add prelocal table to manipulate locally destinated traffic
[project/netifd.git] / system-linux.c
1 /*
2 * netifd - network interface daemon
3 * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
4 * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
5 * Copyright (C) 2013 Steven Barth <steven@midlink.org>
6 * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2
10 * as published by the Free Software Foundation
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 */
17 #define _GNU_SOURCE
18
19 #include <sys/socket.h>
20 #include <sys/ioctl.h>
21 #include <sys/stat.h>
22 #include <sys/syscall.h>
23
24 #include <net/if.h>
25 #include <net/if_arp.h>
26
27 #include <arpa/inet.h>
28 #include <netinet/in.h>
29
30 #include <linux/rtnetlink.h>
31 #include <linux/sockios.h>
32 #include <linux/ip.h>
33 #include <linux/if_link.h>
34 #include <linux/if_vlan.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38 #include <linux/ethtool.h>
39 #include <linux/fib_rules.h>
40 #include <linux/version.h>
41
42 #ifndef RTN_FAILED_POLICY
43 #define RTN_FAILED_POLICY 12
44 #endif
45
46 #ifndef RT_TABLE_PRELOCAL
47 #define RT_TABLE_PRELOCAL 128
48 #endif
49
50 #include <string.h>
51 #include <fcntl.h>
52 #include <glob.h>
53 #include <time.h>
54
55 #include <netlink/msg.h>
56 #include <netlink/attr.h>
57 #include <netlink/socket.h>
58 #include <libubox/uloop.h>
59
60 #include "netifd.h"
61 #include "device.h"
62 #include "system.h"
63
64 struct event_socket {
65 struct uloop_fd uloop;
66 struct nl_sock *sock;
67 int bufsize;
68 };
69
70 static int sock_ioctl = -1;
71 static struct nl_sock *sock_rtnl = NULL;
72
73 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
74 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
75
76 static char dev_buf[256];
77
78 static void
79 handler_nl_event(struct uloop_fd *u, unsigned int events)
80 {
81 struct event_socket *ev = container_of(u, struct event_socket, uloop);
82 int err;
83 socklen_t errlen = sizeof(err);
84
85 if (!u->error) {
86 nl_recvmsgs_default(ev->sock);
87 return;
88 }
89
90 if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
91 goto abort;
92
93 switch(err) {
94 case ENOBUFS:
95 // Increase rx buffer size on netlink socket
96 ev->bufsize *= 2;
97 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
98 goto abort;
99
100 // Request full dump since some info got dropped
101 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
102 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
103 break;
104
105 default:
106 goto abort;
107 }
108 u->error = false;
109 return;
110
111 abort:
112 uloop_fd_delete(&ev->uloop);
113 return;
114 }
115
116 static struct nl_sock *
117 create_socket(int protocol, int groups)
118 {
119 struct nl_sock *sock;
120
121 sock = nl_socket_alloc();
122 if (!sock)
123 return NULL;
124
125 if (groups)
126 nl_join_groups(sock, groups);
127
128 if (nl_connect(sock, protocol))
129 return NULL;
130
131 return sock;
132 }
133
134 static bool
135 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
136 uloop_fd_handler cb, int flags)
137 {
138 ev->sock = create_socket(protocol, groups);
139 if (!ev->sock)
140 return false;
141
142 ev->uloop.fd = nl_socket_get_fd(ev->sock);
143 ev->uloop.cb = cb;
144 if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
145 return false;
146
147 return true;
148 }
149
150 static bool
151 create_event_socket(struct event_socket *ev, int protocol,
152 int (*cb)(struct nl_msg *msg, void *arg))
153 {
154 if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
155 return false;
156
157 // Install the valid custom callback handler
158 nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
159
160 // Disable sequence number checking on event sockets
161 nl_socket_disable_seq_check(ev->sock);
162
163 // Increase rx buffer size to 65K on event sockets
164 ev->bufsize = 65535;
165 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
166 return false;
167
168 return true;
169 }
170
171 static bool
172 system_rtn_aton(const char *src, unsigned int *dst)
173 {
174 char *e;
175 unsigned int n;
176
177 if (!strcmp(src, "local"))
178 n = RTN_LOCAL;
179 else if (!strcmp(src, "nat"))
180 n = RTN_NAT;
181 else if (!strcmp(src, "broadcast"))
182 n = RTN_BROADCAST;
183 else if (!strcmp(src, "anycast"))
184 n = RTN_ANYCAST;
185 else if (!strcmp(src, "multicast"))
186 n = RTN_MULTICAST;
187 else if (!strcmp(src, "prohibit"))
188 n = RTN_PROHIBIT;
189 else if (!strcmp(src, "unreachable"))
190 n = RTN_UNREACHABLE;
191 else if (!strcmp(src, "blackhole"))
192 n = RTN_BLACKHOLE;
193 else if (!strcmp(src, "xresolve"))
194 n = RTN_XRESOLVE;
195 else if (!strcmp(src, "unicast"))
196 n = RTN_UNICAST;
197 else if (!strcmp(src, "throw"))
198 n = RTN_THROW;
199 else if (!strcmp(src, "failed_policy"))
200 n = RTN_FAILED_POLICY;
201 else {
202 n = strtoul(src, &e, 0);
203 if (!e || *e || e == src || n > 255)
204 return false;
205 }
206
207 *dst = n;
208 return true;
209 }
210
211 static bool
212 system_tos_aton(const char *src, unsigned *dst)
213 {
214 char *e;
215
216 *dst = strtoul(src, &e, 16);
217 if (e == src || *e || *dst > 255)
218 return false;
219
220 return true;
221 }
222
223 int system_init(void)
224 {
225 static struct event_socket rtnl_event;
226 static struct event_socket hotplug_event;
227
228 sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
229 system_fd_set_cloexec(sock_ioctl);
230
231 // Prepare socket for routing / address control
232 sock_rtnl = create_socket(NETLINK_ROUTE, 0);
233 if (!sock_rtnl)
234 return -1;
235
236 if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
237 return -1;
238
239 if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
240 handle_hotplug_event, 0))
241 return -1;
242
243 // Receive network link events form kernel
244 nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
245
246 return 0;
247 }
248
249 static void system_set_sysctl(const char *path, const char *val)
250 {
251 int fd;
252
253 fd = open(path, O_WRONLY);
254 if (fd < 0)
255 return;
256
257 if (write(fd, val, strlen(val))) {}
258 close(fd);
259 }
260
261 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
262 {
263 snprintf(dev_buf, sizeof(dev_buf), path, device);
264 system_set_sysctl(dev_buf, val);
265 }
266
267 static void system_set_disable_ipv6(struct device *dev, const char *val)
268 {
269 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
270 }
271
272 static void system_set_rpfilter(struct device *dev, const char *val)
273 {
274 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter", dev->ifname, val);
275 }
276
277 static void system_set_acceptlocal(struct device *dev, const char *val)
278 {
279 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local", dev->ifname, val);
280 }
281
282 static void system_set_igmpversion(struct device *dev, const char *val)
283 {
284 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version", dev->ifname, val);
285 }
286
287 static void system_set_mldversion(struct device *dev, const char *val)
288 {
289 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version", dev->ifname, val);
290 }
291
292 static void system_set_neigh4reachabletime(struct device *dev, const char *val)
293 {
294 system_set_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms", dev->ifname, val);
295 }
296
297 static void system_set_neigh6reachabletime(struct device *dev, const char *val)
298 {
299 system_set_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms", dev->ifname, val);
300 }
301
302 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
303 {
304 int fd = -1, ret = -1;
305
306 fd = open(path, O_RDONLY);
307 if (fd < 0)
308 goto out;
309
310 ssize_t len = read(fd, buf, buf_sz - 1);
311 if (len < 0)
312 goto out;
313
314 ret = buf[len] = 0;
315
316 out:
317 if (fd >= 0)
318 close(fd);
319
320 return ret;
321 }
322
323 static int
324 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
325 {
326 snprintf(dev_buf, sizeof(dev_buf), path, device);
327 return system_get_sysctl(dev_buf, buf, buf_sz);
328 }
329
330 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
331 {
332 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
333 dev->ifname, buf, buf_sz);
334 }
335
336 static int system_get_rpfilter(struct device *dev, char *buf, const size_t buf_sz)
337 {
338 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter",
339 dev->ifname, buf, buf_sz);
340 }
341
342 static int system_get_acceptlocal(struct device *dev, char *buf, const size_t buf_sz)
343 {
344 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local",
345 dev->ifname, buf, buf_sz);
346 }
347
348 static int system_get_igmpversion(struct device *dev, char *buf, const size_t buf_sz)
349 {
350 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version",
351 dev->ifname, buf, buf_sz);
352 }
353
354 static int system_get_mldversion(struct device *dev, char *buf, const size_t buf_sz)
355 {
356 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version",
357 dev->ifname, buf, buf_sz);
358 }
359
360 static int system_get_neigh4reachabletime(struct device *dev, char *buf, const size_t buf_sz)
361 {
362 return system_get_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms",
363 dev->ifname, buf, buf_sz);
364 }
365
366 static int system_get_neigh6reachabletime(struct device *dev, char *buf, const size_t buf_sz)
367 {
368 return system_get_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms",
369 dev->ifname, buf, buf_sz);
370 }
371
372 // Evaluate netlink messages
373 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
374 {
375 struct nlmsghdr *nh = nlmsg_hdr(msg);
376 struct ifinfomsg *ifi = NLMSG_DATA(nh);
377 struct nlattr *nla[__IFLA_MAX];
378 int link_state = 0;
379 char buf[10];
380
381 if (nh->nlmsg_type != RTM_NEWLINK)
382 goto out;
383
384 nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
385 if (!nla[IFLA_IFNAME])
386 goto out;
387
388 struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
389 if (!dev || dev->type->keep_link_status)
390 goto out;
391
392 if (!system_get_dev_sysctl("/sys/class/net/%s/carrier", dev->ifname, buf, sizeof(buf)))
393 link_state = strtoul(buf, NULL, 0);
394
395 device_set_link(dev, link_state ? true : false);
396
397 out:
398 return 0;
399 }
400
401 static void
402 handle_hotplug_msg(char *data, int size)
403 {
404 const char *subsystem = NULL, *interface = NULL;
405 char *cur, *end, *sep;
406 struct device *dev;
407 int skip;
408 bool add;
409
410 if (!strncmp(data, "add@", 4))
411 add = true;
412 else if (!strncmp(data, "remove@", 7))
413 add = false;
414 else
415 return;
416
417 skip = strlen(data) + 1;
418 end = data + size;
419
420 for (cur = data + skip; cur < end; cur += skip) {
421 skip = strlen(cur) + 1;
422
423 sep = strchr(cur, '=');
424 if (!sep)
425 continue;
426
427 *sep = 0;
428 if (!strcmp(cur, "INTERFACE"))
429 interface = sep + 1;
430 else if (!strcmp(cur, "SUBSYSTEM")) {
431 subsystem = sep + 1;
432 if (strcmp(subsystem, "net") != 0)
433 return;
434 }
435 if (subsystem && interface)
436 goto found;
437 }
438 return;
439
440 found:
441 dev = device_get(interface, false);
442 if (!dev)
443 return;
444
445 if (dev->type != &simple_device_type)
446 return;
447
448 if (add && system_if_force_external(dev->ifname))
449 return;
450
451 device_set_present(dev, add);
452 }
453
454 static void
455 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
456 {
457 struct event_socket *ev = container_of(u, struct event_socket, uloop);
458 struct sockaddr_nl nla;
459 unsigned char *buf = NULL;
460 int size;
461
462 while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
463 if (nla.nl_pid == 0)
464 handle_hotplug_msg((char *) buf, size);
465
466 free(buf);
467 }
468 }
469
470 static int system_rtnl_call(struct nl_msg *msg)
471 {
472 int ret;
473
474 ret = nl_send_auto_complete(sock_rtnl, msg);
475 nlmsg_free(msg);
476
477 if (ret < 0)
478 return ret;
479
480 return nl_wait_for_ack(sock_rtnl);
481 }
482
483 int system_bridge_delbr(struct device *bridge)
484 {
485 return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
486 }
487
488 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
489 {
490 struct ifreq ifr;
491
492 memset(&ifr, 0, sizeof(ifr));
493 if (dev)
494 ifr.ifr_ifindex = dev->ifindex;
495 else
496 ifr.ifr_data = data;
497 strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
498 return ioctl(sock_ioctl, cmd, &ifr);
499 }
500
501 static bool system_is_bridge(const char *name, char *buf, int buflen)
502 {
503 struct stat st;
504
505 snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
506 if (stat(buf, &st) < 0)
507 return false;
508
509 return true;
510 }
511
512 static char *system_get_bridge(const char *name, char *buf, int buflen)
513 {
514 char *path;
515 ssize_t len = -1;
516 glob_t gl;
517
518 snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
519 if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
520 return NULL;
521
522 if (gl.gl_pathc > 0)
523 len = readlink(gl.gl_pathv[0], buf, buflen);
524
525 globfree(&gl);
526
527 if (len < 0)
528 return NULL;
529
530 buf[len] = 0;
531 path = strrchr(buf, '/');
532 if (!path)
533 return NULL;
534
535 return path + 1;
536 }
537
538 static void system_bridge_set_wireless(const char *bridge, const char *dev)
539 {
540 snprintf(dev_buf, sizeof(dev_buf),
541 "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
542 bridge, dev);
543 system_set_sysctl(dev_buf, "1");
544 }
545
546 int system_bridge_addif(struct device *bridge, struct device *dev)
547 {
548 char *oldbr;
549 int ret = 0;
550
551 oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
552 if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
553 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
554
555 if (dev->wireless)
556 system_bridge_set_wireless(bridge->ifname, dev->ifname);
557
558 return ret;
559 }
560
561 int system_bridge_delif(struct device *bridge, struct device *dev)
562 {
563 return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
564 }
565
566 int system_if_resolve(struct device *dev)
567 {
568 struct ifreq ifr;
569 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
570 if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
571 return ifr.ifr_ifindex;
572 else
573 return 0;
574 }
575
576 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
577 {
578 struct ifreq ifr;
579
580 memset(&ifr, 0, sizeof(ifr));
581 strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
582 ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
583 ifr.ifr_flags |= add;
584 ifr.ifr_flags &= ~rem;
585 return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
586 }
587
588 struct clear_data {
589 struct nl_msg *msg;
590 struct device *dev;
591 int type;
592 int size;
593 int af;
594 };
595
596
597 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
598 {
599 struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
600
601 return ifa->ifa_index == ifindex;
602 }
603
604 static bool check_route(struct nlmsghdr *hdr, int ifindex)
605 {
606 struct rtmsg *r = NLMSG_DATA(hdr);
607 struct nlattr *tb[__RTA_MAX];
608
609 if (r->rtm_protocol == RTPROT_KERNEL &&
610 r->rtm_family == AF_INET6)
611 return false;
612
613 nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
614 if (!tb[RTA_OIF])
615 return false;
616
617 return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
618 }
619
620 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
621 {
622 return true;
623 }
624
625 static int cb_clear_event(struct nl_msg *msg, void *arg)
626 {
627 struct clear_data *clr = arg;
628 struct nlmsghdr *hdr = nlmsg_hdr(msg);
629 bool (*cb)(struct nlmsghdr *, int ifindex);
630 int type;
631
632 switch(clr->type) {
633 case RTM_GETADDR:
634 type = RTM_DELADDR;
635 if (hdr->nlmsg_type != RTM_NEWADDR)
636 return NL_SKIP;
637
638 cb = check_ifaddr;
639 break;
640 case RTM_GETROUTE:
641 type = RTM_DELROUTE;
642 if (hdr->nlmsg_type != RTM_NEWROUTE)
643 return NL_SKIP;
644
645 cb = check_route;
646 break;
647 case RTM_GETRULE:
648 type = RTM_DELRULE;
649 if (hdr->nlmsg_type != RTM_NEWRULE)
650 return NL_SKIP;
651
652 cb = check_rule;
653 break;
654 default:
655 return NL_SKIP;
656 }
657
658 if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
659 return NL_SKIP;
660
661 if (type == RTM_DELRULE)
662 D(SYSTEM, "Remove a rule\n");
663 else
664 D(SYSTEM, "Remove %s from device %s\n",
665 type == RTM_DELADDR ? "an address" : "a route",
666 clr->dev->ifname);
667 memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
668 hdr = nlmsg_hdr(clr->msg);
669 hdr->nlmsg_type = type;
670 hdr->nlmsg_flags = NLM_F_REQUEST;
671
672 nl_socket_disable_auto_ack(sock_rtnl);
673 nl_send_auto_complete(sock_rtnl, clr->msg);
674 nl_socket_enable_auto_ack(sock_rtnl);
675
676 return NL_SKIP;
677 }
678
679 static int
680 cb_finish_event(struct nl_msg *msg, void *arg)
681 {
682 int *pending = arg;
683 *pending = 0;
684 return NL_STOP;
685 }
686
687 static int
688 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
689 {
690 int *pending = arg;
691 *pending = err->error;
692 return NL_STOP;
693 }
694
695 static void
696 system_if_clear_entries(struct device *dev, int type, int af)
697 {
698 struct clear_data clr;
699 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
700 struct rtmsg rtm = {
701 .rtm_family = af,
702 .rtm_flags = RTM_F_CLONED,
703 };
704 int flags = NLM_F_DUMP;
705 int pending = 1;
706
707 clr.af = af;
708 clr.dev = dev;
709 clr.type = type;
710 switch (type) {
711 case RTM_GETADDR:
712 case RTM_GETRULE:
713 clr.size = sizeof(struct rtgenmsg);
714 break;
715 case RTM_GETROUTE:
716 clr.size = sizeof(struct rtmsg);
717 break;
718 default:
719 return;
720 }
721
722 if (!cb)
723 return;
724
725 clr.msg = nlmsg_alloc_simple(type, flags);
726 if (!clr.msg)
727 goto out;
728
729 nlmsg_append(clr.msg, &rtm, clr.size, 0);
730 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
731 nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
732 nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
733
734 nl_send_auto_complete(sock_rtnl, clr.msg);
735 while (pending > 0)
736 nl_recvmsgs(sock_rtnl, cb);
737
738 nlmsg_free(clr.msg);
739 out:
740 nl_cb_put(cb);
741 }
742
743 /*
744 * Clear bridge (membership) state and bring down device
745 */
746 void system_if_clear_state(struct device *dev)
747 {
748 static char buf[256];
749 char *bridge;
750
751 device_set_ifindex(dev, system_if_resolve(dev));
752 if (dev->external || !dev->ifindex)
753 return;
754
755 system_if_flags(dev->ifname, 0, IFF_UP);
756
757 if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
758 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
759 system_bridge_delbr(dev);
760 return;
761 }
762
763 bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
764 if (bridge) {
765 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
766 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
767 }
768
769 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
770 system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
771 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
772 system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
773 system_set_disable_ipv6(dev, "0");
774 }
775
776 static inline unsigned long
777 sec_to_jiffies(int val)
778 {
779 return (unsigned long) val * 100;
780 }
781
782 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
783 {
784 unsigned long args[4] = {};
785
786 if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
787 return -1;
788
789 args[0] = BRCTL_SET_BRIDGE_STP_STATE;
790 args[1] = !!cfg->stp;
791 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
792
793 args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
794 args[1] = sec_to_jiffies(cfg->forward_delay);
795 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
796
797 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
798 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
799
800 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
801 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
802
803 args[0] = BRCTL_SET_BRIDGE_PRIORITY;
804 args[1] = cfg->priority;
805 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
806
807 if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
808 args[0] = BRCTL_SET_AGEING_TIME;
809 args[1] = sec_to_jiffies(cfg->ageing_time);
810 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
811 }
812
813 if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
814 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
815 args[1] = sec_to_jiffies(cfg->hello_time);
816 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
817 }
818
819 if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
820 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
821 args[1] = sec_to_jiffies(cfg->max_age);
822 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
823 }
824
825 return 0;
826 }
827
828 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
829 {
830 struct nl_msg *msg;
831 struct nlattr *linkinfo, *data;
832 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
833 int i, rv;
834 static const struct {
835 const char *name;
836 enum macvlan_mode val;
837 } modes[] = {
838 { "private", MACVLAN_MODE_PRIVATE },
839 { "vepa", MACVLAN_MODE_VEPA },
840 { "bridge", MACVLAN_MODE_BRIDGE },
841 { "passthru", MACVLAN_MODE_PASSTHRU },
842 };
843
844 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
845
846 if (!msg)
847 return -1;
848
849 nlmsg_append(msg, &iim, sizeof(iim), 0);
850
851 if (cfg->flags & MACVLAN_OPT_MACADDR)
852 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
853 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
854 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
855
856 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
857 goto nla_put_failure;
858
859 nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
860
861 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
862 goto nla_put_failure;
863
864 if (cfg->mode) {
865 for (i = 0; i < ARRAY_SIZE(modes); i++) {
866 if (strcmp(cfg->mode, modes[i].name) != 0)
867 continue;
868
869 nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
870 break;
871 }
872 }
873
874 nla_nest_end(msg, data);
875 nla_nest_end(msg, linkinfo);
876
877 rv = system_rtnl_call(msg);
878 if (rv)
879 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
880
881 return rv;
882
883 nla_put_failure:
884 nlmsg_free(msg);
885 return -ENOMEM;
886 }
887
888 static int system_link_del(const char *ifname)
889 {
890 struct nl_msg *msg;
891 struct ifinfomsg iim = {
892 .ifi_family = AF_UNSPEC,
893 .ifi_index = 0,
894 };
895
896 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
897
898 if (!msg)
899 return -1;
900
901 nlmsg_append(msg, &iim, sizeof(iim), 0);
902 nla_put_string(msg, IFLA_IFNAME, ifname);
903 return system_rtnl_call(msg);
904 }
905
906 int system_macvlan_del(struct device *macvlan)
907 {
908 return system_link_del(macvlan->ifname);
909 }
910
911 static int system_vlan(struct device *dev, int id)
912 {
913 struct vlan_ioctl_args ifr = {
914 .cmd = SET_VLAN_NAME_TYPE_CMD,
915 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
916 };
917
918 ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
919
920 if (id < 0) {
921 ifr.cmd = DEL_VLAN_CMD;
922 ifr.u.VID = 0;
923 } else {
924 ifr.cmd = ADD_VLAN_CMD;
925 ifr.u.VID = id;
926 }
927 strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
928 return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
929 }
930
931 int system_vlan_add(struct device *dev, int id)
932 {
933 return system_vlan(dev, id);
934 }
935
936 int system_vlan_del(struct device *dev)
937 {
938 return system_vlan(dev, -1);
939 }
940
941 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
942 {
943 struct nl_msg *msg;
944 struct nlattr *linkinfo, *data;
945 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
946 int rv;
947
948 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
949
950 if (!msg)
951 return -1;
952
953 nlmsg_append(msg, &iim, sizeof(iim), 0);
954 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
955 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
956
957 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
958 goto nla_put_failure;
959
960 nla_put_string(msg, IFLA_INFO_KIND, "vlan");
961
962 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
963 goto nla_put_failure;
964
965 nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
966
967 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
968 nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
969 #else
970 if(cfg->proto == VLAN_PROTO_8021AD)
971 netifd_log_message(L_WARNING, "%s Your kernel is older than linux 3.10.0, 802.1ad is not supported defaulting to 802.1q", vlandev->type->name);
972 #endif
973
974 nla_nest_end(msg, data);
975 nla_nest_end(msg, linkinfo);
976
977 rv = system_rtnl_call(msg);
978 if (rv)
979 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
980
981 return rv;
982
983 nla_put_failure:
984 nlmsg_free(msg);
985 return -ENOMEM;
986 }
987
988 int system_vlandev_del(struct device *vlandev)
989 {
990 return system_link_del(vlandev->ifname);
991 }
992
993 static void
994 system_if_get_settings(struct device *dev, struct device_settings *s)
995 {
996 struct ifreq ifr;
997 char buf[10];
998
999 memset(&ifr, 0, sizeof(ifr));
1000 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1001
1002 if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
1003 s->mtu = ifr.ifr_mtu;
1004 s->flags |= DEV_OPT_MTU;
1005 }
1006
1007 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
1008 s->txqueuelen = ifr.ifr_qlen;
1009 s->flags |= DEV_OPT_TXQUEUELEN;
1010 }
1011
1012 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
1013 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
1014 s->flags |= DEV_OPT_MACADDR;
1015 }
1016
1017 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
1018 s->ipv6 = !strtoul(buf, NULL, 0);
1019 s->flags |= DEV_OPT_IPV6;
1020 }
1021
1022 if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
1023 s->promisc = ifr.ifr_flags & IFF_PROMISC;
1024 s->flags |= DEV_OPT_PROMISC;
1025 }
1026
1027 if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1028 s->rpfilter = strtoul(buf, NULL, 0);
1029 s->flags |= DEV_OPT_RPFILTER;
1030 }
1031
1032 if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1033 s->acceptlocal = strtoul(buf, NULL, 0);
1034 s->flags |= DEV_OPT_ACCEPTLOCAL;
1035 }
1036
1037 if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1038 s->igmpversion = strtoul(buf, NULL, 0);
1039 s->flags |= DEV_OPT_IGMPVERSION;
1040 }
1041
1042 if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1043 s->mldversion = strtoul(buf, NULL, 0);
1044 s->flags |= DEV_OPT_MLDVERSION;
1045 }
1046
1047 if (!system_get_neigh4reachabletime(dev, buf, sizeof(buf))) {
1048 s->neigh4reachabletime = strtoul(buf, NULL, 0);
1049 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1050 }
1051
1052 if (!system_get_neigh6reachabletime(dev, buf, sizeof(buf))) {
1053 s->neigh6reachabletime = strtoul(buf, NULL, 0);
1054 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1055 }
1056 }
1057
1058 void
1059 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1060 {
1061 struct ifreq ifr;
1062
1063 if (!apply_mask)
1064 return;
1065
1066 memset(&ifr, 0, sizeof(ifr));
1067 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1068 if (s->flags & DEV_OPT_MTU & apply_mask) {
1069 ifr.ifr_mtu = s->mtu;
1070 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1071 s->flags &= ~DEV_OPT_MTU;
1072 }
1073 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1074 ifr.ifr_qlen = s->txqueuelen;
1075 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1076 s->flags &= ~DEV_OPT_TXQUEUELEN;
1077 }
1078 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1079 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1080 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1081 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1082 s->flags &= ~DEV_OPT_MACADDR;
1083 }
1084 if (s->flags & DEV_OPT_IPV6 & apply_mask)
1085 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1086 if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1087 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1088 !s->promisc ? IFF_PROMISC : 0) < 0)
1089 s->flags &= ~DEV_OPT_PROMISC;
1090 }
1091 if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1092 char buf[2];
1093
1094 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1095 system_set_rpfilter(dev, buf);
1096 }
1097 if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1098 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1099 if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1100 char buf[2];
1101
1102 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1103 system_set_igmpversion(dev, buf);
1104 }
1105 if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1106 char buf[2];
1107
1108 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1109 system_set_mldversion(dev, buf);
1110 }
1111 if (s->flags & DEV_OPT_NEIGHREACHABLETIME & apply_mask) {
1112 char buf[12];
1113
1114 snprintf(buf, sizeof(buf), "%d", s->neigh4reachabletime);
1115 system_set_neigh4reachabletime(dev, buf);
1116 snprintf(buf, sizeof(buf), "%d", s->neigh6reachabletime);
1117 system_set_neigh6reachabletime(dev, buf);
1118 }
1119 }
1120
1121 int system_if_up(struct device *dev)
1122 {
1123 system_if_get_settings(dev, &dev->orig_settings);
1124 /* Only keep orig settings based on what needs to be set */
1125 dev->orig_settings.flags &= dev->settings.flags;
1126 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1127 return system_if_flags(dev->ifname, IFF_UP, 0);
1128 }
1129
1130 int system_if_down(struct device *dev)
1131 {
1132 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1133 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1134 return ret;
1135 }
1136
1137 struct if_check_data {
1138 struct device *dev;
1139 int pending;
1140 int ret;
1141 };
1142
1143 #ifndef IFF_LOWER_UP
1144 #define IFF_LOWER_UP 0x10000
1145 #endif
1146
1147 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1148 {
1149 struct nlmsghdr *nh = nlmsg_hdr(msg);
1150 struct ifinfomsg *ifi = NLMSG_DATA(nh);
1151 struct if_check_data *chk = (struct if_check_data *)arg;
1152
1153 if (nh->nlmsg_type != RTM_NEWLINK)
1154 return NL_SKIP;
1155
1156 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1157 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1158
1159 return NL_OK;
1160 }
1161
1162 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1163 {
1164 struct if_check_data *chk = (struct if_check_data *)arg;
1165 chk->pending = 0;
1166 return NL_STOP;
1167 }
1168
1169 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1170 {
1171 struct if_check_data *chk = (struct if_check_data *)arg;
1172
1173 device_set_present(chk->dev, false);
1174 device_set_link(chk->dev, false);
1175 chk->pending = err->error;
1176
1177 return NL_STOP;
1178 }
1179
1180 int system_if_check(struct device *dev)
1181 {
1182 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1183 struct nl_msg *msg;
1184 struct ifinfomsg ifi = {
1185 .ifi_family = AF_UNSPEC,
1186 .ifi_index = 0,
1187 };
1188 struct if_check_data chk = {
1189 .dev = dev,
1190 .pending = 1,
1191 };
1192 int ret = 1;
1193
1194 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1195 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1196 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1197 goto out;
1198
1199 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1200 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1201 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1202
1203 nl_send_auto_complete(sock_rtnl, msg);
1204 while (chk.pending > 0)
1205 nl_recvmsgs(sock_rtnl, cb);
1206
1207 nlmsg_free(msg);
1208 ret = chk.pending;
1209
1210 out:
1211 nl_cb_put(cb);
1212 return ret;
1213 }
1214
1215 struct device *
1216 system_if_get_parent(struct device *dev)
1217 {
1218 char buf[64], *devname;
1219 int ifindex, iflink, len;
1220 FILE *f;
1221
1222 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1223 f = fopen(buf, "r");
1224 if (!f)
1225 return NULL;
1226
1227 len = fread(buf, 1, sizeof(buf) - 1, f);
1228 fclose(f);
1229
1230 if (len <= 0)
1231 return NULL;
1232
1233 buf[len] = 0;
1234 iflink = strtoul(buf, NULL, 0);
1235 ifindex = system_if_resolve(dev);
1236 if (!iflink || iflink == ifindex)
1237 return NULL;
1238
1239 devname = if_indextoname(iflink, buf);
1240 if (!devname)
1241 return NULL;
1242
1243 return device_get(devname, true);
1244 }
1245
1246 static bool
1247 read_string_file(int dir_fd, const char *file, char *buf, int len)
1248 {
1249 bool ret = false;
1250 char *c;
1251 int fd;
1252
1253 fd = openat(dir_fd, file, O_RDONLY);
1254 if (fd < 0)
1255 return false;
1256
1257 retry:
1258 len = read(fd, buf, len - 1);
1259 if (len < 0) {
1260 if (errno == EINTR)
1261 goto retry;
1262 } else if (len > 0) {
1263 buf[len] = 0;
1264
1265 c = strchr(buf, '\n');
1266 if (c)
1267 *c = 0;
1268
1269 ret = true;
1270 }
1271
1272 close(fd);
1273
1274 return ret;
1275 }
1276
1277 static bool
1278 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1279 {
1280 char buf[64];
1281 bool ret = false;
1282
1283 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1284 if (ret)
1285 *val = strtoull(buf, NULL, 0);
1286
1287 return ret;
1288 }
1289
1290 /* Assume advertised flags == supported flags */
1291 static const struct {
1292 uint32_t mask;
1293 const char *name;
1294 } ethtool_link_modes[] = {
1295 { ADVERTISED_10baseT_Half, "10H" },
1296 { ADVERTISED_10baseT_Full, "10F" },
1297 { ADVERTISED_100baseT_Half, "100H" },
1298 { ADVERTISED_100baseT_Full, "100F" },
1299 { ADVERTISED_1000baseT_Half, "1000H" },
1300 { ADVERTISED_1000baseT_Full, "1000F" },
1301 };
1302
1303 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1304 {
1305 int i;
1306 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1307 if (mask & ethtool_link_modes[i].mask)
1308 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1309 }
1310 }
1311
1312 bool
1313 system_if_force_external(const char *ifname)
1314 {
1315 char buf[64];
1316 struct stat s;
1317
1318 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1319 return stat(buf, &s) == 0;
1320 }
1321
1322 int
1323 system_if_dump_info(struct device *dev, struct blob_buf *b)
1324 {
1325 struct ethtool_cmd ecmd;
1326 struct ifreq ifr;
1327 char buf[64], *s;
1328 void *c;
1329 int dir_fd;
1330
1331 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1332 dir_fd = open(buf, O_DIRECTORY);
1333
1334 memset(&ecmd, 0, sizeof(ecmd));
1335 memset(&ifr, 0, sizeof(ifr));
1336 strcpy(ifr.ifr_name, dev->ifname);
1337 ifr.ifr_data = (caddr_t) &ecmd;
1338 ecmd.cmd = ETHTOOL_GSET;
1339
1340 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1341 c = blobmsg_open_array(b, "link-advertising");
1342 system_add_link_modes(b, ecmd.advertising);
1343 blobmsg_close_array(b, c);
1344
1345 c = blobmsg_open_array(b, "link-supported");
1346 system_add_link_modes(b, ecmd.supported);
1347 blobmsg_close_array(b, c);
1348
1349 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1350 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1351 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1352 blobmsg_add_string_buffer(b);
1353 }
1354
1355 close(dir_fd);
1356 return 0;
1357 }
1358
1359 int
1360 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1361 {
1362 const char *const counters[] = {
1363 "collisions", "rx_frame_errors", "tx_compressed",
1364 "multicast", "rx_length_errors", "tx_dropped",
1365 "rx_bytes", "rx_missed_errors", "tx_errors",
1366 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1367 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1368 "rx_dropped", "tx_aborted_errors", "tx_packets",
1369 "rx_errors", "tx_bytes", "tx_window_errors",
1370 "rx_fifo_errors", "tx_carrier_errors",
1371 };
1372 char buf[64];
1373 int stats_dir;
1374 int i;
1375 uint64_t val = 0;
1376
1377 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1378 stats_dir = open(buf, O_DIRECTORY);
1379 if (stats_dir < 0)
1380 return -1;
1381
1382 for (i = 0; i < ARRAY_SIZE(counters); i++)
1383 if (read_uint64_file(stats_dir, counters[i], &val))
1384 blobmsg_add_u64(b, counters[i], val);
1385
1386 close(stats_dir);
1387 return 0;
1388 }
1389
1390 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1391 {
1392 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1393 int alen = v4 ? 4 : 16;
1394 unsigned int flags = 0;
1395 struct ifaddrmsg ifa = {
1396 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1397 .ifa_prefixlen = addr->mask,
1398 .ifa_index = dev->ifindex,
1399 };
1400
1401 struct nl_msg *msg;
1402 if (cmd == RTM_NEWADDR)
1403 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1404
1405 msg = nlmsg_alloc_simple(cmd, flags);
1406 if (!msg)
1407 return -1;
1408
1409 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1410 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1411 if (v4) {
1412 if (addr->broadcast)
1413 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1414 if (addr->point_to_point)
1415 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1416 } else {
1417 time_t now = system_get_rtime();
1418 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1419
1420 if (addr->preferred_until) {
1421 int64_t preferred = addr->preferred_until - now;
1422 if (preferred < 0)
1423 preferred = 0;
1424 else if (preferred > UINT32_MAX)
1425 preferred = UINT32_MAX;
1426
1427 cinfo.ifa_prefered = preferred;
1428 }
1429
1430 if (addr->valid_until) {
1431 int64_t valid = addr->valid_until - now;
1432 if (valid <= 0)
1433 return -1;
1434 else if (valid > UINT32_MAX)
1435 valid = UINT32_MAX;
1436
1437 cinfo.ifa_valid = valid;
1438 }
1439
1440 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1441 }
1442
1443 return system_rtnl_call(msg);
1444 }
1445
1446 int system_add_address(struct device *dev, struct device_addr *addr)
1447 {
1448 return system_addr(dev, addr, RTM_NEWADDR);
1449 }
1450
1451 int system_del_address(struct device *dev, struct device_addr *addr)
1452 {
1453 return system_addr(dev, addr, RTM_DELADDR);
1454 }
1455
1456 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1457 {
1458 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1459 bool have_gw;
1460 unsigned int flags = 0;
1461
1462 if (alen == 4)
1463 have_gw = !!route->nexthop.in.s_addr;
1464 else
1465 have_gw = route->nexthop.in6.s6_addr32[0] ||
1466 route->nexthop.in6.s6_addr32[1] ||
1467 route->nexthop.in6.s6_addr32[2] ||
1468 route->nexthop.in6.s6_addr32[3];
1469
1470 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1471 ? route->table : RT_TABLE_MAIN;
1472
1473 struct rtmsg rtm = {
1474 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1475 .rtm_dst_len = route->mask,
1476 .rtm_src_len = route->sourcemask,
1477 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1478 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1479 .rtm_scope = RT_SCOPE_NOWHERE,
1480 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1481 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1482 };
1483 struct nl_msg *msg;
1484
1485 if (cmd == RTM_NEWROUTE) {
1486 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1487
1488 if (!dev) { // Add null-route
1489 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1490 rtm.rtm_type = RTN_UNREACHABLE;
1491 }
1492 else
1493 rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1494 }
1495
1496 if (route->flags & DEVROUTE_TYPE) {
1497 rtm.rtm_type = route->type;
1498 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1499 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1500 rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1501 rtm.rtm_table = RT_TABLE_LOCAL;
1502 }
1503
1504 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1505 rtm.rtm_scope = RT_SCOPE_HOST;
1506 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1507 rtm.rtm_type == RTN_ANYCAST) {
1508 rtm.rtm_scope = RT_SCOPE_LINK;
1509 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1510 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1511 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1512 dev = NULL;
1513 }
1514 }
1515
1516 msg = nlmsg_alloc_simple(cmd, flags);
1517 if (!msg)
1518 return -1;
1519
1520 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1521
1522 if (route->mask)
1523 nla_put(msg, RTA_DST, alen, &route->addr);
1524
1525 if (route->sourcemask) {
1526 if (rtm.rtm_family == AF_INET)
1527 nla_put(msg, RTA_PREFSRC, alen, &route->source);
1528 else
1529 nla_put(msg, RTA_SRC, alen, &route->source);
1530 }
1531
1532 if (route->metric > 0)
1533 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1534
1535 if (have_gw)
1536 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1537
1538 if (dev)
1539 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1540
1541 if (table >= 256)
1542 nla_put_u32(msg, RTA_TABLE, table);
1543
1544 if (route->flags & DEVROUTE_MTU) {
1545 struct nlattr *metrics;
1546
1547 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1548 goto nla_put_failure;
1549
1550 nla_put_u32(msg, RTAX_MTU, route->mtu);
1551
1552 nla_nest_end(msg, metrics);
1553 }
1554
1555 return system_rtnl_call(msg);
1556
1557 nla_put_failure:
1558 nlmsg_free(msg);
1559 return -ENOMEM;
1560 }
1561
1562 int system_add_route(struct device *dev, struct device_route *route)
1563 {
1564 return system_rt(dev, route, RTM_NEWROUTE);
1565 }
1566
1567 int system_del_route(struct device *dev, struct device_route *route)
1568 {
1569 return system_rt(dev, route, RTM_DELROUTE);
1570 }
1571
1572 int system_flush_routes(void)
1573 {
1574 const char *names[] = {
1575 "/proc/sys/net/ipv4/route/flush",
1576 "/proc/sys/net/ipv6/route/flush"
1577 };
1578 int fd, i;
1579
1580 for (i = 0; i < ARRAY_SIZE(names); i++) {
1581 fd = open(names[i], O_WRONLY);
1582 if (fd < 0)
1583 continue;
1584
1585 if (write(fd, "-1", 2)) {}
1586 close(fd);
1587 }
1588 return 0;
1589 }
1590
1591 bool system_resolve_rt_type(const char *type, unsigned int *id)
1592 {
1593 return system_rtn_aton(type, id);
1594 }
1595
1596 bool system_resolve_rt_table(const char *name, unsigned int *id)
1597 {
1598 FILE *f;
1599 char *e, buf[128];
1600 unsigned int n, table = RT_TABLE_UNSPEC;
1601
1602 /* first try to parse table as number */
1603 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1604 table = n;
1605
1606 /* handle well known aliases */
1607 else if (!strcmp(name, "default"))
1608 table = RT_TABLE_DEFAULT;
1609 else if (!strcmp(name, "main"))
1610 table = RT_TABLE_MAIN;
1611 else if (!strcmp(name, "local"))
1612 table = RT_TABLE_LOCAL;
1613 else if (!strcmp(name, "prelocal"))
1614 table = RT_TABLE_PRELOCAL;
1615
1616 /* try to look up name in /etc/iproute2/rt_tables */
1617 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1618 {
1619 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1620 {
1621 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1622 continue;
1623
1624 n = strtoul(e, NULL, 10);
1625 e = strtok(NULL, " \t\n");
1626
1627 if (e && !strcmp(e, name))
1628 {
1629 table = n;
1630 break;
1631 }
1632 }
1633
1634 fclose(f);
1635 }
1636
1637 if (table == RT_TABLE_UNSPEC)
1638 return false;
1639
1640 *id = table;
1641 return true;
1642 }
1643
1644 bool system_is_default_rt_table(unsigned int id)
1645 {
1646 return (id == RT_TABLE_MAIN);
1647 }
1648
1649 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1650 {
1651 char *e;
1652 unsigned int n;
1653
1654 if (!strcmp(filter, "strict"))
1655 n = 1;
1656 else if (!strcmp(filter, "loose"))
1657 n = 2;
1658 else {
1659 n = strtoul(filter, &e, 0);
1660 if (*e || e == filter || n > 2)
1661 return false;
1662 }
1663
1664 *id = n;
1665 return true;
1666 }
1667
1668 static int system_iprule(struct iprule *rule, int cmd)
1669 {
1670 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1671
1672 struct nl_msg *msg;
1673 struct rtmsg rtm = {
1674 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1675 .rtm_protocol = RTPROT_STATIC,
1676 .rtm_scope = RT_SCOPE_UNIVERSE,
1677 .rtm_table = RT_TABLE_UNSPEC,
1678 .rtm_type = RTN_UNSPEC,
1679 .rtm_flags = 0,
1680 };
1681
1682 if (cmd == RTM_NEWRULE) {
1683 rtm.rtm_type = RTN_UNICAST;
1684 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1685 }
1686
1687 if (rule->invert)
1688 rtm.rtm_flags |= FIB_RULE_INVERT;
1689
1690 if (rule->flags & IPRULE_SRC)
1691 rtm.rtm_src_len = rule->src_mask;
1692
1693 if (rule->flags & IPRULE_DEST)
1694 rtm.rtm_dst_len = rule->dest_mask;
1695
1696 if (rule->flags & IPRULE_TOS)
1697 rtm.rtm_tos = rule->tos;
1698
1699 if (rule->flags & IPRULE_LOOKUP) {
1700 if (rule->lookup < 256)
1701 rtm.rtm_table = rule->lookup;
1702 }
1703
1704 if (rule->flags & IPRULE_ACTION)
1705 rtm.rtm_type = rule->action;
1706 else if (rule->flags & IPRULE_GOTO)
1707 rtm.rtm_type = FR_ACT_GOTO;
1708 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1709 rtm.rtm_type = FR_ACT_NOP;
1710
1711 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1712
1713 if (!msg)
1714 return -1;
1715
1716 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1717
1718 if (rule->flags & IPRULE_IN)
1719 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1720
1721 if (rule->flags & IPRULE_OUT)
1722 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1723
1724 if (rule->flags & IPRULE_SRC)
1725 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1726
1727 if (rule->flags & IPRULE_DEST)
1728 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1729
1730 if (rule->flags & IPRULE_PRIORITY)
1731 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1732 else if (cmd == RTM_NEWRULE)
1733 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1734
1735 if (rule->flags & IPRULE_FWMARK)
1736 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1737
1738 if (rule->flags & IPRULE_FWMASK)
1739 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1740
1741 if (rule->flags & IPRULE_LOOKUP) {
1742 if (rule->lookup >= 256)
1743 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1744 }
1745
1746 if (rule->flags & IPRULE_GOTO)
1747 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1748
1749 return system_rtnl_call(msg);
1750 }
1751
1752 int system_add_iprule(struct iprule *rule)
1753 {
1754 return system_iprule(rule, RTM_NEWRULE);
1755 }
1756
1757 int system_del_iprule(struct iprule *rule)
1758 {
1759 return system_iprule(rule, RTM_DELRULE);
1760 }
1761
1762 int system_flush_iprules(void)
1763 {
1764 int rv = 0;
1765 struct iprule rule;
1766
1767 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1768 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1769
1770 memset(&rule, 0, sizeof(rule));
1771
1772
1773 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1774
1775 rule.priority = 0;
1776 rule.lookup = RT_TABLE_PRELOCAL;
1777 rv |= system_iprule(&rule, RTM_NEWRULE);
1778
1779 rule.priority = 1;
1780 rule.lookup = RT_TABLE_LOCAL;
1781 rv |= system_iprule(&rule, RTM_NEWRULE);
1782
1783 rule.priority = 32766;
1784 rule.lookup = RT_TABLE_MAIN;
1785 rv |= system_iprule(&rule, RTM_NEWRULE);
1786
1787 rule.priority = 32767;
1788 rule.lookup = RT_TABLE_DEFAULT;
1789 rv |= system_iprule(&rule, RTM_NEWRULE);
1790
1791
1792 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1793
1794 rule.priority = 0;
1795 rule.lookup = RT_TABLE_PRELOCAL;
1796 rv |= system_iprule(&rule, RTM_NEWRULE);
1797
1798 rule.priority = 1;
1799 rule.lookup = RT_TABLE_LOCAL;
1800 rv |= system_iprule(&rule, RTM_NEWRULE);
1801
1802 rule.priority = 32766;
1803 rule.lookup = RT_TABLE_MAIN;
1804 rv |= system_iprule(&rule, RTM_NEWRULE);
1805
1806 return rv;
1807 }
1808
1809 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1810 {
1811 return system_rtn_aton(action, id);
1812 }
1813
1814 time_t system_get_rtime(void)
1815 {
1816 struct timespec ts;
1817 struct timeval tv;
1818
1819 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1820 return ts.tv_sec;
1821
1822 if (gettimeofday(&tv, NULL) == 0)
1823 return tv.tv_sec;
1824
1825 return 0;
1826 }
1827
1828 #ifndef IP_DF
1829 #define IP_DF 0x4000
1830 #endif
1831
1832 static int tunnel_ioctl(const char *name, int cmd, void *p)
1833 {
1834 struct ifreq ifr;
1835
1836 memset(&ifr, 0, sizeof(ifr));
1837 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1838 ifr.ifr_ifru.ifru_data = p;
1839 return ioctl(sock_ioctl, cmd, &ifr);
1840 }
1841
1842 #ifdef IFLA_IPTUN_MAX
1843 #define IP6_FLOWINFO_TCLASS htonl(0x0FF00000)
1844 static int system_add_gre_tunnel(const char *name, const char *kind,
1845 const unsigned int link, struct blob_attr **tb, bool v6)
1846 {
1847 struct nl_msg *nlm;
1848 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1849 struct blob_attr *cur;
1850 uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1851 uint16_t iflags = 0, oflags = 0;
1852 uint8_t tos = 0;
1853 int ret = 0, ttl = 64;
1854
1855 nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1856 if (!nlm)
1857 return -1;
1858
1859 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1860 nla_put_string(nlm, IFLA_IFNAME, name);
1861
1862 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1863 if (!linkinfo) {
1864 ret = -ENOMEM;
1865 goto failure;
1866 }
1867
1868 nla_put_string(nlm, IFLA_INFO_KIND, kind);
1869 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1870 if (!infodata) {
1871 ret = -ENOMEM;
1872 goto failure;
1873 }
1874
1875 if (link)
1876 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1877
1878 if ((cur = tb[TUNNEL_ATTR_TTL]))
1879 ttl = blobmsg_get_u32(cur);
1880
1881 nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1882
1883 if ((cur = tb[TUNNEL_ATTR_TOS])) {
1884 char *str = blobmsg_get_string(cur);
1885 if (strcmp(str, "inherit")) {
1886 unsigned uval;
1887
1888 if (!system_tos_aton(str, &uval)) {
1889 ret = -EINVAL;
1890 goto failure;
1891 }
1892
1893 if (v6)
1894 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1895 else
1896 tos = uval;
1897 } else {
1898 if (v6)
1899 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1900 else
1901 tos = 1;
1902 }
1903 }
1904
1905 if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1906 uint8_t icsum, ocsum, iseqno, oseqno;
1907 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1908 &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1909 ret = -EINVAL;
1910 goto failure;
1911 }
1912
1913 if (ikey)
1914 iflags |= GRE_KEY;
1915
1916 if (okey)
1917 oflags |= GRE_KEY;
1918
1919 if (icsum)
1920 iflags |= GRE_CSUM;
1921
1922 if (ocsum)
1923 oflags |= GRE_CSUM;
1924
1925 if (iseqno)
1926 iflags |= GRE_SEQ;
1927
1928 if (oseqno)
1929 oflags |= GRE_SEQ;
1930 }
1931
1932 if (v6) {
1933 struct in6_addr in6buf;
1934 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1935 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1936 ret = -EINVAL;
1937 goto failure;
1938 }
1939 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1940 }
1941
1942 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1943 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1944 ret = -EINVAL;
1945 goto failure;
1946 }
1947 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1948 }
1949 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1950
1951 if (flowinfo)
1952 nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1953
1954 if (flags)
1955 nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
1956 } else {
1957 struct in_addr inbuf;
1958 bool set_df = true;
1959
1960 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1961 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1962 ret = -EINVAL;
1963 goto failure;
1964 }
1965 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
1966 }
1967
1968 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1969 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1970 ret = -EINVAL;
1971 goto failure;
1972 }
1973 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
1974
1975 if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
1976 if (!okey) {
1977 okey = inbuf.s_addr;
1978 oflags |= GRE_KEY;
1979 }
1980
1981 if (!ikey) {
1982 ikey = inbuf.s_addr;
1983 iflags |= GRE_KEY;
1984 }
1985 }
1986 }
1987
1988 if ((cur = tb[TUNNEL_ATTR_DF]))
1989 set_df = blobmsg_get_bool(cur);
1990
1991 /* ttl !=0 and nopmtudisc are incompatible */
1992 if (ttl && !set_df) {
1993 ret = -EINVAL;
1994 goto failure;
1995 }
1996
1997 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
1998
1999 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
2000 }
2001
2002 if (oflags)
2003 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
2004
2005 if (iflags)
2006 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
2007
2008 if (okey)
2009 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
2010
2011 if (ikey)
2012 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
2013
2014 nla_nest_end(nlm, infodata);
2015 nla_nest_end(nlm, linkinfo);
2016
2017 return system_rtnl_call(nlm);
2018
2019 failure:
2020 nlmsg_free(nlm);
2021 return ret;
2022 }
2023 #endif
2024
2025 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
2026 {
2027 struct blob_attr *cur;
2028 bool set_df = true;
2029 struct ip_tunnel_parm p = {
2030 .link = link,
2031 .iph = {
2032 .version = 4,
2033 .ihl = 5,
2034 .protocol = proto,
2035 }
2036 };
2037
2038 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
2039 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
2040 return -EINVAL;
2041
2042 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
2043 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
2044 return -EINVAL;
2045
2046 if ((cur = tb[TUNNEL_ATTR_DF]))
2047 set_df = blobmsg_get_bool(cur);
2048
2049 if ((cur = tb[TUNNEL_ATTR_TTL]))
2050 p.iph.ttl = blobmsg_get_u32(cur);
2051
2052 if ((cur = tb[TUNNEL_ATTR_TOS])) {
2053 char *str = blobmsg_get_string(cur);
2054 if (strcmp(str, "inherit")) {
2055 unsigned uval;
2056
2057 if (!system_tos_aton(str, &uval))
2058 return -EINVAL;
2059
2060 p.iph.tos = uval;
2061 } else
2062 p.iph.tos = 1;
2063 }
2064
2065 p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2066 /* ttl !=0 and nopmtudisc are incompatible */
2067 if (p.iph.ttl && p.iph.frag_off == 0)
2068 return -EINVAL;
2069
2070 strncpy(p.name, name, sizeof(p.name));
2071
2072 switch (p.iph.protocol) {
2073 case IPPROTO_IPIP:
2074 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2075 case IPPROTO_IPV6:
2076 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2077 default:
2078 break;
2079 }
2080 return -1;
2081 }
2082
2083 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2084 {
2085 struct blob_attr *cur;
2086 const char *str;
2087
2088 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2089 return -EINVAL;
2090 str = blobmsg_data(cur);
2091
2092 if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2093 !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2094 return system_link_del(name);
2095 else
2096 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2097 }
2098
2099 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2100 {
2101 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2102
2103 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2104 blob_data(attr), blob_len(attr));
2105
2106 return __system_del_ip_tunnel(name, tb);
2107 }
2108
2109 int system_update_ipv6_mtu(struct device *dev, int mtu)
2110 {
2111 int ret = -1;
2112 char buf[64];
2113 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2114 dev->ifname);
2115
2116 int fd = open(buf, O_RDWR);
2117 ssize_t len = read(fd, buf, sizeof(buf) - 1);
2118 if (len < 0)
2119 goto out;
2120
2121 buf[len] = 0;
2122 ret = atoi(buf);
2123
2124 if (!mtu || ret <= mtu)
2125 goto out;
2126
2127 lseek(fd, 0, SEEK_SET);
2128 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2129 ret = -1;
2130
2131 out:
2132 close(fd);
2133 return ret;
2134 }
2135
2136 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2137 {
2138 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2139 struct blob_attr *cur;
2140 const char *str;
2141
2142 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2143 blob_data(attr), blob_len(attr));
2144
2145 __system_del_ip_tunnel(name, tb);
2146
2147 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2148 return -EINVAL;
2149 str = blobmsg_data(cur);
2150
2151 unsigned int ttl = 0;
2152 if ((cur = tb[TUNNEL_ATTR_TTL])) {
2153 ttl = blobmsg_get_u32(cur);
2154 if (ttl > 255)
2155 return -EINVAL;
2156 }
2157
2158 unsigned int link = 0;
2159 if ((cur = tb[TUNNEL_ATTR_LINK])) {
2160 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2161 if (!iface)
2162 return -EINVAL;
2163
2164 if (iface->l3_dev.dev)
2165 link = iface->l3_dev.dev->ifindex;
2166 }
2167
2168 if (!strcmp(str, "sit")) {
2169 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2170 return -1;
2171
2172 #ifdef SIOCADD6RD
2173 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2174 unsigned int mask;
2175 struct ip_tunnel_6rd p6;
2176
2177 memset(&p6, 0, sizeof(p6));
2178
2179 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2180 &p6.prefix, &mask) || mask > 128)
2181 return -EINVAL;
2182 p6.prefixlen = mask;
2183
2184 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2185 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2186 &p6.relay_prefix, &mask) || mask > 32)
2187 return -EINVAL;
2188 p6.relay_prefixlen = mask;
2189 }
2190
2191 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2192 __system_del_ip_tunnel(name, tb);
2193 return -1;
2194 }
2195 }
2196 #endif
2197 #ifdef IFLA_IPTUN_MAX
2198 } else if (!strcmp(str, "ipip6")) {
2199 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2200 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2201 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2202 int ret = 0;
2203
2204 if (!nlm)
2205 return -1;
2206
2207 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2208 nla_put_string(nlm, IFLA_IFNAME, name);
2209
2210 if (link)
2211 nla_put_u32(nlm, IFLA_LINK, link);
2212
2213 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2214 if (!linkinfo) {
2215 ret = -ENOMEM;
2216 goto failure;
2217 }
2218 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2219 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2220 if (!infodata) {
2221 ret = -ENOMEM;
2222 goto failure;
2223 }
2224
2225 if (link)
2226 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2227
2228 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2229 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2230 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2231
2232 struct in6_addr in6buf;
2233 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2234 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2235 ret = -EINVAL;
2236 goto failure;
2237 }
2238 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2239 }
2240
2241 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2242 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2243 ret = -EINVAL;
2244 goto failure;
2245 }
2246 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2247 }
2248
2249 #ifdef IFLA_IPTUN_FMR_MAX
2250 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2251 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2252
2253 struct blob_attr *fmr;
2254 unsigned rem, fmrcnt = 0;
2255 blobmsg_for_each_attr(fmr, cur, rem) {
2256 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2257 continue;
2258
2259 unsigned ip4len, ip6len, ealen, offset = 6;
2260 char ip6buf[48];
2261 char ip4buf[16];
2262
2263 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2264 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2265 ret = -EINVAL;
2266 goto failure;
2267 }
2268
2269 struct in6_addr ip6prefix;
2270 struct in_addr ip4prefix;
2271 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2272 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2273 ret = -EINVAL;
2274 goto failure;
2275 }
2276
2277 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2278
2279 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2280 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2281 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2282 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2283 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2284 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2285
2286 nla_nest_end(nlm, rule);
2287 }
2288
2289 nla_nest_end(nlm, fmrs);
2290 }
2291 #endif
2292
2293 nla_nest_end(nlm, infodata);
2294 nla_nest_end(nlm, linkinfo);
2295
2296 return system_rtnl_call(nlm);
2297 failure:
2298 nlmsg_free(nlm);
2299 return ret;
2300 } else if (!strcmp(str, "greip")) {
2301 return system_add_gre_tunnel(name, "gre", link, tb, false);
2302 } else if (!strcmp(str, "gretapip")) {
2303 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2304 } else if (!strcmp(str, "greip6")) {
2305 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2306 } else if (!strcmp(str, "gretapip6")) {
2307 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2308 #endif
2309 } else if (!strcmp(str, "ipip")) {
2310 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2311 }
2312 else
2313 return -EINVAL;
2314
2315 return 0;
2316 }