io.c 35.2 KB
Newer Older
1 2 3
/*
 *	BIRD Internet Routing Daemon -- Unix I/O
 *
4
 *	(c) 1998--2004 Martin Mares <mj@ucw.cz>
5
 *      (c) 2004       Ondrej Filip <feela@network.cz>
6 7 8 9
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

10 11 12 13
/* Unfortunately, some glibc versions hide parts of RFC 3542 API
   if _GNU_SOURCE is not defined. */
#define _GNU_SOURCE 1

14 15
#include <stdio.h>
#include <stdlib.h>
16
#include <time.h>
17 18 19 20
#include <sys/time.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/fcntl.h>
21
#include <sys/uio.h>
22
#include <sys/un.h>
23 24
#include <unistd.h>
#include <errno.h>
Ondřej Zajíček's avatar
Ondřej Zajíček committed
25
#include <netinet/in.h>
26
#include <netinet/icmp6.h>
27 28 29 30 31 32

#include "nest/bird.h"
#include "lib/lists.h"
#include "lib/resource.h"
#include "lib/timer.h"
#include "lib/socket.h"
33
#include "lib/event.h"
34
#include "lib/string.h"
35 36 37
#include "nest/iface.h"

#include "lib/unix.h"
38
#include "lib/sysio.h"
39

40
/* Maximum number of calls of tx handler for one socket in one
41 42 43 44 45
 * select iteration. Should be small enough to not monopolize CPU by
 * one protocol instance.
 */
#define MAX_STEPS 4

46 47 48 49 50
/* Maximum number of calls of rx handler for all sockets in one select
   iteration. RX callbacks are often much more costly so we limit
   this to gen small latencies */
#define MAX_RX_STEPS 4

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
/*
 *	Tracked Files
 */

struct rfile {
  resource r;
  FILE *f;
};

static void
rf_free(resource *r)
{
  struct rfile *a = (struct rfile *) r;

  fclose(a->f);
}

static void
rf_dump(resource *r)
{
  struct rfile *a = (struct rfile *) r;

  debug("(FILE *%p)\n", a->f);
}

static struct resclass rf_class = {
  "FILE",
  sizeof(struct rfile),
  rf_free,
80
  rf_dump,
81
  NULL,
82
  NULL
83 84 85
};

void *
86
tracked_fopen(pool *p, char *name, char *mode)
87 88 89 90 91 92 93 94 95 96 97
{
  FILE *f = fopen(name, mode);

  if (f)
    {
      struct rfile *r = ralloc(p, &rf_class);
      r->f = f;
    }
  return f;
}

98 99 100 101 102
/**
 * DOC: Timers
 *
 * Timers are resources which represent a wish of a module to call
 * a function at the specified time. The platform dependent code
Martin Mareš's avatar
Martin Mareš committed
103
 * doesn't guarantee exact timing, only that a timer function
104 105
 * won't be called before the requested time.
 *
106 107 108 109 110
 * In BIRD, time is represented by values of the &bird_clock_t type
 * which are integral numbers interpreted as a relative number of seconds since
 * some fixed time point in past. The current time can be read
 * from variable @now with reasonable accuracy and is monotonic. There is also
 * a current 'absolute' time in variable @now_real reported by OS.
111 112 113 114 115
 *
 * Each timer is described by a &timer structure containing a pointer
 * to the handler function (@hook), data private to this function (@data),
 * time the function should be called at (@expires, 0 for inactive timers),
 * for the other fields see |timer.h|.
116 117 118 119 120 121 122
 */

#define NEAR_TIMER_LIMIT 4

static list near_timers, far_timers;
static bird_clock_t first_far_timer = TIME_INFINITY;

123
/* now must be different from 0, because 0 is a special value in timer->expires */
124
bird_clock_t now = 1, now_real, boot_time;
125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178

static void
update_times_plain(void)
{
  bird_clock_t new_time = time(NULL);
  int delta = new_time - now_real;

  if ((delta >= 0) && (delta < 60))
    now += delta;
  else if (now_real != 0)
   log(L_WARN "Time jump, delta %d s", delta);

  now_real = new_time;
}

static void
update_times_gettime(void)
{
  struct timespec ts;
  int rv;

  rv = clock_gettime(CLOCK_MONOTONIC, &ts);
  if (rv != 0)
    die("clock_gettime: %m");

  if (ts.tv_sec != now) {
    if (ts.tv_sec < now)
      log(L_ERR "Monotonic timer is broken");

    now = ts.tv_sec;
    now_real = time(NULL);
  }
}

static int clock_monotonic_available;

static inline void
update_times(void)
{
  if (clock_monotonic_available)
    update_times_gettime();
  else
    update_times_plain();
}

static inline void
init_times(void)
{
 struct timespec ts;
 clock_monotonic_available = (clock_gettime(CLOCK_MONOTONIC, &ts) == 0);
 if (!clock_monotonic_available)
   log(L_WARN "Monotonic timer is missing");
}

179 180 181 182 183 184 185 186 187 188 189 190 191 192

static void
tm_free(resource *r)
{
  timer *t = (timer *) r;

  tm_stop(t);
}

static void
tm_dump(resource *r)
{
  timer *t = (timer *) r;

193
  debug("(code %p, data %p, ", t->hook, t->data);
194 195 196 197
  if (t->randomize)
    debug("rand %d, ", t->randomize);
  if (t->recurrent)
    debug("recur %d, ", t->recurrent);
198 199 200 201 202 203 204 205 206 207
  if (t->expires)
    debug("expires in %d sec)\n", t->expires - now);
  else
    debug("inactive)\n");
}

static struct resclass tm_class = {
  "Timer",
  sizeof(timer),
  tm_free,
208
  tm_dump,
209
  NULL,
210
  NULL
211 212
};

213 214 215 216 217 218 219 220
/**
 * tm_new - create a timer
 * @p: pool
 *
 * This function creates a new timer resource and returns
 * a pointer to it. To use the timer, you need to fill in
 * the structure fields and call tm_start() to start timing.
 */
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
timer *
tm_new(pool *p)
{
  timer *t = ralloc(p, &tm_class);
  return t;
}

static inline void
tm_insert_near(timer *t)
{
  node *n = HEAD(near_timers);

  while (n->next && (SKIP_BACK(timer, n, n)->expires < t->expires))
    n = n->next;
  insert_node(&t->n, n->prev);
}

238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
/**
 * tm_start - start a timer
 * @t: timer
 * @after: number of seconds the timer should be run after
 *
 * This function schedules the hook function of the timer to
 * be called after @after seconds. If the timer has been already
 * started, it's @expire time is replaced by the new value.
 *
 * You can have set the @randomize field of @t, the timeout
 * will be increased by a random number of seconds chosen
 * uniformly from range 0 .. @randomize.
 *
 * You can call tm_start() from the handler function of the timer
 * to request another run of the timer. Also, you can set the @recurrent
 * field to have the timer re-added automatically with the same timeout.
 */
255 256 257 258 259 260
void
tm_start(timer *t, unsigned after)
{
  bird_clock_t when;

  if (t->randomize)
261
    after += random() % (t->randomize + 1);
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
  when = now + after;
  if (t->expires == when)
    return;
  if (t->expires)
    rem_node(&t->n);
  t->expires = when;
  if (after <= NEAR_TIMER_LIMIT)
    tm_insert_near(t);
  else
    {
      if (!first_far_timer || first_far_timer > when)
	first_far_timer = when;
      add_tail(&far_timers, &t->n);
    }
}

278 279 280 281 282 283 284
/**
 * tm_stop - stop a timer
 * @t: timer
 *
 * This function stops a timer. If the timer is already stopped,
 * nothing happens.
 */
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
void
tm_stop(timer *t)
{
  if (t->expires)
    {
      rem_node(&t->n);
      t->expires = 0;
    }
}

static void
tm_dump_them(char *name, list *l)
{
  node *n;
  timer *t;

  debug("%s timers:\n", name);
  WALK_LIST(n, *l)
    {
      t = SKIP_BACK(timer, n, n);
      debug("%p ", t);
      tm_dump(&t->r);
    }
  debug("\n");
}

void
tm_dump_all(void)
{
  tm_dump_them("Near", &near_timers);
  tm_dump_them("Far", &far_timers);
}

static inline time_t
tm_first_shot(void)
{
  time_t x = first_far_timer;

  if (!EMPTY_LIST(near_timers))
    {
      timer *t = SKIP_BACK(timer, n, HEAD(near_timers));
      if (t->expires < x)
	x = t->expires;
    }
  return x;
}

static void
tm_shot(void)
{
  timer *t;
  node *n, *m;

  if (first_far_timer <= now)
    {
340
      bird_clock_t limit = now + NEAR_TIMER_LIMIT;
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
      first_far_timer = TIME_INFINITY;
      n = HEAD(far_timers);
      while (m = n->next)
	{
	  t = SKIP_BACK(timer, n, n);
	  if (t->expires <= limit)
	    {
	      rem_node(n);
	      tm_insert_near(t);
	    }
	  else if (t->expires < first_far_timer)
	    first_far_timer = t->expires;
	  n = m;
	}
    }
  while ((n = HEAD(near_timers)) -> next)
    {
358
      int delay;
359 360 361 362
      t = SKIP_BACK(timer, n, n);
      if (t->expires > now)
	break;
      rem_node(n);
363
      delay = t->expires - now;
364
      t->expires = 0;
365 366 367 368 369 370 371
      if (t->recurrent)
	{
	  int i = t->recurrent - delay;
	  if (i < 0)
	    i = 0;
	  tm_start(t, i);
	}
372 373 374 375
      t->hook(t);
    }
}

376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399
/**
 * tm_parse_datetime - parse a date and time
 * @x: datetime string
 *
 * tm_parse_datetime() takes a textual representation of
 * a date and time (dd-mm-yyyy hh:mm:ss)
 * and converts it to the corresponding value of type &bird_clock_t.
 */
bird_clock_t
tm_parse_datetime(char *x)
{
  struct tm tm;
  int n;
  time_t t;

  if (sscanf(x, "%d-%d-%d %d:%d:%d%n", &tm.tm_mday, &tm.tm_mon, &tm.tm_year, &tm.tm_hour, &tm.tm_min, &tm.tm_sec, &n) != 6 || x[n])
    return tm_parse_date(x);
  tm.tm_mon--;
  tm.tm_year -= 1900;
  t = mktime(&tm);
  if (t == (time_t) -1)
    return 0;
  return t;
}
400 401 402 403 404 405 406
/**
 * tm_parse_date - parse a date
 * @x: date string
 *
 * tm_parse_date() takes a textual representation of a date (dd-mm-yyyy)
 * and converts it to the corresponding value of type &bird_clock_t.
 */
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
bird_clock_t
tm_parse_date(char *x)
{
  struct tm tm;
  int n;
  time_t t;

  if (sscanf(x, "%d-%d-%d%n", &tm.tm_mday, &tm.tm_mon, &tm.tm_year, &n) != 3 || x[n])
    return 0;
  tm.tm_mon--;
  tm.tm_year -= 1900;
  tm.tm_hour = tm.tm_min = tm.tm_sec = 0;
  t = mktime(&tm);
  if (t == (time_t) -1)
    return 0;
  return t;
}

425 426
static void
tm_format_reltime(char *x, struct tm *tm, bird_clock_t delta)
427
{
428 429
  static char *month_names[12] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun",
				   "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
430

431 432 433 434 435 436
  if (delta < 20*3600)
    bsprintf(x, "%02d:%02d", tm->tm_hour, tm->tm_min);
  else if (delta < 360*86400)
    bsprintf(x, "%s%02d", month_names[tm->tm_mon], tm->tm_mday);
  else
    bsprintf(x, "%d", tm->tm_year+1900);
437 438
}

439 440
#include "conf/conf.h"

441 442 443 444 445
/**
 * tm_format_datetime - convert date and time to textual representation
 * @x: destination buffer of size %TM_DATETIME_BUFFER_SIZE
 * @t: time
 *
446 447
 * This function formats the given relative time value @t to a textual
 * date/time representation (dd-mm-yyyy hh:mm:ss) in real time.
448
 */
449
void
450
tm_format_datetime(char *x, struct timeformat *fmt_spec, bird_clock_t t)
451
{
452
  const char *fmt_used;
453
  struct tm *tm;
454 455
  bird_clock_t delta = now - t;
  t = now_real - delta;
456 457
  tm = localtime(&t);

458 459
  if (fmt_spec->fmt1 == NULL)
    return tm_format_reltime(x, tm, delta);
460

461 462
  if ((fmt_spec->limit == 0) || (delta < fmt_spec->limit))
    fmt_used = fmt_spec->fmt1;
463
  else
464 465 466 467 468
    fmt_used = fmt_spec->fmt2;

  int rv = strftime(x, TM_DATETIME_BUFFER_SIZE, fmt_used, tm);
  if (((rv == 0) && fmt_used[0]) || (rv == TM_DATETIME_BUFFER_SIZE))
    strcpy(x, "<too-long>");
469 470
}

471 472 473 474 475 476 477 478 479 480
/**
 * DOC: Sockets
 *
 * Socket resources represent network connections. Their data structure (&socket)
 * contains a lot of fields defining the exact type of the socket, the local and
 * remote addresses and ports, pointers to socket buffers and finally pointers to
 * hook functions to be called when new data have arrived to the receive buffer
 * (@rx_hook), when the contents of the transmit buffer have been transmitted
 * (@tx_hook) and when an error or connection close occurs (@err_hook).
 *
481
 * Freeing of sockets from inside socket hooks is perfectly safe.
482 483
 */

484 485 486 487
#ifndef SOL_IP
#define SOL_IP IPPROTO_IP
#endif

488 489 490 491
#ifndef SOL_IPV6
#define SOL_IPV6 IPPROTO_IPV6
#endif

492
static list sock_list;
493
static struct birdsock *current_sock;
494
static struct birdsock *stored_sock;
495 496 497 498 499 500 501 502 503 504
static int sock_recalc_fdsets_p;

static inline sock *
sk_next(sock *s)
{
  if (!s->n.next->next)
    return NULL;
  else
    return SKIP_BACK(sock, n, s->n.next);
}
505 506

static void
507
sk_alloc_bufs(sock *s)
508
{
509 510 511 512 513 514 515
  if (!s->rbuf && s->rbsize)
    s->rbuf = s->rbuf_alloc = xmalloc(s->rbsize);
  s->rpos = s->rbuf;
  if (!s->tbuf && s->tbsize)
    s->tbuf = s->tbuf_alloc = xmalloc(s->tbsize);
  s->tpos = s->ttx = s->tbuf;
}
516

517 518 519
static void
sk_free_bufs(sock *s)
{
520
  if (s->rbuf_alloc)
521 522 523 524
    {
      xfree(s->rbuf_alloc);
      s->rbuf = s->rbuf_alloc = NULL;
    }
525
  if (s->tbuf_alloc)
526 527 528 529 530 531 532 533 534 535 536 537
    {
      xfree(s->tbuf_alloc);
      s->tbuf = s->tbuf_alloc = NULL;
    }
}

static void
sk_free(resource *r)
{
  sock *s = (sock *) r;

  sk_free_bufs(s);
538
  if (s->fd >= 0)
539 540
    {
      close(s->fd);
541 542
      if (s == current_sock)
	current_sock = sk_next(s);
543 544
      if (s == stored_sock)
	stored_sock = sk_next(s);
545
      rem_node(&s->n);
546
      sock_recalc_fdsets_p = 1;
547
    }
548 549
}

550 551 552 553 554 555 556
void
sk_reallocate(sock *s)
{
  sk_free_bufs(s);
  sk_alloc_bufs(s);
}

557 558 559 560
static void
sk_dump(resource *r)
{
  sock *s = (sock *) r;
561
  static char *sk_type_names[] = { "TCP<", "TCP>", "TCP", "UDP", "UDP/MC", "IP", "IP/MC", "MAGIC", "UNIX<", "UNIX", "DEL!" };
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578

  debug("(%s, ud=%p, sa=%08x, sp=%d, da=%08x, dp=%d, tos=%d, ttl=%d, if=%s)\n",
	sk_type_names[s->type],
	s->data,
	s->saddr,
	s->sport,
	s->daddr,
	s->dport,
	s->tos,
	s->ttl,
	s->iface ? s->iface->name : "none");
}

static struct resclass sk_class = {
  "Socket",
  sizeof(sock),
  sk_free,
579
  sk_dump,
580
  NULL,
581
  NULL
582 583
};

584 585 586 587 588 589 590
/**
 * sk_new - create a socket
 * @p: pool
 *
 * This function creates a new socket resource. If you want to use it,
 * you need to fill in all the required fields of the structure and
 * call sk_open() to do the actual opening of the socket.
591 592 593
 *
 * The real function name is sock_new(), sk_new() is a macro wrapper
 * to avoid collision with OpenSSL.
594
 */
595
sock *
596
sock_new(pool *p)
597 598 599
{
  sock *s = ralloc(p, &sk_class);
  s->pool = p;
600
  // s->saddr = s->daddr = IPA_NONE;
601
  s->tos = s->priority = s->ttl = -1;
602 603 604 605
  s->fd = -1;
  return s;
}

606 607 608 609 610 611
static void
sk_insert(sock *s)
{
  add_tail(&sock_list, &s->n);
  sock_recalc_fdsets_p = 1;
}
612

613 614
#ifdef IPV6

Ondřej Zajíček's avatar
Ondřej Zajíček committed
615 616
void
fill_in_sockaddr(struct sockaddr_in6 *sa, ip_addr a, struct iface *ifa, unsigned port)
617
{
618
  memset(sa, 0, sizeof (struct sockaddr_in6));
619 620 621
  sa->sin6_family = AF_INET6;
  sa->sin6_port = htons(port);
  sa->sin6_flowinfo = 0;
622 623 624
#ifdef HAVE_SIN_LEN
  sa->sin6_len = sizeof(struct sockaddr_in6);
#endif
625 626
  set_inaddr(&sa->sin6_addr, a);

627 628
  if (ifa && ipa_has_link_scope(a))
    sa->sin6_scope_id = ifa->index;
629 630
}

Ondřej Zajíček's avatar
Ondřej Zajíček committed
631
void
632
get_sockaddr(struct sockaddr_in6 *sa, ip_addr *a, struct iface **ifa, unsigned *port, int check)
633
{
634 635
  if (check && sa->sin6_family != AF_INET6)
    bug("get_sockaddr called for wrong address family (%d)", sa->sin6_family);
636 637 638 639
  if (port)
    *port = ntohs(sa->sin6_port);
  memcpy(a, &sa->sin6_addr, sizeof(*a));
  ipa_ntoh(*a);
640 641 642

  if (ifa && ipa_has_link_scope(*a))
    *ifa = if_find_by_index(sa->sin6_scope_id);
643 644 645 646
}

#else

Ondřej Zajíček's avatar
Ondřej Zajíček committed
647 648
void
fill_in_sockaddr(struct sockaddr_in *sa, ip_addr a, struct iface *ifa, unsigned port)
649
{
650
  memset (sa, 0, sizeof (struct sockaddr_in));
651 652
  sa->sin_family = AF_INET;
  sa->sin_port = htons(port);
653 654 655
#ifdef HAVE_SIN_LEN
  sa->sin_len = sizeof(struct sockaddr_in);
#endif
656 657 658
  set_inaddr(&sa->sin_addr, a);
}

Ondřej Zajíček's avatar
Ondřej Zajíček committed
659
void
660
get_sockaddr(struct sockaddr_in *sa, ip_addr *a, struct iface **ifa, unsigned *port, int check)
661
{
662 663
  if (check && sa->sin_family != AF_INET)
    bug("get_sockaddr called for wrong address family (%d)", sa->sin_family);
664 665
  if (port)
    *port = ntohs(sa->sin_port);
666
  memcpy(a, &sa->sin_addr.s_addr, sizeof(*a));
667
  ipa_ntoh(*a);
668 669
}

670 671
#endif

672 673 674 675

#ifdef IPV6

/* PKTINFO handling is also standardized in IPv6 */
676
#define CMSG_RX_SPACE (CMSG_SPACE(sizeof(struct in6_pktinfo)) + CMSG_SPACE(sizeof(int)))
677 678
#define CMSG_TX_SPACE CMSG_SPACE(sizeof(struct in6_pktinfo))

679 680 681 682 683 684 685 686 687
/*
 * RFC 2292 uses IPV6_PKTINFO for both the socket option and the cmsg
 * type, RFC 3542 changed the socket option to IPV6_RECVPKTINFO. If we
 * don't have IPV6_RECVPKTINFO we suppose the OS implements the older
 * RFC and we use IPV6_PKTINFO.
 */
#ifndef IPV6_RECVPKTINFO
#define IPV6_RECVPKTINFO IPV6_PKTINFO
#endif
688 689 690 691 692 693
/*
 * Same goes for IPV6_HOPLIMIT -> IPV6_RECVHOPLIMIT.
 */
#ifndef IPV6_RECVHOPLIMIT
#define IPV6_RECVHOPLIMIT IPV6_HOPLIMIT
#endif
694

695 696 697 698
static char *
sysio_register_cmsgs(sock *s)
{
  int ok = 1;
699

700
  if ((s->flags & SKF_LADDR_RX) &&
701
      (setsockopt(s->fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, &ok, sizeof(ok)) < 0))
702 703
    return "IPV6_RECVPKTINFO";

704 705 706 707
  if ((s->flags & SKF_TTL_RX) &&
      (setsockopt(s->fd, IPPROTO_IPV6, IPV6_RECVHOPLIMIT, &ok, sizeof(ok)) < 0))
    return "IPV6_RECVHOPLIMIT";

708 709 710 711 712 713 714 715
  return NULL;
}

static void
sysio_process_rx_cmsgs(sock *s, struct msghdr *msg)
{
  struct cmsghdr *cm;
  struct in6_pktinfo *pi = NULL;
716
  int *hlim = NULL;
717 718

  for (cm = CMSG_FIRSTHDR(msg); cm != NULL; cm = CMSG_NXTHDR(msg, cm))
719 720 721 722 723 724 725 726 727 728 729
  {
    if (cm->cmsg_level == IPPROTO_IPV6 && cm->cmsg_type == IPV6_PKTINFO)
      pi = (struct in6_pktinfo *) CMSG_DATA(cm);

    if (cm->cmsg_level == IPPROTO_IPV6 && cm->cmsg_type == IPV6_HOPLIMIT)
      hlim = (int *) CMSG_DATA(cm);
  }

  if (s->flags & SKF_LADDR_RX)
  {
    if (pi)
730
    {
731 732
      get_inaddr(&s->laddr, &pi->ipi6_addr);
      s->lifindex = pi->ipi6_ifindex;
733
    }
734
    else
735 736 737 738
    {
      s->laddr = IPA_NONE;
      s->lifindex = 0;
    }
739 740 741 742
  }

  if (s->flags & SKF_TTL_RX)
    s->ttl = hlim ? *hlim : -1;
743 744 745 746

  return;
}

Ondřej Zajíček's avatar
Ondřej Zajíček committed
747
/*
748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
static void
sysio_prepare_tx_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
{
  struct cmsghdr *cm;
  struct in6_pktinfo *pi;

  if (!(s->flags & SKF_LADDR_TX))
    return;

  msg->msg_control = cbuf;
  msg->msg_controllen = cbuflen;

  cm = CMSG_FIRSTHDR(msg);
  cm->cmsg_level = IPPROTO_IPV6;
  cm->cmsg_type = IPV6_PKTINFO;
  cm->cmsg_len = CMSG_LEN(sizeof(*pi));

  pi = (struct in6_pktinfo *) CMSG_DATA(cm);
  set_inaddr(&pi->ipi6_addr, s->saddr);
  pi->ipi6_ifindex = s->iface ? s->iface->index : 0;

  msg->msg_controllen = cm->cmsg_len;
  return;
}
Ondřej Zajíček's avatar
Ondřej Zajíček committed
772
*/
773 774
#endif

775 776 777 778
static char *
sk_set_ttl_int(sock *s)
{
#ifdef IPV6
779
  if (setsockopt(s->fd, SOL_IPV6, IPV6_UNICAST_HOPS, &s->ttl, sizeof(s->ttl)) < 0)
780 781 782 783 784
    return "IPV6_UNICAST_HOPS";
#else
  if (setsockopt(s->fd, SOL_IP, IP_TTL, &s->ttl, sizeof(s->ttl)) < 0)
    return "IP_TTL";
#ifdef CONFIG_UNIX_DONTROUTE
785
  int one = 1;
786 787 788 789 790 791 792
  if (s->ttl == 1 && setsockopt(s->fd, SOL_SOCKET, SO_DONTROUTE, &one, sizeof(one)) < 0)
    return "SO_DONTROUTE";
#endif 
#endif
  return NULL;
}

793 794 795
#define ERR(x) do { err = x; goto bad; } while(0)
#define WARN(x) log(L_WARN "sk_setup: %s: %m", x)

796 797 798 799
static char *
sk_setup(sock *s)
{
  int fd = s->fd;
800
  char *err = NULL;
801 802 803

  if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0)
    ERR("fcntl(O_NONBLOCK)");
804 805
  if (s->type == SK_UNIX)
    return NULL;
806 807 808 809 810

#ifdef IPV6
  if ((s->tos >= 0) && setsockopt(fd, SOL_IPV6, IPV6_TCLASS, &s->tos, sizeof(s->tos)) < 0)
    WARN("IPV6_TCLASS");
#else
811
  if ((s->tos >= 0) && setsockopt(fd, SOL_IP, IP_TOS, &s->tos, sizeof(s->tos)) < 0)
812
    WARN("IP_TOS");
813
#endif
814

815 816 817
  if (s->priority >= 0)
    sk_set_priority(s, s->priority);

818 819 820 821 822 823
#ifdef IPV6
  int v = 1;
  if ((s->flags & SKF_V6ONLY) && setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &v, sizeof(v)) < 0)
    WARN("IPV6_V6ONLY");
#endif

824 825
  if ((s->ttl >= 0) && (err = sk_set_ttl_int(s)))
    goto bad;
826

827
  err = sysio_register_cmsgs(s);
828 829 830 831
bad:
  return err;
}

832
/**
833
 * sk_set_ttl - set transmit TTL for given socket.
834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
 * @s: socket
 * @ttl: TTL value
 *
 * Set TTL for already opened connections when TTL was not set before.
 * Useful for accepted connections when different ones should have 
 * different TTL.
 *
 * Result: 0 for success, -1 for an error.
 */

int
sk_set_ttl(sock *s, int ttl)
{
  char *err;

  s->ttl = ttl;
  if (err = sk_set_ttl_int(s))
    log(L_ERR "sk_set_ttl: %s: %m", err);

  return (err ? -1 : 0);
}

856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877
/**
 * sk_set_min_ttl - set minimal accepted TTL for given socket.
 * @s: socket
 * @ttl: TTL value
 *
 * Can be used in TTL security implementation
 *
 * Result: 0 for success, -1 for an error.
 */

int
sk_set_min_ttl(sock *s, int ttl)
{
  int err;
#ifdef IPV6
  err = sk_set_min_ttl6(s, ttl);
#else
  err = sk_set_min_ttl4(s, ttl);
#endif

  return err;
}
878 879 880 881 882

/**
 * sk_set_md5_auth - add / remove MD5 security association for given socket.
 * @s: socket
 * @a: IP address of the other side
883
 * @ifa: Interface for link-local IP address
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
 * @passwd: password used for MD5 authentication
 *
 * In TCP MD5 handling code in kernel, there is a set of pairs
 * (address, password) used to choose password according to
 * address of the other side. This function is useful for
 * listening socket, for active sockets it is enough to set
 * s->password field.
 *
 * When called with passwd != NULL, the new pair is added,
 * When called with passwd == NULL, the existing pair is removed.
 *
 * Result: 0 for success, -1 for an error.
 */

int
899
sk_set_md5_auth(sock *s, ip_addr a, struct iface *ifa, char *passwd)
900 901
{
  sockaddr sa;
902
  fill_in_sockaddr(&sa, a, ifa, 0);
903 904 905
  return sk_set_md5_auth_int(s, &sa, passwd);
}

906 907 908 909
int
sk_set_broadcast(sock *s, int enable)
{
  if (setsockopt(s->fd, SOL_SOCKET, SO_BROADCAST, &enable, sizeof(enable)) < 0)
910 911 912 913 914 915
    {
      log(L_ERR "sk_set_broadcast: SO_BROADCAST: %m");
      return -1;
    }

  return 0;
916 917 918 919 920
}


#ifdef IPV6

921 922 923 924 925 926 927 928 929 930 931 932
int
sk_set_ipv6_checksum(sock *s, int offset)
{
  if (setsockopt(s->fd, IPPROTO_IPV6, IPV6_CHECKSUM, &offset, sizeof(offset)) < 0)
    {
      log(L_ERR "sk_set_ipv6_checksum: IPV6_CHECKSUM: %m");
      return -1;
    }

  return 0;
}

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951
int
sk_set_icmp_filter(sock *s, int p1, int p2)
{
  /* a bit of lame interface, but it is here only for Radv */
  struct icmp6_filter f;

  ICMP6_FILTER_SETBLOCKALL(&f);
  ICMP6_FILTER_SETPASS(p1, &f);
  ICMP6_FILTER_SETPASS(p2, &f);

  if (setsockopt(s->fd, IPPROTO_ICMPV6, ICMP6_FILTER, &f, sizeof(f)) < 0)
    {
      log(L_ERR "sk_setup_icmp_filter: ICMP6_FILTER: %m");
      return -1;
    }

  return 0;
}

952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968
int
sk_setup_multicast(sock *s)
{
  char *err;
  int zero = 0;
  int index;

  ASSERT(s->iface && s->iface->addr);

  index = s->iface->index;
  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_HOPS, &s->ttl, sizeof(s->ttl)) < 0)
    ERR("IPV6_MULTICAST_HOPS");
  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_LOOP, &zero, sizeof(zero)) < 0)
    ERR("IPV6_MULTICAST_LOOP");
  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_IF, &index, sizeof(index)) < 0)
    ERR("IPV6_MULTICAST_IF");

969 970 971
  if (err = sysio_bind_to_iface(s))
    goto bad;

972 973 974 975 976 977 978 979 980 981 982
  return 0;

bad:
  log(L_ERR "sk_setup_multicast: %s: %m", err);
  return -1;
}

int
sk_join_group(sock *s, ip_addr maddr)
{
  struct ipv6_mreq mreq;
983

984 985 986 987 988 989 990 991
  set_inaddr(&mreq.ipv6mr_multiaddr, maddr);

#ifdef CONFIG_IPV6_GLIBC_20
  mreq.ipv6mr_ifindex = s->iface->index;
#else
  mreq.ipv6mr_interface = s->iface->index;
#endif

992
  if (setsockopt(s->fd, SOL_IPV6, IPV6_JOIN_GROUP, &mreq, sizeof(mreq)) < 0)
993
    {
994
      log(L_ERR "sk_join_group: IPV6_JOIN_GROUP: %m");
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
      return -1;
    }

  return 0;
}

int
sk_leave_group(sock *s, ip_addr maddr)
{
  struct ipv6_mreq mreq;
	
  set_inaddr(&mreq.ipv6mr_multiaddr, maddr);

#ifdef CONFIG_IPV6_GLIBC_20
  mreq.ipv6mr_ifindex = s->iface->index;
#else
  mreq.ipv6mr_interface = s->iface->index;
#endif

1014
  if (setsockopt(s->fd, SOL_IPV6, IPV6_LEAVE_GROUP, &mreq, sizeof(mreq)) < 0)
1015
    {
1016
      log(L_ERR "sk_leave_group: IPV6_LEAVE_GROUP: %m");
1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
      return -1;
    }

  return 0;
}

#else /* IPV4 */

int
sk_setup_multicast(sock *s)
{
  char *err;

  ASSERT(s->iface && s->iface->addr);

  if (err = sysio_setup_multicast(s))
    {
      log(L_ERR "sk_setup_multicast: %s: %m", err);
      return -1;
    }

  return 0;
}

int
sk_join_group(sock *s, ip_addr maddr)
{
 char *err;

 if (err = sysio_join_group(s, maddr))
    {
      log(L_ERR "sk_join_group: %s: %m", err);
      return -1;
    }

  return 0;
}

int
sk_leave_group(sock *s, ip_addr maddr)
{
 char *err;

 if (err = sysio_leave_group(s, maddr))
    {
      log(L_ERR "sk_leave_group: %s: %m", err);
      return -1;
    }

  return 0;
}

#endif 

1071

1072
static void
1073 1074
sk_tcp_connected(sock *s)
{
1075 1076 1077
  sockaddr lsa;
  int lsa_len = sizeof(lsa);
  if (getsockname(s->fd, (struct sockaddr *) &lsa, &lsa_len) == 0)
1078
    get_sockaddr(&lsa, &s->saddr, &s->iface, &s->sport, 1);
1079

1080 1081
  s->type = SK_TCP;
  sk_alloc_bufs(s);
1082
  s->tx_hook(s);
1083 1084
}

1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
static int
sk_passive_connected(sock *s, struct sockaddr *sa, int al, int type)
{
  int fd = accept(s->fd, sa, &al);
  if (fd >= 0)
    {
      sock *t = sk_new(s->pool);
      char *err;
      t->type = type;
      t->fd = fd;
1095 1096 1097 1098 1099
      t->ttl = s->ttl;
      t->tos = s->tos;
      t->rbsize = s->rbsize;
      t->tbsize = s->tbsize;
      if (type == SK_TCP)
1100 1101 1102 1103
	{
	  sockaddr lsa;
	  int lsa_len = sizeof(lsa);
	  if (getsockname(fd, (struct sockaddr *) &lsa, &lsa_len) == 0)
1104
	    get_sockaddr(&lsa, &t->saddr, &t->iface, &t->sport, 1);
1105

1106
	  get_sockaddr((sockaddr *) sa, &t->daddr, &t->iface, &t->dport, 1);
1107
	}
1108
      sk_insert(t);
1109 1110 1111
      if (err = sk_setup(t))
	{
	  log(L_ERR "Incoming connection: %s: %m", err);
1112 1113
	  rfree(t);
	  return 1;
1114 1115
	}
      sk_alloc_bufs(t);
1116
      s->rx_hook(t, 0);
1117 1118 1119 1120
      return 1;
    }
  else if (errno != EINTR && errno != EAGAIN)
    {
1121
      s->err_hook(s, errno);
1122 1123 1124 1125
    }
  return 0;
}

1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
/**
 * sk_open - open a socket
 * @s: socket
 *
 * This function takes a socket resource created by sk_new() and
 * initialized by the user and binds a corresponding network connection
 * to it.
 *
 * Result: 0 for success, -1 for an error.
 */
1136 1137 1138
int
sk_open(sock *s)
{
1139
  int fd;
1140
  sockaddr sa;
1141 1142 1143 1144 1145 1146 1147 1148
  int one = 1;
  int type = s->type;
  int has_src = ipa_nonzero(s->saddr) || s->sport;
  char *err;

  switch (type)
    {
    case SK_TCP_ACTIVE:
1149 1150
      s->ttx = "";			/* Force s->ttx != s->tpos */
      /* Fall thru */
1151
    case SK_TCP_PASSIVE:
1152
      fd = socket(BIRD_PF, SOCK_STREAM, IPPROTO_TCP);
1153 1154
      break;
    case SK_UDP:
1155
      fd = socket(BIRD_PF, SOCK_DGRAM, IPPROTO_UDP);
1156 1157
      break;
    case SK_IP:
1158
      fd = socket(BIRD_PF, SOCK_RAW, s->dport);
1159
      break;
1160 1161 1162
    case SK_MAGIC:
      fd = s->fd;
      break;
1163
    default:
1164
      bug("sk_open() called for invalid sock type %d", type);
1165 1166 1167 1168 1169 1170 1171
    }
  if (fd < 0)
    die("sk_open: socket: %m");
  s->fd = fd;

  if (err = sk_setup(s))
    goto bad;
1172

1173 1174 1175 1176
  if (has_src)
    {
      int port;

1177
      if (type == SK_IP)
1178 1179 1180 1181 1182 1183 1184
	port = 0;
      else
	{
	  port = s->sport;
	  if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(one)) < 0)
	    ERR("SO_REUSEADDR");
	}
1185
      fill_in_sockaddr(&sa, s->saddr, s->iface, port);
1186 1187 1188
      if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0)
	ERR("bind");
    }
1189
  fill_in_sockaddr(&sa, s->daddr, s->iface, s->dport);
1190 1191 1192 1193 1194 1195 1196 1197

  if (s->password)
    {
      int rv = sk_set_md5_auth_int(s, &sa, s->password);
      if (rv < 0)
	goto bad_no_log;
    }

1198 1199 1200 1201 1202
  switch (type)
    {
    case SK_TCP_ACTIVE:
      if (connect(fd, (struct sockaddr *) &sa, sizeof(sa)) >= 0)
	sk_tcp_connected(s);
1203
      else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS &&
1204
	       errno != ECONNREFUSED && errno != EHOSTUNREACH && errno != ENETUNREACH)
1205 1206 1207 1208 1209 1210
	ERR("connect");
      break;
    case SK_TCP_PASSIVE:
      if (listen(fd, 8))
	ERR("listen");
      break;
1211 1212 1213
    case SK_MAGIC:
      break;
    default:
1214
      sk_alloc_bufs(s);
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
#ifdef IPV6
#ifdef IPV6_MTU_DISCOVER
      {
	int dont = IPV6_PMTUDISC_DONT;
	if (setsockopt(fd, SOL_IPV6, IPV6_MTU_DISCOVER, &dont, sizeof(dont)) < 0)
	  ERR("IPV6_MTU_DISCOVER");
      }
#endif
#else
#ifdef IP_PMTUDISC
      {
	int dont = IP_PMTUDISC_DONT;
	if (setsockopt(fd, SOL_IP, IP_PMTUDISC, &dont, sizeof(dont)) < 0)
	  ERR("IP_PMTUDISC");
      }
#endif
#endif
1232 1233
    }

1234
  sk_insert(s);
1235 1236 1237 1238
  return 0;

bad:
  log(L_ERR "sk_open: %s: %m", err);
1239
bad_no_log:
1240 1241 1242 1243 1244
  close(fd);
  s->fd = -1;
  return -1;
}

1245
void
1246 1247 1248 1249 1250 1251 1252 1253
sk_open_unix(sock *s, char *name)
{
  int fd;
  struct sockaddr_un sa;
  char *err;

  fd = socket(AF_UNIX, SOCK_STREAM, 0);
  if (fd < 0)
1254
    ERR("socket");
1255 1256 1257 1258
  s->fd = fd;
  if (err = sk_setup(s))
    goto bad;
  unlink(name);
1259

1260
  /* Path length checked in test_old_bird() */
1261
  sa.sun_family = AF_UNIX;
1262
  strcpy(sa.sun_path, name);
1263
  if (bind(fd, (struct sockaddr *) &sa, SUN_LEN(&sa)) < 0)
1264 1265 1266
    ERR("bind");
  if (listen(fd, 8))
    ERR("listen");
1267
  sk_insert(s);
1268
  return;
1269

1270
 bad:
1271
  log(L_ERR "sk_open_unix: %s: %m", err);
1272
  die("Unable to create control socket %s", name);
1273 1274
}

1275 1276
static inline void reset_tx_buffer(sock *s) { s->ttx = s->tpos = s->tbuf; }

1277 1278 1279 1280 1281 1282 1283 1284
static int
sk_maybe_write(sock *s)
{
  int e;

  switch (s->type)
    {
    case SK_TCP:
1285
    case SK_MAGIC:
1286
    case SK_UNIX:
1287 1288 1289 1290 1291 1292 1293
      while (s->ttx != s->tpos)
	{
	  e = write(s->fd, s->ttx, s->tpos - s->ttx);
	  if (e < 0)
	    {
	      if (errno != EINTR && errno != EAGAIN)
		{
1294
		  reset_tx_buffer(s);
1295 1296
		  /* EPIPE is just a connection close notification during TX */
		  s->err_hook(s, (errno != EPIPE) ? errno : 0);
1297 1298 1299 1300 1301 1302
		  return -1;
		}
	      return 0;
	    }
	  s->ttx += e;
	}
1303
      reset_tx_buffer(s);
1304 1305 1306 1307 1308 1309
      return 1;
    case SK_UDP:
    case SK_IP:
      {
	if (s->tbuf == s->tpos)
	  return 1;
1310

1311
	sockaddr sa;
1312
	fill_in_sockaddr(&sa, s->daddr, s->iface, s->dport);
1313 1314

	struct iovec iov = {s->tbuf, s->tpos - s->tbuf};
Ondřej Zajíček's avatar
Ondřej Zajíček committed
1315
	// byte cmsg_buf[CMSG_TX_SPACE];
1316 1317 1318 1319 1320

	struct msghdr msg = {
	  .msg_name = &sa,
	  .msg_namelen = sizeof(sa),
	  .msg_iov = &iov,
1321
	  .msg_iovlen = 1};
1322

Ondřej Zajíček's avatar
Ondřej Zajíček committed
1323
	// sysio_prepare_tx_cmsgs(s, &msg, cmsg_buf, sizeof(cmsg_buf));
1324 1325
	e = sendmsg(s->fd, &msg, 0);

1326 1327 1328 1329
	if (e < 0)
	  {
	    if (errno != EINTR && errno != EAGAIN)
	      {
1330
		reset_tx_buffer(s);
1331
		s->err_hook(s, errno);
1332 1333 1334 1335
		return -1;
	      }
	    return 0;
	  }
1336
	reset_tx_buffer(s);
1337 1338 1339
	return 1;
      }
    default:
1340
      bug("sk_maybe_write: unknown socket type %d", s->type);
1341 1342 1343
    }
}

1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
int
sk_rx_ready(sock *s)
{
  fd_set rd, wr;
  struct timeval timo;
  int rv;

  FD_ZERO(&rd);
  FD_ZERO(&wr);
  FD_SET(s->fd, &rd);

  timo.tv_sec = 0;
  timo.tv_usec = 0;

 redo:
  rv = select(s->fd+1, &rd, &wr, NULL, &timo);
  
  if ((rv < 0) && (errno == EINTR || errno == EAGAIN))
    goto redo;

  return rv;
}

1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
/**
 * sk_send - send data to a socket
 * @s: socket
 * @len: number of bytes to send
 *
 * This function sends @len bytes of data prepared in the
 * transmit buffer of the socket @s to the network connection.
 * If the packet can be sent immediately, it does so and returns
 * 1, else it queues the packet for later processing, returns 0
 * and calls the @tx_hook of the socket when the tranmission
 * takes place.
 */
1379 1380 1381 1382 1383 1384 1385 1386
int
sk_send(sock *s, unsigned len)
{
  s->ttx = s->tbuf;
  s->tpos = s->tbuf + len;
  return sk_maybe_write(s);
}

1387 1388 1389 1390 1391 1392 1393
/**
 * sk_send_to - send data to a specific destination
 * @s: socket
 * @len: number of bytes to send
 * @addr: IP address to send the packet to
 * @port: port to send the packet to
 *
1394
 * This is a sk_send() replacement for connection-less packet sockets
1395 1396
 * which allows destination of the packet to be chosen dynamically.
 */
1397 1398 1399
int
sk_send_to(sock *s, unsigned len, ip_addr addr, unsigned port)
{
1400 1401
  s->daddr = addr;
  s->dport = port;
1402 1403 1404 1405 1406
  s->ttx = s->tbuf;
  s->tpos = s->tbuf + len;
  return sk_maybe_write(s);
}

1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
/*
int
sk_send_full(sock *s, unsigned len, struct iface *ifa,
	     ip_addr saddr, ip_addr daddr, unsigned dport)
{
  s->iface = ifa;
  s->saddr = saddr;
  s->daddr = daddr;
  s->dport = dport;
  s->ttx = s->tbuf;
  s->tpos = s->tbuf + len;
  return sk_maybe_write(s);
}
*/

1422 1423 1424 1425 1426 1427 1428
static int
sk_read(sock *s)
{
  switch (s->type)
    {
    case SK_TCP_PASSIVE:
      {
1429
	sockaddr sa;
1430 1431 1432 1433 1434 1435
	return sk_passive_connected(s, (struct sockaddr *) &sa, sizeof(sa), SK_TCP);
      }
    case SK_UNIX_PASSIVE:
      {
	struct sockaddr_un sa;
	return sk_passive_connected(s, (struct sockaddr *) &sa, sizeof(sa), SK_UNIX);
1436 1437
      }
    case SK_TCP:
1438
    case SK_UNIX:
1439 1440 1441 1442 1443 1444
      {
	int c = read(s->fd, s->rpos, s->rbuf + s->rbsize - s->rpos);

	if (c < 0)
	  {
	    if (errno != EINTR && errno != EAGAIN)
1445
	      s->err_hook(s, errno);
1446 1447
	  }
	else if (!c)
1448
	  s->err_hook(s, 0);
1449 1450 1451 1452
	else
	  {
	    s->rpos += c;
	    if (s->rx_hook(s, s->rpos - s->rbuf))
1453 1454 1455 1456 1457
	      {
		/* We need to be careful since the socket could have been deleted by the hook */
		if (current_sock == s)
		  s->rpos = s->rbuf;
	      }
1458 1459 1460 1461
	    return 1;
	  }
	return 0;
      }
1462 1463
    case SK_MAGIC:
      return s->rx_hook(s, 0);
1464 1465
    default:
      {
1466
	sockaddr sa;
1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
	int e;

	struct iovec iov = {s->rbuf, s->rbsize};
	byte cmsg_buf[CMSG_RX_SPACE];

	struct msghdr msg = {
	  .msg_name = &sa,
	  .msg_namelen = sizeof(sa),
	  .msg_iov = &iov,
	  .msg_iovlen = 1,
	  .msg_control = cmsg_buf,
	  .msg_controllen = sizeof(cmsg_buf),
	  .msg_flags = 0};

	e = recvmsg(s->fd, &msg, 0);
1482 1483 1484 1485

	if (e < 0)
	  {
	    if (errno != EINTR && errno != EAGAIN)
1486
	      s->err_hook(s, errno);
1487 1488 1489
	    return 0;
	  }
	s->rpos = s->rbuf + e;
1490
	get_sockaddr(&sa, &s->faddr, NULL, &s->fport, 1);
1491 1492
	sysio_process_rx_cmsgs(s, &msg);

1493 1494 1495 1496 1497 1498
	s->rx_hook(s, e);
	return 1;
      }
    }
}

1499
static int
1500 1501
sk_write(sock *s)
{
1502 1503 1504 1505 1506
  switch (s->type)
    {
    case SK_TCP_ACTIVE:
      {
	sockaddr sa;
1507
	fill_in_sockaddr(&sa, s->daddr, s->iface, s->dport);
Ondřej Filip's avatar
Ondřej Filip committed
1508
	if (connect(s->fd, (struct sockaddr *) &sa, sizeof(sa)) >= 0 || errno == EISCONN)
1509 1510
	  sk_tcp_connected(s);
	else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS)
1511
	  s->err_hook(s, errno);
1512
	return 0;
1513 1514
      }
    default:
1515 1516 1517 1518 1519 1520
      if (s->ttx != s->tpos && sk_maybe_write(s) > 0)
	{
	  s->tx_hook(s);
	  return 1;
	}
      return 0;
1521
    }
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
}

void
sk_dump_all(void)
{
  node *n;
  sock *s;

  debug("Open sockets:\n");
  WALK_LIST(n, sock_list)
    {
      s = SKIP_BACK(sock, n, n);
      debug("%p ", s);
      sk_dump(&s->r);
    }
  debug("\n");
}

#undef ERR
1541
#undef WARN
1542 1543 1544 1545 1546

/*
 *	Main I/O Loop
 */

1547 1548 1549
volatile int async_config_flag;		/* Asynchronous reconfiguration/dump scheduled */
volatile int async_dump_flag;

1550 1551 1552 1553 1554 1555
void
io_init(void)
{
  init_list(&near_timers);
  init_list(&far_timers);
  init_list(&sock_list);
1556
  init_list(&global_event_list);
1557
  krt_io_init();
1558 1559
  init_times();
  update_times();
1560
  boot_time = now;
1561
  srandom((int) now_real);
1562 1563
}

1564 1565 1566
static int short_loops = 0;
#define SHORT_LOOP_MAX 10

1567 1568 1569 1570 1571 1572
void
io_loop(void)
{
  fd_set rd, wr;
  struct timeval timo;
  time_t tout;
1573
  int hi, events;
1574
  sock *s;
1575
  node *n;
1576

1577
  sock_recalc_fdsets_p = 1;
1578 1579
  for(;;)
    {
1580
      events = ev_run_list(&global_event_list);
1581
      update_times();
1582 1583 1584 1585 1586 1587
      tout = tm_first_shot();
      if (tout <= now)
	{
	  tm_shot();
	  continue;
	}
1588
      timo.tv_sec = events ? 0 : MIN(tout - now, 3);
1589
      timo.tv_usec = 0;
1590

1591 1592 1593 1594 1595 1596 1597
      if (sock_recalc_fdsets_p)
	{
	  sock_recalc_fdsets_p = 0;
	  FD_ZERO(&rd);
	  FD_ZERO(&wr);
	}

1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
      hi = 0;
      WALK_LIST(n, sock_list)
	{
	  s = SKIP_BACK(sock, n, n);
	  if (s->rx_hook)
	    {
	      FD_SET(s->fd, &rd);
	      if (s->fd > hi)
		hi = s->fd;
	    }
1608 1609
	  else
	    FD_CLR(s->fd, &rd);
1610 1611 1612 1613 1614 1615
	  if (s->tx_hook && s->ttx != s->tpos)
	    {
	      FD_SET(s->fd, &wr);
	      if (s->fd > hi)
		hi = s->fd;
	    }
1616 1617
	  else
	    FD_CLR(s->fd, &wr);
1618 1619
	}

1620 1621 1622 1623 1624 1625 1626 1627 1628
      /*
       * Yes, this is racy. But even if the signal comes before this test
       * and entering select(), it gets caught on the next timer tick.
       */

      if (async_config_flag)
	{
	  async_config();
	  async_config_flag = 0;
1629
	  continue;
1630 1631 1632 1633 1634
	}
      if (async_dump_flag)
	{
	  async_dump();
	  async_dump_flag = 0;
1635 1636 1637 1638 1639 1640 1641
	  continue;
	}
      if (async_shutdown_flag)
	{
	  async_shutdown();
	  async_shutdown_flag = 0;
	  continue;
1642 1643 1644
	}

      /* And finally enter select() to find active sockets */
1645
      hi = select(hi+1, &rd, &wr, NULL, &timo);
1646

1647 1648 1649 1650 1651 1652 1653 1654
      if (hi < 0)
	{
	  if (errno == EINTR || errno == EAGAIN)
	    continue;
	  die("select: %m");
	}
      if (hi)
	{
1655 1656 1657
	  /* guaranteed to be non-empty */
	  current_sock = SKIP_BACK(sock, n, HEAD(sock_list));

1658
	  while (current_sock)
1659
	    {
1660 1661
	      sock *s = current_sock;
	      int e;
1662 1663 1664 1665
	      int steps;

	      steps = MAX_STEPS;
	      if ((s->type >= SK_MAGIC) && FD_ISSET(s->fd, &rd) && s->rx_hook)
1666 1667
		do
		  {
1668
		    steps--;
1669 1670 1671 1672
		    e = sk_read(s);
		    if (s != current_sock)
		      goto next;
		  }
1673 1674 1675
		while (e && s->rx_hook && steps);

	      steps = MAX_STEPS;
1676 1677 1678
	      if (FD_ISSET(s->fd, &wr))
		do
		  {
1679
		    steps--;
1680 1681 1682 1683
		    e = sk_write(s);
		    if (s != current_sock)
		      goto next;
		  }
1684
		while (e && steps);
1685 1686
	      current_sock = sk_next(s);
	    next: ;
1687