4d0f4c9376a4a8eef51685a7706fb120f7c9cbcc
[project/netifd.git] / system-linux.c
1 /*
2 * netifd - network interface daemon
3 * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
4 * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
5 * Copyright (C) 2013 Steven Barth <steven@midlink.org>
6 * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2
10 * as published by the Free Software Foundation
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 */
17 #define _GNU_SOURCE
18
19 #include <sys/socket.h>
20 #include <sys/ioctl.h>
21 #include <sys/stat.h>
22 #include <sys/syscall.h>
23
24 #include <net/if.h>
25 #include <net/if_arp.h>
26
27 #include <arpa/inet.h>
28 #include <netinet/in.h>
29
30 #include <linux/rtnetlink.h>
31 #include <linux/sockios.h>
32 #include <linux/ip.h>
33 #include <linux/if_link.h>
34 #include <linux/if_vlan.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38 #include <linux/ethtool.h>
39 #include <linux/fib_rules.h>
40 #include <linux/version.h>
41
42 #ifndef RTN_FAILED_POLICY
43 #define RTN_FAILED_POLICY 12
44 #endif
45
46 #include <string.h>
47 #include <fcntl.h>
48 #include <glob.h>
49 #include <time.h>
50
51 #include <netlink/msg.h>
52 #include <netlink/attr.h>
53 #include <netlink/socket.h>
54 #include <libubox/uloop.h>
55
56 #include "netifd.h"
57 #include "device.h"
58 #include "system.h"
59
60 struct event_socket {
61 struct uloop_fd uloop;
62 struct nl_sock *sock;
63 int bufsize;
64 };
65
66 static int sock_ioctl = -1;
67 static struct nl_sock *sock_rtnl = NULL;
68
69 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
70 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
71
72 static char dev_buf[256];
73
74 static void
75 handler_nl_event(struct uloop_fd *u, unsigned int events)
76 {
77 struct event_socket *ev = container_of(u, struct event_socket, uloop);
78 int err;
79 socklen_t errlen = sizeof(err);
80
81 if (!u->error) {
82 nl_recvmsgs_default(ev->sock);
83 return;
84 }
85
86 if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
87 goto abort;
88
89 switch(err) {
90 case ENOBUFS:
91 // Increase rx buffer size on netlink socket
92 ev->bufsize *= 2;
93 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
94 goto abort;
95
96 // Request full dump since some info got dropped
97 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
98 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
99 break;
100
101 default:
102 goto abort;
103 }
104 u->error = false;
105 return;
106
107 abort:
108 uloop_fd_delete(&ev->uloop);
109 return;
110 }
111
112 static struct nl_sock *
113 create_socket(int protocol, int groups)
114 {
115 struct nl_sock *sock;
116
117 sock = nl_socket_alloc();
118 if (!sock)
119 return NULL;
120
121 if (groups)
122 nl_join_groups(sock, groups);
123
124 if (nl_connect(sock, protocol))
125 return NULL;
126
127 return sock;
128 }
129
130 static bool
131 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
132 uloop_fd_handler cb, int flags)
133 {
134 ev->sock = create_socket(protocol, groups);
135 if (!ev->sock)
136 return false;
137
138 ev->uloop.fd = nl_socket_get_fd(ev->sock);
139 ev->uloop.cb = cb;
140 if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
141 return false;
142
143 return true;
144 }
145
146 static bool
147 create_event_socket(struct event_socket *ev, int protocol,
148 int (*cb)(struct nl_msg *msg, void *arg))
149 {
150 if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
151 return false;
152
153 // Install the valid custom callback handler
154 nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
155
156 // Disable sequence number checking on event sockets
157 nl_socket_disable_seq_check(ev->sock);
158
159 // Increase rx buffer size to 65K on event sockets
160 ev->bufsize = 65535;
161 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
162 return false;
163
164 return true;
165 }
166
167 int system_init(void)
168 {
169 static struct event_socket rtnl_event;
170 static struct event_socket hotplug_event;
171
172 sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
173 system_fd_set_cloexec(sock_ioctl);
174
175 // Prepare socket for routing / address control
176 sock_rtnl = create_socket(NETLINK_ROUTE, 0);
177 if (!sock_rtnl)
178 return -1;
179
180 if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
181 return -1;
182
183 if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
184 handle_hotplug_event, 0))
185 return -1;
186
187 // Receive network link events form kernel
188 nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
189
190 return 0;
191 }
192
193 static void system_set_sysctl(const char *path, const char *val)
194 {
195 int fd;
196
197 fd = open(path, O_WRONLY);
198 if (fd < 0)
199 return;
200
201 if (write(fd, val, strlen(val))) {}
202 close(fd);
203 }
204
205 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
206 {
207 snprintf(dev_buf, sizeof(dev_buf), path, device);
208 system_set_sysctl(dev_buf, val);
209 }
210
211 static void system_set_disable_ipv6(struct device *dev, const char *val)
212 {
213 system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
214 }
215
216 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
217 {
218 int fd = -1, ret = -1;
219
220 fd = open(path, O_RDONLY);
221 if (fd < 0)
222 goto out;
223
224 ssize_t len = read(fd, buf, buf_sz - 1);
225 if (len < 0)
226 goto out;
227
228 ret = buf[len] = 0;
229
230 out:
231 if (fd >= 0)
232 close(fd);
233
234 return ret;
235 }
236
237 static int
238 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
239 {
240 snprintf(dev_buf, sizeof(dev_buf), path, device);
241 return system_get_sysctl(dev_buf, buf, buf_sz);
242 }
243
244 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
245 {
246 return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
247 dev->ifname, buf, buf_sz);
248 }
249
250 #ifndef IFF_LOWER_UP
251 #define IFF_LOWER_UP 0x10000
252 #endif
253
254 // Evaluate netlink messages
255 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
256 {
257 struct nlmsghdr *nh = nlmsg_hdr(msg);
258 struct ifinfomsg *ifi = NLMSG_DATA(nh);
259 struct nlattr *nla[__IFLA_MAX];
260
261 if (nh->nlmsg_type != RTM_NEWLINK)
262 goto out;
263
264 nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
265 if (!nla[IFLA_IFNAME])
266 goto out;
267
268 struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
269 if (!dev)
270 goto out;
271
272 device_set_ifindex(dev, ifi->ifi_index);
273 device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
274
275 out:
276 return 0;
277 }
278
279 static void
280 handle_hotplug_msg(char *data, int size)
281 {
282 const char *subsystem = NULL, *interface = NULL;
283 char *cur, *end, *sep;
284 struct device *dev;
285 int skip;
286 bool add;
287
288 if (!strncmp(data, "add@", 4))
289 add = true;
290 else if (!strncmp(data, "remove@", 7))
291 add = false;
292 else
293 return;
294
295 skip = strlen(data) + 1;
296 end = data + size;
297
298 for (cur = data + skip; cur < end; cur += skip) {
299 skip = strlen(cur) + 1;
300
301 sep = strchr(cur, '=');
302 if (!sep)
303 continue;
304
305 *sep = 0;
306 if (!strcmp(cur, "INTERFACE"))
307 interface = sep + 1;
308 else if (!strcmp(cur, "SUBSYSTEM")) {
309 subsystem = sep + 1;
310 if (strcmp(subsystem, "net") != 0)
311 return;
312 }
313 if (subsystem && interface)
314 goto found;
315 }
316 return;
317
318 found:
319 dev = device_get(interface, false);
320 if (!dev)
321 return;
322
323 if (dev->type != &simple_device_type)
324 return;
325
326 if (add && system_if_force_external(dev->ifname))
327 return;
328
329 device_set_present(dev, add);
330 }
331
332 static void
333 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
334 {
335 struct event_socket *ev = container_of(u, struct event_socket, uloop);
336 struct sockaddr_nl nla;
337 unsigned char *buf = NULL;
338 int size;
339
340 while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
341 if (nla.nl_pid == 0)
342 handle_hotplug_msg((char *) buf, size);
343
344 free(buf);
345 }
346 }
347
348 static int system_rtnl_call(struct nl_msg *msg)
349 {
350 int ret;
351
352 ret = nl_send_auto_complete(sock_rtnl, msg);
353 nlmsg_free(msg);
354
355 if (ret < 0)
356 return ret;
357
358 return nl_wait_for_ack(sock_rtnl);
359 }
360
361 int system_bridge_delbr(struct device *bridge)
362 {
363 return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
364 }
365
366 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
367 {
368 struct ifreq ifr;
369
370 memset(&ifr, 0, sizeof(ifr));
371 if (dev)
372 ifr.ifr_ifindex = dev->ifindex;
373 else
374 ifr.ifr_data = data;
375 strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
376 return ioctl(sock_ioctl, cmd, &ifr);
377 }
378
379 static bool system_is_bridge(const char *name, char *buf, int buflen)
380 {
381 struct stat st;
382
383 snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
384 if (stat(buf, &st) < 0)
385 return false;
386
387 return true;
388 }
389
390 static char *system_get_bridge(const char *name, char *buf, int buflen)
391 {
392 char *path;
393 ssize_t len;
394 glob_t gl;
395
396 snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
397 if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
398 return NULL;
399
400 if (gl.gl_pathc == 0)
401 return NULL;
402
403 len = readlink(gl.gl_pathv[0], buf, buflen);
404 if (len < 0)
405 return NULL;
406
407 buf[len] = 0;
408 path = strrchr(buf, '/');
409 if (!path)
410 return NULL;
411
412 return path + 1;
413 }
414
415 int system_bridge_addif(struct device *bridge, struct device *dev)
416 {
417 char *oldbr;
418
419 oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
420 if (oldbr && !strcmp(oldbr, bridge->ifname))
421 return 0;
422
423 return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
424 }
425
426 int system_bridge_delif(struct device *bridge, struct device *dev)
427 {
428 return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
429 }
430
431 static int system_if_resolve(struct device *dev)
432 {
433 struct ifreq ifr;
434 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
435 if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
436 return ifr.ifr_ifindex;
437 else
438 return 0;
439 }
440
441 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
442 {
443 struct ifreq ifr;
444
445 memset(&ifr, 0, sizeof(ifr));
446 strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
447 ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
448 ifr.ifr_flags |= add;
449 ifr.ifr_flags &= ~rem;
450 return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
451 }
452
453 struct clear_data {
454 struct nl_msg *msg;
455 struct device *dev;
456 int type;
457 int size;
458 int af;
459 };
460
461
462 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
463 {
464 struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
465
466 return ifa->ifa_index == ifindex;
467 }
468
469 static bool check_route(struct nlmsghdr *hdr, int ifindex)
470 {
471 struct rtmsg *r = NLMSG_DATA(hdr);
472 struct nlattr *tb[__RTA_MAX];
473
474 if (r->rtm_protocol == RTPROT_KERNEL &&
475 r->rtm_family == AF_INET6)
476 return false;
477
478 nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
479 if (!tb[RTA_OIF])
480 return false;
481
482 return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
483 }
484
485 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
486 {
487 return true;
488 }
489
490 static int cb_clear_event(struct nl_msg *msg, void *arg)
491 {
492 struct clear_data *clr = arg;
493 struct nlmsghdr *hdr = nlmsg_hdr(msg);
494 bool (*cb)(struct nlmsghdr *, int ifindex);
495 int type;
496
497 switch(clr->type) {
498 case RTM_GETADDR:
499 type = RTM_DELADDR;
500 if (hdr->nlmsg_type != RTM_NEWADDR)
501 return NL_SKIP;
502
503 cb = check_ifaddr;
504 break;
505 case RTM_GETROUTE:
506 type = RTM_DELROUTE;
507 if (hdr->nlmsg_type != RTM_NEWROUTE)
508 return NL_SKIP;
509
510 cb = check_route;
511 break;
512 case RTM_GETRULE:
513 type = RTM_DELRULE;
514 if (hdr->nlmsg_type != RTM_NEWRULE)
515 return NL_SKIP;
516
517 cb = check_rule;
518 break;
519 default:
520 return NL_SKIP;
521 }
522
523 if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
524 return NL_SKIP;
525
526 if (type == RTM_DELRULE)
527 D(SYSTEM, "Remove a rule\n");
528 else
529 D(SYSTEM, "Remove %s from device %s\n",
530 type == RTM_DELADDR ? "an address" : "a route",
531 clr->dev->ifname);
532 memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
533 hdr = nlmsg_hdr(clr->msg);
534 hdr->nlmsg_type = type;
535 hdr->nlmsg_flags = NLM_F_REQUEST;
536
537 nl_socket_disable_auto_ack(sock_rtnl);
538 nl_send_auto_complete(sock_rtnl, clr->msg);
539 nl_socket_enable_auto_ack(sock_rtnl);
540
541 return NL_SKIP;
542 }
543
544 static int
545 cb_finish_event(struct nl_msg *msg, void *arg)
546 {
547 int *pending = arg;
548 *pending = 0;
549 return NL_STOP;
550 }
551
552 static int
553 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
554 {
555 int *pending = arg;
556 *pending = err->error;
557 return NL_STOP;
558 }
559
560 static void
561 system_if_clear_entries(struct device *dev, int type, int af)
562 {
563 struct clear_data clr;
564 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
565 struct rtmsg rtm = {
566 .rtm_family = af,
567 .rtm_flags = RTM_F_CLONED,
568 };
569 int flags = NLM_F_DUMP;
570 int pending = 1;
571
572 clr.af = af;
573 clr.dev = dev;
574 clr.type = type;
575 switch (type) {
576 case RTM_GETADDR:
577 case RTM_GETRULE:
578 clr.size = sizeof(struct rtgenmsg);
579 break;
580 case RTM_GETROUTE:
581 clr.size = sizeof(struct rtmsg);
582 break;
583 default:
584 return;
585 }
586
587 if (!cb)
588 return;
589
590 clr.msg = nlmsg_alloc_simple(type, flags);
591 if (!clr.msg)
592 goto out;
593
594 nlmsg_append(clr.msg, &rtm, clr.size, 0);
595 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
596 nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
597 nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
598
599 nl_send_auto_complete(sock_rtnl, clr.msg);
600 while (pending > 0)
601 nl_recvmsgs(sock_rtnl, cb);
602
603 nlmsg_free(clr.msg);
604 out:
605 nl_cb_put(cb);
606 }
607
608 /*
609 * Clear bridge (membership) state and bring down device
610 */
611 void system_if_clear_state(struct device *dev)
612 {
613 static char buf[256];
614 char *bridge;
615
616 device_set_ifindex(dev, system_if_resolve(dev));
617 if (dev->external || !dev->ifindex)
618 return;
619
620 system_if_flags(dev->ifname, 0, IFF_UP);
621
622 if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
623 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
624 system_bridge_delbr(dev);
625 return;
626 }
627
628 bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
629 if (bridge) {
630 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
631 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
632 }
633
634 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
635 system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
636 system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
637 system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
638 system_set_disable_ipv6(dev, "0");
639 }
640
641 static inline unsigned long
642 sec_to_jiffies(int val)
643 {
644 return (unsigned long) val * 100;
645 }
646
647 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
648 {
649 unsigned long args[4] = {};
650
651 if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
652 return -1;
653
654 args[0] = BRCTL_SET_BRIDGE_STP_STATE;
655 args[1] = !!cfg->stp;
656 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
657
658 args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
659 args[1] = sec_to_jiffies(cfg->forward_delay);
660 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
661
662 system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
663 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
664
665 args[0] = BRCTL_SET_BRIDGE_PRIORITY;
666 args[1] = cfg->priority;
667 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
668
669 if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
670 args[0] = BRCTL_SET_AGEING_TIME;
671 args[1] = sec_to_jiffies(cfg->ageing_time);
672 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
673 }
674
675 if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
676 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
677 args[1] = sec_to_jiffies(cfg->hello_time);
678 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
679 }
680
681 if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
682 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
683 args[1] = sec_to_jiffies(cfg->max_age);
684 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
685 }
686
687 return 0;
688 }
689
690 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
691 {
692 struct nl_msg *msg;
693 struct nlattr *linkinfo, *data;
694 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
695 int ifindex = system_if_resolve(dev);
696 int i, rv;
697 static const struct {
698 const char *name;
699 enum macvlan_mode val;
700 } modes[] = {
701 { "private", MACVLAN_MODE_PRIVATE },
702 { "vepa", MACVLAN_MODE_VEPA },
703 { "bridge", MACVLAN_MODE_BRIDGE },
704 { "passthru", MACVLAN_MODE_PASSTHRU },
705 };
706
707 if (ifindex == 0)
708 return -ENOENT;
709
710 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
711
712 if (!msg)
713 return -1;
714
715 nlmsg_append(msg, &iim, sizeof(iim), 0);
716
717 if (cfg->flags & MACVLAN_OPT_MACADDR)
718 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
719 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
720 nla_put_u32(msg, IFLA_LINK, ifindex);
721
722 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
723 goto nla_put_failure;
724
725 nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
726
727 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
728 goto nla_put_failure;
729
730 if (cfg->mode) {
731 for (i = 0; i < ARRAY_SIZE(modes); i++) {
732 if (strcmp(cfg->mode, modes[i].name) != 0)
733 continue;
734
735 nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
736 break;
737 }
738 }
739
740 nla_nest_end(msg, data);
741 nla_nest_end(msg, linkinfo);
742
743 rv = system_rtnl_call(msg);
744 if (rv)
745 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
746
747 return rv;
748
749 nla_put_failure:
750 nlmsg_free(msg);
751 return -ENOMEM;
752 }
753
754 int system_macvlan_del(struct device *macvlan)
755 {
756 struct nl_msg *msg;
757 struct ifinfomsg iim = {
758 .ifi_family = AF_UNSPEC,
759 .ifi_index = 0,
760 };
761
762 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
763
764 if (!msg)
765 return -1;
766
767 nlmsg_append(msg, &iim, sizeof(iim), 0);
768
769 nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
770
771 system_rtnl_call(msg);
772
773 return 0;
774 }
775
776 static int system_vlan(struct device *dev, int id)
777 {
778 struct vlan_ioctl_args ifr = {
779 .cmd = SET_VLAN_NAME_TYPE_CMD,
780 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
781 };
782
783 ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
784
785 if (id < 0) {
786 ifr.cmd = DEL_VLAN_CMD;
787 ifr.u.VID = 0;
788 } else {
789 ifr.cmd = ADD_VLAN_CMD;
790 ifr.u.VID = id;
791 }
792 strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
793 return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
794 }
795
796 int system_vlan_add(struct device *dev, int id)
797 {
798 return system_vlan(dev, id);
799 }
800
801 int system_vlan_del(struct device *dev)
802 {
803 return system_vlan(dev, -1);
804 }
805
806 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
807 {
808 struct nl_msg *msg;
809 struct nlattr *linkinfo, *data;
810 struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
811 int ifindex = system_if_resolve(dev);
812 int rv;
813
814 if (ifindex == 0)
815 return -ENOENT;
816
817 msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
818
819 if (!msg)
820 return -1;
821
822 nlmsg_append(msg, &iim, sizeof(iim), 0);
823 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
824 nla_put_u32(msg, IFLA_LINK, ifindex);
825
826 if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
827 goto nla_put_failure;
828
829 nla_put_string(msg, IFLA_INFO_KIND, "vlan");
830
831 if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
832 goto nla_put_failure;
833
834 nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
835
836 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
837 nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
838 #else
839 if(cfg->proto == VLAN_PROTO_8021AD)
840 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);
841 #endif
842
843 nla_nest_end(msg, data);
844 nla_nest_end(msg, linkinfo);
845
846 rv = system_rtnl_call(msg);
847 if (rv)
848 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
849
850 return rv;
851
852 nla_put_failure:
853 nlmsg_free(msg);
854 return -ENOMEM;
855 }
856
857 int system_vlandev_del(struct device *vlandev)
858 {
859 struct nl_msg *msg;
860 struct ifinfomsg iim = {
861 .ifi_family = AF_UNSPEC,
862 .ifi_index = 0,
863 };
864
865 msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
866
867 if (!msg)
868 return -1;
869
870 nlmsg_append(msg, &iim, sizeof(iim), 0);
871
872 nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
873
874 system_rtnl_call(msg);
875
876 return 0;
877 }
878
879 static void
880 system_if_get_settings(struct device *dev, struct device_settings *s)
881 {
882 struct ifreq ifr;
883 char buf[10];
884
885 memset(&ifr, 0, sizeof(ifr));
886 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
887
888 if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
889 s->mtu = ifr.ifr_mtu;
890 s->flags |= DEV_OPT_MTU;
891 }
892
893 if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
894 s->txqueuelen = ifr.ifr_qlen;
895 s->flags |= DEV_OPT_TXQUEUELEN;
896 }
897
898 if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
899 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
900 s->flags |= DEV_OPT_MACADDR;
901 }
902
903 if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
904 s->ipv6 = !strtoul(buf, NULL, 0);
905 s->flags |= DEV_OPT_IPV6;
906 }
907 }
908
909 void
910 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
911 {
912 struct ifreq ifr;
913
914 if (!apply_mask)
915 return;
916
917 memset(&ifr, 0, sizeof(ifr));
918 strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
919 if (s->flags & DEV_OPT_MTU & apply_mask) {
920 ifr.ifr_mtu = s->mtu;
921 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
922 s->flags &= ~DEV_OPT_MTU;
923 }
924 if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
925 ifr.ifr_qlen = s->txqueuelen;
926 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
927 s->flags &= ~DEV_OPT_TXQUEUELEN;
928 }
929 if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
930 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
931 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
932 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
933 s->flags &= ~DEV_OPT_MACADDR;
934 }
935 if (s->flags & DEV_OPT_IPV6 & apply_mask)
936 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
937 }
938
939 int system_if_up(struct device *dev)
940 {
941 system_if_get_settings(dev, &dev->orig_settings);
942 system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
943 device_set_ifindex(dev, system_if_resolve(dev));
944 return system_if_flags(dev->ifname, IFF_UP, 0);
945 }
946
947 int system_if_down(struct device *dev)
948 {
949 int ret = system_if_flags(dev->ifname, 0, IFF_UP);
950 dev->orig_settings.flags &= dev->settings.flags;
951 system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
952 return ret;
953 }
954
955 struct if_check_data {
956 struct device *dev;
957 int pending;
958 int ret;
959 };
960
961 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
962 {
963 struct nlmsghdr *nh = nlmsg_hdr(msg);
964 struct ifinfomsg *ifi = NLMSG_DATA(nh);
965 struct if_check_data *chk = (struct if_check_data *)arg;
966
967 if (nh->nlmsg_type != RTM_NEWLINK)
968 return NL_SKIP;
969
970 device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
971 device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
972
973 return NL_OK;
974 }
975
976 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
977 {
978 struct if_check_data *chk = (struct if_check_data *)arg;
979 chk->pending = 0;
980 return NL_STOP;
981 }
982
983 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
984 {
985 struct if_check_data *chk = (struct if_check_data *)arg;
986
987 device_set_present(chk->dev, false);
988 device_set_link(chk->dev, false);
989 chk->pending = err->error;
990
991 return NL_STOP;
992 }
993
994 int system_if_check(struct device *dev)
995 {
996 struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
997 struct nl_msg *msg;
998 struct ifinfomsg ifi = {
999 .ifi_family = AF_UNSPEC,
1000 .ifi_index = 0,
1001 };
1002 struct if_check_data chk = {
1003 .dev = dev,
1004 .pending = 1,
1005 };
1006 int ret = 1;
1007
1008 msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1009 if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1010 nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1011 goto out;
1012
1013 nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1014 nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1015 nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1016
1017 nl_send_auto_complete(sock_rtnl, msg);
1018 while (chk.pending > 0)
1019 nl_recvmsgs(sock_rtnl, cb);
1020
1021 nlmsg_free(msg);
1022 ret = chk.pending;
1023
1024 out:
1025 nl_cb_put(cb);
1026 return ret;
1027 }
1028
1029 struct device *
1030 system_if_get_parent(struct device *dev)
1031 {
1032 char buf[64], *devname;
1033 int ifindex, iflink, len;
1034 FILE *f;
1035
1036 snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1037 f = fopen(buf, "r");
1038 if (!f)
1039 return NULL;
1040
1041 len = fread(buf, 1, sizeof(buf) - 1, f);
1042 fclose(f);
1043
1044 if (len <= 0)
1045 return NULL;
1046
1047 buf[len] = 0;
1048 iflink = strtoul(buf, NULL, 0);
1049 ifindex = system_if_resolve(dev);
1050 if (!iflink || iflink == ifindex)
1051 return NULL;
1052
1053 devname = if_indextoname(iflink, buf);
1054 if (!devname)
1055 return NULL;
1056
1057 return device_get(devname, true);
1058 }
1059
1060 static bool
1061 read_string_file(int dir_fd, const char *file, char *buf, int len)
1062 {
1063 bool ret = false;
1064 char *c;
1065 int fd;
1066
1067 fd = openat(dir_fd, file, O_RDONLY);
1068 if (fd < 0)
1069 return false;
1070
1071 retry:
1072 len = read(fd, buf, len - 1);
1073 if (len < 0) {
1074 if (errno == EINTR)
1075 goto retry;
1076 } else if (len > 0) {
1077 buf[len] = 0;
1078
1079 c = strchr(buf, '\n');
1080 if (c)
1081 *c = 0;
1082
1083 ret = true;
1084 }
1085
1086 close(fd);
1087
1088 return ret;
1089 }
1090
1091 static bool
1092 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1093 {
1094 char buf[64];
1095 bool ret = false;
1096
1097 ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1098 if (ret)
1099 *val = strtoull(buf, NULL, 0);
1100
1101 return ret;
1102 }
1103
1104 /* Assume advertised flags == supported flags */
1105 static const struct {
1106 uint32_t mask;
1107 const char *name;
1108 } ethtool_link_modes[] = {
1109 { ADVERTISED_10baseT_Half, "10H" },
1110 { ADVERTISED_10baseT_Full, "10F" },
1111 { ADVERTISED_100baseT_Half, "100H" },
1112 { ADVERTISED_100baseT_Full, "100F" },
1113 { ADVERTISED_1000baseT_Half, "1000H" },
1114 { ADVERTISED_1000baseT_Full, "1000F" },
1115 };
1116
1117 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1118 {
1119 int i;
1120 for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1121 if (mask & ethtool_link_modes[i].mask)
1122 blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1123 }
1124 }
1125
1126 bool
1127 system_if_force_external(const char *ifname)
1128 {
1129 char buf[64];
1130 struct stat s;
1131
1132 snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1133 return stat(buf, &s) == 0;
1134 }
1135
1136 int
1137 system_if_dump_info(struct device *dev, struct blob_buf *b)
1138 {
1139 struct ethtool_cmd ecmd;
1140 struct ifreq ifr;
1141 char buf[64], *s;
1142 void *c;
1143 int dir_fd;
1144
1145 snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1146 dir_fd = open(buf, O_DIRECTORY);
1147
1148 memset(&ecmd, 0, sizeof(ecmd));
1149 memset(&ifr, 0, sizeof(ifr));
1150 strcpy(ifr.ifr_name, dev->ifname);
1151 ifr.ifr_data = (caddr_t) &ecmd;
1152 ecmd.cmd = ETHTOOL_GSET;
1153
1154 if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1155 c = blobmsg_open_array(b, "link-advertising");
1156 system_add_link_modes(b, ecmd.advertising);
1157 blobmsg_close_array(b, c);
1158
1159 c = blobmsg_open_array(b, "link-supported");
1160 system_add_link_modes(b, ecmd.supported);
1161 blobmsg_close_array(b, c);
1162
1163 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1164 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1165 ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1166 blobmsg_add_string_buffer(b);
1167 }
1168
1169 close(dir_fd);
1170 return 0;
1171 }
1172
1173 int
1174 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1175 {
1176 const char *const counters[] = {
1177 "collisions", "rx_frame_errors", "tx_compressed",
1178 "multicast", "rx_length_errors", "tx_dropped",
1179 "rx_bytes", "rx_missed_errors", "tx_errors",
1180 "rx_compressed", "rx_over_errors", "tx_fifo_errors",
1181 "rx_crc_errors", "rx_packets", "tx_heartbeat_errors",
1182 "rx_dropped", "tx_aborted_errors", "tx_packets",
1183 "rx_errors", "tx_bytes", "tx_window_errors",
1184 "rx_fifo_errors", "tx_carrier_errors",
1185 };
1186 char buf[64];
1187 int stats_dir;
1188 int i;
1189 uint64_t val = 0;
1190
1191 snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1192 stats_dir = open(buf, O_DIRECTORY);
1193 if (stats_dir < 0)
1194 return -1;
1195
1196 for (i = 0; i < ARRAY_SIZE(counters); i++)
1197 if (read_uint64_file(stats_dir, counters[i], &val))
1198 blobmsg_add_u64(b, counters[i], val);
1199
1200 close(stats_dir);
1201 return 0;
1202 }
1203
1204 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1205 {
1206 bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1207 int alen = v4 ? 4 : 16;
1208 unsigned int flags = 0;
1209 struct ifaddrmsg ifa = {
1210 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1211 .ifa_prefixlen = addr->mask,
1212 .ifa_index = dev->ifindex,
1213 };
1214
1215 struct nl_msg *msg;
1216 if (cmd == RTM_NEWADDR)
1217 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1218
1219 msg = nlmsg_alloc_simple(cmd, flags);
1220 if (!msg)
1221 return -1;
1222
1223 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1224 nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1225 if (v4) {
1226 if (addr->broadcast)
1227 nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1228 if (addr->point_to_point)
1229 nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1230 } else {
1231 time_t now = system_get_rtime();
1232 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1233
1234 if (addr->preferred_until) {
1235 int64_t preferred = addr->preferred_until - now;
1236 if (preferred < 0)
1237 preferred = 0;
1238 else if (preferred > UINT32_MAX)
1239 preferred = UINT32_MAX;
1240
1241 cinfo.ifa_prefered = preferred;
1242 }
1243
1244 if (addr->valid_until) {
1245 int64_t valid = addr->valid_until - now;
1246 if (valid <= 0)
1247 return -1;
1248 else if (valid > UINT32_MAX)
1249 valid = UINT32_MAX;
1250
1251 cinfo.ifa_valid = valid;
1252 }
1253
1254 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1255 }
1256
1257 return system_rtnl_call(msg);
1258 }
1259
1260 int system_add_address(struct device *dev, struct device_addr *addr)
1261 {
1262 return system_addr(dev, addr, RTM_NEWADDR);
1263 }
1264
1265 int system_del_address(struct device *dev, struct device_addr *addr)
1266 {
1267 return system_addr(dev, addr, RTM_DELADDR);
1268 }
1269
1270 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1271 {
1272 int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1273 bool have_gw;
1274 unsigned int flags = 0;
1275
1276 if (alen == 4)
1277 have_gw = !!route->nexthop.in.s_addr;
1278 else
1279 have_gw = route->nexthop.in6.s6_addr32[0] ||
1280 route->nexthop.in6.s6_addr32[1] ||
1281 route->nexthop.in6.s6_addr32[2] ||
1282 route->nexthop.in6.s6_addr32[3];
1283
1284 unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
1285 (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1286
1287 unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1288 ? route->table : RT_TABLE_MAIN;
1289
1290 struct rtmsg rtm = {
1291 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1292 .rtm_dst_len = route->mask,
1293 .rtm_src_len = route->sourcemask,
1294 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1295 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1296 .rtm_scope = scope,
1297 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1298 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1299 };
1300 struct nl_msg *msg;
1301
1302 if (cmd == RTM_NEWROUTE) {
1303 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1304
1305 if (!dev) { // Add null-route
1306 rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1307 rtm.rtm_type = RTN_UNREACHABLE;
1308 }
1309 }
1310
1311 msg = nlmsg_alloc_simple(cmd, flags);
1312 if (!msg)
1313 return -1;
1314
1315 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1316
1317 if (route->mask)
1318 nla_put(msg, RTA_DST, alen, &route->addr);
1319
1320 if (route->sourcemask)
1321 nla_put(msg, RTA_SRC, alen, &route->source);
1322
1323 if (route->metric > 0)
1324 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1325
1326 if (have_gw)
1327 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1328
1329 if (dev)
1330 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1331
1332 if (table >= 256)
1333 nla_put_u32(msg, RTA_TABLE, table);
1334
1335 if (route->flags & DEVROUTE_MTU) {
1336 struct nlattr *metrics;
1337
1338 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1339 goto nla_put_failure;
1340
1341 nla_put_u32(msg, RTAX_MTU, route->mtu);
1342
1343 nla_nest_end(msg, metrics);
1344 }
1345
1346 return system_rtnl_call(msg);
1347
1348 nla_put_failure:
1349 nlmsg_free(msg);
1350 return -ENOMEM;
1351 }
1352
1353 int system_add_route(struct device *dev, struct device_route *route)
1354 {
1355 return system_rt(dev, route, RTM_NEWROUTE);
1356 }
1357
1358 int system_del_route(struct device *dev, struct device_route *route)
1359 {
1360 return system_rt(dev, route, RTM_DELROUTE);
1361 }
1362
1363 int system_flush_routes(void)
1364 {
1365 const char *names[] = {
1366 "/proc/sys/net/ipv4/route/flush",
1367 "/proc/sys/net/ipv6/route/flush"
1368 };
1369 int fd, i;
1370
1371 for (i = 0; i < ARRAY_SIZE(names); i++) {
1372 fd = open(names[i], O_WRONLY);
1373 if (fd < 0)
1374 continue;
1375
1376 if (write(fd, "-1", 2)) {}
1377 close(fd);
1378 }
1379 return 0;
1380 }
1381
1382 bool system_resolve_rt_table(const char *name, unsigned int *id)
1383 {
1384 FILE *f;
1385 char *e, buf[128];
1386 unsigned int n, table = RT_TABLE_UNSPEC;
1387
1388 /* first try to parse table as number */
1389 if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1390 table = n;
1391
1392 /* handle well known aliases */
1393 else if (!strcmp(name, "default"))
1394 table = RT_TABLE_DEFAULT;
1395 else if (!strcmp(name, "main"))
1396 table = RT_TABLE_MAIN;
1397 else if (!strcmp(name, "local"))
1398 table = RT_TABLE_LOCAL;
1399
1400 /* try to look up name in /etc/iproute2/rt_tables */
1401 else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1402 {
1403 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1404 {
1405 if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1406 continue;
1407
1408 n = strtoul(e, NULL, 10);
1409 e = strtok(NULL, " \t\n");
1410
1411 if (e && !strcmp(e, name))
1412 {
1413 table = n;
1414 break;
1415 }
1416 }
1417
1418 fclose(f);
1419 }
1420
1421 if (table == RT_TABLE_UNSPEC)
1422 return false;
1423
1424 /* do not consider main table special */
1425 if (table == RT_TABLE_MAIN)
1426 table = RT_TABLE_UNSPEC;
1427
1428 *id = table;
1429 return true;
1430 }
1431
1432 static int system_iprule(struct iprule *rule, int cmd)
1433 {
1434 int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1435
1436 struct nl_msg *msg;
1437 struct rtmsg rtm = {
1438 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1439 .rtm_protocol = RTPROT_STATIC,
1440 .rtm_scope = RT_SCOPE_UNIVERSE,
1441 .rtm_table = RT_TABLE_UNSPEC,
1442 .rtm_type = RTN_UNSPEC,
1443 .rtm_flags = 0,
1444 };
1445
1446 if (cmd == RTM_NEWRULE) {
1447 rtm.rtm_type = RTN_UNICAST;
1448 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1449 }
1450
1451 if (rule->invert)
1452 rtm.rtm_flags |= FIB_RULE_INVERT;
1453
1454 if (rule->flags & IPRULE_SRC)
1455 rtm.rtm_src_len = rule->src_mask;
1456
1457 if (rule->flags & IPRULE_DEST)
1458 rtm.rtm_dst_len = rule->dest_mask;
1459
1460 if (rule->flags & IPRULE_TOS)
1461 rtm.rtm_tos = rule->tos;
1462
1463 if (rule->flags & IPRULE_LOOKUP) {
1464 if (rule->lookup < 256)
1465 rtm.rtm_table = rule->lookup;
1466 }
1467
1468 if (rule->flags & IPRULE_ACTION)
1469 rtm.rtm_type = rule->action;
1470 else if (rule->flags & IPRULE_GOTO)
1471 rtm.rtm_type = FR_ACT_GOTO;
1472 else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1473 rtm.rtm_type = FR_ACT_NOP;
1474
1475 msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1476
1477 if (!msg)
1478 return -1;
1479
1480 nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1481
1482 if (rule->flags & IPRULE_IN)
1483 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1484
1485 if (rule->flags & IPRULE_OUT)
1486 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1487
1488 if (rule->flags & IPRULE_SRC)
1489 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1490
1491 if (rule->flags & IPRULE_DEST)
1492 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1493
1494 if (rule->flags & IPRULE_PRIORITY)
1495 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1496 else if (cmd == RTM_NEWRULE)
1497 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1498
1499 if (rule->flags & IPRULE_FWMARK)
1500 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1501
1502 if (rule->flags & IPRULE_FWMASK)
1503 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1504
1505 if (rule->flags & IPRULE_LOOKUP) {
1506 if (rule->lookup >= 256)
1507 nla_put_u32(msg, FRA_TABLE, rule->lookup);
1508 }
1509
1510 if (rule->flags & IPRULE_GOTO)
1511 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1512
1513 return system_rtnl_call(msg);
1514 }
1515
1516 int system_add_iprule(struct iprule *rule)
1517 {
1518 return system_iprule(rule, RTM_NEWRULE);
1519 }
1520
1521 int system_del_iprule(struct iprule *rule)
1522 {
1523 return system_iprule(rule, RTM_DELRULE);
1524 }
1525
1526 int system_flush_iprules(void)
1527 {
1528 int rv = 0;
1529 struct iprule rule;
1530
1531 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1532 system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1533
1534 memset(&rule, 0, sizeof(rule));
1535
1536
1537 rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1538
1539 rule.priority = 0;
1540 rule.lookup = RT_TABLE_LOCAL;
1541 rv |= system_iprule(&rule, RTM_NEWRULE);
1542
1543 rule.priority = 32766;
1544 rule.lookup = RT_TABLE_MAIN;
1545 rv |= system_iprule(&rule, RTM_NEWRULE);
1546
1547 rule.priority = 32767;
1548 rule.lookup = RT_TABLE_DEFAULT;
1549 rv |= system_iprule(&rule, RTM_NEWRULE);
1550
1551
1552 rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1553
1554 rule.priority = 0;
1555 rule.lookup = RT_TABLE_LOCAL;
1556 rv |= system_iprule(&rule, RTM_NEWRULE);
1557
1558 rule.priority = 32766;
1559 rule.lookup = RT_TABLE_MAIN;
1560 rv |= system_iprule(&rule, RTM_NEWRULE);
1561
1562 return rv;
1563 }
1564
1565 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1566 {
1567 char *e;
1568 unsigned int n;
1569
1570 if (!strcmp(action, "local"))
1571 n = RTN_LOCAL;
1572 else if (!strcmp(action, "nat"))
1573 n = RTN_NAT;
1574 else if (!strcmp(action, "broadcast"))
1575 n = RTN_BROADCAST;
1576 else if (!strcmp(action, "anycast"))
1577 n = RTN_ANYCAST;
1578 else if (!strcmp(action, "multicast"))
1579 n = RTN_MULTICAST;
1580 else if (!strcmp(action, "prohibit"))
1581 n = RTN_PROHIBIT;
1582 else if (!strcmp(action, "unreachable"))
1583 n = RTN_UNREACHABLE;
1584 else if (!strcmp(action, "blackhole"))
1585 n = RTN_BLACKHOLE;
1586 else if (!strcmp(action, "xresolve"))
1587 n = RTN_XRESOLVE;
1588 else if (!strcmp(action, "unicast"))
1589 n = RTN_UNICAST;
1590 else if (!strcmp(action, "throw"))
1591 n = RTN_THROW;
1592 else if (!strcmp(action, "failed_policy"))
1593 n = RTN_FAILED_POLICY;
1594 else {
1595 n = strtoul(action, &e, 0);
1596 if (!e || *e || e == action || n > 255)
1597 return false;
1598 }
1599
1600 *id = n;
1601 return true;
1602 }
1603
1604 time_t system_get_rtime(void)
1605 {
1606 struct timespec ts;
1607 struct timeval tv;
1608
1609 if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1610 return ts.tv_sec;
1611
1612 if (gettimeofday(&tv, NULL) == 0)
1613 return tv.tv_sec;
1614
1615 return 0;
1616 }
1617
1618 #ifndef IP_DF
1619 #define IP_DF 0x4000
1620 #endif
1621
1622 static int tunnel_ioctl(const char *name, int cmd, void *p)
1623 {
1624 struct ifreq ifr;
1625
1626 memset(&ifr, 0, sizeof(ifr));
1627 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1628 ifr.ifr_ifru.ifru_data = p;
1629 return ioctl(sock_ioctl, cmd, &ifr);
1630 }
1631
1632 int system_del_ip_tunnel(const char *name)
1633 {
1634 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1635 }
1636
1637 int system_update_ipv6_mtu(struct device *dev, int mtu)
1638 {
1639 int ret = -1;
1640 char buf[64];
1641 snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1642 dev->ifname);
1643
1644 int fd = open(buf, O_RDWR);
1645 ssize_t len = read(fd, buf, sizeof(buf) - 1);
1646 if (len < 0)
1647 goto out;
1648
1649 buf[len] = 0;
1650 ret = atoi(buf);
1651
1652 if (!mtu || ret <= mtu)
1653 goto out;
1654
1655 lseek(fd, 0, SEEK_SET);
1656 if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1657 ret = -1;
1658
1659 out:
1660 close(fd);
1661 return ret;
1662 }
1663
1664 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1665 {
1666 struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1667 struct blob_attr *cur;
1668 bool set_df = true;
1669 const char *str;
1670
1671 system_del_ip_tunnel(name);
1672
1673 blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1674 blob_data(attr), blob_len(attr));
1675
1676 if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1677 return -EINVAL;
1678 str = blobmsg_data(cur);
1679
1680 if ((cur = tb[TUNNEL_ATTR_DF]))
1681 set_df = blobmsg_get_bool(cur);
1682
1683 unsigned int ttl = 0;
1684 if ((cur = tb[TUNNEL_ATTR_TTL])) {
1685 ttl = blobmsg_get_u32(cur);
1686 if (ttl > 255 || (!set_df && ttl))
1687 return -EINVAL;
1688 }
1689
1690 unsigned int link = 0;
1691 if ((cur = tb[TUNNEL_ATTR_LINK])) {
1692 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1693 if (!iface)
1694 return -EINVAL;
1695
1696 if (iface->l3_dev.dev)
1697 link = iface->l3_dev.dev->ifindex;
1698 }
1699
1700 if (!strcmp(str, "sit")) {
1701 struct ip_tunnel_parm p = {
1702 .link = link,
1703 .iph = {
1704 .version = 4,
1705 .ihl = 5,
1706 .frag_off = set_df ? htons(IP_DF) : 0,
1707 .protocol = IPPROTO_IPV6,
1708 .ttl = ttl
1709 }
1710 };
1711
1712 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1713 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1714 return -EINVAL;
1715
1716 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1717 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1718 return -EINVAL;
1719
1720 strncpy(p.name, name, sizeof(p.name));
1721 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1722 return -1;
1723
1724 #ifdef SIOCADD6RD
1725 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1726 unsigned int mask;
1727 struct ip_tunnel_6rd p6;
1728
1729 memset(&p6, 0, sizeof(p6));
1730
1731 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1732 &p6.prefix, &mask) || mask > 128)
1733 return -EINVAL;
1734 p6.prefixlen = mask;
1735
1736 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1737 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1738 &p6.relay_prefix, &mask) || mask > 32)
1739 return -EINVAL;
1740 p6.relay_prefixlen = mask;
1741 }
1742
1743 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1744 system_del_ip_tunnel(name);
1745 return -1;
1746 }
1747 }
1748 #endif
1749 } else if (!strcmp(str, "ipip6")) {
1750 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
1751 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1752
1753 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
1754 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1755 nla_put_string(nlm, IFLA_IFNAME, name);
1756
1757 if (link)
1758 nla_put_u32(nlm, IFLA_LINK, link);
1759
1760 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1761 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
1762 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1763
1764 if (link)
1765 nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
1766
1767 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
1768 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
1769 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
1770
1771 struct in6_addr in6buf;
1772 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1773 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1)
1774 return -EINVAL;
1775 nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
1776 }
1777
1778 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1779 if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1)
1780 return -EINVAL;
1781 nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
1782 }
1783
1784 #ifdef IFLA_IPTUN_FMR_MAX
1785 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
1786 struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
1787
1788 struct blob_attr *fmr;
1789 unsigned rem, fmrcnt = 0;
1790 blobmsg_for_each_attr(fmr, cur, rem) {
1791 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
1792 continue;
1793
1794 unsigned ip4len, ip6len, ealen, offset = 6;
1795 char ip6buf[48];
1796 char ip4buf[16];
1797
1798 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
1799 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5)
1800 return -EINVAL;
1801
1802 struct in6_addr ip6prefix;
1803 struct in_addr ip4prefix;
1804 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
1805 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1)
1806 return -EINVAL;
1807
1808 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
1809
1810 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
1811 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
1812 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
1813 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
1814 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
1815 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
1816
1817 nla_nest_end(nlm, rule);
1818 }
1819
1820 nla_nest_end(nlm, fmrs);
1821 }
1822 #endif
1823
1824 nla_nest_end(nlm, infodata);
1825 nla_nest_end(nlm, linkinfo);
1826
1827 return system_rtnl_call(nlm);
1828 } else
1829 return -EINVAL;
1830
1831 return 0;
1832 }