1 /* Host name resolution and matching.
2 Copyright (C) 1996-2005 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 Foundation, Inc.,
18 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 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. */
38 # include <sys/socket.h>
39 # include <netinet/in.h>
41 # include <arpa/inet.h>
44 # define SET_H_ERRNO(err) ((void)(h_errno = (err)))
46 # define SET_H_ERRNO(err) WSASetLastError (err)
58 # define NO_ADDRESS NO_DATA
61 /* Lists of IP addresses that result from running DNS queries. See
62 lookup_host for details. */
65 int count; /* number of adrresses */
66 ip_address *addresses; /* pointer to the string of addresses */
68 int faulty; /* number of addresses known not to work. */
69 bool connected; /* whether we were able to connect to
70 one of the addresses in the list,
73 int refcount; /* reference count; when it drops to
74 0, the entry is freed. */
77 /* Get the bounds of the address list. */
80 address_list_get_bounds (const struct address_list *al, int *start, int *end)
86 /* Return a pointer to the address at position POS. */
89 address_list_address_at (const struct address_list *al, int pos)
91 assert (pos >= al->faulty && pos < al->count);
92 return al->addresses + pos;
95 /* Return true if AL contains IP, false otherwise. */
98 address_list_contains (const struct address_list *al, const ip_address *ip)
104 for (i = 0; i < al->count; i++)
106 ip_address *cur = al->addresses + i;
107 if (cur->type == IPV4_ADDRESS
108 && (ADDRESS_IPV4_IN_ADDR (cur).s_addr
110 ADDRESS_IPV4_IN_ADDR (ip).s_addr))
116 for (i = 0; i < al->count; i++)
118 ip_address *cur = al->addresses + i;
119 if (cur->type == IPV6_ADDRESS
120 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
121 && ADDRESS_IPV6_SCOPE (cur) == ADDRESS_IPV6_SCOPE (ip)
123 && IN6_ARE_ADDR_EQUAL (&ADDRESS_IPV6_IN6_ADDR (cur),
124 &ADDRESS_IPV6_IN6_ADDR (ip)))
128 #endif /* ENABLE_IPV6 */
134 /* Mark the INDEXth element of AL as faulty, so that the next time
135 this address list is used, the faulty element will be skipped. */
138 address_list_set_faulty (struct address_list *al, int index)
140 /* We assume that the address list is traversed in order, so that a
141 "faulty" attempt is always preceded with all-faulty addresses,
142 and this is how Wget uses it. */
143 assert (index == al->faulty);
146 if (al->faulty >= al->count)
147 /* All addresses have been proven faulty. Since there's not much
148 sense in returning the user an empty address list the next
149 time, we'll rather make them all clean, so that they can be
154 /* Set the "connected" flag to true. This flag used by connect.c to
155 see if the host perhaps needs to be resolved again. */
158 address_list_set_connected (struct address_list *al)
160 al->connected = true;
163 /* Return the value of the "connected" flag. */
166 address_list_connected_p (const struct address_list *al)
168 return al->connected;
173 /* Create an address_list from the addresses in the given struct
176 static struct address_list *
177 address_list_from_addrinfo (const struct addrinfo *ai)
179 struct address_list *al;
180 const struct addrinfo *ptr;
185 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
186 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
191 al = xnew0 (struct address_list);
192 al->addresses = xnew_array (ip_address, cnt);
197 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
198 if (ptr->ai_family == AF_INET6)
200 const struct sockaddr_in6 *sin6 =
201 (const struct sockaddr_in6 *)ptr->ai_addr;
202 ip->type = IPV6_ADDRESS;
203 ADDRESS_IPV6_IN6_ADDR (ip) = sin6->sin6_addr;
204 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
205 ADDRESS_IPV6_SCOPE (ip) = sin6->sin6_scope_id;
209 else if (ptr->ai_family == AF_INET)
211 const struct sockaddr_in *sin =
212 (const struct sockaddr_in *)ptr->ai_addr;
213 ip->type = IPV4_ADDRESS;
214 ADDRESS_IPV4_IN_ADDR (ip) = sin->sin_addr;
217 assert (ip - al->addresses == cnt);
221 #define IS_IPV4(addr) (((const ip_address *) addr)->type == IPV4_ADDRESS)
223 /* Compare two IP addresses by type, giving preference to the IPv4
224 address (sorting it first). In other words, return -1 if ADDR1 is
225 IPv4 and ADDR2 is IPv6, +1 if ADDR1 is IPv6 and ADDR2 is IPv4, and
228 This is intended to be used as the comparator arg to a qsort-like
229 sorting function, which is why it accepts generic pointers. */
232 cmp_prefer_ipv4 (const void *addr1, const void *addr2)
234 return !IS_IPV4 (addr1) - !IS_IPV4 (addr2);
237 #define IS_IPV6(addr) (((const ip_address *) addr)->type == IPV6_ADDRESS)
239 /* Like the above, but give preference to the IPv6 address. */
242 cmp_prefer_ipv6 (const void *addr1, const void *addr2)
244 return !IS_IPV6 (addr1) - !IS_IPV6 (addr2);
247 #else /* not ENABLE_IPV6 */
249 /* Create an address_list from a NULL-terminated vector of IPv4
250 addresses. This kind of vector is returned by gethostbyname. */
252 static struct address_list *
253 address_list_from_ipv4_addresses (char **vec)
256 struct address_list *al = xnew0 (struct address_list);
263 al->addresses = xnew_array (ip_address, count);
267 for (i = 0; i < count; i++)
269 ip_address *ip = &al->addresses[i];
270 ip->type = IPV4_ADDRESS;
271 memcpy (ADDRESS_IPV4_DATA (ip), vec[i], 4);
277 #endif /* not ENABLE_IPV6 */
280 address_list_delete (struct address_list *al)
282 xfree (al->addresses);
286 /* Mark the address list as being no longer in use. This will reduce
287 its reference count which will cause the list to be freed when the
291 address_list_release (struct address_list *al)
294 DEBUGP (("Releasing 0x%0*lx (new refcount %d).\n", PTR_FORMAT (al),
296 if (al->refcount <= 0)
298 DEBUGP (("Deleting unused 0x%0*lx.\n", PTR_FORMAT (al)));
299 address_list_delete (al);
303 /* Versions of gethostbyname and getaddrinfo that support timeout. */
307 struct ghbnwt_context {
308 const char *host_name;
309 struct hostent *hptr;
313 gethostbyname_with_timeout_callback (void *arg)
315 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
316 ctx->hptr = gethostbyname (ctx->host_name);
319 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
320 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
321 The function makes sure that when NULL is returned for reasons
322 other than timeout, errno is reset. */
324 static struct hostent *
325 gethostbyname_with_timeout (const char *host_name, double timeout)
327 struct ghbnwt_context ctx;
328 ctx.host_name = host_name;
329 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
331 SET_H_ERRNO (HOST_NOT_FOUND);
340 /* Print error messages for host errors. */
342 host_errstr (int error)
344 /* Can't use switch since some of these constants can be equal,
345 which makes the compiler complain about duplicate case
347 if (error == HOST_NOT_FOUND
348 || error == NO_RECOVERY
350 || error == NO_ADDRESS)
351 return _("Unknown host");
352 else if (error == TRY_AGAIN)
353 /* Message modeled after what gai_strerror returns in similar
355 return _("Temporary failure in name resolution");
357 return _("Unknown error");
360 #else /* ENABLE_IPV6 */
362 struct gaiwt_context {
365 const struct addrinfo *hints;
366 struct addrinfo **res;
371 getaddrinfo_with_timeout_callback (void *arg)
373 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
374 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
377 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
378 In case of timeout, the EAI_SYSTEM error code is returned and errno
379 is set to ETIMEDOUT. */
382 getaddrinfo_with_timeout (const char *node, const char *service,
383 const struct addrinfo *hints, struct addrinfo **res,
386 struct gaiwt_context ctx;
388 ctx.service = service;
392 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
397 return ctx.exit_code;
400 #endif /* ENABLE_IPV6 */
402 /* Return a textual representation of ADDR, i.e. the dotted quad for
403 IPv4 addresses, and the colon-separated list of hex words (with all
404 zeros omitted, etc.) for IPv6 addresses. */
407 print_address (const ip_address *addr)
412 return inet_ntoa (ADDRESS_IPV4_IN_ADDR (addr));
417 if (!inet_ntop (AF_INET6, &ADDRESS_IPV6_IN6_ADDR (addr),
419 snprintf (buf, sizeof buf, "<error: %s>", strerror (errno));
420 buf[sizeof (buf) - 1] = '\0';
428 /* The following two functions were adapted from glibc. */
431 is_valid_ipv4_address (const char *str, const char *end)
433 bool saw_digit = false;
441 if (ch >= '0' && ch <= '9')
443 val = val * 10 + (ch - '0');
454 else if (ch == '.' && saw_digit)
471 is_valid_ipv6_address (const char *str, const char *end)
473 /* Use lower-case for these to avoid clash with system headers. */
492 /* Leading :: requires some special handling. */
496 if (str == end || *str != ':')
508 /* if ch is a number, add it to val. */
512 val |= XDIGIT_TO_NUM (ch);
519 /* if ch is a colon ... */
532 if (tp > ns_in6addrsz - ns_int16sz)
540 /* if ch is a dot ... */
541 if (ch == '.' && (tp <= ns_in6addrsz - ns_inaddrsz)
542 && is_valid_ipv4_address (curtok, end) == 1)
554 if (tp > ns_in6addrsz - ns_int16sz)
561 if (tp == ns_in6addrsz)
566 if (tp != ns_in6addrsz)
572 /* Simple host cache, used by lookup_host to speed up resolving. The
573 cache doesn't handle TTL because Wget is a fairly short-lived
574 application. Refreshing is attempted when connect fails, though --
575 see connect_to_host. */
577 /* Mapping between known hosts and to lists of their addresses. */
578 static struct hash_table *host_name_addresses_map;
581 /* Return the host's resolved addresses from the cache, if
584 static struct address_list *
585 cache_query (const char *host)
587 struct address_list *al;
588 if (!host_name_addresses_map)
590 al = hash_table_get (host_name_addresses_map, host);
593 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
600 /* Cache the DNS lookup of HOST. Subsequent invocations of
601 lookup_host will return the cached value. */
604 cache_store (const char *host, struct address_list *al)
606 if (!host_name_addresses_map)
607 host_name_addresses_map = make_nocase_string_hash_table (0);
610 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
615 debug_logprintf ("Caching %s =>", host);
616 for (i = 0; i < al->count; i++)
617 debug_logprintf (" %s", print_address (al->addresses + i));
618 debug_logprintf ("\n");
622 /* Remove HOST from the DNS cache. Does nothing is HOST is not in
626 cache_remove (const char *host)
628 struct address_list *al;
629 if (!host_name_addresses_map)
631 al = hash_table_get (host_name_addresses_map, host);
634 address_list_release (al);
635 hash_table_remove (host_name_addresses_map, host);
639 /* Look up HOST in DNS and return a list of IP addresses.
641 This function caches its result so that, if the same host is passed
642 the second time, the addresses are returned without DNS lookup.
643 (Use LH_REFRESH to force lookup, or set opt.dns_cache to 0 to
644 globally disable caching.)
646 The order of the returned addresses is affected by the setting of
647 opt.prefer_family: if it is set to prefer_ipv4, IPv4 addresses are
648 placed at the beginning; if it is prefer_ipv6, IPv6 ones are placed
649 at the beginning; otherwise, the order is left intact. The
650 relative order of addresses with the same family is left
651 undisturbed in either case.
653 FLAGS can be a combination of:
654 LH_SILENT - don't print the "resolving ... done" messages.
655 LH_BIND - resolve addresses for use with bind, which under
656 IPv6 means to use AI_PASSIVE flag to getaddrinfo.
657 Passive lookups are not cached under IPv6.
658 LH_REFRESH - if HOST is cached, remove the entry from the cache
659 and resolve it anew. */
661 struct address_list *
662 lookup_host (const char *host, int flags)
664 struct address_list *al;
665 bool silent = !!(flags & LH_SILENT);
667 bool numeric_address = false;
668 double timeout = opt.dns_timeout;
671 /* If we're not using getaddrinfo, first check if HOST specifies a
672 numeric IPv4 address. Some implementations of gethostbyname
673 (e.g. the Ultrix one and possibly Winsock) don't accept
674 dotted-decimal IPv4 addresses. */
676 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
677 if (addr_ipv4 != (uint32_t) -1)
679 /* No need to cache host->addr relation, just return the
682 vec[0] = (char *)&addr_ipv4;
684 return address_list_from_ipv4_addresses (vec);
687 #else /* ENABLE_IPV6 */
688 /* If we're using getaddrinfo, at least check whether the address is
689 already numeric, in which case there is no need to print the
690 "Resolving..." output. (This comes at no additional cost since
691 the is_valid_ipv*_address are already required for
694 const char *end = host + strlen (host);
695 if (is_valid_ipv4_address (host, end) || is_valid_ipv6_address (host, end))
696 numeric_address = true;
700 /* Cache is normally on, but can be turned off with --no-dns-cache.
701 Don't cache passive lookups under IPv6. */
702 use_cache = opt.dns_cache;
704 if ((flags & LH_BIND) || numeric_address)
708 /* Try to find the host in the cache so we don't need to talk to the
709 resolver. If LH_REFRESH is requested, remove HOST from the cache
713 if (!(flags & LH_REFRESH))
715 al = cache_query (host);
723 /* No luck with the cache; resolve HOST. */
725 if (!silent && !numeric_address)
726 logprintf (LOG_VERBOSE, _("Resolving %s... "), escnonprint (host));
731 struct addrinfo hints, *res;
734 hints.ai_socktype = SOCK_STREAM;
736 hints.ai_family = AF_INET;
737 else if (opt.ipv6_only)
738 hints.ai_family = AF_INET6;
740 /* We tried using AI_ADDRCONFIG, but removed it because: it
741 misinterprets IPv6 loopbacks, it is broken on AIX 5.1, and
742 it's unneeded since we sort the addresses anyway. */
743 hints.ai_family = AF_UNSPEC;
746 hints.ai_flags |= AI_PASSIVE;
748 #ifdef AI_NUMERICHOST
751 /* Where available, the AI_NUMERICHOST hint can prevent costly
752 access to DNS servers. */
753 hints.ai_flags |= AI_NUMERICHOST;
754 timeout = 0; /* no timeout needed when "resolving"
755 numeric hosts -- avoid setting up
756 signal handlers and such. */
760 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, timeout);
761 if (err != 0 || res == NULL)
764 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
765 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
768 al = address_list_from_addrinfo (res);
772 logprintf (LOG_VERBOSE,
773 _("failed: No IPv4/IPv6 addresses for host.\n"));
777 /* Reorder addresses so that IPv4 ones (or IPv6 ones, as per
778 --prefer-family) come first. Sorting is stable so the order of
779 the addresses with the same family is undisturbed. */
780 if (al->count > 1 && opt.prefer_family != prefer_none)
781 stable_sort (al->addresses, al->count, sizeof (ip_address),
782 opt.prefer_family == prefer_ipv4
783 ? cmp_prefer_ipv4 : cmp_prefer_ipv6);
785 #else /* not ENABLE_IPV6 */
787 struct hostent *hptr = gethostbyname_with_timeout (host, timeout);
792 if (errno != ETIMEDOUT)
793 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
794 host_errstr (h_errno));
796 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
800 /* Do older systems have h_addr_list? */
801 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
803 #endif /* not ENABLE_IPV6 */
805 /* Print the addresses determined by DNS lookup, but no more than
807 if (!silent && !numeric_address)
810 int printmax = al->count <= 3 ? al->count : 3;
811 for (i = 0; i < printmax; i++)
813 logputs (LOG_VERBOSE, print_address (al->addresses + i));
814 if (i < printmax - 1)
815 logputs (LOG_VERBOSE, ", ");
817 if (printmax != al->count)
818 logputs (LOG_VERBOSE, ", ...");
819 logputs (LOG_VERBOSE, "\n");
822 /* Cache the lookup information. */
824 cache_store (host, al);
829 /* Determine whether a URL is acceptable to be followed, according to
830 a list of domains to accept. */
832 accept_domain (struct url *u)
834 assert (u->host != NULL);
837 if (!sufmatch ((const char **)opt.domains, u->host))
840 if (opt.exclude_domains)
842 if (sufmatch ((const char **)opt.exclude_domains, u->host))
848 /* Check whether WHAT is matched in LIST, each element of LIST being a
849 pattern to match WHAT against, using backward matching (see
850 match_backwards() in utils.c).
852 If an element of LIST matched, 1 is returned, 0 otherwise. */
854 sufmatch (const char **list, const char *what)
859 for (i = 0; list[i]; i++)
861 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
862 if (TOLOWER (list[i][j]) != TOLOWER (what[k]))
864 /* The domain must be first to reach to beginning. */
872 host_cleanup_mapper (void *key, void *value, void *arg_ignored)
874 struct address_list *al;
876 xfree (key); /* host */
878 al = (struct address_list *)value;
879 assert (al->refcount == 1);
880 address_list_delete (al);
888 if (host_name_addresses_map)
890 hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
891 hash_table_destroy (host_name_addresses_map);
892 host_name_addresses_map = NULL;