed | 356 | netdev_linux_open(struct netdev_dev *netdev_dev_, int ethertype, |
149f577a | 357 | struct netdev **netdevp) |
8b61709d | 358 | { |
5b7448ed | 359 | struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_dev_); |
8b61709d BP |
360 | struct netdev_linux *netdev; |
361 | enum netdev_flags flags; | |
362 | int error; | |
363 | ||
364 | /* Allocate network device. */ | |
ec6fde61 | 365 | netdev = xzalloc(sizeof *netdev); |
49a6a163 | 366 | netdev->fd = -1; |
5b7448ed | 367 | netdev_init(&netdev->netdev, netdev_dev_); |
8b61709d BP |
368 | |
369 | error = netdev_get_flags(&netdev->netdev, &flags); | |
370 | if (error == ENODEV) { | |
371 | goto error; | |
372 | } | |
373 | ||
5b7448ed JG |
374 | if (!strcmp(netdev_dev_get_type(netdev_dev_), "tap")) { |
375 | netdev->fd = netdev_dev->state.tap.fd; | |
5b7448ed | 376 | } else if (ethertype != NETDEV_ETH_TYPE_NONE) { |
8b61709d BP |
377 | struct sockaddr_ll sll; |
378 | int protocol; | |
379 | int ifindex; | |
380 | ||
381 | /* Create file descriptor. */ | |
382 | protocol = (ethertype == NETDEV_ETH_TYPE_ANY ? ETH_P_ALL | |
383 | : ethertype == NETDEV_ETH_TYPE_802_2 ? ETH_P_802_2 | |
384 | : ethertype); | |
5b7448ed JG |
385 | netdev->fd = socket(PF_PACKET, SOCK_RAW, htons(protocol)); |
386 | if (netdev->fd < 0) { | |
8b61709d BP |
387 | error = errno; |
388 | goto error; | |
389 | } | |
8b61709d BP |
390 | |
391 | /* Set non-blocking mode. */ | |
5b7448ed | 392 | error = set_nonblocking(netdev->fd); |
8b61709d BP |
393 | if (error) { |
394 | goto error; | |
395 | } | |
396 | ||
397 | /* Get ethernet device index. */ | |
398 | error = get_ifindex(&netdev->netdev, &ifindex); | |
399 | if (error) { | |
400 | goto error; | |
401 | } | |
402 | ||
403 | /* Bind to specific ethernet device. */ | |
404 | memset(&sll, 0, sizeof sll); | |
405 | sll.sll_family = AF_PACKET; | |
406 | sll.sll_ifindex = ifindex; | |
5b7448ed | 407 | if (bind(netdev->fd, |
8b61709d BP |
408 | (struct sockaddr *) &sll, sizeof sll) < 0) { |
409 | error = errno; | |
5b7448ed | 410 | VLOG_ERR("bind to %s failed: %s", netdev_dev_get_name(netdev_dev_), |
149f577a | 411 | strerror(error)); |
8b61709d BP |
412 | goto error; |
413 | } | |
414 | ||
415 | /* Between the socket() and bind() calls above, the socket receives all | |
416 | * packets of the requested type on all system interfaces. We do not | |
417 | * want to receive that data, but there is no way to avoid it. So we | |
418 | * must now drain out the receive queue. */ | |
5b7448ed | 419 | error = drain_rcvbuf(netdev->fd); |
8b61709d BP |
420 | if (error) { |
421 | goto error; | |
422 | } | |
423 | } | |
424 | ||
425 | *netdevp = &netdev->netdev; | |
426 | return 0; | |
427 | ||
428 | error: | |
149f577a | 429 | netdev_uninit(&netdev->netdev, true); |
8b61709d BP |
430 | return error; |
431 | } | |
432 | ||
433 | /* Closes and destroys 'netdev'. */ | |
434 | static void | |
435 | netdev_linux_close(struct netdev *netdev_) | |
436 | { | |
437 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
438 | ||
49a6a163 | 439 | if (netdev->fd > 0 && strcmp(netdev_get_type(netdev_), "tap")) { |
5b7448ed | 440 | close(netdev->fd); |
8b61709d BP |
441 | } |
442 | free(netdev); | |
443 | } | |
e9e28be3 | 444 | |
8b61709d BP |
445 | /* Initializes 'svec' with a list of the names of all known network devices. */ |
446 | static int | |
447 | netdev_linux_enumerate(struct svec *svec) | |
448 | { | |
449 | struct if_nameindex *names; | |
450 | ||
451 | names = if_nameindex(); | |
452 | if (names) { | |
453 | size_t i; | |
454 | ||
455 | for (i = 0; names[i].if_name != NULL; i++) { | |
456 | svec_add(svec, names[i].if_name); | |
457 | } | |
458 | if_freenameindex(names); | |
459 | return 0; | |
460 | } else { | |
461 | VLOG_WARN("could not obtain list of network device names: %s", | |
462 | strerror(errno)); | |
463 | return errno; | |
464 | } | |
465 | } | |
466 | ||
467 | static int | |
468 | netdev_linux_recv(struct netdev *netdev_, void *data, size_t size) | |
469 | { | |
470 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
471 | ||
5b7448ed | 472 | if (netdev->fd < 0) { |
8b61709d | 473 | /* Device was opened with NETDEV_ETH_TYPE_NONE. */ |
c0e5f6ca | 474 | return -EAGAIN; |
8b61709d BP |
475 | } |
476 | ||
477 | for (;;) { | |
5b7448ed | 478 | ssize_t retval = read(netdev->fd, data, size); |
8b61709d BP |
479 | if (retval >= 0) { |
480 | return retval; | |
481 | } else if (errno != EINTR) { | |
482 | if (errno != EAGAIN) { | |
483 | VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s", | |
484 | strerror(errno), netdev_get_name(netdev_)); | |
485 | } | |
c0e5f6ca | 486 | return -errno; |
8b61709d BP |
487 | } |
488 | } | |
489 | } | |
490 | ||
491 | /* Registers with the poll loop to wake up from the next call to poll_block() | |
492 | * when a packet is ready to be received with netdev_recv() on 'netdev'. */ | |
493 | static void | |
494 | netdev_linux_recv_wait(struct netdev *netdev_) | |
495 | { | |
496 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
5b7448ed JG |
497 | if (netdev->fd >= 0) { |
498 | poll_fd_wait(netdev->fd, POLLIN); | |
8b61709d BP |
499 | } |
500 | } | |
501 | ||
502 | /* Discards all packets waiting to be received from 'netdev'. */ | |
503 | static int | |
504 | netdev_linux_drain(struct netdev *netdev_) | |
505 | { | |
506 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
5b7448ed | 507 | if (netdev->fd < 0) { |
8b61709d | 508 | return 0; |
5b7448ed | 509 | } else if (!strcmp(netdev_get_type(netdev_), "tap")) { |
8b61709d | 510 | struct ifreq ifr; |
149f577a | 511 | int error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, |
8b61709d BP |
512 | SIOCGIFTXQLEN, "SIOCGIFTXQLEN"); |
513 | if (error) { | |
514 | return error; | |
515 | } | |
5b7448ed | 516 | drain_fd(netdev->fd, ifr.ifr_qlen); |
8b61709d BP |
517 | return 0; |
518 | } else { | |
5b7448ed | 519 | return drain_rcvbuf(netdev->fd); |
8b61709d BP |
520 | } |
521 | } | |
522 | ||
523 | /* Sends 'buffer' on 'netdev'. Returns 0 if successful, otherwise a positive | |
524 | * errno value. Returns EAGAIN without blocking if the packet cannot be queued | |
525 | * immediately. Returns EMSGSIZE if a partial packet was transmitted or if | |
526 | * the packet is too big or too small to transmit on the device. | |
527 | * | |
528 | * The caller retains ownership of 'buffer' in all cases. | |
529 | * | |
530 | * The kernel maintains a packet transmission queue, so the caller is not | |
531 | * expected to do additional queuing of packets. */ | |
532 | static int | |
533 | netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) | |
534 | { | |
535 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
536 | ||
537 | /* XXX should support sending even if 'ethertype' was NETDEV_ETH_TYPE_NONE. | |
538 | */ | |
5b7448ed | 539 | if (netdev->fd < 0) { |
8b61709d BP |
540 | return EPIPE; |
541 | } | |
542 | ||
543 | for (;;) { | |
5b7448ed | 544 | ssize_t retval = write(netdev->fd, data, size); |
8b61709d BP |
545 | if (retval < 0) { |
546 | /* The Linux AF_PACKET implementation never blocks waiting for room | |
547 | * for packets, instead returning ENOBUFS. Translate this into | |
548 | * EAGAIN for the caller. */ | |
549 | if (errno == ENOBUFS) { | |
550 | return EAGAIN; | |
551 | } else if (errno == EINTR) { | |
552 | continue; | |
553 | } else if (errno != EAGAIN) { | |
554 | VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s", | |
555 | netdev_get_name(netdev_), strerror(errno)); | |
556 | } | |
557 | return errno; | |
558 | } else if (retval != size) { | |
559 | VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of " | |
560 | "%zu) on %s", retval, size, netdev_get_name(netdev_)); | |
561 | return EMSGSIZE; | |
562 | } else { | |
563 | return 0; | |
564 | } | |
565 | } | |
566 | } | |
567 | ||
568 | /* Registers with the poll loop to wake up from the next call to poll_block() | |
569 | * when the packet transmission queue has sufficient room to transmit a packet | |
570 | * with netdev_send(). | |
571 | * | |
572 | * The kernel maintains a packet transmission queue, so the client is not | |
573 | * expected to do additional queuing of packets. Thus, this function is | |
574 | * unlikely to ever be used. It is included for completeness. */ | |
575 | static void | |
576 | netdev_linux_send_wait(struct netdev *netdev_) | |
577 | { | |
578 | struct netdev_linux *netdev = netdev_linux_cast(netdev_); | |
5b7448ed | 579 | if (netdev->fd < 0) { |
8b61709d | 580 | /* Nothing to do. */ |
5b7448ed JG |
581 | } else if (strcmp(netdev_get_type(netdev_), "tap")) { |
582 | poll_fd_wait(netdev->fd, POLLOUT); | |
8b61709d BP |
583 | } else { |
584 | /* TAP device always accepts packets.*/ | |
585 | poll_immediate_wake(); | |
586 | } | |
587 | } | |
588 | ||
589 | /* Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful, | |
590 | * otherwise a positive errno value. */ | |
591 | static int | |
592 | netdev_linux_set_etheraddr(struct netdev *netdev_, | |
593 | const uint8_t mac[ETH_ADDR_LEN]) | |
594 | { | |
149f577a JG |
595 | struct netdev_dev_linux *netdev_dev = |
596 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
eb395f2e BP |
597 | int error; |
598 | ||
149f577a JG |
599 | if (!(netdev_dev->cache_valid & VALID_ETHERADDR) |
600 | || !eth_addr_equals(netdev_dev->etheraddr, mac)) { | |
eb395f2e BP |
601 | error = set_etheraddr(netdev_get_name(netdev_), ARPHRD_ETHER, mac); |
602 | if (!error) { | |
149f577a JG |
603 | netdev_dev->cache_valid |= VALID_ETHERADDR; |
604 | memcpy(netdev_dev->etheraddr, mac, ETH_ADDR_LEN); | |
eb395f2e BP |
605 | } |
606 | } else { | |
607 | error = 0; | |
8b61709d BP |
608 | } |
609 | return error; | |
610 | } | |
611 | ||
612 | /* Returns a pointer to 'netdev''s MAC address. The caller must not modify or | |
613 | * free the returned buffer. */ | |
614 | static int | |
615 | netdev_linux_get_etheraddr(const struct netdev *netdev_, | |
616 | uint8_t mac[ETH_ADDR_LEN]) | |
617 | { | |
149f577a JG |
618 | struct netdev_dev_linux *netdev_dev = |
619 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
620 | if (!(netdev_dev->cache_valid & VALID_ETHERADDR)) { | |
8b61709d | 621 | int error = get_etheraddr(netdev_get_name(netdev_), |
149f577a | 622 | netdev_dev->etheraddr); |
8b61709d BP |
623 | if (error) { |
624 | return error; | |
625 | } | |
149f577a | 626 | netdev_dev->cache_valid |= VALID_ETHERADDR; |
8b61709d | 627 | } |
149f577a | 628 | memcpy(mac, netdev_dev->etheraddr, ETH_ADDR_LEN); |
8b61709d BP |
629 | return 0; |
630 | } | |
631 | ||
632 | /* Returns the maximum size of transmitted (and received) packets on 'netdev', | |
633 | * in bytes, not including the hardware header; thus, this is typically 1500 | |
634 | * bytes for Ethernet devices. */ | |
635 | static int | |
636 | netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup) | |
637 | { | |
149f577a JG |
638 | struct netdev_dev_linux *netdev_dev = |
639 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
640 | if (!(netdev_dev->cache_valid & VALID_MTU)) { | |
8b61709d BP |
641 | struct ifreq ifr; |
642 | int error; | |
643 | ||
149f577a JG |
644 | error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, |
645 | SIOCGIFMTU, "SIOCGIFMTU"); | |
8b61709d BP |
646 | if (error) { |
647 | return error; | |
648 | } | |
149f577a JG |
649 | netdev_dev->mtu = ifr.ifr_mtu; |
650 | netdev_dev->cache_valid |= VALID_MTU; | |
8b61709d | 651 | } |
149f577a | 652 | *mtup = netdev_dev->mtu; |
8b61709d BP |
653 | return 0; |
654 | } | |
655 | ||
9ab3d9a3 BP |
656 | /* Returns the ifindex of 'netdev', if successful, as a positive number. |
657 | * On failure, returns a negative errno value. */ | |
658 | static int | |
659 | netdev_linux_get_ifindex(const struct netdev *netdev) | |
660 | { | |
661 | int ifindex, error; | |
662 | ||
663 | error = get_ifindex(netdev, &ifindex); | |
664 | return error ? -error : ifindex; | |
665 | } | |
666 | ||
8b61709d BP |
667 | static int |
668 | netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier) | |
669 | { | |
149f577a JG |
670 | struct netdev_dev_linux *netdev_dev = |
671 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
8b61709d BP |
672 | int error = 0; |
673 | char *fn = NULL; | |
674 | int fd = -1; | |
675 | ||
149f577a | 676 | if (!(netdev_dev->cache_valid & VALID_CARRIER)) { |
8b61709d BP |
677 | char line[8]; |
678 | int retval; | |
679 | ||
149f577a JG |
680 | fn = xasprintf("/sys/class/net/%s/carrier", |
681 | netdev_get_name(netdev_)); | |
8b61709d BP |
682 | fd = open(fn, O_RDONLY); |
683 | if (fd < 0) { | |
684 | error = errno; | |
685 | VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(error)); | |
686 | goto exit; | |
687 | } | |
688 | ||
689 | retval = read(fd, line, sizeof line); | |
690 | if (retval < 0) { | |
691 | error = errno; | |
692 | if (error == EINVAL) { | |
693 | /* This is the normal return value when we try to check carrier | |
694 | * if the network device is not up. */ | |
695 | } else { | |
696 | VLOG_WARN_RL(&rl, "%s: read failed: %s", fn, strerror(error)); | |
697 | } | |
698 | goto exit; | |
699 | } else if (retval == 0) { | |
700 | error = EPROTO; | |
701 | VLOG_WARN_RL(&rl, "%s: unexpected end of file", fn); | |
702 | goto exit; | |
703 | } | |
704 | ||
705 | if (line[0] != '0' && line[0] != '1') { | |
706 | error = EPROTO; | |
707 | VLOG_WARN_RL(&rl, "%s: value is %c (expected 0 or 1)", | |
708 | fn, line[0]); | |
709 | goto exit; | |
710 | } | |
149f577a JG |
711 | netdev_dev->carrier = line[0] != '0'; |
712 | netdev_dev->cache_valid |= VALID_CARRIER; | |
8b61709d | 713 | } |
149f577a | 714 | *carrier = netdev_dev->carrier; |
8b61709d BP |
715 | error = 0; |
716 | ||
717 | exit: | |
718 | if (fd >= 0) { | |
719 | close(fd); | |
720 | } | |
721 | free(fn); | |
722 | return error; | |
723 | } | |
724 | ||
725 | /* Check whether we can we use RTM_GETLINK to get network device statistics. | |
726 | * In pre-2.6.19 kernels, this was only available if wireless extensions were | |
727 | * enabled. */ | |
728 | static bool | |
729 | check_for_working_netlink_stats(void) | |
730 | { | |
731 | /* Decide on the netdev_get_stats() implementation to use. Netlink is | |
732 | * preferable, so if that works, we'll use it. */ | |
733 | int ifindex = do_get_ifindex("lo"); | |
734 | if (ifindex < 0) { | |
735 | VLOG_WARN("failed to get ifindex for lo, " | |
736 | "obtaining netdev stats from proc"); | |
737 | return false; | |
738 | } else { | |
739 | struct netdev_stats stats; | |
740 | int error = get_stats_via_netlink(ifindex, &stats); | |
741 | if (!error) { | |
742 | VLOG_DBG("obtaining netdev stats via rtnetlink"); | |
743 | return true; | |
744 | } else { | |
745 | VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats " | |
746 | "via proc (you are probably running a pre-2.6.19 " | |
747 | "kernel)", strerror(error)); | |
748 | return false; | |
749 | } | |
750 | } | |
751 | } | |
752 | ||
8722022c BP |
753 | /* Brings the 'is_internal' and 'is_tap' members of 'netdev_dev' up-to-date. */ |
754 | static void | |
755 | netdev_linux_update_is_pseudo(struct netdev_dev_linux *netdev_dev) | |
756 | { | |
757 | if (!(netdev_dev->cache_valid & VALID_IS_PSEUDO)) { | |
758 | const char *name = netdev_dev_get_name(&netdev_dev->netdev_dev); | |
759 | const char *type = netdev_dev_get_type(&netdev_dev->netdev_dev); | |
760 | ||
761 | netdev_dev->is_tap = !strcmp(type, "tap"); | |
762 | netdev_dev->is_internal = false; | |
763 | if (!netdev_dev->is_tap) { | |
764 | struct ethtool_drvinfo drvinfo; | |
765 | int error; | |
766 | ||
767 | memset(&drvinfo, 0, sizeof drvinfo); | |
768 | error = netdev_linux_do_ethtool(name, | |
769 | (struct ethtool_cmd *)&drvinfo, | |
770 | ETHTOOL_GDRVINFO, | |
771 | "ETHTOOL_GDRVINFO"); | |
772 | ||
773 | if (!error && !strcmp(drvinfo.driver, "openvswitch")) { | |
774 | netdev_dev->is_internal = true; | |
775 | } | |
776 | } | |
777 | ||
778 | netdev_dev->cache_valid |= VALID_IS_PSEUDO; | |
779 | } | |
780 | } | |
781 | ||
8b61709d BP |
782 | /* Retrieves current device stats for 'netdev'. |
783 | * | |
784 | * XXX All of the members of struct netdev_stats are 64 bits wide, but on | |
785 | * 32-bit architectures the Linux network stats are only 32 bits. */ | |
786 | static int | |
149f577a JG |
787 | netdev_linux_get_stats(const struct netdev *netdev_, |
788 | struct netdev_stats *stats) | |
8b61709d | 789 | { |
149f577a JG |
790 | struct netdev_dev_linux *netdev_dev = |
791 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
8b61709d BP |
792 | static int use_netlink_stats = -1; |
793 | int error; | |
fe6b0e03 JG |
794 | struct netdev_stats raw_stats; |
795 | struct netdev_stats *collect_stats = stats; | |
8b61709d BP |
796 | |
797 | COVERAGE_INC(netdev_get_stats); | |
fe6b0e03 | 798 | |
8722022c | 799 | netdev_linux_update_is_pseudo(netdev_dev); |
149f577a | 800 | if (netdev_dev->is_internal) { |
fe6b0e03 JG |
801 | collect_stats = &raw_stats; |
802 | } | |
803 | ||
8b61709d BP |
804 | if (use_netlink_stats < 0) { |
805 | use_netlink_stats = check_for_working_netlink_stats(); | |
806 | } | |
807 | if (use_netlink_stats) { | |
808 | int ifindex; | |
809 | ||
149f577a | 810 | error = get_ifindex(netdev_, &ifindex); |
8b61709d | 811 | if (!error) { |
fe6b0e03 | 812 | error = get_stats_via_netlink(ifindex, collect_stats); |
8b61709d BP |
813 | } |
814 | } else { | |
149f577a | 815 | error = get_stats_via_proc(netdev_get_name(netdev_), collect_stats); |
8b61709d | 816 | } |
fe6b0e03 JG |
817 | |
818 | /* If this port is an internal port then the transmit and receive stats | |
819 | * will appear to be swapped relative to the other ports since we are the | |
820 | * one sending the data, not a remote computer. For consistency, we swap | |
821 | * them back here. */ | |
8722022c | 822 | if (!error && (netdev_dev->is_internal || netdev_dev->is_tap)) { |
fe6b0e03 JG |
823 | stats->rx_packets = raw_stats.tx_packets; |
824 | stats->tx_packets = raw_stats.rx_packets; | |
825 | stats->rx_bytes = raw_stats.tx_bytes; | |
826 | stats->tx_bytes = raw_stats.rx_bytes; | |
827 | stats->rx_errors = raw_stats.tx_errors; | |
828 | stats->tx_errors = raw_stats.rx_errors; | |
829 | stats->rx_dropped = raw_stats.tx_dropped; | |
830 | stats->tx_dropped = raw_stats.rx_dropped; | |
831 | stats->multicast = raw_stats.multicast; | |
832 | stats->collisions = raw_stats.collisions; | |
833 | stats->rx_length_errors = 0; | |
834 | stats->rx_over_errors = 0; | |
835 | stats->rx_crc_errors = 0; | |
836 | stats->rx_frame_errors = 0; | |
837 | stats->rx_fifo_errors = 0; | |
838 | stats->rx_missed_errors = 0; | |
839 | stats->tx_aborted_errors = 0; | |
840 | stats->tx_carrier_errors = 0; | |
841 | stats->tx_fifo_errors = 0; | |
842 | stats->tx_heartbeat_errors = 0; | |
843 | stats->tx_window_errors = 0; | |
844 | } | |
845 | ||
8b61709d BP |
846 | return error; |
847 | } | |
848 | ||
8722022c BP |
849 | static int |
850 | netdev_linux_set_stats(struct netdev *netdev, | |
851 | const struct netdev_stats *stats) | |
852 | { | |
853 | struct netdev_dev_linux *netdev_dev = | |
854 | netdev_dev_linux_cast(netdev_get_dev(netdev)); | |
855 | struct internal_dev_stats dp_dev_stats; | |
856 | struct ifreq ifr; | |
857 | ||
858 | /* We must reject this call if 'netdev' is not an Open vSwitch internal | |
859 | * port, because the ioctl that we are about to execute is in the "device | |
860 | * private ioctls" range, which means that executing it on a device that | |
861 | * is not the type we expect could do any random thing. | |
862 | * | |
863 | * (Amusingly, these ioctl numbers are commented "THESE IOCTLS ARE | |
864 | * _DEPRECATED_ AND WILL DISAPPEAR IN 2.5.X" in linux/sockios.h. I guess | |
865 | * DaveM is a little behind on that.) */ | |
866 | netdev_linux_update_is_pseudo(netdev_dev); | |
867 | if (!netdev_dev->is_internal) { | |
868 | return EOPNOTSUPP; | |
869 | } | |
870 | ||
871 | /* This actually only sets the *offset* that the dp_dev applies, but in our | |
872 | * usage for fake bond devices the dp_dev never has any traffic of it own | |
873 | * so it has the same effect. */ | |
874 | dp_dev_stats.rx_packets = stats->rx_packets; | |
875 | dp_dev_stats.rx_bytes = stats->rx_bytes; | |
876 | dp_dev_stats.tx_packets = stats->tx_packets; | |
877 | dp_dev_stats.tx_bytes = stats->tx_bytes; | |
878 | ifr.ifr_data = (void *) &dp_dev_stats; | |
879 | return netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, | |
880 | INTERNAL_DEV_SET_STATS, | |
881 | "INTERNAL_DEV_SET_STATS"); | |
882 | } | |
883 | ||
8b61709d BP |
884 | /* Stores the features supported by 'netdev' into each of '*current', |
885 | * '*advertised', '*supported', and '*peer' that are non-null. Each value is a | |
886 | * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if | |
7671589a | 887 | * successful, otherwise a positive errno value. */ |
8b61709d BP |
888 | static int |
889 | netdev_linux_get_features(struct netdev *netdev, | |
890 | uint32_t *current, uint32_t *advertised, | |
891 | uint32_t *supported, uint32_t *peer) | |
892 | { | |
893 | struct ethtool_cmd ecmd; | |
894 | int error; | |
895 | ||
896 | memset(&ecmd, 0, sizeof ecmd); | |
0b0544d7 | 897 | error = netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, |
8b61709d BP |
898 | ETHTOOL_GSET, "ETHTOOL_GSET"); |
899 | if (error) { | |
900 | return error; | |
901 | } | |
902 | ||
903 | /* Supported features. */ | |
904 | *supported = 0; | |
905 | if (ecmd.supported & SUPPORTED_10baseT_Half) { | |
906 | *supported |= OFPPF_10MB_HD; | |
907 | } | |
908 | if (ecmd.supported & SUPPORTED_10baseT_Full) { | |
909 | *supported |= OFPPF_10MB_FD; | |
910 | } | |
911 | if (ecmd.supported & SUPPORTED_100baseT_Half) { | |
912 | *supported |= OFPPF_100MB_HD; | |
913 | } | |
914 | if (ecmd.supported & SUPPORTED_100baseT_Full) { | |
915 | *supported |= OFPPF_100MB_FD; | |
916 | } | |
917 | if (ecmd.supported & SUPPORTED_1000baseT_Half) { | |
918 | *supported |= OFPPF_1GB_HD; | |
919 | } | |
920 | if (ecmd.supported & SUPPORTED_1000baseT_Full) { | |
921 | *supported |= OFPPF_1GB_FD; | |
922 | } | |
923 | if (ecmd.supported & SUPPORTED_10000baseT_Full) { | |
924 | *supported |= OFPPF_10GB_FD; | |
925 | } | |
926 | if (ecmd.supported & SUPPORTED_TP) { | |
927 | *supported |= OFPPF_COPPER; | |
928 | } | |
929 | if (ecmd.supported & SUPPORTED_FIBRE) { | |
930 | *supported |= OFPPF_FIBER; | |
931 | } | |
932 | if (ecmd.supported & SUPPORTED_Autoneg) { | |
933 | *supported |= OFPPF_AUTONEG; | |
934 | } | |
935 | if (ecmd.supported & SUPPORTED_Pause) { | |
936 | *supported |= OFPPF_PAUSE; | |
937 | } | |
938 | if (ecmd.supported & SUPPORTED_Asym_Pause) { | |
939 | *supported |= OFPPF_PAUSE_ASYM; | |
940 | } | |
941 | ||
942 | /* Advertised features. */ | |
943 | *advertised = 0; | |
944 | if (ecmd.advertising & ADVERTISED_10baseT_Half) { | |
945 | *advertised |= OFPPF_10MB_HD; | |
946 | } | |
947 | if (ecmd.advertising & ADVERTISED_10baseT_Full) { | |
948 | *advertised |= OFPPF_10MB_FD; | |
949 | } | |
950 | if (ecmd.advertising & ADVERTISED_100baseT_Half) { | |
951 | *advertised |= OFPPF_100MB_HD; | |
952 | } | |
953 | if (ecmd.advertising & ADVERTISED_100baseT_Full) { | |
954 | *advertised |= OFPPF_100MB_FD; | |
955 | } | |
956 | if (ecmd.advertising & ADVERTISED_1000baseT_Half) { | |
957 | *advertised |= OFPPF_1GB_HD; | |
958 | } | |
959 | if (ecmd.advertising & ADVERTISED_1000baseT_Full) { | |
960 | *advertised |= OFPPF_1GB_FD; | |
961 | } | |
962 | if (ecmd.advertising & ADVERTISED_10000baseT_Full) { | |
963 | *advertised |= OFPPF_10GB_FD; | |
964 | } | |
965 | if (ecmd.advertising & ADVERTISED_TP) { | |
966 | *advertised |= OFPPF_COPPER; | |
967 | } | |
968 | if (ecmd.advertising & ADVERTISED_FIBRE) { | |
969 | *advertised |= OFPPF_FIBER; | |
970 | } | |
971 | if (ecmd.advertising & ADVERTISED_Autoneg) { | |
972 | *advertised |= OFPPF_AUTONEG; | |
973 | } | |
974 | if (ecmd.advertising & ADVERTISED_Pause) { | |
975 | *advertised |= OFPPF_PAUSE; | |
976 | } | |
977 | if (ecmd.advertising & ADVERTISED_Asym_Pause) { | |
978 | *advertised |= OFPPF_PAUSE_ASYM; | |
979 | } | |
980 | ||
981 | /* Current settings. */ | |
982 | if (ecmd.speed == SPEED_10) { | |
983 | *current = ecmd.duplex ? OFPPF_10MB_FD : OFPPF_10MB_HD; | |
984 | } else if (ecmd.speed == SPEED_100) { | |
985 | *current = ecmd.duplex ? OFPPF_100MB_FD : OFPPF_100MB_HD; | |
986 | } else if (ecmd.speed == SPEED_1000) { | |
987 | *current = ecmd.duplex ? OFPPF_1GB_FD : OFPPF_1GB_HD; | |
988 | } else if (ecmd.speed == SPEED_10000) { | |
989 | *current = OFPPF_10GB_FD; | |
990 | } else { | |
991 | *current = 0; | |
992 | } | |
993 | ||
994 | if (ecmd.port == PORT_TP) { | |
995 | *current |= OFPPF_COPPER; | |
996 | } else if (ecmd.port == PORT_FIBRE) { | |
997 | *current |= OFPPF_FIBER; | |
998 | } | |
999 | ||
1000 | if (ecmd.autoneg) { | |
1001 | *current |= OFPPF_AUTONEG; | |
1002 | } | |
1003 | ||
1004 | /* Peer advertisements. */ | |
1005 | *peer = 0; /* XXX */ | |
1006 | ||
1007 | return 0; | |
1008 | } | |
1009 | ||
1010 | /* Set the features advertised by 'netdev' to 'advertise'. */ | |
1011 | static int | |
1012 | netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise) | |
1013 | { | |
1014 | struct ethtool_cmd ecmd; | |
1015 | int error; | |
1016 | ||
1017 | memset(&ecmd, 0, sizeof ecmd); | |
0b0544d7 | 1018 | error = netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, |
8b61709d BP |
1019 | ETHTOOL_GSET, "ETHTOOL_GSET"); |
1020 | if (error) { | |
1021 | return error; | |
1022 | } | |
1023 | ||
1024 | ecmd.advertising = 0; | |
1025 | if (advertise & OFPPF_10MB_HD) { | |
1026 | ecmd.advertising |= ADVERTISED_10baseT_Half; | |
1027 | } | |
1028 | if (advertise & OFPPF_10MB_FD) { | |
1029 | ecmd.advertising |= ADVERTISED_10baseT_Full; | |
1030 | } | |
1031 | if (advertise & OFPPF_100MB_HD) { | |
1032 | ecmd.advertising |= ADVERTISED_100baseT_Half; | |
1033 | } | |
1034 | if (advertise & OFPPF_100MB_FD) { | |
1035 | ecmd.advertising |= ADVERTISED_100baseT_Full; | |
1036 | } | |
1037 | if (advertise & OFPPF_1GB_HD) { | |
1038 | ecmd.advertising |= ADVERTISED_1000baseT_Half; | |
1039 | } | |
1040 | if (advertise & OFPPF_1GB_FD) { | |
1041 | ecmd.advertising |= ADVERTISED_1000baseT_Full; | |
1042 | } | |
1043 | if (advertise & OFPPF_10GB_FD) { | |
1044 | ecmd.advertising |= ADVERTISED_10000baseT_Full; | |
1045 | } | |
1046 | if (advertise & OFPPF_COPPER) { | |
1047 | ecmd.advertising |= ADVERTISED_TP; | |
1048 | } | |
1049 | if (advertise & OFPPF_FIBER) { | |
1050 | ecmd.advertising |= ADVERTISED_FIBRE; | |
1051 | } | |
1052 | if (advertise & OFPPF_AUTONEG) { | |
1053 | ecmd.advertising |= ADVERTISED_Autoneg; | |
1054 | } | |
1055 | if (advertise & OFPPF_PAUSE) { | |
1056 | ecmd.advertising |= ADVERTISED_Pause; | |
1057 | } | |
1058 | if (advertise & OFPPF_PAUSE_ASYM) { | |
1059 | ecmd.advertising |= ADVERTISED_Asym_Pause; | |
1060 | } | |
0b0544d7 | 1061 | return netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, |
8b61709d BP |
1062 | ETHTOOL_SSET, "ETHTOOL_SSET"); |
1063 | } | |
1064 | ||
1065 | /* If 'netdev_name' is the name of a VLAN network device (e.g. one created with | |
1066 | * vconfig(8)), sets '*vlan_vid' to the VLAN VID associated with that device | |
1067 | * and returns 0. Otherwise returns a errno value (specifically ENOENT if | |
1068 | * 'netdev_name' is the name of a network device that is not a VLAN device) and | |
1069 | * sets '*vlan_vid' to -1. */ | |
1070 | static int | |
1071 | netdev_linux_get_vlan_vid(const struct netdev *netdev, int *vlan_vid) | |
1072 | { | |
1073 | const char *netdev_name = netdev_get_name(netdev); | |
1074 | struct ds line = DS_EMPTY_INITIALIZER; | |
1075 | FILE *stream = NULL; | |
1076 | int error; | |
1077 | char *fn; | |
1078 | ||
1079 | COVERAGE_INC(netdev_get_vlan_vid); | |
1080 | fn = xasprintf("/proc/net/vlan/%s", netdev_name); | |
1081 | stream = fopen(fn, "r"); | |
1082 | if (!stream) { | |
1083 | error = errno; | |
1084 | goto done; | |
1085 | } | |
1086 | ||
1087 | if (ds_get_line(&line, stream)) { | |
1088 | if (ferror(stream)) { | |
1089 | error = errno; | |
1090 | VLOG_ERR_RL(&rl, "error reading \"%s\": %s", fn, strerror(errno)); | |
1091 | } else { | |
1092 | error = EPROTO; | |
1093 | VLOG_ERR_RL(&rl, "unexpected end of file reading \"%s\"", fn); | |
1094 | } | |
1095 | goto done; | |
1096 | } | |
1097 | ||
1098 | if (!sscanf(ds_cstr(&line), "%*s VID: %d", vlan_vid)) { | |
1099 | error = EPROTO; | |
1100 | VLOG_ERR_RL(&rl, "parse error reading \"%s\" line 1: \"%s\"", | |
1101 | fn, ds_cstr(&line)); | |
1102 | goto done; | |
1103 | } | |
1104 | ||
1105 | error = 0; | |
1106 | ||
1107 | done: | |
1108 | free(fn); | |
1109 | if (stream) { | |
1110 | fclose(stream); | |
1111 | } | |
1112 | ds_destroy(&line); | |
1113 | if (error) { | |
1114 | *vlan_vid = -1; | |
1115 | } | |
1116 | return error; | |
1117 | } | |
1118 | ||
1119 | #define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress" | |
1120 | #define POLICE_CONFIG_CMD "/sbin/tc filter add dev %s parent ffff: protocol ip prio 50 u32 match ip src 0.0.0.0/0 police rate %dkbit burst %dk mtu 65535 drop flowid :1" | |
8b61709d | 1121 | |
8e460221 | 1122 | /* Remove ingress policing from 'netdev'. Returns 0 if successful, otherwise a |
6f42c8ea BP |
1123 | * positive errno value. |
1124 | * | |
1125 | * This function is equivalent to running | |
1126 | * /sbin/tc qdisc del dev %s handle ffff: ingress | |
1127 | * but it is much, much faster. | |
1128 | */ | |
8e460221 BP |
1129 | static int |
1130 | netdev_linux_remove_policing(struct netdev *netdev) | |
1131 | { | |
80a86fbe BP |
1132 | struct netdev_dev_linux *netdev_dev = |
1133 | netdev_dev_linux_cast(netdev_get_dev(netdev)); | |
8e460221 | 1134 | const char *netdev_name = netdev_get_name(netdev); |
8e460221 | 1135 | |
6f42c8ea BP |
1136 | struct ofpbuf request; |
1137 | struct ofpbuf *reply; | |
1138 | struct tcmsg *tcmsg; | |
1139 | struct nl_sock *rtnl_sock; | |
1140 | int ifindex; | |
1141 | int error; | |
1142 | ||
1143 | error = get_ifindex(netdev, &ifindex); | |
1144 | if (error) { | |
1145 | return error; | |
8e460221 | 1146 | } |
6f42c8ea BP |
1147 | |
1148 | error = get_rtnl_sock(&rtnl_sock); | |
1149 | if (error) { | |
1150 | return error; | |
1151 | } | |
1152 | ||
1153 | ofpbuf_init(&request, 0); | |
1154 | nl_msg_put_nlmsghdr(&request, rtnl_sock, sizeof *tcmsg, | |
1155 | RTM_DELQDISC, NLM_F_REQUEST); | |
1156 | tcmsg = ofpbuf_put_zeros(&request, sizeof *tcmsg); | |
1157 | tcmsg->tcm_family = AF_UNSPEC; | |
1158 | tcmsg->tcm_ifindex = ifindex; | |
1159 | tcmsg->tcm_handle = 0xffff0000; | |
1160 | tcmsg->tcm_parent = TC_H_INGRESS; | |
1161 | nl_msg_put_string(&request, TCA_KIND, "ingress"); | |
1162 | nl_msg_put_unspec(&request, TCA_OPTIONS, NULL, 0); | |
1163 | error = nl_sock_transact(rtnl_sock, &request, &reply); | |
1164 | ofpbuf_uninit(&request); | |
1165 | ofpbuf_delete(reply); | |
1166 | if (error && error != ENOENT) { | |
1167 | VLOG_WARN_RL(&rl, "%s: removing policing failed: %s", | |
1168 | netdev_name, strerror(error)); | |
1169 | return error; | |
1170 | } | |
1171 | ||
80a86fbe BP |
1172 | netdev_dev->kbits_rate = 0; |
1173 | netdev_dev->kbits_burst = 0; | |
1174 | netdev_dev->cache_valid |= VALID_POLICING; | |
8e460221 BP |
1175 | return 0; |
1176 | } | |
1177 | ||
8b61709d BP |
1178 | /* Attempts to set input rate limiting (policing) policy. */ |
1179 | static int | |
1180 | netdev_linux_set_policing(struct netdev *netdev, | |
1181 | uint32_t kbits_rate, uint32_t kbits_burst) | |
1182 | { | |
80a86fbe BP |
1183 | struct netdev_dev_linux *netdev_dev = |
1184 | netdev_dev_linux_cast(netdev_get_dev(netdev)); | |
8b61709d BP |
1185 | const char *netdev_name = netdev_get_name(netdev); |
1186 | char command[1024]; | |
1187 | ||
1188 | COVERAGE_INC(netdev_set_policing); | |
8e460221 | 1189 | |
80a86fbe BP |
1190 | kbits_burst = (!kbits_rate ? 0 /* Force to 0 if no rate specified. */ |
1191 | : !kbits_burst ? 1000 /* Default to 1000 kbits if 0. */ | |
1192 | : kbits_burst); /* Stick with user-specified value. */ | |
1193 | ||
1194 | if (netdev_dev->cache_valid & VALID_POLICING | |
1195 | && netdev_dev->kbits_rate == kbits_rate | |
1196 | && netdev_dev->kbits_burst == kbits_burst) { | |
1197 | /* Assume that settings haven't changed since we last set them. */ | |
1198 | return 0; | |
1199 | } | |
1200 | ||
8e460221 | 1201 | netdev_linux_remove_policing(netdev); |
8b61709d | 1202 | if (kbits_rate) { |
8b61709d BP |
1203 | snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name); |
1204 | if (system(command) != 0) { | |
1205 | VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name); | |
1206 | return -1; | |
1207 | } | |
1208 | ||
1209 | snprintf(command, sizeof(command), POLICE_CONFIG_CMD, netdev_name, | |
1210 | kbits_rate, kbits_burst); | |
1211 | if (system(command) != 0) { | |
1212 | VLOG_WARN_RL(&rl, "%s: problem configuring policing", | |
1213 | netdev_name); | |
1214 | return -1; | |
1215 | } | |
80a86fbe BP |
1216 | |
1217 | netdev_dev->kbits_rate = kbits_rate; | |
1218 | netdev_dev->kbits_burst = kbits_burst; | |
1219 | netdev_dev->cache_valid |= VALID_POLICING; | |
8b61709d BP |
1220 | } |
1221 | ||
1222 | return 0; | |
1223 | } | |
1224 | ||
8b61709d | 1225 | static int |
f1acd62b BP |
1226 | netdev_linux_get_in4(const struct netdev *netdev_, |
1227 | struct in_addr *address, struct in_addr *netmask) | |
8b61709d | 1228 | { |
149f577a JG |
1229 | struct netdev_dev_linux *netdev_dev = |
1230 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
1231 | ||
1232 | if (!(netdev_dev->cache_valid & VALID_IN4)) { | |
8b61709d BP |
1233 | int error; |
1234 | ||
149f577a | 1235 | error = netdev_linux_get_ipv4(netdev_, &netdev_dev->address, |
8b61709d BP |
1236 | SIOCGIFADDR, "SIOCGIFADDR"); |
1237 | if (error) { | |
1238 | return error; | |
1239 | } | |
1240 | ||
149f577a | 1241 | error = netdev_linux_get_ipv4(netdev_, &netdev_dev->netmask, |
f1acd62b BP |
1242 | SIOCGIFNETMASK, "SIOCGIFNETMASK"); |
1243 | if (error) { | |
1244 | return error; | |
1245 | } | |
1246 | ||
149f577a | 1247 | netdev_dev->cache_valid |= VALID_IN4; |
8b61709d | 1248 | } |
149f577a JG |
1249 | *address = netdev_dev->address; |
1250 | *netmask = netdev_dev->netmask; | |
f1acd62b | 1251 | return address->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0; |
8b61709d BP |
1252 | } |
1253 | ||
8b61709d | 1254 | static int |
f1acd62b BP |
1255 | netdev_linux_set_in4(struct netdev *netdev_, struct in_addr address, |
1256 | struct in_addr netmask) | |
8b61709d | 1257 | { |
149f577a JG |
1258 | struct netdev_dev_linux *netdev_dev = |
1259 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
8b61709d BP |
1260 | int error; |
1261 | ||
f1acd62b | 1262 | error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", address); |
8b61709d | 1263 | if (!error) { |
149f577a JG |
1264 | netdev_dev->cache_valid |= VALID_IN4; |
1265 | netdev_dev->address = address; | |
1266 | netdev_dev->netmask = netmask; | |
f1acd62b | 1267 | if (address.s_addr != INADDR_ANY) { |
8b61709d | 1268 | error = do_set_addr(netdev_, SIOCSIFNETMASK, |
f1acd62b | 1269 | "SIOCSIFNETMASK", netmask); |
8b61709d BP |
1270 | } |
1271 | } | |
1272 | return error; | |
1273 | } | |
1274 | ||
1275 | static bool | |
1276 | parse_if_inet6_line(const char *line, | |
1277 | struct in6_addr *in6, char ifname[16 + 1]) | |
1278 | { | |
1279 | uint8_t *s6 = in6->s6_addr; | |
1280 | #define X8 "%2"SCNx8 | |
1281 | return sscanf(line, | |
1282 | " "X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 | |
1283 | "%*x %*x %*x %*x %16s\n", | |
1284 | &s6[0], &s6[1], &s6[2], &s6[3], | |
1285 | &s6[4], &s6[5], &s6[6], &s6[7], | |
1286 | &s6[8], &s6[9], &s6[10], &s6[11], | |
1287 | &s6[12], &s6[13], &s6[14], &s6[15], | |
1288 | ifname) == 17; | |
1289 | } | |
1290 | ||
1291 | /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address (if | |
1292 | * 'in6' is non-null) and returns true. Otherwise, returns false. */ | |
1293 | static int | |
1294 | netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6) | |
1295 | { | |
149f577a JG |
1296 | struct netdev_dev_linux *netdev_dev = |
1297 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
1298 | if (!(netdev_dev->cache_valid & VALID_IN6)) { | |
8b61709d BP |
1299 | FILE *file; |
1300 | char line[128]; | |
1301 | ||
149f577a | 1302 | netdev_dev->in6 = in6addr_any; |
8b61709d BP |
1303 | |
1304 | file = fopen("/proc/net/if_inet6", "r"); | |
1305 | if (file != NULL) { | |
1306 | const char *name = netdev_get_name(netdev_); | |
1307 | while (fgets(line, sizeof line, file)) { | |
1308 | struct in6_addr in6; | |
1309 | char ifname[16 + 1]; | |
1310 | if (parse_if_inet6_line(line, &in6, ifname) | |
1311 | && !strcmp(name, ifname)) | |
1312 | { | |
149f577a | 1313 | netdev_dev->in6 = in6; |
8b61709d BP |
1314 | break; |
1315 | } | |
1316 | } | |
1317 | fclose(file); | |
1318 | } | |
149f577a | 1319 | netdev_dev->cache_valid |= VALID_IN6; |
8b61709d | 1320 | } |
149f577a | 1321 | *in6 = netdev_dev->in6; |
8b61709d BP |
1322 | return 0; |
1323 | } | |
1324 | ||
1325 | static void | |
1326 | make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr) | |
1327 | { | |
1328 | struct sockaddr_in sin; | |
1329 | memset(&sin, 0, sizeof sin); | |
1330 | sin.sin_family = AF_INET; | |
1331 | sin.sin_addr = addr; | |
1332 | sin.sin_port = 0; | |
1333 | ||
1334 | memset(sa, 0, sizeof *sa); | |
1335 | memcpy(sa, &sin, sizeof sin); | |
1336 | } | |
1337 | ||
1338 | static int | |
1339 | do_set_addr(struct netdev *netdev, | |
1340 | int ioctl_nr, const char *ioctl_name, struct in_addr addr) | |
1341 | { | |
1342 | struct ifreq ifr; | |
149f577a | 1343 | strncpy(ifr.ifr_name, netdev_get_name(netdev), sizeof ifr.ifr_name); |
8b61709d | 1344 | make_in4_sockaddr(&ifr.ifr_addr, addr); |
149f577a JG |
1345 | |
1346 | return netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, ioctl_nr, | |
1347 | ioctl_name); | |
8b61709d BP |
1348 | } |
1349 | ||
1350 | /* Adds 'router' as a default IP gateway. */ | |
1351 | static int | |
67a4917b | 1352 | netdev_linux_add_router(struct netdev *netdev OVS_UNUSED, struct in_addr router) |
8b61709d BP |
1353 | { |
1354 | struct in_addr any = { INADDR_ANY }; | |
1355 | struct rtentry rt; | |
1356 | int error; | |
1357 | ||
1358 | memset(&rt, 0, sizeof rt); | |
1359 | make_in4_sockaddr(&rt.rt_dst, any); | |
1360 | make_in4_sockaddr(&rt.rt_gateway, router); | |
1361 | make_in4_sockaddr(&rt.rt_genmask, any); | |
1362 | rt.rt_flags = RTF_UP | RTF_GATEWAY; | |
1363 | COVERAGE_INC(netdev_add_router); | |
1364 | error = ioctl(af_inet_sock, SIOCADDRT, &rt) < 0 ? errno : 0; | |
1365 | if (error) { | |
1366 | VLOG_WARN("ioctl(SIOCADDRT): %s", strerror(error)); | |
1367 | } | |
1368 | return error; | |
1369 | } | |
1370 | ||
f1acd62b BP |
1371 | static int |
1372 | netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop, | |
1373 | char **netdev_name) | |
1374 | { | |
1375 | static const char fn[] = "/proc/net/route"; | |
1376 | FILE *stream; | |
1377 | char line[256]; | |
1378 | int ln; | |
1379 | ||
1380 | *netdev_name = NULL; | |
1381 | stream = fopen(fn, "r"); | |
1382 | if (stream == NULL) { | |
1383 | VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno)); | |
1384 | return errno; | |
1385 | } | |
1386 | ||
1387 | ln = 0; | |
1388 | while (fgets(line, sizeof line, stream)) { | |
1389 | if (++ln >= 2) { | |
1390 | char iface[17]; | |
1391 | uint32_t dest, gateway, mask; | |
1392 | int refcnt, metric, mtu; | |
1393 | unsigned int flags, use, window, irtt; | |
1394 | ||
1395 | if (sscanf(line, | |
1396 | "%16s %"SCNx32" %"SCNx32" %04X %d %u %d %"SCNx32 | |
1397 | " %d %u %u\n", | |
1398 | iface, &dest, &gateway, &flags, &refcnt, | |
1399 | &use, &metric, &mask, &mtu, &window, &irtt) != 11) { | |
1400 | ||
1401 | VLOG_WARN_RL(&rl, "%s: could not parse line %d: %s", | |
1402 | fn, ln, line); | |
1403 | continue; | |
1404 | } | |
1405 | if (!(flags & RTF_UP)) { | |
1406 | /* Skip routes that aren't up. */ | |
1407 | continue; | |
1408 | } | |
1409 | ||
1410 | /* The output of 'dest', 'mask', and 'gateway' were given in | |
1411 | * network byte order, so we don't need need any endian | |
1412 | * conversions here. */ | |
1413 | if ((dest & mask) == (host->s_addr & mask)) { | |
1414 | if (!gateway) { | |
1415 | /* The host is directly reachable. */ | |
1416 | next_hop->s_addr = 0; | |
1417 | } else { | |
1418 | /* To reach the host, we must go through a gateway. */ | |
1419 | next_hop->s_addr = gateway; | |
1420 | } | |
1421 | *netdev_name = xstrdup(iface); | |
1422 | fclose(stream); | |
1423 | return 0; | |
1424 | } | |
1425 | } | |
1426 | } | |
1427 | ||
1428 | fclose(stream); | |
1429 | return ENXIO; | |
1430 | } | |
1431 | ||
8b61709d BP |
1432 | /* Looks up the ARP table entry for 'ip' on 'netdev'. If one exists and can be |
1433 | * successfully retrieved, it stores the corresponding MAC address in 'mac' and | |
1434 | * returns 0. Otherwise, it returns a positive errno value; in particular, | |
1435 | * ENXIO indicates that there is not ARP table entry for 'ip' on 'netdev'. */ | |
1436 | static int | |
1437 | netdev_linux_arp_lookup(const struct netdev *netdev, | |
1438 | uint32_t ip, uint8_t mac[ETH_ADDR_LEN]) | |
1439 | { | |
1440 | struct arpreq r; | |
c100e025 | 1441 | struct sockaddr_in sin; |
8b61709d BP |
1442 | int retval; |
1443 | ||
1444 | memset(&r, 0, sizeof r); | |
c100e025 BP |
1445 | sin.sin_family = AF_INET; |
1446 | sin.sin_addr.s_addr = ip; | |
1447 | sin.sin_port = 0; | |
1448 | memcpy(&r.arp_pa, &sin, sizeof sin); | |
8b61709d BP |
1449 | r.arp_ha.sa_family = ARPHRD_ETHER; |
1450 | r.arp_flags = 0; | |
149f577a | 1451 | strncpy(r.arp_dev, netdev_get_name(netdev), sizeof r.arp_dev); |
8b61709d BP |
1452 | COVERAGE_INC(netdev_arp_lookup); |
1453 | retval = ioctl(af_inet_sock, SIOCGARP, &r) < 0 ? errno : 0; | |
1454 | if (!retval) { | |
1455 | memcpy(mac, r.arp_ha.sa_data, ETH_ADDR_LEN); | |
1456 | } else if (retval != ENXIO) { | |
1457 | VLOG_WARN_RL(&rl, "%s: could not look up ARP entry for "IP_FMT": %s", | |
149f577a | 1458 | netdev_get_name(netdev), IP_ARGS(&ip), strerror(retval)); |
8b61709d BP |
1459 | } |
1460 | return retval; | |
1461 | } | |
1462 | ||
1463 | static int | |
1464 | nd_to_iff_flags(enum netdev_flags nd) | |
1465 | { | |
1466 | int iff = 0; | |
1467 | if (nd & NETDEV_UP) { | |
1468 | iff |= IFF_UP; | |
1469 | } | |
1470 | if (nd & NETDEV_PROMISC) { | |
1471 | iff |= IFF_PROMISC; | |
1472 | } | |
1473 | return iff; | |
1474 | } | |
1475 | ||
1476 | static int | |
1477 | iff_to_nd_flags(int iff) | |
1478 | { | |
1479 | enum netdev_flags nd = 0; | |
1480 | if (iff & IFF_UP) { | |
1481 | nd |= NETDEV_UP; | |
1482 | } | |
1483 | if (iff & IFF_PROMISC) { | |
1484 | nd |= NETDEV_PROMISC; | |
1485 | } | |
1486 | return nd; | |
1487 | } | |
1488 | ||
1489 | static int | |
1490 | netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off, | |
1491 | enum netdev_flags on, enum netdev_flags *old_flagsp) | |
1492 | { | |
1493 | int old_flags, new_flags; | |
1494 | int error; | |
1495 | ||
1496 | error = get_flags(netdev, &old_flags); | |
1497 | if (!error) { | |
1498 | *old_flagsp = iff_to_nd_flags(old_flags); | |
1499 | new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on); | |
1500 | if (new_flags != old_flags) { | |
1501 | error = set_flags(netdev, new_flags); | |
1502 | } | |
1503 | } | |
1504 | return error; | |
1505 | } | |
1506 | ||
1507 | static void | |
1508 | poll_notify(struct list *list) | |
1509 | { | |
1510 | struct netdev_linux_notifier *notifier; | |
1511 | LIST_FOR_EACH (notifier, struct netdev_linux_notifier, node, list) { | |
1512 | struct netdev_notifier *n = ¬ifier->notifier; | |
1513 | n->cb(n); | |
1514 | } | |
1515 | } | |
1516 | ||
1517 | static void | |
46097491 | 1518 | netdev_linux_poll_cb(const struct rtnetlink_change *change, |
67a4917b | 1519 | void *aux OVS_UNUSED) |
8b61709d BP |
1520 | { |
1521 | if (change) { | |
1522 | struct list *list = shash_find_data(&netdev_linux_notifiers, | |
1523 | change->ifname); | |
1524 | if (list) { | |
1525 | poll_notify(list); | |
1526 | } | |
1527 | } else { | |
1528 | struct shash_node *node; | |
1529 | SHASH_FOR_EACH (node, &netdev_linux_notifiers) { | |
1530 | poll_notify(node->data); | |
1531 | } | |
1532 | } | |
1533 | } | |
1534 | ||
1535 | static int | |
1536 | netdev_linux_poll_add(struct netdev *netdev, | |
1537 | void (*cb)(struct netdev_notifier *), void *aux, | |
1538 | struct netdev_notifier **notifierp) | |
1539 | { | |
1540 | const char *netdev_name = netdev_get_name(netdev); | |
1541 | struct netdev_linux_notifier *notifier; | |
1542 | struct list *list; | |
1543 | ||
1544 | if (shash_is_empty(&netdev_linux_notifiers)) { | |
46097491 | 1545 | int error = rtnetlink_notifier_register(&netdev_linux_poll_notifier, |
8b61709d BP |
1546 | netdev_linux_poll_cb, NULL); |
1547 | if (error) { | |
1548 | return error; | |
1549 | } | |
1550 | } | |
1551 | ||
1552 | list = shash_find_data(&netdev_linux_notifiers, netdev_name); | |
1553 | if (!list) { | |
1554 | list = xmalloc(sizeof *list); | |
1555 | list_init(list); | |
1556 | shash_add(&netdev_linux_notifiers, netdev_name, list); | |
1557 | } | |
1558 | ||
1559 | notifier = xmalloc(sizeof *notifier); | |
1560 | netdev_notifier_init(¬ifier->notifier, netdev, cb, aux); | |
1561 | list_push_back(list, ¬ifier->node); | |
1562 | *notifierp = ¬ifier->notifier; | |
1563 | return 0; | |
1564 | } | |
1565 | ||
1566 | static void | |
1567 | netdev_linux_poll_remove(struct netdev_notifier *notifier_) | |
1568 | { | |
1569 | struct netdev_linux_notifier *notifier = | |
1570 | CONTAINER_OF(notifier_, struct netdev_linux_notifier, notifier); | |
1571 | struct list *list; | |
1572 | ||
1573 | /* Remove 'notifier' from its list. */ | |
1574 | list = list_remove(¬ifier->node); | |
1575 | if (list_is_empty(list)) { | |
1576 | /* The list is now empty. Remove it from the hash and free it. */ | |
1577 | const char *netdev_name = netdev_get_name(notifier->notifier.netdev); | |
1578 | shash_delete(&netdev_linux_notifiers, | |
1579 | shash_find(&netdev_linux_notifiers, netdev_name)); | |
1580 | free(list); | |
1581 | } | |
1582 | free(notifier); | |
1583 | ||
1584 | /* If that was the last notifier, unregister. */ | |
1585 | if (shash_is_empty(&netdev_linux_notifiers)) { | |
46097491 | 1586 | rtnetlink_notifier_unregister(&netdev_linux_poll_notifier); |
8b61709d BP |
1587 | } |
1588 | } | |
1589 | ||
1590 | const struct netdev_class netdev_linux_class = { | |
149f577a | 1591 | "system", |
8b61709d BP |
1592 | |
1593 | netdev_linux_init, | |
1594 | netdev_linux_run, | |
1595 | netdev_linux_wait, | |
1596 | ||
a740f0de | 1597 | netdev_linux_create_system, |
6c88d577 JP |
1598 | netdev_linux_destroy, |
1599 | NULL, /* reconfigure */ | |
1600 | ||
8b61709d BP |
1601 | netdev_linux_open, |
1602 | netdev_linux_close, | |
1603 | ||
1604 | netdev_linux_enumerate, | |
1605 | ||
1606 | netdev_linux_recv, | |
1607 | netdev_linux_recv_wait, | |
1608 | netdev_linux_drain, | |
1609 | ||
1610 | netdev_linux_send, | |
1611 | netdev_linux_send_wait, | |
1612 | ||
1613 | netdev_linux_set_etheraddr, | |
1614 | netdev_linux_get_etheraddr, | |
1615 | netdev_linux_get_mtu, | |
9ab3d9a3 | 1616 | netdev_linux_get_ifindex, |
8b61709d BP |
1617 | netdev_linux_get_carrier, |
1618 | netdev_linux_get_stats, | |
8722022c | 1619 | netdev_linux_set_stats, |
8b61709d BP |
1620 | |
1621 | netdev_linux_get_features, | |
1622 | netdev_linux_set_advertisements, | |
1623 | netdev_linux_get_vlan_vid, | |
1624 | netdev_linux_set_policing, | |
1625 | ||
1626 | netdev_linux_get_in4, | |
1627 | netdev_linux_set_in4, | |
1628 | netdev_linux_get_in6, | |
1629 | netdev_linux_add_router, | |
f1acd62b | 1630 | netdev_linux_get_next_hop, |
8b61709d BP |
1631 | netdev_linux_arp_lookup, |
1632 | ||
1633 | netdev_linux_update_flags, | |
1634 | ||
1635 | netdev_linux_poll_add, | |
1636 | netdev_linux_poll_remove, | |
1637 | }; | |
1638 | ||
1639 | const struct netdev_class netdev_tap_class = { | |
149f577a | 1640 | "tap", |
8b61709d BP |
1641 | |
1642 | netdev_linux_init, | |
149f577a JG |
1643 | netdev_linux_run, |
1644 | netdev_linux_wait, | |
8b61709d | 1645 | |
a740f0de | 1646 | netdev_linux_create_tap, |
6c88d577 JP |
1647 | netdev_linux_destroy, |
1648 | NULL, /* reconfigure */ | |
1649 | ||
8b61709d BP |
1650 | netdev_linux_open, |
1651 | netdev_linux_close, | |
1652 | ||
149f577a | 1653 | NULL, /* enumerate */ |
8b61709d BP |
1654 | |
1655 | netdev_linux_recv, | |
1656 | netdev_linux_recv_wait, | |
1657 | netdev_linux_drain, | |
1658 | ||
1659 | netdev_linux_send, | |
1660 | netdev_linux_send_wait, | |
1661 | ||
1662 | netdev_linux_set_etheraddr, | |
1663 | netdev_linux_get_etheraddr, | |
1664 | netdev_linux_get_mtu, | |
9ab3d9a3 | 1665 | netdev_linux_get_ifindex, |
8b61709d BP |
1666 | netdev_linux_get_carrier, |
1667 | netdev_linux_get_stats, | |
8722022c | 1668 | NULL, /* set_stats */ |
8b61709d BP |
1669 | |
1670 | netdev_linux_get_features, | |
1671 | netdev_linux_set_advertisements, | |
a740f0de JG |
1672 | netdev_linux_get_vlan_vid, |
1673 | netdev_linux_set_policing, | |
1674 | ||
1675 | netdev_linux_get_in4, | |
1676 | netdev_linux_set_in4, | |
1677 | netdev_linux_get_in6, | |
1678 | netdev_linux_add_router, | |
1679 | netdev_linux_get_next_hop, | |
1680 | netdev_linux_arp_lookup, | |
1681 | ||
1682 | netdev_linux_update_flags, | |
1683 | ||
1684 | netdev_linux_poll_add, | |
1685 | netdev_linux_poll_remove, | |
1686 | }; | |
6f643e49 | 1687 | |
8b61709d BP |
1688 | \f |
1689 | static int | |
1690 | get_stats_via_netlink(int ifindex, struct netdev_stats *stats) | |
1691 | { | |
559843ed BP |
1692 | /* Policy for RTNLGRP_LINK messages. |
1693 | * | |
1694 | * There are *many* more fields in these messages, but currently we only | |
1695 | * care about these fields. */ | |
1696 | static const struct nl_policy rtnlgrp_link_policy[] = { | |
1697 | [IFLA_IFNAME] = { .type = NL_A_STRING, .optional = false }, | |
1698 | [IFLA_STATS] = { .type = NL_A_UNSPEC, .optional = true, | |
1699 | .min_len = sizeof(struct rtnl_link_stats) }, | |
1700 | }; | |
1701 | ||
a5af30fb | 1702 | struct nl_sock *rtnl_sock; |
8b61709d BP |
1703 | struct ofpbuf request; |
1704 | struct ofpbuf *reply; | |
1705 | struct ifinfomsg *ifi; | |
1706 | const struct rtnl_link_stats *rtnl_stats; | |
1707 | struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)]; | |
1708 | int error; | |
1709 | ||
a5af30fb BP |
1710 | error = get_rtnl_sock(&rtnl_sock); |
1711 | if (error) { | |
1712 | return error; | |
8b61709d BP |
1713 | } |
1714 | ||
1715 | ofpbuf_init(&request, 0); | |
1716 | nl_msg_put_nlmsghdr(&request, rtnl_sock, sizeof *ifi, | |
1717 | RTM_GETLINK, NLM_F_REQUEST); | |
1718 | ifi = ofpbuf_put_zeros(&request, sizeof *ifi); | |
1719 | ifi->ifi_family = PF_UNSPEC; | |
1720 | ifi->ifi_index = ifindex; | |
1721 | error = nl_sock_transact(rtnl_sock, &request, &reply); | |
1722 | ofpbuf_uninit(&request); | |
1723 | if (error) { | |
1724 | return error; | |
1725 | } | |
1726 | ||
1727 | if (!nl_policy_parse(reply, NLMSG_HDRLEN + sizeof(struct ifinfomsg), | |
1728 | rtnlgrp_link_policy, | |
1729 | attrs, ARRAY_SIZE(rtnlgrp_link_policy))) { | |
1730 | ofpbuf_delete(reply); | |
1731 | return EPROTO; | |
1732 | } | |
1733 | ||
1734 | if (!attrs[IFLA_STATS]) { | |
1735 | VLOG_WARN_RL(&rl, "RTM_GETLINK reply lacks stats"); | |
576e26d7 | 1736 | ofpbuf_delete(reply); |
8b61709d BP |
1737 | return EPROTO; |
1738 | } | |
1739 | ||
1740 | rtnl_stats = nl_attr_get(attrs[IFLA_STATS]); | |
1741 | stats->rx_packets = rtnl_stats->rx_packets; | |
1742 | stats->tx_packets = rtnl_stats->tx_packets; | |
1743 | stats->rx_bytes = rtnl_stats->rx_bytes; | |
1744 | stats->tx_bytes = rtnl_stats->tx_bytes; | |
1745 | stats->rx_errors = rtnl_stats->rx_errors; | |
1746 | stats->tx_errors = rtnl_stats->tx_errors; | |
1747 | stats->rx_dropped = rtnl_stats->rx_dropped; | |
1748 | stats->tx_dropped = rtnl_stats->tx_dropped; | |
1749 | stats->multicast = rtnl_stats->multicast; | |
1750 | stats->collisions = rtnl_stats->collisions; | |
1751 | stats->rx_length_errors = rtnl_stats->rx_length_errors; | |
1752 | stats->rx_over_errors = rtnl_stats->rx_over_errors; | |
1753 | stats->rx_crc_errors = rtnl_stats->rx_crc_errors; | |
1754 | stats->rx_frame_errors = rtnl_stats->rx_frame_errors; | |
1755 | stats->rx_fifo_errors = rtnl_stats->rx_fifo_errors; | |
1756 | stats->rx_missed_errors = rtnl_stats->rx_missed_errors; | |
1757 | stats->tx_aborted_errors = rtnl_stats->tx_aborted_errors; | |
1758 | stats->tx_carrier_errors = rtnl_stats->tx_carrier_errors; | |
1759 | stats->tx_fifo_errors = rtnl_stats->tx_fifo_errors; | |
1760 | stats->tx_heartbeat_errors = rtnl_stats->tx_heartbeat_errors; | |
1761 | stats->tx_window_errors = rtnl_stats->tx_window_errors; | |
1762 | ||
576e26d7 BP |
1763 | ofpbuf_delete(reply); |
1764 | ||
8b61709d BP |
1765 | return 0; |
1766 | } | |
1767 | ||
1768 | static int | |
1769 | get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats) | |
1770 | { | |
1771 | static const char fn[] = "/proc/net/dev"; | |
1772 | char line[1024]; | |
1773 | FILE *stream; | |
1774 | int ln; | |
1775 | ||
1776 | stream = fopen(fn, "r"); | |
1777 | if (!stream) { | |
1778 | VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno)); | |
1779 | return errno; | |
1780 | } | |
1781 | ||
1782 | ln = 0; | |
1783 | while (fgets(line, sizeof line, stream)) { | |
1784 | if (++ln >= 3) { | |
1785 | char devname[16]; | |
1786 | #define X64 "%"SCNu64 | |
1787 | if (sscanf(line, | |
1788 | " %15[^:]:" | |
1789 | X64 X64 X64 X64 X64 X64 X64 "%*u" | |
1790 | X64 X64 X64 X64 X64 X64 X64 "%*u", | |
1791 | devname, | |
1792 | &stats->rx_bytes, | |
1793 | &stats->rx_packets, | |
1794 | &stats->rx_errors, | |
1795 | &stats->rx_dropped, | |
1796 | &stats->rx_fifo_errors, | |
1797 | &stats->rx_frame_errors, | |
1798 | &stats->multicast, | |
1799 | &stats->tx_bytes, | |
1800 | &stats->tx_packets, | |
1801 | &stats->tx_errors, | |
1802 | &stats->tx_dropped, | |
1803 | &stats->tx_fifo_errors, | |
1804 | &stats->collisions, | |
1805 | &stats->tx_carrier_errors) != 15) { | |
1806 | VLOG_WARN_RL(&rl, "%s:%d: parse error", fn, ln); | |
1807 | } else if (!strcmp(devname, netdev_name)) { | |
1808 | stats->rx_length_errors = UINT64_MAX; | |
1809 | stats->rx_over_errors = UINT64_MAX; | |
1810 | stats->rx_crc_errors = UINT64_MAX; | |
1811 | stats->rx_missed_errors = UINT64_MAX; | |
1812 | stats->tx_aborted_errors = UINT64_MAX; | |
1813 | stats->tx_heartbeat_errors = UINT64_MAX; | |
1814 | stats->tx_window_errors = UINT64_MAX; | |
1815 | fclose(stream); | |
1816 | return 0; | |
1817 | } | |
1818 | } | |
1819 | } | |
1820 | VLOG_WARN_RL(&rl, "%s: no stats for %s", fn, netdev_name); | |
1821 | fclose(stream); | |
1822 | return ENODEV; | |
1823 | } | |
1824 | \f | |
1825 | static int | |
1826 | get_flags(const struct netdev *netdev, int *flags) | |
1827 | { | |
1828 | struct ifreq ifr; | |
1829 | int error; | |
1830 | ||
149f577a JG |
1831 | error = netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, SIOCGIFFLAGS, |
1832 | "SIOCGIFFLAGS"); | |
8b61709d BP |
1833 | *flags = ifr.ifr_flags; |
1834 | return error; | |
1835 | } | |
1836 | ||
1837 | static int | |
1838 | set_flags(struct netdev *netdev, int flags) | |
1839 | { | |
1840 | struct ifreq ifr; | |
1841 | ||
1842 | ifr.ifr_flags = flags; | |
149f577a JG |
1843 | return netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, SIOCSIFFLAGS, |
1844 | "SIOCSIFFLAGS"); | |
8b61709d BP |
1845 | } |
1846 | ||
1847 | static int | |
1848 | do_get_ifindex(const char *netdev_name) | |
1849 | { | |
1850 | struct ifreq ifr; | |
1851 | ||
1852 | strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name); | |
1853 | COVERAGE_INC(netdev_get_ifindex); | |
1854 | if (ioctl(af_inet_sock, SIOCGIFINDEX, &ifr) < 0) { | |
1855 | VLOG_WARN_RL(&rl, "ioctl(SIOCGIFINDEX) on %s device failed: %s", | |
1856 | netdev_name, strerror(errno)); | |
1857 | return -errno; | |
1858 | } | |
1859 | return ifr.ifr_ifindex; | |
1860 | } | |
1861 | ||
1862 | static int | |
1863 | get_ifindex(const struct netdev *netdev_, int *ifindexp) | |
1864 | { | |
149f577a JG |
1865 | struct netdev_dev_linux *netdev_dev = |
1866 | netdev_dev_linux_cast(netdev_get_dev(netdev_)); | |
8b61709d | 1867 | *ifindexp = 0; |
149f577a | 1868 | if (!(netdev_dev->cache_valid & VALID_IFINDEX)) { |
8b61709d BP |
1869 | int ifindex = do_get_ifindex(netdev_get_name(netdev_)); |
1870 | if (ifindex < 0) { | |
1871 | return -ifindex; | |
1872 | } | |
149f577a JG |
1873 | netdev_dev->cache_valid |= VALID_IFINDEX; |
1874 | netdev_dev->ifindex = ifindex; | |
8b61709d | 1875 | } |
149f577a | 1876 | *ifindexp = netdev_dev->ifindex; |
8b61709d BP |
1877 | return 0; |
1878 | } | |
1879 | ||
1880 | static int | |
1881 | get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]) | |
1882 | { | |
1883 | struct ifreq ifr; | |
1884 | int hwaddr_family; | |
1885 | ||
1886 | memset(&ifr, 0, sizeof ifr); | |
1887 | strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name); | |
1888 | COVERAGE_INC(netdev_get_hwaddr); | |
1889 | if (ioctl(af_inet_sock, SIOCGIFHWADDR, &ifr) < 0) { | |
1890 | VLOG_ERR("ioctl(SIOCGIFHWADDR) on %s device failed: %s", | |
1891 | netdev_name, strerror(errno)); | |
1892 | return errno; | |
1893 | } | |
1894 | hwaddr_family = ifr.ifr_hwaddr.sa_family; | |
1895 | if (hwaddr_family != AF_UNSPEC && hwaddr_family != ARPHRD_ETHER) { | |
1896 | VLOG_WARN("%s device has unknown hardware address family %d", | |
1897 | netdev_name, hwaddr_family); | |
1898 | } | |
1899 | memcpy(ea, ifr.ifr_hwaddr.sa_data, ETH_ADDR_LEN); | |
1900 | return 0; | |
1901 | } | |
1902 | ||
1903 | static int | |
1904 | set_etheraddr(const char *netdev_name, int hwaddr_family, | |
1905 | const uint8_t mac[ETH_ADDR_LEN]) | |
1906 | { | |
1907 | struct ifreq ifr; | |
1908 | ||
1909 | memset(&ifr, 0, sizeof ifr); | |
1910 | strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name); | |
1911 | ifr.ifr_hwaddr.sa_family = hwaddr_family; | |
1912 | memcpy(ifr.ifr_hwaddr.sa_data, mac, ETH_ADDR_LEN); | |
1913 | COVERAGE_INC(netdev_set_hwaddr); | |
1914 | if (ioctl(af_inet_sock, SIOCSIFHWADDR, &ifr) < 0) { | |
1915 | VLOG_ERR("ioctl(SIOCSIFHWADDR) on %s device failed: %s", | |
1916 | netdev_name, strerror(errno)); | |
1917 | return errno; | |
1918 | } | |
1919 | return 0; | |
1920 | } | |
1921 | ||
1922 | static int | |
0b0544d7 | 1923 | netdev_linux_do_ethtool(const char *name, struct ethtool_cmd *ecmd, |
8b61709d BP |
1924 | int cmd, const char *cmd_name) |
1925 | { | |
1926 | struct ifreq ifr; | |
1927 | ||
1928 | memset(&ifr, 0, sizeof ifr); | |
0b0544d7 | 1929 | strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name); |
8b61709d BP |
1930 | ifr.ifr_data = (caddr_t) ecmd; |
1931 | ||
1932 | ecmd->cmd = cmd; | |
1933 | COVERAGE_INC(netdev_ethtool); | |
1934 | if (ioctl(af_inet_sock, SIOCETHTOOL, &ifr) == 0) { | |
1935 | return 0; | |
1936 | } else { | |
1937 | if (errno != EOPNOTSUPP) { | |
1938 | VLOG_WARN_RL(&rl, "ethtool command %s on network device %s " | |
0b0544d7 | 1939 | "failed: %s", cmd_name, name, strerror(errno)); |
8b61709d BP |
1940 | } else { |
1941 | /* The device doesn't support this operation. That's pretty | |
1942 | * common, so there's no point in logging anything. */ | |
1943 | } | |
1944 | return errno; | |
1945 | } | |
1946 | } | |
1947 | ||
1948 | static int | |
149f577a JG |
1949 | netdev_linux_do_ioctl(const char *name, struct ifreq *ifr, int cmd, |
1950 | const char *cmd_name) | |
8b61709d | 1951 | { |
149f577a | 1952 | strncpy(ifr->ifr_name, name, sizeof ifr->ifr_name); |
8b61709d | 1953 | if (ioctl(af_inet_sock, cmd, ifr) == -1) { |
149f577a JG |
1954 | VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name, |
1955 | strerror(errno)); | |
8b61709d BP |
1956 | return errno; |
1957 | } | |
1958 | return 0; | |
1959 | } | |
f1acd62b BP |
1960 | |
1961 | static int | |
1962 | netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip, | |
1963 | int cmd, const char *cmd_name) | |
1964 | { | |
1965 | struct ifreq ifr; | |
1966 | int error; | |
1967 | ||
1968 | ifr.ifr_addr.sa_family = AF_INET; | |
149f577a | 1969 | error = netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, cmd, cmd_name); |
f1acd62b BP |
1970 | if (!error) { |
1971 | const struct sockaddr_in *sin = (struct sockaddr_in *) &ifr.ifr_addr; | |
1972 | *ip = sin->sin_addr; | |
1973 | } | |
1974 | return error; | |
1975 | } | |
a5af30fb BP |
1976 | |
1977 | /* Obtains a Netlink routing socket that is not subscribed to any multicast | |
1978 | * groups. Returns 0 if successful, otherwise a positive errno value. Stores | |
1979 | * the socket in '*rtnl_sockp' if successful, otherwise a null pointer. */ | |
1980 | static int | |
1981 | get_rtnl_sock(struct nl_sock **rtnl_sockp) | |
1982 | { | |
1983 | static struct nl_sock *sock; | |
1984 | int error; | |
1985 | ||
1986 | if (!sock) { | |
1987 | error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &sock); | |
1988 | if (error) { | |
1989 | VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s", | |
1990 | strerror(error)); | |
1991 | } | |
1992 | } else { | |
1993 | error = 0; | |
1994 | } | |
1995 | ||
1996 | *rtnl_sockp = sock; | |
1997 | return error; | |
1998 | } |