Home | History | Annotate | Download | only in cookies 
      1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
      2 // Use of this source code is governed by a BSD-style license that can be
      3 // found in the LICENSE file.
      4 
      5 #include "net/cookies/cookie_util.h"
      6 
      7 #include <cstdio>
      8 #include <cstdlib>
      9 
     10 #include "base/logging.h"
     11 #include "base/strings/string_tokenizer.h"
     12 #include "base/strings/string_util.h"
     13 #include "build/build_config.h"
     14 #include "net/base/net_util.h"
     15 #include "net/base/registry_controlled_domains/registry_controlled_domain.h"
     16 #include "url/gurl.h"
     17 
     18 namespace net {
     19 namespace cookie_util {
     20 
     21 bool DomainIsHostOnly(const std::string& domain_string) {
     22   return (domain_string.empty() || domain_string[0] != '.');
     23 }
     24 
     25 std::string GetEffectiveDomain(const std::string& scheme,
     26                                const std::string& host) {
     27   if (scheme == "http" || scheme == "https") {
     28     return registry_controlled_domains::GetDomainAndRegistry(
     29         host, net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
     30   }
     31 
     32   if (!DomainIsHostOnly(host))
     33     return host.substr(1);
     34   return host;
     35 }
     36 
     37 bool GetCookieDomainWithString(const GURL& url,
     38                                const std::string& domain_string,
     39                                std::string* result) {
     40   const std::string url_host(url.host());
     41 
     42   // If no domain was specified in the domain string, default to a host cookie.
     43   // We match IE/Firefox in allowing a domain=IPADDR if it matches the url
     44   // ip address hostname exactly.  It should be treated as a host cookie.
     45   if (domain_string.empty() ||
     46       (url.HostIsIPAddress() && url_host == domain_string)) {
     47     *result = url_host;
     48     DCHECK(DomainIsHostOnly(*result));
     49     return true;
     50   }
     51 
     52   // Get the normalized domain specified in cookie line.
     53   url_canon::CanonHostInfo ignored;
     54   std::string cookie_domain(CanonicalizeHost(domain_string, &ignored));
     55   if (cookie_domain.empty())
     56     return false;
     57   if (cookie_domain[0] != '.')
     58     cookie_domain = "." + cookie_domain;
     59 
     60   // Ensure |url| and |cookie_domain| have the same domain+registry.
     61   const std::string url_scheme(url.scheme());
     62   const std::string url_domain_and_registry(
     63       GetEffectiveDomain(url_scheme, url_host));
     64   if (url_domain_and_registry.empty())
     65     return false;  // IP addresses/intranet hosts can't set domain cookies.
     66   const std::string cookie_domain_and_registry(
     67       GetEffectiveDomain(url_scheme, cookie_domain));
     68   if (url_domain_and_registry != cookie_domain_and_registry)
     69     return false;  // Can't set a cookie on a different domain + registry.
     70 
     71   // Ensure |url_host| is |cookie_domain| or one of its subdomains.  Given that
     72   // we know the domain+registry are the same from the above checks, this is
     73   // basically a simple string suffix check.
     74   const bool is_suffix = (url_host.length() < cookie_domain.length()) ?
     75       (cookie_domain != ("." + url_host)) :
     76       (url_host.compare(url_host.length() - cookie_domain.length(),
     77                         cookie_domain.length(), cookie_domain) != 0);
     78   if (is_suffix)
     79     return false;
     80 
     81   *result = cookie_domain;
     82   return true;
     83 }
     84 
     85 // Parse a cookie expiration time.  We try to be lenient, but we need to
     86 // assume some order to distinguish the fields.  The basic rules:
     87 //  - The month name must be present and prefix the first 3 letters of the
     88 //    full month name (jan for January, jun for June).
     89 //  - If the year is <= 2 digits, it must occur after the day of month.
     90 //  - The time must be of the format hh:mm:ss.
     91 // An average cookie expiration will look something like this:
     92 //   Sat, 15-Apr-17 21:01:22 GMT
     93 base::Time ParseCookieTime(const std::string& time_string) {
     94   static const char* kMonths[] = { "jan", "feb", "mar", "apr", "may", "jun",
     95                                    "jul", "aug", "sep", "oct", "nov", "dec" };
     96   static const int kMonthsLen = arraysize(kMonths);
     97   // We want to be pretty liberal, and support most non-ascii and non-digit
     98   // characters as a delimiter.  We can't treat : as a delimiter, because it
     99   // is the delimiter for hh:mm:ss, and we want to keep this field together.
    100   // We make sure to include - and +, since they could prefix numbers.
    101   // If the cookie attribute came in in quotes (ex expires="XXX"), the quotes
    102   // will be preserved, and we will get them here.  So we make sure to include
    103   // quote characters, and also \ for anything that was internally escaped.
    104   static const char* kDelimiters = "\t !\"#$%&'()*+,-./;<=>?@[\\]^_`{|}~";
    105 
    106   base::Time::Exploded exploded = {0};
    107 
    108   base::StringTokenizer tokenizer(time_string, kDelimiters);
    109 
    110   bool found_day_of_month = false;
    111   bool found_month = false;
    112   bool found_time = false;
    113   bool found_year = false;
    114 
    115   while (tokenizer.GetNext()) {
    116     const std::string token = tokenizer.token();
    117     DCHECK(!token.empty());
    118     bool numerical = IsAsciiDigit(token[0]);
    119 
    120     // String field
    121     if (!numerical) {
    122       if (!found_month) {
    123         for (int i = 0; i < kMonthsLen; ++i) {
    124           // Match prefix, so we could match January, etc
    125           if (base::strncasecmp(token.c_str(), kMonths[i], 3) == 0) {
    126             exploded.month = i + 1;
    127             found_month = true;
    128             break;
    129           }
    130         }
    131       } else {
    132         // If we've gotten here, it means we've already found and parsed our
    133         // month, and we have another string, which we would expect to be the
    134         // the time zone name.  According to the RFC and my experiments with
    135         // how sites format their expirations, we don't have much of a reason
    136         // to support timezones.  We don't want to ever barf on user input,
    137         // but this DCHECK should pass for well-formed data.
    138         // DCHECK(token == "GMT");
    139       }
    140     // Numeric field w/ a colon
    141     } else if (token.find(':') != std::string::npos) {
    142       if (!found_time &&
    143 #ifdef COMPILER_MSVC
    144           sscanf_s(
    145 #else
    146           sscanf(
    147 #endif
    148                  token.c_str(), "%2u:%2u:%2u", &exploded.hour,
    149                  &exploded.minute, &exploded.second) == 3) {
    150         found_time = true;
    151       } else {
    152         // We should only ever encounter one time-like thing.  If we're here,
    153         // it means we've found a second, which shouldn't happen.  We keep
    154         // the first.  This check should be ok for well-formed input:
    155         // NOTREACHED();
    156       }
    157     // Numeric field
    158     } else {
    159       // Overflow with atoi() is unspecified, so we enforce a max length.
    160       if (!found_day_of_month && token.length() <= 2) {
    161         exploded.day_of_month = atoi(token.c_str());
    162         found_day_of_month = true;
    163       } else if (!found_year && token.length() <= 5) {
    164         exploded.year = atoi(token.c_str());
    165         found_year = true;
    166       } else {
    167         // If we're here, it means we've either found an extra numeric field,
    168         // or a numeric field which was too long.  For well-formed input, the
    169         // following check would be reasonable:
    170         // NOTREACHED();
    171       }
    172     }
    173   }
    174 
    175   if (!found_day_of_month || !found_month || !found_time || !found_year) {
    176     // We didn't find all of the fields we need.  For well-formed input, the
    177     // following check would be reasonable:
    178     // NOTREACHED() << "Cookie parse expiration failed: " << time_string;
    179     return base::Time();
    180   }
    181 
    182   // Normalize the year to expand abbreviated years to the full year.
    183   if (exploded.year >= 69 && exploded.year <= 99)
    184     exploded.year += 1900;
    185   if (exploded.year >= 0 && exploded.year <= 68)
    186     exploded.year += 2000;
    187 
    188   // If our values are within their correct ranges, we got our time.
    189   if (exploded.day_of_month >= 1 && exploded.day_of_month <= 31 &&
    190       exploded.month >= 1 && exploded.month <= 12 &&
    191       exploded.year >= 1601 && exploded.year <= 30827 &&
    192       exploded.hour <= 23 && exploded.minute <= 59 && exploded.second <= 59) {
    193     return base::Time::FromUTCExploded(exploded);
    194   }
    195 
    196   // One of our values was out of expected range.  For well-formed input,
    197   // the following check would be reasonable:
    198   // NOTREACHED() << "Cookie exploded expiration failed: " << time_string;
    199 
    200   return base::Time();
    201 }
    202 
    203 GURL CookieOriginToURL(const std::string& domain, bool is_https) {
    204   if (domain.empty())
    205     return GURL();
    206 
    207   const std::string scheme = is_https ? "https" : "http";
    208   const std::string host = domain[0] == '.' ? domain.substr(1) : domain;
    209   return GURL(scheme + "://" + host);
    210 }
    211 
    212 }  // namespace cookie_utils
    213 }  // namespace net
    214 
    215