netlink.c 40.3 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 *	BIRD -- Linux Netlink Interface
 *
 *	(c) 1999--2000 Martin Mares <mj@ucw.cz>
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

#include <stdio.h>
10
#include <unistd.h>
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
#include <fcntl.h>
#include <sys/socket.h>
#include <sys/uio.h>
#include <errno.h>

#undef LOCAL_DEBUG

#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"
#include "lib/socket.h"
#include "lib/string.h"
27
#include "lib/hash.h"
28 29 30 31 32 33 34
#include "conf/conf.h"

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

35

36 37 38 39
#ifndef MSG_TRUNC			/* Hack: Several versions of glibc miss this one :( */
#define MSG_TRUNC 0x20
#endif

40 41 42 43
#ifndef IFA_FLAGS
#define IFA_FLAGS 8
#endif

44 45 46 47
#ifndef IFF_LOWER_UP
#define IFF_LOWER_UP 0x10000
#endif

48 49 50 51 52
#ifndef RTA_TABLE
#define RTA_TABLE  15
#endif


53 54 55 56 57 58 59 60 61
#ifdef IPV6
#define krt_ecmp6(X) 1
#else
#define krt_ecmp6(X) 0
#endif

/*
 * Structure nl_parse_state keeps state of received route processing. Ideally,
 * we could just independently parse received Netlink messages and immediately
62 63 64 65
 * propagate received routes to the rest of BIRD, but older Linux kernel (before
 * version 4.11) represents and announces IPv6 ECMP routes not as one route with
 * multiple next hops (like RTA_MULTIPATH in IPv4 ECMP), but as a sequence of
 * routes with the same prefix. More recent kernels work as with IPv4.
66 67 68 69 70 71
 *
 * Therefore, BIRD keeps currently processed route in nl_parse_state structure
 * and postpones its propagation until we expect it to be final; i.e., when
 * non-matching route is received or when the scan ends. When another matching
 * route is received, it is merged with the already processed route to form an
 * ECMP route. Note that merging is done only for IPv6 (merge == 1), but the
72 73
 * postponing is done in both cases (for simplicity). All IPv4 routes or IPv6
 * routes with RTA_MULTIPATH set are just considered non-matching.
74 75 76 77
 *
 * This is ignored for asynchronous notifications (every notification is handled
 * as a separate route). It is not an issue for our routes, as we ignore such
 * notifications anyways. But importing alien IPv6 ECMP routes does not work
78 79 80 81
 * properly with older kernels.
 *
 * Whatever the kernel version is, IPv6 ECMP routes are sent as multiple routes
 * for the same prefix.
82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
 */

struct nl_parse_state
{
  struct linpool *pool;
  int scan;
  int merge;

  net *net;
  rta *attrs;
  struct krt_proto *proto;
  s8 new;
  s8 krt_src;
  u8 krt_type;
  u8 krt_proto;
  u32 krt_metric;
};

100 101 102 103 104 105 106 107 108 109
/*
 *	Synchronous Netlink interface
 */

struct nl_sock
{
  int fd;
  u32 seq;
  byte *rx_buffer;			/* Receive buffer */
  struct nlmsghdr *last_hdr;		/* Recently received packet */
Pavel Tvrdík's avatar
Pavel Tvrdík committed
110
  uint last_size;
111 112 113 114
};

#define NL_RX_SIZE 8192

115 116 117 118 119 120 121
#define NL_OP_DELETE	0
#define NL_OP_ADD	(NLM_F_CREATE|NLM_F_EXCL)
#define NL_OP_REPLACE	(NLM_F_CREATE|NLM_F_REPLACE)
#define NL_OP_APPEND	(NLM_F_CREATE|NLM_F_APPEND)

static linpool *nl_linpool;

122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
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 */

static void
nl_open_sock(struct nl_sock *nl)
{
  if (nl->fd < 0)
    {
      nl->fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
      if (nl->fd < 0)
	die("Unable to open rtnetlink socket: %m");
      nl->seq = now;
      nl->rx_buffer = xmalloc(NL_RX_SIZE);
      nl->last_hdr = NULL;
      nl->last_size = 0;
    }
}

static void
nl_open(void)
{
  nl_open_sock(&nl_scan);
  nl_open_sock(&nl_req);
}

static void
nl_send(struct nl_sock *nl, struct nlmsghdr *nh)
{
  struct sockaddr_nl sa;

  memset(&sa, 0, sizeof(sa));
  sa.nl_family = AF_NETLINK;
  nh->nlmsg_pid = 0;
  nh->nlmsg_seq = ++(nl->seq);
  if (sendto(nl->fd, nh, nh->nlmsg_len, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0)
    die("rtnetlink sendto: %m");
  nl->last_hdr = NULL;
}

static void
162
nl_request_dump(int af, int cmd)
163 164 165 166
{
  struct {
    struct nlmsghdr nh;
    struct rtgenmsg g;
167 168 169 170 171 172
  } req = {
    .nh.nlmsg_type = cmd,
    .nh.nlmsg_len = sizeof(req),
    .nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP,
    .g.rtgen_family = af
  };
173 174 175 176 177 178 179 180 181 182 183 184
  nl_send(&nl_scan, &req.nh);
}

static struct nlmsghdr *
nl_get_reply(struct nl_sock *nl)
{
  for(;;)
    {
      if (!nl->last_hdr)
	{
	  struct iovec iov = { nl->rx_buffer, NL_RX_SIZE };
	  struct sockaddr_nl sa;
185 186 187 188 189 190
	  struct msghdr m = {
	    .msg_name = &sa,
	    .msg_namelen = sizeof(sa),
	    .msg_iov = &iov,
	    .msg_iovlen = 1,
	  };
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
	  int x = recvmsg(nl->fd, &m, 0);
	  if (x < 0)
	    die("nl_get_reply: %m");
	  if (sa.nl_pid)		/* It isn't from the kernel */
	    {
	      DBG("Non-kernel packet\n");
	      continue;
	    }
	  nl->last_size = x;
	  nl->last_hdr = (void *) nl->rx_buffer;
	  if (m.msg_flags & MSG_TRUNC)
	    bug("nl_get_reply: got truncated reply which should be impossible");
	}
      if (NLMSG_OK(nl->last_hdr, nl->last_size))
	{
	  struct nlmsghdr *h = nl->last_hdr;
	  nl->last_hdr = NLMSG_NEXT(h, nl->last_size);
	  if (h->nlmsg_seq != nl->seq)
	    {
	      log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)",
		  h->nlmsg_seq, nl->seq);
	      continue;
	    }
	  return h;
	}
      if (nl->last_size)
	log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl->last_size);
      nl->last_hdr = NULL;
    }
}

222
static struct tbf rl_netlink_err = TBF_DEFAULT_LOG_LIMITS;
223 224

static int
225
nl_error(struct nlmsghdr *h, int ignore_esrch)
226 227 228 229 230 231 232 233 234 235 236
{
  struct nlmsgerr *e;
  int ec;

  if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr)))
    {
      log(L_WARN "Netlink: Truncated error message received");
      return ENOBUFS;
    }
  e = (struct nlmsgerr *) NLMSG_DATA(h);
  ec = -e->error;
237
  if (ec && !(ignore_esrch && (ec == ESRCH)))
238 239 240 241 242 243 244 245 246 247 248 249 250
    log_rl(&rl_netlink_err, L_WARN "Netlink: %s", strerror(ec));
  return ec;
}

static struct nlmsghdr *
nl_get_scan(void)
{
  struct nlmsghdr *h = nl_get_reply(&nl_scan);

  if (h->nlmsg_type == NLMSG_DONE)
    return NULL;
  if (h->nlmsg_type == NLMSG_ERROR)
    {
251
      nl_error(h, 0);
252 253 254 255 256 257
      return NULL;
    }
  return h;
}

static int
258
nl_exchange(struct nlmsghdr *pkt, int ignore_esrch)
259 260 261 262 263 264 265 266 267 268 269
{
  struct nlmsghdr *h;

  nl_send(&nl_req, pkt);
  for(;;)
    {
      h = nl_get_reply(&nl_req);
      if (h->nlmsg_type == NLMSG_ERROR)
	break;
      log(L_WARN "nl_exchange: Unexpected reply received");
    }
270
  return nl_error(h, ignore_esrch) ? -1 : 0;
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
}

/*
 *	Netlink attributes
 */

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);
}

291 292 293 294 295 296 297 298 299 300 301 302
struct nl_want_attrs {
  u8 defined:1;
  u8 checksize:1;
  u8 size;
};


#define BIRD_IFLA_MAX (IFLA_WIRELESS+1)

static struct nl_want_attrs ifla_attr_want[BIRD_IFLA_MAX] = {
  [IFLA_IFNAME]	  = { 1, 0, 0 },
  [IFLA_MTU]	  = { 1, 1, sizeof(u32) },
Ondřej Zajíček's avatar
Ondřej Zajíček committed
303
  [IFLA_MASTER]	  = { 1, 1, sizeof(u32) },
304 305 306 307
  [IFLA_WIRELESS] = { 1, 0, 0 },
};


308
#define BIRD_IFA_MAX  (IFA_FLAGS+1)
309 310 311 312 313 314

#ifndef IPV6
static struct nl_want_attrs ifa_attr_want4[BIRD_IFA_MAX] = {
  [IFA_ADDRESS]	  = { 1, 1, sizeof(ip4_addr) },
  [IFA_LOCAL]	  = { 1, 1, sizeof(ip4_addr) },
  [IFA_BROADCAST] = { 1, 1, sizeof(ip4_addr) },
315
  [IFA_FLAGS]	  = { 1, 1, sizeof(u32) },
316 317 318 319 320
};
#else
static struct nl_want_attrs ifa_attr_want6[BIRD_IFA_MAX] = {
  [IFA_ADDRESS]	  = { 1, 1, sizeof(ip6_addr) },
  [IFA_LOCAL]	  = { 1, 1, sizeof(ip6_addr) },
321
  [IFA_FLAGS]	  = { 1, 1, sizeof(u32) },
322 323 324 325 326 327
};
#endif


#define BIRD_RTA_MAX  (RTA_TABLE+1)

328
#ifndef IPV6
329 330 331
static struct nl_want_attrs mpnh_attr_want4[BIRD_RTA_MAX] = {
  [RTA_GATEWAY]	  = { 1, 1, sizeof(ip4_addr) },
};
332 333 334 335 336
#else
static struct nl_want_attrs mpnh_attr_want6[BIRD_RTA_MAX] = {
  [RTA_GATEWAY]	  = { 1, 1, sizeof(ip6_addr) },
};
#endif
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358

#ifndef IPV6
static struct nl_want_attrs rtm_attr_want4[BIRD_RTA_MAX] = {
  [RTA_DST]	  = { 1, 1, sizeof(ip4_addr) },
  [RTA_OIF]	  = { 1, 1, sizeof(u32) },
  [RTA_GATEWAY]	  = { 1, 1, sizeof(ip4_addr) },
  [RTA_PRIORITY]  = { 1, 1, sizeof(u32) },
  [RTA_PREFSRC]	  = { 1, 1, sizeof(ip4_addr) },
  [RTA_METRICS]	  = { 1, 0, 0 },
  [RTA_MULTIPATH] = { 1, 0, 0 },
  [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
  [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
};
#else
static struct nl_want_attrs rtm_attr_want6[BIRD_RTA_MAX] = {
  [RTA_DST]	  = { 1, 1, sizeof(ip6_addr) },
  [RTA_IIF]	  = { 1, 1, sizeof(u32) },
  [RTA_OIF]	  = { 1, 1, sizeof(u32) },
  [RTA_GATEWAY]	  = { 1, 1, sizeof(ip6_addr) },
  [RTA_PRIORITY]  = { 1, 1, sizeof(u32) },
  [RTA_PREFSRC]	  = { 1, 1, sizeof(ip6_addr) },
  [RTA_METRICS]	  = { 1, 0, 0 },
359
  [RTA_MULTIPATH] = { 1, 0, 0 },
360 361 362 363 364 365
  [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
  [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
};
#endif


366
static int
367
nl_parse_attrs(struct rtattr *a, struct nl_want_attrs *want, struct rtattr **k, int ksize)
368 369 370
{
  int max = ksize / sizeof(struct rtattr *);
  bzero(k, ksize);
371 372

  for ( ; RTA_OK(a, nl_attr_len); a = RTA_NEXT(a, nl_attr_len))
373
    {
374 375 376 377 378 379 380 381 382 383
      if ((a->rta_type >= max) || !want[a->rta_type].defined)
	continue;

      if (want[a->rta_type].checksize && (RTA_PAYLOAD(a) != want[a->rta_type].size))
	{
	  log(L_ERR "nl_parse_attrs: Malformed message received");
	  return 0;
	}

      k[a->rta_type] = a;
384
    }
385

386 387 388 389 390
  if (nl_attr_len)
    {
      log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len);
      return 0;
    }
391 392

  return 1;
393 394
}

395
static inline u32 rta_get_u32(struct rtattr *a)
Ondřej Zajíček's avatar
Ondřej Zajíček committed
396 397 398 399 400 401 402 403 404
{ return *(u32 *) RTA_DATA(a); }

static inline ip4_addr rta_get_ip4(struct rtattr *a)
{ return ip4_ntoh(*(ip4_addr *) RTA_DATA(a)); }

static inline ip6_addr rta_get_ip6(struct rtattr *a)
{ return ip6_ntoh(*(ip6_addr *) RTA_DATA(a)); }


405 406
struct rtattr *
nl_add_attr(struct nlmsghdr *h, uint bufsize, uint code, const void *data, uint dlen)
407
{
408 409
  uint pos = NLMSG_ALIGN(h->nlmsg_len);
  uint len = RTA_LENGTH(dlen);
410 411 412

  if (pos + len > bufsize)
    bug("nl_add_attr: packet buffer overflow");
413 414

  struct rtattr *a = (struct rtattr *)((char *)h + pos);
415 416 417
  a->rta_type = code;
  a->rta_len = len;
  h->nlmsg_len = pos + len;
418 419 420 421 422

  if (dlen > 0)
    memcpy(RTA_DATA(a), data, dlen);

  return a;
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437
}

static inline void
nl_add_attr_u32(struct nlmsghdr *h, unsigned bufsize, int code, u32 data)
{
  nl_add_attr(h, bufsize, code, &data, 4);
}

static inline void
nl_add_attr_ipa(struct nlmsghdr *h, unsigned bufsize, int code, ip_addr ipa)
{
  ipa_hton(ipa);
  nl_add_attr(h, bufsize, code, &ipa, sizeof(ipa));
}

438 439 440 441 442
static inline struct rtattr *
nl_open_attr(struct nlmsghdr *h, uint bufsize, uint code)
{
  return nl_add_attr(h, bufsize, code, NULL, 0);
}
443 444

static inline void
445
nl_close_attr(struct nlmsghdr *h, struct rtattr *a)
446
{
447
  a->rta_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)a;
448 449
}

450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
static inline struct rtnexthop *
nl_open_nexthop(struct nlmsghdr *h, uint bufsize)
{
  uint pos = NLMSG_ALIGN(h->nlmsg_len);
  uint len = RTNH_LENGTH(0);

  if (pos + len > bufsize)
    bug("nl_open_nexthop: packet buffer overflow");

  h->nlmsg_len = pos + len;

  return (void *)h + pos;
}

static inline void
nl_close_nexthop(struct nlmsghdr *h, struct rtnexthop *nh)
{
  nh->rtnh_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)nh;
}
469 470 471 472

static void
nl_add_multipath(struct nlmsghdr *h, unsigned bufsize, struct mpnh *nh)
{
473 474
  struct rtattr *a = nl_open_attr(h, bufsize, RTA_MULTIPATH);

475
  for (; nh; nh = nh->next)
476 477
  {
    struct rtnexthop *rtnh = nl_open_nexthop(h, bufsize);
478

479 480 481
    rtnh->rtnh_flags = 0;
    rtnh->rtnh_hops = nh->weight;
    rtnh->rtnh_ifindex = nh->iface->index;
482

483
    nl_add_attr_ipa(h, bufsize, RTA_GATEWAY, nh->gw);
484

485 486 487 488 489
    nl_close_nexthop(h, rtnh);
  }

  nl_close_attr(h, a);
}
490 491

static struct mpnh *
492
nl_parse_multipath(struct krt_proto *p, struct rtattr *ra, int af)
493 494 495 496 497 498
{
  /* Temporary buffer for multicast nexthops */
  static struct mpnh *nh_buffer;
  static int nh_buf_size;	/* in number of structures */
  static int nh_buf_used;

499
  struct rtattr *a[BIRD_RTA_MAX];
500 501
  struct rtnexthop *nh = RTA_DATA(ra);
  struct mpnh *rv, *first, **last;
502
  unsigned len = RTA_PAYLOAD(ra);
503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529

  first = NULL;
  last = &first;
  nh_buf_used = 0;

  while (len)
    {
      /* Use RTNH_OK(nh,len) ?? */
      if ((len < sizeof(*nh)) || (len < nh->rtnh_len))
	return NULL;

      if (nh_buf_used == nh_buf_size)
      {
	nh_buf_size = nh_buf_size ? (nh_buf_size * 2) : 4;
	nh_buffer = xrealloc(nh_buffer, nh_buf_size * sizeof(struct mpnh));
      }
      *last = rv = nh_buffer + nh_buf_used++;
      rv->next = NULL;
      last = &(rv->next);

      rv->weight = nh->rtnh_hops;
      rv->iface = if_find_by_index(nh->rtnh_ifindex);
      if (!rv->iface)
	return NULL;

      /* Nonexistent RTNH_PAYLOAD ?? */
      nl_attr_len = nh->rtnh_len - RTNH_LENGTH(0);
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
      switch (af)
        {
#ifndef IPV6
	case AF_INET:
	  if (!nl_parse_attrs(RTNH_DATA(nh), mpnh_attr_want4, a, sizeof(a)))
	    return NULL;
	  break;
#else
	case AF_INET6:
	  if (!nl_parse_attrs(RTNH_DATA(nh), mpnh_attr_want6, a, sizeof(a)))
	    return NULL;
	  break;
#endif
	default:
	  return NULL;
	}

547 548
      if (a[RTA_GATEWAY])
	{
549
	  memcpy(&rv->gw, RTA_DATA(a[RTA_GATEWAY]), sizeof(rv->gw));
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
	  ipa_ntoh(rv->gw);

	  neighbor *ng = neigh_find2(&p->p, &rv->gw, rv->iface,
				     (nh->rtnh_flags & RTNH_F_ONLINK) ? NEF_ONLINK : 0);
	  if (!ng || (ng->scope == SCOPE_HOST))
	    return NULL;
	}
      else
	return NULL;

      len -= NLMSG_ALIGN(nh->rtnh_len);
      nh = RTNH_NEXT(nh);
    }

  return first;
}

567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
static void
nl_add_metrics(struct nlmsghdr *h, uint bufsize, u32 *metrics, int max)
{
  struct rtattr *a = nl_open_attr(h, bufsize, RTA_METRICS);
  int t;

  for (t = 1; t < max; t++)
    if (metrics[0] & (1 << t))
      nl_add_attr_u32(h, bufsize, t, metrics[t]);

  nl_close_attr(h, a);
}

static int
nl_parse_metrics(struct rtattr *hdr, u32 *metrics, int max)
{
  struct rtattr *a = RTA_DATA(hdr);
  int len = RTA_PAYLOAD(hdr);

  metrics[0] = 0;
  for (; RTA_OK(a, len); a = RTA_NEXT(a, len))
  {
    if (a->rta_type == RTA_UNSPEC)
      continue;

    if (a->rta_type >= max)
      continue;

    if (RTA_PAYLOAD(a) != 4)
      return -1;

    metrics[0] |= 1 << a->rta_type;
Ondřej Zajíček's avatar
Ondřej Zajíček committed
599
    metrics[a->rta_type] = rta_get_u32(a);
600 601 602 603 604 605 606 607
  }

  if (len > 0)
    return -1;

  return 0;
}

608 609 610 611 612 613 614 615 616

/*
 *	Scanning of interfaces
 */

static void
nl_parse_link(struct nlmsghdr *h, int scan)
{
  struct ifinfomsg *i;
617
  struct rtattr *a[BIRD_IFLA_MAX];
618 619 620 621
  int new = h->nlmsg_type == RTM_NEWLINK;
  struct iface f = {};
  struct iface *ifi;
  char *name;
Ondřej Zajíček's avatar
Ondřej Zajíček committed
622
  u32 mtu, master = 0;
Pavel Tvrdík's avatar
Pavel Tvrdík committed
623
  uint fl;
624

625
  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), ifla_attr_want, a, sizeof(a)))
626
    return;
627
  if (!a[IFLA_IFNAME] || (RTA_PAYLOAD(a[IFLA_IFNAME]) < 2) || !a[IFLA_MTU])
628
    {
629 630 631 632 633 634 635 636 637 638
      /*
       * IFLA_IFNAME and IFLA_MTU are required, in fact, but there may also come
       * a message with IFLA_WIRELESS set, where (e.g.) no IFLA_IFNAME exists.
       * We simply ignore all such messages with IFLA_WIRELESS without notice.
       */

      if (a[IFLA_WIRELESS])
	return;

      log(L_ERR "KIF: Malformed message received");
639 640
      return;
    }
641

642
  name = RTA_DATA(a[IFLA_IFNAME]);
Ondřej Zajíček's avatar
Ondřej Zajíček committed
643
  mtu = rta_get_u32(a[IFLA_MTU]);
644

Ondřej Zajíček's avatar
Ondřej Zajíček committed
645 646 647
  if (a[IFLA_MASTER])
    master = rta_get_u32(a[IFLA_MASTER]);

648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
  ifi = if_find_by_index(i->ifi_index);
  if (!new)
    {
      DBG("KIF: IF%d(%s) goes down\n", i->ifi_index, name);
      if (!ifi)
	return;

      if_delete(ifi);
    }
  else
    {
      DBG("KIF: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags);
      if (ifi && strncmp(ifi->name, name, sizeof(ifi->name)-1))
	if_delete(ifi);

      strncpy(f.name, name, sizeof(f.name)-1);
      f.index = i->ifi_index;
      f.mtu = mtu;

Ondřej Zajíček's avatar
Ondřej Zajíček committed
667 668 669
      f.master_index = master;
      f.master = if_find_by_index(master);

670 671 672 673 674 675 676 677 678 679 680 681 682
      fl = i->ifi_flags;
      if (fl & IFF_UP)
	f.flags |= IF_ADMIN_UP;
      if (fl & IFF_LOWER_UP)
	f.flags |= IF_LINK_UP;
      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 */
683

684 685 686
      if (fl & IFF_MULTICAST)
	f.flags |= IF_MULTICAST;

687 688 689 690
      ifi = if_update(&f);

      if (!scan)
	if_end_partial_update(ifi);
691 692 693 694
    }
}

static void
695
nl_parse_addr(struct nlmsghdr *h, int scan)
696 697
{
  struct ifaddrmsg *i;
698
  struct rtattr *a[BIRD_IFA_MAX];
699 700 701 702
  int new = h->nlmsg_type == RTM_NEWADDR;
  struct ifa ifa;
  struct iface *ifi;
  int scope;
703
  u32 ifa_flags;
704

705
  if (!(i = nl_checkin(h, sizeof(*i))))
706
    return;
707 708 709 710 711 712 713 714 715 716 717 718 719

  switch (i->ifa_family)
    {
#ifndef IPV6
      case AF_INET:
	if (!nl_parse_attrs(IFA_RTA(i), ifa_attr_want4, a, sizeof(a)))
	  return;
	if (!a[IFA_LOCAL])
	  {
	    log(L_ERR "KIF: Malformed message received (missing IFA_LOCAL)");
	    return;
	  }
	break;
720
#else
721 722 723 724
      case AF_INET6:
	if (!nl_parse_attrs(IFA_RTA(i), ifa_attr_want6, a, sizeof(a)))
	  return;
	break;
725
#endif
726 727 728 729 730
      default:
	return;
    }

  if (!a[IFA_ADDRESS])
731
    {
732
      log(L_ERR "KIF: Malformed message received (missing IFA_ADDRESS)");
733 734 735
      return;
    }

736 737 738 739 740
  if (a[IFA_FLAGS])
    ifa_flags = rta_get_u32(a[IFA_FLAGS]);
  else
    ifa_flags = i->ifa_flags;

741 742 743 744 745 746 747 748 749
  ifi = if_find_by_index(i->ifa_index);
  if (!ifi)
    {
      log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index);
      return;
    }

  bzero(&ifa, sizeof(ifa));
  ifa.iface = ifi;
750
  if (ifa_flags & IFA_F_SECONDARY)
751 752
    ifa.flags |= IA_SECONDARY;

753 754 755 756 757 758
#ifdef IPV6
  /* Ignore tentative addresses silently */
  if (ifa_flags & IFA_F_TENTATIVE)
    return;
#endif

759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
  /* IFA_LOCAL can be unset for IPv6 interfaces */
  memcpy(&ifa.ip, RTA_DATA(a[IFA_LOCAL] ? : a[IFA_ADDRESS]), sizeof(ifa.ip));
  ipa_ntoh(ifa.ip);
  ifa.pxlen = i->ifa_prefixlen;
  if (i->ifa_prefixlen > BITS_PER_IP_ADDRESS)
    {
      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)
    {
      ip_addr addr;
      memcpy(&addr, RTA_DATA(a[IFA_ADDRESS]), sizeof(addr));
      ipa_ntoh(addr);
      ifa.prefix = ifa.brd = addr;

      /* It is either a host address or a peer address */
      if (ipa_equal(ifa.ip, addr))
	ifa.flags |= IA_HOST;
      else
	{
	  ifa.flags |= IA_PEER;
	  ifa.opposite = addr;
	}
    }
  else
    {
      ip_addr netmask = ipa_mkmask(ifa.pxlen);
      ifa.prefix = ipa_and(ifa.ip, netmask);
      ifa.brd = ipa_or(ifa.ip, ipa_not(netmask));
      if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 1)
	ifa.opposite = ipa_opposite_m1(ifa.ip);

#ifndef IPV6
      if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 2)
	ifa.opposite = ipa_opposite_m2(ifa.ip);

      if ((ifi->flags & IF_BROADCAST) && a[IFA_BROADCAST])
	{
	  ip_addr xbrd;
	  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;
	  else if (ifi->flags & IF_TMP_DOWN) /* Complain only during the first scan */
	    log(L_ERR "KIF: Invalid broadcast address %I for %s", xbrd, ifi->name);
	}
#endif
    }

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

  DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %I/%d, brd %I, opp %I\n",
      ifi->index, ifi->name,
      new ? "added" : "removed",
      ifa.ip, ifa.flags, ifa.prefix, ifa.pxlen, ifa.brd, ifa.opposite);
821

822 823 824 825
  if (new)
    ifa_update(&ifa);
  else
    ifa_delete(&ifa);
826 827 828

  if (!scan)
    if_end_partial_update(ifi);
829 830 831 832 833 834 835 836 837
}

void
kif_do_scan(struct kif_proto *p UNUSED)
{
  struct nlmsghdr *h;

  if_start_update();

838
  nl_request_dump(AF_UNSPEC, RTM_GETLINK);
839 840 841 842 843 844
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK)
      nl_parse_link(h, 1);
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

Ondřej Zajíček's avatar
Ondřej Zajíček committed
845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
  /* Re-resolve master interface for slaves */
  struct iface *i;
  WALK_LIST(i, iface_list)
    if (i->master_index)
    {
      struct iface f = {
	.flags = i->flags,
	.mtu = i->mtu,
	.index = i->index,
	.master_index = i->master_index,
	.master = if_find_by_index(i->master_index)
      };

      if (f.master != i->master)
      {
	memcpy(f.name, i->name, sizeof(f.name));
	if_update(&f);
      }
    }

865
  nl_request_dump(BIRD_AF, RTM_GETADDR);
866 867
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR)
868
      nl_parse_addr(h, 1);
869 870 871 872 873 874 875 876 877 878
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

  if_end_update();
}

/*
 *	Routes
 */

879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895
static inline u32
krt_table_id(struct krt_proto *p)
{
  return KRT_CF->sys.table_id;
}

static HASH(struct krt_proto) nl_table_map;

#define RTH_FN(k)	u32_hash(k)
#define RTH_EQ(k1,k2)	k1 == k2
#define RTH_KEY(p)	krt_table_id(p)
#define RTH_NEXT(p)	p->sys.hash_next

#define RTH_REHASH		rth_rehash
#define RTH_PARAMS		/8, *2, 2, 2, 6, 20

HASH_DEFINE_REHASH_FN(RTH, struct krt_proto)
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926

int
krt_capable(rte *e)
{
  rta *a = e->attrs;

  if (a->cast != RTC_UNICAST)
    return 0;

  switch (a->dest)
    {
    case RTD_ROUTER:
    case RTD_DEVICE:
      if (a->iface == NULL)
	return 0;
    case RTD_BLACKHOLE:
    case RTD_UNREACHABLE:
    case RTD_PROHIBIT:
    case RTD_MULTIPATH:
      break;
    default:
      return 0;
    }
  return 1;
}

static inline int
nh_bufsize(struct mpnh *nh)
{
  int rv = 0;
  for (; nh != NULL; nh = nh->next)
927
    rv += RTNH_LENGTH(RTA_LENGTH(sizeof(ip_addr)));
928 929 930 931
  return rv;
}

static int
932
nl_send_route(struct krt_proto *p, rte *e, struct ea_list *eattrs, int op, int dest, ip_addr gw, struct iface *iface)
933 934 935 936
{
  eattr *ea;
  net *net = e->net;
  rta *a = e->attrs;
937 938
  u32 priority = 0;

939 940 941
  struct {
    struct nlmsghdr h;
    struct rtmsg r;
942 943 944 945 946
    char buf[0];
  } *r;

  uint rsize = sizeof(*r) + 128 + KRT_METRICS_MAX*8 + nh_bufsize(a->nexthops);
  r = alloca(rsize);
947

948
  DBG("nl_send_route(%I/%d,op=%x)\n", net->n.prefix, net->n.pxlen, op);
949

950 951 952 953 954
  bzero(&r->h, sizeof(r->h));
  bzero(&r->r, sizeof(r->r));
  r->h.nlmsg_type = op ? RTM_NEWROUTE : RTM_DELROUTE;
  r->h.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
  r->h.nlmsg_flags = op | NLM_F_REQUEST | NLM_F_ACK;
955

956 957 958 959 960
  r->r.rtm_family = BIRD_AF;
  r->r.rtm_dst_len = net->n.pxlen;
  r->r.rtm_protocol = RTPROT_BIRD;
  r->r.rtm_scope = RT_SCOPE_NOWHERE;
  nl_add_attr_ipa(&r->h, rsize, RTA_DST, net->n.prefix);
961

962 963 964 965 966 967 968
  /*
   * Strange behavior for RTM_DELROUTE:
   * 1) rtm_family is ignored in IPv6, works for IPv4
   * 2) not setting RTA_PRIORITY is different from setting default value (on IPv6)
   * 3) not setting RTA_PRIORITY is equivalent to setting 0, which is wildcard
   */

969
  if (krt_table_id(p) < 256)
970
    r->r.rtm_table = krt_table_id(p);
971
  else
972
    nl_add_attr_u32(&r->h, rsize, RTA_TABLE, krt_table_id(p));
973

974 975 976 977 978 979 980 981
  if (a->source == RTS_DUMMY)
    priority = e->u.krt.metric;
  else if (KRT_CF->sys.metric)
    priority = KRT_CF->sys.metric;
  else if ((op != NL_OP_DELETE) && (ea = ea_find(eattrs, EA_KRT_METRIC)))
    priority = ea->u.data;

  if (priority)
982
    nl_add_attr_u32(&r->h, rsize, RTA_PRIORITY, priority);
983

984 985 986
  /* For route delete, we do not specify remaining route attributes */
  if (op == NL_OP_DELETE)
    goto dest;
987

988 989
  /* Default scope is LINK for device routes, UNIVERSE otherwise */
  if (ea = ea_find(eattrs, EA_KRT_SCOPE))
990
    r->r.rtm_scope = ea->u.data;
991
  else
992
    r->r.rtm_scope = (dest == RTD_DEVICE) ? RT_SCOPE_LINK : RT_SCOPE_UNIVERSE;
993

994
  if (ea = ea_find(eattrs, EA_KRT_PREFSRC))
995
    nl_add_attr_ipa(&r->h, rsize, RTA_PREFSRC, *(ip_addr *)ea->u.ptr->data);
996 997

  if (ea = ea_find(eattrs, EA_KRT_REALM))
998
    nl_add_attr_u32(&r->h, rsize, RTA_FLOW, ea->u.data);
999

1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012

  u32 metrics[KRT_METRICS_MAX];
  metrics[0] = 0;

  struct ea_walk_state ews = { .eattrs = eattrs };
  while (ea = ea_walk(&ews, EA_KRT_METRICS, KRT_METRICS_MAX))
  {
    int id = ea->id - EA_KRT_METRICS;
    metrics[0] |= 1 << id;
    metrics[id] = ea->u.data;
  }

  if (metrics[0])
1013
    nl_add_metrics(&r->h, rsize, metrics, KRT_METRICS_MAX);
1014 1015


1016
dest:
1017
  /* a->iface != NULL checked in krt_capable() for router and device routes */
1018
  switch (dest)
1019 1020
    {
    case RTD_ROUTER:
1021 1022 1023
      r->r.rtm_type = RTN_UNICAST;
      nl_add_attr_u32(&r->h, rsize, RTA_OIF, iface->index);
      nl_add_attr_ipa(&r->h, rsize, RTA_GATEWAY, gw);
1024 1025
      break;
    case RTD_DEVICE:
1026 1027
      r->r.rtm_type = RTN_UNICAST;
      nl_add_attr_u32(&r->h, rsize, RTA_OIF, iface->index);
1028 1029
      break;
    case RTD_BLACKHOLE:
1030
      r->r.rtm_type = RTN_BLACKHOLE;
1031 1032
      break;
    case RTD_UNREACHABLE:
1033
      r->r.rtm_type = RTN_UNREACHABLE;
1034 1035
      break;
    case RTD_PROHIBIT:
1036
      r->r.rtm_type = RTN_PROHIBIT;
1037 1038
      break;
    case RTD_MULTIPATH:
1039 1040
      r->r.rtm_type = RTN_UNICAST;
      nl_add_multipath(&r->h, rsize, a->nexthops);
1041
      break;
1042 1043
    case RTD_NONE:
      break;
1044 1045 1046 1047
    default:
      bug("krt_capable inconsistent with nl_send_route");
    }

1048
  /* Ignore missing for DELETE */
1049
  return nl_exchange(&r->h, (op == NL_OP_DELETE));
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
}

static inline int
nl_add_rte(struct krt_proto *p, rte *e, struct ea_list *eattrs)
{
  rta *a = e->attrs;
  int err = 0;

  if (krt_ecmp6(p) && (a->dest == RTD_MULTIPATH))
  {
    struct mpnh *nh = a->nexthops;

    err = nl_send_route(p, e, eattrs, NL_OP_ADD, RTD_ROUTER, nh->gw, nh->iface);
    if (err < 0)
      return err;

    for (nh = nh->next; nh; nh = nh->next)
      err += nl_send_route(p, e, eattrs, NL_OP_APPEND, RTD_ROUTER, nh->gw, nh->iface);

    return err;
  }

  return nl_send_route(p, e, eattrs, NL_OP_ADD, a->dest, a->gw, a->iface);
}

static inline int
nl_delete_rte(struct krt_proto *p, rte *e, struct ea_list *eattrs)
{
  int err = 0;

  /* For IPv6, we just repeatedly request DELETE until we get error */
  do
    err = nl_send_route(p, e, eattrs, NL_OP_DELETE, RTD_NONE, IPA_NONE, NULL);
  while (krt_ecmp6(p) && !err);

  return err;
1086 1087 1088 1089 1090 1091 1092 1093
}

void
krt_replace_rte(struct krt_proto *p, net *n, rte *new, rte *old, struct ea_list *eattrs)
{
  int err = 0;

  /*
1094 1095 1096 1097 1098 1099 1100 1101
   * We could use NL_OP_REPLACE, but route replace on Linux has some problems:
   *
   * 1) Does not check for matching rtm_protocol
   * 2) Has broken semantics for IPv6 ECMP
   * 3) Crashes some kernel version when used for IPv6 ECMP
   *
   * So we use NL_OP_DELETE and then NL_OP_ADD. We also do not trust the old
   * route value, so we do not try to optimize IPv6 ECMP reconfigurations.
1102 1103 1104
   */

  if (old)
1105
    nl_delete_rte(p, old, eattrs);
1106 1107

  if (new)
1108
    err = nl_add_rte(p, new, eattrs);
1109 1110 1111 1112 1113 1114 1115 1116

  if (err < 0)
    n->n.flags |= KRF_SYNC_ERROR;
  else
    n->n.flags &= ~KRF_SYNC_ERROR;
}


1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
static inline struct mpnh *
nl_alloc_mpnh(struct nl_parse_state *s, ip_addr gw, struct iface *iface, byte weight)
{
  struct mpnh *nh = lp_alloc(s->pool, sizeof(struct mpnh));

  nh->gw = gw;
  nh->iface = iface;
  nh->next = NULL;
  nh->weight = weight;

  return nh;
}

static int
nl_mergable_route(struct nl_parse_state *s, net *net, struct krt_proto *p, uint priority, uint krt_type)
{
  /* Route merging must be active */
  if (!s->merge)
    return 0;

  /* Saved and new route must have same network, proto/table, and priority */
  if ((s->net != net) || (s->proto != p) || (s->krt_metric != priority))
    return 0;

  /* Both must be regular unicast routes */
  if ((s->krt_type != RTN_UNICAST) || (krt_type != RTN_UNICAST))
    return 0;

  return 1;
}

static void
nl_announce_route(struct nl_parse_state *s)
{
  rte *e = rte_get_temp(s->attrs);
  e->net = s->net;
  e->u.krt.src = s->krt_src;
  e->u.krt.proto = s->krt_proto;
  e->u.krt.seen = 0;
  e->u.krt.best = 0;
  e->u.krt.metric = s->krt_metric;

  if (s->scan)
    krt_got_route(s->proto, e);
  else
    krt_got_route_async(s->proto, e, s->new);

  s->net = NULL;
  s->attrs = NULL;
  s->proto = NULL;
  lp_flush(s->pool);
}

static inline void
nl_parse_begin(struct nl_parse_state *s, int scan, int merge)
{
  memset(s, 0, sizeof (struct nl_parse_state));
  s->pool = nl_linpool;
  s->scan = scan;
  s->merge = merge;
}

static inline void
nl_parse_end(struct nl_parse_state *s)
{
  if (s->net)
    nl_announce_route(s);
}


1187 1188 1189
#define SKIP(ARG...) do { DBG("KRT: Ignoring route - " ARG); return; } while(0)

static void
1190
nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
1191 1192 1193
{
  struct krt_proto *p;
  struct rtmsg *i;
1194
  struct rtattr *a[BIRD_RTA_MAX];
1195 1196 1197 1198
  int new = h->nlmsg_type == RTM_NEWROUTE;

  ip_addr dst = IPA_NONE;
  u32 oif = ~0;
1199
  u32 table;
1200
  u32 priority = 0;
1201
  u32 def_scope = RT_SCOPE_UNIVERSE;
1202 1203
  int src;

1204
  if (!(i = nl_checkin(h, sizeof(*i))))
1205
    return;
1206 1207

  switch (i->rtm_family)
1208
    {
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
#ifndef IPV6
      case AF_INET:
	if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want4, a, sizeof(a)))
	  return;
	break;
#else
      case AF_INET6:
	if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want6, a, sizeof(a)))
	  return;
	break;
#endif
      default:
	return;
1222 1223 1224 1225 1226 1227 1228 1229 1230
    }

  if (a[RTA_DST])
    {
      memcpy(&dst, RTA_DATA(a[RTA_DST]), sizeof(dst));
      ipa_ntoh(dst);
    }

  if (a[RTA_OIF])
Ondřej Zajíček's avatar
Ondřej Zajíček committed
1231
    oif = rta_get_u32(a[RTA_OIF]);
1232

1233 1234 1235 1236 1237 1238 1239
  if (a[RTA_TABLE])
    table = rta_get_u32(a[RTA_TABLE]);
  else
    table = i->rtm_table;

  p = HASH_FIND(nl_table_map, RTH, table); /* Do we know this table? */
  DBG("KRT: Got %I/%d, type=%d, oif=%d, table=%d, prid=%d, proto=%s\n", dst, i->rtm_dst_len, i->rtm_type, oif, table, i->rtm_protocol, p ? p->p.name : "(none)");
1240
  if (!p)
1241
    SKIP("unknown table %d\n", table);
1242 1243 1244 1245 1246 1247 1248 1249 1250

#ifdef IPV6
  if (a[RTA_IIF])
    SKIP("IIF set\n");
#else
  if (i->rtm_tos != 0)			/* We don't support TOS */
    SKIP("TOS %02x\n", i->rtm_tos);
#endif

1251
  if (s->scan && !new)
1252 1253
    SKIP("RTM_DELROUTE in scan\n");

1254 1255 1256
  if (a[RTA_PRIORITY])
    priority = rta_get_u32(a[RTA_PRIORITY]);

1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
  int c = ipa_classify_net(dst);
  if ((c < 0) || !(c & IADDR_HOST) || ((c & IADDR_SCOPE_MASK) <= SCOPE_LINK))
    SKIP("strange class/scope\n");

  switch (i->rtm_protocol)
    {
    case RTPROT_UNSPEC:
      SKIP("proto unspec\n");

    case RTPROT_REDIRECT:
      src = KRT_SRC_REDIRECT;
      break;

    case RTPROT_KERNEL:
      src = KRT_SRC_KERNEL;
      return;

    case RTPROT_BIRD:
1275
      if (!s->scan)
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
	SKIP("echo\n");
      src = KRT_SRC_BIRD;
      break;

    case RTPROT_BOOT:
    default:
      src = KRT_SRC_ALIEN;
    }

  net *net = net_get(p->p.table, dst, i->rtm_dst_len);

1287 1288 1289 1290 1291 1292 1293 1294
  if (s->net && !nl_mergable_route(s, net, p, priority, i->rtm_type))
    nl_announce_route(s);

  rta *ra = lp_allocz(s->pool, sizeof(rta));
  ra->src = p->p.main_source;
  ra->source = RTS_INHERIT;
  ra->scope = SCOPE_UNIVERSE;
  ra->cast = RTC_UNICAST;
1295 1296 1297 1298 1299

  switch (i->rtm_type)
    {
    case RTN_UNICAST:

1300
      if (a[RTA_MULTIPATH])
1301
	{
1302
	  ra->dest = RTD_MULTIPATH;
1303
	  ra->nexthops = nl_parse_multipath(p, a[RTA_MULTIPATH], i->rtm_family);
1304
	  if (!ra->nexthops)
1305 1306 1307 1308 1309
	    {
	      log(L_ERR "KRT: Received strange multipath route %I/%d",
		  net->n.prefix, net->n.pxlen);
	      return;
	    }
1310

1311 1312 1313
	  break;
	}

1314 1315
      ra->iface = if_find_by_index(oif);
      if (!ra->iface)
1316 1317 1318 1319 1320 1321 1322 1323 1324
	{
	  log(L_ERR "KRT: Received route %I/%d with unknown ifindex %u",
	      net->n.prefix, net->n.pxlen, oif);
	  return;
	}

      if (a[RTA_GATEWAY])
	{
	  neighbor *ng;
1325 1326 1327
	  ra->dest = RTD_ROUTER;
	  memcpy(&ra->gw, RTA_DATA(a[RTA_GATEWAY]), sizeof(ra->gw));
	  ipa_ntoh(ra->gw);
1328

1329
#ifdef IPV6
1330
	  /* Silently skip strange 6to4 routes */
1331
	  if (ipa_in_net(ra->gw, IPA_NONE, 96))
1332
	    return;
1333
#endif
1334

1335
	  ng = neigh_find2(&p->p, &ra->gw, ra->iface,
1336 1337 1338 1339
			   (i->rtm_flags & RTNH_F_ONLINK) ? NEF_ONLINK : 0);
	  if (!ng || (ng->scope == SCOPE_HOST))
	    {
	      log(L_ERR "KRT: Received route %I/%d with strange next-hop %I",
1340
		  net->n.prefix, net->n.pxlen, ra->gw);
1341 1342 1343 1344 1345
	      return;
	    }
	}
      else
	{
1346
	  ra->dest = RTD_DEVICE;
1347
	  def_scope = RT_SCOPE_LINK;
1348 1349 1350 1351
	}

      break;
    case RTN_BLACKHOLE:
1352
      ra->dest = RTD_BLACKHOLE;
1353 1354
      break;
    case RTN_UNREACHABLE:
1355
      ra->dest = RTD_UNREACHABLE;
1356 1357
      break;
    case RTN_PROHIBIT:
1358
      ra->dest = RTD_PROHIBIT;
1359 1360 1361 1362 1363 1364 1365
      break;
    /* FIXME: What about RTN_THROW? */
    default:
      SKIP("type %d\n", i->rtm_type);
      return;
    }

1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
  if (i->rtm_scope != def_scope)
    {
      ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr));
      ea->next = ra->eattrs;
      ra->eattrs = ea;
      ea->flags = EALF_SORTED;
      ea->count = 1;
      ea->attrs[0].id = EA_KRT_SCOPE;
      ea->attrs[0].flags = 0;
      ea->attrs[0].type = EAF_TYPE_INT;
      ea->attrs[0].u.data = i->rtm_scope;
    }

1379 1380 1381 1382 1383 1384
  if (a[RTA_PREFSRC])
    {
      ip_addr ps;
      memcpy(&ps, RTA_DATA(a[RTA_PREFSRC]), sizeof(ps));
      ipa_ntoh(ps);

1385 1386 1387
      ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr));
      ea->next = ra->eattrs;
      ra->eattrs = ea;
1388 1389 1390 1391 1392
      ea->flags = EALF_SORTED;
      ea->count = 1;
      ea->attrs[0].id = EA_KRT_PREFSRC;
      ea->attrs[0].flags = 0;
      ea->attrs[0].type = EAF_TYPE_IP_ADDRESS;
1393
      ea->attrs[0].u.ptr = lp_alloc(s->pool, sizeof(struct adata) + sizeof(ps));
1394 1395 1396 1397 1398 1399
      ea->attrs[0].u.ptr->length = sizeof(ps);
      memcpy(ea->attrs[0].u.ptr->data, &ps, sizeof(ps));
    }

  if (a[RTA_FLOW])
    {
1400 1401 1402
      ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr));
      ea->next = ra->eattrs;
      ra->eattrs = ea;
1403 1404 1405 1406 1407
      ea->flags = EALF_SORTED;
      ea->count = 1;
      ea->attrs[0].id = EA_KRT_REALM;
      ea->attrs[0].flags = 0;
      ea->attrs[0].type = EAF_TYPE_INT;
Ondřej Zajíček's avatar
Ondřej Zajíček committed
1408
      ea->attrs[0].u.data = rta_get_u32(a[RTA_FLOW]);
1409 1410
    }

1411 1412 1413
  if (a[RTA_METRICS])
    {
      u32 metrics[KRT_METRICS_MAX];
1414
      ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + KRT_METRICS_MAX * sizeof(eattr));
1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
      int t, n = 0;

      if (nl_parse_metrics(a[RTA_METRICS], metrics, ARRAY_SIZE(metrics)) < 0)
        {
	  log(L_ERR "KRT: Received route %I/%d with strange RTA_METRICS attribute",
	      net->n.prefix, net->n.pxlen);
	  return;
	}

      for (t = 1; t < KRT_METRICS_MAX; t++)
	if (metrics[0] & (1 << t))
	  {
	    ea->attrs[n].id = EA_CODE(EAP_KRT, KRT_METRICS_OFFSET + t);
	    ea->attrs[n].flags = 0;
	    ea->attrs[n].type = EAF_TYPE_INT; /* FIXME: Some are EAF_TYPE_BITFIELD */
	    ea->attrs[n].u.data = metrics[t];
	    n++;
	  }

      if (n > 0)
        {
1436
	  ea->next = ra->eattrs;
1437 1438
	  ea->flags = EALF_SORTED;
	  ea->count = n;
1439
	  ra->eattrs = ea;
1440 1441 1442
	}
    }

1443 1444
  /*
   * Ideally, now we would send the received route to the rest of kernel code.
1445 1446 1447 1448
   * But IPv6 ECMP routes before 4.11 are sent as a sequence of routes, so we
   * postpone it and merge next hops until the end of the sequence. Note that
   * proper multipath updates are rejected by nl_mergable_route(), so it is
   * always the first case for them.
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
   */

  if (!s->net)
  {
    /* Store the new route */
    s->net = net;
    s->attrs = ra;
    s->proto = p;
    s->new = new;
    s->krt_src = src;
    s->krt_type = i->rtm_type;
    s->krt_proto = i->rtm_protocol;
    s->krt_metric = priority;
  }
1463
  else
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
  {
    /* Merge next hops with the stored route */
    rta *a = s->attrs;

    if (a->dest != RTD_MULTIPATH)
    {
      a->dest = RTD_MULTIPATH;
      a->nexthops = nl_alloc_mpnh(s, a->gw, a->iface, 0);
    }

    mpnh_insert(&a->nexthops, nl_alloc_mpnh(s, ra->gw, ra->iface, 0));
  }
1476 1477 1478 1479 1480 1481
}

void
krt_do_scan(struct krt_proto *p UNUSED)	/* CONFIG_ALL_TABLES_AT_ONCE => p is NULL */
{
  struct nlmsghdr *h;
1482 1483 1484
  struct nl_parse_state s;

  nl_parse_begin(&s, 1, krt_ecmp6(p));
1485

1486
  nl_request_dump(BIRD_AF, RTM_GETROUTE);
1487 1488
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)
1489
      nl_parse_route(&s, h);
1490 1491
    else
      log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type);
1492 1493

  nl_parse_end(&s);
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
}

/*
 *	Asynchronous Netlink interface
 */

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

static void
nl_async_msg(struct nlmsghdr *h)
{
1506 1507
  struct nl_parse_state s;

1508 1509 1510 1511 1512
  switch (h->nlmsg_type)
    {
    case RTM_NEWROUTE:
    case RTM_DELROUTE:
      DBG("KRT: Received async route notification (%d)\n", h->nlmsg_type);
1513 1514 1515
      nl_parse_begin(&s, 0, 0);
      nl_parse_route(&s, h);
      nl_parse_end(&s);
1516 1517 1518 1519
      break;
    case RTM_NEWLINK:
    case RTM_DELLINK:
      DBG("KRT: Received async link notification (%d)\n", h->nlmsg_type);
1520 1521
      if (kif_proto)
	nl_parse_link(h, 0);
1522 1523 1524 1525
      break;
    case RTM_NEWADDR:
    case RTM_DELADDR:
      DBG("KRT: Received async address notification (%d)\n", h->nlmsg_type);
1526 1527
      if (kif_proto)
	nl_parse_addr(h, 0);
1528 1529 1530 1531 1532 1533 1534
      break;
    default:
      DBG("KRT: Received unknown async notification (%d)\n", h->nlmsg_type);
    }
}

static int
1535
nl_async_hook(sock *sk, uint size UNUSED)
1536 1537 1538
{
  struct iovec iov = { nl_async_rx_buffer, NL_RX_SIZE };
  struct sockaddr_nl sa;
1539 1540 1541 1542 1543 1544
  struct msghdr m = {
    .msg_name = &sa,
    .msg_namelen = sizeof(sa),
    .msg_iov = &iov,
    .msg_iovlen = 1,
  };
1545 1546
  struct nlmsghdr *h;
  int x;
Pavel Tvrdík's avatar
Pavel Tvrdík committed
1547
  uint len;
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558

  x = recvmsg(sk->fd, &m, 0);
  if (x < 0)
    {
      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.
	   */
1559
	  log(L_WARN "Kernel dropped some netlink messages, will resync on next scan.");
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
	  return 1;	/* More data are likely to be ready */
	}
      else if (errno != EWOULDBLOCK)
	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))
    {
      nl_async_msg(h);
      h = NLMSG_NEXT(h, len);
    }
  if (len)
    log(L_WARN "nl_async_hook: Found packet remnant of size %d", len);
  return 1;
}

1588 1589 1590 1591 1592 1593
static void
nl_async_err_hook(sock *sk, int e UNUSED)
{
  nl_async_hook(sk, 0);
}

1594 1595 1596 1597 1598 1599 1600
static void
nl_open_async(void)
{
  sock *sk;
  struct sockaddr_nl sa;
  int fd;

1601
  if (nl_async_sk)
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
    return;

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

  fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
  if (fd < 0)
    {
      log(L_ERR "Unable to open asynchronous rtnetlink socket: %m");
      return;
    }

  bzero(&sa, sizeof(sa));
  sa.nl_family = AF_NETLINK;
#ifdef IPV6
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV6_IFADDR | RTMGRP_IPV6_ROUTE;
#else
  sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE;
#endif
  if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0)
    {
      log(L_ERR "Unable to bind asynchronous rtnetlink socket: %m");
1623
      close(fd);
1624 1625 1626
      return;
    }

1627 1628
  nl_async_rx_buffer = xmalloc(NL_RX_SIZE);

1629 1630 1631
  sk = nl_async_sk = sk_new(krt_pool);
  sk->type = SK_MAGIC;
  sk->rx_hook = nl_async_hook;
1632
  sk->err_hook = nl_async_err_hook;
1633
  sk->fd = fd;
1634
  if (sk_open(sk) < 0)
1635 1636 1637
    bug("Netlink: sk_open failed");
}

1638

1639 1640 1641 1642 1643
/*
 *	Interface to the UNIX krt module
 */

void
1644 1645
krt_sys_io_init(void)
{
1646
  nl_linpool = lp_new(krt_pool, 4080);
1647 1648 1649 1650
  HASH_INIT(nl_table_map, krt_pool, 6);
}

int
1651
krt_sys_start(struct krt_proto *p)
1652
{
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
  struct krt_proto *old = HASH_FIND(nl_table_map, RTH, krt_table_id(p));

  if (old)
    {
      log(L_ERR "%s: Kernel table %u already registered by %s",
	  p->p.name, krt_table_id(p), old->p.name);
      return 0;
    }

  HASH_INSERT2(nl_table_map, RTH, krt_pool, p);
1663 1664 1665

  nl_open();
  nl_open_async();
1666 1667

  return 1;
1668 1669 1670
}

void
1671
krt_sys_shutdown(struct krt_proto *p)
1672
{
1673
  HASH_REMOVE2(nl_table_map, RTH, krt_pool, p);
1674 1675 1676 1677 1678
}

int
krt_sys_reconfigure(struct krt_proto *p UNUSED, struct krt_config *n, struct krt_config *o)
{
1679
  return (n->sys.table_id == o->sys.table_id) && (n->sys.metric == o->sys.metric);
1680 1681 1682 1683 1684 1685
}

void
krt_sys_init_config(struct krt_config *cf)
{
  cf->sys.table_id = RT_TABLE_MAIN;
1686
  cf->sys.metric = 0;
1687 1688 1689 1690 1691 1692
}

void
krt_sys_copy_config(struct krt_config *d, struct krt_config *s)
{
  d->sys.table_id = s->sys.table_id;
1693
  d->sys.metric = s->sys.metric;
1694 1695
}

1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
static const char *krt_metrics_names[KRT_METRICS_MAX] = {
  NULL, "lock", "mtu", "window", "rtt", "rttvar", "sstresh", "cwnd", "advmss",
  "reordering", "hoplimit", "initcwnd", "features", "rto_min", "initrwnd", "quickack"
};

static const char *krt_features_names[KRT_FEATURES_MAX] = {
  "ecn", NULL, NULL, "allfrag"
};

int
krt_sys_get_attr(eattr *a, byte *buf, int buflen UNUSED)
{
  switch (a->id)
  {
  case EA_KRT_PREFSRC:
    bsprintf(buf, "prefsrc");
    return GA_NAME;

  case EA_KRT_REALM:
    bsprintf(buf, "realm");
    return GA_NAME;

1718 1719 1720 1721
  case EA_KRT_SCOPE:
    bsprintf(buf, "scope");
    return GA_NAME;

1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
  case EA_KRT_LOCK:
    buf += bsprintf(buf, "lock:");
    ea_format_bitfield(a, buf, buflen, krt_metrics_names, 2, KRT_METRICS_MAX);
    return GA_FULL;

  case EA_KRT_FEATURES:
    buf += bsprintf(buf, "features:");
    ea_format_bitfield(a, buf, buflen, krt_features_names, 0, KRT_FEATURES_MAX);
    return GA_FULL;

  default:;
    int id = (int)EA_ID(a->id) - KRT_METRICS_OFFSET;
    if (id > 0 && id < KRT_METRICS_MAX)
    {
      bsprintf(buf, "%s", krt_metrics_names[id]);
      return GA_NAME;
    }

    return GA_UNKNOWN;
  }
}

1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755


void
kif_sys_start(struct kif_proto *p UNUSED)
{
  nl_open();
  nl_open_async();
}

void
kif_sys_shutdown(struct kif_proto *p UNUSED)
{
1756
}