chrome_frame/html_utils.cc

   1 // Copyright (c) 2009 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 "chrome_frame/utils.h"
  13
  14 const wchar_t kQuotes[] = L"\"'";
  15 const char kXFrameOptionsHeader[] = "X-Frame-Options";
  16 const char kXFrameOptionsValueAllowAll[] = "allowall";
  17
  18 HTMLScanner::StringRange::StringRange() {
  19 }
  20
  21 HTMLScanner::StringRange::StringRange(StrPos start, StrPos end)
  22     : start_(start), end_(end) {
  23 }
  24
  25 bool HTMLScanner::StringRange::LowerCaseEqualsASCII(const char* other) const {
  26   return ::LowerCaseEqualsASCII(start_, end_, other);
  27 }
  28
  29 bool HTMLScanner::StringRange::Equals(const wchar_t* other) const {
  30   int ret = wcsncmp(&start_[0], other, end_ - start_);
  31   if (ret == 0)
  32     ret = (other[end_ - start_] == L'\0') ? 0 : -1;
  33   return ret == 0;
  34 }
  35
  36 std::wstring HTMLScanner::StringRange::Copy() const {
  37   return std::wstring(start_, end_);
  38 }
  39
  40 bool HTMLScanner::StringRange::GetTagName(std::wstring* tag_name) const {
  41   if (*start_ != L'<') {
  42     LOG(ERROR) << "Badly formatted tag found";
  43     return false;
  44   }
  45
  46   StrPos name_start = start_;
  47   name_start++;
  48   while (name_start < end_ && IsWhitespace(*name_start))
  49     name_start++;
  50
  51   if (name_start >= end_) {
  52     // We seem to have a degenerate tag (i.e. <   >). Return false here.
  53     return false;
  54   }
  55
  56   StrPos name_end = name_start + 1;
  57   while (name_end < end_ && !IsWhitespace(*name_end))
  58     name_end++;
  59
  60   if (name_end > end_) {
  61     // This looks like an improperly formatted tab ('<foo'). Return false here.
  62     return false;
  63   }
  64
  65   tag_name->assign(name_start, name_end);
  66   return true;
  67 }
  68
  69
  70 bool HTMLScanner::StringRange::GetTagAttribute(const wchar_t* attribute_name,
  71     StringRange* attribute_value) const {
  72   if (NULL == attribute_name || NULL == attribute_value) {
  73     NOTREACHED();
  74     return false;
  75   }
  76
  77   // Use this so we can use the convenience method LowerCaseEqualsASCII()
  78   // from string_util.h.
  79   std::string search_name_ascii(WideToASCII(attribute_name));
  80
  81   WStringTokenizer tokenizer(start_, end_, L" =/");
  82   tokenizer.set_options(WStringTokenizer::RETURN_DELIMS);
  83
  84   // Set up the quote chars so that we get quoted attribute values as single
  85   // tokens.
  86   tokenizer.set_quote_chars(L"\"'");
  87
  88   const bool PARSE_STATE_NAME = true;
  89   const bool PARSE_STATE_VALUE = false;
  90   bool parse_state = PARSE_STATE_NAME;
  91
  92   // Used to skip the first token, which is the tag name.
  93   bool first_token_skipped = false;
  94
  95   // This is set during a loop iteration in which an '=' sign was spotted.
  96   // It is used to filter out degenerate tags such as:
  97   // <meta foo==bar>
  98   bool last_token_was_delim = false;
  99
 100   // Set this if the attribute name has been found that we might then
 101   // pick up the value in the next loop iteration.
 102   bool attribute_name_found = false;
 103
 104   while (tokenizer.GetNext()) {
 105     // If we have a whitespace delimiter, just keep going. Cases of this should
 106     // be reduced by the CollapseWhitespace call. If we have an '=' character,
 107     // we update our state and reiterate.
 108     if (tokenizer.token_is_delim()) {
 109       if (*tokenizer.token_begin() == L'=') {
 110         if (last_token_was_delim) {
 111           // Looks like we have a badly formed tag, just stop parsing now.
 112           return false;
 113         }
 114         parse_state = !parse_state;
 115         last_token_was_delim = true;
 116       }
 117       continue;
 118     }
 119
 120     last_token_was_delim = false;
 121
 122     // The first non-delimiter token is the tag name, which we don't want.
 123     if (!first_token_skipped) {
 124       first_token_skipped = true;
 125       continue;
 126     }
 127
 128     if (PARSE_STATE_NAME == parse_state) {
 129       // We have a tag name, check to see if it matches our target name:
 130       if (::LowerCaseEqualsASCII(tokenizer.token_begin(), tokenizer.token_end(),
 131                                  search_name_ascii.c_str())) {
 132         attribute_name_found = true;
 133         continue;
 134       }
 135     } else if (PARSE_STATE_VALUE == parse_state && attribute_name_found) {
 136       attribute_value->start_ = tokenizer.token_begin();
 137       attribute_value->end_ = tokenizer.token_end();
 138
 139       // Unquote the attribute value if need be.
 140       attribute_value->UnQuote();
 141
 142       return true;
 143     } else if (PARSE_STATE_VALUE == parse_state) {
 144       // If we haven't found the attribute name we want yet, ignore this token
 145       // and go back to looking for our name.
 146       parse_state = PARSE_STATE_NAME;
 147     }
 148   }
 149
 150   return false;
 151 }
 152
 153 bool HTMLScanner::StringRange::UnQuote() {
 154   if (start_ + 2 > end_) {
 155     // String's too short to be quoted, bail.
 156     return false;
 157   }
 158
 159   if ((*start_ == L'\'' && *(end_ - 1) == L'\'') ||
 160       (*start_ == L'"' && *(end_ - 1) == L'"')) {
 161     start_ = start_ + 1;
 162     end_ = end_ - 1;
 163     return true;
 164   }
 165
 166   return false;
 167 }
 168
 169 HTMLScanner::HTMLScanner(const wchar_t* html_string)
 170     : html_string_(CollapseWhitespace(html_string, true)),
 171       quotes_(kQuotes) {
 172 }
 173
 174 void HTMLScanner::GetTagsByName(const wchar_t* name, StringRangeList* tag_list,
 175                                 const wchar_t* stop_tag) {
 176   DCHECK(NULL != name);
 177   DCHECK(NULL != tag_list);
 178   DCHECK(NULL != stop_tag);
 179
 180   StringRange remaining_html(html_string_.begin(), html_string_.end());
 181
 182   std::wstring search_name(name);
 183   TrimWhitespace(search_name, TRIM_ALL, &search_name);
 184
 185   // Use this so we can use the convenience method LowerCaseEqualsASCII()
 186   // from string_util.h.
 187   std::string search_name_ascii(WideToASCII(search_name));
 188   std::string stop_tag_ascii(WideToASCII(stop_tag));
 189
 190   StringRange current_tag;
 191   std::wstring current_name;
 192   while (NextTag(&remaining_html, &current_tag)) {
 193     if (current_tag.GetTagName(&current_name)) {
 194       if (LowerCaseEqualsASCII(current_name, search_name_ascii.c_str())) {
 195         tag_list->push_back(current_tag);
 196       } else if (LowerCaseEqualsASCII(current_name, stop_tag_ascii.c_str())) {
 197         // We hit the stop tag so it's time to go home.
 198         break;
 199       }
 200     }
 201   }
 202 }
 203
 204 struct ScanState {
 205   bool in_quote;
 206   bool in_escape;
 207   wchar_t quote_char;
 208   ScanState() : in_quote(false), in_escape(false) {}
 209 };
 210
 211 bool HTMLScanner::IsQuote(wchar_t c) {
 212   return quotes_.find(c) != std::wstring::npos;
 213 }
 214
 215 bool HTMLScanner::IsHTMLCommentClose(StringRange* html_string, StrPos pos) {
 216   if (pos < html_string->end_ && pos > html_string->start_ + 2 &&
 217       *pos == L'>') {
 218     return *(pos-1) == L'-' && *(pos-2) == L'-';
 219   }
 220   return false;
 221 }
 222
 223 bool HTMLScanner::NextTag(StringRange* html_string, StringRange* tag) {
 224   DCHECK(NULL != html_string);
 225   DCHECK(NULL != tag);
 226
 227   tag->start_ = html_string->start_;
 228   while (tag->start_ < html_string->end_ && *tag->start_ != L'<') {
 229     tag->start_++;
 230   }
 231
 232   // we went past the end of the string.
 233   if (tag->start_ >= html_string->end_) {
 234     return false;
 235   }
 236
 237   tag->end_ = tag->start_ + 1;
 238
 239   // Get the tag name to see if we are in an HTML comment. If we are, then
 240   // don't consider quotes. This should work for example:
 241   // <!-- foo ' --> <meta foo='bar'>
 242   std::wstring tag_name;
 243   StringRange start_range(tag->start_, html_string->end_);
 244   start_range.GetTagName(&tag_name);
 245   if (StartsWith(tag_name, L"!--", true)) {
 246     // We're inside a comment tag, keep going until we get out of it.
 247     while (tag->end_ < html_string->end_ &&
 248            !IsHTMLCommentClose(html_string, tag->end_)) {
 249       tag->end_++;
 250     }
 251   } else {
 252     // Properly handle quoted strings within non-comment tags by maintaining
 253     // some state while scanning. Specifically, we have to maintain state on
 254     // whether we are inside a string, what the string terminating character
 255     // will be and whether we are inside an escape sequence.
 256     ScanState state;
 257     while (tag->end_ < html_string->end_) {
 258       if (state.in_quote) {
 259         if (state.in_escape) {
 260           state.in_escape = false;
 261         } else if (*tag->end_ == '\\') {
 262           state.in_escape = true;
 263         } else if (*tag->end_ == state.quote_char) {
 264           state.in_quote = false;
 265         }
 266       } else {
 267         state.in_quote = IsQuote(state.quote_char = *tag->end_);
 268       }
 269
 270       if (!state.in_quote && *tag->end_ == L'>') {
 271         break;
 272       }
 273       tag->end_++;
 274     }
 275   }
 276
 277   // We hit the end_ but found no matching tag closure. Consider this an
 278   // incomplete tag and do not report it.
 279   if (tag->end_ >= html_string->end_)
 280     return false;
 281
 282   // Modify html_string to point to just beyond the end_ of the current tag.
 283   html_string->start_ = tag->end_ + 1;
 284
 285   return true;
 286 }
 287
 288 namespace http_utils {
 289
 290 const char kChromeFrameUserAgent[] = "chromeframe";
 291
 292 const char* GetChromeFrameUserAgent() {
 293   static char cf_user_agent[100] = {0};
 294   if (!cf_user_agent[0]) {
 295     _pAtlModule->m_csStaticDataInitAndTypeInfo.Lock();
 296     if (!cf_user_agent[0]) {
 297       uint32 high_version = 0, low_version = 0;
 298       GetModuleVersion(reinterpret_cast<HMODULE>(&__ImageBase), &high_version,
 299                        &low_version);
 300       wsprintfA(cf_user_agent, "%s/%i.%i.%i.%i", kChromeFrameUserAgent,
 301                 HIWORD(high_version), LOWORD(high_version),
 302                 HIWORD(low_version), LOWORD(low_version));
 303     }
 304     _pAtlModule->m_csStaticDataInitAndTypeInfo.Unlock();
 305   }
 306   return cf_user_agent;
 307 }
 308
 309 std::string AddChromeFrameToUserAgentValue(const std::string& value) {
 310   if (value.empty()) {
 311     DLOG(WARNING) << "empty user agent value";
 312     return "";
 313   }
 314
 315   DCHECK_EQ(false, StartsWithASCII(value, "User-Agent:", true));
 316
 317   if (value.find(kChromeFrameUserAgent) != std::string::npos) {
 318     // Our user agent has already been added.
 319     return value;
 320   }
 321
 322   std::string ret(value);
 323   ret += " ";
 324   ret += GetChromeFrameUserAgent();
 325
 326   return ret;
 327 }
 328
 329 std::string GetDefaultUserAgentHeaderWithCFTag() {
 330   std::string ua(GetDefaultUserAgent());
 331   return "User-Agent: " + AddChromeFrameToUserAgentValue(ua);
 332 }
 333
 334 std::string GetDefaultUserAgent() {
 335   std::string ret;
 336   DWORD size = MAX_PATH;  // NOLINT
 337   HRESULT hr = E_OUTOFMEMORY;
 338   for (int retries = 1; hr == E_OUTOFMEMORY && retries <= 10; ++retries) {
 339     hr = ::ObtainUserAgentString(0, WriteInto(&ret, size + 1), &size);
 340     if (hr == E_OUTOFMEMORY) {
 341       size = MAX_PATH * retries;
 342     } else if (SUCCEEDED(hr)) {
 343       // Truncate the extra allocation.
 344       DCHECK(size > 0);  // NOLINT
 345       ret.resize(size - sizeof(char));  // NOLINT
 346     }
 347   }
 348
 349   if (FAILED(hr)) {
 350     NOTREACHED() << StringPrintf("ObtainUserAgentString==0x%08X", hr);
 351     return "";
 352   } else {
 353     DCHECK(ret.length() == lstrlenA(ret.c_str()));
 354   }
 355
 356   return ret;
 357 }
 358
 359 bool HasFrameBustingHeader(const std::string& http_headers) {
 360   net::HttpUtil::HeadersIterator it(
 361       http_headers.begin(), http_headers.end(), "\r\n");
 362   while (it.GetNext()) {
 363     if (it.name() == kXFrameOptionsHeader) {
 364       std::string allow_all(kXFrameOptionsValueAllowAll);
 365       if (it.values_end() - it.values_begin() != allow_all.length() ||
 366           !std::equal(it.values_begin(), it.values_end(),
 367               allow_all.begin(),
 368               CaseInsensitiveCompareASCII<const char>())) {
 369         return true;
 370       }
 371     }
 372   }
 373
 374   return false;
 375 }
 376
 377 }  // namespace http_utils