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>
47 # include <sys/socket.h>
48 # include <netinet/in.h>
50 # include <arpa/inet.h>
53 # define SET_H_ERRNO(err) ((void)(h_errno = (err)))
55 # define SET_H_ERRNO(err) WSASetLastError (err)
65 #include "connect.h" /* for socket_has_inet6 */
78 # define NO_ADDRESS NO_DATA
81 /* Lists of IP addresses that result from running DNS queries. See
82 lookup_host for details. */
85 int count; /* number of adrresses */
86 ip_address *addresses; /* pointer to the string of addresses */
88 int faulty; /* number of addresses known not to work. */
89 int connected; /* whether we were able to connect to
90 one of the addresses in the list,
93 int refcount; /* reference count; when it drops to
94 0, the entry is freed. */
97 /* Get the bounds of the address list. */
100 address_list_get_bounds (const struct address_list *al, int *start, int *end)
106 /* Return a pointer to the address at position POS. */
109 address_list_address_at (const struct address_list *al, int pos)
111 assert (pos >= al->faulty && pos < al->count);
112 return al->addresses + pos;
115 /* Return non-zero if AL contains IP, zero otherwise. */
118 address_list_contains (const struct address_list *al, const ip_address *ip)
124 for (i = 0; i < al->count; i++)
126 ip_address *cur = al->addresses + i;
127 if (cur->type == IPV4_ADDRESS
128 && (ADDRESS_IPV4_IN_ADDR (cur).s_addr
130 ADDRESS_IPV4_IN_ADDR (ip).s_addr))
136 for (i = 0; i < al->count; i++)
138 ip_address *cur = al->addresses + i;
139 if (cur->type == IPV6_ADDRESS
140 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
141 && ADDRESS_IPV6_SCOPE (cur) == ADDRESS_IPV6_SCOPE (ip)
143 && IN6_ARE_ADDR_EQUAL (&ADDRESS_IPV6_IN6_ADDR (cur),
144 &ADDRESS_IPV6_IN6_ADDR (ip)))
148 #endif /* ENABLE_IPV6 */
154 /* Mark the INDEXth element of AL as faulty, so that the next time
155 this address list is used, the faulty element will be skipped. */
158 address_list_set_faulty (struct address_list *al, int index)
160 /* We assume that the address list is traversed in order, so that a
161 "faulty" attempt is always preceded with all-faulty addresses,
162 and this is how Wget uses it. */
163 assert (index == al->faulty);
166 if (al->faulty >= al->count)
167 /* All addresses have been proven faulty. Since there's not much
168 sense in returning the user an empty address list the next
169 time, we'll rather make them all clean, so that they can be
174 /* Set the "connected" flag to true. This flag used by connect.c to
175 see if the host perhaps needs to be resolved again. */
178 address_list_set_connected (struct address_list *al)
183 /* Return the value of the "connected" flag. */
186 address_list_connected_p (const struct address_list *al)
188 return al->connected;
193 /* Create an address_list from the addresses in the given struct
196 static struct address_list *
197 address_list_from_addrinfo (const struct addrinfo *ai)
199 struct address_list *al;
200 const struct addrinfo *ptr;
205 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
206 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
211 al = xnew0 (struct address_list);
212 al->addresses = xnew_array (ip_address, cnt);
217 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
218 if (ptr->ai_family == AF_INET6)
220 const struct sockaddr_in6 *sin6 =
221 (const struct sockaddr_in6 *)ptr->ai_addr;
222 ip->type = IPV6_ADDRESS;
223 ADDRESS_IPV6_IN6_ADDR (ip) = sin6->sin6_addr;
224 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
225 ADDRESS_IPV6_SCOPE (ip) = sin6->sin6_scope_id;
229 else if (ptr->ai_family == AF_INET)
231 const struct sockaddr_in *sin =
232 (const struct sockaddr_in *)ptr->ai_addr;
233 ip->type = IPV4_ADDRESS;
234 ADDRESS_IPV4_IN_ADDR (ip) = sin->sin_addr;
237 assert (ip - al->addresses == cnt);
241 #define IS_IPV4(addr) (((const ip_address *) addr)->type == IPV4_ADDRESS)
243 /* Compare two IP addresses by type, giving preference to the IPv4
244 address (sorting it first). In other words, return -1 if ADDR1 is
245 IPv4 and ADDR2 is IPv6, +1 if ADDR1 is IPv6 and ADDR2 is IPv4, and
248 This is intended to be used as the comparator arg to a qsort-like
249 sorting function, which is why it accepts generic pointers. */
252 cmp_prefer_ipv4 (const void *addr1, const void *addr2)
254 return !IS_IPV4 (addr1) - !IS_IPV4 (addr2);
257 #define IS_IPV6(addr) (((const ip_address *) addr)->type == IPV6_ADDRESS)
259 /* Like the above, but give preference to the IPv6 address. */
262 cmp_prefer_ipv6 (const void *addr1, const void *addr2)
264 return !IS_IPV6 (addr1) - !IS_IPV6 (addr2);
267 #else /* not ENABLE_IPV6 */
269 /* Create an address_list from a NULL-terminated vector of IPv4
270 addresses. This kind of vector is returned by gethostbyname. */
272 static struct address_list *
273 address_list_from_ipv4_addresses (char **vec)
276 struct address_list *al = xnew0 (struct address_list);
283 al->addresses = xnew_array (ip_address, count);
287 for (i = 0; i < count; i++)
289 ip_address *ip = &al->addresses[i];
290 ip->type = IPV4_ADDRESS;
291 memcpy (ADDRESS_IPV4_DATA (ip), vec[i], 4);
297 #endif /* not ENABLE_IPV6 */
300 address_list_delete (struct address_list *al)
302 xfree (al->addresses);
306 /* Mark the address list as being no longer in use. This will reduce
307 its reference count which will cause the list to be freed when the
311 address_list_release (struct address_list *al)
314 DEBUGP (("Releasing %p (new refcount %d).\n", al, al->refcount));
315 if (al->refcount <= 0)
317 DEBUGP (("Deleting unused %p.\n", al));
318 address_list_delete (al);
322 /* Versions of gethostbyname and getaddrinfo that support timeout. */
326 struct ghbnwt_context {
327 const char *host_name;
328 struct hostent *hptr;
332 gethostbyname_with_timeout_callback (void *arg)
334 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
335 ctx->hptr = gethostbyname (ctx->host_name);
338 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
339 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
340 The function makes sure that when NULL is returned for reasons
341 other than timeout, errno is reset. */
343 static struct hostent *
344 gethostbyname_with_timeout (const char *host_name, double timeout)
346 struct ghbnwt_context ctx;
347 ctx.host_name = host_name;
348 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
350 SET_H_ERRNO (HOST_NOT_FOUND);
359 /* Print error messages for host errors. */
361 host_errstr (int error)
363 /* Can't use switch since some of these constants can be equal,
364 which makes the compiler complain about duplicate case
366 if (error == HOST_NOT_FOUND
367 || error == NO_RECOVERY
369 || error == NO_ADDRESS)
370 return _("Unknown host");
371 else if (error == TRY_AGAIN)
372 /* Message modeled after what gai_strerror returns in similar
374 return _("Temporary failure in name resolution");
376 return _("Unknown error");
379 #else /* ENABLE_IPV6 */
381 struct gaiwt_context {
384 const struct addrinfo *hints;
385 struct addrinfo **res;
390 getaddrinfo_with_timeout_callback (void *arg)
392 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
393 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
396 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
397 In case of timeout, the EAI_SYSTEM error code is returned and errno
398 is set to ETIMEDOUT. */
401 getaddrinfo_with_timeout (const char *node, const char *service,
402 const struct addrinfo *hints, struct addrinfo **res,
405 struct gaiwt_context ctx;
407 ctx.service = service;
411 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
416 return ctx.exit_code;
419 #endif /* ENABLE_IPV6 */
421 /* Pretty-print ADDR. When compiled without IPv6, this is the same as
422 inet_ntoa. With IPv6, it either prints an IPv6 address or an IPv4
426 pretty_print_address (const ip_address *addr)
431 return inet_ntoa (ADDRESS_IPV4_IN_ADDR (addr));
435 static char buf[128];
436 inet_ntop (AF_INET6, &ADDRESS_IPV6_IN6_ADDR (addr), buf, sizeof (buf));
438 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
440 /* append "%SCOPE_ID" for all ?non-global? addresses */
441 char *p = buf + strlen (buf);
443 number_to_string (p, ADDRESS_IPV6_SCOPE (addr));
447 buf[sizeof (buf) - 1] = '\0';
455 /* The following two functions were adapted from glibc. */
458 is_valid_ipv4_address (const char *str, const char *end)
468 if (ch >= '0' && ch <= '9')
470 val = val * 10 + (ch - '0');
481 else if (ch == '.' && saw_digit == 1)
498 is_valid_ipv6_address (const char *str, const char *end)
500 /* Use lower-case for these to avoid clash with system headers. */
519 /* Leading :: requires some special handling. */
523 if (str == end || *str != ':')
535 /* if ch is a number, add it to val. */
539 val |= XDIGIT_TO_NUM (ch);
546 /* if ch is a colon ... */
559 if (tp > ns_in6addrsz - ns_int16sz)
567 /* if ch is a dot ... */
568 if (ch == '.' && (tp <= ns_in6addrsz - ns_inaddrsz)
569 && is_valid_ipv4_address (curtok, end) == 1)
581 if (tp > ns_in6addrsz - ns_int16sz)
588 if (tp == ns_in6addrsz)
593 if (tp != ns_in6addrsz)
599 /* Simple host cache, used by lookup_host to speed up resolving. The
600 cache doesn't handle TTL because Wget is a fairly short-lived
601 application. Refreshing is attempted when connect fails, though --
602 see connect_to_host. */
604 /* Mapping between known hosts and to lists of their addresses. */
605 static struct hash_table *host_name_addresses_map;
608 /* Return the host's resolved addresses from the cache, if
611 static struct address_list *
612 cache_query (const char *host)
614 struct address_list *al;
615 if (!host_name_addresses_map)
617 al = hash_table_get (host_name_addresses_map, host);
620 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
627 /* Cache the DNS lookup of HOST. Subsequent invocations of
628 lookup_host will return the cached value. */
631 cache_store (const char *host, struct address_list *al)
633 if (!host_name_addresses_map)
634 host_name_addresses_map = make_nocase_string_hash_table (0);
637 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
643 debug_logprintf ("Caching %s =>", host);
644 for (i = 0; i < al->count; i++)
645 debug_logprintf (" %s", pretty_print_address (al->addresses + i));
646 debug_logprintf ("\n");
651 /* Remove HOST from the DNS cache. Does nothing is HOST is not in
655 cache_remove (const char *host)
657 struct address_list *al;
658 if (!host_name_addresses_map)
660 al = hash_table_get (host_name_addresses_map, host);
663 address_list_release (al);
664 hash_table_remove (host_name_addresses_map, host);
668 /* Look up HOST in DNS and return a list of IP addresses.
670 This function caches its result so that, if the same host is passed
671 the second time, the addresses are returned without DNS lookup.
672 (Use LH_REFRESH to force lookup, or set opt.dns_cache to 0 to
673 globally disable caching.)
675 The order of the returned addresses is affected by the setting of
676 opt.prefer_family: if it is set to prefer_ipv4, IPv4 addresses are
677 placed at the beginning; if it is prefer_ipv6, IPv6 ones are placed
678 at the beginning; otherwise, the order is left intact. The
679 relative order of addresses with the same family is left
680 undisturbed in either case.
682 FLAGS can be a combination of:
683 LH_SILENT - don't print the "resolving ... done" messages.
684 LH_BIND - resolve addresses for use with bind, which under
685 IPv6 means to use AI_PASSIVE flag to getaddrinfo.
686 Passive lookups are not cached under IPv6.
687 LH_REFRESH - if HOST is cached, remove the entry from the cache
688 and resolve it anew. */
690 struct address_list *
691 lookup_host (const char *host, int flags)
693 struct address_list *al;
694 int silent = flags & LH_SILENT;
696 int numeric_address = 0;
697 double timeout = opt.dns_timeout;
700 /* If we're not using getaddrinfo, first check if HOST specifies a
701 numeric IPv4 address. Some implementations of gethostbyname
702 (e.g. the Ultrix one and possibly Winsock) don't accept
703 dotted-decimal IPv4 addresses. */
705 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
706 if (addr_ipv4 != (uint32_t) -1)
708 /* No need to cache host->addr relation, just return the
711 vec[0] = (char *)&addr_ipv4;
713 return address_list_from_ipv4_addresses (vec);
716 #else /* ENABLE_IPV6 */
717 /* If we're using getaddrinfo, at least check whether the address is
718 already numeric, in which case there is no need to print the
719 "Resolving..." output. (This comes at no additional cost since
720 the is_valid_ipv*_address are already required for
723 const char *end = host + strlen (host);
724 if (is_valid_ipv4_address (host, end) || is_valid_ipv6_address (host, end))
729 /* Cache is normally on, but can be turned off with --no-dns-cache.
730 Don't cache passive lookups under IPv6. */
731 use_cache = opt.dns_cache;
733 if ((flags & LH_BIND) || numeric_address)
737 /* Try to find the host in the cache so we don't need to talk to the
738 resolver. If LH_REFRESH is requested, remove HOST from the cache
742 if (!(flags & LH_REFRESH))
744 al = cache_query (host);
752 /* No luck with the cache; resolve HOST. */
754 if (!silent && !numeric_address)
755 logprintf (LOG_VERBOSE, _("Resolving %s... "), escnonprint (host));
760 struct addrinfo hints, *res;
763 hints.ai_socktype = SOCK_STREAM;
765 hints.ai_family = AF_INET;
766 else if (opt.ipv6_only)
767 hints.ai_family = AF_INET6;
770 hints.ai_family = AF_UNSPEC;
772 hints.ai_flags |= AI_ADDRCONFIG;
774 /* On systems without AI_ADDRCONFIG, emulate it by manually
775 checking whether the system supports IPv6 sockets. */
776 if (!socket_has_inet6 ())
777 hints.ai_family = AF_INET;
781 hints.ai_flags |= AI_PASSIVE;
783 #ifdef AI_NUMERICHOST
786 /* Where available, the AI_NUMERICHOST hint can prevent costly
787 access to DNS servers. */
788 hints.ai_flags |= AI_NUMERICHOST;
789 timeout = 0; /* no timeout needed when "resolving"
790 numeric hosts -- avoid setting up
791 signal handlers and such. */
795 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, timeout);
796 if (err != 0 || res == NULL)
799 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
800 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
803 al = address_list_from_addrinfo (res);
807 logprintf (LOG_VERBOSE,
808 _("failed: No IPv4/IPv6 addresses for host.\n"));
812 /* Reorder addresses so that IPv4 ones (or IPv6 ones, as per
813 --prefer-family) come first. Sorting is stable so the order of
814 the addresses with the same family is undisturbed. */
815 if (al->count > 1 && opt.prefer_family != prefer_none)
816 stable_sort (al->addresses, al->count, sizeof (ip_address),
817 opt.prefer_family == prefer_ipv4
818 ? cmp_prefer_ipv4 : cmp_prefer_ipv6);
820 #else /* not ENABLE_IPV6 */
822 struct hostent *hptr = gethostbyname_with_timeout (host, timeout);
827 if (errno != ETIMEDOUT)
828 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
829 host_errstr (h_errno));
831 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
835 /* Do older systems have h_addr_list? */
836 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
838 #endif /* not ENABLE_IPV6 */
840 /* Print the addresses determined by DNS lookup, but no more than
842 if (!silent && !numeric_address)
845 int printmax = al->count <= 3 ? al->count : 3;
846 for (i = 0; i < printmax; i++)
848 logprintf (LOG_VERBOSE, "%s",
849 pretty_print_address (al->addresses + i));
850 if (i < printmax - 1)
851 logputs (LOG_VERBOSE, ", ");
853 if (printmax != al->count)
854 logputs (LOG_VERBOSE, ", ...");
855 logputs (LOG_VERBOSE, "\n");
858 /* Cache the lookup information. */
860 cache_store (host, al);
865 /* Determine whether a URL is acceptable to be followed, according to
866 a list of domains to accept. */
868 accept_domain (struct url *u)
870 assert (u->host != NULL);
873 if (!sufmatch ((const char **)opt.domains, u->host))
876 if (opt.exclude_domains)
878 if (sufmatch ((const char **)opt.exclude_domains, u->host))
884 /* Check whether WHAT is matched in LIST, each element of LIST being a
885 pattern to match WHAT against, using backward matching (see
886 match_backwards() in utils.c).
888 If an element of LIST matched, 1 is returned, 0 otherwise. */
890 sufmatch (const char **list, const char *what)
895 for (i = 0; list[i]; i++)
897 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
898 if (TOLOWER (list[i][j]) != TOLOWER (what[k]))
900 /* The domain must be first to reach to beginning. */
908 host_cleanup_mapper (void *key, void *value, void *arg_ignored)
910 struct address_list *al;
912 xfree (key); /* host */
914 al = (struct address_list *)value;
915 assert (al->refcount == 1);
916 address_list_delete (al);
924 if (host_name_addresses_map)
926 hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
927 hash_table_destroy (host_name_addresses_map);
928 host_name_addresses_map = NULL;