bridge: allow setting hash_max value
[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 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
1023 s->txqueuelen = ifr.ifr_qlen;
1024 s->flags |= DEV_OPT_TXQUEUELEN;
1025 }
1026
1027 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
1028 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
1029 s->flags |= DEV_OPT_MACADDR;
1030 }
1031
1032 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
1033 s->ipv6 = !strtoul(buf, NULL, 0);
1034 s->flags |= DEV_OPT_IPV6;
1035 }
1036
1037 if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
1038 s->promisc = ifr.ifr_flags & IFF_PROMISC;
1039 s->flags |= DEV_OPT_PROMISC;
1040 }
1041
1042 if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1043 s->rpfilter = strtoul(buf, NULL, 0);
1044 s->flags |= DEV_OPT_RPFILTER;
1045 }
1046
1047 if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1048 s->acceptlocal = strtoul(buf, NULL, 0);
1049 s->flags |= DEV_OPT_ACCEPTLOCAL;
1050 }
1051
1052 if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1053 s->igmpversion = strtoul(buf, NULL, 0);
1054 s->flags |= DEV_OPT_IGMPVERSION;
1055 }
1056
1057 if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1058 s->mldversion = strtoul(buf, NULL, 0);
1059 s->flags |= DEV_OPT_MLDVERSION;
1060 }
1061
1062 if (!system_get_neigh4reachabletime(dev, buf, sizeof(buf))) {
1063 s->neigh4reachabletime = strtoul(buf, NULL, 0);
1064 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1065 }
1066
1067 if (!system_get_neigh6reachabletime(dev, buf, sizeof(buf))) {
1068 s->neigh6reachabletime = strtoul(buf, NULL, 0);
1069 s->flags |= DEV_OPT_NEIGHREACHABLETIME;
1070 }
1071 }
1072
1073 static void
1074 system_if_set_rps_xps_val(const char *path, int val)
1075 {
1076 char val_buf[8];
1077 glob_t gl;
1078 int i;
1079
1080 if (glob(path, 0, NULL, &gl))
1081 return;
1082
1083 snprintf(val_buf, sizeof(val_buf), "%x", val);
1084 for (i = 0; i < gl.gl_pathc; i++)
1085 system_set_sysctl(gl.gl_pathv[i], val_buf);
1086 }
1087
1088 static void
1089 system_if_apply_rps_xps(struct device *dev, struct device_settings *s)
1090 {
1091 long n_cpus = sysconf(_SC_NPROCESSORS_ONLN);
1092 int val;
1093
1094 if (n_cpus < 2)
1095 return;
1096
1097 val = (1 << n_cpus) - 1;
1098 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/rps_cpus", dev->ifname);
1099 system_if_set_rps_xps_val(dev_buf, s->rps ? val : 0);
1100
1101 snprintf(dev_buf, sizeof(dev_buf), "/sys/class/net/%s/queues/*/xps_cpus", dev->ifname);
1102 system_if_set_rps_xps_val(dev_buf, s->xps ? val : 0);
1103 }
1104
1105 void
1106 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1107 {
1108 struct ifreq ifr;
1109
1110 memset(&ifr, 0, sizeof(ifr));
1111 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1112 if (s->flags & DEV_OPT_MTU & apply_mask) {
1113 ifr.ifr_mtu = s->mtu;
1114 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1115 s->flags &= ~DEV_OPT_MTU;
1116 }
1117 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1118 ifr.ifr_qlen = s->txqueuelen;
1119 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1120 s->flags &= ~DEV_OPT_TXQUEUELEN;
1121 }
1122 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1123 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1124 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1125 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1126 s->flags &= ~DEV_OPT_MACADDR;
1127 }
1128 if (s->flags & DEV_OPT_IPV6 & apply_mask)
1129 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1130 if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1131 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1132 !s->promisc ? IFF_PROMISC : 0) < 0)
1133 s->flags &= ~DEV_OPT_PROMISC;
1134 }
1135 if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1136 char buf[2];
1137
1138 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1139 system_set_rpfilter(dev, buf);
1140 }
1141 if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1142 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1143 if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1144 char buf[2];
1145
1146 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1147 system_set_igmpversion(dev, buf);
1148 }
1149 if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1150 char buf[2];
1151
1152 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1153 system_set_mldversion(dev, buf);
1154 }
1155 if (s->flags & DEV_OPT_NEIGHREACHABLETIME & apply_mask) {
1156 char buf[12];
1157
1158 snprintf(buf, sizeof(buf), "%d", s->neigh4reachabletime);
1159 system_set_neigh4reachabletime(dev, buf);
1160 snprintf(buf, sizeof(buf), "%d", s->neigh6reachabletime);
1161 system_set_neigh6reachabletime(dev, buf);
1162 }
1163
1164 system_if_apply_rps_xps(dev, s);
1165 }
1166
1167 int system_if_up(struct device *dev)
1168 {
1169 system_if_get_settings(dev, &dev->orig_settings);
1170 /* Only keep orig settings based on what needs to be set */
1171 dev->orig_settings.flags &= dev->settings.flags;
1172 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1173 return system_if_flags(dev->ifname, IFF_UP, 0);
1174 }
1175
1176 int system_if_down(struct device *dev)
1177 {
1178 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1179 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1180 return ret;
1181 }
1182
1183 struct if_check_data {
1184 struct device *dev;
1185 int pending;
1186 int ret;
1187 };
1188
1189 #ifndef IFF_LOWER_UP
1190 #define IFF_LOWER_UP 0x10000
1191 #endif
1192
1193 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1194 {
1195 struct nlmsghdr *nh = nlmsg_hdr(msg);
1196 struct ifinfomsg *ifi = NLMSG_DATA(nh);
1197 struct if_check_data *chk = (struct if_check_data *)arg;
1198
1199 if (nh->nlmsg_type != RTM_NEWLINK)
1200 return NL_SKIP;
1201
1202 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1203 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1204
1205 return NL_OK;
1206 }
1207
1208 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1209 {
1210 struct if_check_data *chk = (struct if_check_data *)arg;
1211 chk->pending = 0;
1212 return NL_STOP;
1213 }
1214
1215 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1216 {
1217 struct if_check_data *chk = (struct if_check_data *)arg;
1218
1219 device_set_present(chk->dev, false);
1220 device_set_link(chk->dev, false);
1221 chk->pending = err->error;
1222
1223 return NL_STOP;
1224 }
1225
1226 int system_if_check(struct device *dev)
1227 {
1228 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1229 struct nl_msg *msg;
1230 struct ifinfomsg ifi = {
1231 .ifi_family = AF_UNSPEC,
1232 .ifi_index = 0,
1233 };
1234 struct if_check_data chk = {
1235 .dev = dev,
1236 .pending = 1,
1237 };
1238 int ret = 1;
1239
1240 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1241 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1242 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1243 goto out;
1244
1245 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1246 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1247 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1248
1249 nl_send_auto_complete(sock_rtnl, msg);
1250 while (chk.pending > 0)
1251 nl_recvmsgs(sock_rtnl, cb);
1252
1253 nlmsg_free(msg);
1254 ret = chk.pending;
1255
1256 out:
1257 nl_cb_put(cb);
1258 return ret;
1259 }
1260
1261 struct device *
1262 system_if_get_parent(struct device *dev)
1263 {
1264 char buf[64], *devname;
1265 int ifindex, iflink, len;
1266 FILE *f;
1267
1268 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1269 f = fopen(buf, "r");
1270 if (!f)
1271 return NULL;
1272
1273 len = fread(buf, 1, sizeof(buf) - 1, f);
1274 fclose(f);
1275
1276 if (len <= 0)
1277 return NULL;
1278
1279 buf[len] = 0;
1280 iflink = strtoul(buf, NULL, 0);
1281 ifindex = system_if_resolve(dev);
1282 if (!iflink || iflink == ifindex)
1283 return NULL;
1284
1285 devname = if_indextoname(iflink, buf);
1286 if (!devname)
1287 return NULL;
1288
1289 return device_get(devname, true);
1290 }
1291
1292 static bool
1293 read_string_file(int dir_fd, const char *file, char *buf, int len)
1294 {
1295 bool ret = false;
1296 char *c;
1297 int fd;
1298
1299 fd = openat(dir_fd, file, O_RDONLY);
1300 if (fd < 0)
1301 return false;
1302
1303 retry:
1304 len = read(fd, buf, len - 1);
1305 if (len < 0) {
1306 if (errno == EINTR)
1307 goto retry;
1308 } else if (len > 0) {
1309 buf[len] = 0;
1310
1311 c = strchr(buf, '\n');
1312 if (c)
1313 *c = 0;
1314
1315 ret = true;
1316 }
1317
1318 close(fd);
1319
1320 return ret;
1321 }
1322
1323 static bool
1324 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1325 {
1326 char buf[64];
1327 bool ret = false;
1328
1329 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1330 if (ret)
1331 *val = strtoull(buf, NULL, 0);
1332
1333 return ret;
1334 }
1335
1336 /* Assume advertised flags == supported flags */
1337 static const struct {
1338 uint32_t mask;
1339 const char *name;
1340 } ethtool_link_modes[] = {
1341 { ADVERTISED_10baseT_Half, "10H" },
1342 { ADVERTISED_10baseT_Full, "10F" },
1343 { ADVERTISED_100baseT_Half, "100H" },
1344 { ADVERTISED_100baseT_Full, "100F" },
1345 { ADVERTISED_1000baseT_Half, "1000H" },
1346 { ADVERTISED_1000baseT_Full, "1000F" },
1347 };
1348
1349 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1350 {
1351 int i;
1352 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1353 if (mask & ethtool_link_modes[i].mask)
1354 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1355 }
1356 }
1357
1358 bool
1359 system_if_force_external(const char *ifname)
1360 {
1361 char buf[64];
1362 struct stat s;
1363
1364 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1365 return stat(buf, &s) == 0;
1366 }
1367
1368 int
1369 system_if_dump_info(struct device *dev, struct blob_buf *b)
1370 {
1371 struct ethtool_cmd ecmd;
1372 struct ifreq ifr;
1373 char buf[64], *s;
1374 void *c;
1375 int dir_fd;
1376
1377 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1378 dir_fd = open(buf, O_DIRECTORY);
1379
1380 memset(&ecmd, 0, sizeof(ecmd));
1381 memset(&ifr, 0, sizeof(ifr));
1382 strcpy(ifr.ifr_name, dev->ifname);
1383 ifr.ifr_data = (caddr_t) &ecmd;
1384 ecmd.cmd = ETHTOOL_GSET;
1385
1386 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1387 c = blobmsg_open_array(b, "link-advertising");
1388 system_add_link_modes(b, ecmd.advertising);
1389 blobmsg_close_array(b, c);
1390
1391 c = blobmsg_open_array(b, "link-supported");
1392 system_add_link_modes(b, ecmd.supported);
1393 blobmsg_close_array(b, c);
1394
1395 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1396 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1397 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1398 blobmsg_add_string_buffer(b);
1399 }
1400
1401 close(dir_fd);
1402 return 0;
1403 }
1404
1405 int
1406 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1407 {
1408 const char *const counters[] = {
1409 "collisions", "rx_frame_errors", "tx_compressed",
1410 "multicast", "rx_length_errors", "tx_dropped",
1411 "rx_bytes", "rx_missed_errors", "tx_errors",
1412 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1413 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1414 "rx_dropped", "tx_aborted_errors", "tx_packets",
1415 "rx_errors", "tx_bytes", "tx_window_errors",
1416 "rx_fifo_errors", "tx_carrier_errors",
1417 };
1418 char buf[64];
1419 int stats_dir;
1420 int i;
1421 uint64_t val = 0;
1422
1423 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1424 stats_dir = open(buf, O_DIRECTORY);
1425 if (stats_dir < 0)
1426 return -1;
1427
1428 for (i = 0; i < ARRAY_SIZE(counters); i++)
1429 if (read_uint64_file(stats_dir, counters[i], &val))
1430 blobmsg_add_u64(b, counters[i], val);
1431
1432 close(stats_dir);
1433 return 0;
1434 }
1435
1436 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1437 {
1438 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1439 int alen = v4 ? 4 : 16;
1440 unsigned int flags = 0;
1441 struct ifaddrmsg ifa = {
1442 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1443 .ifa_prefixlen = addr->mask,
1444 .ifa_index = dev->ifindex,
1445 };
1446
1447 struct nl_msg *msg;
1448 if (cmd == RTM_NEWADDR)
1449 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1450
1451 msg = nlmsg_alloc_simple(cmd, flags);
1452 if (!msg)
1453 return -1;
1454
1455 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1456 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1457 if (v4) {
1458 if (addr->broadcast)
1459 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1460 if (addr->point_to_point)
1461 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1462 } else {
1463 time_t now = system_get_rtime();
1464 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1465
1466 if (addr->preferred_until) {
1467 int64_t preferred = addr->preferred_until - now;
1468 if (preferred < 0)
1469 preferred = 0;
1470 else if (preferred > UINT32_MAX)
1471 preferred = UINT32_MAX;
1472
1473 cinfo.ifa_prefered = preferred;
1474 }
1475
1476 if (addr->valid_until) {
1477 int64_t valid = addr->valid_until - now;
1478 if (valid <= 0)
1479 return -1;
1480 else if (valid > UINT32_MAX)
1481 valid = UINT32_MAX;
1482
1483 cinfo.ifa_valid = valid;
1484 }
1485
1486 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1487
1488 if (cmd == RTM_NEWADDR && (addr->flags & DEVADDR_OFFLINK))
1489 nla_put_u32(msg, IFA_FLAGS, IFA_F_NOPREFIXROUTE);
1490 }
1491
1492 return system_rtnl_call(msg);
1493 }
1494
1495 int system_add_address(struct device *dev, struct device_addr *addr)
1496 {
1497 return system_addr(dev, addr, RTM_NEWADDR);
1498 }
1499
1500 int system_del_address(struct device *dev, struct device_addr *addr)
1501 {
1502 return system_addr(dev, addr, RTM_DELADDR);
1503 }
1504
1505 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1506 {
1507 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1508 bool have_gw;
1509 unsigned int flags = 0;
1510
1511 if (alen == 4)
1512 have_gw = !!route->nexthop.in.s_addr;
1513 else
1514 have_gw = route->nexthop.in6.s6_addr32[0] ||
1515 route->nexthop.in6.s6_addr32[1] ||
1516 route->nexthop.in6.s6_addr32[2] ||
1517 route->nexthop.in6.s6_addr32[3];
1518
1519 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1520 ? route->table : RT_TABLE_MAIN;
1521
1522 struct rtmsg rtm = {
1523 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1524 .rtm_dst_len = route->mask,
1525 .rtm_src_len = route->sourcemask,
1526 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1527 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1528 .rtm_scope = RT_SCOPE_NOWHERE,
1529 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1530 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1531 };
1532 struct nl_msg *msg;
1533
1534 if (cmd == RTM_NEWROUTE) {
1535 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1536
1537 if (!dev) { // Add null-route
1538 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1539 rtm.rtm_type = RTN_UNREACHABLE;
1540 }
1541 else
1542 rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1543 }
1544
1545 if (route->flags & DEVROUTE_TYPE) {
1546 rtm.rtm_type = route->type;
1547 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1548 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1549 rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1550 rtm.rtm_table = RT_TABLE_LOCAL;
1551 }
1552
1553 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1554 rtm.rtm_scope = RT_SCOPE_HOST;
1555 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1556 rtm.rtm_type == RTN_ANYCAST) {
1557 rtm.rtm_scope = RT_SCOPE_LINK;
1558 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1559 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1560 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1561 dev = NULL;
1562 }
1563 }
1564
1565 msg = nlmsg_alloc_simple(cmd, flags);
1566 if (!msg)
1567 return -1;
1568
1569 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1570
1571 if (route->mask)
1572 nla_put(msg, RTA_DST, alen, &route->addr);
1573
1574 if (route->sourcemask) {
1575 if (rtm.rtm_family == AF_INET)
1576 nla_put(msg, RTA_PREFSRC, alen, &route->source);
1577 else
1578 nla_put(msg, RTA_SRC, alen, &route->source);
1579 }
1580
1581 if (route->metric > 0)
1582 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1583
1584 if (have_gw)
1585 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1586
1587 if (dev)
1588 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1589
1590 if (table >= 256)
1591 nla_put_u32(msg, RTA_TABLE, table);
1592
1593 if (route->flags & DEVROUTE_MTU) {
1594 struct nlattr *metrics;
1595
1596 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1597 goto nla_put_failure;
1598
1599 nla_put_u32(msg, RTAX_MTU, route->mtu);
1600
1601 nla_nest_end(msg, metrics);
1602 }
1603
1604 return system_rtnl_call(msg);
1605
1606 nla_put_failure:
1607 nlmsg_free(msg);
1608 return -ENOMEM;
1609 }
1610
1611 int system_add_route(struct device *dev, struct device_route *route)
1612 {
1613 return system_rt(dev, route, RTM_NEWROUTE);
1614 }
1615
1616 int system_del_route(struct device *dev, struct device_route *route)
1617 {
1618 return system_rt(dev, route, RTM_DELROUTE);
1619 }
1620
1621 int system_flush_routes(void)
1622 {
1623 const char *names[] = {
1624 "/proc/sys/net/ipv4/route/flush",
1625 "/proc/sys/net/ipv6/route/flush"
1626 };
1627 int fd, i;
1628
1629 for (i = 0; i < ARRAY_SIZE(names); i++) {
1630 fd = open(names[i], O_WRONLY);
1631 if (fd < 0)
1632 continue;
1633
1634 if (write(fd, "-1", 2)) {}
1635 close(fd);
1636 }
1637 return 0;
1638 }
1639
1640 bool system_resolve_rt_type(const char *type, unsigned int *id)
1641 {
1642 return system_rtn_aton(type, id);
1643 }
1644
1645 bool system_resolve_rt_table(const char *name, unsigned int *id)
1646 {
1647 FILE *f;
1648 char *e, buf[128];
1649 unsigned int n, table = RT_TABLE_UNSPEC;
1650
1651 /* first try to parse table as number */
1652 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1653 table = n;
1654
1655 /* handle well known aliases */
1656 else if (!strcmp(name, "default"))
1657 table = RT_TABLE_DEFAULT;
1658 else if (!strcmp(name, "main"))
1659 table = RT_TABLE_MAIN;
1660 else if (!strcmp(name, "local"))
1661 table = RT_TABLE_LOCAL;
1662 else if (!strcmp(name, "prelocal"))
1663 table = RT_TABLE_PRELOCAL;
1664
1665 /* try to look up name in /etc/iproute2/rt_tables */
1666 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1667 {
1668 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1669 {
1670 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1671 continue;
1672
1673 n = strtoul(e, NULL, 10);
1674 e = strtok(NULL, " \t\n");
1675
1676 if (e && !strcmp(e, name))
1677 {
1678 table = n;
1679 break;
1680 }
1681 }
1682
1683 fclose(f);
1684 }
1685
1686 if (table == RT_TABLE_UNSPEC)
1687 return false;
1688
1689 *id = table;
1690 return true;
1691 }
1692
1693 bool system_is_default_rt_table(unsigned int id)
1694 {
1695 return (id == RT_TABLE_MAIN);
1696 }
1697
1698 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1699 {
1700 char *e;
1701 unsigned int n;
1702
1703 if (!strcmp(filter, "strict"))
1704 n = 1;
1705 else if (!strcmp(filter, "loose"))
1706 n = 2;
1707 else {
1708 n = strtoul(filter, &e, 0);
1709 if (*e || e == filter || n > 2)
1710 return false;
1711 }
1712
1713 *id = n;
1714 return true;
1715 }
1716
1717 static int system_iprule(struct iprule *rule, int cmd)
1718 {
1719 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1720
1721 struct nl_msg *msg;
1722 struct rtmsg rtm = {
1723 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1724 .rtm_protocol = RTPROT_STATIC,
1725 .rtm_scope = RT_SCOPE_UNIVERSE,
1726 .rtm_table = RT_TABLE_UNSPEC,
1727 .rtm_type = RTN_UNSPEC,
1728 .rtm_flags = 0,
1729 };
1730
1731 if (cmd == RTM_NEWRULE) {
1732 rtm.rtm_type = RTN_UNICAST;
1733 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1734 }
1735
1736 if (rule->invert)
1737 rtm.rtm_flags |= FIB_RULE_INVERT;
1738
1739 if (rule->flags & IPRULE_SRC)
1740 rtm.rtm_src_len = rule->src_mask;
1741
1742 if (rule->flags & IPRULE_DEST)
1743 rtm.rtm_dst_len = rule->dest_mask;
1744
1745 if (rule->flags & IPRULE_TOS)
1746 rtm.rtm_tos = rule->tos;
1747
1748 if (rule->flags & IPRULE_LOOKUP) {
1749 if (rule->lookup < 256)
1750 rtm.rtm_table = rule->lookup;
1751 }
1752
1753 if (rule->flags & IPRULE_ACTION)
1754 rtm.rtm_type = rule->action;
1755 else if (rule->flags & IPRULE_GOTO)
1756 rtm.rtm_type = FR_ACT_GOTO;
1757 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1758 rtm.rtm_type = FR_ACT_NOP;
1759
1760 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1761
1762 if (!msg)
1763 return -1;
1764
1765 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1766
1767 if (rule->flags & IPRULE_IN)
1768 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1769
1770 if (rule->flags & IPRULE_OUT)
1771 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1772
1773 if (rule->flags & IPRULE_SRC)
1774 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1775
1776 if (rule->flags & IPRULE_DEST)
1777 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1778
1779 if (rule->flags & IPRULE_PRIORITY)
1780 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1781 else if (cmd == RTM_NEWRULE)
1782 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1783
1784 if (rule->flags & IPRULE_FWMARK)
1785 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1786
1787 if (rule->flags & IPRULE_FWMASK)
1788 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1789
1790 if (rule->flags & IPRULE_LOOKUP) {
1791 if (rule->lookup >= 256)
1792 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1793 }
1794
1795 if (rule->flags & IPRULE_GOTO)
1796 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1797
1798 return system_rtnl_call(msg);
1799 }
1800
1801 int system_add_iprule(struct iprule *rule)
1802 {
1803 return system_iprule(rule, RTM_NEWRULE);
1804 }
1805
1806 int system_del_iprule(struct iprule *rule)
1807 {
1808 return system_iprule(rule, RTM_DELRULE);
1809 }
1810
1811 int system_flush_iprules(void)
1812 {
1813 int rv = 0;
1814 struct iprule rule;
1815
1816 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1817 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1818
1819 memset(&rule, 0, sizeof(rule));
1820
1821
1822 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1823
1824 rule.priority = 0;
1825 rule.lookup = RT_TABLE_PRELOCAL;
1826 rv |= system_iprule(&rule, RTM_NEWRULE);
1827
1828 rule.priority = 1;
1829 rule.lookup = RT_TABLE_LOCAL;
1830 rv |= system_iprule(&rule, RTM_NEWRULE);
1831
1832 rule.priority = 32766;
1833 rule.lookup = RT_TABLE_MAIN;
1834 rv |= system_iprule(&rule, RTM_NEWRULE);
1835
1836 rule.priority = 32767;
1837 rule.lookup = RT_TABLE_DEFAULT;
1838 rv |= system_iprule(&rule, RTM_NEWRULE);
1839
1840
1841 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1842
1843 rule.priority = 0;
1844 rule.lookup = RT_TABLE_PRELOCAL;
1845 rv |= system_iprule(&rule, RTM_NEWRULE);
1846
1847 rule.priority = 1;
1848 rule.lookup = RT_TABLE_LOCAL;
1849 rv |= system_iprule(&rule, RTM_NEWRULE);
1850
1851 rule.priority = 32766;
1852 rule.lookup = RT_TABLE_MAIN;
1853 rv |= system_iprule(&rule, RTM_NEWRULE);
1854
1855 return rv;
1856 }
1857
1858 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1859 {
1860 return system_rtn_aton(action, id);
1861 }
1862
1863 time_t system_get_rtime(void)
1864 {
1865 struct timespec ts;
1866 struct timeval tv;
1867
1868 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1869 return ts.tv_sec;
1870
1871 if (gettimeofday(&tv, NULL) == 0)
1872 return tv.tv_sec;
1873
1874 return 0;
1875 }
1876
1877 #ifndef IP_DF
1878 #define IP_DF 0x4000
1879 #endif
1880
1881 static int tunnel_ioctl(const char *name, int cmd, void *p)
1882 {
1883 struct ifreq ifr;
1884
1885 memset(&ifr, 0, sizeof(ifr));
1886 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1887 ifr.ifr_ifru.ifru_data = p;
1888 return ioctl(sock_ioctl, cmd, &ifr);
1889 }
1890
1891 #ifdef IFLA_IPTUN_MAX
1892 #define IP6_FLOWINFO_TCLASS htonl(0x0FF00000)
1893 static int system_add_gre_tunnel(const char *name, const char *kind,
1894 const unsigned int link, struct blob_attr **tb, bool v6)
1895 {
1896 struct nl_msg *nlm;
1897 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1898 struct blob_attr *cur;
1899 uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1900 uint16_t iflags = 0, oflags = 0;
1901 uint8_t tos = 0;
1902 int ret = 0, ttl = 64;
1903
1904 nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1905 if (!nlm)
1906 return -1;
1907
1908 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1909 nla_put_string(nlm, IFLA_IFNAME, name);
1910
1911 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1912 if (!linkinfo) {
1913 ret = -ENOMEM;
1914 goto failure;
1915 }
1916
1917 nla_put_string(nlm, IFLA_INFO_KIND, kind);
1918 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1919 if (!infodata) {
1920 ret = -ENOMEM;
1921 goto failure;
1922 }
1923
1924 if (link)
1925 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1926
1927 if ((cur = tb[TUNNEL_ATTR_TTL]))
1928 ttl = blobmsg_get_u32(cur);
1929
1930 nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1931
1932 if ((cur = tb[TUNNEL_ATTR_TOS])) {
1933 char *str = blobmsg_get_string(cur);
1934 if (strcmp(str, "inherit")) {
1935 unsigned uval;
1936
1937 if (!system_tos_aton(str, &uval)) {
1938 ret = -EINVAL;
1939 goto failure;
1940 }
1941
1942 if (v6)
1943 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1944 else
1945 tos = uval;
1946 } else {
1947 if (v6)
1948 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1949 else
1950 tos = 1;
1951 }
1952 }
1953
1954 if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1955 uint8_t icsum, ocsum, iseqno, oseqno;
1956 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1957 &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1958 ret = -EINVAL;
1959 goto failure;
1960 }
1961
1962 if (ikey)
1963 iflags |= GRE_KEY;
1964
1965 if (okey)
1966 oflags |= GRE_KEY;
1967
1968 if (icsum)
1969 iflags |= GRE_CSUM;
1970
1971 if (ocsum)
1972 oflags |= GRE_CSUM;
1973
1974 if (iseqno)
1975 iflags |= GRE_SEQ;
1976
1977 if (oseqno)
1978 oflags |= GRE_SEQ;
1979 }
1980
1981 if (v6) {
1982 struct in6_addr in6buf;
1983 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1984 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1985 ret = -EINVAL;
1986 goto failure;
1987 }
1988 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1989 }
1990
1991 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1992 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1993 ret = -EINVAL;
1994 goto failure;
1995 }
1996 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1997 }
1998 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1999
2000 if (flowinfo)
2001 nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
2002
2003 if (flags)
2004 nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
2005 } else {
2006 struct in_addr inbuf;
2007 bool set_df = true;
2008
2009 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2010 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2011 ret = -EINVAL;
2012 goto failure;
2013 }
2014 nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
2015 }
2016
2017 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2018 if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
2019 ret = -EINVAL;
2020 goto failure;
2021 }
2022 nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
2023
2024 if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
2025 if (!okey) {
2026 okey = inbuf.s_addr;
2027 oflags |= GRE_KEY;
2028 }
2029
2030 if (!ikey) {
2031 ikey = inbuf.s_addr;
2032 iflags |= GRE_KEY;
2033 }
2034 }
2035 }
2036
2037 if ((cur = tb[TUNNEL_ATTR_DF]))
2038 set_df = blobmsg_get_bool(cur);
2039
2040 /* ttl !=0 and nopmtudisc are incompatible */
2041 if (ttl && !set_df) {
2042 ret = -EINVAL;
2043 goto failure;
2044 }
2045
2046 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
2047
2048 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
2049 }
2050
2051 if (oflags)
2052 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
2053
2054 if (iflags)
2055 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
2056
2057 if (okey)
2058 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
2059
2060 if (ikey)
2061 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
2062
2063 nla_nest_end(nlm, infodata);
2064 nla_nest_end(nlm, linkinfo);
2065
2066 return system_rtnl_call(nlm);
2067
2068 failure:
2069 nlmsg_free(nlm);
2070 return ret;
2071 }
2072 #endif
2073
2074 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
2075 {
2076 struct blob_attr *cur;
2077 bool set_df = true;
2078 struct ip_tunnel_parm p = {
2079 .link = link,
2080 .iph = {
2081 .version = 4,
2082 .ihl = 5,
2083 .protocol = proto,
2084 }
2085 };
2086
2087 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
2088 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
2089 return -EINVAL;
2090
2091 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
2092 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
2093 return -EINVAL;
2094
2095 if ((cur = tb[TUNNEL_ATTR_DF]))
2096 set_df = blobmsg_get_bool(cur);
2097
2098 if ((cur = tb[TUNNEL_ATTR_TTL]))
2099 p.iph.ttl = blobmsg_get_u32(cur);
2100
2101 if ((cur = tb[TUNNEL_ATTR_TOS])) {
2102 char *str = blobmsg_get_string(cur);
2103 if (strcmp(str, "inherit")) {
2104 unsigned uval;
2105
2106 if (!system_tos_aton(str, &uval))
2107 return -EINVAL;
2108
2109 p.iph.tos = uval;
2110 } else
2111 p.iph.tos = 1;
2112 }
2113
2114 p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2115 /* ttl !=0 and nopmtudisc are incompatible */
2116 if (p.iph.ttl && p.iph.frag_off == 0)
2117 return -EINVAL;
2118
2119 strncpy(p.name, name, sizeof(p.name));
2120
2121 switch (p.iph.protocol) {
2122 case IPPROTO_IPIP:
2123 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2124 case IPPROTO_IPV6:
2125 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2126 default:
2127 break;
2128 }
2129 return -1;
2130 }
2131
2132 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2133 {
2134 struct blob_attr *cur;
2135 const char *str;
2136
2137 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2138 return -EINVAL;
2139 str = blobmsg_data(cur);
2140
2141 if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2142 !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2143 return system_link_del(name);
2144 else
2145 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2146 }
2147
2148 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2149 {
2150 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2151
2152 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2153 blob_data(attr), blob_len(attr));
2154
2155 return __system_del_ip_tunnel(name, tb);
2156 }
2157
2158 int system_update_ipv6_mtu(struct device *dev, int mtu)
2159 {
2160 int ret = -1;
2161 char buf[64];
2162 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2163 dev->ifname);
2164
2165 int fd = open(buf, O_RDWR);
2166 ssize_t len = read(fd, buf, sizeof(buf) - 1);
2167 if (len < 0)
2168 goto out;
2169
2170 buf[len] = 0;
2171 ret = atoi(buf);
2172
2173 if (!mtu || ret <= mtu)
2174 goto out;
2175
2176 lseek(fd, 0, SEEK_SET);
2177 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2178 ret = -1;
2179
2180 out:
2181 close(fd);
2182 return ret;
2183 }
2184
2185 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2186 {
2187 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2188 struct blob_attr *cur;
2189 const char *str;
2190
2191 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2192 blob_data(attr), blob_len(attr));
2193
2194 __system_del_ip_tunnel(name, tb);
2195
2196 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2197 return -EINVAL;
2198 str = blobmsg_data(cur);
2199
2200 unsigned int ttl = 0;
2201 if ((cur = tb[TUNNEL_ATTR_TTL])) {
2202 ttl = blobmsg_get_u32(cur);
2203 if (ttl > 255)
2204 return -EINVAL;
2205 }
2206
2207 unsigned int link = 0;
2208 if ((cur = tb[TUNNEL_ATTR_LINK])) {
2209 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2210 if (!iface)
2211 return -EINVAL;
2212
2213 if (iface->l3_dev.dev)
2214 link = iface->l3_dev.dev->ifindex;
2215 }
2216
2217 if (!strcmp(str, "sit")) {
2218 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2219 return -1;
2220
2221 #ifdef SIOCADD6RD
2222 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2223 unsigned int mask;
2224 struct ip_tunnel_6rd p6;
2225
2226 memset(&p6, 0, sizeof(p6));
2227
2228 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2229 &p6.prefix, &mask) || mask > 128)
2230 return -EINVAL;
2231 p6.prefixlen = mask;
2232
2233 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2234 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2235 &p6.relay_prefix, &mask) || mask > 32)
2236 return -EINVAL;
2237 p6.relay_prefixlen = mask;
2238 }
2239
2240 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2241 __system_del_ip_tunnel(name, tb);
2242 return -1;
2243 }
2244 }
2245 #endif
2246 #ifdef IFLA_IPTUN_MAX
2247 } else if (!strcmp(str, "ipip6")) {
2248 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2249 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2250 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2251 int ret = 0;
2252
2253 if (!nlm)
2254 return -1;
2255
2256 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2257 nla_put_string(nlm, IFLA_IFNAME, name);
2258
2259 if (link)
2260 nla_put_u32(nlm, IFLA_LINK, link);
2261
2262 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2263 if (!linkinfo) {
2264 ret = -ENOMEM;
2265 goto failure;
2266 }
2267 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2268 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2269 if (!infodata) {
2270 ret = -ENOMEM;
2271 goto failure;
2272 }
2273
2274 if (link)
2275 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2276
2277 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2278 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2279 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2280
2281 struct in6_addr in6buf;
2282 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2283 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2284 ret = -EINVAL;
2285 goto failure;
2286 }
2287 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2288 }
2289
2290 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2291 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2292 ret = -EINVAL;
2293 goto failure;
2294 }
2295 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2296 }
2297
2298 #ifdef IFLA_IPTUN_FMR_MAX
2299 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2300 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2301
2302 struct blob_attr *fmr;
2303 unsigned rem, fmrcnt = 0;
2304 blobmsg_for_each_attr(fmr, cur, rem) {
2305 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2306 continue;
2307
2308 unsigned ip4len, ip6len, ealen, offset = 6;
2309 char ip6buf[48];
2310 char ip4buf[16];
2311
2312 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2313 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2314 ret = -EINVAL;
2315 goto failure;
2316 }
2317
2318 struct in6_addr ip6prefix;
2319 struct in_addr ip4prefix;
2320 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2321 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2322 ret = -EINVAL;
2323 goto failure;
2324 }
2325
2326 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2327
2328 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2329 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2330 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2331 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2332 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2333 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2334
2335 nla_nest_end(nlm, rule);
2336 }
2337
2338 nla_nest_end(nlm, fmrs);
2339 }
2340 #endif
2341
2342 nla_nest_end(nlm, infodata);
2343 nla_nest_end(nlm, linkinfo);
2344
2345 return system_rtnl_call(nlm);
2346 failure:
2347 nlmsg_free(nlm);
2348 return ret;
2349 } else if (!strcmp(str, "greip")) {
2350 return system_add_gre_tunnel(name, "gre", link, tb, false);
2351 } else if (!strcmp(str, "gretapip")) {
2352 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2353 } else if (!strcmp(str, "greip6")) {
2354 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2355 } else if (!strcmp(str, "gretapip6")) {
2356 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2357 #endif
2358 } else if (!strcmp(str, "ipip")) {
2359 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2360 }
2361 else
2362 return -EINVAL;
2363
2364 return 0;
2365 }