diff options
Diffstat (limited to 'src/cookies.c')
| -rw-r--r-- | src/cookies.c | 1539 |
1 files changed, 1539 insertions, 0 deletions
diff --git a/src/cookies.c b/src/cookies.c new file mode 100644 index 0000000..8d903d7 --- /dev/null +++ b/src/cookies.c @@ -0,0 +1,1539 @@ +/* Support for cookies. + Copyright (C) 2001-2011, 2015, 2018-2022 Free Software Foundation, + Inc. + +This file is part of GNU Wget. + +GNU Wget is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3 of the License, or (at +your option) any later version. + +GNU Wget is distributed in the hope that it will be useful, but +WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +General Public License for more details. + +You should have received a copy of the GNU General Public License +along with Wget. If not, see <http://www.gnu.org/licenses/>. + +Additional permission under GNU GPL version 3 section 7 + +If you modify this program, or any covered work, by linking or +combining it with the OpenSSL project's OpenSSL library (or a +modified version of that library), containing parts covered by the +terms of the OpenSSL or SSLeay licenses, the Free Software Foundation +grants you additional permission to convey the resulting work. +Corresponding Source for a non-source form of such a combination +shall include the source code for the parts of OpenSSL used as well +as that of the covered work. */ + +/* Written by Hrvoje Niksic. Parts are loosely inspired by the + cookie patch submitted by Tomasz Wegrzanowski. + + This implements the client-side cookie support, as specified + (loosely) by Netscape's "preliminary specification", currently + available at: + + http://wp.netscape.com/newsref/std/cookie_spec.html + + rfc2109 is not supported because of its incompatibilities with the + above widely-used specification. rfc2965 is entirely ignored, + since popular client software doesn't implement it, and even the + sites that do send Set-Cookie2 also emit Set-Cookie for + compatibility. */ + +#include "wget.h" + +#include <stdint.h> +#include <stdio.h> +#include <string.h> +#include <stdlib.h> +#include <assert.h> +#include <errno.h> +#include <time.h> +#ifdef HAVE_LIBPSL +# include <libpsl.h> +#endif +#include "utils.h" +#include "hash.h" +#include "cookies.h" +#include "http.h" /* for http_atotm */ +#include "c-strcase.h" + + +/* Declarations of `struct cookie' and the most basic functions. */ + +/* Cookie jar serves as cookie storage and a means of retrieving + cookies efficiently. All cookies with the same domain are stored + in a linked list called "chain". A cookie chain can be reached by + looking up the domain in the cookie jar's chains_by_domain table. + + For example, to reach all the cookies under google.com, one must + execute hash_table_get(jar->chains_by_domain, "google.com"). Of + course, when sending a cookie to `www.google.com', one must search + for cookies that belong to either `www.google.com' or `google.com' + -- but the point is that the code doesn't need to go through *all* + the cookies. */ + +struct cookie_jar { + /* Cookie chains indexed by domain. */ + struct hash_table *chains; + + int cookie_count; /* number of cookies in the jar. */ +}; + +/* Value set by entry point functions, so that the low-level + routines don't need to call time() all the time. */ +static time_t cookies_now; + +struct cookie_jar * +cookie_jar_new (void) +{ + struct cookie_jar *jar = xnew (struct cookie_jar); + jar->chains = make_nocase_string_hash_table (0); + jar->cookie_count = 0; + return jar; +} + +struct cookie { + char *domain; /* domain of the cookie */ + int port; /* port number */ + char *path; /* path prefix of the cookie */ + + unsigned discard_requested :1;/* whether cookie was created to + request discarding another + cookie. */ + + unsigned secure :1; /* whether cookie should be + transmitted over non-https + connections. */ + unsigned domain_exact :1; /* whether DOMAIN must match as a + whole. */ + + unsigned permanent :1; /* whether the cookie should outlive + the session. */ + time_t expiry_time; /* time when the cookie expires, 0 + means undetermined. */ + + char *attr; /* cookie attribute name */ + char *value; /* cookie attribute value */ + + struct cookie *next; /* used for chaining of cookies in the + same domain. */ +}; + +#define PORT_ANY (-1) + +/* Allocate and return a new, empty cookie structure. */ + +static struct cookie * +cookie_new (void) +{ + struct cookie *cookie = xnew0 (struct cookie); + + /* Both cookie->permanent and cookie->expiry_time are now 0. This + means that the cookie doesn't expire, but is only valid for this + session (i.e. not written out to disk). */ + + cookie->port = PORT_ANY; + return cookie; +} + +/* Non-zero if the cookie has expired. Assumes cookies_now has been + set by one of the entry point functions. */ + +static bool +cookie_expired_p (const struct cookie *c) +{ + return c->expiry_time != 0 && c->expiry_time < cookies_now; +} + +/* Deallocate COOKIE and its components. */ + +static void +delete_cookie (struct cookie *cookie) +{ + xfree (cookie->domain); + xfree (cookie->path); + xfree (cookie->attr); + xfree (cookie->value); + xfree (cookie); +} + +/* Functions for storing cookies. + + All cookies can be reached beginning with jar->chains. The key in + that table is the domain name, and the value is a linked list of + all cookies from that domain. Every new cookie is placed on the + head of the list. */ + +/* Find and return a cookie in JAR whose domain, path, and attribute + name correspond to COOKIE. If found, PREVPTR will point to the + location of the cookie previous in chain, or NULL if the found + cookie is the head of a chain. + + If no matching cookie is found, return NULL. */ + +static struct cookie * +find_matching_cookie (struct cookie_jar *jar, struct cookie *cookie, + struct cookie **prevptr) +{ + struct cookie *chain, *prev; + + chain = hash_table_get (jar->chains, cookie->domain); + if (!chain) + goto nomatch; + + prev = NULL; + for (; chain; prev = chain, chain = chain->next) + if (0 == strcmp (cookie->path, chain->path) + && 0 == strcmp (cookie->attr, chain->attr) + && cookie->port == chain->port) + { + *prevptr = prev; + return chain; + } + + nomatch: + *prevptr = NULL; + return NULL; +} + +/* Store COOKIE to the jar. + + This is done by placing COOKIE at the head of its chain. However, + if COOKIE matches a cookie already in memory, as determined by + find_matching_cookie, the old cookie is unlinked and destroyed. + + The key of each chain's hash table entry is allocated only the + first time; next hash_table_put's reuse the same key. */ + +static void +store_cookie (struct cookie_jar *jar, struct cookie *cookie) +{ + struct cookie *chain_head; + char *chain_key; + + if (hash_table_get_pair (jar->chains, cookie->domain, + &chain_key, &chain_head)) + { + /* A chain of cookies in this domain already exists. Check for + duplicates -- if an extant cookie exactly matches our domain, + port, path, and name, replace it. */ + struct cookie *prev; + struct cookie *victim = find_matching_cookie (jar, cookie, &prev); + + if (victim) + { + /* Remove VICTIM from the chain. COOKIE will be placed at + the head. */ + if (prev) + { + prev->next = victim->next; + cookie->next = chain_head; + } + else + { + /* prev is NULL; apparently VICTIM was at the head of + the chain. This place will be taken by COOKIE, so + all we need to do is: */ + cookie->next = victim->next; + } + delete_cookie (victim); + --jar->cookie_count; + DEBUGP (("Deleted old cookie (to be replaced.)\n")); + } + else + cookie->next = chain_head; + } + else + { + /* We are now creating the chain. Use a copy of cookie->domain + as the key for the life-time of the chain. Using + cookie->domain would be unsafe because the life-time of the + chain may exceed the life-time of the cookie. (Cookies may + be deleted from the chain by this very function.) */ + cookie->next = NULL; + chain_key = xstrdup (cookie->domain); + } + + hash_table_put (jar->chains, chain_key, cookie); + ++jar->cookie_count; + + IF_DEBUG + { + time_t exptime = cookie->expiry_time; + DEBUGP (("\nStored cookie %s %d%s %s <%s> <%s> [expiry %s] %s %s\n", + cookie->domain, cookie->port, + cookie->port == PORT_ANY ? " (ANY)" : "", + cookie->path, + cookie->permanent ? "permanent" : "session", + cookie->secure ? "secure" : "insecure", + cookie->expiry_time ? datetime_str (exptime) : "none", + cookie->attr, cookie->value)); + } +} + +/* Discard a cookie matching COOKIE's domain, port, path, and + attribute name. This gets called when we encounter a cookie whose + expiry date is in the past, or whose max-age is set to 0. The + former corresponds to netscape cookie spec, while the latter is + specified by rfc2109. */ + +static void +discard_matching_cookie (struct cookie_jar *jar, struct cookie *cookie) +{ + struct cookie *prev, *victim; + + if (!hash_table_count (jar->chains)) + /* No elements == nothing to discard. */ + return; + + victim = find_matching_cookie (jar, cookie, &prev); + if (victim) + { + if (prev) + /* Simply unchain the victim. */ + prev->next = victim->next; + else + { + /* VICTIM was head of its chain. We need to place a new + cookie at the head. */ + char *chain_key = NULL; + int res; + + res = hash_table_get_pair (jar->chains, victim->domain, + &chain_key, NULL); + + if (res == 0) + { + logprintf (LOG_VERBOSE, _("Unable to get cookie for %s\n"), + victim->domain); + } + if (!victim->next) + { + /* VICTIM was the only cookie in the chain. Destroy the + chain and deallocate the chain key. */ + hash_table_remove (jar->chains, victim->domain); + xfree (chain_key); + } + else + hash_table_put (jar->chains, chain_key, victim->next); + } + delete_cookie (victim); + DEBUGP (("Discarded old cookie.\n")); + } +} + +/* Functions for parsing the `Set-Cookie' header, and creating new + cookies from the wire. */ + +#define TOKEN_IS(token, string_literal) \ + BOUNDED_EQUAL_NO_CASE (token.b, token.e, string_literal) + +#define TOKEN_NON_EMPTY(token) (token.b != NULL && token.b != token.e) + +/* Parse the contents of the `Set-Cookie' header. The header looks + like this: + + name1=value1; name2=value2; ... + + Trailing semicolon is optional; spaces are allowed between all + tokens. Additionally, values may be quoted. + + A new cookie is returned upon success, NULL otherwise. + + The first name-value pair will be used to set the cookie's + attribute name and value. Subsequent parameters will be checked + against field names such as `domain', `path', etc. Recognized + fields will be parsed and the corresponding members of COOKIE + filled. */ + +static struct cookie * +parse_set_cookie (const char *set_cookie, bool silent) +{ + const char *ptr = set_cookie; + struct cookie *cookie = cookie_new (); + param_token name, value; + + if (!extract_param (&ptr, &name, &value, ';', NULL)) + goto error; + if (!value.b) + goto error; + + /* If the value is quoted, do not modify it. */ + if (*(value.b - 1) == '"') + value.b--; + if (*value.e == '"') + value.e++; + + cookie->attr = strdupdelim (name.b, name.e); + cookie->value = strdupdelim (value.b, value.e); + + while (extract_param (&ptr, &name, &value, ';', NULL)) + { + if (TOKEN_IS (name, "domain")) + { + if (!TOKEN_NON_EMPTY (value)) + goto error; + xfree (cookie->domain); + /* Strictly speaking, we should set cookie->domain_exact if the + domain doesn't begin with a dot. But many sites set the + domain to "foo.com" and expect "subhost.foo.com" to get the + cookie, and it apparently works in browsers. */ + if (*value.b == '.') + ++value.b; + cookie->domain = strdupdelim (value.b, value.e); + } + else if (TOKEN_IS (name, "path")) + { + if (!TOKEN_NON_EMPTY (value)) + goto error; + xfree (cookie->path); + cookie->path = strdupdelim (value.b, value.e); + } + else if (TOKEN_IS (name, "expires")) + { + char value_copy[128]; + size_t value_len = value.e - value.b; + time_t expires; + + if (!TOKEN_NON_EMPTY (value) || value_len >= sizeof (value_copy)) + goto error; + + memcpy (value_copy, value.b, value_len); + value_copy[value_len] = 0; + + /* Check if expiration spec is valid. + If not, assume default (cookie doesn't expire, but valid only for + this session.) */ + expires = http_atotm (value_copy); + if (expires != (time_t) -1) + { + cookie->permanent = 1; + cookie->expiry_time = expires; + /* According to netscape's specification, expiry time in + the past means that discarding of a matching cookie + is requested. */ + if (cookie->expiry_time < cookies_now) + cookie->discard_requested = 1; + } + } + else if (TOKEN_IS (name, "max-age")) + { + double maxage = -1; + char value_copy[32]; + size_t value_len = value.e - value.b; + + if (!TOKEN_NON_EMPTY (value) || value_len >= sizeof (value_copy)) + goto error; + + memcpy (value_copy, value.b, value_len); + value_copy[value_len] = 0; + + sscanf (value_copy, "%lf", &maxage); + if (maxage == -1) + /* something went wrong. */ + goto error; + cookie->permanent = 1; + cookie->expiry_time = cookies_now + (time_t) maxage; + + /* According to rfc2109, a cookie with max-age of 0 means that + discarding of a matching cookie is requested. */ + if (maxage == 0) + cookie->discard_requested = 1; + } + else if (TOKEN_IS (name, "secure")) + { + /* ignore value completely */ + cookie->secure = 1; + } + /* else: Ignore unrecognized attribute. */ + } + if (*ptr) + /* extract_param has encountered a syntax error */ + goto error; + + /* The cookie has been successfully constructed; return it. */ + return cookie; + + error: + if (!silent) + logprintf (LOG_NOTQUIET, + _("Syntax error in Set-Cookie: %s at position %d.\n"), + quotearg_style (escape_quoting_style, set_cookie), + (int) (ptr - set_cookie)); + delete_cookie (cookie); + return NULL; +} + +#undef TOKEN_IS +#undef TOKEN_NON_EMPTY + +/* Sanity checks. These are important, otherwise it is possible for + mailcious attackers to destroy important cookie information and/or + violate your privacy. */ + + +#define REQUIRE_DIGITS(p) do { \ + if (!c_isdigit (*p)) \ + return false; \ + for (++p; c_isdigit (*p); p++) \ + ; \ +} while (0) + +#define REQUIRE_DOT(p) do { \ + if (*p++ != '.') \ + return false; \ +} while (0) + +/* Check whether ADDR matches <digits>.<digits>.<digits>.<digits>. + + We don't want to call network functions like inet_addr() because + all we need is a check, preferably one that is small, fast, and + well-defined. */ + +static bool +numeric_address_p (const char *addr) +{ + const char *p = addr; + + REQUIRE_DIGITS (p); /* A */ + REQUIRE_DOT (p); /* . */ + REQUIRE_DIGITS (p); /* B */ + REQUIRE_DOT (p); /* . */ + REQUIRE_DIGITS (p); /* C */ + REQUIRE_DOT (p); /* . */ + REQUIRE_DIGITS (p); /* D */ + + if (*p != '\0') + return false; + return true; +} + +/* Check whether COOKIE_DOMAIN is an appropriate domain for HOST. + Originally I tried to make the check compliant with rfc2109, but + the sites deviated too often, so I had to fall back to "tail + matching", as defined by the original Netscape's cookie spec. + + Wget now uses libpsl to check domain names against a public suffix + list to see if they are valid. However, since we don't provide a + psl on our own, if libpsl is compiled without a public suffix list, + fall back to using the original "tail matching" heuristic. Also if + libpsl is unable to convert the domain to lowercase, which means that + it doesn't have any runtime conversion support, we again fall back to + "tail matching" since libpsl states the results are unpredictable with + upper case strings. + */ + +#ifdef HAVE_LIBPSL +static psl_ctx_t *psl; +#endif + +static bool +check_domain_match (const char *cookie_domain, const char *host) +{ +#ifdef HAVE_LIBPSL + static int init_psl; + char *cookie_domain_lower = NULL; + char *host_lower = NULL; + int is_acceptable; + + DEBUGP (("cdm: 1\n")); + if (!init_psl) + { + init_psl = 1; + +#ifdef HAVE_PSL_LATEST + if ((psl = psl_latest (NULL))) + goto have_psl; + + DEBUGP (("\nPSL: Failed to load any PSL data. " + "Falling back to insecure heuristics.\n")); +#else + if ((psl = psl_builtin ()) && !psl_builtin_outdated ()) + goto have_psl; + + DEBUGP (("\nPSL: built-in data outdated. " + "Trying to load data from %s.\n", + quote (psl_builtin_filename ()))); + + if ((psl = psl_load_file (psl_builtin_filename ()))) + goto have_psl; + + DEBUGP (("\nPSL: %s not found or not readable. " + "Falling back to built-in data.\n", + quote (psl_builtin_filename ()))); + + if (!(psl = psl_builtin ())) + { + DEBUGP (("\nPSL: libpsl not built with a public suffix list. " + "Falling back to insecure heuristics.\n")); + goto no_psl; + } +#endif + } + else if (!psl) + goto no_psl; + +have_psl: + if (psl_str_to_utf8lower (cookie_domain, NULL, NULL, &cookie_domain_lower) == PSL_SUCCESS && + psl_str_to_utf8lower (host, NULL, NULL, &host_lower) == PSL_SUCCESS) + { + is_acceptable = psl_is_cookie_domain_acceptable (psl, host_lower, cookie_domain_lower); + } + else + { + DEBUGP (("libpsl unable to parse domain name. " + "Falling back to simple heuristics.\n")); + goto no_psl; + } + + xfree (cookie_domain_lower); + xfree (host_lower); + + return is_acceptable == 1; + +no_psl: + /* Cleanup the PSL pointers first */ + xfree (cookie_domain_lower); + xfree (host_lower); +#endif + + /* For efficiency make some elementary checks first */ + DEBUGP (("cdm: 2\n")); + + /* For the sake of efficiency, check for exact match first. */ + if (0 == strcasecmp (cookie_domain, host)) + return true; + + DEBUGP (("cdm: 3\n")); + + /* HOST must match the tail of cookie_domain. */ + if (!match_tail (host, cookie_domain, true)) + return false; + + /* We know that COOKIE_DOMAIN is a subset of HOST; however, we must + make sure that somebody is not trying to set the cookie for a + subdomain shared by many entities. For example, "company.co.uk" + must not be allowed to set a cookie for ".co.uk". On the other + hand, "sso.redhat.de" should be able to set a cookie for + ".redhat.de". + + The only marginally sane way to handle this I can think of is to + reject on the basis of the length of the second-level domain name + (but when the top-level domain is unknown), with the assumption + that those of three or less characters could be reserved. For + example: + + .co.org -> works because the TLD is known + .co.uk -> doesn't work because "co" is only two chars long + .com.au -> doesn't work because "com" is only 3 chars long + .cnn.uk -> doesn't work because "cnn" is also only 3 chars long (ugh) + .cnn.de -> doesn't work for the same reason (ugh!!) + .abcd.de -> works because "abcd" is 4 chars long + .img.cnn.de -> works because it's not trying to set the 2nd level domain + .cnn.co.uk -> works for the same reason + + That should prevent misuse, while allowing reasonable usage. If + someone knows of a better way to handle this, please let me + know. */ + { + const char *p = cookie_domain; + int dccount = 1; /* number of domain components */ + int ldcl = 0; /* last domain component length */ + int nldcl = 0; /* next to last domain component length */ + int out; + if (*p == '.') + /* Ignore leading period in this calculation. */ + ++p; + DEBUGP (("cdm: 4\n")); + for (out = 0; !out; p++) + switch (*p) + { + case '\0': + out = 1; + break; + case '.': + if (ldcl == 0) + /* Empty domain component found -- the domain is invalid. */ + return false; + if (*(p + 1) == '\0') + { + /* Tolerate trailing '.' by not treating the domain as + one ending with an empty domain component. */ + out = 1; + break; + } + nldcl = ldcl; + ldcl = 0; + ++dccount; + break; + default: + ++ldcl; + } + + DEBUGP (("cdm: 5\n")); + + if (dccount < 2) + return false; + + DEBUGP (("cdm: 6\n")); + + if (dccount == 2) + { + size_t i; + int known_toplevel = false; + static const char *known_toplevel_domains[] = { + ".com", ".edu", ".net", ".org", ".gov", ".mil", ".int" + }; + for (i = 0; i < countof (known_toplevel_domains); i++) + if (match_tail (cookie_domain, known_toplevel_domains[i], true)) + { + known_toplevel = true; + break; + } + if (!known_toplevel && nldcl <= 3) + return false; + } + } + + DEBUGP (("cdm: 7\n")); + + /* Don't allow the host "foobar.com" to set a cookie for domain + "bar.com". */ + if (*cookie_domain != '.') + { + int dlen = strlen (cookie_domain); + int hlen = strlen (host); + /* cookie host: hostname.foobar.com */ + /* desired domain: bar.com */ + /* '.' must be here in host-> ^ */ + if (hlen > dlen && host[hlen - dlen - 1] != '.') + return false; + } + + DEBUGP (("cdm: 8\n")); + + return true; +} + +static int path_matches (const char *, const char *); + +/* Check whether PATH begins with COOKIE_PATH. */ + +static bool +check_path_match (const char *cookie_path, const char *path) +{ + return path_matches (path, cookie_path) != 0; +} + +/* Process the HTTP `Set-Cookie' header. This results in storing the + cookie or discarding a matching one, or ignoring it completely, all + depending on the contents. */ + +void +cookie_handle_set_cookie (struct cookie_jar *jar, + const char *host, int port, + const char *path, const char *set_cookie) +{ + struct cookie *cookie; + cookies_now = time (NULL); + char buf[1024], *tmp; + size_t pathlen = strlen(path); + + /* Wget's paths don't begin with '/' (blame rfc1808), but cookie + usage assumes /-prefixed paths. Until the rest of Wget is fixed, + simply prepend slash to PATH. */ + if (pathlen < sizeof (buf) - 1) + tmp = buf; + else + tmp = xmalloc (pathlen + 2); + + *tmp = '/'; + memcpy (tmp + 1, path, pathlen + 1); + path = tmp; + + cookie = parse_set_cookie (set_cookie, false); + if (!cookie) + goto out; + + /* Sanitize parts of cookie. */ + + if (!cookie->domain) + { + cookie->domain = xstrdup (host); + cookie->domain_exact = 1; + /* Set the port, but only if it's non-default. */ + if (port != 80 && port != 443) + cookie->port = port; + } + else + { + if (!check_domain_match (cookie->domain, host)) + { + logprintf (LOG_NOTQUIET, + _("Cookie coming from %s attempted to set domain to "), + quotearg_style (escape_quoting_style, host)); + logprintf (LOG_NOTQUIET, + _("%s\n"), + quotearg_style (escape_quoting_style, cookie->domain)); + cookie->discard_requested = true; + } + } + + if (!cookie->path) + { + /* The cookie doesn't set path: set it to the URL path, sans the + file part ("/dir/file" truncated to "/dir/"). */ + char *trailing_slash = strrchr (path, '/'); + if (trailing_slash) + cookie->path = strdupdelim (path, trailing_slash + 1); + else + /* no slash in the string -- can this even happen? */ + cookie->path = xstrdup (path); + } + else + { + /* The cookie sets its own path; verify that it is legal. */ + if (!check_path_match (cookie->path, path)) + { + DEBUGP (("Attempt to fake the path: %s, %s\n", + cookie->path, path)); + goto out; + } + } + + /* Now store the cookie, or discard an existing cookie, if + discarding was requested. */ + + if (cookie->discard_requested) + { + discard_matching_cookie (jar, cookie); + goto out; + } + + store_cookie (jar, cookie); + if (tmp != buf) + xfree (tmp); + return; + + out: + if (cookie) + delete_cookie (cookie); + if (tmp != buf) + xfree (tmp); +} + +/* Support for sending out cookies in HTTP requests, based on + previously stored cookies. Entry point is + `build_cookies_request'. */ + +/* Return a count of how many times CHR occurs in STRING. */ + +static int +count_char (const char *string, char chr) +{ + const char *p; + int count = 0; + for (p = string; *p; p++) + if (*p == chr) + ++count; + return count; +} + +/* Find the cookie chains whose domains match HOST and store them to + DEST. + + A cookie chain is the head of a list of cookies that belong to a + host/domain. Given HOST "img.search.xemacs.org", this function + will return the chains for "img.search.xemacs.org", + "search.xemacs.org", and "xemacs.org" -- those of them that exist + (if any), that is. + + DEST should be large enough to accept (in the worst case) as many + elements as there are domain components of HOST. */ + +static int +find_chains_of_host (struct cookie_jar *jar, const char *host, + struct cookie *dest[]) +{ + int dest_count = 0; + int passes, passcnt; + + /* Bail out quickly if there are no cookies in the jar. */ + if (!hash_table_count (jar->chains)) + return 0; + + if (numeric_address_p (host)) + /* If host is an IP address, only check for the exact match. */ + passes = 1; + else + /* Otherwise, check all the subdomains except the top-level (last) + one. As a domain with N components has N-1 dots, the number of + passes equals the number of dots. */ + passes = count_char (host, '.'); + + passcnt = 0; + + /* Find chains that match HOST, starting with exact match and + progressing to less specific domains. For instance, given HOST + fly.srk.fer.hr, first look for fly.srk.fer.hr's chain, then + srk.fer.hr's, then fer.hr's. */ + while (1) + { + struct cookie *chain = hash_table_get (jar->chains, host); + if (chain) + dest[dest_count++] = chain; + if (++passcnt >= passes) + break; + host = strchr (host, '.') + 1; + } + + return dest_count; +} + +/* If FULL_PATH begins with PREFIX, return the length of PREFIX, zero + otherwise. */ + +static int +path_matches (const char *full_path, const char *prefix) +{ + int len = strlen (prefix); + + if (0 != strncmp (full_path, prefix, len)) + /* FULL_PATH doesn't begin with PREFIX. */ + return 0; + + /* Length of PREFIX determines the quality of the match. */ + return len + 1; +} + +/* Return true if COOKIE matches the provided parameters of the URL + being downloaded: HOST, PORT, PATH, and SECFLAG. + + If PATH_GOODNESS is non-NULL, store the "path goodness" value + there. That value is a measure of how closely COOKIE matches PATH, + used for ordering cookies. */ + +static bool +cookie_matches_url (const struct cookie *cookie, + const char *host, int port, const char *path, + bool secflag, int *path_goodness) +{ + int pg; + + if (cookie_expired_p (cookie)) + /* Ignore stale cookies. Don't bother unchaining the cookie at + this point -- Wget is a relatively short-lived application, and + stale cookies will not be saved by `save_cookies'. On the + other hand, this function should be as efficient as + possible. */ + return false; + + if (cookie->secure && !secflag) + /* Don't transmit secure cookies over insecure connections. */ + return false; + if (cookie->port != PORT_ANY && cookie->port != port) + return false; + + /* If exact domain match is required, verify that cookie's domain is + equal to HOST. If not, assume success on the grounds of the + cookie's chain having been found by find_chains_of_host. */ + if (cookie->domain_exact + && 0 != strcasecmp (host, cookie->domain)) + return false; + + pg = path_matches (path, cookie->path); + if (pg == 0) + return false; + + if (path_goodness) + /* If the caller requested path_goodness, we return it. This is + an optimization, so that the caller doesn't need to call + path_matches() again. */ + *path_goodness = pg; + return true; +} + +/* A structure that points to a cookie, along with the additional + information about the cookie's "goodness". This allows us to sort + the cookies when returning them to the server, as required by the + spec. */ + +struct weighed_cookie { + struct cookie *cookie; + int domain_goodness; + int path_goodness; +}; + +/* Comparator used for uniquifying the list. */ + +static int +equality_comparator (const void *p1, const void *p2) +{ + struct weighed_cookie *wc1 = (struct weighed_cookie *)p1; + struct weighed_cookie *wc2 = (struct weighed_cookie *)p2; + + int namecmp = strcmp (wc1->cookie->attr, wc2->cookie->attr); + int valuecmp = strcmp (wc1->cookie->value, wc2->cookie->value); + + /* We only really care whether both name and value are equal. We + return them in this order only for consistency... */ + return namecmp ? namecmp : valuecmp; +} + +/* Eliminate duplicate cookies. "Duplicate cookies" are any two + cookies with the same attr name and value. Whenever a duplicate + pair is found, one of the cookies is removed. */ + +static int +eliminate_dups (struct weighed_cookie *outgoing, int count) +{ + struct weighed_cookie *h; /* hare */ + struct weighed_cookie *t; /* tortoise */ + struct weighed_cookie *end = outgoing + count; + + /* We deploy a simple uniquify algorithm: first sort the array + according to our sort criteria, then copy it to itself, comparing + each cookie to its neighbor and ignoring the duplicates. */ + + qsort (outgoing, count, sizeof (struct weighed_cookie), equality_comparator); + + /* "Hare" runs through all the entries in the array, followed by + "tortoise". If a duplicate is found, the hare skips it. + Non-duplicate entries are copied to the tortoise ptr. */ + + for (h = t = outgoing; h < end; h++) + { + if (h != end - 1) + { + struct cookie *c0 = h[0].cookie; + struct cookie *c1 = h[1].cookie; + if (!strcmp (c0->attr, c1->attr) && !strcmp (c0->value, c1->value)) + continue; /* ignore the duplicate */ + } + + /* If the hare has advanced past the tortoise (because of + previous dups), make sure the values get copied. Otherwise, + no copying is necessary. */ + if (h != t) + *t++ = *h; + else + t++; + } + return t - outgoing; +} + +/* Comparator used for sorting by quality. */ + +static int +goodness_comparator (const void *p1, const void *p2) +{ + struct weighed_cookie *wc1 = (struct weighed_cookie *)p1; + struct weighed_cookie *wc2 = (struct weighed_cookie *)p2; + + /* Subtractions take `wc2' as the first argument becauase we want a + sort in *decreasing* order of goodness. */ + int dgdiff = wc2->domain_goodness - wc1->domain_goodness; + int pgdiff = wc2->path_goodness - wc1->path_goodness; + + /* Sort by domain goodness; if these are the same, sort by path + goodness. (The sorting order isn't really specified; maybe it + should be the other way around.) */ + return dgdiff ? dgdiff : pgdiff; +} + +/* Generate a `Cookie' header for a request that goes to HOST:PORT and + requests PATH from the server. The resulting string is allocated + with `malloc', and the caller is responsible for freeing it. If no + cookies pertain to this request, i.e. no cookie header should be + generated, NULL is returned. */ + +char * +cookie_header (struct cookie_jar *jar, const char *host, + int port, const char *path, bool secflag) +{ + struct cookie *chains[32]; + int chain_count; + + struct cookie *cookie; + struct weighed_cookie *outgoing; + size_t count, i, ocnt; + char *result = NULL; + int result_size, pos; + char pathbuf[1024]; + + /* First, find the cookie chains whose domains match HOST. */ + + /* Allocate room for find_chains_of_host to write to. The number of + chains can at most equal the number of subdomains, hence + 1+<number of dots>. We ignore cookies with more than 32 labels. */ + chain_count = 1 + count_char (host, '.'); + if (chain_count > (int) countof (chains)) + return NULL; + chain_count = find_chains_of_host (jar, host, chains); + + /* No cookies for this host. */ + if (chain_count <= 0) + return NULL; + + /* Wget's paths don't begin with '/' (blame rfc1808), but cookie + usage assumes /-prefixed paths. Until the rest of Wget is fixed, + simply prepend slash to PATH. */ + { + char *tmp; + size_t pathlen = strlen(path); + + if (pathlen < sizeof (pathbuf) - 1) + tmp = pathbuf; + else + tmp = xmalloc (pathlen + 2); + + *tmp = '/'; + memcpy (tmp + 1, path, pathlen + 1); + path = tmp; + } + + cookies_now = time (NULL); + + /* Now extract from the chains those cookies that match our host + (for domain_exact cookies), port (for cookies with port other + than PORT_ANY), etc. See matching_cookie for details. */ + + /* Count the number of matching cookies. */ + count = 0; + for (i = 0; i < (unsigned) chain_count; i++) + for (cookie = chains[i]; cookie; cookie = cookie->next) + if (cookie_matches_url (cookie, host, port, path, secflag, NULL)) + ++count; + if (!count) + goto out; /* no cookies matched */ + + /* Allocate the array. */ + if (count > SIZE_MAX / sizeof (struct weighed_cookie)) + goto out; /* unable to process so many cookies */ + outgoing = xmalloc (count * sizeof (struct weighed_cookie)); + + /* Fill the array with all the matching cookies from the chains that + match HOST. */ + ocnt = 0; + for (i = 0; i < (unsigned) chain_count; i++) + for (cookie = chains[i]; cookie; cookie = cookie->next) + { + int pg; + if (!cookie_matches_url (cookie, host, port, path, secflag, &pg)) + continue; + outgoing[ocnt].cookie = cookie; + outgoing[ocnt].domain_goodness = strlen (cookie->domain); + outgoing[ocnt].path_goodness = pg; + ++ocnt; + } + assert (ocnt == count); + + /* Eliminate duplicate cookies; that is, those whose name and value + are the same. */ + count = eliminate_dups (outgoing, count); + + /* Sort the array so that best-matching domains come first, and + that, within one domain, best-matching paths come first. */ + qsort (outgoing, count, sizeof (struct weighed_cookie), goodness_comparator); + + /* Count the space the name=value pairs will take. */ + result_size = 0; + for (i = 0; i < count; i++) + { + struct cookie *c = outgoing[i].cookie; + /* name=value */ + result_size += strlen (c->attr) + 1 + strlen (c->value); + } + + /* Allocate output buffer: + name=value pairs -- result_size + "; " separators -- (count - 1) * 2 + \0 terminator -- 1 */ + result_size = result_size + (count - 1) * 2 + 1; + result = xmalloc (result_size); + pos = 0; + for (i = 0; i < count; i++) + { + struct cookie *c = outgoing[i].cookie; + int namlen = strlen (c->attr); + int vallen = strlen (c->value); + + memcpy (result + pos, c->attr, namlen); + pos += namlen; + result[pos++] = '='; + memcpy (result + pos, c->value, vallen); + pos += vallen; + if (i < count - 1) + { + result[pos++] = ';'; + result[pos++] = ' '; + } + } + result[pos++] = '\0'; + xfree (outgoing); + assert (pos == result_size); + +out: + if (path != pathbuf) + xfree (path); + +return result; +} + +/* Support for loading and saving cookies. The format used for + loading and saving should be the format of the `cookies.txt' file + used by Netscape and Mozilla, at least the Unix versions. + (Apparently IE can export cookies in that format as well.) The + format goes like this: + + DOMAIN DOMAIN-FLAG PATH SECURE-FLAG TIMESTAMP ATTR-NAME ATTR-VALUE + + DOMAIN -- cookie domain, optionally followed by :PORT + DOMAIN-FLAG -- whether all hosts in the domain match + PATH -- cookie path + SECURE-FLAG -- whether cookie requires secure connection + TIMESTAMP -- expiry timestamp, number of seconds since epoch + ATTR-NAME -- name of the cookie attribute + ATTR-VALUE -- value of the cookie attribute (empty if absent) + + The fields are separated by TABs. All fields are mandatory, except + for ATTR-VALUE. The `-FLAG' fields are boolean, their legal values + being "TRUE" and "FALSE'. Empty lines, lines consisting of + whitespace only, and comment lines (beginning with # optionally + preceded by whitespace) are ignored. + + Example line from cookies.txt (split in two lines for readability): + + .google.com TRUE / FALSE 2147368447 \ + PREF ID=34bb47565bbcd47b:LD=en:NR=20:TM=985172580:LM=985739012 + +*/ + +/* If the region [B, E) ends with :<digits>, parse the number, return + it, and store new boundary (location of the `:') to DOMAIN_E_PTR. + If port is not specified, return 0. */ + +static int +domain_port (const char *domain_b, const char *domain_e, + const char **domain_e_ptr) +{ + int port = 0; + const char *p; + const char *colon = memchr (domain_b, ':', domain_e - domain_b); + if (!colon) + return 0; + for (p = colon + 1; p < domain_e && c_isdigit (*p); p++) + port = 10 * port + (*p - '0'); + if (p < domain_e) + /* Garbage following port number. */ + return 0; + *domain_e_ptr = colon; + return port; +} + +#define GET_WORD(p, b, e) do { \ + b = p; \ + while (*p && *p != '\t') \ + ++p; \ + e = p; \ + if (b == e || !*p) \ + goto next; \ + ++p; \ +} while (0) + +/* Load cookies from FILE. */ + +void +cookie_jar_load (struct cookie_jar *jar, const char *file) +{ + char *line = NULL; + size_t bufsize = 0; + + FILE *fp = fopen (file, "r"); + if (!fp) + { + logprintf (LOG_NOTQUIET, _("Cannot open cookies file %s: %s\n"), + quote (file), strerror (errno)); + return; + } + + cookies_now = time (NULL); + + while (getline (&line, &bufsize, fp) > 0) + { + struct cookie *cookie; + char *p = line; + + double expiry; + int port; + + char *domain_b = NULL, *domain_e = NULL; + char *domflag_b = NULL, *domflag_e = NULL; + char *path_b = NULL, *path_e = NULL; + char *secure_b = NULL, *secure_e = NULL; + char *expires_b = NULL, *expires_e = NULL; + char *name_b = NULL, *name_e = NULL; + char *value_b = NULL, *value_e = NULL; + + /* Skip leading white-space. */ + while (*p && c_isspace (*p)) + ++p; + /* Ignore empty lines. */ + if (!*p || *p == '#') + continue; + + GET_WORD (p, domain_b, domain_e); + GET_WORD (p, domflag_b, domflag_e); + GET_WORD (p, path_b, path_e); + GET_WORD (p, secure_b, secure_e); + GET_WORD (p, expires_b, expires_e); + GET_WORD (p, name_b, name_e); + + /* Don't use GET_WORD for value because it ends with newline, + not TAB. */ + value_b = p; + value_e = p + strlen (p); + if (value_e > value_b && value_e[-1] == '\n') + --value_e; + if (value_e > value_b && value_e[-1] == '\r') + --value_e; + /* Empty values are legal (I think), so don't bother checking. */ + + cookie = cookie_new (); + + cookie->attr = strdupdelim (name_b, name_e); + cookie->value = strdupdelim (value_b, value_e); + cookie->path = strdupdelim (path_b, path_e); + cookie->secure = BOUNDED_EQUAL (secure_b, secure_e, "TRUE"); + + /* Curl source says, quoting Andre Garcia: "flag: A TRUE/FALSE + value indicating if all machines within a given domain can + access the variable. This value is set automatically by the + browser, depending on the value set for the domain." */ + cookie->domain_exact = !BOUNDED_EQUAL (domflag_b, domflag_e, "TRUE"); + + /* DOMAIN needs special treatment because we might need to + extract the port. */ + port = domain_port (domain_b, domain_e, (const char **)&domain_e); + if (port) + cookie->port = port; + + if (*domain_b == '.') + ++domain_b; /* remove leading dot internally */ + cookie->domain = strdupdelim (domain_b, domain_e); + + /* safe default in case EXPIRES field is garbled. */ + expiry = (double)cookies_now - 1; + + /* I don't like changing the line, but it's safe here. (line is + malloced.) */ + *expires_e = '\0'; + sscanf (expires_b, "%lf", &expiry); + + if (expiry == 0) + { + /* EXPIRY can be 0 for session cookies saved because the + user specified `--keep-session-cookies' in the past. + They remain session cookies, and will be saved only if + the user has specified `keep-session-cookies' again. */ + } + else + { + if (expiry < cookies_now) + goto abort_cookie; /* ignore stale cookie. */ + cookie->expiry_time = (time_t) expiry; + cookie->permanent = 1; + } + + store_cookie (jar, cookie); + + next: + continue; + + abort_cookie: + delete_cookie (cookie); + } + + xfree(line); + fclose (fp); +} + +/* Save cookies, in format described above, to FILE. */ + +void +cookie_jar_save (struct cookie_jar *jar, const char *file) +{ + FILE *fp; + hash_table_iterator iter; + + DEBUGP (("Saving cookies to %s.\n", file)); + + cookies_now = time (NULL); + + fp = fopen (file, "w"); + if (!fp) + { + logprintf (LOG_NOTQUIET, _("Cannot open cookies file %s: %s\n"), + quote (file), strerror (errno)); + return; + } + + fputs ("# HTTP Cookie File\n", fp); + fprintf (fp, "# Generated by Wget on %s.\n", datetime_str (cookies_now)); + fputs ("# Edit at your own risk.\n\n", fp); + + for (hash_table_iterate (jar->chains, &iter); + hash_table_iter_next (&iter); + ) + { + const char *domain = iter.key; + struct cookie *cookie = iter.value; + for (; cookie; cookie = cookie->next) + { + if (!cookie->permanent && !opt.keep_session_cookies) + continue; + if (cookie_expired_p (cookie)) + continue; + if (!cookie->domain_exact) + fputc ('.', fp); + fputs (domain, fp); + if (cookie->port != PORT_ANY) + fprintf (fp, ":%d", cookie->port); + fprintf (fp, "\t%s\t%s\t%s\t%.0f\t%s\t%s\n", + cookie->domain_exact ? "FALSE" : "TRUE", + cookie->path, cookie->secure ? "TRUE" : "FALSE", + (double)cookie->expiry_time, + cookie->attr, cookie->value); + if (ferror (fp)) + goto out; + } + } + out: + if (ferror (fp)) + logprintf (LOG_NOTQUIET, _("Error writing to %s: %s\n"), + quote (file), strerror (errno)); + if (fclose (fp) < 0) + logprintf (LOG_NOTQUIET, _("Error closing %s: %s\n"), + quote (file), strerror (errno)); + + DEBUGP (("Done saving cookies.\n")); +} + +/* Clean up cookie-related data. */ + +void +cookie_jar_delete (struct cookie_jar *jar) +{ + /* Iterate over chains (indexed by domain) and free them. */ + hash_table_iterator iter; + for (hash_table_iterate (jar->chains, &iter); hash_table_iter_next (&iter); ) + { + struct cookie *chain = iter.value; + xfree (iter.key); + /* Then all cookies in this chain. */ + while (chain) + { + struct cookie *next = chain->next; + delete_cookie (chain); + chain = next; + } + } + hash_table_destroy (jar->chains); + xfree (jar); + +#ifdef HAVE_LIBPSL + psl_free (psl); + psl = NULL; +#endif +} + +/* Test cases. Currently this is only tests parse_set_cookies. To + use, recompile Wget with -DTEST_COOKIES and call test_cookies() + from main. */ + +#ifdef TEST_COOKIES +void +test_cookies (void) +{ + /* Tests expected to succeed: */ + static struct { + const char *data; + const char *results[10]; + } tests_succ[] = { + { "arg=value", {"arg", "value", NULL} }, + { "arg1=value1;arg2=value2", {"arg1", "value1", "arg2", "value2", NULL} }, + { "arg1=value1; arg2=value2", {"arg1", "value1", "arg2", "value2", NULL} }, + { "arg1=value1; arg2=value2;", {"arg1", "value1", "arg2", "value2", NULL} }, + { "arg1=value1; arg2=value2; ", {"arg1", "value1", "arg2", "value2", NULL} }, + { "arg1=\"value1\"; arg2=\"\"", {"arg1", "value1", "arg2", "", NULL} }, + { "arg=", {"arg", "", NULL} }, + { "arg1=; arg2=", {"arg1", "", "arg2", "", NULL} }, + { "arg1 = ; arg2= ", {"arg1", "", "arg2", "", NULL} }, + }; + + /* Tests expected to fail: */ + static char *tests_fail[] = { + ";", + "arg=\"unterminated", + "=empty-name", + "arg1=;=another-empty-name", + }; + int i; + + for (i = 0; i < countof (tests_succ); i++) + { + int ind; + const char *data = tests_succ[i].data; + const char **expected = tests_succ[i].results; + struct cookie *c; + + c = parse_set_cookie (data, true); + if (!c) + { + printf ("NULL cookie returned for valid data: %s\n", data); + continue; + } + + /* Test whether extract_param handles these cases correctly. */ + { + param_token name, value; + const char *ptr = data; + int j = 0; + while (extract_param (&ptr, &name, &value, ';', NULL)) + { + char *n = strdupdelim (name.b, name.e); + char *v = strdupdelim (value.b, value.e); + if (!expected[j]) + { + printf ("Too many parameters for '%s'\n", data); + break; + } + if (0 != strcmp (expected[j], n)) + printf ("Invalid name %d for '%s' (expected '%s', got '%s')\n", + j / 2 + 1, data, expected[j], n); + if (0 != strcmp (expected[j + 1], v)) + printf ("Invalid value %d for '%s' (expected '%s', got '%s')\n", + j / 2 + 1, data, expected[j + 1], v); + j += 2; + xfree (n); + xfree (v); + } + if (expected[j]) + printf ("Too few parameters for '%s'\n", data); + } + } + + for (i = 0; i < countof (tests_fail); i++) + { + struct cookie *c; + char *data = tests_fail[i]; + c = parse_set_cookie (data, true); + if (c) + printf ("Failed to report error on invalid data: %s\n", data); + } +} +#endif /* TEST_COOKIES */ |