chrome_frame/html_utils.cc

   1 // Copyright (c) 2010 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 "chrome_frame/html_utils.h"
   6
   7 #include <atlbase.h>
   8 #include <urlmon.h>
   9
  10 #include "base/string_util.h"
  11 #include "base/string_tokenizer.h"
  12 #include "base/stringprintf.h"
  13 #include "chrome/common/chrome_version_info.h"
  14 #include "chrome_frame/utils.h"
  15 #include "net/base/net_util.h"
  16 #include "webkit/glue/user_agent.h"
  17
  18 const wchar_t kQuotes[] = L"\"'";
  19 const char kXFrameOptionsHeader[] = "X-Frame-Options";
  20 const char kXFrameOptionsValueAllowAll[] = "allowall";
  21
  22 HTMLScanner::StringRange::StringRange() {
  23 }
  24
  25 HTMLScanner::StringRange::StringRange(StrPos start, StrPos end)
  26     : start_(start), end_(end) {
  27 }
  28
  29 bool HTMLScanner::StringRange::LowerCaseEqualsASCII(const char* other) const {
  30   return ::LowerCaseEqualsASCII(start_, end_, other);
  31 }
  32
  33 bool HTMLScanner::StringRange::Equals(const wchar_t* other) const {
  34   int ret = wcsncmp(&start_[0], other, end_ - start_);
  35   if (ret == 0)
  36     ret = (other[end_ - start_] == L'\0') ? 0 : -1;
  37   return ret == 0;
  38 }
  39
  40 std::wstring HTMLScanner::StringRange::Copy() const {
  41   return std::wstring(start_, end_);
  42 }
  43
  44 bool HTMLScanner::StringRange::GetTagName(std::wstring* tag_name) const {
  45   if (*start_ != L'<') {
  46     LOG(ERROR) << "Badly formatted tag found";
  47     return false;
  48   }
  49
  50   StrPos name_start = start_;
  51   name_start++;
  52   while (name_start < end_ && IsWhitespace(*name_start))
  53     name_start++;
  54
  55   if (name_start >= end_) {
  56     // We seem to have a degenerate tag (i.e. <   >). Return false here.
  57     return false;
  58   }
  59
  60   StrPos name_end = name_start + 1;
  61   while (name_end < end_ && !IsWhitespace(*name_end))
  62     name_end++;
  63
  64   if (name_end > end_) {
  65     // This looks like an improperly formatted tab ('<foo'). Return false here.
  66     return false;
  67   }
  68
  69   tag_name->assign(name_start, name_end);
  70   return true;
  71 }
  72
  73
  74 bool HTMLScanner::StringRange::GetTagAttribute(const wchar_t* attribute_name,
  75     StringRange* attribute_value) const {
  76   if (NULL == attribute_name || NULL == attribute_value) {
  77     NOTREACHED();
  78     return false;
  79   }
  80
  81   // Use this so we can use the convenience method LowerCaseEqualsASCII()
  82   // from string_util.h.
  83   std::string search_name_ascii(WideToASCII(attribute_name));
  84
  85   WStringTokenizer tokenizer(start_, end_, L" =/");
  86   tokenizer.set_options(WStringTokenizer::RETURN_DELIMS);
  87
  88   // Set up the quote chars so that we get quoted attribute values as single
  89   // tokens.
  90   tokenizer.set_quote_chars(L"\"'");
  91
  92   const bool PARSE_STATE_NAME = true;
  93   const bool PARSE_STATE_VALUE = false;
  94   bool parse_state = PARSE_STATE_NAME;
  95
  96   // Used to skip the first token, which is the tag name.
  97   bool first_token_skipped = false;
  98
  99   // This is set during a loop iteration in which an '=' sign was spotted.
 100   // It is used to filter out degenerate tags such as:
 101   // <meta foo==bar>
 102   bool last_token_was_delim = false;
 103
 104   // Set this if the attribute name has been found that we might then
 105   // pick up the value in the next loop iteration.
 106   bool attribute_name_found = false;
 107
 108   while (tokenizer.GetNext()) {
 109     // If we have a whitespace delimiter, just keep going. Cases of this should
 110     // be reduced by the CollapseWhitespace call. If we have an '=' character,
 111     // we update our state and reiterate.
 112     if (tokenizer.token_is_delim()) {
 113       if (*tokenizer.token_begin() == L'=') {
 114         if (last_token_was_delim) {
 115           // Looks like we have a badly formed tag, just stop parsing now.
 116           return false;
 117         }
 118         parse_state = !parse_state;
 119         last_token_was_delim = true;
 120       }
 121       continue;
 122     }
 123
 124     last_token_was_delim = false;
 125
 126     // The first non-delimiter token is the tag name, which we don't want.
 127     if (!first_token_skipped) {
 128       first_token_skipped = true;
 129       continue;
 130     }
 131
 132     if (PARSE_STATE_NAME == parse_state) {
 133       // We have a tag name, check to see if it matches our target name:
 134       if (::LowerCaseEqualsASCII(tokenizer.token_begin(), tokenizer.token_end(),
 135                                  search_name_ascii.c_str())) {
 136         attribute_name_found = true;
 137         continue;
 138       }
 139     } else if (PARSE_STATE_VALUE == parse_state && attribute_name_found) {
 140       attribute_value->start_ = tokenizer.token_begin();
 141       attribute_value->end_ = tokenizer.token_end();
 142
 143       // Unquote the attribute value if need be.
 144       attribute_value->UnQuote();
 145
 146       return true;
 147     } else if (PARSE_STATE_VALUE == parse_state) {
 148       // If we haven't found the attribute name we want yet, ignore this token
 149       // and go back to looking for our name.
 150       parse_state = PARSE_STATE_NAME;
 151     }
 152   }
 153
 154   return false;
 155 }
 156
 157 bool HTMLScanner::StringRange::UnQuote() {
 158   if (start_ + 2 > end_) {
 159     // String's too short to be quoted, bail.
 160     return false;
 161   }
 162
 163   if ((*start_ == L'\'' && *(end_ - 1) == L'\'') ||
 164       (*start_ == L'"' && *(end_ - 1) == L'"')) {
 165     start_ = start_ + 1;
 166     end_ = end_ - 1;
 167     return true;
 168   }
 169
 170   return false;
 171 }
 172
 173 HTMLScanner::HTMLScanner(const wchar_t* html_string)
 174     : html_string_(CollapseWhitespace(html_string, true)),
 175       quotes_(kQuotes) {
 176 }
 177
 178 void HTMLScanner::GetTagsByName(const wchar_t* name, StringRangeList* tag_list,
 179                                 const wchar_t* stop_tag) {
 180   DCHECK(NULL != name);
 181   DCHECK(NULL != tag_list);
 182   DCHECK(NULL != stop_tag);
 183
 184   StringRange remaining_html(html_string_.begin(), html_string_.end());
 185
 186   std::wstring search_name(name);
 187   TrimWhitespace(search_name, TRIM_ALL, &search_name);
 188
 189   // Use this so we can use the convenience method LowerCaseEqualsASCII()
 190   // from string_util.h.
 191   std::string search_name_ascii(WideToASCII(search_name));
 192   std::string stop_tag_ascii(WideToASCII(stop_tag));
 193
 194   StringRange current_tag;
 195   std::wstring current_name;
 196   while (NextTag(&remaining_html, &current_tag)) {
 197     if (current_tag.GetTagName(&current_name)) {
 198       if (LowerCaseEqualsASCII(current_name, search_name_ascii.c_str())) {
 199         tag_list->push_back(current_tag);
 200       } else if (LowerCaseEqualsASCII(current_name, stop_tag_ascii.c_str())) {
 201         // We hit the stop tag so it's time to go home.
 202         break;
 203       }
 204     }
 205   }
 206 }
 207
 208 struct ScanState {
 209   bool in_quote;
 210   bool in_escape;
 211   wchar_t quote_char;
 212   ScanState() : in_quote(false), in_escape(false) {}
 213 };
 214
 215 bool HTMLScanner::IsQuote(wchar_t c) {
 216   return quotes_.find(c) != std::wstring::npos;
 217 }
 218
 219 bool HTMLScanner::IsHTMLCommentClose(const StringRange* html_string,
 220                                      StrPos pos) {
 221   if (pos < html_string->end_ && pos > html_string->start_ + 2 &&
 222       *pos == L'>') {
 223     return *(pos-1) == L'-' && *(pos-2) == L'-';
 224   }
 225   return false;
 226 }
 227
 228 bool HTMLScanner::IsIEConditionalCommentClose(const StringRange* html_string,
 229                                               StrPos pos) {
 230   if (pos < html_string->end_ && pos > html_string->start_ + 2 &&
 231       *pos == L'>') {
 232     return *(pos-1) == L']';
 233   }
 234   return false;
 235 }
 236
 237
 238 bool HTMLScanner::NextTag(StringRange* html_string, StringRange* tag) {
 239   DCHECK(NULL != html_string);
 240   DCHECK(NULL != tag);
 241
 242   tag->start_ = html_string->start_;
 243   while (tag->start_ < html_string->end_ && *tag->start_ != L'<') {
 244     tag->start_++;
 245   }
 246
 247   // we went past the end of the string.
 248   if (tag->start_ >= html_string->end_) {
 249     return false;
 250   }
 251
 252   tag->end_ = tag->start_ + 1;
 253
 254   // Get the tag name to see if we are in an HTML comment. If we are, then
 255   // don't consider quotes. This should work for example:
 256   // <!-- foo ' --> <meta foo='bar'>
 257   std::wstring tag_name;
 258   StringRange start_range(tag->start_, html_string->end_);
 259   start_range.GetTagName(&tag_name);
 260   if (StartsWith(tag_name, L"!--[if", true)) {
 261     // This looks like the beginning of an IE conditional comment, scan until
 262     // we hit the end which always looks like ']>'. For now we disregard the
 263     // contents of the condition, and always assume true.
 264     // TODO(robertshield): Optionally support the grammar defined by
 265     // http://msdn.microsoft.com/en-us/library/ms537512(VS.85).aspx#syntax.
 266     while (tag->end_ < html_string->end_ &&
 267            !IsIEConditionalCommentClose(html_string, tag->end_)) {
 268       tag->end_++;
 269     }
 270   } else if (StartsWith(tag_name, L"!--", true)) {
 271     // We're inside a comment tag which ends in '-->'. Keep going until we
 272     // reach the end.
 273     while (tag->end_ < html_string->end_ &&
 274            !IsHTMLCommentClose(html_string, tag->end_)) {
 275       tag->end_++;
 276     }
 277   } else if (StartsWith(tag_name, L"![endif", true)) {
 278     // We're inside the closing tag of an IE conditional comment which ends in
 279     // either '-->' of ']>'. Keep going until we reach the end.
 280     while (tag->end_ < html_string->end_ &&
 281            !IsIEConditionalCommentClose(html_string, tag->end_) &&
 282            !IsHTMLCommentClose(html_string, tag->end_)) {
 283       tag->end_++;
 284     }
 285   } else {
 286     // Properly handle quoted strings within non-comment tags by maintaining
 287     // some state while scanning. Specifically, we have to maintain state on
 288     // whether we are inside a string, what the string terminating character
 289     // will be and whether we are inside an escape sequence.
 290     ScanState state;
 291     while (tag->end_ < html_string->end_) {
 292       if (state.in_quote) {
 293         if (state.in_escape) {
 294           state.in_escape = false;
 295         } else if (*tag->end_ == '\\') {
 296           state.in_escape = true;
 297         } else if (*tag->end_ == state.quote_char) {
 298           state.in_quote = false;
 299         }
 300       } else {
 301         state.in_quote = IsQuote(state.quote_char = *tag->end_);
 302       }
 303
 304       if (!state.in_quote && *tag->end_ == L'>') {
 305         break;
 306       }
 307       tag->end_++;
 308     }
 309   }
 310
 311   // We hit the end_ but found no matching tag closure. Consider this an
 312   // incomplete tag and do not report it.
 313   if (tag->end_ >= html_string->end_)
 314     return false;
 315
 316   // Modify html_string to point to just beyond the end_ of the current tag.
 317   html_string->start_ = tag->end_ + 1;
 318
 319   return true;
 320 }
 321
 322 namespace http_utils {
 323
 324 const char kChromeFrameUserAgent[] = "chromeframe";
 325 static char g_cf_user_agent[100] = {0};
 326 static char g_chrome_user_agent[255] = {0};
 327
 328 const char* GetChromeFrameUserAgent() {
 329   if (!g_cf_user_agent[0]) {
 330     _pAtlModule->m_csStaticDataInitAndTypeInfo.Lock();
 331     if (!g_cf_user_agent[0]) {
 332       uint32 high_version = 0, low_version = 0;
 333       GetModuleVersion(reinterpret_cast<HMODULE>(&__ImageBase), &high_version,
 334                        &low_version);
 335       wsprintfA(g_cf_user_agent, "%s/%i.%i.%i.%i", kChromeFrameUserAgent,
 336                 HIWORD(high_version), LOWORD(high_version),
 337                 HIWORD(low_version), LOWORD(low_version));
 338     }
 339     _pAtlModule->m_csStaticDataInitAndTypeInfo.Unlock();
 340   }
 341   return g_cf_user_agent;
 342 }
 343
 344 std::string AddChromeFrameToUserAgentValue(const std::string& value) {
 345   if (value.empty()) {
 346     DLOG(WARNING) << "empty user agent value";
 347     return "";
 348   }
 349
 350   DCHECK_EQ(false, StartsWithASCII(value, "User-Agent:", true));
 351
 352   if (value.find(kChromeFrameUserAgent) != std::string::npos) {
 353     // Our user agent has already been added.
 354     return value;
 355   }
 356
 357   std::string ret(value);
 358   std::string::size_type insert_position = ret.find(')');
 359   if (insert_position != std::string::npos) {
 360     if (insert_position > 1 && isalnum(ret[insert_position - 1]))
 361       ret.insert(insert_position++, ";");
 362     ret.insert(insert_position++, " ");
 363     ret.insert(insert_position, GetChromeFrameUserAgent());
 364   } else {
 365     ret += " ";
 366     ret += GetChromeFrameUserAgent();
 367   }
 368
 369   return ret;
 370 }
 371
 372 std::string GetDefaultUserAgentHeaderWithCFTag() {
 373   std::string ua(GetDefaultUserAgent());
 374   return "User-Agent: " + AddChromeFrameToUserAgentValue(ua);
 375 }
 376
 377 const char* GetChromeUserAgent() {
 378   if (!g_chrome_user_agent[0]) {
 379     _pAtlModule->m_csStaticDataInitAndTypeInfo.Lock();
 380     if (!g_chrome_user_agent[0]) {
 381       std::string ua;
 382
 383       chrome::VersionInfo version_info;
 384       std::string product("Chrome/");
 385       product += version_info.is_valid() ? version_info.Version()
 386                                          : "0.0.0.0";
 387
 388       ua = webkit_glue::BuildUserAgentHelper(false, product);
 389
 390       DCHECK(ua.length() < arraysize(g_chrome_user_agent));
 391       lstrcpynA(g_chrome_user_agent, ua.c_str(),
 392                 arraysize(g_chrome_user_agent) - 1);
 393     }
 394     _pAtlModule->m_csStaticDataInitAndTypeInfo.Unlock();
 395   }
 396   return g_chrome_user_agent;
 397 }
 398
 399 std::string GetDefaultUserAgent() {
 400   std::string ret;
 401   DWORD size = MAX_PATH;
 402   HRESULT hr = E_OUTOFMEMORY;
 403   for (int retries = 1; hr == E_OUTOFMEMORY && retries <= 10; ++retries) {
 404     hr = ::ObtainUserAgentString(0, WriteInto(&ret, size + 1), &size);
 405     if (hr == E_OUTOFMEMORY) {
 406       size = MAX_PATH * retries;
 407     } else if (SUCCEEDED(hr)) {
 408       // Truncate the extra allocation.
 409       DCHECK_GT(size, 0U);
 410       ret.resize(size - 1);
 411     }
 412   }
 413
 414   if (FAILED(hr)) {
 415     NOTREACHED() << base::StringPrintf("ObtainUserAgentString==0x%08X", hr);
 416     return std::string();
 417   }
 418
 419   return ret;
 420 }
 421
 422 bool HasFrameBustingHeader(const std::string& http_headers) {
 423   // NOTE: We cannot use net::GetSpecificHeader() here since when there are
 424   // multiple instances of a header that returns the first value seen, and we
 425   // need to look at all instances.
 426   net::HttpUtil::HeadersIterator it(http_headers.begin(), http_headers.end(),
 427                                     "\r\n");
 428   while (it.GetNext()) {
 429     if (!lstrcmpiA(it.name().c_str(), kXFrameOptionsHeader) &&
 430         lstrcmpiA(it.values().c_str(), kXFrameOptionsValueAllowAll))
 431       return true;
 432   }
 433   return false;
 434 }
 435
 436 std::string GetHttpHeaderFromHeaderList(const std::string& header,
 437                                         const std::string& headers) {
 438   net::HttpUtil::HeadersIterator it(headers.begin(), headers.end(), "\r\n");
 439   while (it.GetNext()) {
 440     if (!lstrcmpiA(it.name().c_str(), header.c_str()))
 441       return std::string(it.values_begin(), it.values_end());
 442   }
 443   return std::string();
 444 }
 445
 446 }  // namespace http_utils