Statistics
| Branch: | Revision:

iof-bird-daemon / sysdep / linux / netlink / netlink.c @ df49d4e1

History | View | Annotate | Download (20.8 KB)

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

    
9
#include <stdio.h>
10
#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
#include "nest/bird.h"
19
#include "nest/route.h"
20
#include "nest/protocol.h"
21
#include "nest/iface.h"
22
#include "lib/timer.h"
23
#include "lib/unix.h"
24
#include "lib/krt.h"
25
#include "lib/socket.h"
26
#include "lib/string.h"
27
#include "conf/conf.h"
28

    
29
#include <asm/types.h>
30
#include <linux/netlink.h>
31
#include <linux/rtnetlink.h>
32

    
33
#ifndef MSG_TRUNC                        /* Hack: Several versions of glibc miss this one :( */
34
#define MSG_TRUNC 0x20
35
#endif
36

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

    
41
static int nl_sync_fd = -1;                /* Unix socket for synchronous netlink actions */
42
static u32 nl_sync_seq;                        /* Sequence number of last request sent */
43

    
44
static byte *nl_rx_buffer;                /* Receive buffer */
45
#define NL_RX_SIZE 8192
46

    
47
static struct nlmsghdr *nl_last_hdr;        /* Recently received packet */
48
static unsigned int nl_last_size;
49

    
50
static void
51
nl_open(void)
52
{
53
  if (nl_sync_fd < 0)
54
    {
55
      nl_sync_fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
56
      if (nl_sync_fd < 0)
57
        die("Unable to open rtnetlink socket: %m");
58
      nl_sync_seq = now;
59
      nl_rx_buffer = xmalloc(NL_RX_SIZE);
60
    }
61
}
62

    
63
static void
64
nl_send(struct nlmsghdr *nh)
65
{
66
  struct sockaddr_nl sa;
67

    
68
  memset(&sa, 0, sizeof(sa));
69
  sa.nl_family = AF_NETLINK;
70
  nh->nlmsg_pid = 0;
71
  nh->nlmsg_seq = ++nl_sync_seq;
72
  if (sendto(nl_sync_fd, nh, nh->nlmsg_len, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0)
73
    die("rtnetlink sendto: %m");
74
  nl_last_hdr = NULL;
75
}
76

    
77
static void
78
nl_request_dump(int cmd)
79
{
80
  struct {
81
    struct nlmsghdr nh;
82
    struct rtgenmsg g;
83
  } req;
84
  req.nh.nlmsg_type = cmd;
85
  req.nh.nlmsg_len = sizeof(req);
86
  req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP;
87
  req.g.rtgen_family = BIRD_PF;
88
  nl_send(&req.nh);
89
}
90

    
91
static struct nlmsghdr *
92
nl_get_reply(void)
93
{
94
  for(;;)
95
    {
96
      if (!nl_last_hdr)
97
        {
98
          struct iovec iov = { nl_rx_buffer, NL_RX_SIZE };
99
          struct sockaddr_nl sa;
100
          struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
101
          int x = recvmsg(nl_sync_fd, &m, 0);
102
          if (x < 0)
103
            die("nl_get_reply: %m");
104
          if (sa.nl_pid)                /* It isn't from the kernel */
105
            {
106
              DBG("Non-kernel packet\n");
107
              continue;
108
            }
109
          nl_last_size = x;
110
          nl_last_hdr = (void *) nl_rx_buffer;
111
          if (m.msg_flags & MSG_TRUNC)
112
            bug("nl_get_reply: got truncated reply which should be impossible");
113
        }
114
      if (NLMSG_OK(nl_last_hdr, nl_last_size))
115
        {
116
          struct nlmsghdr *h = nl_last_hdr;
117
          if (h->nlmsg_seq != nl_sync_seq)
118
            {
119
              log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)",
120
                  h->nlmsg_seq, nl_sync_seq);
121
              continue;
122
            }
123
          nl_last_hdr = NLMSG_NEXT(h, nl_last_size);
124
          return h;
125
        }
126
      if (nl_last_size)
127
        log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl_last_size);
128
      nl_last_hdr = NULL;
129
    }
130
}
131

    
132
static int
133
nl_error(struct nlmsghdr *h)
134
{
135
  struct nlmsgerr *e;
136
  int ec;
137

    
138
  if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr)))
139
    {
140
      log(L_WARN "Netlink: Truncated error message received");
141
      return ENOBUFS;
142
    }
143
  e = (struct nlmsgerr *) NLMSG_DATA(h);
144
  ec = -e->error;
145
  if (ec)
146
    log(L_WARN "Netlink: %s", strerror(ec));
147
  return ec;
148
}
149

    
150
static struct nlmsghdr *
151
nl_get_scan(void)
152
{
153
  struct nlmsghdr *h = nl_get_reply();
154

    
155
  if (h->nlmsg_type == NLMSG_DONE)
156
    return NULL;
157
  if (h->nlmsg_type == NLMSG_ERROR)
158
    {
159
      nl_error(h);
160
      return NULL;
161
    }
162
  return h;
163
}
164

    
165
static int
166
nl_exchange(struct nlmsghdr *pkt)
167
{
168
  struct nlmsghdr *h;
169

    
170
  nl_send(pkt);
171
  for(;;)
172
    {
173
      h = nl_get_reply();
174
      if (h->nlmsg_type == NLMSG_ERROR)
175
        break;
176
      log(L_WARN "nl_exchange: Unexpected reply received");
177
    }
178
  return nl_error(h);
179
}
180

    
181
/*
182
 *        Netlink attributes
183
 */
184

    
185
static int nl_attr_len;
186

    
187
static void *
188
nl_checkin(struct nlmsghdr *h, int lsize)
189
{
190
  nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize);
191
  if (nl_attr_len < 0)
192
    {
193
      log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len);
194
      return NULL;
195
    }
196
  return NLMSG_DATA(h);
197
}
198

    
199
static int
200
nl_parse_attrs(struct rtattr *a, struct rtattr **k, int ksize)
201
{
202
  int max = ksize / sizeof(struct rtattr *);
203
  bzero(k, ksize);
204
  while (RTA_OK(a, nl_attr_len))
205
    {
206
      if (a->rta_type < max)
207
        k[a->rta_type] = a;
208
      a = RTA_NEXT(a, nl_attr_len);
209
    }
210
  if (nl_attr_len)
211
    {
212
      log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len);
213
      return 0;
214
    }
215
  else
216
    return 1;
217
}
218

    
219
static void
220
nl_add_attr_u32(struct nlmsghdr *h, unsigned maxsize, int code, u32 data)
221
{
222
  unsigned len = RTA_LENGTH(4);
223
  struct rtattr *a;
224

    
225
  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
226
    bug("nl_add_attr32: packet buffer overflow");
227
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
228
  a->rta_type = code;
229
  a->rta_len = len;
230
  memcpy(RTA_DATA(a), &data, 4);
231
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
232
}
233

    
234
static void
235
nl_add_attr_ipa(struct nlmsghdr *h, unsigned maxsize, int code, ip_addr ipa)
236
{
237
  unsigned len = RTA_LENGTH(sizeof(ipa));
238
  struct rtattr *a;
239

    
240
  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
241
    bug("nl_add_attr_ipa: packet buffer overflow");
242
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
243
  a->rta_type = code;
244
  a->rta_len = len;
245
  ipa_hton(ipa);
246
  memcpy(RTA_DATA(a), &ipa, sizeof(ipa));
247
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
248
}
249

    
250
/*
251
 *        Scanning of interfaces
252
 */
253

    
254
static void
255
nl_parse_link(struct nlmsghdr *h, int scan)
256
{
257
  struct ifinfomsg *i;
258
  struct rtattr *a[IFLA_STATS+1];
259
  int new = h->nlmsg_type == RTM_NEWLINK;
260
  struct iface f;
261
  struct iface *ifi;
262
  char *name;
263
  u32 mtu;
264
  unsigned int fl;
265

    
266
  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), a, sizeof(a)))
267
    return;
268
  if (!a[IFLA_IFNAME] || RTA_PAYLOAD(a[IFLA_IFNAME]) < 2 ||
269
      !a[IFLA_MTU] || RTA_PAYLOAD(a[IFLA_MTU]) != 4)
270
    {
271
      log(L_ERR "nl_parse_link: Malformed message received");
272
      return;
273
    }
274
  name = RTA_DATA(a[IFLA_IFNAME]);
275
  memcpy(&mtu, RTA_DATA(a[IFLA_MTU]), sizeof(u32));
276

    
277
  ifi = if_find_by_index(i->ifi_index);
278
  if (!new)
279
    {
280
      DBG("KIF: IF%d(%s) goes down\n", i->ifi_index, name);
281
      if (ifi && !scan)
282
        {
283
          memcpy(&f, ifi, sizeof(struct iface));
284
          f.flags |= IF_ADMIN_DOWN;
285
          if_update(&f);
286
        }
287
    }
288
  else
289
    {
290
      DBG("KIF: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags);
291
      if (ifi)
292
        memcpy(&f, ifi, sizeof(f));
293
      else
294
        {
295
          bzero(&f, sizeof(f));
296
          f.index = i->ifi_index;
297
        }
298
      strncpy(f.name, RTA_DATA(a[IFLA_IFNAME]), sizeof(f.name)-1);
299
      f.mtu = mtu;
300
      f.flags = 0;
301
      fl = i->ifi_flags;
302
      if (fl & IFF_UP)
303
        f.flags |= IF_LINK_UP;
304
      if (fl & IFF_LOOPBACK)                /* Loopback */
305
        f.flags |= IF_MULTIACCESS | IF_LOOPBACK | IF_IGNORE;
306
      else if (fl & IFF_POINTOPOINT)        /* PtP */
307
        f.flags |= IF_MULTICAST;
308
      else if (fl & IFF_BROADCAST)        /* Broadcast */
309
        f.flags |= IF_MULTIACCESS | IF_BROADCAST | IF_MULTICAST;
310
      else
311
        f.flags |= IF_MULTIACCESS;        /* NBMA */
312
      if_update(&f);
313
    }
314
}
315

    
316
static void
317
nl_parse_addr(struct nlmsghdr *h)
318
{
319
  struct ifaddrmsg *i;
320
  struct rtattr *a[IFA_ANYCAST+1];
321
  int new = h->nlmsg_type == RTM_NEWADDR;
322
  struct ifa ifa;
323
  struct iface *ifi;
324
  int scope;
325

    
326
  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFA_RTA(i), a, sizeof(a)))
327
    return;
328
  if (i->ifa_family != BIRD_AF)
329
    return;
330
  if (!a[IFA_ADDRESS] || RTA_PAYLOAD(a[IFA_ADDRESS]) != sizeof(ip_addr)
331
#ifdef IPV6
332
      || a[IFA_LOCAL] && RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
333
#else
334
      || !a[IFA_LOCAL] || RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
335
      || (a[IFA_BROADCAST] && RTA_PAYLOAD(a[IFA_BROADCAST]) != sizeof(ip_addr))
336
#endif
337
      )
338
    {
339
      log(L_ERR "nl_parse_addr: Malformed message received");
340
      return;
341
    }
342

    
343
#ifdef IPV6
344
  if (i->ifa_scope == RT_SCOPE_LINK)
345
    return;
346
#endif
347

    
348
  ifi = if_find_by_index(i->ifa_index);
349
  if (!ifi)
350
    {
351
      log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index);
352
      return;
353
    }
354

    
355
  bzero(&ifa, sizeof(ifa));
356
  ifa.iface = ifi;
357
  if (i->ifa_flags & IFA_F_SECONDARY)
358
    ifa.flags |= IA_SECONDARY;
359

    
360
  /* IFA_LOCAL can be unset for IPv6 interfaces */
361
  memcpy(&ifa.ip, RTA_DATA(a[IFA_LOCAL] ? : a[IFA_ADDRESS]), sizeof(ifa.ip));
362
  ipa_ntoh(ifa.ip);
363
  ifa.pxlen = i->ifa_prefixlen;
364
  if (i->ifa_prefixlen > BITS_PER_IP_ADDRESS ||
365
      i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 1)
366
    {
367
      log(L_ERR "KIF: Invalid prefix length for interface %s: %d", ifi->name, i->ifa_prefixlen);
368
      new = 0;
369
    }
370
  if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS)
371
    {
372
      ifa.flags |= IA_UNNUMBERED;
373
      memcpy(&ifa.opposite, RTA_DATA(a[IFA_ADDRESS]), sizeof(ifa.opposite));
374
      ipa_ntoh(ifa.opposite);
375
      ifa.prefix = ifa.brd = ifa.opposite;
376
    }
377
  else
378
    {
379
      ip_addr netmask = ipa_mkmask(ifa.pxlen);
380
#ifndef IPV6
381
      if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 2)
382
        ifa.opposite = ipa_opposite(ifa.ip);
383
      if ((ifi->flags & IF_BROADCAST) && a[IFA_BROADCAST])
384
        {
385
          memcpy(&ifa.brd, RTA_DATA(a[IFA_BROADCAST]), sizeof(ifa.brd));
386
          ipa_ntoh(ifa.brd);
387
        }
388
      else
389
        ifa.brd = ipa_or(ifa.ip, ipa_not(netmask));
390
#endif
391
      ifa.prefix = ipa_and(ifa.ip, netmask);
392
    }
393

    
394
  scope = ipa_classify(ifa.ip);
395
  if (scope < 0)
396
    {
397
      log(L_ERR "KIF: Invalid interface address %I", ifa.ip);
398
      return;
399
    }
400
  ifa.scope = scope & IADDR_SCOPE_MASK;
401

    
402
  DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %I/%d, brd %I, opp %I\n",
403
      ifi->index, ifi->name,
404
      new ? "added" : "removed",
405
      ifa.ip, ifa.flags, ifa.prefix, ifa.pxlen, ifa.brd, ifa.opposite);
406
  if (new)
407
    ifa_update(&ifa);
408
  else
409
    ifa_delete(&ifa);
410
}
411

    
412
void
413
krt_if_scan(struct kif_proto *p)
414
{
415
  struct nlmsghdr *h;
416

    
417
  if_start_update();
418

    
419
  nl_request_dump(RTM_GETLINK);
420
  while (h = nl_get_scan())
421
    if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK)
422
      nl_parse_link(h, 1);
423
    else
424
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);
425

    
426
  nl_request_dump(RTM_GETADDR);
427
  while (h = nl_get_scan())
428
    if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR)
429
      nl_parse_addr(h);
430
    else
431
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);
432

    
433
  if_end_update();
434
}
435

    
436
/*
437
 *        Routes
438
 */
439

    
440
static struct krt_proto *nl_table_map[NL_NUM_TABLES];
441

    
442
int
443
krt_capable(rte *e)
444
{
445
  rta *a = e->attrs;
446

    
447
  if (a->cast != RTC_UNICAST
448
#if 0
449
      && a->cast != RTC_ANYCAST
450
#endif
451
      )
452
    return 0;
453
  if (a->source == RTS_DEVICE)        /* Kernel takes care of device routes itself */
454
    return 0;
455
  switch (a->dest)
456
    {
457
    case RTD_ROUTER:
458
    case RTD_DEVICE:
459
    case RTD_BLACKHOLE:
460
    case RTD_UNREACHABLE:
461
    case RTD_PROHIBIT:
462
      break;
463
    default:
464
      return 0;
465
    }
466
  return 1;
467
}
468

    
469
static void
470
nl_send_route(struct krt_proto *p, rte *e, int new)
471
{
472
  net *net = e->net;
473
  rta *a = e->attrs;
474
  struct {
475
    struct nlmsghdr h;
476
    struct rtmsg r;
477
    char buf[128];
478
  } r;
479

    
480
  DBG("nl_send_route(%I/%d,new=%d)\n", net->n.prefix, net->n.pxlen, new);
481

    
482
  bzero(&r.h, sizeof(r.h));
483
  bzero(&r.r, sizeof(r.r));
484
  r.h.nlmsg_type = new ? RTM_NEWROUTE : RTM_DELROUTE;
485
  r.h.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
486
  r.h.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | (new ? NLM_F_CREATE|NLM_F_REPLACE : 0);
487

    
488
  r.r.rtm_family = BIRD_AF;
489
  r.r.rtm_dst_len = net->n.pxlen;
490
  r.r.rtm_tos = 0;
491
  r.r.rtm_table = KRT_CF->scan.table_id;
492
  r.r.rtm_protocol = RTPROT_BIRD;
493
  r.r.rtm_scope = RT_SCOPE_UNIVERSE;
494
  nl_add_attr_ipa(&r.h, sizeof(r), RTA_DST, net->n.prefix);
495
  switch (a->dest)
496
    {
497
    case RTD_ROUTER:
498
      r.r.rtm_type = RTN_UNICAST;
499
      nl_add_attr_ipa(&r.h, sizeof(r), RTA_GATEWAY, a->gw);
500
      break;
501
    case RTD_DEVICE:
502
      r.r.rtm_type = RTN_UNICAST;
503
      nl_add_attr_u32(&r.h, sizeof(r), RTA_OIF, a->iface->index);
504
      break;
505
    case RTD_BLACKHOLE:
506
      r.r.rtm_type = RTN_BLACKHOLE;
507
      break;
508
    case RTD_UNREACHABLE:
509
      r.r.rtm_type = RTN_UNREACHABLE;
510
      break;
511
    case RTD_PROHIBIT:
512
      r.r.rtm_type = RTN_PROHIBIT;
513
      break;
514
    default:
515
      bug("krt_capable inconsistent with nl_send_route");
516
    }
517

    
518
  nl_exchange(&r.h);
519
}
520

    
521
void
522
krt_set_notify(struct krt_proto *p, net *n, rte *new, rte *old)
523
{
524
  if (old && new)
525
    {
526
      /*
527
       *  We should check whether priority and TOS is identical as well,
528
       *  but we don't use these and default value is always equal to default value. :-)
529
       */
530
      nl_send_route(p, new, 1);
531
    }
532
  else
533
    {
534
      if (old)
535
        {
536
          if (!old->attrs->iface || (old->attrs->iface->flags & IF_UP))
537
            nl_send_route(p, old, 0);
538
          /* else the kernel has already flushed it */
539
        }
540
      if (new)
541
        nl_send_route(p, new, 1);
542
    }
543
}
544

    
545
struct iface *
546
krt_temp_iface(struct krt_proto *p, unsigned index)
547
{
548
  struct iface *i, *j;
549

    
550
  WALK_LIST(i, p->scan.temp_ifs)
551
    if (i->index == index)
552
      return i;
553
  i = mb_allocz(p->p.pool, sizeof(struct iface));
554
  if (j = if_find_by_index(index))
555
    strcpy(i->name, j->name);
556
  else
557
    strcpy(i->name, "?");
558
  i->index = index;
559
  add_tail(&p->scan.temp_ifs, &i->n);
560
  return i;
561
}
562

    
563
static void
564
nl_parse_route(struct nlmsghdr *h, int scan)
565
{
566
  struct krt_proto *p;
567
  struct rtmsg *i;
568
  struct rtattr *a[RTA_CACHEINFO+1];
569
  int new = h->nlmsg_type == RTM_NEWROUTE;
570
  ip_addr dst;
571
  rta ra;
572
  rte *e;
573
  net *net;
574
  u32 oif;
575
  int src;
576

    
577
  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(RTM_RTA(i), a, sizeof(a)))
578
    return;
579
  if (i->rtm_family != BIRD_AF)
580
    return;
581
  if ((a[RTA_DST] && RTA_PAYLOAD(a[RTA_DST]) != sizeof(ip_addr)) ||
582
      (a[RTA_OIF] && RTA_PAYLOAD(a[RTA_OIF]) != 4) ||
583
      (a[RTA_PRIORITY] && RTA_PAYLOAD(a[RTA_PRIORITY]) != 4) ||
584
#ifdef IPV6
585
      (a[RTA_IIF] && RTA_PAYLOAD(a[RTA_IIF]) != 4) ||
586
#endif
587
      (a[RTA_GATEWAY] && RTA_PAYLOAD(a[RTA_GATEWAY]) != sizeof(ip_addr)))
588
    {
589
      log(L_ERR "nl_parse_route: Malformed message received");
590
      return;
591
    }
592

    
593
  p = nl_table_map[i->rtm_table];        /* Do we know this table? */
594
  if (!p)
595
    return;
596

    
597
#ifdef IPV6
598
  if (a[RTA_IIF])
599
    {
600
      DBG("KRT: Ignoring route with IIF set\n");
601
      return;
602
    }
603
#else
604
  if (i->rtm_tos != 0)                        /* We don't support TOS */
605
    {
606
      DBG("KRT: Ignoring route with TOS %02x\n", i->rtm_tos);
607
      return;
608
    }
609
#endif
610

    
611
  if (scan && !new)
612
    {
613
      DBG("KRT: Ignoring route deletion\n");
614
      return;
615
    }
616

    
617
  if (a[RTA_DST])
618
    {
619
      memcpy(&dst, RTA_DATA(a[RTA_DST]), sizeof(dst));
620
      ipa_ntoh(dst);
621
    }
622
  else
623
    dst = IPA_NONE;
624
  if (a[RTA_OIF])
625
    memcpy(&oif, RTA_DATA(a[RTA_OIF]), sizeof(oif));
626
  else
627
    oif = ~0;
628

    
629
  DBG("Got %I/%d, type=%d, oif=%d, table=%d, proto=%s\n", dst, i->rtm_dst_len, i->rtm_type, oif, i->rtm_table, p->p.name);
630

    
631
  switch (i->rtm_protocol)
632
    {
633
    case RTPROT_REDIRECT:
634
      src = KRT_SRC_REDIRECT;
635
      break;
636
    case RTPROT_KERNEL:
637
      DBG("Route originated in kernel, ignoring\n");
638
      return;
639
    case RTPROT_BIRD:
640
      if (!scan)
641
        {
642
          DBG("Echo of our own route, ignoring\n");
643
          return;
644
        }
645
      src = KRT_SRC_BIRD;
646
      break;
647
    default:
648
      src = KRT_SRC_ALIEN;
649
    }
650

    
651
  net = net_get(p->p.table, dst, i->rtm_dst_len);
652
  ra.proto = &p->p;
653
  ra.source = RTS_INHERIT;
654
  ra.scope = SCOPE_UNIVERSE;
655
  ra.cast = RTC_UNICAST;
656
  ra.flags = ra.aflags = 0;
657
  ra.from = IPA_NONE;
658
  ra.gw = IPA_NONE;
659
  ra.iface = NULL;
660
  ra.eattrs = NULL;
661

    
662
  switch (i->rtm_type)
663
    {
664
    case RTN_UNICAST:
665
      if (oif == ~0U)
666
        {
667
          log(L_ERR "KRT: Mysterious route with no OIF (%I/%d)", net->n.prefix, net->n.pxlen);
668
          return;
669
        }
670
      if (a[RTA_GATEWAY])
671
        {
672
          neighbor *ng;
673
          ra.dest = RTD_ROUTER;
674
          memcpy(&ra.gw, RTA_DATA(a[RTA_GATEWAY]), sizeof(ra.gw));
675
          ipa_ntoh(ra.gw);
676
          ng = neigh_find(&p->p, &ra.gw, 0);
677
          if (ng && ng->scope)
678
            ra.iface = ng->iface;
679
          else
680
            /* FIXME: Remove this warning? Handle it somehow... */
681
            log(L_WARN "Kernel told us to use non-neighbor %I for %I/%d", ra.gw, net->n.prefix, net->n.pxlen);
682
        }
683
      else
684
        {
685
          ra.dest = RTD_DEVICE;
686
          ra.iface = krt_temp_iface(p, oif);
687
        }
688
      break;
689
    case RTN_BLACKHOLE:
690
      ra.dest = RTD_BLACKHOLE;
691
      break;
692
    case RTN_UNREACHABLE:
693
      ra.dest = RTD_UNREACHABLE;
694
      break;
695
    case RTN_PROHIBIT:
696
      ra.dest = RTD_PROHIBIT;
697
      break;
698
    /* FIXME: What about RTN_THROW? */
699
    default:
700
      DBG("KRT: Ignoring route with type=%d\n", i->rtm_type);
701
      return;
702
    }
703

    
704
  if (i->rtm_scope != RT_SCOPE_UNIVERSE)
705
    {
706
      DBG("KRT: Ignoring route with scope=%d\n", i->rtm_scope);
707
      return;
708
    }
709

    
710
  e = rte_get_temp(&ra);
711
  e->net = net;
712
  e->u.krt.src = src;
713
  e->u.krt.proto = i->rtm_protocol;
714
  e->u.krt.type = i->rtm_type;
715
  if (a[RTA_PRIORITY])
716
    memcpy(&e->u.krt.metric, RTA_DATA(a[RTA_PRIORITY]), sizeof(e->u.krt.metric));
717
  else
718
    e->u.krt.metric = 0;
719
  if (scan)
720
    krt_got_route(p, e);
721
  else
722
    krt_got_route_async(p, e, new);
723
}
724

    
725
void
726
krt_scan_fire(struct krt_proto *p)        /* CONFIG_ALL_TABLES_AT_ONCE => p is NULL */
727
{
728
  struct nlmsghdr *h;
729

    
730
  nl_request_dump(RTM_GETROUTE);
731
  while (h = nl_get_scan())
732
    if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)
733
      nl_parse_route(h, 1);
734
    else
735
      log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type);
736
}
737

    
738
/*
739
 *        Asynchronous Netlink interface
740
 */
741

    
742
static sock *nl_async_sk;                /* BIRD socket for asynchronous notifications */
743
static byte *nl_async_rx_buffer;        /* Receive buffer */
744

    
745
static void
746
nl_async_msg(struct nlmsghdr *h)
747
{
748
  switch (h->nlmsg_type)
749
    {
750
    case RTM_NEWROUTE:
751
    case RTM_DELROUTE:
752
      DBG("KRT: Received async route notification (%d)\n", h->nlmsg_type);
753
      nl_parse_route(h, 0);
754
      break;
755
    case RTM_NEWLINK:
756
    case RTM_DELLINK:
757
      DBG("KRT: Received async link notification (%d)\n", h->nlmsg_type);
758
      nl_parse_link(h, 0);
759
      break;
760
    case RTM_NEWADDR:
761
    case RTM_DELADDR:
762
      DBG("KRT: Received async address notification (%d)\n", h->nlmsg_type);
763
      nl_parse_addr(h);
764
      break;
765
    default:
766
      DBG("KRT: Received unknown async notification (%d)\n", h->nlmsg_type);
767
    }
768
}
769

    
770
static int
771
nl_async_hook(sock *sk, int size)
772
{
773
  struct iovec iov = { nl_async_rx_buffer, NL_RX_SIZE };
774
  struct sockaddr_nl sa;
775
  struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
776
  struct nlmsghdr *h;
777
  int x;
778
  unsigned int len;
779

    
780
  nl_last_hdr = NULL;                /* Discard packets accidentally remaining in the rxbuf */
781
  x = recvmsg(sk->fd, &m, 0);
782
  if (x < 0)
783
    {
784
      if (errno == ENOBUFS)
785
        {
786
          /*
787
           *  Netlink reports some packets have been thrown away.
788
           *  One day we might react to it by asking for route table
789
           *  scan in near future.
790
           */
791
          return 1;        /* More data are likely to be ready */
792
        }
793
      else if (errno != EWOULDBLOCK)
794
        log(L_ERR "Netlink recvmsg: %m");
795
      return 0;
796
    }
797
  if (sa.nl_pid)                /* It isn't from the kernel */
798
    {
799
      DBG("Non-kernel packet\n");
800
      return 1;
801
    }
802
  h = (void *) nl_async_rx_buffer;
803
  len = x;
804
  if (m.msg_flags & MSG_TRUNC)
805
    {
806
      log(L_WARN "Netlink got truncated asynchronous message");
807
      return 1;
808
    }
809
  while (NLMSG_OK(h, len))
810
    {
811
      nl_async_msg(h);
812
      h = NLMSG_NEXT(h, len);
813
    }
814
  if (len)
815
    log(L_WARN "nl_async_hook: Found packet remnant of size %d", len);
816
  return 1;
817
}
818

    
819
static void
820
nl_open_async(void)
821
{
822
  sock *sk;
823
  struct sockaddr_nl sa;
824
  int fd;
825
  static int nl_open_tried = 0;
826

    
827
  if (nl_open_tried)
828
    return;
829
  nl_open_tried = 1;
830

    
831
  DBG("KRT: Opening async netlink socket\n");
832

    
833
  fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
834
  if (fd < 0)
835
    {
836
      log(L_ERR "Unable to open asynchronous rtnetlink socket: %m");
837
      return;
838
    }
839

    
840
  bzero(&sa, sizeof(sa));
841
  sa.nl_family = AF_NETLINK;
842
#ifdef IPV6
843
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV6_IFADDR | RTMGRP_IPV6_ROUTE;
844
#else
845
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE;
846
#endif
847
  if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0)
848
    {
849
      log(L_ERR "Unable to bind asynchronous rtnetlink socket: %m");
850
      return;
851
    }
852

    
853
  sk = nl_async_sk = sk_new(krt_pool);
854
  sk->type = SK_MAGIC;
855
  sk->rx_hook = nl_async_hook;
856
  sk->fd = fd;
857
  if (sk_open(sk))
858
    bug("Netlink: sk_open failed");
859

    
860
  if (!nl_async_rx_buffer)
861
    nl_async_rx_buffer = xmalloc(NL_RX_SIZE);
862
}
863

    
864
/*
865
 *        Interface to the UNIX krt module
866
 */
867

    
868
static u8 nl_cf_table[(NL_NUM_TABLES+7) / 8];
869

    
870
void
871
krt_scan_preconfig(struct config *c)
872
{
873
  bzero(&nl_cf_table, sizeof(nl_cf_table));
874
}
875

    
876
void
877
krt_scan_postconfig(struct krt_config *x)
878
{
879
  int id = x->scan.table_id;
880

    
881
  if (nl_cf_table[id/8] & (1 << (id%8)))
882
    cf_error("Multiple kernel syncers defined for table #%d", id);
883
  nl_cf_table[id/8] |= (1 << (id%8));
884
}
885

    
886
void
887
krt_scan_construct(struct krt_config *x)
888
{
889
#ifndef IPV6
890
  x->scan.table_id = RT_TABLE_MAIN;
891
#endif
892
}
893

    
894
void
895
krt_scan_start(struct krt_proto *p, int first)
896
{
897
  init_list(&p->scan.temp_ifs);
898
  nl_table_map[KRT_CF->scan.table_id] = p;
899
  if (first)
900
    {
901
      nl_open();
902
      nl_open_async();
903
    }
904
}
905

    
906
void
907
krt_scan_shutdown(struct krt_proto *p, int last)
908
{
909
}
910

    
911
void
912
krt_if_start(struct kif_proto *p)
913
{
914
  nl_open();
915
  nl_open_async();
916
}