1 /* Host name resolution and matching.
2 Copyright (C) 1995, 1996, 1997, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
44 #include <sys/types.h>
48 # define SET_H_ERRNO(err) WSASetLastError (err)
50 # include <sys/socket.h>
51 # include <netinet/in.h>
53 # include <arpa/inet.h>
56 # define SET_H_ERRNO(err) ((void)(h_errno = (err)))
60 #define NO_ADDRESS NO_DATA
63 #ifdef HAVE_SYS_UTSNAME_H
64 # include <sys/utsname.h>
84 /* Mapping between known hosts and to lists of their addresses. */
86 static struct hash_table *host_name_addresses_map;
88 /* Lists of addresses. This should eventually be extended to handle
92 int count; /* number of adrresses */
93 ip_address *addresses; /* pointer to the string of addresses */
95 int faulty; /* number of addresses known not to work. */
96 int from_cache; /* whether this entry was pulled from
97 cache or freshly looked up. */
99 int refcount; /* reference count; when it drops to
100 0, the entry is freed. */
103 /* Get the bounds of the address list. */
106 address_list_get_bounds (const struct address_list *al, int *start, int *end)
112 /* Return whether this address list entry has been obtained from the
116 address_list_cached_p (const struct address_list *al)
118 return al->from_cache;
121 /* Return a pointer to the address at position POS. */
124 address_list_address_at (const struct address_list *al, int pos)
126 assert (pos >= al->faulty && pos < al->count);
127 return al->addresses + pos;
130 /* Check whether two address lists have all their IPs in common. */
133 address_list_match_all (const struct address_list *al1,
134 const struct address_list *al2)
141 if (al1->count != al2->count)
144 /* For the comparison to be complete, we'd need to sort the IP
145 addresses first. But that's not necessary because this is only
146 used as an optimization. */
149 /* In the non-IPv6 case, there is only one address type, so we can
150 compare the whole array with memcmp. */
151 return 0 == memcmp (al1->addresses, al2->addresses,
152 al1->count * sizeof (ip_address));
153 #else /* ENABLE_IPV6 */
154 for (i = 0; i < al1->count; ++i)
156 const ip_address *ip1 = &al1->addresses[i];
157 const ip_address *ip2 = &al2->addresses[i];
159 if (ip1->type != ip2->type)
165 if (ADDRESS_IPV4_IN_ADDR (ip1).s_addr
167 ADDRESS_IPV4_IN_ADDR (ip2).s_addr)
171 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
172 if (ADDRESS_IPV6_SCOPE (ip1) != ADDRESS_IPV6_SCOPE (ip2))
174 #endif /* HAVE_SOCKADDR_IN6_SCOPE_ID */
175 if (!IN6_ARE_ADDR_EQUAL (&ADDRESS_IPV6_IN6_ADDR (ip1),
176 &ADDRESS_IPV6_IN6_ADDR (ip2)))
184 #endif /* ENABLE_IPV6 */
187 /* Mark the INDEXth element of AL as faulty, so that the next time
188 this address list is used, the faulty element will be skipped. */
191 address_list_set_faulty (struct address_list *al, int index)
193 /* We assume that the address list is traversed in order, so that a
194 "faulty" attempt is always preceded with all-faulty addresses,
195 and this is how Wget uses it. */
196 assert (index == al->faulty);
199 if (al->faulty >= al->count)
200 /* All addresses have been proven faulty. Since there's not much
201 sense in returning the user an empty address list the next
202 time, we'll rather make them all clean, so that they can be
209 * address_list_from_addrinfo
211 * This function transform an addrinfo links list in and address_list.
214 * addrinfo* Linked list of addrinfo
217 * address_list* New allocated address_list
219 static struct address_list *
220 address_list_from_addrinfo (const struct addrinfo *ai)
222 struct address_list *al;
223 const struct addrinfo *ptr;
228 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
229 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
234 al = xnew0 (struct address_list);
235 al->addresses = xnew_array (ip_address, cnt);
240 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
241 if (ptr->ai_family == AF_INET6)
243 const struct sockaddr_in6 *sin6 =
244 (const struct sockaddr_in6 *)ptr->ai_addr;
245 ip->type = IPV6_ADDRESS;
246 ADDRESS_IPV6_IN6_ADDR (ip) = sin6->sin6_addr;
247 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
248 ADDRESS_IPV6_SCOPE (ip) = sin6->sin6_scope_id;
252 else if (ptr->ai_family == AF_INET)
254 const struct sockaddr_in *sin =
255 (const struct sockaddr_in *)ptr->ai_addr;
256 ip->type = IPV4_ADDRESS;
257 ADDRESS_IPV4_IN_ADDR (ip) = sin->sin_addr;
260 assert (ip - al->addresses == cnt);
264 /* Create an address_list from a NULL-terminated vector of IPv4
265 addresses. This kind of vector is returned by gethostbyname. */
267 static struct address_list *
268 address_list_from_ipv4_addresses (char **vec)
271 struct address_list *al = xnew0 (struct address_list);
278 al->addresses = xnew_array (ip_address, count);
282 for (i = 0; i < count; i++)
284 ip_address *ip = &al->addresses[i];
285 ip->type = IPV4_ADDRESS;
286 memcpy (ADDRESS_IPV4_DATA (ip), vec[i], 4);
294 address_list_delete (struct address_list *al)
296 xfree (al->addresses);
301 address_list_release (struct address_list *al)
304 DEBUGP (("Releasing %p (new refcount %d).\n", al, al->refcount));
305 if (al->refcount <= 0)
307 DEBUGP (("Deleting unused %p.\n", al));
308 address_list_delete (al);
312 /* Versions of gethostbyname and getaddrinfo that support timeout. */
316 struct ghbnwt_context {
317 const char *host_name;
318 struct hostent *hptr;
322 gethostbyname_with_timeout_callback (void *arg)
324 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
325 ctx->hptr = gethostbyname (ctx->host_name);
328 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
329 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
330 The function makes sure that when NULL is returned for reasons
331 other than timeout, errno is reset. */
333 static struct hostent *
334 gethostbyname_with_timeout (const char *host_name, double timeout)
336 struct ghbnwt_context ctx;
337 ctx.host_name = host_name;
338 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
340 SET_H_ERRNO (HOST_NOT_FOUND);
349 #else /* ENABLE_IPV6 */
351 struct gaiwt_context {
354 const struct addrinfo *hints;
355 struct addrinfo **res;
360 getaddrinfo_with_timeout_callback (void *arg)
362 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
363 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
366 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
367 In case of timeout, the EAI_SYSTEM error code is returned and errno
368 is set to ETIMEDOUT. */
371 getaddrinfo_with_timeout (const char *node, const char *service,
372 const struct addrinfo *hints, struct addrinfo **res,
375 struct gaiwt_context ctx;
377 ctx.service = service;
381 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
386 return ctx.exit_code;
389 #endif /* ENABLE_IPV6 */
391 /* Pretty-print ADDR. When compiled without IPv6, this is the same as
392 inet_ntoa. With IPv6, it either prints an IPv6 address or an IPv4
396 pretty_print_address (const ip_address *addr)
401 return inet_ntoa (ADDRESS_IPV4_IN_ADDR (addr));
405 static char buf[128];
406 inet_ntop (AF_INET6, &ADDRESS_IPV6_IN6_ADDR (addr), buf, sizeof (buf));
408 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
410 /* append "%SCOPE_ID" for all ?non-global? addresses */
411 char *p = buf + strlen (buf);
413 number_to_string (p, ADDRESS_IPV6_SCOPE (addr));
417 buf[sizeof (buf) - 1] = '\0';
426 /* Add host name HOST with the address ADDR_TEXT to the cache.
427 ADDR_LIST is a NULL-terminated list of addresses, as in struct
431 cache_host_lookup (const char *host, struct address_list *al)
433 if (!host_name_addresses_map)
434 host_name_addresses_map = make_nocase_string_hash_table (0);
437 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
443 debug_logprintf ("Caching %s =>", host);
444 for (i = 0; i < al->count; i++)
445 debug_logprintf (" %s", pretty_print_address (al->addresses + i));
446 debug_logprintf ("\n");
451 /* Remove HOST from Wget's DNS cache. Does nothing is HOST is not in
455 forget_host_lookup (const char *host)
457 struct address_list *al = hash_table_get (host_name_addresses_map, host);
460 address_list_release (al);
461 hash_table_remove (host_name_addresses_map, host);
465 /* Look up HOST in DNS and return a list of IP addresses.
467 This function caches its result so that, if the same host is passed
468 the second time, the addresses are returned without the DNS lookup.
469 If you want to force lookup, call forget_host_lookup() prior to
470 this function, or set opt.dns_cache to 0 to globally disable
473 FLAGS can be a combination of:
474 LH_SILENT - don't print the "resolving ... done" message.
475 LH_IPV4_ONLY - return only IPv4 addresses.
476 LH_IPV6_ONLY - return only IPv6 addresses. */
478 struct address_list *
479 lookup_host (const char *host, int flags)
481 struct address_list *al = NULL;
485 struct addrinfo hints, *res;
488 hints.ai_socktype = SOCK_STREAM;
490 /* Should we inspect opt.<something> directly? */
491 if (flags & LH_IPV4_ONLY)
492 hints.ai_family = AF_INET;
493 else if (flags & LH_IPV6_ONLY)
494 hints.ai_family = AF_INET6;
496 hints.ai_family = AF_UNSPEC;
499 /* First, try to check whether the address is already a numeric
500 address, in which case we don't want to cache it or bother with
501 setting up timeouts. Plus, old (e.g. Ultrix) implementations of
502 gethostbyname can't handle numeric addresses (!).
504 Where getaddrinfo is available, we do it using the AI_NUMERICHOST
505 flag. Without IPv6, we use inet_addr. */
508 hints.ai_flags = AI_NUMERICHOST;
509 if (flags & LH_PASSIVE)
510 hints.ai_flags |= AI_PASSIVE;
512 /* No need to specify timeout, as we're not resolving HOST, but
513 merely translating it from the presentation (ASCII) to network
515 err = getaddrinfo (host, NULL, &hints, &res);
516 if (err == 0 && res != NULL)
518 al = address_list_from_addrinfo (res);
524 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
525 if (addr_ipv4 != (uint32_t) -1)
527 /* The return value of inet_addr is in network byte order, so
528 we can just copy it to IP. */
530 vec[0] = (char *)&addr_ipv4;
532 return address_list_from_ipv4_addresses (vec);
537 /* Then, try to find the host in the cache. */
539 if (host_name_addresses_map)
541 al = hash_table_get (host_name_addresses_map, host);
544 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
551 if (!(flags & LH_SILENT))
552 logprintf (LOG_VERBOSE, _("Resolving %s... "), host);
554 /* Host name lookup goes on below. */
558 if (flags & LH_PASSIVE)
559 hints.ai_flags |= AI_PASSIVE;
561 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, opt.dns_timeout);
562 if (err != 0 || res == NULL)
564 if (!(flags & LH_SILENT))
565 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
566 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
569 al = address_list_from_addrinfo (res);
573 struct hostent *hptr = gethostbyname_with_timeout (host, opt.dns_timeout);
576 if (!(flags & LH_SILENT))
578 if (errno != ETIMEDOUT)
579 logprintf (LOG_VERBOSE, _("failed: %s.\n"), herrmsg (h_errno));
581 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
585 assert (hptr->h_length == 4);
586 /* Do older systems have h_addr_list? */
587 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
591 /* Print the addresses determined by DNS lookup, but no more than
593 if (!(flags & LH_SILENT))
596 int printmax = al->count <= 3 ? al->count : 3;
597 for (i = 0; i < printmax; i++)
599 logprintf (LOG_VERBOSE, "%s",
600 pretty_print_address (al->addresses + i));
601 if (i < printmax - 1)
602 logputs (LOG_VERBOSE, ", ");
604 if (printmax != al->count)
605 logputs (LOG_VERBOSE, ", ...");
606 logputs (LOG_VERBOSE, "\n");
609 /* Cache the lookup information. */
611 cache_host_lookup (host, al);
616 /* Determine whether a URL is acceptable to be followed, according to
617 a list of domains to accept. */
619 accept_domain (struct url *u)
621 assert (u->host != NULL);
624 if (!sufmatch ((const char **)opt.domains, u->host))
627 if (opt.exclude_domains)
629 if (sufmatch ((const char **)opt.exclude_domains, u->host))
635 /* Check whether WHAT is matched in LIST, each element of LIST being a
636 pattern to match WHAT against, using backward matching (see
637 match_backwards() in utils.c).
639 If an element of LIST matched, 1 is returned, 0 otherwise. */
641 sufmatch (const char **list, const char *what)
646 for (i = 0; list[i]; i++)
648 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
649 if (TOLOWER (list[i][j]) != TOLOWER (what[k]))
651 /* The domain must be first to reach to beginning. */
658 /* Print error messages for host errors. */
662 /* Can't use switch since some constants are equal (at least on my
663 system), and the compiler signals "duplicate case value". */
664 if (error == HOST_NOT_FOUND
665 || error == NO_RECOVERY
667 || error == NO_ADDRESS
668 || error == TRY_AGAIN)
669 return _("Host not found");
671 return _("Unknown error");
675 host_cleanup_mapper (void *key, void *value, void *arg_ignored)
677 struct address_list *al;
679 xfree (key); /* host */
681 al = (struct address_list *)value;
682 assert (al->refcount == 1);
683 address_list_delete (al);
691 if (host_name_addresses_map)
693 hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
694 hash_table_destroy (host_name_addresses_map);
695 host_name_addresses_map = NULL;