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 # define SET_H_ERRNO(err) WSASetLastError (err)
64 # define NO_ADDRESS NO_DATA
67 #if !HAVE_DECL_H_ERRNO && !defined(WINDOWS)
72 /* Lists of IP addresses that result from running DNS queries. See
73 lookup_host for details. */
76 int count; /* number of adrresses */
77 ip_address *addresses; /* pointer to the string of addresses */
79 int faulty; /* number of addresses known not to work. */
80 bool connected; /* whether we were able to connect to
81 one of the addresses in the list,
84 int refcount; /* reference count; when it drops to
85 0, the entry is freed. */
88 /* Get the bounds of the address list. */
91 address_list_get_bounds (const struct address_list *al, int *start, int *end)
97 /* Return a pointer to the address at position POS. */
100 address_list_address_at (const struct address_list *al, int pos)
102 assert (pos >= al->faulty && pos < al->count);
103 return al->addresses + pos;
106 /* Return true if AL contains IP, false otherwise. */
109 address_list_contains (const struct address_list *al, const ip_address *ip)
115 for (i = 0; i < al->count; i++)
117 ip_address *cur = al->addresses + i;
118 if (cur->family == AF_INET
119 && (cur->data.d4.s_addr == ip->data.d4.s_addr))
125 for (i = 0; i < al->count; i++)
127 ip_address *cur = al->addresses + i;
128 if (cur->family == AF_INET6
129 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
130 && cur->ipv6_scope == ip->ipv6_scope
132 && IN6_ARE_ADDR_EQUAL (&cur->data.d6, &ip->data.d6))
136 #endif /* ENABLE_IPV6 */
142 /* Mark the INDEXth element of AL as faulty, so that the next time
143 this address list is used, the faulty element will be skipped. */
146 address_list_set_faulty (struct address_list *al, int index)
148 /* We assume that the address list is traversed in order, so that a
149 "faulty" attempt is always preceded with all-faulty addresses,
150 and this is how Wget uses it. */
151 assert (index == al->faulty);
154 if (al->faulty >= al->count)
155 /* All addresses have been proven faulty. Since there's not much
156 sense in returning the user an empty address list the next
157 time, we'll rather make them all clean, so that they can be
162 /* Set the "connected" flag to true. This flag used by connect.c to
163 see if the host perhaps needs to be resolved again. */
166 address_list_set_connected (struct address_list *al)
168 al->connected = true;
171 /* Return the value of the "connected" flag. */
174 address_list_connected_p (const struct address_list *al)
176 return al->connected;
181 /* Create an address_list from the addresses in the given struct
184 static struct address_list *
185 address_list_from_addrinfo (const struct addrinfo *ai)
187 struct address_list *al;
188 const struct addrinfo *ptr;
193 for (ptr = ai; ptr != NULL ; ptr = ptr->ai_next)
194 if (ptr->ai_family == AF_INET || ptr->ai_family == AF_INET6)
199 al = xnew0 (struct address_list);
200 al->addresses = xnew_array (ip_address, cnt);
205 for (ptr = ai; ptr != NULL; ptr = ptr->ai_next)
206 if (ptr->ai_family == AF_INET6)
208 const struct sockaddr_in6 *sin6 =
209 (const struct sockaddr_in6 *)ptr->ai_addr;
210 ip->family = AF_INET6;
211 ip->data.d6 = sin6->sin6_addr;
212 #ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
213 ip->ipv6_scope = sin6->sin6_scope_id;
217 else if (ptr->ai_family == AF_INET)
219 const struct sockaddr_in *sin =
220 (const struct sockaddr_in *)ptr->ai_addr;
221 ip->family = AF_INET;
222 ip->data.d4 = sin->sin_addr;
225 assert (ip - al->addresses == cnt);
229 #define IS_IPV4(addr) (((const ip_address *) addr)->family == AF_INET)
231 /* Compare two IP addresses by family, giving preference to the IPv4
232 address (sorting it first). In other words, return -1 if ADDR1 is
233 IPv4 and ADDR2 is IPv6, +1 if ADDR1 is IPv6 and ADDR2 is IPv4, and
236 This is intended to be used as the comparator arg to a qsort-like
237 sorting function, which is why it accepts generic pointers. */
240 cmp_prefer_ipv4 (const void *addr1, const void *addr2)
242 return !IS_IPV4 (addr1) - !IS_IPV4 (addr2);
245 #define IS_IPV6(addr) (((const ip_address *) addr)->family == AF_INET6)
247 /* Like the above, but give preference to the IPv6 address. */
250 cmp_prefer_ipv6 (const void *addr1, const void *addr2)
252 return !IS_IPV6 (addr1) - !IS_IPV6 (addr2);
255 #else /* not ENABLE_IPV6 */
257 /* Create an address_list from a NULL-terminated vector of IPv4
258 addresses. This kind of vector is returned by gethostbyname. */
260 static struct address_list *
261 address_list_from_ipv4_addresses (char **vec)
264 struct address_list *al = xnew0 (struct address_list);
271 al->addresses = xnew_array (ip_address, count);
275 for (i = 0; i < count; i++)
277 ip_address *ip = &al->addresses[i];
278 ip->family = AF_INET;
279 memcpy (IP_INADDR_DATA (ip), vec[i], 4);
285 #endif /* not ENABLE_IPV6 */
288 address_list_delete (struct address_list *al)
290 xfree (al->addresses);
294 /* Mark the address list as being no longer in use. This will reduce
295 its reference count which will cause the list to be freed when the
299 address_list_release (struct address_list *al)
302 DEBUGP (("Releasing 0x%0*lx (new refcount %d).\n", PTR_FORMAT (al),
304 if (al->refcount <= 0)
306 DEBUGP (("Deleting unused 0x%0*lx.\n", PTR_FORMAT (al)));
307 address_list_delete (al);
311 /* Versions of gethostbyname and getaddrinfo that support timeout. */
315 struct ghbnwt_context {
316 const char *host_name;
317 struct hostent *hptr;
321 gethostbyname_with_timeout_callback (void *arg)
323 struct ghbnwt_context *ctx = (struct ghbnwt_context *)arg;
324 ctx->hptr = gethostbyname (ctx->host_name);
327 /* Just like gethostbyname, except it times out after TIMEOUT seconds.
328 In case of timeout, NULL is returned and errno is set to ETIMEDOUT.
329 The function makes sure that when NULL is returned for reasons
330 other than timeout, errno is reset. */
332 static struct hostent *
333 gethostbyname_with_timeout (const char *host_name, double timeout)
335 struct ghbnwt_context ctx;
336 ctx.host_name = host_name;
337 if (run_with_timeout (timeout, gethostbyname_with_timeout_callback, &ctx))
339 SET_H_ERRNO (HOST_NOT_FOUND);
348 /* Print error messages for host errors. */
350 host_errstr (int error)
352 /* Can't use switch since some of these constants can be equal,
353 which makes the compiler complain about duplicate case
355 if (error == HOST_NOT_FOUND
356 || error == NO_RECOVERY
358 || error == NO_ADDRESS)
359 return _("Unknown host");
360 else if (error == TRY_AGAIN)
361 /* Message modeled after what gai_strerror returns in similar
363 return _("Temporary failure in name resolution");
365 return _("Unknown error");
368 #else /* ENABLE_IPV6 */
370 struct gaiwt_context {
373 const struct addrinfo *hints;
374 struct addrinfo **res;
379 getaddrinfo_with_timeout_callback (void *arg)
381 struct gaiwt_context *ctx = (struct gaiwt_context *)arg;
382 ctx->exit_code = getaddrinfo (ctx->node, ctx->service, ctx->hints, ctx->res);
385 /* Just like getaddrinfo, except it times out after TIMEOUT seconds.
386 In case of timeout, the EAI_SYSTEM error code is returned and errno
387 is set to ETIMEDOUT. */
390 getaddrinfo_with_timeout (const char *node, const char *service,
391 const struct addrinfo *hints, struct addrinfo **res,
394 struct gaiwt_context ctx;
396 ctx.service = service;
400 if (run_with_timeout (timeout, getaddrinfo_with_timeout_callback, &ctx))
405 return ctx.exit_code;
408 #endif /* ENABLE_IPV6 */
410 /* Return a textual representation of ADDR, i.e. the dotted quad for
411 IPv4 addresses, and the colon-separated list of hex words (with all
412 zeros omitted, etc.) for IPv6 addresses. */
415 print_address (const ip_address *addr)
419 if (!inet_ntop (addr->family, IP_INADDR_DATA (addr), buf, sizeof buf))
420 snprintf (buf, sizeof buf, "<error: %s>", strerror (errno));
423 return inet_ntoa (addr->data.d4);
427 /* The following two functions were adapted from glibc's
428 implementation of inet_pton, written by Paul Vixie. */
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)
727 char *str = NULL, *name;
729 if (opt.enable_iri && (name = idn_decode ((char *) host)) != NULL)
731 int len = strlen (host) + strlen (name) + 4;
733 snprintf (str, len, "%s (%s)", name, host);
738 logprintf (LOG_VERBOSE, _("Resolving %s... "),
739 quotearg_style (escape_quoting_style, str ? str : host));
748 struct addrinfo hints, *res;
751 hints.ai_socktype = SOCK_STREAM;
753 hints.ai_family = AF_INET;
754 else if (opt.ipv6_only)
755 hints.ai_family = AF_INET6;
757 /* We tried using AI_ADDRCONFIG, but removed it because: it
758 misinterprets IPv6 loopbacks, it is broken on AIX 5.1, and
759 it's unneeded since we sort the addresses anyway. */
760 hints.ai_family = AF_UNSPEC;
763 hints.ai_flags |= AI_PASSIVE;
765 #ifdef AI_NUMERICHOST
768 /* Where available, the AI_NUMERICHOST hint can prevent costly
769 access to DNS servers. */
770 hints.ai_flags |= AI_NUMERICHOST;
771 timeout = 0; /* no timeout needed when "resolving"
772 numeric hosts -- avoid setting up
773 signal handlers and such. */
777 err = getaddrinfo_with_timeout (host, NULL, &hints, &res, timeout);
778 if (err != 0 || res == NULL)
781 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
782 err != EAI_SYSTEM ? gai_strerror (err) : strerror (errno));
785 al = address_list_from_addrinfo (res);
789 logprintf (LOG_VERBOSE,
790 _("failed: No IPv4/IPv6 addresses for host.\n"));
794 /* Reorder addresses so that IPv4 ones (or IPv6 ones, as per
795 --prefer-family) come first. Sorting is stable so the order of
796 the addresses with the same family is undisturbed. */
797 if (al->count > 1 && opt.prefer_family != prefer_none)
798 stable_sort (al->addresses, al->count, sizeof (ip_address),
799 opt.prefer_family == prefer_ipv4
800 ? cmp_prefer_ipv4 : cmp_prefer_ipv6);
802 #else /* not ENABLE_IPV6 */
804 struct hostent *hptr = gethostbyname_with_timeout (host, timeout);
809 if (errno != ETIMEDOUT)
810 logprintf (LOG_VERBOSE, _("failed: %s.\n"),
811 host_errstr (h_errno));
813 logputs (LOG_VERBOSE, _("failed: timed out.\n"));
817 /* Do older systems have h_addr_list? */
818 al = address_list_from_ipv4_addresses (hptr->h_addr_list);
820 #endif /* not ENABLE_IPV6 */
822 /* Print the addresses determined by DNS lookup, but no more than
824 if (!silent && !numeric_address)
827 int printmax = al->count <= 3 ? al->count : 3;
828 for (i = 0; i < printmax; i++)
830 logputs (LOG_VERBOSE, print_address (al->addresses + i));
831 if (i < printmax - 1)
832 logputs (LOG_VERBOSE, ", ");
834 if (printmax != al->count)
835 logputs (LOG_VERBOSE, ", ...");
836 logputs (LOG_VERBOSE, "\n");
839 /* Cache the lookup information. */
841 cache_store (host, al);
846 /* Determine whether a URL is acceptable to be followed, according to
847 a list of domains to accept. */
849 accept_domain (struct url *u)
851 assert (u->host != NULL);
854 if (!sufmatch ((const char **)opt.domains, u->host))
857 if (opt.exclude_domains)
859 if (sufmatch ((const char **)opt.exclude_domains, u->host))
865 /* Check whether WHAT is matched in LIST, each element of LIST being a
866 pattern to match WHAT against, using backward matching (see
867 match_backwards() in utils.c).
869 If an element of LIST matched, 1 is returned, 0 otherwise. */
871 sufmatch (const char **list, const char *what)
876 for (i = 0; list[i]; i++)
878 if (list[i][0] == '\0')
881 for (j = strlen (list[i]), k = lw; j >= 0 && k >= 0; j--, k--)
882 if (c_tolower (list[i][j]) != c_tolower (what[k]))
884 /* The domain must be first to reach to beginning. */
894 if (host_name_addresses_map)
896 hash_table_iterator iter;
897 for (hash_table_iterate (host_name_addresses_map, &iter);
898 hash_table_iter_next (&iter);
901 char *host = iter.key;
902 struct address_list *al = iter.value;
904 assert (al->refcount == 1);
905 address_list_delete (al);
907 hash_table_destroy (host_name_addresses_map);
908 host_name_addresses_map = NULL;