1 /* Support for cookies.
2 Copyright (C) 2001, 2002 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 (at
9 your option) any later version.
11 GNU Wget is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 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. */
30 /* Written by Hrvoje Niksic. Parts are loosely inspired by cookie
31 code submitted by Tomasz Wegrzanowski.
33 TODO: Implement limits on cookie-related sizes, such as max. cookie
34 size, max. number of cookies, etc. Add more "cookie jar" methods,
35 such as methods to over stored cookies, to clear temporary cookies,
36 to perform intelligent auto-saving, etc. Ultimately support
37 `Set-Cookie2' and `Cookie2' headers. */
56 /* This should *really* be in a .h file! */
57 time_t http_atotm PARAMS ((const char *));
59 /* Declarations of `struct cookie' and the most basic functions. */
62 /* Hash table that maps domain names to cookie chains. A "cookie
63 chain" is a linked list of cookies that belong to the same
65 struct hash_table *chains_by_domain;
67 int cookie_count; /* number of cookies in the jar. */
70 /* Value set by entry point functions, so that the low-level
71 routines don't need to call time() all the time. */
77 struct cookie_jar *jar = xmalloc (sizeof (struct cookie_jar));
78 jar->chains_by_domain = make_nocase_string_hash_table (0);
79 jar->cookie_count = 0;
84 char *domain; /* domain of the cookie */
85 int port; /* port number */
86 char *path; /* path prefix of the cookie */
87 int secure; /* whether cookie should be
88 transmitted over non-https
90 int permanent; /* whether the cookie should outlive
92 time_t expiry_time; /* time when the cookie expires */
93 int discard_requested; /* whether cookie was created to
94 request discarding another
97 char *attr; /* cookie attribute name */
98 char *value; /* cookie attribute value */
100 struct cookie_jar *jar; /* pointer back to the cookie jar, for
102 struct cookie *next; /* used for chaining of cookies in the
106 #define PORT_ANY (-1)
107 #define COOKIE_EXPIRED_P(c) ((c)->expiry_time != 0 && (c)->expiry_time < cookies_now)
109 /* Allocate and return a new, empty cookie structure. */
111 static struct cookie *
114 struct cookie *cookie = xmalloc (sizeof (struct cookie));
115 memset (cookie, '\0', sizeof (struct cookie));
117 /* Both cookie->permanent and cookie->expiry_time are now 0. By
118 default, we assume that the cookie is non-permanent and valid
119 until the end of the session. */
121 cookie->port = PORT_ANY;
125 /* Deallocate COOKIE and its components. */
128 delete_cookie (struct cookie *cookie)
130 FREE_MAYBE (cookie->domain);
131 FREE_MAYBE (cookie->path);
132 FREE_MAYBE (cookie->attr);
133 FREE_MAYBE (cookie->value);
137 /* Functions for storing cookies.
139 All cookies can be reached beginning with jar->chains_by_domain.
140 The key in that table is the domain name, and the value is a linked
141 list of all cookies from that domain. Every new cookie is placed
142 on the head of the list. */
144 /* Find and return a cookie in JAR whose domain, path, and attribute
145 name correspond to COOKIE. If found, PREVPTR will point to the
146 location of the cookie previous in chain, or NULL if the found
147 cookie is the head of a chain.
149 If no matching cookie is found, return NULL. */
151 static struct cookie *
152 find_matching_cookie (struct cookie_jar *jar, struct cookie *cookie,
153 struct cookie **prevptr)
155 struct cookie *chain, *prev;
157 chain = hash_table_get (jar->chains_by_domain, cookie->domain);
162 for (; chain; prev = chain, chain = chain->next)
163 if (0 == strcmp (cookie->path, chain->path)
164 && 0 == strcmp (cookie->attr, chain->attr)
165 && cookie->port == chain->port)
176 /* Store COOKIE to the jar.
178 This is done by placing COOKIE at the head of its chain. However,
179 if COOKIE matches a cookie already in memory, as determined by
180 find_matching_cookie, the old cookie is unlinked and destroyed.
182 The key of each chain's hash table entry is allocated only the
183 first time; next hash_table_put's reuse the same key. */
186 store_cookie (struct cookie_jar *jar, struct cookie *cookie)
188 struct cookie *chain_head;
191 if (hash_table_get_pair (jar->chains_by_domain, cookie->domain,
192 &chain_key, &chain_head))
194 /* A chain of cookies in this domain already exists. Check for
195 duplicates -- if an extant cookie exactly matches our domain,
196 port, path, and name, replace it. */
198 struct cookie *victim = find_matching_cookie (jar, cookie, &prev);
202 /* Remove VICTIM from the chain. COOKIE will be placed at
206 prev->next = victim->next;
207 cookie->next = chain_head;
211 /* prev is NULL; apparently VICTIM was at the head of
212 the chain. This place will be taken by COOKIE, so
213 all we need to do is: */
214 cookie->next = victim->next;
216 delete_cookie (victim);
218 DEBUGP (("Deleted old cookie (to be replaced.)\n"));
221 cookie->next = chain_head;
225 /* We are now creating the chain. Allocate the string that will
226 be used as a key. It is unsafe to use cookie->domain for
227 that, because it might get deallocated by the above code at
230 chain_key = xstrdup (cookie->domain);
233 hash_table_put (jar->chains_by_domain, chain_key, cookie);
236 DEBUGP (("\nStored cookie %s %d%s %s %s %d %s %s %s\n",
237 cookie->domain, cookie->port,
238 cookie->port == PORT_ANY ? " (ANY)" : "",
240 cookie->permanent ? "permanent" : "nonpermanent",
243 ? asctime (localtime (&cookie->expiry_time)) : "<undefined>",
244 cookie->attr, cookie->value));
247 /* Discard a cookie matching COOKIE's domain, port, path, and
248 attribute name. This gets called when we encounter a cookie whose
249 expiry date is in the past, or whose max-age is set to 0. The
250 former corresponds to netscape cookie spec, while the latter is
251 specified by rfc2109. */
254 discard_matching_cookie (struct cookie_jar *jar, struct cookie *cookie)
256 struct cookie *prev, *victim;
258 if (!hash_table_count (jar->chains_by_domain))
259 /* No elements == nothing to discard. */
262 victim = find_matching_cookie (jar, cookie, &prev);
266 /* Simply unchain the victim. */
267 prev->next = victim->next;
270 /* VICTIM was head of its chain. We need to place a new
271 cookie at the head. */
272 char *chain_key = NULL;
275 res = hash_table_get_pair (jar->chains_by_domain, victim->domain,
280 /* VICTIM was the only cookie in the chain. Destroy the
281 chain and deallocate the chain key. */
282 hash_table_remove (jar->chains_by_domain, victim->domain);
286 hash_table_put (jar->chains_by_domain, chain_key, victim->next);
288 delete_cookie (victim);
289 DEBUGP (("Discarded old cookie.\n"));
293 /* Functions for parsing the `Set-Cookie' header, and creating new
294 cookies from the wire. */
297 #define NAME_IS(string_literal) \
298 BOUNDED_EQUAL_NO_CASE (name_b, name_e, string_literal)
300 #define VALUE_EXISTS (value_b && value_e)
302 #define VALUE_NON_EMPTY (VALUE_EXISTS && (value_b != value_e))
304 /* Update the appropriate cookie field. [name_b, name_e) are expected
305 to delimit the attribute name, while [value_b, value_e) (optional)
306 should delimit the attribute value.
308 When called the first time, it will set the cookie's attribute name
309 and value. After that, it will check the attribute name for
310 special fields such as `domain', `path', etc. Where appropriate,
311 it will parse the values of the fields it recognizes and fill the
312 corresponding fields in COOKIE.
314 Returns 1 on success. Returns zero in case a syntax error is
315 found; such a cookie should be discarded. */
318 update_cookie_field (struct cookie *cookie,
319 const char *name_b, const char *name_e,
320 const char *value_b, const char *value_e)
322 assert (name_b != NULL && name_e != NULL);
328 cookie->attr = strdupdelim (name_b, name_e);
329 cookie->value = strdupdelim (value_b, value_e);
333 if (NAME_IS ("domain"))
335 if (!VALUE_NON_EMPTY)
337 FREE_MAYBE (cookie->domain);
338 cookie->domain = strdupdelim (value_b, value_e);
341 else if (NAME_IS ("path"))
343 if (!VALUE_NON_EMPTY)
345 FREE_MAYBE (cookie->path);
346 cookie->path = strdupdelim (value_b, value_e);
349 else if (NAME_IS ("expires"))
354 if (!VALUE_NON_EMPTY)
356 BOUNDED_TO_ALLOCA (value_b, value_e, value_copy);
358 expires = http_atotm (value_copy);
361 cookie->permanent = 1;
362 cookie->expiry_time = (time_t)expires;
365 /* Error in expiration spec. Assume default (cookie valid for
369 /* According to netscape's specification, expiry time in the
370 past means that discarding of a matching cookie is
372 if (cookie->expiry_time < cookies_now)
373 cookie->discard_requested = 1;
377 else if (NAME_IS ("max-age"))
382 if (!VALUE_NON_EMPTY)
384 BOUNDED_TO_ALLOCA (value_b, value_e, value_copy);
386 sscanf (value_copy, "%lf", &maxage);
388 /* something went wrong. */
390 cookie->permanent = 1;
391 cookie->expiry_time = cookies_now + maxage;
393 /* According to rfc2109, a cookie with max-age of 0 means that
394 discarding of a matching cookie is requested. */
396 cookie->discard_requested = 1;
400 else if (NAME_IS ("secure"))
402 /* ignore value completely */
407 /* Unrecognized attribute; ignore it. */
413 /* Returns non-zero for characters that are legal in the name of an
414 attribute. This used to allow only alphanumerics, '-', and '_',
415 but we need to be more lenient because a number of sites wants to
416 use weirder attribute names. rfc2965 "informally specifies"
417 attribute name (token) as "a sequence of non-special, non-white
418 space characters". So we allow everything except the stuff we know
421 #define ATTR_NAME_CHAR(c) ((c) > 32 && (c) < 127 \
422 && (c) != '"' && (c) != '=' \
423 && (c) != ';' && (c) != ',')
425 /* Fetch the next character without doing anything special if CH gets
426 set to 0. (The code executed next is expected to handle it.) */
428 #define FETCH1(ch, ptr) do { \
432 /* Like FETCH1, but jumps to `eof' label if CH gets set to 0. */
434 #define FETCH(ch, ptr) do { \
440 /* Parse the contents of the `Set-Cookie' header. The header looks
443 name1=value1; name2=value2; ...
445 Trailing semicolon is optional; spaces are allowed between all
446 tokens. Additionally, values may be quoted.
448 A new cookie is returned upon success, NULL otherwise. The
449 function `update_cookie_field' is used to update the fields of the
450 newly created cookie structure. */
452 static struct cookie *
453 parse_set_cookies (const char *sc)
455 struct cookie *cookie = cookie_new ();
457 enum { S_NAME_PRE, S_NAME, S_NAME_POST,
458 S_VALUE_PRE, S_VALUE, S_VALUE_TRAILSPACE_MAYBE,
459 S_QUOTED_VALUE, S_QUOTED_VALUE_POST,
461 S_DONE, S_ERROR } state = S_NAME_PRE;
466 const char *name_b = NULL, *name_e = NULL;
467 const char *value_b = NULL, *value_e = NULL;
469 while (state != S_DONE && state != S_ERROR)
474 /* Strip whitespace preceding the name. */
477 while (c && ISSPACE (c));
480 else if (ATTR_NAME_CHAR (c))
487 /* empty attr name not allowed */
491 if (ATTR_NAME_CHAR (c))
493 else if (!c || c == ';' || c == '=' || ISSPACE (c))
504 else if (!c || c == ';')
506 value_b = value_e = NULL;
507 state = S_ATTR_ACTION;
524 state = S_QUOTED_VALUE;
526 else if (c == ';' || c == '\0')
528 value_b = value_e = p - 1;
529 state = S_ATTR_ACTION;
539 if (c == ';' || c == '\0')
543 state = S_ATTR_ACTION;
545 else if (ISSPACE (c))
549 state = S_VALUE_TRAILSPACE_MAYBE;
553 value_e = NULL; /* no trailing space */
557 case S_VALUE_TRAILSPACE_MAYBE:
568 state = S_QUOTED_VALUE_POST;
573 case S_QUOTED_VALUE_POST:
575 state = S_ATTR_ACTION;
576 else if (ISSPACE (c))
583 int legal = update_cookie_field (cookie, name_b, name_e,
588 BOUNDED_TO_ALLOCA (name_b, name_e, name);
589 logprintf (LOG_NOTQUIET,
590 _("Error in Set-Cookie, field `%s'"), name);
599 /* handled by loop condition */
606 delete_cookie (cookie);
607 if (state == S_ERROR)
608 logprintf (LOG_NOTQUIET, _("Syntax error in Set-Cookie at character `%c'.\n"), c);
614 delete_cookie (cookie);
615 logprintf (LOG_NOTQUIET,
616 _("Syntax error in Set-Cookie: premature end of string.\n"));
620 /* Sanity checks. These are important, otherwise it is possible for
621 mailcious attackers to destroy important cookie information and/or
622 violate your privacy. */
625 #define REQUIRE_DIGITS(p) do { \
628 for (++p; ISDIGIT (*p); p++) \
632 #define REQUIRE_DOT(p) do { \
637 /* Check whether ADDR matches <digits>.<digits>.<digits>.<digits>.
639 We don't want to call network functions like inet_addr() because all
640 we need is a check, preferrably one that is small, fast, and
644 numeric_address_p (const char *addr)
646 const char *p = addr;
648 REQUIRE_DIGITS (p); /* A */
649 REQUIRE_DOT (p); /* . */
650 REQUIRE_DIGITS (p); /* B */
651 REQUIRE_DOT (p); /* . */
652 REQUIRE_DIGITS (p); /* C */
653 REQUIRE_DOT (p); /* . */
654 REQUIRE_DIGITS (p); /* D */
661 /* Check whether COOKIE_DOMAIN is an appropriate domain for HOST.
662 Originally I tried to make the check compliant with rfc2109, but
663 the sites deviated too often, so I had to fall back to "tail
664 matching", as defined by the original Netscape's cookie spec. */
667 check_domain_match (const char *cookie_domain, const char *host)
671 /* Numeric address requires exact match. It also requires HOST to
673 if (numeric_address_p (cookie_domain))
674 return 0 == strcmp (cookie_domain, host);
678 /* For the sake of efficiency, check for exact match first. */
679 if (!strcasecmp (cookie_domain, host))
684 /* HOST must match the tail of cookie_domain. */
685 if (!match_tail (host, cookie_domain, 1))
688 /* We know that COOKIE_DOMAIN is a subset of HOST; however, we must
689 make sure that somebody is not trying to set the cookie for a
690 subdomain shared by many entities. For example, "company.co.uk"
691 must not be allowed to set a cookie for ".co.uk". On the other
692 hand, "sso.redhat.de" should be able to set a cookie for
695 The only marginally sane way to handle this I can think of is to
696 reject on the basis of the length of the second-level domain name
697 (but when the top-level domain is unknown), with the assumption
698 that those of three or less characters could be reserved. For
701 .co.org -> works because the TLD is known
702 .co.uk -> doesn't work because "co" is only two chars long
703 .com.au -> doesn't work because "com" is only 3 chars long
704 .cnn.uk -> doesn't work because "cnn" is also only 3 chars long (ugh)
705 .cnn.de -> doesn't work for the same reason (ugh!!)
706 .abcd.de -> works because "abcd" is 4 chars long
707 .img.cnn.de -> works because it's not trying to set the 2nd level domain
708 .cnn.co.uk -> works for the same reason
710 That should prevent misuse, while allowing reasonable usage. If
711 someone knows of a better way to handle this, please let me
714 const char *p = cookie_domain;
715 int dccount = 1; /* number of domain components */
716 int ldcl = 0; /* last domain component length */
717 int nldcl = 0; /* next to last domain component length */
720 /* Ignore leading period in this calculation. */
723 for (out = 0; !out; p++)
731 /* Empty domain component found -- the domain is invalid. */
733 if (*(p + 1) == '\0')
735 /* Tolerate trailing '.' by not treating the domain as
736 one ending with an empty domain component. */
758 int known_toplevel = 0;
759 static char *known_toplevel_domains[] = {
760 ".com", ".edu", ".net", ".org", ".gov", ".mil", ".int"
762 for (i = 0; i < ARRAY_SIZE (known_toplevel_domains); i++)
763 if (match_tail (cookie_domain, known_toplevel_domains[i], 1))
768 if (!known_toplevel && nldcl <= 3)
775 /* Don't allow domain "bar.com" to match host "foobar.com". */
776 if (*cookie_domain != '.')
778 int dlen = strlen (cookie_domain);
779 int hlen = strlen (host);
780 /* cookie host: hostname.foobar.com */
781 /* desired domain: bar.com */
782 /* '.' must be here in host-> ^ */
783 if (hlen > dlen && host[hlen - dlen - 1] != '.')
792 static int path_matches PARAMS ((const char *, const char *));
794 /* Check whether PATH begins with COOKIE_PATH. */
797 check_path_match (const char *cookie_path, const char *path)
799 return path_matches (path, cookie_path);
802 /* Process the HTTP `Set-Cookie' header. This results in storing the
803 cookie or discarding a matching one, or ignoring it completely, all
804 depending on the contents. */
807 cookie_jar_process_set_cookie (struct cookie_jar *jar,
808 const char *host, int port,
809 const char *path, const char *set_cookie)
811 struct cookie *cookie;
812 cookies_now = time (NULL);
814 cookie = parse_set_cookies (set_cookie);
818 /* Sanitize parts of cookie. */
823 cookie->domain = xstrdup (host);
828 if (!check_domain_match (cookie->domain, host))
830 logprintf (LOG_NOTQUIET,
831 "Cookie coming from %s attempted to set domain to %s\n",
832 host, cookie->domain);
837 cookie->path = xstrdup (path);
840 if (!check_path_match (cookie->path, path))
842 DEBUGP (("Attempt to fake the path: %s, %s\n",
843 cookie->path, path));
848 if (cookie->discard_requested)
850 discard_matching_cookie (jar, cookie);
854 store_cookie (jar, cookie);
859 delete_cookie (cookie);
862 /* Support for sending out cookies in HTTP requests, based on
863 previously stored cookies. Entry point is
864 `build_cookies_request'. */
866 /* Store CHAIN to STORE if there is room in STORE. If not, inrecement
867 COUNT anyway, so that when the function is done, we end up with the
868 exact count of how much place we actually need. */
870 #define STORE_CHAIN(st_chain, st_store, st_size, st_count) do { \
871 if (st_count < st_size) \
872 store[st_count] = st_chain; \
876 /* Store cookie chains that match HOST. Since more than one chain can
877 match, the matches are written to STORE. No more than SIZE matches
878 are written; if more matches are present, return the number of
879 chains that would have been written. */
882 find_matching_chains (struct cookie_jar *jar, const char *host,
883 struct cookie *store[], int size)
885 struct cookie *chain;
890 if (!hash_table_count (jar->chains_by_domain))
893 STRDUP_ALLOCA (hash_key, host);
895 /* Look for an exact match. */
896 chain = hash_table_get (jar->chains_by_domain, hash_key);
898 STORE_CHAIN (chain, store, size, count);
900 dot_count = count_char (host, '.');
902 /* Match less and less specific domains. For instance, given
903 fly.srk.fer.hr, we match .srk.fer.hr, then .fer.hr. */
904 while (dot_count-- > 1)
906 /* Note: we operate directly on hash_key (in form host:port)
907 because we don't want to allocate new hash keys in a
909 char *p = strchr (hash_key, '.');
911 chain = hash_table_get (jar->chains_by_domain, p);
913 STORE_CHAIN (chain, store, size, count);
919 /* If FULL_PATH begins with PREFIX, return the length of PREFIX, zero
923 path_matches (const char *full_path, const char *prefix)
928 /* Wget's HTTP paths do not begin with '/' (the URL code treats it
929 as a separator), but the '/' is assumed when matching against
934 len = strlen (prefix);
936 if (0 != strncmp (full_path, prefix, len))
937 /* FULL_PATH doesn't begin with PREFIX. */
940 /* Length of PREFIX determines the quality of the match. */
944 /* Return non-zero iff COOKIE matches the given PATH, PORT, and
945 security flag. HOST is not a flag because it is assumed that the
946 cookie comes from the correct chain.
948 If PATH_GOODNESS is non-NULL, store the "path goodness" there. The
949 said goodness is a measure of how well COOKIE matches PATH. It is
950 used for ordering cookies. */
953 matching_cookie (const struct cookie *cookie, const char *path, int port,
954 int connection_secure_p, int *path_goodness)
958 if (COOKIE_EXPIRED_P (cookie))
959 /* Ignore stale cookies. Don't bother unchaining the cookie at
960 this point -- Wget is a relatively short-lived application, and
961 stale cookies will not be saved by `save_cookies'. On the
962 other hand, this function should be as efficient as
966 if (cookie->secure && !connection_secure_p)
967 /* Don't transmit secure cookies over an insecure connection. */
969 if (cookie->port != PORT_ANY && cookie->port != port)
971 pg = path_matches (path, cookie->path);
976 /* If the caller requested path_goodness, we return it. This is
977 an optimization, so that the caller doesn't need to call
978 path_matches() again. */
983 struct weighed_cookie {
984 struct cookie *cookie;
989 /* Comparator used for uniquifying the list. */
992 equality_comparator (const void *p1, const void *p2)
994 struct weighed_cookie *wc1 = (struct weighed_cookie *)p1;
995 struct weighed_cookie *wc2 = (struct weighed_cookie *)p2;
997 int namecmp = strcmp (wc1->cookie->attr, wc2->cookie->attr);
998 int valuecmp = strcmp (wc1->cookie->value, wc2->cookie->value);
1000 /* We only really care whether both name and value are equal. We
1001 return them in this order only for consistency... */
1002 return namecmp ? namecmp : valuecmp;
1005 /* Eliminate duplicate cookies. "Duplicate cookies" are any two
1006 cookies whose name and value are the same. Whenever a duplicate
1007 pair is found, one of the cookies is removed. */
1010 eliminate_dups (struct weighed_cookie *outgoing, int count)
1014 /* We deploy a simple uniquify algorithm: first sort the array
1015 according to our sort criterion, then uniquify it by comparing
1016 each cookie with its neighbor. */
1018 qsort (outgoing, count, sizeof (struct weighed_cookie), equality_comparator);
1020 for (i = 0; i < count - 1; i++)
1022 struct cookie *c1 = outgoing[i].cookie;
1023 struct cookie *c2 = outgoing[i + 1].cookie;
1024 if (!strcmp (c1->attr, c2->attr) && !strcmp (c1->value, c2->value))
1026 /* c1 and c2 are the same; get rid of c2. */
1028 /* move all ptrs from positions [i + 1, count) to i. */
1029 memmove (outgoing + i, outgoing + i + 1,
1030 (count - (i + 1)) * sizeof (struct weighed_cookie));
1031 /* We decrement i to counter the ++i above. Remember that
1032 we've just removed the element in front of us; we need to
1033 remain in place to check whether outgoing[i] matches what
1034 used to be outgoing[i + 2]. */
1042 /* Comparator used for sorting by quality. */
1045 goodness_comparator (const void *p1, const void *p2)
1047 struct weighed_cookie *wc1 = (struct weighed_cookie *)p1;
1048 struct weighed_cookie *wc2 = (struct weighed_cookie *)p2;
1050 /* Subtractions take `wc2' as the first argument becauase we want a
1051 sort in *decreasing* order of goodness. */
1052 int dgdiff = wc2->domain_goodness - wc1->domain_goodness;
1053 int pgdiff = wc2->path_goodness - wc1->path_goodness;
1055 /* Sort by domain goodness; if these are the same, sort by path
1056 goodness. (The sorting order isn't really specified; maybe it
1057 should be the other way around.) */
1058 return dgdiff ? dgdiff : pgdiff;
1061 /* Generate a `Cookie' header for a request that goes to HOST:PORT and
1062 requests PATH from the server. The resulting string is allocated
1063 with `malloc', and the caller is responsible for freeing it. If no
1064 cookies pertain to this request, i.e. no cookie header should be
1065 generated, NULL is returned. */
1068 cookie_jar_generate_cookie_header (struct cookie_jar *jar, const char *host,
1069 int port, const char *path,
1070 int connection_secure_p)
1072 struct cookie *chain_default_store[20];
1073 struct cookie **all_chains = chain_default_store;
1074 int chain_store_size = ARRAY_SIZE (chain_default_store);
1077 struct cookie *cookie;
1078 struct weighed_cookie *outgoing;
1081 int result_size, pos;
1084 chain_count = find_matching_chains (jar, host, all_chains, chain_store_size);
1085 if (chain_count > chain_store_size)
1087 /* It's extremely unlikely that more than 20 chains will ever
1088 match. But since find_matching_chains reports the exact size
1089 it needs, it's easy to not have the limitation, so we
1091 all_chains = alloca (chain_count * sizeof (struct cookie *));
1092 chain_store_size = chain_count;
1099 cookies_now = time (NULL);
1101 /* Count the number of cookies whose path matches. */
1103 for (i = 0; i < chain_count; i++)
1104 for (cookie = all_chains[i]; cookie; cookie = cookie->next)
1105 if (matching_cookie (cookie, path, port, connection_secure_p, NULL))
1108 /* No matching cookies. */
1111 /* Allocate the array. */
1112 outgoing = alloca (count * sizeof (struct weighed_cookie));
1114 /* Fill the array with all the matching cookies from all the
1117 for (i = 0; i < chain_count; i++)
1118 for (cookie = all_chains[i]; cookie; cookie = cookie->next)
1121 if (!matching_cookie (cookie, path, port, connection_secure_p, &pg))
1123 outgoing[ocnt].cookie = cookie;
1124 outgoing[ocnt].domain_goodness = strlen (cookie->domain);
1125 outgoing[ocnt].path_goodness = pg;
1128 assert (ocnt == count);
1130 /* Eliminate duplicate cookies; that is, those whose name and value
1132 count = eliminate_dups (outgoing, count);
1134 /* Sort the array so that best-matching domains come first, and
1135 that, within one domain, best-matching paths come first. */
1136 qsort (outgoing, count, sizeof (struct weighed_cookie), goodness_comparator);
1138 /* Count the space the name=value pairs will take. */
1140 for (i = 0; i < count; i++)
1142 struct cookie *c = outgoing[i].cookie;
1144 result_size += strlen (c->attr) + 1 + strlen (c->value);
1147 /* Allocate output buffer:
1149 name=value pairs -- result_size
1150 "; " separators -- (count - 1) * 2
1151 \r\n line ending -- 2
1152 \0 terminator -- 1 */
1153 result_size = 8 + result_size + (count - 1) * 2 + 2 + 1;
1154 result = xmalloc (result_size);
1156 strcpy (result, "Cookie: ");
1158 for (i = 0; i < count; i++)
1160 struct cookie *c = outgoing[i].cookie;
1161 int namlen = strlen (c->attr);
1162 int vallen = strlen (c->value);
1164 memcpy (result + pos, c->attr, namlen);
1166 result[pos++] = '=';
1167 memcpy (result + pos, c->value, vallen);
1171 result[pos++] = ';';
1172 result[pos++] = ' ';
1175 result[pos++] = '\r';
1176 result[pos++] = '\n';
1177 result[pos++] = '\0';
1178 assert (pos == result_size);
1182 /* Support for loading and saving cookies. The format used for
1183 loading and saving roughly matches the format of `cookies.txt' file
1184 used by Netscape and Mozilla, at least the Unix versions. The
1185 format goes like this:
1187 DOMAIN DOMAIN-FLAG PATH SECURE-FLAG TIMESTAMP ATTR-NAME ATTR-VALUE
1189 DOMAIN -- cookie domain, optionally followed by :PORT
1190 DOMAIN-FLAG -- whether all hosts in the domain match
1192 SECURE-FLAG -- whether cookie requires secure connection
1193 TIMESTAMP -- expiry timestamp, number of seconds since epoch
1194 ATTR-NAME -- name of the cookie attribute
1195 ATTR-VALUE -- value of the cookie attribute (empty if absent)
1197 The fields are separated by TABs (but Wget's loader recognizes any
1198 whitespace). All fields are mandatory, except for ATTR-VALUE. The
1199 `-FLAG' fields are boolean, their legal values being "TRUE" and
1200 "FALSE'. Empty lines, lines consisting of whitespace only, and
1201 comment lines (beginning with # optionally preceded by whitespace)
1204 Example line from cookies.txt (split in two lines for readability):
1206 .google.com TRUE / FALSE 2147368447 \
1207 PREF ID=34bb47565bbcd47b:LD=en:NR=20:TM=985172580:LM=985739012
1209 DOMAIN-FLAG is currently not honored by Wget. The cookies whose
1210 domain begins with `.' are treated as if DOMAIN-FLAG were true,
1211 while all other cookies are treated as if it were FALSE. */
1214 /* If the region [B, E) ends with :<digits>, parse the number, return
1215 it, and store new boundary (location of the `:') to DOMAIN_E_PTR.
1216 If port is not specified, return 0. */
1219 domain_port (const char *domain_b, const char *domain_e,
1220 const char **domain_e_ptr)
1224 const char *colon = memchr (domain_b, ':', domain_e - domain_b);
1227 for (p = colon + 1; p < domain_e && ISDIGIT (*p); p++)
1228 port = 10 * port + (*p - '0');
1230 /* Garbage following port number. */
1232 *domain_e_ptr = colon;
1236 #define SKIP_WS(p) do { \
1237 while (*p && ISSPACE (*p)) \
1241 #define SET_WORD_BOUNDARIES(p, b, e) do { \
1245 while (*p && !ISSPACE (*p)) \
1252 /* Load cookies from FILE. */
1255 cookie_jar_load (struct cookie_jar *jar, const char *file)
1258 FILE *fp = fopen (file, "r");
1261 logprintf (LOG_NOTQUIET, "Cannot open cookies file `%s': %s\n",
1262 file, strerror (errno));
1265 cookies_now = time (NULL);
1267 for (; ((line = read_whole_line (fp)) != NULL); xfree (line))
1269 struct cookie *cookie;
1275 char *domain_b = NULL, *domain_e = NULL;
1276 char *ignore_b = NULL, *ignore_e = NULL;
1277 char *path_b = NULL, *path_e = NULL;
1278 char *secure_b = NULL, *secure_e = NULL;
1279 char *expires_b = NULL, *expires_e = NULL;
1280 char *name_b = NULL, *name_e = NULL;
1281 char *value_b = NULL, *value_e = NULL;
1285 if (!*p || *p == '#')
1289 SET_WORD_BOUNDARIES (p, domain_b, domain_e);
1290 SET_WORD_BOUNDARIES (p, ignore_b, ignore_e);
1291 SET_WORD_BOUNDARIES (p, path_b, path_e);
1292 SET_WORD_BOUNDARIES (p, secure_b, secure_e);
1293 SET_WORD_BOUNDARIES (p, expires_b, expires_e);
1294 SET_WORD_BOUNDARIES (p, name_b, name_e);
1296 /* Don't use SET_WORD_BOUNDARIES for value because it may
1297 contain whitespace. Instead, set value_e to the end of line,
1298 modulo trailing space (this will skip the line separator.) */
1301 value_e = p + strlen (p);
1302 while (value_e > value_b && ISSPACE (*(value_e - 1)))
1304 if (value_b == value_e)
1305 /* Hmm, should we check for empty value? I guess that's
1306 legal, so I leave it. */
1309 cookie = cookie_new ();
1311 cookie->attr = strdupdelim (name_b, name_e);
1312 cookie->value = strdupdelim (value_b, value_e);
1313 cookie->path = strdupdelim (path_b, path_e);
1315 if (BOUNDED_EQUAL (secure_b, secure_e, "TRUE"))
1318 /* DOMAIN needs special treatment because we might need to
1319 extract the port. */
1320 port = domain_port (domain_b, domain_e, (const char **)&domain_e);
1322 cookie->port = port;
1323 cookie->domain = strdupdelim (domain_b, domain_e);
1325 /* safe default in case EXPIRES field is garbled. */
1326 expiry = (double)cookies_now - 1;
1328 /* I don't like changing the line, but it's completely safe.
1329 (line is malloced.) */
1331 sscanf (expires_b, "%lf", &expiry);
1332 if (expiry < cookies_now)
1333 /* ignore stale cookie. */
1335 cookie->expiry_time = expiry;
1337 /* If the cookie has survived being saved into an external file,
1338 it is obviously permanent. */
1339 cookie->permanent = 1;
1341 store_cookie (jar, cookie);
1347 delete_cookie (cookie);
1352 /* Mapper for save_cookies callable by hash_table_map. VALUE points
1353 to the head in a chain of cookies. The function prints the entire
1357 save_cookies_mapper (void *key, void *value, void *arg)
1359 FILE *fp = (FILE *)arg;
1360 char *domain = (char *)key;
1361 struct cookie *chain = (struct cookie *)value;
1362 for (; chain; chain = chain->next)
1364 if (!chain->permanent)
1366 if (COOKIE_EXPIRED_P (chain))
1369 if (chain->port != PORT_ANY)
1370 fprintf (fp, ":%d", chain->port);
1371 fprintf (fp, "\t%s\t%s\t%s\t%.0f\t%s\t%s\n",
1372 *domain == '.' ? "TRUE" : "FALSE",
1373 chain->path, chain->secure ? "TRUE" : "FALSE",
1374 (double)chain->expiry_time,
1375 chain->attr, chain->value);
1377 return 1; /* stop mapping */
1382 /* Save cookies, in format described above, to FILE. */
1385 cookie_jar_save (struct cookie_jar *jar, const char *file)
1389 DEBUGP (("Saving cookies to %s.\n", file));
1391 cookies_now = time (NULL);
1393 fp = fopen (file, "w");
1396 logprintf (LOG_NOTQUIET, _("Cannot open cookies file `%s': %s\n"),
1397 file, strerror (errno));
1401 fputs ("# HTTP cookie file.\n", fp);
1402 fprintf (fp, "# Generated by Wget on %s.\n", datetime_str (NULL));
1403 fputs ("# Edit at your own risk.\n\n", fp);
1405 hash_table_map (jar->chains_by_domain, save_cookies_mapper, fp);
1408 logprintf (LOG_NOTQUIET, _("Error writing to `%s': %s\n"),
1409 file, strerror (errno));
1411 if (fclose (fp) < 0)
1412 logprintf (LOG_NOTQUIET, _("Error closing `%s': %s\n"),
1413 file, strerror (errno));
1415 DEBUGP (("Done saving cookies.\n"));
1418 /* Destroy all the elements in the chain and unhook it from the cookie
1419 jar. This is written in the form of a callback to hash_table_map
1420 and used by cookie_jar_delete to delete all the cookies in a
1424 nuke_cookie_chain (void *value, void *key, void *arg)
1426 char *chain_key = (char *)value;
1427 struct cookie *chain = (struct cookie *)key;
1428 struct cookie_jar *jar = (struct cookie_jar *)arg;
1430 /* Remove the chain from the table and free the key. */
1431 hash_table_remove (jar->chains_by_domain, chain_key);
1434 /* Then delete all the cookies in the chain. */
1437 struct cookie *next = chain->next;
1438 delete_cookie (chain);
1446 /* Clean up cookie-related data. */
1449 cookie_jar_delete (struct cookie_jar *jar)
1451 hash_table_map (jar->chains_by_domain, nuke_cookie_chain, jar);
1452 hash_table_destroy (jar->chains_by_domain);