net/cookies/cookie_util.cc

// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#include "net/cookies/cookie_util.h"

#include <cstdio>
#include <cstdlib>

#include "base/logging.h"
#include "base/strings/string_tokenizer.h"
#include "base/strings/string_util.h"
#include "build/build_config.h"
#include "net/base/net_util.h"
#include "net/base/registry_controlled_domains/registry_controlled_domain.h"
#include "url/gurl.h"

namespace net {
namespace cookie_util {

bool DomainIsHostOnly(const std::string& domain_string) {
  return (domain_string.empty() || domain_string[0] != '.');
}

std::string GetEffectiveDomain(const std::string& scheme,
                               const std::string& host) {
  if (scheme == "http" || scheme == "https") {
    return registry_controlled_domains::GetDomainAndRegistry(
        host,
        registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES);
  }

  if (!DomainIsHostOnly(host))
    return host.substr(1);
  return host;
}

bool GetCookieDomainWithString(const GURL& url,
                               const std::string& domain_string,
                               std::string* result) {
  const std::string url_host(url.host());

  // If no domain was specified in the domain string, default to a host cookie.
  // We match IE/Firefox in allowing a domain=IPADDR if it matches the url
  // ip address hostname exactly.  It should be treated as a host cookie.
  if (domain_string.empty() ||
      (url.HostIsIPAddress() && url_host == domain_string)) {
    *result = url_host;
    DCHECK(DomainIsHostOnly(*result));
    return true;
  }

  // Get the normalized domain specified in cookie line.
  url::CanonHostInfo ignored;
  std::string cookie_domain(CanonicalizeHost(domain_string, &ignored));
  if (cookie_domain.empty())
    return false;
  if (cookie_domain[0] != '.')
    cookie_domain = "." + cookie_domain;

  // Ensure |url| and |cookie_domain| have the same domain+registry.
  const std::string url_scheme(url.scheme());
  const std::string url_domain_and_registry(
      GetEffectiveDomain(url_scheme, url_host));
  if (url_domain_and_registry.empty())
    return false;  // IP addresses/intranet hosts can't set domain cookies.
  const std::string cookie_domain_and_registry(
      GetEffectiveDomain(url_scheme, cookie_domain));
  if (url_domain_and_registry != cookie_domain_and_registry)
    return false;  // Can't set a cookie on a different domain + registry.

  // Ensure |url_host| is |cookie_domain| or one of its subdomains.  Given that
  // we know the domain+registry are the same from the above checks, this is
  // basically a simple string suffix check.
  const bool is_suffix = (url_host.length() < cookie_domain.length()) ?
      (cookie_domain != ("." + url_host)) :
      (url_host.compare(url_host.length() - cookie_domain.length(),
                        cookie_domain.length(), cookie_domain) != 0);
  if (is_suffix)
    return false;

  *result = cookie_domain;
  return true;
}

// Parse a cookie expiration time.  We try to be lenient, but we need to
// assume some order to distinguish the fields.  The basic rules:
//  - The month name must be present and prefix the first 3 letters of the
//    full month name (jan for January, jun for June).
//  - If the year is <= 2 digits, it must occur after the day of month.
//  - The time must be of the format hh:mm:ss.
// An average cookie expiration will look something like this:
//   Sat, 15-Apr-17 21:01:22 GMT
base::Time ParseCookieTime(const std::string& time_string) {
  static const char* kMonths[] = { "jan", "feb", "mar", "apr", "may", "jun",
                                   "jul", "aug", "sep", "oct", "nov", "dec" };
  static const int kMonthsLen = arraysize(kMonths);
  // We want to be pretty liberal, and support most non-ascii and non-digit
  // characters as a delimiter.  We can't treat : as a delimiter, because it
  // is the delimiter for hh:mm:ss, and we want to keep this field together.
  // We make sure to include - and +, since they could prefix numbers.
  // If the cookie attribute came in in quotes (ex expires="XXX"), the quotes
  // will be preserved, and we will get them here.  So we make sure to include
  // quote characters, and also \ for anything that was internally escaped.
  static const char* kDelimiters = "\t !\"#$%&'()*+,-./;<=>?@[\\]^_`{|}~";

  base::Time::Exploded exploded = {0};

  base::StringTokenizer tokenizer(time_string, kDelimiters);

  bool found_day_of_month = false;
  bool found_month = false;
  bool found_time = false;
  bool found_year = false;

  while (tokenizer.GetNext()) {
    const std::string token = tokenizer.token();
    DCHECK(!token.empty());
    bool numerical = IsAsciiDigit(token[0]);

    // String field
    if (!numerical) {
      if (!found_month) {
        for (int i = 0; i < kMonthsLen; ++i) {
          // Match prefix, so we could match January, etc
          if (base::strncasecmp(token.c_str(), kMonths[i], 3) == 0) {
            exploded.month = i + 1;
            found_month = true;
            break;
          }
        }
      } else {
        // If we've gotten here, it means we've already found and parsed our
        // month, and we have another string, which we would expect to be the
        // the time zone name.  According to the RFC and my experiments with
        // how sites format their expirations, we don't have much of a reason
        // to support timezones.  We don't want to ever barf on user input,
        // but this DCHECK should pass for well-formed data.
        // DCHECK(token == "GMT");
      }
    // Numeric field w/ a colon
    } else if (token.find(':') != std::string::npos) {
      if (!found_time &&
#ifdef COMPILER_MSVC
          sscanf_s(
#else
          sscanf(
#endif
                 token.c_str(), "%2u:%2u:%2u", &exploded.hour,
                 &exploded.minute, &exploded.second) == 3) {
        found_time = true;
      } else {
        // We should only ever encounter one time-like thing.  If we're here,
        // it means we've found a second, which shouldn't happen.  We keep
        // the first.  This check should be ok for well-formed input:
        // NOTREACHED();
      }
    // Numeric field
    } else {
      // Overflow with atoi() is unspecified, so we enforce a max length.
      if (!found_day_of_month && token.length() <= 2) {
        exploded.day_of_month = atoi(token.c_str());
        found_day_of_month = true;
      } else if (!found_year && token.length() <= 5) {
        exploded.year = atoi(token.c_str());
        found_year = true;
      } else {
        // If we're here, it means we've either found an extra numeric field,
        // or a numeric field which was too long.  For well-formed input, the
        // following check would be reasonable:
        // NOTREACHED();
      }
    }
  }

  if (!found_day_of_month || !found_month || !found_time || !found_year) {
    // We didn't find all of the fields we need.  For well-formed input, the
    // following check would be reasonable:
    // NOTREACHED() << "Cookie parse expiration failed: " << time_string;
    return base::Time();
  }

  // Normalize the year to expand abbreviated years to the full year.
  if (exploded.year >= 69 && exploded.year <= 99)
    exploded.year += 1900;
  if (exploded.year >= 0 && exploded.year <= 68)
    exploded.year += 2000;

  // If our values are within their correct ranges, we got our time.
  if (exploded.day_of_month >= 1 && exploded.day_of_month <= 31 &&
      exploded.month >= 1 && exploded.month <= 12 &&
      exploded.year >= 1601 && exploded.year <= 30827 &&
      exploded.hour <= 23 && exploded.minute <= 59 && exploded.second <= 59) {
    return base::Time::FromUTCExploded(exploded);
  }

  // One of our values was out of expected range.  For well-formed input,
  // the following check would be reasonable:
  // NOTREACHED() << "Cookie exploded expiration failed: " << time_string;

  return base::Time();
}

GURL CookieOriginToURL(const std::string& domain, bool is_https) {
  if (domain.empty())
    return GURL();

  const std::string scheme = is_https ? "https" : "http";
  const std::string host = domain[0] == '.' ? domain.substr(1) : domain;
  return GURL(scheme + "://" + host);
}

void ParseRequestCookieLine(const std::string& header_value,
                            ParsedRequestCookies* parsed_cookies) {
  std::string::const_iterator i = header_value.begin();
  while (i != header_value.end()) {
    // Here we are at the beginning of a cookie.

    // Eat whitespace.
    while (i != header_value.end() && *i == ' ') ++i;
    if (i == header_value.end()) return;

    // Find cookie name.
    std::string::const_iterator cookie_name_beginning = i;
    while (i != header_value.end() && *i != '=') ++i;
    base::StringPiece cookie_name(cookie_name_beginning, i);

    // Find cookie value.
    base::StringPiece cookie_value;
    // Cookies may have no value, in this case '=' may or may not be there.
    if (i != header_value.end() && i + 1 != header_value.end()) {
      ++i;  // Skip '='.
      std::string::const_iterator cookie_value_beginning = i;
      if (*i == '"') {
        ++i;  // Skip '"'.
        while (i != header_value.end() && *i != '"') ++i;
        if (i == header_value.end()) return;
        ++i;  // Skip '"'.
        cookie_value = base::StringPiece(cookie_value_beginning, i);
        // i points to character after '"', potentially a ';'.
      } else {
        while (i != header_value.end() && *i != ';') ++i;
        cookie_value = base::StringPiece(cookie_value_beginning, i);
        // i points to ';' or end of string.
      }
    }
    parsed_cookies->push_back(std::make_pair(cookie_name, cookie_value));
    // Eat ';'.
    if (i != header_value.end()) ++i;
  }
}

std::string SerializeRequestCookieLine(
    const ParsedRequestCookies& parsed_cookies) {
  std::string buffer;
  for (ParsedRequestCookies::const_iterator i = parsed_cookies.begin();
       i != parsed_cookies.end(); ++i) {
    if (!buffer.empty())
      buffer.append("; ");
    buffer.append(i->first.begin(), i->first.end());
    buffer.push_back('=');
    buffer.append(i->second.begin(), i->second.end());
  }
  return buffer;
}

}  // namespace cookie_utils
}  // namespace net