1 /* Host name resolution and matching.
2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3 2005, 2006, 2007, 2008, 2009, 2010, 2011 Free Software Foundation,
6 This file is part of GNU Wget.
8 GNU Wget is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 GNU Wget is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with Wget. If not, see <http://www.gnu.org/licenses/>.
21 Additional permission under GNU GPL version 3 section 7
23 If you modify this program, or any covered work, by linking or
24 combining it with the OpenSSL project's OpenSSL library (or a
25 modified version of that library), containing parts covered by the
26 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
27 grants you additional permission to convey the resulting work.
28 Corresponding Source for a non-source form of such a combination
29 shall include the source code for the parts of OpenSSL used as well
30 as that of the covered work. */
40 # include <sys/types.h>
41 # include <sys/socket.h>
42 # include <netinet/in.h>
44 # include <arpa/inet.h>
48 # else /* def __VMS */
50 # endif /* def __VMS [else] */
51 # define SET_H_ERRNO(err) ((void)(h_errno = (err)))
53 # include <winsock2.h>
54 # include <ws2tcpip.h>
55 # define SET_H_ERRNO(err) WSASetLastError (err)
66 # define NO_ADDRESS NO_DATA
69 #if !HAVE_DECL_H_ERRNO && !defined(WINDOWS)
74 /* Lists of IP addresses that result from running DNS queries. See
75 lookup_host for details. */
78 int count; /* number of adrresses */
79 ip_address *addresses; /* pointer to the string of addresses */
81 int faulty; /* number of addresses known not to work. */
82 bool connected; /* whether we were able to connect to
83 one of the addresses in the list,
86 int refcount; /* reference count; when it drops to
87 0, the entry is freed. */
90 /* Get the bounds of the address list. */
93 address_list_get_bounds (const struct address_list *al, int *start, int *end)
99 /* Return a pointer to the address at position POS. */
102 address_list_address_at (const struct address_list *al, int pos)
104 assert (pos >= al->faulty && pos < al->count);
105 return al->addresses + pos;
108 /* Return true if AL contains IP, false otherwise. */
111 address_list_contains (const struct address_list *al, const ip_address *ip)
117 for (i = 0; i < al->count; i++)
119 ip_address *cur = al->addresses + i;
120 if (cur->family == AF_INET
121 && (cur->data.d4.s_addr == ip->data.d4.s_addr))
127 for (i = 0; i < al->count; i++)
129 ip_address *cur = al->addresses + i;
130 if (cur->family == AF_INET6
131 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
132 && cur->ipv6_scope == ip->ipv6_scope
134 && IN6_ARE_ADDR_EQUAL (&cur->data.d6, &ip->data.d6))
138 #endif /* ENABLE_IPV6 */
144 /* Mark the INDEXth element of AL as faulty, so that the next time
145 this address list is used, the faulty element will be skipped. */
148 address_list_set_faulty (struct address_list *al, int index)
150 /* We assume that the address list is traversed in order, so that a
151 "faulty" attempt is always preceded with all-faulty addresses,
152 and this is how Wget uses it. */
153 assert (index == al->faulty);
156 if (al->faulty >= al->count)
157 /* All addresses have been proven faulty. Since there's not much
158 sense in returning the user an empty address list the next
159 time, we'll rather make them all clean, so that they can be
164 /* Set the "connected" flag to true. This flag used by connect.c to
165 see if the host perhaps needs to be resolved again. */
168 address_list_set_connected (struct address_list *al)
170 al->connected = true;
173 /* Return the value of the "connected" flag. */
176 address_list_connected_p (const struct address_list *al)
178 return al->connected;
183 /* Create an address_list from the addresses in the given struct
186 static struct address_list *
187 address_list_from_addrinfo (const struct addrinfo *ai)
189 struct address_list *al;
190 const struct addrinfo *ptr;
195 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
196 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
201 al = xnew0 (struct address_list);
202 al->addresses = xnew_array (ip_address, cnt);
207 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
208 if (ptr->ai_family == AF_INET6)
210 const struct sockaddr_in6 *sin6 =
211 (const struct sockaddr_in6 *)ptr->ai_addr;
212 ip->family = AF_INET6;
213 ip->data.d6 = sin6->sin6_addr;
214 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
215 ip->ipv6_scope = sin6->sin6_scope_id;
219 else if (ptr->ai_family == AF_INET)
221 const struct sockaddr_in *sin =
222 (const struct sockaddr_in *)ptr->ai_addr;
223 ip->family = AF_INET;
224 ip->data.d4 = sin->sin_addr;
227 assert (ip - al->addresses == cnt);
231 #define IS_IPV4(addr) (((const ip_address *) addr)->family == AF_INET)
233 /* Compare two IP addresses by family, giving preference to the IPv4
234 address (sorting it first). In other words, return -1 if ADDR1 is
235 IPv4 and ADDR2 is IPv6, +1 if ADDR1 is IPv6 and ADDR2 is IPv4, and
238 This is intended to be used as the comparator arg to a qsort-like
239 sorting function, which is why it accepts generic pointers. */
242 cmp_prefer_ipv4 (const void *addr1, const void *addr2)
244 return !IS_IPV4 (addr1) - !IS_IPV4 (addr2);
247 #define IS_IPV6(addr) (((const ip_address *) addr)->family == AF_INET6)
249 /* Like the above, but give preference to the IPv6 address. */
252 cmp_prefer_ipv6 (const void *addr1, const void *addr2)
254 return !IS_IPV6 (addr1) - !IS_IPV6 (addr2);
257 #else /* not ENABLE_IPV6 */
259 /* Create an address_list from a NULL-terminated vector of IPv4
260 addresses. This kind of vector is returned by gethostbyname. */
262 static struct address_list *
263 address_list_from_ipv4_addresses (char **vec)
266 struct address_list *al = xnew0 (struct address_list);
273 al->addresses = xnew_array (ip_address, count);
277 for (i = 0; i < count; i++)
279 ip_address *ip = &al->addresses[i];
280 ip->family = AF_INET;
281 memcpy (IP_INADDR_DATA (ip), vec[i], 4);
287 #endif /* not ENABLE_IPV6 */
290 address_list_delete (struct address_list *al)
292 xfree (al->addresses);
296 /* Mark the address list as being no longer in use. This will reduce
297 its reference count which will cause the list to be freed when the
301 address_list_release (struct address_list *al)
304 DEBUGP (("Releasing 0x%0*lx (new refcount %d).\n", PTR_FORMAT (al),
306 if (al->refcount <= 0)
308 DEBUGP (("Deleting unused 0x%0*lx.\n", PTR_FORMAT (al)));
309 address_list_delete (al);
313 /* Versions of gethostbyname and getaddrinfo that support timeout. */
317 struct ghbnwt_context {
318 const char *host_name;
319 struct hostent *hptr;
323 gethostbyname_with_timeout_callback (void *arg)
325 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
326 ctx->hptr = gethostbyname (ctx->host_name);
329 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
330 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
331 The function makes sure that when NULL is returned for reasons
332 other than timeout, errno is reset. */
334 static struct hostent *
335 gethostbyname_with_timeout (const char *host_name, double timeout)
337 struct ghbnwt_context ctx;
338 ctx.host_name = host_name;
339 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
341 SET_H_ERRNO (HOST_NOT_FOUND);
350 /* Print error messages for host errors. */
352 host_errstr (int error)
354 /* Can't use switch since some of these constants can be equal,
355 which makes the compiler complain about duplicate case
357 if (error == HOST_NOT_FOUND
358 || error == NO_RECOVERY
360 || error == NO_ADDRESS)
361 return _("Unknown host");
362 else if (error == TRY_AGAIN)
363 /* Message modeled after what gai_strerror returns in similar
365 return _("Temporary failure in name resolution");
367 return _("Unknown error");
370 #else /* ENABLE_IPV6 */
372 struct gaiwt_context {
375 const struct addrinfo *hints;
376 struct addrinfo **res;
381 getaddrinfo_with_timeout_callback (void *arg)
383 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
384 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
387 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
388 In case of timeout, the EAI_SYSTEM error code is returned and errno
389 is set to ETIMEDOUT. */
392 getaddrinfo_with_timeout (const char *node, const char *service,
393 const struct addrinfo *hints, struct addrinfo **res,
396 struct gaiwt_context ctx;
398 ctx.service = service;
402 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
407 return ctx.exit_code;
410 #endif /* ENABLE_IPV6 */
412 /* Return a textual representation of ADDR, i.e. the dotted quad for
413 IPv4 addresses, and the colon-separated list of hex words (with all
414 zeros omitted, etc.) for IPv6 addresses. */
417 print_address (const ip_address *addr)
421 if (!inet_ntop (addr->family, IP_INADDR_DATA (addr), buf, sizeof buf))
422 snprintf (buf, sizeof buf, "<error: %s>", strerror (errno));
425 return inet_ntoa (addr->data.d4);
429 /* The following two functions were adapted from glibc's
430 implementation of inet_pton, written by Paul Vixie. */
433 is_valid_ipv4_address (const char *str, const char *end)
435 bool saw_digit = false;
443 if (ch >= '0' && ch <= '9')
445 val = val * 10 + (ch - '0');
456 else if (ch == '.' && saw_digit)
473 is_valid_ipv6_address (const char *str, const char *end)
475 /* Use lower-case for these to avoid clash with system headers. */
494 /* Leading :: requires some special handling. */
498 if (str == end || *str != ':')
510 /* if ch is a number, add it to val. */
514 val |= XDIGIT_TO_NUM (ch);
521 /* if ch is a colon ... */
534 if (tp > ns_in6addrsz - ns_int16sz)
542 /* if ch is a dot ... */
543 if (ch == '.' && (tp <= ns_in6addrsz - ns_inaddrsz)
544 && is_valid_ipv4_address (curtok, end) == 1)
556 if (tp > ns_in6addrsz - ns_int16sz)
563 if (tp == ns_in6addrsz)
568 if (tp != ns_in6addrsz)
574 /* Simple host cache, used by lookup_host to speed up resolving. The
575 cache doesn't handle TTL because Wget is a fairly short-lived
576 application. Refreshing is attempted when connect fails, though --
577 see connect_to_host. */
579 /* Mapping between known hosts and to lists of their addresses. */
580 static struct hash_table *host_name_addresses_map;
583 /* Return the host's resolved addresses from the cache, if
586 static struct address_list *
587 cache_query (const char *host)
589 struct address_list *al;
590 if (!host_name_addresses_map)
592 al = hash_table_get (host_name_addresses_map, host);
595 DEBUGP (("Found %s in host_name_addresses_map (%p)\n", host, al));
602 /* Cache the DNS lookup of HOST. Subsequent invocations of
603 lookup_host will return the cached value. */
606 cache_store (const char *host, struct address_list *al)
608 if (!host_name_addresses_map)
609 host_name_addresses_map = make_nocase_string_hash_table (0);
612 hash_table_put (host_name_addresses_map, xstrdup_lower (host), al);
617 debug_logprintf ("Caching %s =>", host);
618 for (i = 0; i < al->count; i++)
619 debug_logprintf (" %s", print_address (al->addresses + i));
620 debug_logprintf ("\n");
624 /* Remove HOST from the DNS cache. Does nothing is HOST is not in
628 cache_remove (const char *host)
630 struct address_list *al;
631 if (!host_name_addresses_map)
633 al = hash_table_get (host_name_addresses_map, host);
636 address_list_release (al);
637 hash_table_remove (host_name_addresses_map, host);
641 /* Look up HOST in DNS and return a list of IP addresses.
643 This function caches its result so that, if the same host is passed
644 the second time, the addresses are returned without DNS lookup.
645 (Use LH_REFRESH to force lookup, or set opt.dns_cache to 0 to
646 globally disable caching.)
648 The order of the returned addresses is affected by the setting of
649 opt.prefer_family: if it is set to prefer_ipv4, IPv4 addresses are
650 placed at the beginning; if it is prefer_ipv6, IPv6 ones are placed
651 at the beginning; otherwise, the order is left intact. The
652 relative order of addresses with the same family is left
653 undisturbed in either case.
655 FLAGS can be a combination of:
656 LH_SILENT - don't print the "resolving ... done" messages.
657 LH_BIND - resolve addresses for use with bind, which under
658 IPv6 means to use AI_PASSIVE flag to getaddrinfo.
659 Passive lookups are not cached under IPv6.
660 LH_REFRESH - if HOST is cached, remove the entry from the cache
661 and resolve it anew. */
663 struct address_list *
664 lookup_host (const char *host, int flags)
666 struct address_list *al;
667 bool silent = !!(flags & LH_SILENT);
669 bool numeric_address = false;
670 double timeout = opt.dns_timeout;
673 /* If we're not using getaddrinfo, first check if HOST specifies a
674 numeric IPv4 address. Some implementations of gethostbyname
675 (e.g. the Ultrix one and possibly Winsock) don't accept
676 dotted-decimal IPv4 addresses. */
678 uint32_t addr_ipv4 = (uint32_t)inet_addr (host);
679 if (addr_ipv4 != (uint32_t) -1)
681 /* No need to cache host->addr relation, just return the
684 vec[0] = (char *)&addr_ipv4;
686 return address_list_from_ipv4_addresses (vec);
689 #else /* ENABLE_IPV6 */
690 /* If we're using getaddrinfo, at least check whether the address is
691 already numeric, in which case there is no need to print the
692 "Resolving..." output. (This comes at no additional cost since
693 the is_valid_ipv*_address are already required for
696 const char *end = host + strlen (host);
697 if (is_valid_ipv4_address (host, end) || is_valid_ipv6_address (host, end))
698 numeric_address = true;
702 /* Cache is normally on, but can be turned off with --no-dns-cache.
703 Don't cache passive lookups under IPv6. */
704 use_cache = opt.dns_cache;
706 if ((flags & LH_BIND) || numeric_address)
710 /* Try to find the host in the cache so we don't need to talk to the
711 resolver. If LH_REFRESH is requested, remove HOST from the cache
715 if (!(flags & LH_REFRESH))
717 al = cache_query (host);
725 /* No luck with the cache; resolve HOST. */
727 if (!silent && !numeric_address)
729 char *str = NULL, *name;
731 if (opt.enable_iri && (name = idn_decode ((char *) host)) != NULL)
733 int len = strlen (host) + strlen (name) + 4;
735 snprintf (str, len, "%s (%s)", name, host);
740 logprintf (LOG_VERBOSE, _("Resolving %s... "),
741 quotearg_style (escape_quoting_style, str ? str : host));
750 struct addrinfo hints, *res;
753 hints.ai_socktype = SOCK_STREAM;
755 hints.ai_family = AF_INET;
756 else if (opt.ipv6_only)
757 hints.ai_family = AF_INET6;
759 /* We tried using AI_ADDRCONFIG, but removed it because: it
760 misinterprets IPv6 loopbacks, it is broken on AIX 5.1, and
761 it's unneeded since we sort the addresses anyway. */
762 hints.ai_family = AF_UNSPEC;
765 hints.ai_flags |= AI_PASSIVE;
767 #ifdef AI_NUMERICHOST
770 /* Where available, the AI_NUMERICHOST hint can prevent costly
771 access to DNS servers. */
772 hints.ai_flags |= AI_NUMERICHOST;
773 timeout = 0; /* no timeout needed when "resolving"
774 numeric hosts -- avoid setting up
775 signal handlers and such. */
779 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, timeout);
780 if (err != 0 || res == NULL)
783 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
784 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
787 al = address_list_from_addrinfo (res);
791 logprintf (LOG_VERBOSE,
792 _("failed: No IPv4/IPv6 addresses for host.\n"));
796 /* Reorder addresses so that IPv4 ones (or IPv6 ones, as per
797 --prefer-family) come first. Sorting is stable so the order of
798 the addresses with the same family is undisturbed. */
799 if (al->count > 1 && opt.prefer_family != prefer_none)
800 stable_sort (al->addresses, al->count, sizeof (ip_address),
801 opt.prefer_family == prefer_ipv4
802 ? cmp_prefer_ipv4 : cmp_prefer_ipv6);
804 #else /* not ENABLE_IPV6 */
806 struct hostent *hptr = gethostbyname_with_timeout (host, timeout);
811 if (errno != ETIMEDOUT)
812 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
813 host_errstr (h_errno));
815 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
819 /* Do older systems have h_addr_list? */
820 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
822 #endif /* not ENABLE_IPV6 */
824 /* Print the addresses determined by DNS lookup, but no more than
825 three if show_all_dns_entries is not specified. */
826 if (!silent && !numeric_address)
829 int printmax = al->count;
831 if (! opt.show_all_dns_entries)
834 for (i = 0; i < printmax; i++)
836 logputs (LOG_VERBOSE, print_address (al->addresses + i));
837 if (i < printmax - 1)
838 logputs (LOG_VERBOSE, ", ");
840 if (printmax != al->count)
841 logputs (LOG_VERBOSE, ", ...");
842 logputs (LOG_VERBOSE, "\n");
845 /* Cache the lookup information. */
847 cache_store (host, al);
852 /* Determine whether a URL is acceptable to be followed, according to
853 a list of domains to accept. */
855 accept_domain (struct url *u)
857 assert (u->host != NULL);
860 if (!sufmatch ((const char **)opt.domains, u->host))
863 if (opt.exclude_domains)
865 if (sufmatch ((const char **)opt.exclude_domains, u->host))
871 /* Check whether WHAT is matched in LIST, each element of LIST being a
872 pattern to match WHAT against, using backward matching (see
873 match_backwards() in utils.c).
875 If an element of LIST matched, 1 is returned, 0 otherwise. */
877 sufmatch (const char **list, const char *what)
882 for (i = 0; list[i]; i++)
884 if (list[i][0] == '\0')
887 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
888 if (c_tolower (list[i][j]) != c_tolower (what[k]))
890 /* The domain must be first to reach to beginning. */
900 if (host_name_addresses_map)
902 hash_table_iterator iter;
903 for (hash_table_iterate (host_name_addresses_map, &iter);
904 hash_table_iter_next (&iter);
907 char *host = iter.key;
908 struct address_list *al = iter.value;
910 assert (al->refcount == 1);
911 address_list_delete (al);
913 hash_table_destroy (host_name_addresses_map);
914 host_name_addresses_map = NULL;