extensions/common/url_pattern.cc

   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 "extensions/common/url_pattern.h"
   6
   7 #include <ostream>
   8
   9 #include "base/strings/pattern.h"
  10 #include "base/strings/string_number_conversions.h"
  11 #include "base/strings/string_piece.h"
  12 #include "base/strings/string_split.h"
  13 #include "base/strings/string_util.h"
  14 #include "base/strings/stringprintf.h"
  15 #include "content/public/common/url_constants.h"
  16 #include "extensions/common/constants.h"
  17 #include "net/base/registry_controlled_domains/registry_controlled_domain.h"
  18 #include "url/gurl.h"
  19 #include "url/url_util.h"
  20
  21 const char URLPattern::kAllUrlsPattern[] = "<all_urls>";
  22
  23 namespace {
  24
  25 // TODO(aa): What about more obscure schemes like data: and javascript: ?
  26 // Note: keep this array in sync with kValidSchemeMasks.
  27 const char* kValidSchemes[] = {
  28     url::kHttpScheme,
  29     url::kHttpsScheme,
  30     url::kFileScheme,
  31     url::kFtpScheme,
  32     content::kChromeUIScheme,
  33     extensions::kExtensionScheme,
  34     url::kFileSystemScheme,
  35 };
  36
  37 const int kValidSchemeMasks[] = {
  38   URLPattern::SCHEME_HTTP,
  39   URLPattern::SCHEME_HTTPS,
  40   URLPattern::SCHEME_FILE,
  41   URLPattern::SCHEME_FTP,
  42   URLPattern::SCHEME_CHROMEUI,
  43   URLPattern::SCHEME_EXTENSION,
  44   URLPattern::SCHEME_FILESYSTEM,
  45 };
  46
  47 static_assert(arraysize(kValidSchemes) == arraysize(kValidSchemeMasks),
  48               "must keep these arrays in sync");
  49
  50 const char kParseSuccess[] = "Success.";
  51 const char kParseErrorMissingSchemeSeparator[] = "Missing scheme separator.";
  52 const char kParseErrorInvalidScheme[] = "Invalid scheme.";
  53 const char kParseErrorWrongSchemeType[] = "Wrong scheme type.";
  54 const char kParseErrorEmptyHost[] = "Host can not be empty.";
  55 const char kParseErrorInvalidHostWildcard[] = "Invalid host wildcard.";
  56 const char kParseErrorEmptyPath[] = "Empty path.";
  57 const char kParseErrorInvalidPort[] = "Invalid port.";
  58 const char kParseErrorInvalidHost[] = "Invalid host.";
  59
  60 // Message explaining each URLPattern::ParseResult.
  61 const char* const kParseResultMessages[] = {
  62   kParseSuccess,
  63   kParseErrorMissingSchemeSeparator,
  64   kParseErrorInvalidScheme,
  65   kParseErrorWrongSchemeType,
  66   kParseErrorEmptyHost,
  67   kParseErrorInvalidHostWildcard,
  68   kParseErrorEmptyPath,
  69   kParseErrorInvalidPort,
  70   kParseErrorInvalidHost,
  71 };
  72
  73 static_assert(URLPattern::NUM_PARSE_RESULTS == arraysize(kParseResultMessages),
  74               "must add message for each parse result");
  75
  76 const char kPathSeparator[] = "/";
  77
  78 bool IsStandardScheme(const std::string& scheme) {
  79   // "*" gets the same treatment as a standard scheme.
  80   if (scheme == "*")
  81     return true;
  82
  83   return url::IsStandard(scheme.c_str(),
  84                          url::Component(0, static_cast<int>(scheme.length())));
  85 }
  86
  87 bool IsValidPortForScheme(const std::string& scheme, const std::string& port) {
  88   if (port == "*")
  89     return true;
  90
  91   // Only accept non-wildcard ports if the scheme uses ports.
  92   if (url::DefaultPortForScheme(scheme.c_str(), scheme.length()) ==
  93       url::PORT_UNSPECIFIED) {
  94     return false;
  95   }
  96
  97   int parsed_port = url::PORT_UNSPECIFIED;
  98   if (!base::StringToInt(port, &parsed_port))
  99     return false;
 100   return (parsed_port >= 0) && (parsed_port < 65536);
 101 }
 102
 103 // Returns |path| with the trailing wildcard stripped if one existed.
 104 //
 105 // The functions that rely on this (OverlapsWith and Contains) are only
 106 // called for the patterns inside URLPatternSet. In those cases, we know that
 107 // the path will have only a single wildcard at the end. This makes figuring
 108 // out overlap much easier. It seems like there is probably a computer-sciency
 109 // way to solve the general case, but we don't need that yet.
 110 std::string StripTrailingWildcard(const std::string& path) {
 111   size_t wildcard_index = path.find('*');
 112   size_t path_last = path.size() - 1;
 113   return wildcard_index == path_last ? path.substr(0, path_last) : path;
 114 }
 115
 116 }  // namespace
 117
 118 // static
 119 bool URLPattern::IsValidSchemeForExtensions(const std::string& scheme) {
 120   for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
 121     if (scheme == kValidSchemes[i])
 122       return true;
 123   }
 124   return false;
 125 }
 126
 127 URLPattern::URLPattern()
 128     : valid_schemes_(SCHEME_NONE),
 129       match_all_urls_(false),
 130       match_subdomains_(false),
 131       port_("*") {}
 132
 133 URLPattern::URLPattern(int valid_schemes)
 134     : valid_schemes_(valid_schemes),
 135       match_all_urls_(false),
 136       match_subdomains_(false),
 137       port_("*") {}
 138
 139 URLPattern::URLPattern(int valid_schemes, const std::string& pattern)
 140     // Strict error checking is used, because this constructor is only
 141     // appropriate when we know |pattern| is valid.
 142     : valid_schemes_(valid_schemes),
 143       match_all_urls_(false),
 144       match_subdomains_(false),
 145       port_("*") {
 146   ParseResult result = Parse(pattern);
 147   if (PARSE_SUCCESS != result)
 148     NOTREACHED() << "URLPattern invalid: " << pattern << " result " << result;
 149 }
 150
 151 URLPattern::~URLPattern() {
 152 }
 153
 154 bool URLPattern::operator<(const URLPattern& other) const {
 155   return GetAsString() < other.GetAsString();
 156 }
 157
 158 bool URLPattern::operator>(const URLPattern& other) const {
 159   return GetAsString() > other.GetAsString();
 160 }
 161
 162 bool URLPattern::operator==(const URLPattern& other) const {
 163   return GetAsString() == other.GetAsString();
 164 }
 165
 166 std::ostream& operator<<(std::ostream& out, const URLPattern& url_pattern) {
 167   return out << '"' << url_pattern.GetAsString() << '"';
 168 }
 169
 170 URLPattern::ParseResult URLPattern::Parse(const std::string& pattern) {
 171   spec_.clear();
 172   SetMatchAllURLs(false);
 173   SetMatchSubdomains(false);
 174   SetPort("*");
 175
 176   // Special case pattern to match every valid URL.
 177   if (pattern == kAllUrlsPattern) {
 178     SetMatchAllURLs(true);
 179     return PARSE_SUCCESS;
 180   }
 181
 182   // Parse out the scheme.
 183   size_t scheme_end_pos = pattern.find(url::kStandardSchemeSeparator);
 184   bool has_standard_scheme_separator = true;
 185
 186   // Some urls also use ':' alone as the scheme separator.
 187   if (scheme_end_pos == std::string::npos) {
 188     scheme_end_pos = pattern.find(':');
 189     has_standard_scheme_separator = false;
 190   }
 191
 192   if (scheme_end_pos == std::string::npos)
 193     return PARSE_ERROR_MISSING_SCHEME_SEPARATOR;
 194
 195   if (!SetScheme(pattern.substr(0, scheme_end_pos)))
 196     return PARSE_ERROR_INVALID_SCHEME;
 197
 198   bool standard_scheme = IsStandardScheme(scheme_);
 199   if (standard_scheme != has_standard_scheme_separator)
 200     return PARSE_ERROR_WRONG_SCHEME_SEPARATOR;
 201
 202   // Advance past the scheme separator.
 203   scheme_end_pos +=
 204       (standard_scheme ? strlen(url::kStandardSchemeSeparator) : 1);
 205   if (scheme_end_pos >= pattern.size())
 206     return PARSE_ERROR_EMPTY_HOST;
 207
 208   // Parse out the host and path.
 209   size_t host_start_pos = scheme_end_pos;
 210   size_t path_start_pos = 0;
 211
 212   if (!standard_scheme) {
 213     path_start_pos = host_start_pos;
 214   } else if (scheme_ == url::kFileScheme) {
 215     size_t host_end_pos = pattern.find(kPathSeparator, host_start_pos);
 216     if (host_end_pos == std::string::npos) {
 217       // Allow hostname omission.
 218       // e.g. file://* is interpreted as file:///*,
 219       // file://foo* is interpreted as file:///foo*.
 220       path_start_pos = host_start_pos - 1;
 221     } else {
 222       // Ignore hostname if scheme is file://.
 223       // e.g. file://localhost/foo is equal to file:///foo.
 224       path_start_pos = host_end_pos;
 225     }
 226   } else {
 227     size_t host_end_pos = pattern.find(kPathSeparator, host_start_pos);
 228
 229     // Host is required.
 230     if (host_start_pos == host_end_pos)
 231       return PARSE_ERROR_EMPTY_HOST;
 232
 233     if (host_end_pos == std::string::npos)
 234       return PARSE_ERROR_EMPTY_PATH;
 235
 236     host_ = pattern.substr(host_start_pos, host_end_pos - host_start_pos);
 237
 238     // The first component can optionally be '*' to match all subdomains.
 239     std::vector<std::string> host_components = base::SplitString(
 240         host_, ".", base::TRIM_WHITESPACE, base::SPLIT_WANT_ALL);
 241
 242     // Could be empty if the host only consists of whitespace characters.
 243     if (host_components.empty() ||
 244         (host_components.size() == 1 && host_components[0].empty()))
 245       return PARSE_ERROR_EMPTY_HOST;
 246
 247     if (host_components[0] == "*") {
 248       match_subdomains_ = true;
 249       host_components.erase(host_components.begin(),
 250                             host_components.begin() + 1);
 251     }
 252     host_ = base::JoinString(host_components, ".");
 253
 254     path_start_pos = host_end_pos;
 255   }
 256
 257   SetPath(pattern.substr(path_start_pos));
 258
 259   size_t port_pos = host_.find(':');
 260   if (port_pos != std::string::npos) {
 261     if (!SetPort(host_.substr(port_pos + 1)))
 262       return PARSE_ERROR_INVALID_PORT;
 263     host_ = host_.substr(0, port_pos);
 264   }
 265
 266   // No other '*' can occur in the host, though. This isn't necessary, but is
 267   // done as a convenience to developers who might otherwise be confused and
 268   // think '*' works as a glob in the host.
 269   if (host_.find('*') != std::string::npos)
 270     return PARSE_ERROR_INVALID_HOST_WILDCARD;
 271
 272   // Null characters are not allowed in hosts.
 273   if (host_.find('\0') != std::string::npos)
 274     return PARSE_ERROR_INVALID_HOST;
 275
 276   return PARSE_SUCCESS;
 277 }
 278
 279 void URLPattern::SetValidSchemes(int valid_schemes) {
 280   spec_.clear();
 281   valid_schemes_ = valid_schemes;
 282 }
 283
 284 void URLPattern::SetHost(const std::string& host) {
 285   spec_.clear();
 286   host_ = host;
 287 }
 288
 289 void URLPattern::SetMatchAllURLs(bool val) {
 290   spec_.clear();
 291   match_all_urls_ = val;
 292
 293   if (val) {
 294     match_subdomains_ = true;
 295     scheme_ = "*";
 296     host_.clear();
 297     SetPath("/*");
 298   }
 299 }
 300
 301 void URLPattern::SetMatchSubdomains(bool val) {
 302   spec_.clear();
 303   match_subdomains_ = val;
 304 }
 305
 306 bool URLPattern::SetScheme(const std::string& scheme) {
 307   spec_.clear();
 308   scheme_ = scheme;
 309   if (scheme_ == "*") {
 310     valid_schemes_ &= (SCHEME_HTTP | SCHEME_HTTPS);
 311   } else if (!IsValidScheme(scheme_)) {
 312     return false;
 313   }
 314   return true;
 315 }
 316
 317 bool URLPattern::IsValidScheme(const std::string& scheme) const {
 318   if (valid_schemes_ == SCHEME_ALL)
 319     return true;
 320
 321   for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
 322     if (scheme == kValidSchemes[i] && (valid_schemes_ & kValidSchemeMasks[i]))
 323       return true;
 324   }
 325
 326   return false;
 327 }
 328
 329 void URLPattern::SetPath(const std::string& path) {
 330   spec_.clear();
 331   path_ = path;
 332   path_escaped_ = path_;
 333   base::ReplaceSubstringsAfterOffset(&path_escaped_, 0, "\\", "\\\\");
 334   base::ReplaceSubstringsAfterOffset(&path_escaped_, 0, "?", "\\?");
 335 }
 336
 337 bool URLPattern::SetPort(const std::string& port) {
 338   spec_.clear();
 339   if (IsValidPortForScheme(scheme_, port)) {
 340     port_ = port;
 341     return true;
 342   }
 343   return false;
 344 }
 345
 346 bool URLPattern::MatchesURL(const GURL& test) const {
 347   const GURL* test_url = &test;
 348   bool has_inner_url = test.inner_url() != NULL;
 349
 350   if (has_inner_url) {
 351     if (!test.SchemeIsFileSystem())
 352       return false;  // The only nested URLs we handle are filesystem URLs.
 353     test_url = test.inner_url();
 354   }
 355
 356   if (!MatchesScheme(test_url->scheme()))
 357     return false;
 358
 359   if (match_all_urls_)
 360     return true;
 361
 362   std::string path_for_request = test.PathForRequest();
 363   if (has_inner_url)
 364     path_for_request = test_url->path() + path_for_request;
 365
 366   return MatchesSecurityOriginHelper(*test_url) &&
 367          MatchesPath(path_for_request);
 368 }
 369
 370 bool URLPattern::MatchesSecurityOrigin(const GURL& test) const {
 371   const GURL* test_url = &test;
 372   bool has_inner_url = test.inner_url() != NULL;
 373
 374   if (has_inner_url) {
 375     if (!test.SchemeIsFileSystem())
 376       return false;  // The only nested URLs we handle are filesystem URLs.
 377     test_url = test.inner_url();
 378   }
 379
 380   if (!MatchesScheme(test_url->scheme()))
 381     return false;
 382
 383   if (match_all_urls_)
 384     return true;
 385
 386   return MatchesSecurityOriginHelper(*test_url);
 387 }
 388
 389 bool URLPattern::MatchesScheme(const std::string& test) const {
 390   if (!IsValidScheme(test))
 391     return false;
 392
 393   return scheme_ == "*" || test == scheme_;
 394 }
 395
 396 bool URLPattern::MatchesHost(const std::string& host) const {
 397   std::string test(url::kHttpScheme);
 398   test += url::kStandardSchemeSeparator;
 399   test += host;
 400   test += "/";
 401   return MatchesHost(GURL(test));
 402 }
 403
 404 bool URLPattern::MatchesHost(const GURL& test) const {
 405   // If the hosts are exactly equal, we have a match.
 406   if (test.host() == host_)
 407     return true;
 408
 409   // If we're matching subdomains, and we have no host in the match pattern,
 410   // that means that we're matching all hosts, which means we have a match no
 411   // matter what the test host is.
 412   if (match_subdomains_ && host_.empty())
 413     return true;
 414
 415   // Otherwise, we can only match if our match pattern matches subdomains.
 416   if (!match_subdomains_)
 417     return false;
 418
 419   // We don't do subdomain matching against IP addresses, so we can give up now
 420   // if the test host is an IP address.
 421   if (test.HostIsIPAddress())
 422     return false;
 423
 424   // Check if the test host is a subdomain of our host.
 425   if (test.host().length() <= (host_.length() + 1))
 426     return false;
 427
 428   if (test.host().compare(test.host().length() - host_.length(),
 429                           host_.length(), host_) != 0)
 430     return false;
 431
 432   return test.host()[test.host().length() - host_.length() - 1] == '.';
 433 }
 434
 435 bool URLPattern::ImpliesAllHosts() const {
 436   // Check if it matches all urls or is a pattern like http://*/*.
 437   if (match_all_urls_ ||
 438       (match_subdomains_ && host_.empty() && port_ == "*" && path_ == "/*")) {
 439     return true;
 440   }
 441
 442   // If this doesn't even match subdomains, it can't possibly imply all hosts.
 443   if (!match_subdomains_)
 444     return false;
 445
 446   // If |host_| is a recognized TLD, this will be 0. We don't include private
 447   // TLDs, so that, e.g., *.appspot.com does not imply all hosts.
 448   size_t registry_length = net::registry_controlled_domains::GetRegistryLength(
 449       host_,
 450       net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
 451       net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
 452   // If there was more than just a TLD in the host (e.g., *.foobar.com), it
 453   // doesn't imply all hosts.
 454   if (registry_length > 0)
 455     return false;
 456
 457   // At this point the host could either be just a TLD ("com") or some unknown
 458   // TLD-like string ("notatld"). To disambiguate between them construct a
 459   // fake URL, and check the registry. This returns 0 if the TLD is
 460   // unrecognized, or the length of the recognized TLD.
 461   registry_length = net::registry_controlled_domains::GetRegistryLength(
 462       base::StringPrintf("foo.%s", host_.c_str()),
 463       net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
 464       net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
 465   // If we recognized this TLD, then this is a pattern like *.com, and it
 466   // should imply all hosts. Otherwise, this doesn't imply all hosts.
 467   return registry_length > 0;
 468 }
 469
 470 bool URLPattern::MatchesSingleOrigin() const {
 471   // Strictly speaking, the port is part of the origin, but in URLPattern it
 472   // defaults to *. It's not very interesting anyway, so leave it out.
 473   return !ImpliesAllHosts() && scheme_ != "*" && !match_subdomains_;
 474 }
 475
 476 bool URLPattern::MatchesPath(const std::string& test) const {
 477   // Make the behaviour of OverlapsWith consistent with MatchesURL, which is
 478   // need to match hosted apps on e.g. 'google.com' also run on 'google.com/'.
 479   if (test + "/*" == path_escaped_)
 480     return true;
 481
 482   return base::MatchPattern(test, path_escaped_);
 483 }
 484
 485 const std::string& URLPattern::GetAsString() const {
 486   if (!spec_.empty())
 487     return spec_;
 488
 489   if (match_all_urls_) {
 490     spec_ = kAllUrlsPattern;
 491     return spec_;
 492   }
 493
 494   bool standard_scheme = IsStandardScheme(scheme_);
 495
 496   std::string spec = scheme_ +
 497       (standard_scheme ? url::kStandardSchemeSeparator : ":");
 498
 499   if (scheme_ != url::kFileScheme && standard_scheme) {
 500     if (match_subdomains_) {
 501       spec += "*";
 502       if (!host_.empty())
 503         spec += ".";
 504     }
 505
 506     if (!host_.empty())
 507       spec += host_;
 508
 509     if (port_ != "*") {
 510       spec += ":";
 511       spec += port_;
 512     }
 513   }
 514
 515   if (!path_.empty())
 516     spec += path_;
 517
 518   spec_ = spec;
 519   return spec_;
 520 }
 521
 522 bool URLPattern::OverlapsWith(const URLPattern& other) const {
 523   if (match_all_urls() || other.match_all_urls())
 524     return true;
 525   return (MatchesAnyScheme(other.GetExplicitSchemes()) ||
 526           other.MatchesAnyScheme(GetExplicitSchemes()))
 527       && (MatchesHost(other.host()) || other.MatchesHost(host()))
 528       && (MatchesPortPattern(other.port()) || other.MatchesPortPattern(port()))
 529       && (MatchesPath(StripTrailingWildcard(other.path())) ||
 530           other.MatchesPath(StripTrailingWildcard(path())));
 531 }
 532
 533 bool URLPattern::Contains(const URLPattern& other) const {
 534   if (match_all_urls())
 535     return true;
 536   return MatchesAllSchemes(other.GetExplicitSchemes()) &&
 537          MatchesHost(other.host()) &&
 538          (!other.match_subdomains_ || match_subdomains_) &&
 539          MatchesPortPattern(other.port()) &&
 540          MatchesPath(StripTrailingWildcard(other.path()));
 541 }
 542
 543 bool URLPattern::MatchesAnyScheme(
 544     const std::vector<std::string>& schemes) const {
 545   for (std::vector<std::string>::const_iterator i = schemes.begin();
 546        i != schemes.end(); ++i) {
 547     if (MatchesScheme(*i))
 548       return true;
 549   }
 550
 551   return false;
 552 }
 553
 554 bool URLPattern::MatchesAllSchemes(
 555     const std::vector<std::string>& schemes) const {
 556   for (std::vector<std::string>::const_iterator i = schemes.begin();
 557        i != schemes.end(); ++i) {
 558     if (!MatchesScheme(*i))
 559       return false;
 560   }
 561
 562   return true;
 563 }
 564
 565 bool URLPattern::MatchesSecurityOriginHelper(const GURL& test) const {
 566   // Ignore hostname if scheme is file://.
 567   if (scheme_ != url::kFileScheme && !MatchesHost(test))
 568     return false;
 569
 570   if (!MatchesPortPattern(base::IntToString(test.EffectiveIntPort())))
 571     return false;
 572
 573   return true;
 574 }
 575
 576 bool URLPattern::MatchesPortPattern(const std::string& port) const {
 577   return port_ == "*" || port_ == port;
 578 }
 579
 580 std::vector<std::string> URLPattern::GetExplicitSchemes() const {
 581   std::vector<std::string> result;
 582
 583   if (scheme_ != "*" && !match_all_urls_ && IsValidScheme(scheme_)) {
 584     result.push_back(scheme_);
 585     return result;
 586   }
 587
 588   for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
 589     if (MatchesScheme(kValidSchemes[i])) {
 590       result.push_back(kValidSchemes[i]);
 591     }
 592   }
 593
 594   return result;
 595 }
 596
 597 std::vector<URLPattern> URLPattern::ConvertToExplicitSchemes() const {
 598   std::vector<std::string> explicit_schemes = GetExplicitSchemes();
 599   std::vector<URLPattern> result;
 600
 601   for (std::vector<std::string>::const_iterator i = explicit_schemes.begin();
 602        i != explicit_schemes.end(); ++i) {
 603     URLPattern temp = *this;
 604     temp.SetScheme(*i);
 605     temp.SetMatchAllURLs(false);
 606     result.push_back(temp);
 607   }
 608
 609   return result;
 610 }
 611
 612 // static
 613 const char* URLPattern::GetParseResultString(
 614     URLPattern::ParseResult parse_result) {
 615   return kParseResultMessages[parse_result];
 616 }