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