Add mtu6 option to override IPv6 MTU
[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_addr.h>
34 #include <linux/if_link.h>
35 #include <linux/if_vlan.h>
36 #include <linux/if_bridge.h>
37 #include <linux/if_tunnel.h>
38 #include <linux/ip6_tunnel.h>
39 #include <linux/ethtool.h>
40 #include <linux/fib_rules.h>
41 #include <linux/version.h>
42
43 #ifndef RTN_FAILED_POLICY
44 #define RTN_FAILED_POLICY 12
45 #endif
46
47 #ifndef RT_TABLE_PRELOCAL
48 #define RT_TABLE_PRELOCAL 128
49 #endif
50
51 #ifndef IFA_F_NOPREFIXROUTE
52 #define IFA_F_NOPREFIXROUTE 0x200
53 #endif
54
55 #ifndef IFA_FLAGS
56 #define IFA_FLAGS (IFA_MULTICAST + 1)
57 #endif
58
59
60 #include <string.h>
61 #include <fcntl.h>
62 #include <glob.h>
63 #include <time.h>
64 #include <unistd.h>
65
66 #include <netlink/msg.h>
67 #include <netlink/attr.h>
68 #include <netlink/socket.h>
69 #include <libubox/uloop.h>
70
71 #include "netifd.h"
72 #include "device.h"
73 #include "system.h"
74
75 struct event_socket {
76 struct uloop_fd uloop;
77 struct nl_sock *sock;
78 int bufsize;
79 };
80
81 static int sock_ioctl = -1;
82 static struct nl_sock *sock_rtnl = NULL;
83
84 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
85 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
86
87 static char dev_buf[256];
88
89 static void
90 handler_nl_event(struct uloop_fd *u, unsigned int events)
91 {
92 struct event_socket *ev = container_of(u, struct event_socket, uloop);
93 int err;
94 socklen_t errlen = sizeof(err);
95
96 if (!u->error) {
97 nl_recvmsgs_default(ev->sock);
98 return;
99 }
100
101 if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
102 goto abort;
103
104 switch(err) {
105 case ENOBUFS:
106 // Increase rx buffer size on netlink socket
107 ev->bufsize *= 2;
108 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
109 goto abort;
110
111 // Request full dump since some info got dropped
112 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
113 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
114 break;
115
116 default:
117 goto abort;
118 }
119 u->error = false;
120 return;
121
122 abort:
123 uloop_fd_delete(&ev->uloop);
124 return;
125 }
126
127 static struct nl_sock *
128 create_socket(int protocol, int groups)
129 {
130 struct nl_sock *sock;
131
132 sock = nl_socket_alloc();
133 if (!sock)
134 return NULL;
135
136 if (groups)
137 nl_join_groups(sock, groups);
138
139 if (nl_connect(sock, protocol))
140 return NULL;
141
142 return sock;
143 }
144
145 static bool
146 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
147 uloop_fd_handler cb, int flags)
148 {
149 ev->sock = create_socket(protocol, groups);
150 if (!ev->sock)
151 return false;
152
153 ev->uloop.fd = nl_socket_get_fd(ev->sock);
154 ev->uloop.cb = cb;
155 if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
156 return false;
157
158 return true;
159 }
160
161 static bool
162 create_event_socket(struct event_socket *ev, int protocol,
163 int (*cb)(struct nl_msg *msg, void *arg))
164 {
165 if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
166 return false;
167
168 // Install the valid custom callback handler
169 nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
170
171 // Disable sequence number checking on event sockets
172 nl_socket_disable_seq_check(ev->sock);
173
174 // Increase rx buffer size to 65K on event sockets
175 ev->bufsize = 65535;
176 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
177 return false;
178
179 return true;
180 }
181
182 static bool
183 system_rtn_aton(const char *src, unsigned int *dst)
184 {
185 char *e;
186 unsigned int n;
187
188 if (!strcmp(src, "local"))
189 n = RTN_LOCAL;
190 else if (!strcmp(src, "nat"))
191 n = RTN_NAT;
192 else if (!strcmp(src, "broadcast"))
193 n = RTN_BROADCAST;
194 else if (!strcmp(src, "anycast"))
195 n = RTN_ANYCAST;
196 else if (!strcmp(src, "multicast"))
197 n = RTN_MULTICAST;
198 else if (!strcmp(src, "prohibit"))
199 n = RTN_PROHIBIT;
200 else if (!strcmp(src, "unreachable"))
201 n = RTN_UNREACHABLE;
202 else if (!strcmp(src, "blackhole"))
203 n = RTN_BLACKHOLE;
204 else if (!strcmp(src, "xresolve"))
205 n = RTN_XRESOLVE;
206 else if (!strcmp(src, "unicast"))
207 n = RTN_UNICAST;
208 else if (!strcmp(src, "throw"))
209 n = RTN_THROW;
210 else if (!strcmp(src, "failed_policy"))
211 n = RTN_FAILED_POLICY;
212 else {
213 n = strtoul(src, &e, 0);
214 if (!e || *e || e == src || n > 255)
215 return false;
216 }
217
218 *dst = n;
219 return true;
220 }
221
222 static bool
223 system_tos_aton(const char *src, unsigned *dst)
224 {
225 char *e;
226
227 *dst = strtoul(src, &e, 16);
228 if (e == src || *e || *dst > 255)
229 return false;
230
231 return true;
232 }
233
234 int system_init(void)
235 {
236 static struct event_socket rtnl_event;
237 static struct event_socket hotplug_event;
238
239 sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
240 system_fd_set_cloexec(sock_ioctl);
241
242 // Prepare socket for routing / address control
243 sock_rtnl = create_socket(NETLINK_ROUTE, 0);
244 if (!sock_rtnl)
245 return -1;
246
247 if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
248 return -1;
249
250 if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
251 handle_hotplug_event, 0))
252 return -1;
253
254 // Receive network link events form kernel
255 nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
256
257 return 0;
258 }
259
260 static void system_set_sysctl(const char *path, const char *val)
261 {
262 int fd;
263
264 fd = open(path, O_WRONLY);
265 if (fd < 0)
266 return;
267
268 if (write(fd, val, strlen(val))) {}
269 close(fd);
270 }
271
272 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
273 {
274 snprintf(dev_buf, sizeof(dev_buf), path, device);
275 system_set_sysctl(dev_buf, val);
276 }
277
278 static void system_set_disable_ipv6(struct device *dev, const char *val)
279 {
280 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
281 }
282
283 static void system_set_rpfilter(struct device *dev, const char *val)
284 {
285 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter", dev->ifname, val);
286 }
287
288 static void system_set_acceptlocal(struct device *dev, const char *val)
289 {
290 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local", dev->ifname, val);
291 }
292
293 static void system_set_igmpversion(struct device *dev, const char *val)
294 {
295 system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version", dev->ifname, val);
296 }
297
298 static void system_set_mldversion(struct device *dev, const char *val)
299 {
300 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version", dev->ifname, val);
301 }
302
303 static void system_set_neigh4reachabletime(struct device *dev, const char *val)
304 {
305 system_set_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms", dev->ifname, val);
306 }
307
308 static void system_set_neigh6reachabletime(struct device *dev, const char *val)
309 {
310 system_set_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms", dev->ifname, val);
311 }
312
313 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
314 {
315 int fd = -1, ret = -1;
316
317 fd = open(path, O_RDONLY);
318 if (fd < 0)
319 goto out;
320
321 ssize_t len = read(fd, buf, buf_sz - 1);
322 if (len < 0)
323 goto out;
324
325 ret = buf[len] = 0;
326
327 out:
328 if (fd >= 0)
329 close(fd);
330
331 return ret;
332 }
333
334 static int
335 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
336 {
337 snprintf(dev_buf, sizeof(dev_buf), path, device);
338 return system_get_sysctl(dev_buf, buf, buf_sz);
339 }
340
341 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
342 {
343 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
344 dev->ifname, buf, buf_sz);
345 }
346
347 static int system_get_rpfilter(struct device *dev, char *buf, const size_t buf_sz)
348 {
349 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter",
350 dev->ifname, buf, buf_sz);
351 }
352
353 static int system_get_acceptlocal(struct device *dev, char *buf, const size_t buf_sz)
354 {
355 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local",
356 dev->ifname, buf, buf_sz);
357 }
358
359 static int system_get_igmpversion(struct device *dev, char *buf, const size_t buf_sz)
360 {
361 return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version",
362 dev->ifname, buf, buf_sz);
363 }
364
365 static int system_get_mldversion(struct device *dev, char *buf, const size_t buf_sz)
366 {
367 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version",
368 dev->ifname, buf, buf_sz);
369 }
370
371 static int system_get_neigh4reachabletime(struct device *dev, char *buf, const size_t buf_sz)
372 {
373 return system_get_dev_sysctl("/proc/sys/net/ipv4/neigh/%s/base_reachable_time_ms",
374 dev->ifname, buf, buf_sz);
375 }
376
377 static int system_get_neigh6reachabletime(struct device *dev, char *buf, const size_t buf_sz)
378 {
379 return system_get_dev_sysctl("/proc/sys/net/ipv6/neigh/%s/base_reachable_time_ms",
380 dev->ifname, buf, buf_sz);
381 }
382
383 // Evaluate netlink messages
384 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
385 {
386 struct nlmsghdr *nh = nlmsg_hdr(msg);
387 struct nlattr *nla[__IFLA_MAX];
388 int link_state = 0;
389 char buf[10];
390
391 if (nh->nlmsg_type != RTM_NEWLINK)
392 goto out;
393
394 nlmsg_parse(nh, sizeof(struct ifinfomsg), nla, __IFLA_MAX - 1, NULL);
395 if (!nla[IFLA_IFNAME])
396 goto out;
397
398 struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
399 if (!dev || dev->type->keep_link_status)
400 goto out;
401
402 if (!system_get_dev_sysctl("/sys/class/net/%s/carrier", dev->ifname, buf, sizeof(buf)))
403 link_state = strtoul(buf, NULL, 0);
404
405 device_set_link(dev, link_state ? true : false);
406
407 out:
408 return 0;
409 }
410
411 static void
412 handle_hotplug_msg(char *data, int size)
413 {
414 const char *subsystem = NULL, *interface = NULL;
415 char *cur, *end, *sep;
416 struct device *dev;
417 int skip;
418 bool add;
419
420 if (!strncmp(data, "add@", 4))
421 add = true;
422 else if (!strncmp(data, "remove@", 7))
423 add = false;
424 else
425 return;
426
427 skip = strlen(data) + 1;
428 end = data + size;
429
430 for (cur = data + skip; cur < end; cur += skip) {
431 skip = strlen(cur) + 1;
432
433 sep = strchr(cur, '=');
434 if (!sep)
435 continue;
436
437 *sep = 0;
438 if (!strcmp(cur, "INTERFACE"))
439 interface = sep + 1;
440 else if (!strcmp(cur, "SUBSYSTEM")) {
441 subsystem = sep + 1;
442 if (strcmp(subsystem, "net") != 0)
443 return;
444 }
445 if (subsystem && interface)
446 goto found;
447 }
448 return;
449
450 found:
451 dev = device_get(interface, false);
452 if (!dev)
453 return;
454
455 if (dev->type != &simple_device_type)
456 return;
457
458 if (add && system_if_force_external(dev->ifname))
459 return;
460
461 device_set_present(dev, add);
462 }
463
464 static void
465 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
466 {
467 struct event_socket *ev = container_of(u, struct event_socket, uloop);
468 struct sockaddr_nl nla;
469 unsigned char *buf = NULL;
470 int size;
471
472 while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
473 if (nla.nl_pid == 0)
474 handle_hotplug_msg((char *) buf, size);
475
476 free(buf);
477 }
478 }
479
480 static int system_rtnl_call(struct nl_msg *msg)
481 {
482 int ret;
483
484 ret = nl_send_auto_complete(sock_rtnl, msg);
485 nlmsg_free(msg);
486
487 if (ret < 0)
488 return ret;
489
490 return nl_wait_for_ack(sock_rtnl);
491 }
492
493 int system_bridge_delbr(struct device *bridge)
494 {
495 return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
496 }
497
498 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
499 {
500 struct ifreq ifr;
501
502 memset(&ifr, 0, sizeof(ifr));
503 if (dev)
504 ifr.ifr_ifindex = dev->ifindex;
505 else
506 ifr.ifr_data = data;
507 strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
508 return ioctl(sock_ioctl, cmd, &ifr);
509 }
510
511 static bool system_is_bridge(const char *name, char *buf, int buflen)
512 {
513 struct stat st;
514
515 snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
516 if (stat(buf, &st) < 0)
517 return false;
518
519 return true;
520 }
521
522 static char *system_get_bridge(const char *name, char *buf, int buflen)
523 {
524 char *path;
525 ssize_t len = -1;
526 glob_t gl;
527
528 snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
529 if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
530 return NULL;
531
532 if (gl.gl_pathc > 0)
533 len = readlink(gl.gl_pathv[0], buf, buflen);
534
535 globfree(&gl);
536
537 if (len < 0)
538 return NULL;
539
540 buf[len] = 0;
541 path = strrchr(buf, '/');
542 if (!path)
543 return NULL;
544
545 return path + 1;
546 }
547
548 static void system_bridge_set_wireless(const char *bridge, const char *dev)
549 {
550 snprintf(dev_buf, sizeof(dev_buf),
551 "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
552 bridge, dev);
553 system_set_sysctl(dev_buf, "1");
554 }
555
556 int system_bridge_addif(struct device *bridge, struct device *dev)
557 {
558 char *oldbr;
559 int ret = 0;
560
561 oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
562 if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
563 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
564
565 if (dev->wireless)
566 system_bridge_set_wireless(bridge->ifname, dev->ifname);
567
568 return ret;
569 }
570
571 int system_bridge_delif(struct device *bridge, struct device *dev)
572 {
573 return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
574 }
575
576 int system_if_resolve(struct device *dev)
577 {
578 struct ifreq ifr;
579 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
580 if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
581 return ifr.ifr_ifindex;
582 else
583 return 0;
584 }
585
586 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
587 {
588 struct ifreq ifr;
589
590 memset(&ifr, 0, sizeof(ifr));
591 strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
592 ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
593 ifr.ifr_flags |= add;
594 ifr.ifr_flags &= ~rem;
595 return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
596 }
597
598 struct clear_data {
599 struct nl_msg *msg;
600 struct device *dev;
601 int type;
602 int size;
603 int af;
604 };
605
606
607 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
608 {
609 struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
610
611 return ifa->ifa_index == ifindex;
612 }
613
614 static bool check_route(struct nlmsghdr *hdr, int ifindex)
615 {
616 struct rtmsg *r = NLMSG_DATA(hdr);
617 struct nlattr *tb[__RTA_MAX];
618
619 if (r->rtm_protocol == RTPROT_KERNEL &&
620 r->rtm_family == AF_INET6)
621 return false;
622
623 nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
624 if (!tb[RTA_OIF])
625 return false;
626
627 return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
628 }
629
630 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
631 {
632 return true;
633 }
634
635 static int cb_clear_event(struct nl_msg *msg, void *arg)
636 {
637 struct clear_data *clr = arg;
638 struct nlmsghdr *hdr = nlmsg_hdr(msg);
639 bool (*cb)(struct nlmsghdr *, int ifindex);
640 int type;
641
642 switch(clr->type) {
643 case RTM_GETADDR:
644 type = RTM_DELADDR;
645 if (hdr->nlmsg_type != RTM_NEWADDR)
646 return NL_SKIP;
647
648 cb = check_ifaddr;
649 break;
650 case RTM_GETROUTE:
651 type = RTM_DELROUTE;
652 if (hdr->nlmsg_type != RTM_NEWROUTE)
653 return NL_SKIP;
654
655 cb = check_route;
656 break;
657 case RTM_GETRULE:
658 type = RTM_DELRULE;
659 if (hdr->nlmsg_type != RTM_NEWRULE)
660 return NL_SKIP;
661
662 cb = check_rule;
663 break;
664 default:
665 return NL_SKIP;
666 }
667
668 if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
669 return NL_SKIP;
670
671 if (type == RTM_DELRULE)
672 D(SYSTEM, "Remove a rule\n");
673 else
674 D(SYSTEM, "Remove %s from device %s\n",
675 type == RTM_DELADDR ? "an address" : "a route",
676 clr->dev->ifname);
677 memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
678 hdr = nlmsg_hdr(clr->msg);
679 hdr->nlmsg_type = type;
680 hdr->nlmsg_flags = NLM_F_REQUEST;
681
682 nl_socket_disable_auto_ack(sock_rtnl);
683 nl_send_auto_complete(sock_rtnl, clr->msg);
684 nl_socket_enable_auto_ack(sock_rtnl);
685
686 return NL_SKIP;
687 }
688
689 static int
690 cb_finish_event(struct nl_msg *msg, void *arg)
691 {
692 int *pending = arg;
693 *pending = 0;
694 return NL_STOP;
695 }
696
697 static int
698 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
699 {
700 int *pending = arg;
701 *pending = err->error;
702 return NL_STOP;
703 }
704
705 static void
706 system_if_clear_entries(struct device *dev, int type, int af)
707 {
708 struct clear_data clr;
709 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
710 struct rtmsg rtm = {
711 .rtm_family = af,
712 .rtm_flags = RTM_F_CLONED,
713 };
714 int flags = NLM_F_DUMP;
715 int pending = 1;
716
717 clr.af = af;
718 clr.dev = dev;
719 clr.type = type;
720 switch (type) {
721 case RTM_GETADDR:
722 case RTM_GETRULE:
723 clr.size = sizeof(struct rtgenmsg);
724 break;
725 case RTM_GETROUTE:
726 clr.size = sizeof(struct rtmsg);
727 break;
728 default:
729 return;
730 }
731
732 if (!cb)
733 return;
734
735 clr.msg = nlmsg_alloc_simple(type, flags);
736 if (!clr.msg)
737 goto out;
738
739 nlmsg_append(clr.msg, &rtm, clr.size, 0);
740 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
741 nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
742 nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
743
744 nl_send_auto_complete(sock_rtnl, clr.msg);
745 while (pending > 0)
746 nl_recvmsgs(sock_rtnl, cb);
747
748 nlmsg_free(clr.msg);
749 out:
750 nl_cb_put(cb);
751 }
752
753 /*
754 * Clear bridge (membership) state and bring down device
755 */
756 void system_if_clear_state(struct device *dev)
757 {
758 static char buf[256];
759 char *bridge;
760
761 device_set_ifindex(dev, system_if_resolve(dev));
762 if (dev->external || !dev->ifindex)
763 return;
764
765 system_if_flags(dev->ifname, 0, IFF_UP);
766
767 if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
768 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
769 system_bridge_delbr(dev);
770 return;
771 }
772
773 bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
774 if (bridge) {
775 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
776 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
777 }
778
779 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
780 system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
781 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
782 system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
783 system_set_disable_ipv6(dev, "0");
784 }
785
786 static inline unsigned long
787 sec_to_jiffies(int val)
788 {
789 return (unsigned long) val * 100;
790 }
791
792 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
793 {
794 char buf[64];
795 unsigned long args[4] = {};
796
797 if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
798 return -1;
799
800 args[0] = BRCTL_SET_BRIDGE_STP_STATE;
801 args[1] = !!cfg->stp;
802 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
803
804 args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
805 args[1] = sec_to_jiffies(cfg->forward_delay);
806 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
807
808 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
809 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
810
811 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
812 bridge->ifname, cfg->multicast_querier ? "1" : "0");
813
814 snprintf(buf, sizeof(buf), "%i", cfg->hash_max);
815 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/hash_max",
816 bridge->ifname, buf);
817
818 args[0] = BRCTL_SET_BRIDGE_PRIORITY;
819 args[1] = cfg->priority;
820 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
821
822 if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
823 args[0] = BRCTL_SET_AGEING_TIME;
824 args[1] = sec_to_jiffies(cfg->ageing_time);
825 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
826 }
827
828 if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
829 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
830 args[1] = sec_to_jiffies(cfg->hello_time);
831 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
832 }
833
834 if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
835 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
836 args[1] = sec_to_jiffies(cfg->max_age);
837 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
838 }
839
840 return 0;
841 }
842
843 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
844 {
845 struct nl_msg *msg;
846 struct nlattr *linkinfo, *data;
847 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
848 int i, rv;
849 static const struct {
850 const char *name;
851 enum macvlan_mode val;
852 } modes[] = {
853 { "private", MACVLAN_MODE_PRIVATE },
854 { "vepa", MACVLAN_MODE_VEPA },
855 { "bridge", MACVLAN_MODE_BRIDGE },
856 { "passthru", MACVLAN_MODE_PASSTHRU },
857 };
858
859 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
860
861 if (!msg)
862 return -1;
863
864 nlmsg_append(msg, &iim, sizeof(iim), 0);
865
866 if (cfg->flags & MACVLAN_OPT_MACADDR)
867 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
868 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
869 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
870
871 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
872 goto nla_put_failure;
873
874 nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
875
876 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
877 goto nla_put_failure;
878
879 if (cfg->mode) {
880 for (i = 0; i < ARRAY_SIZE(modes); i++) {
881 if (strcmp(cfg->mode, modes[i].name) != 0)
882 continue;
883
884 nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
885 break;
886 }
887 }
888
889 nla_nest_end(msg, data);
890 nla_nest_end(msg, linkinfo);
891
892 rv = system_rtnl_call(msg);
893 if (rv)
894 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
895
896 return rv;
897
898 nla_put_failure:
899 nlmsg_free(msg);
900 return -ENOMEM;
901 }
902
903 static int system_link_del(const char *ifname)
904 {
905 struct nl_msg *msg;
906 struct ifinfomsg iim = {
907 .ifi_family = AF_UNSPEC,
908 .ifi_index = 0,
909 };
910
911 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
912
913 if (!msg)
914 return -1;
915
916 nlmsg_append(msg, &iim, sizeof(iim), 0);
917 nla_put_string(msg, IFLA_IFNAME, ifname);
918 return system_rtnl_call(msg);
919 }
920
921 int system_macvlan_del(struct device *macvlan)
922 {
923 return system_link_del(macvlan->ifname);
924 }
925
926 static int system_vlan(struct device *dev, int id)
927 {
928 struct vlan_ioctl_args ifr = {
929 .cmd = SET_VLAN_NAME_TYPE_CMD,
930 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
931 };
932
933 ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
934
935 if (id < 0) {
936 ifr.cmd = DEL_VLAN_CMD;
937 ifr.u.VID = 0;
938 } else {
939 ifr.cmd = ADD_VLAN_CMD;
940 ifr.u.VID = id;
941 }
942 strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
943 return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
944 }
945
946 int system_vlan_add(struct device *dev, int id)
947 {
948 return system_vlan(dev, id);
949 }
950
951 int system_vlan_del(struct device *dev)
952 {
953 return system_vlan(dev, -1);
954 }
955
956 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
957 {
958 struct nl_msg *msg;
959 struct nlattr *linkinfo, *data;
960 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
961 int rv;
962
963 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
964
965 if (!msg)
966 return -1;
967
968 nlmsg_append(msg, &iim, sizeof(iim), 0);
969 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
970 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
971
972 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
973 goto nla_put_failure;
974
975 nla_put_string(msg, IFLA_INFO_KIND, "vlan");
976
977 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
978 goto nla_put_failure;
979
980 nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
981
982 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
983 nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
984 #else
985 if(cfg->proto == VLAN_PROTO_8021AD)
986 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);
987 #endif
988
989 nla_nest_end(msg, data);
990 nla_nest_end(msg, linkinfo);
991
992 rv = system_rtnl_call(msg);
993 if (rv)
994 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
995
996 return rv;
997
998 nla_put_failure:
999 nlmsg_free(msg);
1000 return -ENOMEM;
1001 }
1002
1003 int system_vlandev_del(struct device *vlandev)
1004 {
1005 return system_link_del(vlandev->ifname);
1006 }
1007
1008 static void
1009 system_if_get_settings(struct device *dev, struct device_settings *s)
1010 {
1011 struct ifreq ifr;
1012 char buf[10];
1013
1014 memset(&ifr, 0, sizeof(ifr));
1015 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1016
1017 if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
1018 s->mtu = ifr.ifr_mtu;
1019 s->flags |= DEV_OPT_MTU;
1020 }
1021
1022 s->mtu6 = system_update_ipv6_mtu(dev, 0);
1023 if (s->mtu6 > 0)
1024 s->flags |= DEV_OPT_MTU6;
1025
1026 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
1027 s->txqueuelen = ifr.ifr_qlen;
1028 s->flags |= DEV_OPT_TXQUEUELEN;
1029 }
1030
1031 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
1032 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
1033 s->flags |= DEV_OPT_MACADDR;
1034 }
1035
1036 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
1037 s->ipv6 = !strtoul(buf, NULL, 0);
1038 s->flags |= DEV_OPT_IPV6;
1039 }
1040
1041 if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
1042 s->promisc = ifr.ifr_flags & IFF_PROMISC;
1043 s->flags |= DEV_OPT_PROMISC;
1044 }
1045
1046 if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1047 s->rpfilter = strtoul(buf, NULL, 0);
1048 s->flags |= DEV_OPT_RPFILTER;
1049 }
1050
1051 if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1052 s->acceptlocal = strtoul(buf, NULL, 0);
1053 s->flags |= DEV_OPT_ACCEPTLOCAL;
1054 }
1055
1056 if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1057 s->igmpversion = strtoul(buf, NULL, 0);
1058 s->flags |= DEV_OPT_IGMPVERSION;
1059 }
1060
1061 if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1062 s->mldversion = strtoul(buf, NULL, 0);
1063 s->flags |= DEV_OPT_MLDVERSION;
1064 }
1065
1066 if (!system_get_neigh4reachabletime(dev, buf, sizeof(buf))) {
1067 s->neigh4reachabletime = strtoul(buf, NULL, 0);
1068 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1069 }
1070
1071 if (!system_get_neigh6reachabletime(dev, buf, sizeof(buf))) {
1072 s->neigh6reachabletime = strtoul(buf, NULL, 0);
1073 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1074 }
1075 }
1076
1077 static void
1078 system_if_set_rps_xps_val(const char *path, int val)
1079 {
1080 char val_buf[8];
1081 glob_t gl;
1082 int i;
1083
1084 if (glob(path, 0, NULL, &gl))
1085 return;
1086
1087 snprintf(val_buf, sizeof(val_buf), "%x", val);
1088 for (i = 0; i < gl.gl_pathc; i++)
1089 system_set_sysctl(gl.gl_pathv[i], val_buf);
1090 }
1091
1092 static void
1093 system_if_apply_rps_xps(struct device *dev, struct device_settings *s)
1094 {
1095 long n_cpus = sysconf(_SC_NPROCESSORS_ONLN);
1096 int val;
1097
1098 if (n_cpus < 2)
1099 return;
1100
1101 val = (1 << n_cpus) - 1;
1102 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/rps_cpus", dev->ifname);
1103 system_if_set_rps_xps_val(dev_buf, s->rps ? val : 0);
1104
1105 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/xps_cpus", dev->ifname);
1106 system_if_set_rps_xps_val(dev_buf, s->xps ? val : 0);
1107 }
1108
1109 void
1110 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1111 {
1112 struct ifreq ifr;
1113
1114 memset(&ifr, 0, sizeof(ifr));
1115 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1116 if (s->flags & DEV_OPT_MTU & apply_mask) {
1117 ifr.ifr_mtu = s->mtu;
1118 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1119 s->flags &= ~DEV_OPT_MTU;
1120 }
1121 if (s->flags & DEV_OPT_MTU6 & apply_mask) {
1122 system_update_ipv6_mtu(dev, s->mtu6);
1123 }
1124 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1125 ifr.ifr_qlen = s->txqueuelen;
1126 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1127 s->flags &= ~DEV_OPT_TXQUEUELEN;
1128 }
1129 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1130 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1131 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1132 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1133 s->flags &= ~DEV_OPT_MACADDR;
1134 }
1135 if (s->flags & DEV_OPT_IPV6 & apply_mask)
1136 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1137 if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1138 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1139 !s->promisc ? IFF_PROMISC : 0) < 0)
1140 s->flags &= ~DEV_OPT_PROMISC;
1141 }
1142 if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1143 char buf[2];
1144
1145 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1146 system_set_rpfilter(dev, buf);
1147 }
1148 if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1149 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1150 if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1151 char buf[2];
1152
1153 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1154 system_set_igmpversion(dev, buf);
1155 }
1156 if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1157 char buf[2];
1158
1159 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1160 system_set_mldversion(dev, buf);
1161 }
1162 if (s->flags & DEV_OPT_NEIGHREACHABLETIME & apply_mask) {
1163 char buf[12];
1164
1165 snprintf(buf, sizeof(buf), "%d", s->neigh4reachabletime);
1166 system_set_neigh4reachabletime(dev, buf);
1167 snprintf(buf, sizeof(buf), "%d", s->neigh6reachabletime);
1168 system_set_neigh6reachabletime(dev, buf);
1169 }
1170
1171 system_if_apply_rps_xps(dev, s);
1172 }
1173
1174 int system_if_up(struct device *dev)
1175 {
1176 system_if_get_settings(dev, &dev->orig_settings);
1177 /* Only keep orig settings based on what needs to be set */
1178 dev->orig_settings.flags &= dev->settings.flags;
1179 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1180 return system_if_flags(dev->ifname, IFF_UP, 0);
1181 }
1182
1183 int system_if_down(struct device *dev)
1184 {
1185 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1186 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1187 return ret;
1188 }
1189
1190 struct if_check_data {
1191 struct device *dev;
1192 int pending;
1193 int ret;
1194 };
1195
1196 #ifndef IFF_LOWER_UP
1197 #define IFF_LOWER_UP 0x10000
1198 #endif
1199
1200 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1201 {
1202 struct nlmsghdr *nh = nlmsg_hdr(msg);
1203 struct ifinfomsg *ifi = NLMSG_DATA(nh);
1204 struct if_check_data *chk = (struct if_check_data *)arg;
1205
1206 if (nh->nlmsg_type != RTM_NEWLINK)
1207 return NL_SKIP;
1208
1209 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1210 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1211
1212 return NL_OK;
1213 }
1214
1215 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1216 {
1217 struct if_check_data *chk = (struct if_check_data *)arg;
1218 chk->pending = 0;
1219 return NL_STOP;
1220 }
1221
1222 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1223 {
1224 struct if_check_data *chk = (struct if_check_data *)arg;
1225
1226 device_set_present(chk->dev, false);
1227 device_set_link(chk->dev, false);
1228 chk->pending = err->error;
1229
1230 return NL_STOP;
1231 }
1232
1233 int system_if_check(struct device *dev)
1234 {
1235 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1236 struct nl_msg *msg;
1237 struct ifinfomsg ifi = {
1238 .ifi_family = AF_UNSPEC,
1239 .ifi_index = 0,
1240 };
1241 struct if_check_data chk = {
1242 .dev = dev,
1243 .pending = 1,
1244 };
1245 int ret = 1;
1246
1247 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1248 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1249 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1250 goto out;
1251
1252 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1253 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1254 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1255
1256 nl_send_auto_complete(sock_rtnl, msg);
1257 while (chk.pending > 0)
1258 nl_recvmsgs(sock_rtnl, cb);
1259
1260 nlmsg_free(msg);
1261 ret = chk.pending;
1262
1263 out:
1264 nl_cb_put(cb);
1265 return ret;
1266 }
1267
1268 struct device *
1269 system_if_get_parent(struct device *dev)
1270 {
1271 char buf[64], *devname;
1272 int ifindex, iflink, len;
1273 FILE *f;
1274
1275 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1276 f = fopen(buf, "r");
1277 if (!f)
1278 return NULL;
1279
1280 len = fread(buf, 1, sizeof(buf) - 1, f);
1281 fclose(f);
1282
1283 if (len <= 0)
1284 return NULL;
1285
1286 buf[len] = 0;
1287 iflink = strtoul(buf, NULL, 0);
1288 ifindex = system_if_resolve(dev);
1289 if (!iflink || iflink == ifindex)
1290 return NULL;
1291
1292 devname = if_indextoname(iflink, buf);
1293 if (!devname)
1294 return NULL;
1295
1296 return device_get(devname, true);
1297 }
1298
1299 static bool
1300 read_string_file(int dir_fd, const char *file, char *buf, int len)
1301 {
1302 bool ret = false;
1303 char *c;
1304 int fd;
1305
1306 fd = openat(dir_fd, file, O_RDONLY);
1307 if (fd < 0)
1308 return false;
1309
1310 retry:
1311 len = read(fd, buf, len - 1);
1312 if (len < 0) {
1313 if (errno == EINTR)
1314 goto retry;
1315 } else if (len > 0) {
1316 buf[len] = 0;
1317
1318 c = strchr(buf, '\n');
1319 if (c)
1320 *c = 0;
1321
1322 ret = true;
1323 }
1324
1325 close(fd);
1326
1327 return ret;
1328 }
1329
1330 static bool
1331 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1332 {
1333 char buf[64];
1334 bool ret = false;
1335
1336 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1337 if (ret)
1338 *val = strtoull(buf, NULL, 0);
1339
1340 return ret;
1341 }
1342
1343 /* Assume advertised flags == supported flags */
1344 static const struct {
1345 uint32_t mask;
1346 const char *name;
1347 } ethtool_link_modes[] = {
1348 { ADVERTISED_10baseT_Half, "10H" },
1349 { ADVERTISED_10baseT_Full, "10F" },
1350 { ADVERTISED_100baseT_Half, "100H" },
1351 { ADVERTISED_100baseT_Full, "100F" },
1352 { ADVERTISED_1000baseT_Half, "1000H" },
1353 { ADVERTISED_1000baseT_Full, "1000F" },
1354 };
1355
1356 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1357 {
1358 int i;
1359 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1360 if (mask & ethtool_link_modes[i].mask)
1361 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1362 }
1363 }
1364
1365 bool
1366 system_if_force_external(const char *ifname)
1367 {
1368 char buf[64];
1369 struct stat s;
1370
1371 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1372 return stat(buf, &s) == 0;
1373 }
1374
1375 int
1376 system_if_dump_info(struct device *dev, struct blob_buf *b)
1377 {
1378 struct ethtool_cmd ecmd;
1379 struct ifreq ifr;
1380 char buf[64], *s;
1381 void *c;
1382 int dir_fd;
1383
1384 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1385 dir_fd = open(buf, O_DIRECTORY);
1386
1387 memset(&ecmd, 0, sizeof(ecmd));
1388 memset(&ifr, 0, sizeof(ifr));
1389 strcpy(ifr.ifr_name, dev->ifname);
1390 ifr.ifr_data = (caddr_t) &ecmd;
1391 ecmd.cmd = ETHTOOL_GSET;
1392
1393 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1394 c = blobmsg_open_array(b, "link-advertising");
1395 system_add_link_modes(b, ecmd.advertising);
1396 blobmsg_close_array(b, c);
1397
1398 c = blobmsg_open_array(b, "link-supported");
1399 system_add_link_modes(b, ecmd.supported);
1400 blobmsg_close_array(b, c);
1401
1402 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1403 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1404 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1405 blobmsg_add_string_buffer(b);
1406 }
1407
1408 close(dir_fd);
1409 return 0;
1410 }
1411
1412 int
1413 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1414 {
1415 const char *const counters[] = {
1416 "collisions", "rx_frame_errors", "tx_compressed",
1417 "multicast", "rx_length_errors", "tx_dropped",
1418 "rx_bytes", "rx_missed_errors", "tx_errors",
1419 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1420 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1421 "rx_dropped", "tx_aborted_errors", "tx_packets",
1422 "rx_errors", "tx_bytes", "tx_window_errors",
1423 "rx_fifo_errors", "tx_carrier_errors",
1424 };
1425 char buf[64];
1426 int stats_dir;
1427 int i;
1428 uint64_t val = 0;
1429
1430 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1431 stats_dir = open(buf, O_DIRECTORY);
1432 if (stats_dir < 0)
1433 return -1;
1434
1435 for (i = 0; i < ARRAY_SIZE(counters); i++)
1436 if (read_uint64_file(stats_dir, counters[i], &val))
1437 blobmsg_add_u64(b, counters[i], val);
1438
1439 close(stats_dir);
1440 return 0;
1441 }
1442
1443 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1444 {
1445 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1446 int alen = v4 ? 4 : 16;
1447 unsigned int flags = 0;
1448 struct ifaddrmsg ifa = {
1449 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1450 .ifa_prefixlen = addr->mask,
1451 .ifa_index = dev->ifindex,
1452 };
1453
1454 struct nl_msg *msg;
1455 if (cmd == RTM_NEWADDR)
1456 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1457
1458 msg = nlmsg_alloc_simple(cmd, flags);
1459 if (!msg)
1460 return -1;
1461
1462 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1463 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1464 if (v4) {
1465 if (addr->broadcast)
1466 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1467 if (addr->point_to_point)
1468 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1469 } else {
1470 time_t now = system_get_rtime();
1471 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1472
1473 if (addr->preferred_until) {
1474 int64_t preferred = addr->preferred_until - now;
1475 if (preferred < 0)
1476 preferred = 0;
1477 else if (preferred > UINT32_MAX)
1478 preferred = UINT32_MAX;
1479
1480 cinfo.ifa_prefered = preferred;
1481 }
1482
1483 if (addr->valid_until) {
1484 int64_t valid = addr->valid_until - now;
1485 if (valid <= 0)
1486 return -1;
1487 else if (valid > UINT32_MAX)
1488 valid = UINT32_MAX;
1489
1490 cinfo.ifa_valid = valid;
1491 }
1492
1493 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1494
1495 if (cmd == RTM_NEWADDR && (addr->flags & DEVADDR_OFFLINK))
1496 nla_put_u32(msg, IFA_FLAGS, IFA_F_NOPREFIXROUTE);
1497 }
1498
1499 return system_rtnl_call(msg);
1500 }
1501
1502 int system_add_address(struct device *dev, struct device_addr *addr)
1503 {
1504 return system_addr(dev, addr, RTM_NEWADDR);
1505 }
1506
1507 int system_del_address(struct device *dev, struct device_addr *addr)
1508 {
1509 return system_addr(dev, addr, RTM_DELADDR);
1510 }
1511
1512 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1513 {
1514 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1515 bool have_gw;
1516 unsigned int flags = 0;
1517
1518 if (alen == 4)
1519 have_gw = !!route->nexthop.in.s_addr;
1520 else
1521 have_gw = route->nexthop.in6.s6_addr32[0] ||
1522 route->nexthop.in6.s6_addr32[1] ||
1523 route->nexthop.in6.s6_addr32[2] ||
1524 route->nexthop.in6.s6_addr32[3];
1525
1526 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1527 ? route->table : RT_TABLE_MAIN;
1528
1529 struct rtmsg rtm = {
1530 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1531 .rtm_dst_len = route->mask,
1532 .rtm_src_len = route->sourcemask,
1533 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1534 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1535 .rtm_scope = RT_SCOPE_NOWHERE,
1536 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1537 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1538 };
1539 struct nl_msg *msg;
1540
1541 if (cmd == RTM_NEWROUTE) {
1542 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1543
1544 if (!dev) { // Add null-route
1545 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1546 rtm.rtm_type = RTN_UNREACHABLE;
1547 }
1548 else
1549 rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1550 }
1551
1552 if (route->flags & DEVROUTE_TYPE) {
1553 rtm.rtm_type = route->type;
1554 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1555 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1556 rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1557 rtm.rtm_table = RT_TABLE_LOCAL;
1558 }
1559
1560 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1561 rtm.rtm_scope = RT_SCOPE_HOST;
1562 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1563 rtm.rtm_type == RTN_ANYCAST) {
1564 rtm.rtm_scope = RT_SCOPE_LINK;
1565 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1566 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1567 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1568 dev = NULL;
1569 }
1570 }
1571
1572 msg = nlmsg_alloc_simple(cmd, flags);
1573 if (!msg)
1574 return -1;
1575
1576 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1577
1578 if (route->mask)
1579 nla_put(msg, RTA_DST, alen, &route->addr);
1580
1581 if (route->sourcemask) {
1582 if (rtm.rtm_family == AF_INET)
1583 nla_put(msg, RTA_PREFSRC, alen, &route->source);
1584 else
1585 nla_put(msg, RTA_SRC, alen, &route->source);
1586 }
1587
1588 if (route->metric > 0)
1589 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1590
1591 if (have_gw)
1592 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1593
1594 if (dev)
1595 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1596
1597 if (table >= 256)
1598 nla_put_u32(msg, RTA_TABLE, table);
1599
1600 if (route->flags & DEVROUTE_MTU) {
1601 struct nlattr *metrics;
1602
1603 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1604 goto nla_put_failure;
1605
1606 nla_put_u32(msg, RTAX_MTU, route->mtu);
1607
1608 nla_nest_end(msg, metrics);
1609 }
1610
1611 return system_rtnl_call(msg);
1612
1613 nla_put_failure:
1614 nlmsg_free(msg);
1615 return -ENOMEM;
1616 }
1617
1618 int system_add_route(struct device *dev, struct device_route *route)
1619 {
1620 return system_rt(dev, route, RTM_NEWROUTE);
1621 }
1622
1623 int system_del_route(struct device *dev, struct device_route *route)
1624 {
1625 return system_rt(dev, route, RTM_DELROUTE);
1626 }
1627
1628 int system_flush_routes(void)
1629 {
1630 const char *names[] = {
1631 "/proc/sys/net/ipv4/route/flush",
1632 "/proc/sys/net/ipv6/route/flush"
1633 };
1634 int fd, i;
1635
1636 for (i = 0; i < ARRAY_SIZE(names); i++) {
1637 fd = open(names[i], O_WRONLY);
1638 if (fd < 0)
1639 continue;
1640
1641 if (write(fd, "-1", 2)) {}
1642 close(fd);
1643 }
1644 return 0;
1645 }
1646
1647 bool system_resolve_rt_type(const char *type, unsigned int *id)
1648 {
1649 return system_rtn_aton(type, id);
1650 }
1651
1652 bool system_resolve_rt_table(const char *name, unsigned int *id)
1653 {
1654 FILE *f;
1655 char *e, buf[128];
1656 unsigned int n, table = RT_TABLE_UNSPEC;
1657
1658 /* first try to parse table as number */
1659 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1660 table = n;
1661
1662 /* handle well known aliases */
1663 else if (!strcmp(name, "default"))
1664 table = RT_TABLE_DEFAULT;
1665 else if (!strcmp(name, "main"))
1666 table = RT_TABLE_MAIN;
1667 else if (!strcmp(name, "local"))
1668 table = RT_TABLE_LOCAL;
1669 else if (!strcmp(name, "prelocal"))
1670 table = RT_TABLE_PRELOCAL;
1671
1672 /* try to look up name in /etc/iproute2/rt_tables */
1673 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1674 {
1675 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1676 {
1677 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1678 continue;
1679
1680 n = strtoul(e, NULL, 10);
1681 e = strtok(NULL, " \t\n");
1682
1683 if (e && !strcmp(e, name))
1684 {
1685 table = n;
1686 break;
1687 }
1688 }
1689
1690 fclose(f);
1691 }
1692
1693 if (table == RT_TABLE_UNSPEC)
1694 return false;
1695
1696 *id = table;
1697 return true;
1698 }
1699
1700 bool system_is_default_rt_table(unsigned int id)
1701 {
1702 return (id == RT_TABLE_MAIN);
1703 }
1704
1705 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1706 {
1707 char *e;
1708 unsigned int n;
1709
1710 if (!strcmp(filter, "strict"))
1711 n = 1;
1712 else if (!strcmp(filter, "loose"))
1713 n = 2;
1714 else {
1715 n = strtoul(filter, &e, 0);
1716 if (*e || e == filter || n > 2)
1717 return false;
1718 }
1719
1720 *id = n;
1721 return true;
1722 }
1723
1724 static int system_iprule(struct iprule *rule, int cmd)
1725 {
1726 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1727
1728 struct nl_msg *msg;
1729 struct rtmsg rtm = {
1730 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1731 .rtm_protocol = RTPROT_STATIC,
1732 .rtm_scope = RT_SCOPE_UNIVERSE,
1733 .rtm_table = RT_TABLE_UNSPEC,
1734 .rtm_type = RTN_UNSPEC,
1735 .rtm_flags = 0,
1736 };
1737
1738 if (cmd == RTM_NEWRULE) {
1739 rtm.rtm_type = RTN_UNICAST;
1740 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1741 }
1742
1743 if (rule->invert)
1744 rtm.rtm_flags |= FIB_RULE_INVERT;
1745
1746 if (rule->flags & IPRULE_SRC)
1747 rtm.rtm_src_len = rule->src_mask;
1748
1749 if (rule->flags & IPRULE_DEST)
1750 rtm.rtm_dst_len = rule->dest_mask;
1751
1752 if (rule->flags & IPRULE_TOS)
1753 rtm.rtm_tos = rule->tos;
1754
1755 if (rule->flags & IPRULE_LOOKUP) {
1756 if (rule->lookup < 256)
1757 rtm.rtm_table = rule->lookup;
1758 }
1759
1760 if (rule->flags & IPRULE_ACTION)
1761 rtm.rtm_type = rule->action;
1762 else if (rule->flags & IPRULE_GOTO)
1763 rtm.rtm_type = FR_ACT_GOTO;
1764 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1765 rtm.rtm_type = FR_ACT_NOP;
1766
1767 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1768
1769 if (!msg)
1770 return -1;
1771
1772 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1773
1774 if (rule->flags & IPRULE_IN)
1775 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1776
1777 if (rule->flags & IPRULE_OUT)
1778 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1779
1780 if (rule->flags & IPRULE_SRC)
1781 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1782
1783 if (rule->flags & IPRULE_DEST)
1784 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1785
1786 if (rule->flags & IPRULE_PRIORITY)
1787 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1788 else if (cmd == RTM_NEWRULE)
1789 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1790
1791 if (rule->flags & IPRULE_FWMARK)
1792 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1793
1794 if (rule->flags & IPRULE_FWMASK)
1795 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1796
1797 if (rule->flags & IPRULE_LOOKUP) {
1798 if (rule->lookup >= 256)
1799 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1800 }
1801
1802 if (rule->flags & IPRULE_GOTO)
1803 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1804
1805 return system_rtnl_call(msg);
1806 }
1807
1808 int system_add_iprule(struct iprule *rule)
1809 {
1810 return system_iprule(rule, RTM_NEWRULE);
1811 }
1812
1813 int system_del_iprule(struct iprule *rule)
1814 {
1815 return system_iprule(rule, RTM_DELRULE);
1816 }
1817
1818 int system_flush_iprules(void)
1819 {
1820 int rv = 0;
1821 struct iprule rule;
1822
1823 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1824 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1825
1826 memset(&rule, 0, sizeof(rule));
1827
1828
1829 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1830
1831 rule.priority = 0;
1832 rule.lookup = RT_TABLE_PRELOCAL;
1833 rv |= system_iprule(&rule, RTM_NEWRULE);
1834
1835 rule.priority = 1;
1836 rule.lookup = RT_TABLE_LOCAL;
1837 rv |= system_iprule(&rule, RTM_NEWRULE);
1838
1839 rule.priority = 32766;
1840 rule.lookup = RT_TABLE_MAIN;
1841 rv |= system_iprule(&rule, RTM_NEWRULE);
1842
1843 rule.priority = 32767;
1844 rule.lookup = RT_TABLE_DEFAULT;
1845 rv |= system_iprule(&rule, RTM_NEWRULE);
1846
1847
1848 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1849
1850 rule.priority = 0;
1851 rule.lookup = RT_TABLE_PRELOCAL;
1852 rv |= system_iprule(&rule, RTM_NEWRULE);
1853
1854 rule.priority = 1;
1855 rule.lookup = RT_TABLE_LOCAL;
1856 rv |= system_iprule(&rule, RTM_NEWRULE);
1857
1858 rule.priority = 32766;
1859 rule.lookup = RT_TABLE_MAIN;
1860 rv |= system_iprule(&rule, RTM_NEWRULE);
1861
1862 return rv;
1863 }
1864
1865 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1866 {
1867 return system_rtn_aton(action, id);
1868 }
1869
1870 time_t system_get_rtime(void)
1871 {
1872 struct timespec ts;
1873 struct timeval tv;
1874
1875 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1876 return ts.tv_sec;
1877
1878 if (gettimeofday(&tv, NULL) == 0)
1879 return tv.tv_sec;
1880
1881 return 0;
1882 }
1883
1884 #ifndef IP_DF
1885 #define IP_DF 0x4000
1886 #endif
1887
1888 static int tunnel_ioctl(const char *name, int cmd, void *p)
1889 {
1890 struct ifreq ifr;
1891
1892 memset(&ifr, 0, sizeof(ifr));
1893 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1894 ifr.ifr_ifru.ifru_data = p;
1895 return ioctl(sock_ioctl, cmd, &ifr);
1896 }
1897
1898 #ifdef IFLA_IPTUN_MAX
1899 #define IP6_FLOWINFO_TCLASS htonl(0x0FF00000)
1900 static int system_add_gre_tunnel(const char *name, const char *kind,
1901 const unsigned int link, struct blob_attr **tb, bool v6)
1902 {
1903 struct nl_msg *nlm;
1904 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1905 struct blob_attr *cur;
1906 uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1907 uint16_t iflags = 0, oflags = 0;
1908 uint8_t tos = 0;
1909 int ret = 0, ttl = 64;
1910
1911 nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1912 if (!nlm)
1913 return -1;
1914
1915 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1916 nla_put_string(nlm, IFLA_IFNAME, name);
1917
1918 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1919 if (!linkinfo) {
1920 ret = -ENOMEM;
1921 goto failure;
1922 }
1923
1924 nla_put_string(nlm, IFLA_INFO_KIND, kind);
1925 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1926 if (!infodata) {
1927 ret = -ENOMEM;
1928 goto failure;
1929 }
1930
1931 if (link)
1932 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1933
1934 if ((cur = tb[TUNNEL_ATTR_TTL]))
1935 ttl = blobmsg_get_u32(cur);
1936
1937 nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1938
1939 if ((cur = tb[TUNNEL_ATTR_TOS])) {
1940 char *str = blobmsg_get_string(cur);
1941 if (strcmp(str, "inherit")) {
1942 unsigned uval;
1943
1944 if (!system_tos_aton(str, &uval)) {
1945 ret = -EINVAL;
1946 goto failure;
1947 }
1948
1949 if (v6)
1950 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1951 else
1952 tos = uval;
1953 } else {
1954 if (v6)
1955 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1956 else
1957 tos = 1;
1958 }
1959 }
1960
1961 if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1962 uint8_t icsum, ocsum, iseqno, oseqno;
1963 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1964 &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1965 ret = -EINVAL;
1966 goto failure;
1967 }
1968
1969 if (ikey)
1970 iflags |= GRE_KEY;
1971
1972 if (okey)
1973 oflags |= GRE_KEY;
1974
1975 if (icsum)
1976 iflags |= GRE_CSUM;
1977
1978 if (ocsum)
1979 oflags |= GRE_CSUM;
1980
1981 if (iseqno)
1982 iflags |= GRE_SEQ;
1983
1984 if (oseqno)
1985 oflags |= GRE_SEQ;
1986 }
1987
1988 if (v6) {
1989 struct in6_addr in6buf;
1990 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1991 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1992 ret = -EINVAL;
1993 goto failure;
1994 }
1995 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1996 }
1997
1998 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1999 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2000 ret = -EINVAL;
2001 goto failure;
2002 }
2003 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
2004 }
2005 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
2006
2007 if (flowinfo)
2008 nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
2009
2010 if (flags)
2011 nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
2012 } else {
2013 struct in_addr inbuf;
2014 bool set_df = true;
2015
2016 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2017 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2018 ret = -EINVAL;
2019 goto failure;
2020 }
2021 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
2022 }
2023
2024 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2025 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2026 ret = -EINVAL;
2027 goto failure;
2028 }
2029 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
2030
2031 if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
2032 if (!okey) {
2033 okey = inbuf.s_addr;
2034 oflags |= GRE_KEY;
2035 }
2036
2037 if (!ikey) {
2038 ikey = inbuf.s_addr;
2039 iflags |= GRE_KEY;
2040 }
2041 }
2042 }
2043
2044 if ((cur = tb[TUNNEL_ATTR_DF]))
2045 set_df = blobmsg_get_bool(cur);
2046
2047 /* ttl !=0 and nopmtudisc are incompatible */
2048 if (ttl && !set_df) {
2049 ret = -EINVAL;
2050 goto failure;
2051 }
2052
2053 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
2054
2055 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
2056 }
2057
2058 if (oflags)
2059 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
2060
2061 if (iflags)
2062 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
2063
2064 if (okey)
2065 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
2066
2067 if (ikey)
2068 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
2069
2070 nla_nest_end(nlm, infodata);
2071 nla_nest_end(nlm, linkinfo);
2072
2073 return system_rtnl_call(nlm);
2074
2075 failure:
2076 nlmsg_free(nlm);
2077 return ret;
2078 }
2079 #endif
2080
2081 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
2082 {
2083 struct blob_attr *cur;
2084 bool set_df = true;
2085 struct ip_tunnel_parm p = {
2086 .link = link,
2087 .iph = {
2088 .version = 4,
2089 .ihl = 5,
2090 .protocol = proto,
2091 }
2092 };
2093
2094 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
2095 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
2096 return -EINVAL;
2097
2098 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
2099 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
2100 return -EINVAL;
2101
2102 if ((cur = tb[TUNNEL_ATTR_DF]))
2103 set_df = blobmsg_get_bool(cur);
2104
2105 if ((cur = tb[TUNNEL_ATTR_TTL]))
2106 p.iph.ttl = blobmsg_get_u32(cur);
2107
2108 if ((cur = tb[TUNNEL_ATTR_TOS])) {
2109 char *str = blobmsg_get_string(cur);
2110 if (strcmp(str, "inherit")) {
2111 unsigned uval;
2112
2113 if (!system_tos_aton(str, &uval))
2114 return -EINVAL;
2115
2116 p.iph.tos = uval;
2117 } else
2118 p.iph.tos = 1;
2119 }
2120
2121 p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2122 /* ttl !=0 and nopmtudisc are incompatible */
2123 if (p.iph.ttl && p.iph.frag_off == 0)
2124 return -EINVAL;
2125
2126 strncpy(p.name, name, sizeof(p.name));
2127
2128 switch (p.iph.protocol) {
2129 case IPPROTO_IPIP:
2130 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2131 case IPPROTO_IPV6:
2132 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2133 default:
2134 break;
2135 }
2136 return -1;
2137 }
2138
2139 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2140 {
2141 struct blob_attr *cur;
2142 const char *str;
2143
2144 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2145 return -EINVAL;
2146 str = blobmsg_data(cur);
2147
2148 if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2149 !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2150 return system_link_del(name);
2151 else
2152 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2153 }
2154
2155 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2156 {
2157 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2158
2159 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2160 blob_data(attr), blob_len(attr));
2161
2162 return __system_del_ip_tunnel(name, tb);
2163 }
2164
2165 int system_update_ipv6_mtu(struct device *dev, int mtu)
2166 {
2167 int ret = -1;
2168 char buf[64];
2169 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2170 dev->ifname);
2171
2172 int fd = open(buf, O_RDWR);
2173
2174 if (!mtu) {
2175 ssize_t len = read(fd, buf, sizeof(buf) - 1);
2176 if (len < 0)
2177 goto out;
2178
2179 buf[len] = 0;
2180 ret = atoi(buf);
2181 } else {
2182 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) > 0)
2183 ret = mtu;
2184 }
2185
2186 out:
2187 close(fd);
2188 return ret;
2189 }
2190
2191 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2192 {
2193 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2194 struct blob_attr *cur;
2195 const char *str;
2196
2197 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2198 blob_data(attr), blob_len(attr));
2199
2200 __system_del_ip_tunnel(name, tb);
2201
2202 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2203 return -EINVAL;
2204 str = blobmsg_data(cur);
2205
2206 unsigned int ttl = 0;
2207 if ((cur = tb[TUNNEL_ATTR_TTL])) {
2208 ttl = blobmsg_get_u32(cur);
2209 if (ttl > 255)
2210 return -EINVAL;
2211 }
2212
2213 unsigned int link = 0;
2214 if ((cur = tb[TUNNEL_ATTR_LINK])) {
2215 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2216 if (!iface)
2217 return -EINVAL;
2218
2219 if (iface->l3_dev.dev)
2220 link = iface->l3_dev.dev->ifindex;
2221 }
2222
2223 if (!strcmp(str, "sit")) {
2224 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2225 return -1;
2226
2227 #ifdef SIOCADD6RD
2228 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2229 unsigned int mask;
2230 struct ip_tunnel_6rd p6;
2231
2232 memset(&p6, 0, sizeof(p6));
2233
2234 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2235 &p6.prefix, &mask) || mask > 128)
2236 return -EINVAL;
2237 p6.prefixlen = mask;
2238
2239 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2240 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2241 &p6.relay_prefix, &mask) || mask > 32)
2242 return -EINVAL;
2243 p6.relay_prefixlen = mask;
2244 }
2245
2246 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2247 __system_del_ip_tunnel(name, tb);
2248 return -1;
2249 }
2250 }
2251 #endif
2252 #ifdef IFLA_IPTUN_MAX
2253 } else if (!strcmp(str, "ipip6")) {
2254 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2255 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2256 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2257 int ret = 0;
2258
2259 if (!nlm)
2260 return -1;
2261
2262 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2263 nla_put_string(nlm, IFLA_IFNAME, name);
2264
2265 if (link)
2266 nla_put_u32(nlm, IFLA_LINK, link);
2267
2268 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2269 if (!linkinfo) {
2270 ret = -ENOMEM;
2271 goto failure;
2272 }
2273 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2274 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2275 if (!infodata) {
2276 ret = -ENOMEM;
2277 goto failure;
2278 }
2279
2280 if (link)
2281 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2282
2283 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2284 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2285 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2286
2287 struct in6_addr in6buf;
2288 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2289 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2290 ret = -EINVAL;
2291 goto failure;
2292 }
2293 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2294 }
2295
2296 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2297 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2298 ret = -EINVAL;
2299 goto failure;
2300 }
2301 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2302 }
2303
2304 #ifdef IFLA_IPTUN_FMR_MAX
2305 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2306 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2307
2308 struct blob_attr *fmr;
2309 unsigned rem, fmrcnt = 0;
2310 blobmsg_for_each_attr(fmr, cur, rem) {
2311 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2312 continue;
2313
2314 unsigned ip4len, ip6len, ealen, offset = 6;
2315 char ip6buf[48];
2316 char ip4buf[16];
2317
2318 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2319 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2320 ret = -EINVAL;
2321 goto failure;
2322 }
2323
2324 struct in6_addr ip6prefix;
2325 struct in_addr ip4prefix;
2326 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2327 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2328 ret = -EINVAL;
2329 goto failure;
2330 }
2331
2332 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2333
2334 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2335 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2336 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2337 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2338 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2339 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2340
2341 nla_nest_end(nlm, rule);
2342 }
2343
2344 nla_nest_end(nlm, fmrs);
2345 }
2346 #endif
2347
2348 nla_nest_end(nlm, infodata);
2349 nla_nest_end(nlm, linkinfo);
2350
2351 return system_rtnl_call(nlm);
2352 failure:
2353 nlmsg_free(nlm);
2354 return ret;
2355 } else if (!strcmp(str, "greip")) {
2356 return system_add_gre_tunnel(name, "gre", link, tb, false);
2357 } else if (!strcmp(str, "gretapip")) {
2358 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2359 } else if (!strcmp(str, "greip6")) {
2360 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2361 } else if (!strcmp(str, "gretapip6")) {
2362 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2363 #endif
2364 } else if (!strcmp(str, "ipip")) {
2365 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2366 }
2367 else
2368 return -EINVAL;
2369
2370 return 0;
2371 }