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