netlink.c 21.5 KB
Newer Older
1
2
3
/*
 *	BIRD -- Linux Netlink Interface
 *
4
 *	(c) 1999--2000 Martin Mares <mj@ucw.cz>
5
6
7
8
9
10
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

#include <stdio.h>
#include <fcntl.h>
11
#include <net/if.h>
12
#include <sys/socket.h>
13
#include <sys/uio.h>
14
#include <errno.h>
15

16
#undef LOCAL_DEBUG
17

18
19
20
21
22
23
24
#include "nest/bird.h"
#include "nest/route.h"
#include "nest/protocol.h"
#include "nest/iface.h"
#include "lib/timer.h"
#include "lib/unix.h"
#include "lib/krt.h"
25
#include "lib/socket.h"
26
#include "lib/string.h"
27
#include "conf/conf.h"
28
29
30
31
32

#include <asm/types.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>

Martin Mareš's avatar
Martin Mareš committed
33
#ifndef MSG_TRUNC			/* Hack: Several versions of glibc miss this one :( */
34
35
#define MSG_TRUNC 0x20
#endif
36

37
38
39
40
/*
 *	Synchronous Netlink interface
 */

41
42
43
44
struct nl_sock
{
  int fd;
  u32 seq;
45
46
47
  byte *rx_buffer;			/* Receive buffer */
  struct nlmsghdr *last_hdr;		/* Recently received packet */
  unsigned int last_size;
48
49
};

50
#define NL_RX_SIZE 8192
51

52
53
54
static struct nl_sock nl_scan = {.fd = -1};	/* Netlink socket for synchronous scan */
static struct nl_sock nl_req  = {.fd = -1};	/* Netlink socket for requests */

55

56
static void
57
nl_open_sock(struct nl_sock *nl)
58
{
59
  if (nl->fd < 0)
60
    {
61
62
      nl->fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
      if (nl->fd < 0)
63
	die("Unable to open rtnetlink socket: %m");
64
      nl->seq = now;
65
66
67
      nl->rx_buffer = xmalloc(NL_RX_SIZE);
      nl->last_hdr = NULL;
      nl->last_size = 0;
68
69
70
    }
}

71
static void
72
73
74
75
76
77
78
79
nl_open(void)
{
  nl_open_sock(&nl_scan);
  nl_open_sock(&nl_req);
}

static void
nl_send(struct nl_sock *nl, struct nlmsghdr *nh)
80
81
82
83
84
85
{
  struct sockaddr_nl sa;

  memset(&sa, 0, sizeof(sa));
  sa.nl_family = AF_NETLINK;
  nh->nlmsg_pid = 0;
86
87
  nh->nlmsg_seq = ++(nl->seq);
  if (sendto(nl->fd, nh, nh->nlmsg_len, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0)
88
    die("rtnetlink sendto: %m");
89
  nl->last_hdr = NULL;
90
91
92
93
94
95
96
97
98
99
}

static void
nl_request_dump(int cmd)
{
  struct {
    struct nlmsghdr nh;
    struct rtgenmsg g;
  } req;
  req.nh.nlmsg_type = cmd;
100
  req.nh.nlmsg_len = sizeof(req);
101
  req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP;
102
  req.g.rtgen_family = BIRD_PF;
103
  nl_send(&nl_scan, &req.nh);
104
105
106
}

static struct nlmsghdr *
107
nl_get_reply(struct nl_sock *nl)
108
109
110
{
  for(;;)
    {
111
      if (!nl->last_hdr)
112
	{
113
	  struct iovec iov = { nl->rx_buffer, NL_RX_SIZE };
114
115
	  struct sockaddr_nl sa;
	  struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
116
	  int x = recvmsg(nl->fd, &m, 0);
117
118
	  if (x < 0)
	    die("nl_get_reply: %m");
119
120
121
122
123
	  if (sa.nl_pid)		/* It isn't from the kernel */
	    {
	      DBG("Non-kernel packet\n");
	      continue;
	    }
124
125
	  nl->last_size = x;
	  nl->last_hdr = (void *) nl->rx_buffer;
126
127
128
	  if (m.msg_flags & MSG_TRUNC)
	    bug("nl_get_reply: got truncated reply which should be impossible");
	}
129
      if (NLMSG_OK(nl->last_hdr, nl->last_size))
130
	{
131
132
	  struct nlmsghdr *h = nl->last_hdr;
	  nl->last_hdr = NLMSG_NEXT(h, nl->last_size);
133
	  if (h->nlmsg_seq != nl->seq)
134
135
	    {
	      log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)",
136
		  h->nlmsg_seq, nl->seq);
137
138
139
140
	      continue;
	    }
	  return h;
	}
141
142
143
      if (nl->last_size)
	log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl->last_size);
      nl->last_hdr = NULL;
144
145
146
    }
}

147
148
static struct rate_limit rl_netlink_err;

149
static int
150
151
nl_error(struct nlmsghdr *h)
{
152
153
154
  struct nlmsgerr *e;
  int ec;

155
  if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr)))
156
157
158
159
160
161
162
    {
      log(L_WARN "Netlink: Truncated error message received");
      return ENOBUFS;
    }
  e = (struct nlmsgerr *) NLMSG_DATA(h);
  ec = -e->error;
  if (ec)
163
    log_rl(&rl_netlink_err, L_WARN "Netlink: %s", strerror(ec));
164
  return ec;
165
166
167
168
169
}

static struct nlmsghdr *
nl_get_scan(void)
{
170
  struct nlmsghdr *h = nl_get_reply(&nl_scan);
171
172
173
174
175

  if (h->nlmsg_type == NLMSG_DONE)
    return NULL;
  if (h->nlmsg_type == NLMSG_ERROR)
    {
176
      nl_error(h);
177
178
179
180
181
      return NULL;
    }
  return h;
}

182
183
184
185
186
static int
nl_exchange(struct nlmsghdr *pkt)
{
  struct nlmsghdr *h;

187
  nl_send(&nl_req, pkt);
188
189
  for(;;)
    {
190
      h = nl_get_reply(&nl_req);
191
192
193
194
195
196
197
      if (h->nlmsg_type == NLMSG_ERROR)
	break;
      log(L_WARN "nl_exchange: Unexpected reply received");
    }
  return nl_error(h);
}

198
/*
199
 *	Netlink attributes
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
 */

static int nl_attr_len;

static void *
nl_checkin(struct nlmsghdr *h, int lsize)
{
  nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize);
  if (nl_attr_len < 0)
    {
      log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len);
      return NULL;
    }
  return NLMSG_DATA(h);
}

static int
nl_parse_attrs(struct rtattr *a, struct rtattr **k, int ksize)
{
  int max = ksize / sizeof(struct rtattr *);
  bzero(k, ksize);
  while (RTA_OK(a, nl_attr_len))
    {
      if (a->rta_type < max)
	k[a->rta_type] = a;
      a = RTA_NEXT(a, nl_attr_len);
    }
  if (nl_attr_len)
    {
      log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len);
      return 0;
    }
  else
    return 1;
}

236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
static void
nl_add_attr_u32(struct nlmsghdr *h, unsigned maxsize, int code, u32 data)
{
  unsigned len = RTA_LENGTH(4);
  struct rtattr *a;

  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
    bug("nl_add_attr32: packet buffer overflow");
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
  a->rta_type = code;
  a->rta_len = len;
  memcpy(RTA_DATA(a), &data, 4);
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
}

static void
nl_add_attr_ipa(struct nlmsghdr *h, unsigned maxsize, int code, ip_addr ipa)
{
  unsigned len = RTA_LENGTH(sizeof(ipa));
  struct rtattr *a;

  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
    bug("nl_add_attr_ipa: packet buffer overflow");
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
  a->rta_type = code;
  a->rta_len = len;
262
  ipa_hton(ipa);
263
264
265
266
  memcpy(RTA_DATA(a), &ipa, sizeof(ipa));
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
}

267
268
269
270
271
/*
 *	Scanning of interfaces
 */

static void
272
nl_parse_link(struct nlmsghdr *h, int scan)
273
274
{
  struct ifinfomsg *i;
275
  struct rtattr *a[IFLA_WIRELESS+1];
276
277
278
279
280
281
  int new = h->nlmsg_type == RTM_NEWLINK;
  struct iface f;
  struct iface *ifi;
  char *name;
  u32 mtu;
  unsigned int fl;
282
283
284

  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), a, sizeof(a)))
    return;
285
286
287
  if (!a[IFLA_IFNAME] || RTA_PAYLOAD(a[IFLA_IFNAME]) < 2 ||
      !a[IFLA_MTU] || RTA_PAYLOAD(a[IFLA_MTU]) != 4)
    {
288
      if (scan || !a[IFLA_WIRELESS])
289
        log(L_ERR "nl_parse_link: Malformed message received");
290
291
292
293
294
295
296
297
      return;
    }
  name = RTA_DATA(a[IFLA_IFNAME]);
  memcpy(&mtu, RTA_DATA(a[IFLA_MTU]), sizeof(u32));

  ifi = if_find_by_index(i->ifi_index);
  if (!new)
    {
298
      DBG("KIF: IF%d(%s) goes down\n", i->ifi_index, name);
299
300
301
302
303
304
305
306
307
      if (ifi && !scan)
	{
	  memcpy(&f, ifi, sizeof(struct iface));
	  f.flags |= IF_ADMIN_DOWN;
	  if_update(&f);
	}
    }
  else
    {
308
      DBG("KIF: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags);
309
310
311
312
313
314
315
316
317
318
319
320
321
      if (ifi)
	memcpy(&f, ifi, sizeof(f));
      else
	{
	  bzero(&f, sizeof(f));
	  f.index = i->ifi_index;
	}
      strncpy(f.name, RTA_DATA(a[IFLA_IFNAME]), sizeof(f.name)-1);
      f.mtu = mtu;
      f.flags = 0;
      fl = i->ifi_flags;
      if (fl & IFF_UP)
	f.flags |= IF_LINK_UP;
322
323
324
325
326
327
328
329
      if (fl & IFF_LOOPBACK)		/* Loopback */
	f.flags |= IF_MULTIACCESS | IF_LOOPBACK | IF_IGNORE;
      else if (fl & IFF_POINTOPOINT)	/* PtP */
	f.flags |= IF_MULTICAST;
      else if (fl & IFF_BROADCAST)	/* Broadcast */
	f.flags |= IF_MULTIACCESS | IF_BROADCAST | IF_MULTICAST;
      else
	f.flags |= IF_MULTIACCESS;	/* NBMA */
330
331
      if_update(&f);
    }
332
333
334
335
336
337
338
}

static void
nl_parse_addr(struct nlmsghdr *h)
{
  struct ifaddrmsg *i;
  struct rtattr *a[IFA_ANYCAST+1];
339
  int new = h->nlmsg_type == RTM_NEWADDR;
340
  struct ifa ifa;
341
  struct iface *ifi;
342
  int scope;
343
344
345

  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFA_RTA(i), a, sizeof(a)))
    return;
346
  if (i->ifa_family != BIRD_AF)
347
    return;
348
349
350
351
352
353
354
355
  if (!a[IFA_ADDRESS] || RTA_PAYLOAD(a[IFA_ADDRESS]) != sizeof(ip_addr)
#ifdef IPV6
      || a[IFA_LOCAL] && RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
#else
      || !a[IFA_LOCAL] || RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
      || (a[IFA_BROADCAST] && RTA_PAYLOAD(a[IFA_BROADCAST]) != sizeof(ip_addr))
#endif
      )
356
357
358
359
360
361
362
363
    {
      log(L_ERR "nl_parse_addr: Malformed message received");
      return;
    }

  ifi = if_find_by_index(i->ifa_index);
  if (!ifi)
    {
364
      log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index);
365
366
367
      return;
    }

368
369
370
371
  bzero(&ifa, sizeof(ifa));
  ifa.iface = ifi;
  if (i->ifa_flags & IFA_F_SECONDARY)
    ifa.flags |= IA_SECONDARY;
372

373
374
  /* IFA_LOCAL can be unset for IPv6 interfaces */
  memcpy(&ifa.ip, RTA_DATA(a[IFA_LOCAL] ? : a[IFA_ADDRESS]), sizeof(ifa.ip));
375
  ipa_ntoh(ifa.ip);
376
  ifa.pxlen = i->ifa_prefixlen;
377
378
  if (i->ifa_prefixlen > BITS_PER_IP_ADDRESS ||
      i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 1)
379
    {
380
381
382
383
384
385
      log(L_ERR "KIF: Invalid prefix length for interface %s: %d", ifi->name, i->ifa_prefixlen);
      new = 0;
    }
  if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS)
    {
      ifa.flags |= IA_UNNUMBERED;
386
      memcpy(&ifa.opposite, RTA_DATA(a[IFA_ADDRESS]), sizeof(ifa.opposite));
387
      ipa_ntoh(ifa.opposite);
388
      ifa.prefix = ifa.brd = ifa.opposite;
389
    }
390
  else
391
    {
392
      ip_addr netmask = ipa_mkmask(ifa.pxlen);
393
394
395
      ip_addr xbrd;
      ifa.prefix = ipa_and(ifa.ip, netmask);
      ifa.brd = ipa_or(ifa.ip, ipa_not(netmask));
396
#ifndef IPV6
397
      if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 2)
398
	ifa.opposite = ipa_opposite(ifa.ip, i->ifa_prefixlen);
399
400
      if ((ifi->flags & IF_BROADCAST) && a[IFA_BROADCAST])
	{
401
402
403
404
	  memcpy(&xbrd, RTA_DATA(a[IFA_BROADCAST]), sizeof(xbrd));
	  ipa_ntoh(xbrd);
	  if (ipa_equal(xbrd, ifa.prefix) || ipa_equal(xbrd, ifa.brd))
	    ifa.brd = xbrd;
405
	  else if (ifi->flags & IF_TMP_DOWN) /* Complain only during the first scan */
406
	    log(L_ERR "KIF: Invalid broadcast address %I for %s", xbrd, ifi->name);
407
	}
408
#endif
409
    }
410

411
412
413
  scope = ipa_classify(ifa.ip);
  if (scope < 0)
    {
414
      log(L_ERR "KIF: Invalid interface address %I for %s", ifa.ip, ifi->name);
415
416
417
418
      return;
    }
  ifa.scope = scope & IADDR_SCOPE_MASK;

419
  DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %I/%d, brd %I, opp %I\n",
420
421
      ifi->index, ifi->name,
      new ? "added" : "removed",
422
      ifa.ip, ifa.flags, ifa.prefix, ifa.pxlen, ifa.brd, ifa.opposite);
423
424
425
426
  if (new)
    ifa_update(&ifa);
  else
    ifa_delete(&ifa);
427
428
}

429
void
430
krt_if_scan(struct kif_proto *p UNUSED)
431
432
433
{
  struct nlmsghdr *h;

434
435
  if_start_update();

436
437
438
  nl_request_dump(RTM_GETLINK);
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK)
439
      nl_parse_link(h, 1);
440
441
442
443
444
445
446
447
448
449
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

  nl_request_dump(RTM_GETADDR);
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR)
      nl_parse_addr(h);
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

450
  if_end_update();
451
452
453
}

/*
454
 *	Routes
455
456
 */

457
static struct krt_proto *nl_table_map[NL_NUM_TABLES];
458

459
int
460
461
krt_capable(rte *e)
{
Martin Mareš's avatar
Martin Mareš committed
462
463
  rta *a = e->attrs;

464
  if (a->cast != RTC_UNICAST
465
#if 0
466
467
468
      && a->cast != RTC_ANYCAST
#endif
      )
Martin Mareš's avatar
Martin Mareš committed
469
    return 0;
470
471
  if (a->source == RTS_DEVICE)	/* Kernel takes care of device routes itself */
    return 0;
Martin Mareš's avatar
Martin Mareš committed
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
  switch (a->dest)
    {
    case RTD_ROUTER:
    case RTD_DEVICE:
    case RTD_BLACKHOLE:
    case RTD_UNREACHABLE:
    case RTD_PROHIBIT:
      break;
    default:
      return 0;
    }
  return 1;
}

static void
487
nl_send_route(struct krt_proto *p, rte *e, int new)
Martin Mareš's avatar
Martin Mareš committed
488
{
489
490
491
492
493
494
495
496
  net *net = e->net;
  rta *a = e->attrs;
  struct {
    struct nlmsghdr h;
    struct rtmsg r;
    char buf[128];
  } r;

497
498
  DBG("nl_send_route(%I/%d,new=%d)\n", net->n.prefix, net->n.pxlen, new);

499
500
501
502
  bzero(&r.h, sizeof(r.h));
  bzero(&r.r, sizeof(r.r));
  r.h.nlmsg_type = new ? RTM_NEWROUTE : RTM_DELROUTE;
  r.h.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
503
  r.h.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | (new ? NLM_F_CREATE|NLM_F_EXCL : 0);
504

505
  r.r.rtm_family = BIRD_AF;
506
  r.r.rtm_dst_len = net->n.pxlen;
507
508
  r.r.rtm_tos = 0;
  r.r.rtm_table = KRT_CF->scan.table_id;
509
  r.r.rtm_protocol = RTPROT_BIRD;
510
  r.r.rtm_scope = RT_SCOPE_UNIVERSE;
511
512
513
514
515
516
517
518
  nl_add_attr_ipa(&r.h, sizeof(r), RTA_DST, net->n.prefix);
  switch (a->dest)
    {
    case RTD_ROUTER:
      r.r.rtm_type = RTN_UNICAST;
      nl_add_attr_ipa(&r.h, sizeof(r), RTA_GATEWAY, a->gw);
      break;
    case RTD_DEVICE:
519
520
      if (!a->iface)
	return;
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
      r.r.rtm_type = RTN_UNICAST;
      nl_add_attr_u32(&r.h, sizeof(r), RTA_OIF, a->iface->index);
      break;
    case RTD_BLACKHOLE:
      r.r.rtm_type = RTN_BLACKHOLE;
      break;
    case RTD_UNREACHABLE:
      r.r.rtm_type = RTN_UNREACHABLE;
      break;
    case RTD_PROHIBIT:
      r.r.rtm_type = RTN_PROHIBIT;
      break;
    default:
      bug("krt_capable inconsistent with nl_send_route");
    }
Martin Mareš's avatar
Martin Mareš committed
536

537
  nl_exchange(&r.h);
538
}
539

540
void
541
krt_set_notify(struct krt_proto *p, net *n UNUSED, rte *new, rte *old)
542
{
543
544
  if (old)
    nl_send_route(p, old, 0);
545

546
547
  if (new)
    nl_send_route(p, new, 1);
Martin Mareš's avatar
Martin Mareš committed
548
549
}

550
static struct iface *
Martin Mareš's avatar
Martin Mareš committed
551
552
553
554
555
556
557
krt_temp_iface(struct krt_proto *p, unsigned index)
{
  struct iface *i, *j;

  WALK_LIST(i, p->scan.temp_ifs)
    if (i->index == index)
      return i;
558
  i = mb_allocz(p->p.pool, sizeof(struct iface));
Martin Mareš's avatar
Martin Mareš committed
559
560
561
562
563
564
565
566
567
568
  if (j = if_find_by_index(index))
    strcpy(i->name, j->name);
  else
    strcpy(i->name, "?");
  i->index = index;
  add_tail(&p->scan.temp_ifs, &i->n);
  return i;
}

static void
569
nl_parse_route(struct nlmsghdr *h, int scan)
Martin Mareš's avatar
Martin Mareš committed
570
{
571
  struct krt_proto *p;
Martin Mareš's avatar
Martin Mareš committed
572
573
574
575
576
577
578
579
  struct rtmsg *i;
  struct rtattr *a[RTA_CACHEINFO+1];
  int new = h->nlmsg_type == RTM_NEWROUTE;
  ip_addr dst;
  rta ra;
  rte *e;
  net *net;
  u32 oif;
580
  int src;
Martin Mareš's avatar
Martin Mareš committed
581
582
583

  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(RTM_RTA(i), a, sizeof(a)))
    return;
584
  if (i->rtm_family != BIRD_AF)
Martin Mareš's avatar
Martin Mareš committed
585
586
587
    return;
  if ((a[RTA_DST] && RTA_PAYLOAD(a[RTA_DST]) != sizeof(ip_addr)) ||
      (a[RTA_OIF] && RTA_PAYLOAD(a[RTA_OIF]) != 4) ||
588
      (a[RTA_PRIORITY] && RTA_PAYLOAD(a[RTA_PRIORITY]) != 4) ||
589
590
591
#ifdef IPV6
      (a[RTA_IIF] && RTA_PAYLOAD(a[RTA_IIF]) != 4) ||
#endif
Martin Mareš's avatar
Martin Mareš committed
592
593
594
595
596
597
      (a[RTA_GATEWAY] && RTA_PAYLOAD(a[RTA_GATEWAY]) != sizeof(ip_addr)))
    {
      log(L_ERR "nl_parse_route: Malformed message received");
      return;
    }

598
599
  p = nl_table_map[i->rtm_table];	/* Do we know this table? */
  if (!p)
Martin Mareš's avatar
Martin Mareš committed
600
    return;
601

602
603
604
605
606
607
608
#ifdef IPV6
  if (a[RTA_IIF])
    {
      DBG("KRT: Ignoring route with IIF set\n");
      return;
    }
#else
609
610
611
612
613
  if (i->rtm_tos != 0)			/* We don't support TOS */
    {
      DBG("KRT: Ignoring route with TOS %02x\n", i->rtm_tos);
      return;
    }
614
#endif
615

616
  if (scan && !new)
Martin Mareš's avatar
Martin Mareš committed
617
618
619
620
621
622
623
624
    {
      DBG("KRT: Ignoring route deletion\n");
      return;
    }

  if (a[RTA_DST])
    {
      memcpy(&dst, RTA_DATA(a[RTA_DST]), sizeof(dst));
625
      ipa_ntoh(dst);
Martin Mareš's avatar
Martin Mareš committed
626
627
628
629
630
631
632
633
    }
  else
    dst = IPA_NONE;
  if (a[RTA_OIF])
    memcpy(&oif, RTA_DATA(a[RTA_OIF]), sizeof(oif));
  else
    oif = ~0;

634
  DBG("Got %I/%d, type=%d, oif=%d, table=%d, prid=%d, proto=%s\n", dst, i->rtm_dst_len, i->rtm_type, oif, i->rtm_table, i->rtm_protocol, p->p.name);
Martin Mareš's avatar
Martin Mareš committed
635

636
637
638
639
640
641
642
643
644
  switch (i->rtm_protocol)
    {
    case RTPROT_REDIRECT:
      src = KRT_SRC_REDIRECT;
      break;
    case RTPROT_KERNEL:
      DBG("Route originated in kernel, ignoring\n");
      return;
    case RTPROT_BIRD:
645
646
647
648
#ifdef IPV6
    case RTPROT_BOOT:
      /* Current Linux kernels don't remember rtm_protocol for IPv6 routes and supply RTPROT_BOOT instead */
#endif
649
650
651
652
653
654
655
656
657
658
659
      if (!scan)
	{
	  DBG("Echo of our own route, ignoring\n");
	  return;
	}
      src = KRT_SRC_BIRD;
      break;
    default:
      src = KRT_SRC_ALIEN;
    }

660
  net = net_get(p->p.table, dst, i->rtm_dst_len);
Martin Mareš's avatar
Martin Mareš committed
661
662
  ra.proto = &p->p;
  ra.source = RTS_INHERIT;
663
  ra.scope = SCOPE_UNIVERSE;
Martin Mareš's avatar
Martin Mareš committed
664
  ra.cast = RTC_UNICAST;
665
  ra.flags = ra.aflags = 0;
Martin Mareš's avatar
Martin Mareš committed
666
667
668
  ra.from = IPA_NONE;
  ra.gw = IPA_NONE;
  ra.iface = NULL;
669
  ra.eattrs = NULL;
Martin Mareš's avatar
Martin Mareš committed
670
671
672
673
674
675
676
677
678
679
680
681
682
683

  switch (i->rtm_type)
    {
    case RTN_UNICAST:
      if (oif == ~0U)
	{
	  log(L_ERR "KRT: Mysterious route with no OIF (%I/%d)", net->n.prefix, net->n.pxlen);
	  return;
	}
      if (a[RTA_GATEWAY])
	{
	  neighbor *ng;
	  ra.dest = RTD_ROUTER;
	  memcpy(&ra.gw, RTA_DATA(a[RTA_GATEWAY]), sizeof(ra.gw));
684
	  ipa_ntoh(ra.gw);
685
686
687
688
689
690
691
692
693
694
695
696

	  if (i->rtm_flags & RTNH_F_ONLINK)
	    {
	      /* route with 'onlink' attribute */
	      ra.iface = if_find_by_index(oif);
	      if (ra.iface == NULL)
		{
		  log(L_WARN "Kernel told us to use unknown interface %u for %I/%d",
		      oif, net->n.prefix, net->n.pxlen);
		  return;
		}
	    }
Martin Mareš's avatar
Martin Mareš committed
697
	  else
698
	    {
699
700
701
702
703
704
705
706
707
	      /* standard route */
	      ng = neigh_find(&p->p, &ra.gw, 0);
	      if (ng && ng->scope)
		ra.iface = ng->iface;
	      else
		{
		  log(L_WARN "Kernel told us to use non-neighbor %I for %I/%d", ra.gw, net->n.prefix, net->n.pxlen);
		  return;
		}
708
	    }
Martin Mareš's avatar
Martin Mareš committed
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
	}
      else
	{
	  ra.dest = RTD_DEVICE;
	  ra.iface = krt_temp_iface(p, oif);
	}
      break;
    case RTN_BLACKHOLE:
      ra.dest = RTD_BLACKHOLE;
      break;
    case RTN_UNREACHABLE:
      ra.dest = RTD_UNREACHABLE;
      break;
    case RTN_PROHIBIT:
      ra.dest = RTD_PROHIBIT;
      break;
    /* FIXME: What about RTN_THROW? */
    default:
      DBG("KRT: Ignoring route with type=%d\n", i->rtm_type);
      return;
    }
730

731
  if (i->rtm_scope != RT_SCOPE_UNIVERSE)
732
733
734
735
736
    {
      DBG("KRT: Ignoring route with scope=%d\n", i->rtm_scope);
      return;
    }

Martin Mareš's avatar
Martin Mareš committed
737
738
  e = rte_get_temp(&ra);
  e->net = net;
739
740
741
742
743
744
745
  e->u.krt.src = src;
  e->u.krt.proto = i->rtm_protocol;
  e->u.krt.type = i->rtm_type;
  if (a[RTA_PRIORITY])
    memcpy(&e->u.krt.metric, RTA_DATA(a[RTA_PRIORITY]), sizeof(e->u.krt.metric));
  else
    e->u.krt.metric = 0;
746
747
748
749
  if (scan)
    krt_got_route(p, e);
  else
    krt_got_route_async(p, e, new);
750
751
752
}

void
753
krt_scan_fire(struct krt_proto *p UNUSED)	/* CONFIG_ALL_TABLES_AT_ONCE => p is NULL */
754
{
Martin Mareš's avatar
Martin Mareš committed
755
756
757
758
759
  struct nlmsghdr *h;

  nl_request_dump(RTM_GETROUTE);
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)
760
      nl_parse_route(h, 1);
Martin Mareš's avatar
Martin Mareš committed
761
762
    else
      log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type);
763
764
765
}

/*
766
 *	Asynchronous Netlink interface
767
768
 */

769
static sock *nl_async_sk;		/* BIRD socket for asynchronous notifications */
770
771
772
static byte *nl_async_rx_buffer;	/* Receive buffer */

static void
773
nl_async_msg(struct nlmsghdr *h)
774
775
776
777
778
779
{
  switch (h->nlmsg_type)
    {
    case RTM_NEWROUTE:
    case RTM_DELROUTE:
      DBG("KRT: Received async route notification (%d)\n", h->nlmsg_type);
780
      nl_parse_route(h, 0);
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
      break;
    case RTM_NEWLINK:
    case RTM_DELLINK:
      DBG("KRT: Received async link notification (%d)\n", h->nlmsg_type);
      nl_parse_link(h, 0);
      break;
    case RTM_NEWADDR:
    case RTM_DELADDR:
      DBG("KRT: Received async address notification (%d)\n", h->nlmsg_type);
      nl_parse_addr(h);
      break;
    default:
      DBG("KRT: Received unknown async notification (%d)\n", h->nlmsg_type);
    }
}
796

797
static int
798
nl_async_hook(sock *sk, int size UNUSED)
799
{
800
801
802
803
804
805
806
807
808
809
  struct iovec iov = { nl_async_rx_buffer, NL_RX_SIZE };
  struct sockaddr_nl sa;
  struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
  struct nlmsghdr *h;
  int x;
  unsigned int len;

  x = recvmsg(sk->fd, &m, 0);
  if (x < 0)
    {
810
811
812
813
814
815
816
817
818
819
      if (errno == ENOBUFS)
	{
	  /*
	   *  Netlink reports some packets have been thrown away.
	   *  One day we might react to it by asking for route table
	   *  scan in near future.
	   */
	  return 1;	/* More data are likely to be ready */
	}
      else if (errno != EWOULDBLOCK)
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
	log(L_ERR "Netlink recvmsg: %m");
      return 0;
    }
  if (sa.nl_pid)		/* It isn't from the kernel */
    {
      DBG("Non-kernel packet\n");
      return 1;
    }
  h = (void *) nl_async_rx_buffer;
  len = x;
  if (m.msg_flags & MSG_TRUNC)
    {
      log(L_WARN "Netlink got truncated asynchronous message");
      return 1;
    }
  while (NLMSG_OK(h, len))
    {
837
      nl_async_msg(h);
838
839
840
841
842
      h = NLMSG_NEXT(h, len);
    }
  if (len)
    log(L_WARN "nl_async_hook: Found packet remnant of size %d", len);
  return 1;
843
844
}

845
static void
846
nl_open_async(void)
847
848
849
{
  sock *sk;
  struct sockaddr_nl sa;
850
  int fd;
851
852
853
854
855
  static int nl_open_tried = 0;

  if (nl_open_tried)
    return;
  nl_open_tried = 1;
856

857
858
  DBG("KRT: Opening async netlink socket\n");

859
860
861
  fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
  if (fd < 0)
    {
862
      log(L_ERR "Unable to open asynchronous rtnetlink socket: %m");
863
864
      return;
    }
865
866
867

  bzero(&sa, sizeof(sa));
  sa.nl_family = AF_NETLINK;
868
869
870
#ifdef IPV6
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV6_IFADDR | RTMGRP_IPV6_ROUTE;
#else
871
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE;
872
#endif
873
874
  if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0)
    {
875
      log(L_ERR "Unable to bind asynchronous rtnetlink socket: %m");
876
877
878
      return;
    }

879
  sk = nl_async_sk = sk_new(krt_pool);
880
881
882
883
884
885
886
887
  sk->type = SK_MAGIC;
  sk->rx_hook = nl_async_hook;
  sk->fd = fd;
  if (sk_open(sk))
    bug("Netlink: sk_open failed");

  if (!nl_async_rx_buffer)
    nl_async_rx_buffer = xmalloc(NL_RX_SIZE);
888
889
}

890
891
892
/*
 *	Interface to the UNIX krt module
 */
893

894
static u8 nl_cf_table[(NL_NUM_TABLES+7) / 8];
895

896
void
897
krt_scan_preconfig(struct config *c UNUSED)
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
{
  bzero(&nl_cf_table, sizeof(nl_cf_table));
}

void
krt_scan_postconfig(struct krt_config *x)
{
  int id = x->scan.table_id;

  if (nl_cf_table[id/8] & (1 << (id%8)))
    cf_error("Multiple kernel syncers defined for table #%d", id);
  nl_cf_table[id/8] |= (1 << (id%8));
}

void
krt_scan_construct(struct krt_config *x)
914
{
915
#ifndef IPV6
916
  x->scan.table_id = RT_TABLE_MAIN;
917
918
#else
  x->scan.table_id = 254;
919
#endif
920
921
}

922
void
923
krt_scan_start(struct krt_proto *p, int first)
924
{
Martin Mareš's avatar
Martin Mareš committed
925
  init_list(&p->scan.temp_ifs);
926
927
928
929
  nl_table_map[KRT_CF->scan.table_id] = p;
  if (first)
    {
      nl_open();
930
      nl_open_async();
931
    }
932
933
934
}

void
935
krt_scan_shutdown(struct krt_proto *p UNUSED, int last UNUSED)
936
937
{
}
938
939

void
940
krt_if_start(struct kif_proto *p UNUSED)
941
942
{
  nl_open();
943
  nl_open_async();
944
}