0ba729e826fc584a652bbb55a81eafbe5d8297ef
[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 unsigned long args[4] = {};
795
796 if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
797 return -1;
798
799 args[0] = BRCTL_SET_BRIDGE_STP_STATE;
800 args[1] = !!cfg->stp;
801 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
802
803 args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
804 args[1] = sec_to_jiffies(cfg->forward_delay);
805 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
806
807 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
808 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
809
810 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
811 bridge->ifname, cfg->multicast_querier ? "1" : "0");
812
813 args[0] = BRCTL_SET_BRIDGE_PRIORITY;
814 args[1] = cfg->priority;
815 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
816
817 if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
818 args[0] = BRCTL_SET_AGEING_TIME;
819 args[1] = sec_to_jiffies(cfg->ageing_time);
820 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
821 }
822
823 if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
824 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
825 args[1] = sec_to_jiffies(cfg->hello_time);
826 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
827 }
828
829 if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
830 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
831 args[1] = sec_to_jiffies(cfg->max_age);
832 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
833 }
834
835 return 0;
836 }
837
838 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
839 {
840 struct nl_msg *msg;
841 struct nlattr *linkinfo, *data;
842 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
843 int i, rv;
844 static const struct {
845 const char *name;
846 enum macvlan_mode val;
847 } modes[] = {
848 { "private", MACVLAN_MODE_PRIVATE },
849 { "vepa", MACVLAN_MODE_VEPA },
850 { "bridge", MACVLAN_MODE_BRIDGE },
851 { "passthru", MACVLAN_MODE_PASSTHRU },
852 };
853
854 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
855
856 if (!msg)
857 return -1;
858
859 nlmsg_append(msg, &iim, sizeof(iim), 0);
860
861 if (cfg->flags & MACVLAN_OPT_MACADDR)
862 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
863 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
864 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
865
866 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
867 goto nla_put_failure;
868
869 nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
870
871 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
872 goto nla_put_failure;
873
874 if (cfg->mode) {
875 for (i = 0; i < ARRAY_SIZE(modes); i++) {
876 if (strcmp(cfg->mode, modes[i].name) != 0)
877 continue;
878
879 nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
880 break;
881 }
882 }
883
884 nla_nest_end(msg, data);
885 nla_nest_end(msg, linkinfo);
886
887 rv = system_rtnl_call(msg);
888 if (rv)
889 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
890
891 return rv;
892
893 nla_put_failure:
894 nlmsg_free(msg);
895 return -ENOMEM;
896 }
897
898 static int system_link_del(const char *ifname)
899 {
900 struct nl_msg *msg;
901 struct ifinfomsg iim = {
902 .ifi_family = AF_UNSPEC,
903 .ifi_index = 0,
904 };
905
906 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
907
908 if (!msg)
909 return -1;
910
911 nlmsg_append(msg, &iim, sizeof(iim), 0);
912 nla_put_string(msg, IFLA_IFNAME, ifname);
913 return system_rtnl_call(msg);
914 }
915
916 int system_macvlan_del(struct device *macvlan)
917 {
918 return system_link_del(macvlan->ifname);
919 }
920
921 static int system_vlan(struct device *dev, int id)
922 {
923 struct vlan_ioctl_args ifr = {
924 .cmd = SET_VLAN_NAME_TYPE_CMD,
925 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
926 };
927
928 ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
929
930 if (id < 0) {
931 ifr.cmd = DEL_VLAN_CMD;
932 ifr.u.VID = 0;
933 } else {
934 ifr.cmd = ADD_VLAN_CMD;
935 ifr.u.VID = id;
936 }
937 strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
938 return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
939 }
940
941 int system_vlan_add(struct device *dev, int id)
942 {
943 return system_vlan(dev, id);
944 }
945
946 int system_vlan_del(struct device *dev)
947 {
948 return system_vlan(dev, -1);
949 }
950
951 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
952 {
953 struct nl_msg *msg;
954 struct nlattr *linkinfo, *data;
955 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
956 int rv;
957
958 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
959
960 if (!msg)
961 return -1;
962
963 nlmsg_append(msg, &iim, sizeof(iim), 0);
964 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
965 nla_put_u32(msg, IFLA_LINK, dev->ifindex);
966
967 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
968 goto nla_put_failure;
969
970 nla_put_string(msg, IFLA_INFO_KIND, "vlan");
971
972 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
973 goto nla_put_failure;
974
975 nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
976
977 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
978 nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
979 #else
980 if(cfg->proto == VLAN_PROTO_8021AD)
981 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);
982 #endif
983
984 nla_nest_end(msg, data);
985 nla_nest_end(msg, linkinfo);
986
987 rv = system_rtnl_call(msg);
988 if (rv)
989 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
990
991 return rv;
992
993 nla_put_failure:
994 nlmsg_free(msg);
995 return -ENOMEM;
996 }
997
998 int system_vlandev_del(struct device *vlandev)
999 {
1000 return system_link_del(vlandev->ifname);
1001 }
1002
1003 static void
1004 system_if_get_settings(struct device *dev, struct device_settings *s)
1005 {
1006 struct ifreq ifr;
1007 char buf[10];
1008
1009 memset(&ifr, 0, sizeof(ifr));
1010 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1011
1012 if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
1013 s->mtu = ifr.ifr_mtu;
1014 s->flags |= DEV_OPT_MTU;
1015 }
1016
1017 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
1018 s->txqueuelen = ifr.ifr_qlen;
1019 s->flags |= DEV_OPT_TXQUEUELEN;
1020 }
1021
1022 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
1023 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
1024 s->flags |= DEV_OPT_MACADDR;
1025 }
1026
1027 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
1028 s->ipv6 = !strtoul(buf, NULL, 0);
1029 s->flags |= DEV_OPT_IPV6;
1030 }
1031
1032 if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
1033 s->promisc = ifr.ifr_flags & IFF_PROMISC;
1034 s->flags |= DEV_OPT_PROMISC;
1035 }
1036
1037 if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1038 s->rpfilter = strtoul(buf, NULL, 0);
1039 s->flags |= DEV_OPT_RPFILTER;
1040 }
1041
1042 if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1043 s->acceptlocal = strtoul(buf, NULL, 0);
1044 s->flags |= DEV_OPT_ACCEPTLOCAL;
1045 }
1046
1047 if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1048 s->igmpversion = strtoul(buf, NULL, 0);
1049 s->flags |= DEV_OPT_IGMPVERSION;
1050 }
1051
1052 if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1053 s->mldversion = strtoul(buf, NULL, 0);
1054 s->flags |= DEV_OPT_MLDVERSION;
1055 }
1056
1057 if (!system_get_neigh4reachabletime(dev, buf, sizeof(buf))) {
1058 s->neigh4reachabletime = strtoul(buf, NULL, 0);
1059 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1060 }
1061
1062 if (!system_get_neigh6reachabletime(dev, buf, sizeof(buf))) {
1063 s->neigh6reachabletime = strtoul(buf, NULL, 0);
1064 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1065 }
1066 }
1067
1068 static void
1069 system_if_set_rps_xps_val(const char *path, int val)
1070 {
1071 char val_buf[8];
1072 glob_t gl;
1073 int i;
1074
1075 if (glob(path, 0, NULL, &gl))
1076 return;
1077
1078 snprintf(val_buf, sizeof(val_buf), "%x", val);
1079 for (i = 0; i < gl.gl_pathc; i++)
1080 system_set_sysctl(gl.gl_pathv[i], val_buf);
1081 }
1082
1083 static void
1084 system_if_apply_rps_xps(struct device *dev, struct device_settings *s)
1085 {
1086 long n_cpus = sysconf(_SC_NPROCESSORS_ONLN);
1087 int val;
1088
1089 if (n_cpus < 2)
1090 return;
1091
1092 val = (1 << n_cpus) - 1;
1093 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/rps_cpus", dev->ifname);
1094 system_if_set_rps_xps_val(dev_buf, s->rps ? val : 0);
1095
1096 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/xps_cpus", dev->ifname);
1097 system_if_set_rps_xps_val(dev_buf, s->xps ? val : 0);
1098 }
1099
1100 void
1101 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1102 {
1103 struct ifreq ifr;
1104
1105 memset(&ifr, 0, sizeof(ifr));
1106 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1107 if (s->flags & DEV_OPT_MTU & apply_mask) {
1108 ifr.ifr_mtu = s->mtu;
1109 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1110 s->flags &= ~DEV_OPT_MTU;
1111 }
1112 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1113 ifr.ifr_qlen = s->txqueuelen;
1114 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1115 s->flags &= ~DEV_OPT_TXQUEUELEN;
1116 }
1117 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1118 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1119 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1120 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1121 s->flags &= ~DEV_OPT_MACADDR;
1122 }
1123 if (s->flags & DEV_OPT_IPV6 & apply_mask)
1124 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1125 if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1126 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1127 !s->promisc ? IFF_PROMISC : 0) < 0)
1128 s->flags &= ~DEV_OPT_PROMISC;
1129 }
1130 if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1131 char buf[2];
1132
1133 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1134 system_set_rpfilter(dev, buf);
1135 }
1136 if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1137 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1138 if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1139 char buf[2];
1140
1141 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1142 system_set_igmpversion(dev, buf);
1143 }
1144 if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1145 char buf[2];
1146
1147 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1148 system_set_mldversion(dev, buf);
1149 }
1150 if (s->flags & DEV_OPT_NEIGHREACHABLETIME & apply_mask) {
1151 char buf[12];
1152
1153 snprintf(buf, sizeof(buf), "%d", s->neigh4reachabletime);
1154 system_set_neigh4reachabletime(dev, buf);
1155 snprintf(buf, sizeof(buf), "%d", s->neigh6reachabletime);
1156 system_set_neigh6reachabletime(dev, buf);
1157 }
1158
1159 system_if_apply_rps_xps(dev, s);
1160 }
1161
1162 int system_if_up(struct device *dev)
1163 {
1164 system_if_get_settings(dev, &dev->orig_settings);
1165 /* Only keep orig settings based on what needs to be set */
1166 dev->orig_settings.flags &= dev->settings.flags;
1167 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1168 return system_if_flags(dev->ifname, IFF_UP, 0);
1169 }
1170
1171 int system_if_down(struct device *dev)
1172 {
1173 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1174 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1175 return ret;
1176 }
1177
1178 struct if_check_data {
1179 struct device *dev;
1180 int pending;
1181 int ret;
1182 };
1183
1184 #ifndef IFF_LOWER_UP
1185 #define IFF_LOWER_UP 0x10000
1186 #endif
1187
1188 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1189 {
1190 struct nlmsghdr *nh = nlmsg_hdr(msg);
1191 struct ifinfomsg *ifi = NLMSG_DATA(nh);
1192 struct if_check_data *chk = (struct if_check_data *)arg;
1193
1194 if (nh->nlmsg_type != RTM_NEWLINK)
1195 return NL_SKIP;
1196
1197 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1198 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1199
1200 return NL_OK;
1201 }
1202
1203 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1204 {
1205 struct if_check_data *chk = (struct if_check_data *)arg;
1206 chk->pending = 0;
1207 return NL_STOP;
1208 }
1209
1210 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1211 {
1212 struct if_check_data *chk = (struct if_check_data *)arg;
1213
1214 device_set_present(chk->dev, false);
1215 device_set_link(chk->dev, false);
1216 chk->pending = err->error;
1217
1218 return NL_STOP;
1219 }
1220
1221 int system_if_check(struct device *dev)
1222 {
1223 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1224 struct nl_msg *msg;
1225 struct ifinfomsg ifi = {
1226 .ifi_family = AF_UNSPEC,
1227 .ifi_index = 0,
1228 };
1229 struct if_check_data chk = {
1230 .dev = dev,
1231 .pending = 1,
1232 };
1233 int ret = 1;
1234
1235 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1236 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1237 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1238 goto out;
1239
1240 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1241 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1242 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1243
1244 nl_send_auto_complete(sock_rtnl, msg);
1245 while (chk.pending > 0)
1246 nl_recvmsgs(sock_rtnl, cb);
1247
1248 nlmsg_free(msg);
1249 ret = chk.pending;
1250
1251 out:
1252 nl_cb_put(cb);
1253 return ret;
1254 }
1255
1256 struct device *
1257 system_if_get_parent(struct device *dev)
1258 {
1259 char buf[64], *devname;
1260 int ifindex, iflink, len;
1261 FILE *f;
1262
1263 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1264 f = fopen(buf, "r");
1265 if (!f)
1266 return NULL;
1267
1268 len = fread(buf, 1, sizeof(buf) - 1, f);
1269 fclose(f);
1270
1271 if (len <= 0)
1272 return NULL;
1273
1274 buf[len] = 0;
1275 iflink = strtoul(buf, NULL, 0);
1276 ifindex = system_if_resolve(dev);
1277 if (!iflink || iflink == ifindex)
1278 return NULL;
1279
1280 devname = if_indextoname(iflink, buf);
1281 if (!devname)
1282 return NULL;
1283
1284 return device_get(devname, true);
1285 }
1286
1287 static bool
1288 read_string_file(int dir_fd, const char *file, char *buf, int len)
1289 {
1290 bool ret = false;
1291 char *c;
1292 int fd;
1293
1294 fd = openat(dir_fd, file, O_RDONLY);
1295 if (fd < 0)
1296 return false;
1297
1298 retry:
1299 len = read(fd, buf, len - 1);
1300 if (len < 0) {
1301 if (errno == EINTR)
1302 goto retry;
1303 } else if (len > 0) {
1304 buf[len] = 0;
1305
1306 c = strchr(buf, '\n');
1307 if (c)
1308 *c = 0;
1309
1310 ret = true;
1311 }
1312
1313 close(fd);
1314
1315 return ret;
1316 }
1317
1318 static bool
1319 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1320 {
1321 char buf[64];
1322 bool ret = false;
1323
1324 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1325 if (ret)
1326 *val = strtoull(buf, NULL, 0);
1327
1328 return ret;
1329 }
1330
1331 /* Assume advertised flags == supported flags */
1332 static const struct {
1333 uint32_t mask;
1334 const char *name;
1335 } ethtool_link_modes[] = {
1336 { ADVERTISED_10baseT_Half, "10H" },
1337 { ADVERTISED_10baseT_Full, "10F" },
1338 { ADVERTISED_100baseT_Half, "100H" },
1339 { ADVERTISED_100baseT_Full, "100F" },
1340 { ADVERTISED_1000baseT_Half, "1000H" },
1341 { ADVERTISED_1000baseT_Full, "1000F" },
1342 };
1343
1344 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1345 {
1346 int i;
1347 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1348 if (mask & ethtool_link_modes[i].mask)
1349 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1350 }
1351 }
1352
1353 bool
1354 system_if_force_external(const char *ifname)
1355 {
1356 char buf[64];
1357 struct stat s;
1358
1359 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1360 return stat(buf, &s) == 0;
1361 }
1362
1363 int
1364 system_if_dump_info(struct device *dev, struct blob_buf *b)
1365 {
1366 struct ethtool_cmd ecmd;
1367 struct ifreq ifr;
1368 char buf[64], *s;
1369 void *c;
1370 int dir_fd;
1371
1372 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1373 dir_fd = open(buf, O_DIRECTORY);
1374
1375 memset(&ecmd, 0, sizeof(ecmd));
1376 memset(&ifr, 0, sizeof(ifr));
1377 strcpy(ifr.ifr_name, dev->ifname);
1378 ifr.ifr_data = (caddr_t) &ecmd;
1379 ecmd.cmd = ETHTOOL_GSET;
1380
1381 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1382 c = blobmsg_open_array(b, "link-advertising");
1383 system_add_link_modes(b, ecmd.advertising);
1384 blobmsg_close_array(b, c);
1385
1386 c = blobmsg_open_array(b, "link-supported");
1387 system_add_link_modes(b, ecmd.supported);
1388 blobmsg_close_array(b, c);
1389
1390 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1391 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1392 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1393 blobmsg_add_string_buffer(b);
1394 }
1395
1396 close(dir_fd);
1397 return 0;
1398 }
1399
1400 int
1401 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1402 {
1403 const char *const counters[] = {
1404 "collisions", "rx_frame_errors", "tx_compressed",
1405 "multicast", "rx_length_errors", "tx_dropped",
1406 "rx_bytes", "rx_missed_errors", "tx_errors",
1407 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1408 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1409 "rx_dropped", "tx_aborted_errors", "tx_packets",
1410 "rx_errors", "tx_bytes", "tx_window_errors",
1411 "rx_fifo_errors", "tx_carrier_errors",
1412 };
1413 char buf[64];
1414 int stats_dir;
1415 int i;
1416 uint64_t val = 0;
1417
1418 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1419 stats_dir = open(buf, O_DIRECTORY);
1420 if (stats_dir < 0)
1421 return -1;
1422
1423 for (i = 0; i < ARRAY_SIZE(counters); i++)
1424 if (read_uint64_file(stats_dir, counters[i], &val))
1425 blobmsg_add_u64(b, counters[i], val);
1426
1427 close(stats_dir);
1428 return 0;
1429 }
1430
1431 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1432 {
1433 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1434 int alen = v4 ? 4 : 16;
1435 unsigned int flags = 0;
1436 struct ifaddrmsg ifa = {
1437 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1438 .ifa_prefixlen = addr->mask,
1439 .ifa_index = dev->ifindex,
1440 };
1441
1442 struct nl_msg *msg;
1443 if (cmd == RTM_NEWADDR)
1444 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1445
1446 msg = nlmsg_alloc_simple(cmd, flags);
1447 if (!msg)
1448 return -1;
1449
1450 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1451 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1452 if (v4) {
1453 if (addr->broadcast)
1454 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1455 if (addr->point_to_point)
1456 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1457 } else {
1458 time_t now = system_get_rtime();
1459 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1460
1461 if (addr->preferred_until) {
1462 int64_t preferred = addr->preferred_until - now;
1463 if (preferred < 0)
1464 preferred = 0;
1465 else if (preferred > UINT32_MAX)
1466 preferred = UINT32_MAX;
1467
1468 cinfo.ifa_prefered = preferred;
1469 }
1470
1471 if (addr->valid_until) {
1472 int64_t valid = addr->valid_until - now;
1473 if (valid <= 0)
1474 return -1;
1475 else if (valid > UINT32_MAX)
1476 valid = UINT32_MAX;
1477
1478 cinfo.ifa_valid = valid;
1479 }
1480
1481 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1482
1483 if (cmd == RTM_NEWADDR && (addr->flags & DEVADDR_OFFLINK))
1484 nla_put_u32(msg, IFA_FLAGS, IFA_F_NOPREFIXROUTE);
1485 }
1486
1487 return system_rtnl_call(msg);
1488 }
1489
1490 int system_add_address(struct device *dev, struct device_addr *addr)
1491 {
1492 return system_addr(dev, addr, RTM_NEWADDR);
1493 }
1494
1495 int system_del_address(struct device *dev, struct device_addr *addr)
1496 {
1497 return system_addr(dev, addr, RTM_DELADDR);
1498 }
1499
1500 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1501 {
1502 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1503 bool have_gw;
1504 unsigned int flags = 0;
1505
1506 if (alen == 4)
1507 have_gw = !!route->nexthop.in.s_addr;
1508 else
1509 have_gw = route->nexthop.in6.s6_addr32[0] ||
1510 route->nexthop.in6.s6_addr32[1] ||
1511 route->nexthop.in6.s6_addr32[2] ||
1512 route->nexthop.in6.s6_addr32[3];
1513
1514 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1515 ? route->table : RT_TABLE_MAIN;
1516
1517 struct rtmsg rtm = {
1518 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1519 .rtm_dst_len = route->mask,
1520 .rtm_src_len = route->sourcemask,
1521 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1522 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1523 .rtm_scope = RT_SCOPE_NOWHERE,
1524 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1525 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1526 };
1527 struct nl_msg *msg;
1528
1529 if (cmd == RTM_NEWROUTE) {
1530 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1531
1532 if (!dev) { // Add null-route
1533 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1534 rtm.rtm_type = RTN_UNREACHABLE;
1535 }
1536 else
1537 rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1538 }
1539
1540 if (route->flags & DEVROUTE_TYPE) {
1541 rtm.rtm_type = route->type;
1542 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1543 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1544 rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1545 rtm.rtm_table = RT_TABLE_LOCAL;
1546 }
1547
1548 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1549 rtm.rtm_scope = RT_SCOPE_HOST;
1550 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1551 rtm.rtm_type == RTN_ANYCAST) {
1552 rtm.rtm_scope = RT_SCOPE_LINK;
1553 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1554 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1555 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1556 dev = NULL;
1557 }
1558 }
1559
1560 msg = nlmsg_alloc_simple(cmd, flags);
1561 if (!msg)
1562 return -1;
1563
1564 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1565
1566 if (route->mask)
1567 nla_put(msg, RTA_DST, alen, &route->addr);
1568
1569 if (route->sourcemask) {
1570 if (rtm.rtm_family == AF_INET)
1571 nla_put(msg, RTA_PREFSRC, alen, &route->source);
1572 else
1573 nla_put(msg, RTA_SRC, alen, &route->source);
1574 }
1575
1576 if (route->metric > 0)
1577 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1578
1579 if (have_gw)
1580 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1581
1582 if (dev)
1583 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1584
1585 if (table >= 256)
1586 nla_put_u32(msg, RTA_TABLE, table);
1587
1588 if (route->flags & DEVROUTE_MTU) {
1589 struct nlattr *metrics;
1590
1591 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1592 goto nla_put_failure;
1593
1594 nla_put_u32(msg, RTAX_MTU, route->mtu);
1595
1596 nla_nest_end(msg, metrics);
1597 }
1598
1599 return system_rtnl_call(msg);
1600
1601 nla_put_failure:
1602 nlmsg_free(msg);
1603 return -ENOMEM;
1604 }
1605
1606 int system_add_route(struct device *dev, struct device_route *route)
1607 {
1608 return system_rt(dev, route, RTM_NEWROUTE);
1609 }
1610
1611 int system_del_route(struct device *dev, struct device_route *route)
1612 {
1613 return system_rt(dev, route, RTM_DELROUTE);
1614 }
1615
1616 int system_flush_routes(void)
1617 {
1618 const char *names[] = {
1619 "/proc/sys/net/ipv4/route/flush",
1620 "/proc/sys/net/ipv6/route/flush"
1621 };
1622 int fd, i;
1623
1624 for (i = 0; i < ARRAY_SIZE(names); i++) {
1625 fd = open(names[i], O_WRONLY);
1626 if (fd < 0)
1627 continue;
1628
1629 if (write(fd, "-1", 2)) {}
1630 close(fd);
1631 }
1632 return 0;
1633 }
1634
1635 bool system_resolve_rt_type(const char *type, unsigned int *id)
1636 {
1637 return system_rtn_aton(type, id);
1638 }
1639
1640 bool system_resolve_rt_table(const char *name, unsigned int *id)
1641 {
1642 FILE *f;
1643 char *e, buf[128];
1644 unsigned int n, table = RT_TABLE_UNSPEC;
1645
1646 /* first try to parse table as number */
1647 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1648 table = n;
1649
1650 /* handle well known aliases */
1651 else if (!strcmp(name, "default"))
1652 table = RT_TABLE_DEFAULT;
1653 else if (!strcmp(name, "main"))
1654 table = RT_TABLE_MAIN;
1655 else if (!strcmp(name, "local"))
1656 table = RT_TABLE_LOCAL;
1657 else if (!strcmp(name, "prelocal"))
1658 table = RT_TABLE_PRELOCAL;
1659
1660 /* try to look up name in /etc/iproute2/rt_tables */
1661 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1662 {
1663 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1664 {
1665 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1666 continue;
1667
1668 n = strtoul(e, NULL, 10);
1669 e = strtok(NULL, " \t\n");
1670
1671 if (e && !strcmp(e, name))
1672 {
1673 table = n;
1674 break;
1675 }
1676 }
1677
1678 fclose(f);
1679 }
1680
1681 if (table == RT_TABLE_UNSPEC)
1682 return false;
1683
1684 *id = table;
1685 return true;
1686 }
1687
1688 bool system_is_default_rt_table(unsigned int id)
1689 {
1690 return (id == RT_TABLE_MAIN);
1691 }
1692
1693 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1694 {
1695 char *e;
1696 unsigned int n;
1697
1698 if (!strcmp(filter, "strict"))
1699 n = 1;
1700 else if (!strcmp(filter, "loose"))
1701 n = 2;
1702 else {
1703 n = strtoul(filter, &e, 0);
1704 if (*e || e == filter || n > 2)
1705 return false;
1706 }
1707
1708 *id = n;
1709 return true;
1710 }
1711
1712 static int system_iprule(struct iprule *rule, int cmd)
1713 {
1714 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1715
1716 struct nl_msg *msg;
1717 struct rtmsg rtm = {
1718 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1719 .rtm_protocol = RTPROT_STATIC,
1720 .rtm_scope = RT_SCOPE_UNIVERSE,
1721 .rtm_table = RT_TABLE_UNSPEC,
1722 .rtm_type = RTN_UNSPEC,
1723 .rtm_flags = 0,
1724 };
1725
1726 if (cmd == RTM_NEWRULE) {
1727 rtm.rtm_type = RTN_UNICAST;
1728 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1729 }
1730
1731 if (rule->invert)
1732 rtm.rtm_flags |= FIB_RULE_INVERT;
1733
1734 if (rule->flags & IPRULE_SRC)
1735 rtm.rtm_src_len = rule->src_mask;
1736
1737 if (rule->flags & IPRULE_DEST)
1738 rtm.rtm_dst_len = rule->dest_mask;
1739
1740 if (rule->flags & IPRULE_TOS)
1741 rtm.rtm_tos = rule->tos;
1742
1743 if (rule->flags & IPRULE_LOOKUP) {
1744 if (rule->lookup < 256)
1745 rtm.rtm_table = rule->lookup;
1746 }
1747
1748 if (rule->flags & IPRULE_ACTION)
1749 rtm.rtm_type = rule->action;
1750 else if (rule->flags & IPRULE_GOTO)
1751 rtm.rtm_type = FR_ACT_GOTO;
1752 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1753 rtm.rtm_type = FR_ACT_NOP;
1754
1755 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1756
1757 if (!msg)
1758 return -1;
1759
1760 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1761
1762 if (rule->flags & IPRULE_IN)
1763 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1764
1765 if (rule->flags & IPRULE_OUT)
1766 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1767
1768 if (rule->flags & IPRULE_SRC)
1769 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1770
1771 if (rule->flags & IPRULE_DEST)
1772 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1773
1774 if (rule->flags & IPRULE_PRIORITY)
1775 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1776 else if (cmd == RTM_NEWRULE)
1777 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1778
1779 if (rule->flags & IPRULE_FWMARK)
1780 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1781
1782 if (rule->flags & IPRULE_FWMASK)
1783 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1784
1785 if (rule->flags & IPRULE_LOOKUP) {
1786 if (rule->lookup >= 256)
1787 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1788 }
1789
1790 if (rule->flags & IPRULE_GOTO)
1791 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1792
1793 return system_rtnl_call(msg);
1794 }
1795
1796 int system_add_iprule(struct iprule *rule)
1797 {
1798 return system_iprule(rule, RTM_NEWRULE);
1799 }
1800
1801 int system_del_iprule(struct iprule *rule)
1802 {
1803 return system_iprule(rule, RTM_DELRULE);
1804 }
1805
1806 int system_flush_iprules(void)
1807 {
1808 int rv = 0;
1809 struct iprule rule;
1810
1811 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1812 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1813
1814 memset(&rule, 0, sizeof(rule));
1815
1816
1817 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1818
1819 rule.priority = 0;
1820 rule.lookup = RT_TABLE_PRELOCAL;
1821 rv |= system_iprule(&rule, RTM_NEWRULE);
1822
1823 rule.priority = 1;
1824 rule.lookup = RT_TABLE_LOCAL;
1825 rv |= system_iprule(&rule, RTM_NEWRULE);
1826
1827 rule.priority = 32766;
1828 rule.lookup = RT_TABLE_MAIN;
1829 rv |= system_iprule(&rule, RTM_NEWRULE);
1830
1831 rule.priority = 32767;
1832 rule.lookup = RT_TABLE_DEFAULT;
1833 rv |= system_iprule(&rule, RTM_NEWRULE);
1834
1835
1836 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1837
1838 rule.priority = 0;
1839 rule.lookup = RT_TABLE_PRELOCAL;
1840 rv |= system_iprule(&rule, RTM_NEWRULE);
1841
1842 rule.priority = 1;
1843 rule.lookup = RT_TABLE_LOCAL;
1844 rv |= system_iprule(&rule, RTM_NEWRULE);
1845
1846 rule.priority = 32766;
1847 rule.lookup = RT_TABLE_MAIN;
1848 rv |= system_iprule(&rule, RTM_NEWRULE);
1849
1850 return rv;
1851 }
1852
1853 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1854 {
1855 return system_rtn_aton(action, id);
1856 }
1857
1858 time_t system_get_rtime(void)
1859 {
1860 struct timespec ts;
1861 struct timeval tv;
1862
1863 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1864 return ts.tv_sec;
1865
1866 if (gettimeofday(&tv, NULL) == 0)
1867 return tv.tv_sec;
1868
1869 return 0;
1870 }
1871
1872 #ifndef IP_DF
1873 #define IP_DF 0x4000
1874 #endif
1875
1876 static int tunnel_ioctl(const char *name, int cmd, void *p)
1877 {
1878 struct ifreq ifr;
1879
1880 memset(&ifr, 0, sizeof(ifr));
1881 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1882 ifr.ifr_ifru.ifru_data = p;
1883 return ioctl(sock_ioctl, cmd, &ifr);
1884 }
1885
1886 #ifdef IFLA_IPTUN_MAX
1887 #define IP6_FLOWINFO_TCLASS htonl(0x0FF00000)
1888 static int system_add_gre_tunnel(const char *name, const char *kind,
1889 const unsigned int link, struct blob_attr **tb, bool v6)
1890 {
1891 struct nl_msg *nlm;
1892 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1893 struct blob_attr *cur;
1894 uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1895 uint16_t iflags = 0, oflags = 0;
1896 uint8_t tos = 0;
1897 int ret = 0, ttl = 64;
1898
1899 nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1900 if (!nlm)
1901 return -1;
1902
1903 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1904 nla_put_string(nlm, IFLA_IFNAME, name);
1905
1906 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1907 if (!linkinfo) {
1908 ret = -ENOMEM;
1909 goto failure;
1910 }
1911
1912 nla_put_string(nlm, IFLA_INFO_KIND, kind);
1913 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1914 if (!infodata) {
1915 ret = -ENOMEM;
1916 goto failure;
1917 }
1918
1919 if (link)
1920 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1921
1922 if ((cur = tb[TUNNEL_ATTR_TTL]))
1923 ttl = blobmsg_get_u32(cur);
1924
1925 nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1926
1927 if ((cur = tb[TUNNEL_ATTR_TOS])) {
1928 char *str = blobmsg_get_string(cur);
1929 if (strcmp(str, "inherit")) {
1930 unsigned uval;
1931
1932 if (!system_tos_aton(str, &uval)) {
1933 ret = -EINVAL;
1934 goto failure;
1935 }
1936
1937 if (v6)
1938 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1939 else
1940 tos = uval;
1941 } else {
1942 if (v6)
1943 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1944 else
1945 tos = 1;
1946 }
1947 }
1948
1949 if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1950 uint8_t icsum, ocsum, iseqno, oseqno;
1951 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1952 &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1953 ret = -EINVAL;
1954 goto failure;
1955 }
1956
1957 if (ikey)
1958 iflags |= GRE_KEY;
1959
1960 if (okey)
1961 oflags |= GRE_KEY;
1962
1963 if (icsum)
1964 iflags |= GRE_CSUM;
1965
1966 if (ocsum)
1967 oflags |= GRE_CSUM;
1968
1969 if (iseqno)
1970 iflags |= GRE_SEQ;
1971
1972 if (oseqno)
1973 oflags |= GRE_SEQ;
1974 }
1975
1976 if (v6) {
1977 struct in6_addr in6buf;
1978 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1979 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1980 ret = -EINVAL;
1981 goto failure;
1982 }
1983 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1984 }
1985
1986 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1987 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1988 ret = -EINVAL;
1989 goto failure;
1990 }
1991 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1992 }
1993 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1994
1995 if (flowinfo)
1996 nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1997
1998 if (flags)
1999 nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
2000 } else {
2001 struct in_addr inbuf;
2002 bool set_df = true;
2003
2004 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2005 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2006 ret = -EINVAL;
2007 goto failure;
2008 }
2009 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
2010 }
2011
2012 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2013 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2014 ret = -EINVAL;
2015 goto failure;
2016 }
2017 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
2018
2019 if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
2020 if (!okey) {
2021 okey = inbuf.s_addr;
2022 oflags |= GRE_KEY;
2023 }
2024
2025 if (!ikey) {
2026 ikey = inbuf.s_addr;
2027 iflags |= GRE_KEY;
2028 }
2029 }
2030 }
2031
2032 if ((cur = tb[TUNNEL_ATTR_DF]))
2033 set_df = blobmsg_get_bool(cur);
2034
2035 /* ttl !=0 and nopmtudisc are incompatible */
2036 if (ttl && !set_df) {
2037 ret = -EINVAL;
2038 goto failure;
2039 }
2040
2041 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
2042
2043 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
2044 }
2045
2046 if (oflags)
2047 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
2048
2049 if (iflags)
2050 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
2051
2052 if (okey)
2053 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
2054
2055 if (ikey)
2056 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
2057
2058 nla_nest_end(nlm, infodata);
2059 nla_nest_end(nlm, linkinfo);
2060
2061 return system_rtnl_call(nlm);
2062
2063 failure:
2064 nlmsg_free(nlm);
2065 return ret;
2066 }
2067 #endif
2068
2069 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
2070 {
2071 struct blob_attr *cur;
2072 bool set_df = true;
2073 struct ip_tunnel_parm p = {
2074 .link = link,
2075 .iph = {
2076 .version = 4,
2077 .ihl = 5,
2078 .protocol = proto,
2079 }
2080 };
2081
2082 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
2083 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
2084 return -EINVAL;
2085
2086 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
2087 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
2088 return -EINVAL;
2089
2090 if ((cur = tb[TUNNEL_ATTR_DF]))
2091 set_df = blobmsg_get_bool(cur);
2092
2093 if ((cur = tb[TUNNEL_ATTR_TTL]))
2094 p.iph.ttl = blobmsg_get_u32(cur);
2095
2096 if ((cur = tb[TUNNEL_ATTR_TOS])) {
2097 char *str = blobmsg_get_string(cur);
2098 if (strcmp(str, "inherit")) {
2099 unsigned uval;
2100
2101 if (!system_tos_aton(str, &uval))
2102 return -EINVAL;
2103
2104 p.iph.tos = uval;
2105 } else
2106 p.iph.tos = 1;
2107 }
2108
2109 p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2110 /* ttl !=0 and nopmtudisc are incompatible */
2111 if (p.iph.ttl && p.iph.frag_off == 0)
2112 return -EINVAL;
2113
2114 strncpy(p.name, name, sizeof(p.name));
2115
2116 switch (p.iph.protocol) {
2117 case IPPROTO_IPIP:
2118 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2119 case IPPROTO_IPV6:
2120 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2121 default:
2122 break;
2123 }
2124 return -1;
2125 }
2126
2127 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2128 {
2129 struct blob_attr *cur;
2130 const char *str;
2131
2132 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2133 return -EINVAL;
2134 str = blobmsg_data(cur);
2135
2136 if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2137 !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2138 return system_link_del(name);
2139 else
2140 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2141 }
2142
2143 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2144 {
2145 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2146
2147 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2148 blob_data(attr), blob_len(attr));
2149
2150 return __system_del_ip_tunnel(name, tb);
2151 }
2152
2153 int system_update_ipv6_mtu(struct device *dev, int mtu)
2154 {
2155 int ret = -1;
2156 char buf[64];
2157 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2158 dev->ifname);
2159
2160 int fd = open(buf, O_RDWR);
2161 ssize_t len = read(fd, buf, sizeof(buf) - 1);
2162 if (len < 0)
2163 goto out;
2164
2165 buf[len] = 0;
2166 ret = atoi(buf);
2167
2168 if (!mtu || ret <= mtu)
2169 goto out;
2170
2171 lseek(fd, 0, SEEK_SET);
2172 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2173 ret = -1;
2174
2175 out:
2176 close(fd);
2177 return ret;
2178 }
2179
2180 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2181 {
2182 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2183 struct blob_attr *cur;
2184 const char *str;
2185
2186 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2187 blob_data(attr), blob_len(attr));
2188
2189 __system_del_ip_tunnel(name, tb);
2190
2191 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2192 return -EINVAL;
2193 str = blobmsg_data(cur);
2194
2195 unsigned int ttl = 0;
2196 if ((cur = tb[TUNNEL_ATTR_TTL])) {
2197 ttl = blobmsg_get_u32(cur);
2198 if (ttl > 255)
2199 return -EINVAL;
2200 }
2201
2202 unsigned int link = 0;
2203 if ((cur = tb[TUNNEL_ATTR_LINK])) {
2204 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2205 if (!iface)
2206 return -EINVAL;
2207
2208 if (iface->l3_dev.dev)
2209 link = iface->l3_dev.dev->ifindex;
2210 }
2211
2212 if (!strcmp(str, "sit")) {
2213 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2214 return -1;
2215
2216 #ifdef SIOCADD6RD
2217 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2218 unsigned int mask;
2219 struct ip_tunnel_6rd p6;
2220
2221 memset(&p6, 0, sizeof(p6));
2222
2223 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2224 &p6.prefix, &mask) || mask > 128)
2225 return -EINVAL;
2226 p6.prefixlen = mask;
2227
2228 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2229 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2230 &p6.relay_prefix, &mask) || mask > 32)
2231 return -EINVAL;
2232 p6.relay_prefixlen = mask;
2233 }
2234
2235 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2236 __system_del_ip_tunnel(name, tb);
2237 return -1;
2238 }
2239 }
2240 #endif
2241 #ifdef IFLA_IPTUN_MAX
2242 } else if (!strcmp(str, "ipip6")) {
2243 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2244 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2245 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2246 int ret = 0;
2247
2248 if (!nlm)
2249 return -1;
2250
2251 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2252 nla_put_string(nlm, IFLA_IFNAME, name);
2253
2254 if (link)
2255 nla_put_u32(nlm, IFLA_LINK, link);
2256
2257 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2258 if (!linkinfo) {
2259 ret = -ENOMEM;
2260 goto failure;
2261 }
2262 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2263 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2264 if (!infodata) {
2265 ret = -ENOMEM;
2266 goto failure;
2267 }
2268
2269 if (link)
2270 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2271
2272 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2273 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2274 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2275
2276 struct in6_addr in6buf;
2277 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2278 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2279 ret = -EINVAL;
2280 goto failure;
2281 }
2282 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2283 }
2284
2285 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2286 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2287 ret = -EINVAL;
2288 goto failure;
2289 }
2290 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2291 }
2292
2293 #ifdef IFLA_IPTUN_FMR_MAX
2294 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2295 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2296
2297 struct blob_attr *fmr;
2298 unsigned rem, fmrcnt = 0;
2299 blobmsg_for_each_attr(fmr, cur, rem) {
2300 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2301 continue;
2302
2303 unsigned ip4len, ip6len, ealen, offset = 6;
2304 char ip6buf[48];
2305 char ip4buf[16];
2306
2307 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2308 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2309 ret = -EINVAL;
2310 goto failure;
2311 }
2312
2313 struct in6_addr ip6prefix;
2314 struct in_addr ip4prefix;
2315 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2316 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2317 ret = -EINVAL;
2318 goto failure;
2319 }
2320
2321 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2322
2323 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2324 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2325 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2326 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2327 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2328 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2329
2330 nla_nest_end(nlm, rule);
2331 }
2332
2333 nla_nest_end(nlm, fmrs);
2334 }
2335 #endif
2336
2337 nla_nest_end(nlm, infodata);
2338 nla_nest_end(nlm, linkinfo);
2339
2340 return system_rtnl_call(nlm);
2341 failure:
2342 nlmsg_free(nlm);
2343 return ret;
2344 } else if (!strcmp(str, "greip")) {
2345 return system_add_gre_tunnel(name, "gre", link, tb, false);
2346 } else if (!strcmp(str, "gretapip")) {
2347 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2348 } else if (!strcmp(str, "greip6")) {
2349 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2350 } else if (!strcmp(str, "gretapip6")) {
2351 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2352 #endif
2353 } else if (!strcmp(str, "ipip")) {
2354 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2355 }
2356 else
2357 return -EINVAL;
2358
2359 return 0;
2360 }