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 0x%0*lx (new refcount %d).\n", PTR_FORMAT (al),
316 if (al->refcount <= 0)
318 DEBUGP (("Deleting unused 0x%0*lx.\n", PTR_FORMAT (al)));
319 address_list_delete (al);
323 /* Versions of gethostbyname and getaddrinfo that support timeout. */
327 struct ghbnwt_context {
328 const char *host_name;
329 struct hostent *hptr;
333 gethostbyname_with_timeout_callback (void *arg)
335 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
336 ctx->hptr = gethostbyname (ctx->host_name);
339 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
340 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
341 The function makes sure that when NULL is returned for reasons
342 other than timeout, errno is reset. */
344 static struct hostent *
345 gethostbyname_with_timeout (const char *host_name, double timeout)
347 struct ghbnwt_context ctx;
348 ctx.host_name = host_name;
349 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
351 SET_H_ERRNO (HOST_NOT_FOUND);
360 /* Print error messages for host errors. */
362 host_errstr (int error)
364 /* Can't use switch since some of these constants can be equal,
365 which makes the compiler complain about duplicate case
367 if (error == HOST_NOT_FOUND
368 || error == NO_RECOVERY
370 || error == NO_ADDRESS)
371 return _("Unknown host");
372 else if (error == TRY_AGAIN)
373 /* Message modeled after what gai_strerror returns in similar
375 return _("Temporary failure in name resolution");
377 return _("Unknown error");
380 #else /* ENABLE_IPV6 */
382 struct gaiwt_context {
385 const struct addrinfo *hints;
386 struct addrinfo **res;
391 getaddrinfo_with_timeout_callback (void *arg)
393 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
394 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
397 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
398 In case of timeout, the EAI_SYSTEM error code is returned and errno
399 is set to ETIMEDOUT. */
402 getaddrinfo_with_timeout (const char *node, const char *service,
403 const struct addrinfo *hints, struct addrinfo **res,
406 struct gaiwt_context ctx;
408 ctx.service = service;
412 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
417 return ctx.exit_code;
420 #endif /* ENABLE_IPV6 */
422 /* Pretty-print ADDR. When compiled without IPv6, this is the same as
423 inet_ntoa. With IPv6, it either prints an IPv6 address or an IPv4
427 pretty_print_address (const ip_address *addr)
432 return inet_ntoa (ADDRESS_IPV4_IN_ADDR (addr));
436 static char buf[128];
437 inet_ntop (AF_INET6, &ADDRESS_IPV6_IN6_ADDR (addr), buf, sizeof (buf));
439 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
441 /* append "%SCOPE_ID" for all ?non-global? addresses */
442 char *p = buf + strlen (buf);
444 number_to_string (p, ADDRESS_IPV6_SCOPE (addr));
448 buf[sizeof (buf) - 1] = '\0';
456 /* The following two functions were adapted from glibc. */
459 is_valid_ipv4_address (const char *str, const char *end)
469 if (ch >= '0' && ch <= '9')
471 val = val * 10 + (ch - '0');
482 else if (ch == '.' && saw_digit == 1)
499 is_valid_ipv6_address (const char *str, const char *end)
501 /* Use lower-case for these to avoid clash with system headers. */
520 /* Leading :: requires some special handling. */
524 if (str == end || *str != ':')
536 /* if ch is a number, add it to val. */
540 val |= XDIGIT_TO_NUM (ch);
547 /* if ch is a colon ... */
560 if (tp > ns_in6addrsz - ns_int16sz)
568 /* if ch is a dot ... */
569 if (ch == '.' && (tp <= ns_in6addrsz - ns_inaddrsz)
570 && is_valid_ipv4_address (curtok, end) == 1)
582 if (tp > ns_in6addrsz - ns_int16sz)
589 if (tp == ns_in6addrsz)
594 if (tp != ns_in6addrsz)
600 /* Simple host cache, used by lookup_host to speed up resolving. The
601 cache doesn't handle TTL because Wget is a fairly short-lived
602 application. Refreshing is attempted when connect fails, though --
603 see connect_to_host. */
605 /* Mapping between known hosts and to lists of their addresses. */
606 static struct hash_table *host_name_addresses_map;
609 /* Return the host's resolved addresses from the cache, if
612 static struct address_list *
613 cache_query (const char *host)
615 struct address_list *al;
616 if (!host_name_addresses_map)
618 al = hash_table_get (host_name_addresses_map, host);
621 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
628 /* Cache the DNS lookup of HOST. Subsequent invocations of
629 lookup_host will return the cached value. */
632 cache_store (const char *host, struct address_list *al)
634 if (!host_name_addresses_map)
635 host_name_addresses_map = make_nocase_string_hash_table (0);
638 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
644 debug_logprintf ("Caching %s =>", host);
645 for (i = 0; i < al->count; i++)
646 debug_logprintf (" %s", pretty_print_address (al->addresses + i));
647 debug_logprintf ("\n");
652 /* Remove HOST from the DNS cache. Does nothing is HOST is not in
656 cache_remove (const char *host)
658 struct address_list *al;
659 if (!host_name_addresses_map)
661 al = hash_table_get (host_name_addresses_map, host);
664 address_list_release (al);
665 hash_table_remove (host_name_addresses_map, host);
669 /* Look up HOST in DNS and return a list of IP addresses.
671 This function caches its result so that, if the same host is passed
672 the second time, the addresses are returned without DNS lookup.
673 (Use LH_REFRESH to force lookup, or set opt.dns_cache to 0 to
674 globally disable caching.)
676 The order of the returned addresses is affected by the setting of
677 opt.prefer_family: if it is set to prefer_ipv4, IPv4 addresses are
678 placed at the beginning; if it is prefer_ipv6, IPv6 ones are placed
679 at the beginning; otherwise, the order is left intact. The
680 relative order of addresses with the same family is left
681 undisturbed in either case.
683 FLAGS can be a combination of:
684 LH_SILENT - don't print the "resolving ... done" messages.
685 LH_BIND - resolve addresses for use with bind, which under
686 IPv6 means to use AI_PASSIVE flag to getaddrinfo.
687 Passive lookups are not cached under IPv6.
688 LH_REFRESH - if HOST is cached, remove the entry from the cache
689 and resolve it anew. */
691 struct address_list *
692 lookup_host (const char *host, int flags)
694 struct address_list *al;
695 int silent = flags & LH_SILENT;
697 int numeric_address = 0;
698 double timeout = opt.dns_timeout;
701 /* If we're not using getaddrinfo, first check if HOST specifies a
702 numeric IPv4 address. Some implementations of gethostbyname
703 (e.g. the Ultrix one and possibly Winsock) don't accept
704 dotted-decimal IPv4 addresses. */
706 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
707 if (addr_ipv4 != (uint32_t) -1)
709 /* No need to cache host->addr relation, just return the
712 vec[0] = (char *)&addr_ipv4;
714 return address_list_from_ipv4_addresses (vec);
717 #else /* ENABLE_IPV6 */
718 /* If we're using getaddrinfo, at least check whether the address is
719 already numeric, in which case there is no need to print the
720 "Resolving..." output. (This comes at no additional cost since
721 the is_valid_ipv*_address are already required for
724 const char *end = host + strlen (host);
725 if (is_valid_ipv4_address (host, end) || is_valid_ipv6_address (host, end))
730 /* Cache is normally on, but can be turned off with --no-dns-cache.
731 Don't cache passive lookups under IPv6. */
732 use_cache = opt.dns_cache;
734 if ((flags & LH_BIND) || numeric_address)
738 /* Try to find the host in the cache so we don't need to talk to the
739 resolver. If LH_REFRESH is requested, remove HOST from the cache
743 if (!(flags & LH_REFRESH))
745 al = cache_query (host);
753 /* No luck with the cache; resolve HOST. */
755 if (!silent && !numeric_address)
756 logprintf (LOG_VERBOSE, _("Resolving %s... "), escnonprint (host));
761 struct addrinfo hints, *res;
764 hints.ai_socktype = SOCK_STREAM;
766 hints.ai_family = AF_INET;
767 else if (opt.ipv6_only)
768 hints.ai_family = AF_INET6;
771 hints.ai_family = AF_UNSPEC;
773 hints.ai_flags |= AI_ADDRCONFIG;
775 /* On systems without AI_ADDRCONFIG, emulate it by manually
776 checking whether the system supports IPv6 sockets. */
777 if (!socket_has_inet6 ())
778 hints.ai_family = AF_INET;
782 hints.ai_flags |= AI_PASSIVE;
784 #ifdef AI_NUMERICHOST
787 /* Where available, the AI_NUMERICHOST hint can prevent costly
788 access to DNS servers. */
789 hints.ai_flags |= AI_NUMERICHOST;
790 timeout = 0; /* no timeout needed when "resolving"
791 numeric hosts -- avoid setting up
792 signal handlers and such. */
796 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, timeout);
797 if (err != 0 || res == NULL)
800 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
801 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
804 al = address_list_from_addrinfo (res);
808 logprintf (LOG_VERBOSE,
809 _("failed: No IPv4/IPv6 addresses for host.\n"));
813 /* Reorder addresses so that IPv4 ones (or IPv6 ones, as per
814 --prefer-family) come first. Sorting is stable so the order of
815 the addresses with the same family is undisturbed. */
816 if (al->count > 1 && opt.prefer_family != prefer_none)
817 stable_sort (al->addresses, al->count, sizeof (ip_address),
818 opt.prefer_family == prefer_ipv4
819 ? cmp_prefer_ipv4 : cmp_prefer_ipv6);
821 #else /* not ENABLE_IPV6 */
823 struct hostent *hptr = gethostbyname_with_timeout (host, timeout);
828 if (errno != ETIMEDOUT)
829 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
830 host_errstr (h_errno));
832 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
836 /* Do older systems have h_addr_list? */
837 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
839 #endif /* not ENABLE_IPV6 */
841 /* Print the addresses determined by DNS lookup, but no more than
843 if (!silent && !numeric_address)
846 int printmax = al->count <= 3 ? al->count : 3;
847 for (i = 0; i < printmax; i++)
849 logprintf (LOG_VERBOSE, "%s",
850 pretty_print_address (al->addresses + i));
851 if (i < printmax - 1)
852 logputs (LOG_VERBOSE, ", ");
854 if (printmax != al->count)
855 logputs (LOG_VERBOSE, ", ...");
856 logputs (LOG_VERBOSE, "\n");
859 /* Cache the lookup information. */
861 cache_store (host, al);
866 /* Determine whether a URL is acceptable to be followed, according to
867 a list of domains to accept. */
869 accept_domain (struct url *u)
871 assert (u->host != NULL);
874 if (!sufmatch ((const char **)opt.domains, u->host))
877 if (opt.exclude_domains)
879 if (sufmatch ((const char **)opt.exclude_domains, u->host))
885 /* Check whether WHAT is matched in LIST, each element of LIST being a
886 pattern to match WHAT against, using backward matching (see
887 match_backwards() in utils.c).
889 If an element of LIST matched, 1 is returned, 0 otherwise. */
891 sufmatch (const char **list, const char *what)
896 for (i = 0; list[i]; i++)
898 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
899 if (TOLOWER (list[i][j]) != TOLOWER (what[k]))
901 /* The domain must be first to reach to beginning. */
909 host_cleanup_mapper (void *key, void *value, void *arg_ignored)
911 struct address_list *al;
913 xfree (key); /* host */
915 al = (struct address_list *)value;
916 assert (al->refcount == 1);
917 address_list_delete (al);
925 if (host_name_addresses_map)
927 hash_table_map (host_name_addresses_map, host_cleanup_mapper, NULL);
928 hash_table_destroy (host_name_addresses_map);
929 host_name_addresses_map = NULL;