linux: add compat-define for IFA_F_NOPREFIXROUTE
[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 #ifndef IFA_F_NOPREFIXROUTE
51 #define IFA_F_NOPREFIXROUTE 0x200
52 #endif
53
54 #include <string.h>
55 #include <fcntl.h>
56 #include <glob.h>
57 #include <time.h>
58 #include <unistd.h>
59
60 #include <netlink/msg.h>
61 #include <netlink/attr.h>
62 #include <netlink/socket.h>
63 #include <libubox/uloop.h>
64
65 #include "netifd.h"
66 #include "device.h"
67 #include "system.h"
68
69 struct event_socket {
70 struct uloop_fd uloop;
71 struct nl_sock *sock;
72 int bufsize;
73 };
74
75 static int sock_ioctl = -1;
76 static struct nl_sock *sock_rtnl = NULL;
77
78 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
79 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
80
81 static char dev_buf[256];
82
83 static void
84 handler_nl_event(struct uloop_fd *u, unsigned int events)
85 {
86 struct event_socket *ev = container_of(u, struct event_socket, uloop);
87 int err;
88 socklen_t errlen = sizeof(err);
89
90 if (!u->error) {
91 nl_recvmsgs_default(ev->sock);
92 return;
93 }
94
95 if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
96 goto abort;
97
98 switch(err) {
99 case ENOBUFS:
100 // Increase rx buffer size on netlink socket
101 ev->bufsize *= 2;
102 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
103 goto abort;
104
105 // Request full dump since some info got dropped
106 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
107 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
108 break;
109
110 default:
111 goto abort;
112 }
113 u->error = false;
114 return;
115
116 abort:
117 uloop_fd_delete(&ev->uloop);
118 return;
119 }
120
121 static struct nl_sock *
122 create_socket(int protocol, int groups)
123 {
124 struct nl_sock *sock;
125
126 sock = nl_socket_alloc();
127 if (!sock)
128 return NULL;
129
130 if (groups)
131 nl_join_groups(sock, groups);
132
133 if (nl_connect(sock, protocol))
134 return NULL;
135
136 return sock;
137 }
138
139 static bool
140 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
141 uloop_fd_handler cb, int flags)
142 {
143 ev->sock = create_socket(protocol, groups);
144 if (!ev->sock)
145 return false;
146
147 ev->uloop.fd = nl_socket_get_fd(ev->sock);
148 ev->uloop.cb = cb;
149 if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
150 return false;
151
152 return true;
153 }
154
155 static bool
156 create_event_socket(struct event_socket *ev, int protocol,
157 int (*cb)(struct nl_msg *msg, void *arg))
158 {
159 if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
160 return false;
161
162 // Install the valid custom callback handler
163 nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
164
165 // Disable sequence number checking on event sockets
166 nl_socket_disable_seq_check(ev->sock);
167
168 // Increase rx buffer size to 65K on event sockets
169 ev->bufsize = 65535;
170 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
171 return false;
172
173 return true;
174 }
175
176 static bool
177 system_rtn_aton(const char *src, unsigned int *dst)
178 {
179 char *e;
180 unsigned int n;
181
182 if (!strcmp(src, "local"))
183 n = RTN_LOCAL;
184 else if (!strcmp(src, "nat"))
185 n = RTN_NAT;
186 else if (!strcmp(src, "broadcast"))
187 n = RTN_BROADCAST;
188 else if (!strcmp(src, "anycast"))
189 n = RTN_ANYCAST;
190 else if (!strcmp(src, "multicast"))
191 n = RTN_MULTICAST;
192 else if (!strcmp(src, "prohibit"))
193 n = RTN_PROHIBIT;
194 else if (!strcmp(src, "unreachable"))
195 n = RTN_UNREACHABLE;
196 else if (!strcmp(src, "blackhole"))
197 n = RTN_BLACKHOLE;
198 else if (!strcmp(src, "xresolve"))
199 n = RTN_XRESOLVE;
200 else if (!strcmp(src, "unicast"))
201 n = RTN_UNICAST;
202 else if (!strcmp(src, "throw"))
203 n = RTN_THROW;
204 else if (!strcmp(src, "failed_policy"))
205 n = RTN_FAILED_POLICY;
206 else {
207 n = strtoul(src, &e, 0);
208 if (!e || *e || e == src || n > 255)
209 return false;
210 }
211
212 *dst = n;
213 return true;
214 }
215
216 static bool
217 system_tos_aton(const char *src, unsigned *dst)
218 {
219 char *e;
220
221 *dst = strtoul(src, &e, 16);
222 if (e == src || *e || *dst > 255)
223 return false;
224
225 return true;
226 }
227
228 int system_init(void)
229 {
230 static struct event_socket rtnl_event;
231 static struct event_socket hotplug_event;
232
233 sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
234 system_fd_set_cloexec(sock_ioctl);
235
236 // Prepare socket for routing / address control
237 sock_rtnl = create_socket(NETLINK_ROUTE, 0);
238 if (!sock_rtnl)
239 return -1;
240
241 if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
242 return -1;
243
244 if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
245 handle_hotplug_event, 0))
246 return -1;
247
248 // Receive network link events form kernel
249 nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
250
251 return 0;
252 }
253
254 static void system_set_sysctl(const char *path, const char *val)
255 {
256 int fd;
257
258 fd = open(path, O_WRONLY);
259 if (fd < 0)
260 return;
261
262 if (write(fd, val, strlen(val))) {}
263 close(fd);
264 }
265
266 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
267 {
268 snprintf(dev_buf, sizeof(dev_buf), path, device);
269 system_set_sysctl(dev_buf, val);
270 }
271
272 static void system_set_disable_ipv6(struct device *dev, const char *val)
273 {
274 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
275 }
276
277 static void system_set_rpfilter(struct device *dev, const char *val)
278 {
279 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter", dev->ifname, val);
280 }
281
282 static void system_set_acceptlocal(struct device *dev, const char *val)
283 {
284 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local", dev->ifname, val);
285 }
286
287 static void system_set_igmpversion(struct device *dev, const char *val)
288 {
289 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version", dev->ifname, val);
290 }
291
292 static void system_set_mldversion(struct device *dev, const char *val)
293 {
294 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version", dev->ifname, val);
295 }
296
297 static void system_set_neigh4reachabletime(struct device *dev, const char *val)
298 {
299 system_set_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms", dev->ifname, val);
300 }
301
302 static void system_set_neigh6reachabletime(struct device *dev, const char *val)
303 {
304 system_set_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms", dev->ifname, val);
305 }
306
307 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
308 {
309 int fd = -1, ret = -1;
310
311 fd = open(path, O_RDONLY);
312 if (fd < 0)
313 goto out;
314
315 ssize_t len = read(fd, buf, buf_sz - 1);
316 if (len < 0)
317 goto out;
318
319 ret = buf[len] = 0;
320
321 out:
322 if (fd >= 0)
323 close(fd);
324
325 return ret;
326 }
327
328 static int
329 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
330 {
331 snprintf(dev_buf, sizeof(dev_buf), path, device);
332 return system_get_sysctl(dev_buf, buf, buf_sz);
333 }
334
335 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
336 {
337 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
338 dev->ifname, buf, buf_sz);
339 }
340
341 static int system_get_rpfilter(struct device *dev, char *buf, const size_t buf_sz)
342 {
343 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter",
344 dev->ifname, buf, buf_sz);
345 }
346
347 static int system_get_acceptlocal(struct device *dev, char *buf, const size_t buf_sz)
348 {
349 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local",
350 dev->ifname, buf, buf_sz);
351 }
352
353 static int system_get_igmpversion(struct device *dev, char *buf, const size_t buf_sz)
354 {
355 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version",
356 dev->ifname, buf, buf_sz);
357 }
358
359 static int system_get_mldversion(struct device *dev, char *buf, const size_t buf_sz)
360 {
361 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version",
362 dev->ifname, buf, buf_sz);
363 }
364
365 static int system_get_neigh4reachabletime(struct device *dev, char *buf, const size_t buf_sz)
366 {
367 return system_get_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms",
368 dev->ifname, buf, buf_sz);
369 }
370
371 static int system_get_neigh6reachabletime(struct device *dev, char *buf, const size_t buf_sz)
372 {
373 return system_get_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms",
374 dev->ifname, buf, buf_sz);
375 }
376
377 // Evaluate netlink messages
378 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
379 {
380 struct nlmsghdr *nh = nlmsg_hdr(msg);
381 struct nlattr *nla[__IFLA_MAX];
382 int link_state = 0;
383 char buf[10];
384
385 if (nh->nlmsg_type != RTM_NEWLINK)
386 goto out;
387
388 nlmsg_parse(nh, sizeof(struct ifinfomsg), nla, __IFLA_MAX - 1, NULL);
389 if (!nla[IFLA_IFNAME])
390 goto out;
391
392 struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
393 if (!dev || dev->type->keep_link_status)
394 goto out;
395
396 if (!system_get_dev_sysctl("/sys/class/net/%s/carrier", dev->ifname, buf, sizeof(buf)))
397 link_state = strtoul(buf, NULL, 0);
398
399 device_set_link(dev, link_state ? true : false);
400
401 out:
402 return 0;
403 }
404
405 static void
406 handle_hotplug_msg(char *data, int size)
407 {
408 const char *subsystem = NULL, *interface = NULL;
409 char *cur, *end, *sep;
410 struct device *dev;
411 int skip;
412 bool add;
413
414 if (!strncmp(data, "add@", 4))
415 add = true;
416 else if (!strncmp(data, "remove@", 7))
417 add = false;
418 else
419 return;
420
421 skip = strlen(data) + 1;
422 end = data + size;
423
424 for (cur = data + skip; cur < end; cur += skip) {
425 skip = strlen(cur) + 1;
426
427 sep = strchr(cur, '=');
428 if (!sep)
429 continue;
430
431 *sep = 0;
432 if (!strcmp(cur, "INTERFACE"))
433 interface = sep + 1;
434 else if (!strcmp(cur, "SUBSYSTEM")) {
435 subsystem = sep + 1;
436 if (strcmp(subsystem, "net") != 0)
437 return;
438 }
439 if (subsystem && interface)
440 goto found;
441 }
442 return;
443
444 found:
445 dev = device_get(interface, false);
446 if (!dev)
447 return;
448
449 if (dev->type != &simple_device_type)
450 return;
451
452 if (add && system_if_force_external(dev->ifname))
453 return;
454
455 device_set_present(dev, add);
456 }
457
458 static void
459 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
460 {
461 struct event_socket *ev = container_of(u, struct event_socket, uloop);
462 struct sockaddr_nl nla;
463 unsigned char *buf = NULL;
464 int size;
465
466 while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
467 if (nla.nl_pid == 0)
468 handle_hotplug_msg((char *) buf, size);
469
470 free(buf);
471 }
472 }
473
474 static int system_rtnl_call(struct nl_msg *msg)
475 {
476 int ret;
477
478 ret = nl_send_auto_complete(sock_rtnl, msg);
479 nlmsg_free(msg);
480
481 if (ret < 0)
482 return ret;
483
484 return nl_wait_for_ack(sock_rtnl);
485 }
486
487 int system_bridge_delbr(struct device *bridge)
488 {
489 return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
490 }
491
492 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
493 {
494 struct ifreq ifr;
495
496 memset(&ifr, 0, sizeof(ifr));
497 if (dev)
498 ifr.ifr_ifindex = dev->ifindex;
499 else
500 ifr.ifr_data = data;
501 strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
502 return ioctl(sock_ioctl, cmd, &ifr);
503 }
504
505 static bool system_is_bridge(const char *name, char *buf, int buflen)
506 {
507 struct stat st;
508
509 snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
510 if (stat(buf, &st) < 0)
511 return false;
512
513 return true;
514 }
515
516 static char *system_get_bridge(const char *name, char *buf, int buflen)
517 {
518 char *path;
519 ssize_t len = -1;
520 glob_t gl;
521
522 snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
523 if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
524 return NULL;
525
526 if (gl.gl_pathc > 0)
527 len = readlink(gl.gl_pathv[0], buf, buflen);
528
529 globfree(&gl);
530
531 if (len < 0)
532 return NULL;
533
534 buf[len] = 0;
535 path = strrchr(buf, '/');
536 if (!path)
537 return NULL;
538
539 return path + 1;
540 }
541
542 static void system_bridge_set_wireless(const char *bridge, const char *dev)
543 {
544 snprintf(dev_buf, sizeof(dev_buf),
545 "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
546 bridge, dev);
547 system_set_sysctl(dev_buf, "1");
548 }
549
550 int system_bridge_addif(struct device *bridge, struct device *dev)
551 {
552 char *oldbr;
553 int ret = 0;
554
555 oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
556 if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
557 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
558
559 if (dev->wireless)
560 system_bridge_set_wireless(bridge->ifname, dev->ifname);
561
562 return ret;
563 }
564
565 int system_bridge_delif(struct device *bridge, struct device *dev)
566 {
567 return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
568 }
569
570 int system_if_resolve(struct device *dev)
571 {
572 struct ifreq ifr;
573 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
574 if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
575 return ifr.ifr_ifindex;
576 else
577 return 0;
578 }
579
580 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
581 {
582 struct ifreq ifr;
583
584 memset(&ifr, 0, sizeof(ifr));
585 strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
586 ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
587 ifr.ifr_flags |= add;
588 ifr.ifr_flags &= ~rem;
589 return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
590 }
591
592 struct clear_data {
593 struct nl_msg *msg;
594 struct device *dev;
595 int type;
596 int size;
597 int af;
598 };
599
600
601 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
602 {
603 struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
604
605 return ifa->ifa_index == ifindex;
606 }
607
608 static bool check_route(struct nlmsghdr *hdr, int ifindex)
609 {
610 struct rtmsg *r = NLMSG_DATA(hdr);
611 struct nlattr *tb[__RTA_MAX];
612
613 if (r->rtm_protocol == RTPROT_KERNEL &&
614 r->rtm_family == AF_INET6)
615 return false;
616
617 nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
618 if (!tb[RTA_OIF])
619 return false;
620
621 return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
622 }
623
624 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
625 {
626 return true;
627 }
628
629 static int cb_clear_event(struct nl_msg *msg, void *arg)
630 {
631 struct clear_data *clr = arg;
632 struct nlmsghdr *hdr = nlmsg_hdr(msg);
633 bool (*cb)(struct nlmsghdr *, int ifindex);
634 int type;
635
636 switch(clr->type) {
637 case RTM_GETADDR:
638 type = RTM_DELADDR;
639 if (hdr->nlmsg_type != RTM_NEWADDR)
640 return NL_SKIP;
641
642 cb = check_ifaddr;
643 break;
644 case RTM_GETROUTE:
645 type = RTM_DELROUTE;
646 if (hdr->nlmsg_type != RTM_NEWROUTE)
647 return NL_SKIP;
648
649 cb = check_route;
650 break;
651 case RTM_GETRULE:
652 type = RTM_DELRULE;
653 if (hdr->nlmsg_type != RTM_NEWRULE)
654 return NL_SKIP;
655
656 cb = check_rule;
657 break;
658 default:
659 return NL_SKIP;
660 }
661
662 if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
663 return NL_SKIP;
664
665 if (type == RTM_DELRULE)
666 D(SYSTEM, "Remove a rule\n");
667 else
668 D(SYSTEM, "Remove %s from device %s\n",
669 type == RTM_DELADDR ? "an address" : "a route",
670 clr->dev->ifname);
671 memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
672 hdr = nlmsg_hdr(clr->msg);
673 hdr->nlmsg_type = type;
674 hdr->nlmsg_flags = NLM_F_REQUEST;
675
676 nl_socket_disable_auto_ack(sock_rtnl);
677 nl_send_auto_complete(sock_rtnl, clr->msg);
678 nl_socket_enable_auto_ack(sock_rtnl);
679
680 return NL_SKIP;
681 }
682
683 static int
684 cb_finish_event(struct nl_msg *msg, void *arg)
685 {
686 int *pending = arg;
687 *pending = 0;
688 return NL_STOP;
689 }
690
691 static int
692 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
693 {
694 int *pending = arg;
695 *pending = err->error;
696 return NL_STOP;
697 }
698
699 static void
700 system_if_clear_entries(struct device *dev, int type, int af)
701 {
702 struct clear_data clr;
703 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
704 struct rtmsg rtm = {
705 .rtm_family = af,
706 .rtm_flags = RTM_F_CLONED,
707 };
708 int flags = NLM_F_DUMP;
709 int pending = 1;
710
711 clr.af = af;
712 clr.dev = dev;
713 clr.type = type;
714 switch (type) {
715 case RTM_GETADDR:
716 case RTM_GETRULE:
717 clr.size = sizeof(struct rtgenmsg);
718 break;
719 case RTM_GETROUTE:
720 clr.size = sizeof(struct rtmsg);
721 break;
722 default:
723 return;
724 }
725
726 if (!cb)
727 return;
728
729 clr.msg = nlmsg_alloc_simple(type, flags);
730 if (!clr.msg)
731 goto out;
732
733 nlmsg_append(clr.msg, &rtm, clr.size, 0);
734 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
735 nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
736 nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
737
738 nl_send_auto_complete(sock_rtnl, clr.msg);
739 while (pending > 0)
740 nl_recvmsgs(sock_rtnl, cb);
741
742 nlmsg_free(clr.msg);
743 out:
744 nl_cb_put(cb);
745 }
746
747 /*
748 * Clear bridge (membership) state and bring down device
749 */
750 void system_if_clear_state(struct device *dev)
751 {
752 static char buf[256];
753 char *bridge;
754
755 device_set_ifindex(dev, system_if_resolve(dev));
756 if (dev->external || !dev->ifindex)
757 return;
758
759 system_if_flags(dev->ifname, 0, IFF_UP);
760
761 if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
762 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
763 system_bridge_delbr(dev);
764 return;
765 }
766
767 bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
768 if (bridge) {
769 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
770 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
771 }
772
773 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
774 system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
775 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
776 system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
777 system_set_disable_ipv6(dev, "0");
778 }
779
780 static inline unsigned long
781 sec_to_jiffies(int val)
782 {
783 return (unsigned long) val * 100;
784 }
785
786 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
787 {
788 unsigned long args[4] = {};
789
790 if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
791 return -1;
792
793 args[0] = BRCTL_SET_BRIDGE_STP_STATE;
794 args[1] = !!cfg->stp;
795 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
796
797 args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
798 args[1] = sec_to_jiffies(cfg->forward_delay);
799 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
800
801 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
802 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
803
804 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
805 bridge->ifname, cfg->multicast_querier ? "1" : "0");
806
807 args[0] = BRCTL_SET_BRIDGE_PRIORITY;
808 args[1] = cfg->priority;
809 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
810
811 if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
812 args[0] = BRCTL_SET_AGEING_TIME;
813 args[1] = sec_to_jiffies(cfg->ageing_time);
814 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
815 }
816
817 if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
818 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
819 args[1] = sec_to_jiffies(cfg->hello_time);
820 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
821 }
822
823 if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
824 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
825 args[1] = sec_to_jiffies(cfg->max_age);
826 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
827 }
828
829 return 0;
830 }
831
832 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
833 {
834 struct nl_msg *msg;
835 struct nlattr *linkinfo, *data;
836 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
837 int i, rv;
838 static const struct {
839 const char *name;
840 enum macvlan_mode val;
841 } modes[] = {
842 { "private", MACVLAN_MODE_PRIVATE },
843 { "vepa", MACVLAN_MODE_VEPA },
844 { "bridge", MACVLAN_MODE_BRIDGE },
845 { "passthru", MACVLAN_MODE_PASSTHRU },
846 };
847
848 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
849
850 if (!msg)
851 return -1;
852
853 nlmsg_append(msg, &iim, sizeof(iim), 0);
854
855 if (cfg->flags & MACVLAN_OPT_MACADDR)
856 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
857 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
858 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
859
860 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
861 goto nla_put_failure;
862
863 nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
864
865 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
866 goto nla_put_failure;
867
868 if (cfg->mode) {
869 for (i = 0; i < ARRAY_SIZE(modes); i++) {
870 if (strcmp(cfg->mode, modes[i].name) != 0)
871 continue;
872
873 nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
874 break;
875 }
876 }
877
878 nla_nest_end(msg, data);
879 nla_nest_end(msg, linkinfo);
880
881 rv = system_rtnl_call(msg);
882 if (rv)
883 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
884
885 return rv;
886
887 nla_put_failure:
888 nlmsg_free(msg);
889 return -ENOMEM;
890 }
891
892 static int system_link_del(const char *ifname)
893 {
894 struct nl_msg *msg;
895 struct ifinfomsg iim = {
896 .ifi_family = AF_UNSPEC,
897 .ifi_index = 0,
898 };
899
900 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
901
902 if (!msg)
903 return -1;
904
905 nlmsg_append(msg, &iim, sizeof(iim), 0);
906 nla_put_string(msg, IFLA_IFNAME, ifname);
907 return system_rtnl_call(msg);
908 }
909
910 int system_macvlan_del(struct device *macvlan)
911 {
912 return system_link_del(macvlan->ifname);
913 }
914
915 static int system_vlan(struct device *dev, int id)
916 {
917 struct vlan_ioctl_args ifr = {
918 .cmd = SET_VLAN_NAME_TYPE_CMD,
919 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
920 };
921
922 ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
923
924 if (id < 0) {
925 ifr.cmd = DEL_VLAN_CMD;
926 ifr.u.VID = 0;
927 } else {
928 ifr.cmd = ADD_VLAN_CMD;
929 ifr.u.VID = id;
930 }
931 strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
932 return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
933 }
934
935 int system_vlan_add(struct device *dev, int id)
936 {
937 return system_vlan(dev, id);
938 }
939
940 int system_vlan_del(struct device *dev)
941 {
942 return system_vlan(dev, -1);
943 }
944
945 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
946 {
947 struct nl_msg *msg;
948 struct nlattr *linkinfo, *data;
949 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
950 int rv;
951
952 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
953
954 if (!msg)
955 return -1;
956
957 nlmsg_append(msg, &iim, sizeof(iim), 0);
958 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
959 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
960
961 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
962 goto nla_put_failure;
963
964 nla_put_string(msg, IFLA_INFO_KIND, "vlan");
965
966 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
967 goto nla_put_failure;
968
969 nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
970
971 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
972 nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
973 #else
974 if(cfg->proto == VLAN_PROTO_8021AD)
975 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);
976 #endif
977
978 nla_nest_end(msg, data);
979 nla_nest_end(msg, linkinfo);
980
981 rv = system_rtnl_call(msg);
982 if (rv)
983 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
984
985 return rv;
986
987 nla_put_failure:
988 nlmsg_free(msg);
989 return -ENOMEM;
990 }
991
992 int system_vlandev_del(struct device *vlandev)
993 {
994 return system_link_del(vlandev->ifname);
995 }
996
997 static void
998 system_if_get_settings(struct device *dev, struct device_settings *s)
999 {
1000 struct ifreq ifr;
1001 char buf[10];
1002
1003 memset(&ifr, 0, sizeof(ifr));
1004 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1005
1006 if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
1007 s->mtu = ifr.ifr_mtu;
1008 s->flags |= DEV_OPT_MTU;
1009 }
1010
1011 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
1012 s->txqueuelen = ifr.ifr_qlen;
1013 s->flags |= DEV_OPT_TXQUEUELEN;
1014 }
1015
1016 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
1017 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
1018 s->flags |= DEV_OPT_MACADDR;
1019 }
1020
1021 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
1022 s->ipv6 = !strtoul(buf, NULL, 0);
1023 s->flags |= DEV_OPT_IPV6;
1024 }
1025
1026 if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
1027 s->promisc = ifr.ifr_flags & IFF_PROMISC;
1028 s->flags |= DEV_OPT_PROMISC;
1029 }
1030
1031 if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1032 s->rpfilter = strtoul(buf, NULL, 0);
1033 s->flags |= DEV_OPT_RPFILTER;
1034 }
1035
1036 if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1037 s->acceptlocal = strtoul(buf, NULL, 0);
1038 s->flags |= DEV_OPT_ACCEPTLOCAL;
1039 }
1040
1041 if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1042 s->igmpversion = strtoul(buf, NULL, 0);
1043 s->flags |= DEV_OPT_IGMPVERSION;
1044 }
1045
1046 if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1047 s->mldversion = strtoul(buf, NULL, 0);
1048 s->flags |= DEV_OPT_MLDVERSION;
1049 }
1050
1051 if (!system_get_neigh4reachabletime(dev, buf, sizeof(buf))) {
1052 s->neigh4reachabletime = strtoul(buf, NULL, 0);
1053 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1054 }
1055
1056 if (!system_get_neigh6reachabletime(dev, buf, sizeof(buf))) {
1057 s->neigh6reachabletime = strtoul(buf, NULL, 0);
1058 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1059 }
1060 }
1061
1062 static void
1063 system_if_set_rps_xps_val(const char *path, int val)
1064 {
1065 char val_buf[8];
1066 glob_t gl;
1067 int i;
1068
1069 if (glob(path, 0, NULL, &gl))
1070 return;
1071
1072 snprintf(val_buf, sizeof(val_buf), "%x", val);
1073 for (i = 0; i < gl.gl_pathc; i++)
1074 system_set_sysctl(gl.gl_pathv[i], val_buf);
1075 }
1076
1077 static void
1078 system_if_apply_rps_xps(struct device *dev, struct device_settings *s)
1079 {
1080 long n_cpus = sysconf(_SC_NPROCESSORS_ONLN);
1081 int val;
1082
1083 if (n_cpus < 2)
1084 return;
1085
1086 val = (1 << n_cpus) - 1;
1087 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/rps_cpus", dev->ifname);
1088 system_if_set_rps_xps_val(dev_buf, s->rps ? val : 0);
1089
1090 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/xps_cpus", dev->ifname);
1091 system_if_set_rps_xps_val(dev_buf, s->xps ? val : 0);
1092 }
1093
1094 void
1095 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1096 {
1097 struct ifreq ifr;
1098
1099 memset(&ifr, 0, sizeof(ifr));
1100 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1101 if (s->flags & DEV_OPT_MTU & apply_mask) {
1102 ifr.ifr_mtu = s->mtu;
1103 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1104 s->flags &= ~DEV_OPT_MTU;
1105 }
1106 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1107 ifr.ifr_qlen = s->txqueuelen;
1108 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1109 s->flags &= ~DEV_OPT_TXQUEUELEN;
1110 }
1111 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1112 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1113 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1114 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1115 s->flags &= ~DEV_OPT_MACADDR;
1116 }
1117 if (s->flags & DEV_OPT_IPV6 & apply_mask)
1118 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1119 if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1120 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1121 !s->promisc ? IFF_PROMISC : 0) < 0)
1122 s->flags &= ~DEV_OPT_PROMISC;
1123 }
1124 if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1125 char buf[2];
1126
1127 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1128 system_set_rpfilter(dev, buf);
1129 }
1130 if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1131 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1132 if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1133 char buf[2];
1134
1135 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1136 system_set_igmpversion(dev, buf);
1137 }
1138 if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1139 char buf[2];
1140
1141 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1142 system_set_mldversion(dev, buf);
1143 }
1144 if (s->flags & DEV_OPT_NEIGHREACHABLETIME & apply_mask) {
1145 char buf[12];
1146
1147 snprintf(buf, sizeof(buf), "%d", s->neigh4reachabletime);
1148 system_set_neigh4reachabletime(dev, buf);
1149 snprintf(buf, sizeof(buf), "%d", s->neigh6reachabletime);
1150 system_set_neigh6reachabletime(dev, buf);
1151 }
1152
1153 system_if_apply_rps_xps(dev, s);
1154 }
1155
1156 int system_if_up(struct device *dev)
1157 {
1158 system_if_get_settings(dev, &dev->orig_settings);
1159 /* Only keep orig settings based on what needs to be set */
1160 dev->orig_settings.flags &= dev->settings.flags;
1161 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1162 return system_if_flags(dev->ifname, IFF_UP, 0);
1163 }
1164
1165 int system_if_down(struct device *dev)
1166 {
1167 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1168 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1169 return ret;
1170 }
1171
1172 struct if_check_data {
1173 struct device *dev;
1174 int pending;
1175 int ret;
1176 };
1177
1178 #ifndef IFF_LOWER_UP
1179 #define IFF_LOWER_UP 0x10000
1180 #endif
1181
1182 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1183 {
1184 struct nlmsghdr *nh = nlmsg_hdr(msg);
1185 struct ifinfomsg *ifi = NLMSG_DATA(nh);
1186 struct if_check_data *chk = (struct if_check_data *)arg;
1187
1188 if (nh->nlmsg_type != RTM_NEWLINK)
1189 return NL_SKIP;
1190
1191 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1192 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1193
1194 return NL_OK;
1195 }
1196
1197 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1198 {
1199 struct if_check_data *chk = (struct if_check_data *)arg;
1200 chk->pending = 0;
1201 return NL_STOP;
1202 }
1203
1204 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1205 {
1206 struct if_check_data *chk = (struct if_check_data *)arg;
1207
1208 device_set_present(chk->dev, false);
1209 device_set_link(chk->dev, false);
1210 chk->pending = err->error;
1211
1212 return NL_STOP;
1213 }
1214
1215 int system_if_check(struct device *dev)
1216 {
1217 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1218 struct nl_msg *msg;
1219 struct ifinfomsg ifi = {
1220 .ifi_family = AF_UNSPEC,
1221 .ifi_index = 0,
1222 };
1223 struct if_check_data chk = {
1224 .dev = dev,
1225 .pending = 1,
1226 };
1227 int ret = 1;
1228
1229 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1230 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1231 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1232 goto out;
1233
1234 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1235 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1236 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1237
1238 nl_send_auto_complete(sock_rtnl, msg);
1239 while (chk.pending > 0)
1240 nl_recvmsgs(sock_rtnl, cb);
1241
1242 nlmsg_free(msg);
1243 ret = chk.pending;
1244
1245 out:
1246 nl_cb_put(cb);
1247 return ret;
1248 }
1249
1250 struct device *
1251 system_if_get_parent(struct device *dev)
1252 {
1253 char buf[64], *devname;
1254 int ifindex, iflink, len;
1255 FILE *f;
1256
1257 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1258 f = fopen(buf, "r");
1259 if (!f)
1260 return NULL;
1261
1262 len = fread(buf, 1, sizeof(buf) - 1, f);
1263 fclose(f);
1264
1265 if (len <= 0)
1266 return NULL;
1267
1268 buf[len] = 0;
1269 iflink = strtoul(buf, NULL, 0);
1270 ifindex = system_if_resolve(dev);
1271 if (!iflink || iflink == ifindex)
1272 return NULL;
1273
1274 devname = if_indextoname(iflink, buf);
1275 if (!devname)
1276 return NULL;
1277
1278 return device_get(devname, true);
1279 }
1280
1281 static bool
1282 read_string_file(int dir_fd, const char *file, char *buf, int len)
1283 {
1284 bool ret = false;
1285 char *c;
1286 int fd;
1287
1288 fd = openat(dir_fd, file, O_RDONLY);
1289 if (fd < 0)
1290 return false;
1291
1292 retry:
1293 len = read(fd, buf, len - 1);
1294 if (len < 0) {
1295 if (errno == EINTR)
1296 goto retry;
1297 } else if (len > 0) {
1298 buf[len] = 0;
1299
1300 c = strchr(buf, '\n');
1301 if (c)
1302 *c = 0;
1303
1304 ret = true;
1305 }
1306
1307 close(fd);
1308
1309 return ret;
1310 }
1311
1312 static bool
1313 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1314 {
1315 char buf[64];
1316 bool ret = false;
1317
1318 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1319 if (ret)
1320 *val = strtoull(buf, NULL, 0);
1321
1322 return ret;
1323 }
1324
1325 /* Assume advertised flags == supported flags */
1326 static const struct {
1327 uint32_t mask;
1328 const char *name;
1329 } ethtool_link_modes[] = {
1330 { ADVERTISED_10baseT_Half, "10H" },
1331 { ADVERTISED_10baseT_Full, "10F" },
1332 { ADVERTISED_100baseT_Half, "100H" },
1333 { ADVERTISED_100baseT_Full, "100F" },
1334 { ADVERTISED_1000baseT_Half, "1000H" },
1335 { ADVERTISED_1000baseT_Full, "1000F" },
1336 };
1337
1338 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1339 {
1340 int i;
1341 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1342 if (mask & ethtool_link_modes[i].mask)
1343 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1344 }
1345 }
1346
1347 bool
1348 system_if_force_external(const char *ifname)
1349 {
1350 char buf[64];
1351 struct stat s;
1352
1353 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1354 return stat(buf, &s) == 0;
1355 }
1356
1357 int
1358 system_if_dump_info(struct device *dev, struct blob_buf *b)
1359 {
1360 struct ethtool_cmd ecmd;
1361 struct ifreq ifr;
1362 char buf[64], *s;
1363 void *c;
1364 int dir_fd;
1365
1366 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1367 dir_fd = open(buf, O_DIRECTORY);
1368
1369 memset(&ecmd, 0, sizeof(ecmd));
1370 memset(&ifr, 0, sizeof(ifr));
1371 strcpy(ifr.ifr_name, dev->ifname);
1372 ifr.ifr_data = (caddr_t) &ecmd;
1373 ecmd.cmd = ETHTOOL_GSET;
1374
1375 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1376 c = blobmsg_open_array(b, "link-advertising");
1377 system_add_link_modes(b, ecmd.advertising);
1378 blobmsg_close_array(b, c);
1379
1380 c = blobmsg_open_array(b, "link-supported");
1381 system_add_link_modes(b, ecmd.supported);
1382 blobmsg_close_array(b, c);
1383
1384 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1385 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1386 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1387 blobmsg_add_string_buffer(b);
1388 }
1389
1390 close(dir_fd);
1391 return 0;
1392 }
1393
1394 int
1395 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1396 {
1397 const char *const counters[] = {
1398 "collisions", "rx_frame_errors", "tx_compressed",
1399 "multicast", "rx_length_errors", "tx_dropped",
1400 "rx_bytes", "rx_missed_errors", "tx_errors",
1401 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1402 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1403 "rx_dropped", "tx_aborted_errors", "tx_packets",
1404 "rx_errors", "tx_bytes", "tx_window_errors",
1405 "rx_fifo_errors", "tx_carrier_errors",
1406 };
1407 char buf[64];
1408 int stats_dir;
1409 int i;
1410 uint64_t val = 0;
1411
1412 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1413 stats_dir = open(buf, O_DIRECTORY);
1414 if (stats_dir < 0)
1415 return -1;
1416
1417 for (i = 0; i < ARRAY_SIZE(counters); i++)
1418 if (read_uint64_file(stats_dir, counters[i], &val))
1419 blobmsg_add_u64(b, counters[i], val);
1420
1421 close(stats_dir);
1422 return 0;
1423 }
1424
1425 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1426 {
1427 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1428 int alen = v4 ? 4 : 16;
1429 unsigned int flags = 0;
1430 struct ifaddrmsg ifa = {
1431 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1432 .ifa_prefixlen = addr->mask,
1433 .ifa_index = dev->ifindex,
1434 .ifa_flags = (addr->flags & DEVADDR_OFFLINK) ? IFA_F_NOPREFIXROUTE : 0,
1435 };
1436
1437 struct nl_msg *msg;
1438 if (cmd == RTM_NEWADDR)
1439 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1440
1441 msg = nlmsg_alloc_simple(cmd, flags);
1442 if (!msg)
1443 return -1;
1444
1445 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1446 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1447 if (v4) {
1448 if (addr->broadcast)
1449 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1450 if (addr->point_to_point)
1451 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1452 } else {
1453 time_t now = system_get_rtime();
1454 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1455
1456 if (addr->preferred_until) {
1457 int64_t preferred = addr->preferred_until - now;
1458 if (preferred < 0)
1459 preferred = 0;
1460 else if (preferred > UINT32_MAX)
1461 preferred = UINT32_MAX;
1462
1463 cinfo.ifa_prefered = preferred;
1464 }
1465
1466 if (addr->valid_until) {
1467 int64_t valid = addr->valid_until - now;
1468 if (valid <= 0)
1469 return -1;
1470 else if (valid > UINT32_MAX)
1471 valid = UINT32_MAX;
1472
1473 cinfo.ifa_valid = valid;
1474 }
1475
1476 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1477 }
1478
1479 return system_rtnl_call(msg);
1480 }
1481
1482 int system_add_address(struct device *dev, struct device_addr *addr)
1483 {
1484 return system_addr(dev, addr, RTM_NEWADDR);
1485 }
1486
1487 int system_del_address(struct device *dev, struct device_addr *addr)
1488 {
1489 return system_addr(dev, addr, RTM_DELADDR);
1490 }
1491
1492 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1493 {
1494 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1495 bool have_gw;
1496 unsigned int flags = 0;
1497
1498 if (alen == 4)
1499 have_gw = !!route->nexthop.in.s_addr;
1500 else
1501 have_gw = route->nexthop.in6.s6_addr32[0] ||
1502 route->nexthop.in6.s6_addr32[1] ||
1503 route->nexthop.in6.s6_addr32[2] ||
1504 route->nexthop.in6.s6_addr32[3];
1505
1506 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1507 ? route->table : RT_TABLE_MAIN;
1508
1509 struct rtmsg rtm = {
1510 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1511 .rtm_dst_len = route->mask,
1512 .rtm_src_len = route->sourcemask,
1513 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1514 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1515 .rtm_scope = RT_SCOPE_NOWHERE,
1516 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1517 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1518 };
1519 struct nl_msg *msg;
1520
1521 if (cmd == RTM_NEWROUTE) {
1522 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1523
1524 if (!dev) { // Add null-route
1525 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1526 rtm.rtm_type = RTN_UNREACHABLE;
1527 }
1528 else
1529 rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1530 }
1531
1532 if (route->flags & DEVROUTE_TYPE) {
1533 rtm.rtm_type = route->type;
1534 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1535 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1536 rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1537 rtm.rtm_table = RT_TABLE_LOCAL;
1538 }
1539
1540 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1541 rtm.rtm_scope = RT_SCOPE_HOST;
1542 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1543 rtm.rtm_type == RTN_ANYCAST) {
1544 rtm.rtm_scope = RT_SCOPE_LINK;
1545 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1546 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1547 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1548 dev = NULL;
1549 }
1550 }
1551
1552 msg = nlmsg_alloc_simple(cmd, flags);
1553 if (!msg)
1554 return -1;
1555
1556 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1557
1558 if (route->mask)
1559 nla_put(msg, RTA_DST, alen, &route->addr);
1560
1561 if (route->sourcemask) {
1562 if (rtm.rtm_family == AF_INET)
1563 nla_put(msg, RTA_PREFSRC, alen, &route->source);
1564 else
1565 nla_put(msg, RTA_SRC, alen, &route->source);
1566 }
1567
1568 if (route->metric > 0)
1569 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1570
1571 if (have_gw)
1572 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1573
1574 if (dev)
1575 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1576
1577 if (table >= 256)
1578 nla_put_u32(msg, RTA_TABLE, table);
1579
1580 if (route->flags & DEVROUTE_MTU) {
1581 struct nlattr *metrics;
1582
1583 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1584 goto nla_put_failure;
1585
1586 nla_put_u32(msg, RTAX_MTU, route->mtu);
1587
1588 nla_nest_end(msg, metrics);
1589 }
1590
1591 return system_rtnl_call(msg);
1592
1593 nla_put_failure:
1594 nlmsg_free(msg);
1595 return -ENOMEM;
1596 }
1597
1598 int system_add_route(struct device *dev, struct device_route *route)
1599 {
1600 return system_rt(dev, route, RTM_NEWROUTE);
1601 }
1602
1603 int system_del_route(struct device *dev, struct device_route *route)
1604 {
1605 return system_rt(dev, route, RTM_DELROUTE);
1606 }
1607
1608 int system_flush_routes(void)
1609 {
1610 const char *names[] = {
1611 "/proc/sys/net/ipv4/route/flush",
1612 "/proc/sys/net/ipv6/route/flush"
1613 };
1614 int fd, i;
1615
1616 for (i = 0; i < ARRAY_SIZE(names); i++) {
1617 fd = open(names[i], O_WRONLY);
1618 if (fd < 0)
1619 continue;
1620
1621 if (write(fd, "-1", 2)) {}
1622 close(fd);
1623 }
1624 return 0;
1625 }
1626
1627 bool system_resolve_rt_type(const char *type, unsigned int *id)
1628 {
1629 return system_rtn_aton(type, id);
1630 }
1631
1632 bool system_resolve_rt_table(const char *name, unsigned int *id)
1633 {
1634 FILE *f;
1635 char *e, buf[128];
1636 unsigned int n, table = RT_TABLE_UNSPEC;
1637
1638 /* first try to parse table as number */
1639 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1640 table = n;
1641
1642 /* handle well known aliases */
1643 else if (!strcmp(name, "default"))
1644 table = RT_TABLE_DEFAULT;
1645 else if (!strcmp(name, "main"))
1646 table = RT_TABLE_MAIN;
1647 else if (!strcmp(name, "local"))
1648 table = RT_TABLE_LOCAL;
1649 else if (!strcmp(name, "prelocal"))
1650 table = RT_TABLE_PRELOCAL;
1651
1652 /* try to look up name in /etc/iproute2/rt_tables */
1653 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1654 {
1655 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1656 {
1657 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1658 continue;
1659
1660 n = strtoul(e, NULL, 10);
1661 e = strtok(NULL, " \t\n");
1662
1663 if (e && !strcmp(e, name))
1664 {
1665 table = n;
1666 break;
1667 }
1668 }
1669
1670 fclose(f);
1671 }
1672
1673 if (table == RT_TABLE_UNSPEC)
1674 return false;
1675
1676 *id = table;
1677 return true;
1678 }
1679
1680 bool system_is_default_rt_table(unsigned int id)
1681 {
1682 return (id == RT_TABLE_MAIN);
1683 }
1684
1685 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1686 {
1687 char *e;
1688 unsigned int n;
1689
1690 if (!strcmp(filter, "strict"))
1691 n = 1;
1692 else if (!strcmp(filter, "loose"))
1693 n = 2;
1694 else {
1695 n = strtoul(filter, &e, 0);
1696 if (*e || e == filter || n > 2)
1697 return false;
1698 }
1699
1700 *id = n;
1701 return true;
1702 }
1703
1704 static int system_iprule(struct iprule *rule, int cmd)
1705 {
1706 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1707
1708 struct nl_msg *msg;
1709 struct rtmsg rtm = {
1710 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1711 .rtm_protocol = RTPROT_STATIC,
1712 .rtm_scope = RT_SCOPE_UNIVERSE,
1713 .rtm_table = RT_TABLE_UNSPEC,
1714 .rtm_type = RTN_UNSPEC,
1715 .rtm_flags = 0,
1716 };
1717
1718 if (cmd == RTM_NEWRULE) {
1719 rtm.rtm_type = RTN_UNICAST;
1720 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1721 }
1722
1723 if (rule->invert)
1724 rtm.rtm_flags |= FIB_RULE_INVERT;
1725
1726 if (rule->flags & IPRULE_SRC)
1727 rtm.rtm_src_len = rule->src_mask;
1728
1729 if (rule->flags & IPRULE_DEST)
1730 rtm.rtm_dst_len = rule->dest_mask;
1731
1732 if (rule->flags & IPRULE_TOS)
1733 rtm.rtm_tos = rule->tos;
1734
1735 if (rule->flags & IPRULE_LOOKUP) {
1736 if (rule->lookup < 256)
1737 rtm.rtm_table = rule->lookup;
1738 }
1739
1740 if (rule->flags & IPRULE_ACTION)
1741 rtm.rtm_type = rule->action;
1742 else if (rule->flags & IPRULE_GOTO)
1743 rtm.rtm_type = FR_ACT_GOTO;
1744 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1745 rtm.rtm_type = FR_ACT_NOP;
1746
1747 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1748
1749 if (!msg)
1750 return -1;
1751
1752 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1753
1754 if (rule->flags & IPRULE_IN)
1755 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1756
1757 if (rule->flags & IPRULE_OUT)
1758 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1759
1760 if (rule->flags & IPRULE_SRC)
1761 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1762
1763 if (rule->flags & IPRULE_DEST)
1764 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1765
1766 if (rule->flags & IPRULE_PRIORITY)
1767 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1768 else if (cmd == RTM_NEWRULE)
1769 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1770
1771 if (rule->flags & IPRULE_FWMARK)
1772 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1773
1774 if (rule->flags & IPRULE_FWMASK)
1775 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1776
1777 if (rule->flags & IPRULE_LOOKUP) {
1778 if (rule->lookup >= 256)
1779 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1780 }
1781
1782 if (rule->flags & IPRULE_GOTO)
1783 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1784
1785 return system_rtnl_call(msg);
1786 }
1787
1788 int system_add_iprule(struct iprule *rule)
1789 {
1790 return system_iprule(rule, RTM_NEWRULE);
1791 }
1792
1793 int system_del_iprule(struct iprule *rule)
1794 {
1795 return system_iprule(rule, RTM_DELRULE);
1796 }
1797
1798 int system_flush_iprules(void)
1799 {
1800 int rv = 0;
1801 struct iprule rule;
1802
1803 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1804 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1805
1806 memset(&rule, 0, sizeof(rule));
1807
1808
1809 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1810
1811 rule.priority = 0;
1812 rule.lookup = RT_TABLE_PRELOCAL;
1813 rv |= system_iprule(&rule, RTM_NEWRULE);
1814
1815 rule.priority = 1;
1816 rule.lookup = RT_TABLE_LOCAL;
1817 rv |= system_iprule(&rule, RTM_NEWRULE);
1818
1819 rule.priority = 32766;
1820 rule.lookup = RT_TABLE_MAIN;
1821 rv |= system_iprule(&rule, RTM_NEWRULE);
1822
1823 rule.priority = 32767;
1824 rule.lookup = RT_TABLE_DEFAULT;
1825 rv |= system_iprule(&rule, RTM_NEWRULE);
1826
1827
1828 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1829
1830 rule.priority = 0;
1831 rule.lookup = RT_TABLE_PRELOCAL;
1832 rv |= system_iprule(&rule, RTM_NEWRULE);
1833
1834 rule.priority = 1;
1835 rule.lookup = RT_TABLE_LOCAL;
1836 rv |= system_iprule(&rule, RTM_NEWRULE);
1837
1838 rule.priority = 32766;
1839 rule.lookup = RT_TABLE_MAIN;
1840 rv |= system_iprule(&rule, RTM_NEWRULE);
1841
1842 return rv;
1843 }
1844
1845 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1846 {
1847 return system_rtn_aton(action, id);
1848 }
1849
1850 time_t system_get_rtime(void)
1851 {
1852 struct timespec ts;
1853 struct timeval tv;
1854
1855 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1856 return ts.tv_sec;
1857
1858 if (gettimeofday(&tv, NULL) == 0)
1859 return tv.tv_sec;
1860
1861 return 0;
1862 }
1863
1864 #ifndef IP_DF
1865 #define IP_DF 0x4000
1866 #endif
1867
1868 static int tunnel_ioctl(const char *name, int cmd, void *p)
1869 {
1870 struct ifreq ifr;
1871
1872 memset(&ifr, 0, sizeof(ifr));
1873 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1874 ifr.ifr_ifru.ifru_data = p;
1875 return ioctl(sock_ioctl, cmd, &ifr);
1876 }
1877
1878 #ifdef IFLA_IPTUN_MAX
1879 #define IP6_FLOWINFO_TCLASS htonl(0x0FF00000)
1880 static int system_add_gre_tunnel(const char *name, const char *kind,
1881 const unsigned int link, struct blob_attr **tb, bool v6)
1882 {
1883 struct nl_msg *nlm;
1884 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1885 struct blob_attr *cur;
1886 uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1887 uint16_t iflags = 0, oflags = 0;
1888 uint8_t tos = 0;
1889 int ret = 0, ttl = 64;
1890
1891 nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1892 if (!nlm)
1893 return -1;
1894
1895 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1896 nla_put_string(nlm, IFLA_IFNAME, name);
1897
1898 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1899 if (!linkinfo) {
1900 ret = -ENOMEM;
1901 goto failure;
1902 }
1903
1904 nla_put_string(nlm, IFLA_INFO_KIND, kind);
1905 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1906 if (!infodata) {
1907 ret = -ENOMEM;
1908 goto failure;
1909 }
1910
1911 if (link)
1912 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1913
1914 if ((cur = tb[TUNNEL_ATTR_TTL]))
1915 ttl = blobmsg_get_u32(cur);
1916
1917 nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1918
1919 if ((cur = tb[TUNNEL_ATTR_TOS])) {
1920 char *str = blobmsg_get_string(cur);
1921 if (strcmp(str, "inherit")) {
1922 unsigned uval;
1923
1924 if (!system_tos_aton(str, &uval)) {
1925 ret = -EINVAL;
1926 goto failure;
1927 }
1928
1929 if (v6)
1930 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1931 else
1932 tos = uval;
1933 } else {
1934 if (v6)
1935 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1936 else
1937 tos = 1;
1938 }
1939 }
1940
1941 if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1942 uint8_t icsum, ocsum, iseqno, oseqno;
1943 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1944 &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1945 ret = -EINVAL;
1946 goto failure;
1947 }
1948
1949 if (ikey)
1950 iflags |= GRE_KEY;
1951
1952 if (okey)
1953 oflags |= GRE_KEY;
1954
1955 if (icsum)
1956 iflags |= GRE_CSUM;
1957
1958 if (ocsum)
1959 oflags |= GRE_CSUM;
1960
1961 if (iseqno)
1962 iflags |= GRE_SEQ;
1963
1964 if (oseqno)
1965 oflags |= GRE_SEQ;
1966 }
1967
1968 if (v6) {
1969 struct in6_addr in6buf;
1970 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1971 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1972 ret = -EINVAL;
1973 goto failure;
1974 }
1975 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1976 }
1977
1978 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1979 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1980 ret = -EINVAL;
1981 goto failure;
1982 }
1983 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1984 }
1985 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1986
1987 if (flowinfo)
1988 nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1989
1990 if (flags)
1991 nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
1992 } else {
1993 struct in_addr inbuf;
1994 bool set_df = true;
1995
1996 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1997 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1998 ret = -EINVAL;
1999 goto failure;
2000 }
2001 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
2002 }
2003
2004 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2005 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2006 ret = -EINVAL;
2007 goto failure;
2008 }
2009 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
2010
2011 if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
2012 if (!okey) {
2013 okey = inbuf.s_addr;
2014 oflags |= GRE_KEY;
2015 }
2016
2017 if (!ikey) {
2018 ikey = inbuf.s_addr;
2019 iflags |= GRE_KEY;
2020 }
2021 }
2022 }
2023
2024 if ((cur = tb[TUNNEL_ATTR_DF]))
2025 set_df = blobmsg_get_bool(cur);
2026
2027 /* ttl !=0 and nopmtudisc are incompatible */
2028 if (ttl && !set_df) {
2029 ret = -EINVAL;
2030 goto failure;
2031 }
2032
2033 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
2034
2035 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
2036 }
2037
2038 if (oflags)
2039 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
2040
2041 if (iflags)
2042 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
2043
2044 if (okey)
2045 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
2046
2047 if (ikey)
2048 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
2049
2050 nla_nest_end(nlm, infodata);
2051 nla_nest_end(nlm, linkinfo);
2052
2053 return system_rtnl_call(nlm);
2054
2055 failure:
2056 nlmsg_free(nlm);
2057 return ret;
2058 }
2059 #endif
2060
2061 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
2062 {
2063 struct blob_attr *cur;
2064 bool set_df = true;
2065 struct ip_tunnel_parm p = {
2066 .link = link,
2067 .iph = {
2068 .version = 4,
2069 .ihl = 5,
2070 .protocol = proto,
2071 }
2072 };
2073
2074 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
2075 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
2076 return -EINVAL;
2077
2078 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
2079 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
2080 return -EINVAL;
2081
2082 if ((cur = tb[TUNNEL_ATTR_DF]))
2083 set_df = blobmsg_get_bool(cur);
2084
2085 if ((cur = tb[TUNNEL_ATTR_TTL]))
2086 p.iph.ttl = blobmsg_get_u32(cur);
2087
2088 if ((cur = tb[TUNNEL_ATTR_TOS])) {
2089 char *str = blobmsg_get_string(cur);
2090 if (strcmp(str, "inherit")) {
2091 unsigned uval;
2092
2093 if (!system_tos_aton(str, &uval))
2094 return -EINVAL;
2095
2096 p.iph.tos = uval;
2097 } else
2098 p.iph.tos = 1;
2099 }
2100
2101 p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2102 /* ttl !=0 and nopmtudisc are incompatible */
2103 if (p.iph.ttl && p.iph.frag_off == 0)
2104 return -EINVAL;
2105
2106 strncpy(p.name, name, sizeof(p.name));
2107
2108 switch (p.iph.protocol) {
2109 case IPPROTO_IPIP:
2110 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2111 case IPPROTO_IPV6:
2112 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2113 default:
2114 break;
2115 }
2116 return -1;
2117 }
2118
2119 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2120 {
2121 struct blob_attr *cur;
2122 const char *str;
2123
2124 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2125 return -EINVAL;
2126 str = blobmsg_data(cur);
2127
2128 if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2129 !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2130 return system_link_del(name);
2131 else
2132 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2133 }
2134
2135 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2136 {
2137 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2138
2139 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2140 blob_data(attr), blob_len(attr));
2141
2142 return __system_del_ip_tunnel(name, tb);
2143 }
2144
2145 int system_update_ipv6_mtu(struct device *dev, int mtu)
2146 {
2147 int ret = -1;
2148 char buf[64];
2149 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2150 dev->ifname);
2151
2152 int fd = open(buf, O_RDWR);
2153 ssize_t len = read(fd, buf, sizeof(buf) - 1);
2154 if (len < 0)
2155 goto out;
2156
2157 buf[len] = 0;
2158 ret = atoi(buf);
2159
2160 if (!mtu || ret <= mtu)
2161 goto out;
2162
2163 lseek(fd, 0, SEEK_SET);
2164 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2165 ret = -1;
2166
2167 out:
2168 close(fd);
2169 return ret;
2170 }
2171
2172 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2173 {
2174 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2175 struct blob_attr *cur;
2176 const char *str;
2177
2178 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2179 blob_data(attr), blob_len(attr));
2180
2181 __system_del_ip_tunnel(name, tb);
2182
2183 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2184 return -EINVAL;
2185 str = blobmsg_data(cur);
2186
2187 unsigned int ttl = 0;
2188 if ((cur = tb[TUNNEL_ATTR_TTL])) {
2189 ttl = blobmsg_get_u32(cur);
2190 if (ttl > 255)
2191 return -EINVAL;
2192 }
2193
2194 unsigned int link = 0;
2195 if ((cur = tb[TUNNEL_ATTR_LINK])) {
2196 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2197 if (!iface)
2198 return -EINVAL;
2199
2200 if (iface->l3_dev.dev)
2201 link = iface->l3_dev.dev->ifindex;
2202 }
2203
2204 if (!strcmp(str, "sit")) {
2205 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2206 return -1;
2207
2208 #ifdef SIOCADD6RD
2209 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2210 unsigned int mask;
2211 struct ip_tunnel_6rd p6;
2212
2213 memset(&p6, 0, sizeof(p6));
2214
2215 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2216 &p6.prefix, &mask) || mask > 128)
2217 return -EINVAL;
2218 p6.prefixlen = mask;
2219
2220 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2221 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2222 &p6.relay_prefix, &mask) || mask > 32)
2223 return -EINVAL;
2224 p6.relay_prefixlen = mask;
2225 }
2226
2227 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2228 __system_del_ip_tunnel(name, tb);
2229 return -1;
2230 }
2231 }
2232 #endif
2233 #ifdef IFLA_IPTUN_MAX
2234 } else if (!strcmp(str, "ipip6")) {
2235 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2236 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2237 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2238 int ret = 0;
2239
2240 if (!nlm)
2241 return -1;
2242
2243 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2244 nla_put_string(nlm, IFLA_IFNAME, name);
2245
2246 if (link)
2247 nla_put_u32(nlm, IFLA_LINK, link);
2248
2249 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2250 if (!linkinfo) {
2251 ret = -ENOMEM;
2252 goto failure;
2253 }
2254 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2255 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2256 if (!infodata) {
2257 ret = -ENOMEM;
2258 goto failure;
2259 }
2260
2261 if (link)
2262 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2263
2264 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2265 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2266 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2267
2268 struct in6_addr in6buf;
2269 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2270 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2271 ret = -EINVAL;
2272 goto failure;
2273 }
2274 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2275 }
2276
2277 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2278 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2279 ret = -EINVAL;
2280 goto failure;
2281 }
2282 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2283 }
2284
2285 #ifdef IFLA_IPTUN_FMR_MAX
2286 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2287 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2288
2289 struct blob_attr *fmr;
2290 unsigned rem, fmrcnt = 0;
2291 blobmsg_for_each_attr(fmr, cur, rem) {
2292 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2293 continue;
2294
2295 unsigned ip4len, ip6len, ealen, offset = 6;
2296 char ip6buf[48];
2297 char ip4buf[16];
2298
2299 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2300 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2301 ret = -EINVAL;
2302 goto failure;
2303 }
2304
2305 struct in6_addr ip6prefix;
2306 struct in_addr ip4prefix;
2307 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2308 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2309 ret = -EINVAL;
2310 goto failure;
2311 }
2312
2313 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2314
2315 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2316 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2317 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2318 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2319 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2320 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2321
2322 nla_nest_end(nlm, rule);
2323 }
2324
2325 nla_nest_end(nlm, fmrs);
2326 }
2327 #endif
2328
2329 nla_nest_end(nlm, infodata);
2330 nla_nest_end(nlm, linkinfo);
2331
2332 return system_rtnl_call(nlm);
2333 failure:
2334 nlmsg_free(nlm);
2335 return ret;
2336 } else if (!strcmp(str, "greip")) {
2337 return system_add_gre_tunnel(name, "gre", link, tb, false);
2338 } else if (!strcmp(str, "gretapip")) {
2339 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2340 } else if (!strcmp(str, "greip6")) {
2341 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2342 } else if (!strcmp(str, "gretapip6")) {
2343 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2344 #endif
2345 } else if (!strcmp(str, "ipip")) {
2346 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2347 }
2348 else
2349 return -EINVAL;
2350
2351 return 0;
2352 }