• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 // The rules for header parsing were borrowed from Firefox:
6 // http://lxr.mozilla.org/seamonkey/source/netwerk/protocol/http/src/nsHttpResponseHead.cpp
7 // The rules for parsing content-types were also borrowed from Firefox:
8 // http://lxr.mozilla.org/mozilla/source/netwerk/base/src/nsURLHelper.cpp#834
9 
10 #include "net/http/http_response_headers.h"
11 
12 #include <algorithm>
13 
14 #include "base/format_macros.h"
15 #include "base/logging.h"
16 #include "base/metrics/histogram.h"
17 #include "base/pickle.h"
18 #include "base/strings/string_number_conversions.h"
19 #include "base/strings/string_piece.h"
20 #include "base/strings/string_util.h"
21 #include "base/strings/stringprintf.h"
22 #include "base/time/time.h"
23 #include "base/values.h"
24 #include "net/base/escape.h"
25 #include "net/http/http_byte_range.h"
26 #include "net/http/http_log_util.h"
27 #include "net/http/http_util.h"
28 
29 using base::StringPiece;
30 using base::Time;
31 using base::TimeDelta;
32 
33 namespace net {
34 
35 //-----------------------------------------------------------------------------
36 
37 namespace {
38 
39 // These headers are RFC 2616 hop-by-hop headers;
40 // not to be stored by caches.
41 const char* const kHopByHopResponseHeaders[] = {
42   "connection",
43   "proxy-connection",
44   "keep-alive",
45   "trailer",
46   "transfer-encoding",
47   "upgrade"
48 };
49 
50 // These headers are challenge response headers;
51 // not to be stored by caches.
52 const char* const kChallengeResponseHeaders[] = {
53   "www-authenticate",
54   "proxy-authenticate"
55 };
56 
57 // These headers are cookie setting headers;
58 // not to be stored by caches or disclosed otherwise.
59 const char* const kCookieResponseHeaders[] = {
60   "set-cookie",
61   "set-cookie2"
62 };
63 
64 // By default, do not cache Strict-Transport-Security or Public-Key-Pins.
65 // This avoids erroneously re-processing them on page loads from cache ---
66 // they are defined to be valid only on live and error-free HTTPS
67 // connections.
68 const char* const kSecurityStateHeaders[] = {
69   "strict-transport-security",
70   "public-key-pins"
71 };
72 
73 // These response headers are not copied from a 304/206 response to the cached
74 // response headers.  This list is based on Mozilla's nsHttpResponseHead.cpp.
75 const char* const kNonUpdatedHeaders[] = {
76   "connection",
77   "proxy-connection",
78   "keep-alive",
79   "www-authenticate",
80   "proxy-authenticate",
81   "trailer",
82   "transfer-encoding",
83   "upgrade",
84   "etag",
85   "x-frame-options",
86   "x-xss-protection",
87 };
88 
89 // Some header prefixes mean "Don't copy this header from a 304 response.".
90 // Rather than listing all the relevant headers, we can consolidate them into
91 // this list:
92 const char* const kNonUpdatedHeaderPrefixes[] = {
93   "content-",
94   "x-content-",
95   "x-webkit-"
96 };
97 
ShouldUpdateHeader(const std::string::const_iterator & name_begin,const std::string::const_iterator & name_end)98 bool ShouldUpdateHeader(const std::string::const_iterator& name_begin,
99                         const std::string::const_iterator& name_end) {
100   for (size_t i = 0; i < arraysize(kNonUpdatedHeaders); ++i) {
101     if (LowerCaseEqualsASCII(name_begin, name_end, kNonUpdatedHeaders[i]))
102       return false;
103   }
104   for (size_t i = 0; i < arraysize(kNonUpdatedHeaderPrefixes); ++i) {
105     if (StartsWithASCII(std::string(name_begin, name_end),
106                         kNonUpdatedHeaderPrefixes[i], false))
107       return false;
108   }
109   return true;
110 }
111 
CheckDoesNotHaveEmbededNulls(const std::string & str)112 void CheckDoesNotHaveEmbededNulls(const std::string& str) {
113   // Care needs to be taken when adding values to the raw headers string to
114   // make sure it does not contain embeded NULLs. Any embeded '\0' may be
115   // understood as line terminators and change how header lines get tokenized.
116   CHECK(str.find('\0') == std::string::npos);
117 }
118 
119 }  // namespace
120 
121 const char HttpResponseHeaders::kContentRange[] = "Content-Range";
122 
123 struct HttpResponseHeaders::ParsedHeader {
124   // A header "continuation" contains only a subsequent value for the
125   // preceding header.  (Header values are comma separated.)
is_continuationnet::HttpResponseHeaders::ParsedHeader126   bool is_continuation() const { return name_begin == name_end; }
127 
128   std::string::const_iterator name_begin;
129   std::string::const_iterator name_end;
130   std::string::const_iterator value_begin;
131   std::string::const_iterator value_end;
132 };
133 
134 //-----------------------------------------------------------------------------
135 
HttpResponseHeaders(const std::string & raw_input)136 HttpResponseHeaders::HttpResponseHeaders(const std::string& raw_input)
137     : response_code_(-1) {
138   Parse(raw_input);
139 
140   // The most important thing to do with this histogram is find out
141   // the existence of unusual HTTP status codes.  As it happens
142   // right now, there aren't double-constructions of response headers
143   // using this constructor, so our counts should also be accurate,
144   // without instantiating the histogram in two places.  It is also
145   // important that this histogram not collect data in the other
146   // constructor, which rebuilds an histogram from a pickle, since
147   // that would actually create a double call between the original
148   // HttpResponseHeader that was serialized, and initialization of the
149   // new object from that pickle.
150   UMA_HISTOGRAM_CUSTOM_ENUMERATION("Net.HttpResponseCode",
151                                    HttpUtil::MapStatusCodeForHistogram(
152                                        response_code_),
153                                    // Note the third argument is only
154                                    // evaluated once, see macro
155                                    // definition for details.
156                                    HttpUtil::GetStatusCodesForHistogram());
157 }
158 
HttpResponseHeaders(const Pickle & pickle,PickleIterator * iter)159 HttpResponseHeaders::HttpResponseHeaders(const Pickle& pickle,
160                                          PickleIterator* iter)
161     : response_code_(-1) {
162   std::string raw_input;
163   if (pickle.ReadString(iter, &raw_input))
164     Parse(raw_input);
165 }
166 
Persist(Pickle * pickle,PersistOptions options)167 void HttpResponseHeaders::Persist(Pickle* pickle, PersistOptions options) {
168   if (options == PERSIST_RAW) {
169     pickle->WriteString(raw_headers_);
170     return;  // Done.
171   }
172 
173   HeaderSet filter_headers;
174 
175   // Construct set of headers to filter out based on options.
176   if ((options & PERSIST_SANS_NON_CACHEABLE) == PERSIST_SANS_NON_CACHEABLE)
177     AddNonCacheableHeaders(&filter_headers);
178 
179   if ((options & PERSIST_SANS_COOKIES) == PERSIST_SANS_COOKIES)
180     AddCookieHeaders(&filter_headers);
181 
182   if ((options & PERSIST_SANS_CHALLENGES) == PERSIST_SANS_CHALLENGES)
183     AddChallengeHeaders(&filter_headers);
184 
185   if ((options & PERSIST_SANS_HOP_BY_HOP) == PERSIST_SANS_HOP_BY_HOP)
186     AddHopByHopHeaders(&filter_headers);
187 
188   if ((options & PERSIST_SANS_RANGES) == PERSIST_SANS_RANGES)
189     AddHopContentRangeHeaders(&filter_headers);
190 
191   if ((options & PERSIST_SANS_SECURITY_STATE) == PERSIST_SANS_SECURITY_STATE)
192     AddSecurityStateHeaders(&filter_headers);
193 
194   std::string blob;
195   blob.reserve(raw_headers_.size());
196 
197   // This copies the status line w/ terminator null.
198   // Note raw_headers_ has embedded nulls instead of \n,
199   // so this just copies the first header line.
200   blob.assign(raw_headers_.c_str(), strlen(raw_headers_.c_str()) + 1);
201 
202   for (size_t i = 0; i < parsed_.size(); ++i) {
203     DCHECK(!parsed_[i].is_continuation());
204 
205     // Locate the start of the next header.
206     size_t k = i;
207     while (++k < parsed_.size() && parsed_[k].is_continuation()) {}
208     --k;
209 
210     std::string header_name(parsed_[i].name_begin, parsed_[i].name_end);
211     StringToLowerASCII(&header_name);
212 
213     if (filter_headers.find(header_name) == filter_headers.end()) {
214       // Make sure there is a null after the value.
215       blob.append(parsed_[i].name_begin, parsed_[k].value_end);
216       blob.push_back('\0');
217     }
218 
219     i = k;
220   }
221   blob.push_back('\0');
222 
223   pickle->WriteString(blob);
224 }
225 
Update(const HttpResponseHeaders & new_headers)226 void HttpResponseHeaders::Update(const HttpResponseHeaders& new_headers) {
227   DCHECK(new_headers.response_code() == 304 ||
228          new_headers.response_code() == 206);
229 
230   // Copy up to the null byte.  This just copies the status line.
231   std::string new_raw_headers(raw_headers_.c_str());
232   new_raw_headers.push_back('\0');
233 
234   HeaderSet updated_headers;
235 
236   // NOTE: we write the new headers then the old headers for convenience.  The
237   // order should not matter.
238 
239   // Figure out which headers we want to take from new_headers:
240   for (size_t i = 0; i < new_headers.parsed_.size(); ++i) {
241     const HeaderList& new_parsed = new_headers.parsed_;
242 
243     DCHECK(!new_parsed[i].is_continuation());
244 
245     // Locate the start of the next header.
246     size_t k = i;
247     while (++k < new_parsed.size() && new_parsed[k].is_continuation()) {}
248     --k;
249 
250     const std::string::const_iterator& name_begin = new_parsed[i].name_begin;
251     const std::string::const_iterator& name_end = new_parsed[i].name_end;
252     if (ShouldUpdateHeader(name_begin, name_end)) {
253       std::string name(name_begin, name_end);
254       StringToLowerASCII(&name);
255       updated_headers.insert(name);
256 
257       // Preserve this header line in the merged result, making sure there is
258       // a null after the value.
259       new_raw_headers.append(name_begin, new_parsed[k].value_end);
260       new_raw_headers.push_back('\0');
261     }
262 
263     i = k;
264   }
265 
266   // Now, build the new raw headers.
267   MergeWithHeaders(new_raw_headers, updated_headers);
268 }
269 
MergeWithHeaders(const std::string & raw_headers,const HeaderSet & headers_to_remove)270 void HttpResponseHeaders::MergeWithHeaders(const std::string& raw_headers,
271                                            const HeaderSet& headers_to_remove) {
272   std::string new_raw_headers(raw_headers);
273   for (size_t i = 0; i < parsed_.size(); ++i) {
274     DCHECK(!parsed_[i].is_continuation());
275 
276     // Locate the start of the next header.
277     size_t k = i;
278     while (++k < parsed_.size() && parsed_[k].is_continuation()) {}
279     --k;
280 
281     std::string name(parsed_[i].name_begin, parsed_[i].name_end);
282     StringToLowerASCII(&name);
283     if (headers_to_remove.find(name) == headers_to_remove.end()) {
284       // It's ok to preserve this header in the final result.
285       new_raw_headers.append(parsed_[i].name_begin, parsed_[k].value_end);
286       new_raw_headers.push_back('\0');
287     }
288 
289     i = k;
290   }
291   new_raw_headers.push_back('\0');
292 
293   // Make this object hold the new data.
294   raw_headers_.clear();
295   parsed_.clear();
296   Parse(new_raw_headers);
297 }
298 
RemoveHeader(const std::string & name)299 void HttpResponseHeaders::RemoveHeader(const std::string& name) {
300   // Copy up to the null byte.  This just copies the status line.
301   std::string new_raw_headers(raw_headers_.c_str());
302   new_raw_headers.push_back('\0');
303 
304   std::string lowercase_name(name);
305   StringToLowerASCII(&lowercase_name);
306   HeaderSet to_remove;
307   to_remove.insert(lowercase_name);
308   MergeWithHeaders(new_raw_headers, to_remove);
309 }
310 
RemoveHeaderLine(const std::string & name,const std::string & value)311 void HttpResponseHeaders::RemoveHeaderLine(const std::string& name,
312                                            const std::string& value) {
313   std::string name_lowercase(name);
314   StringToLowerASCII(&name_lowercase);
315 
316   std::string new_raw_headers(GetStatusLine());
317   new_raw_headers.push_back('\0');
318 
319   new_raw_headers.reserve(raw_headers_.size());
320 
321   void* iter = NULL;
322   std::string old_header_name;
323   std::string old_header_value;
324   while (EnumerateHeaderLines(&iter, &old_header_name, &old_header_value)) {
325     std::string old_header_name_lowercase(name);
326     StringToLowerASCII(&old_header_name_lowercase);
327 
328     if (name_lowercase == old_header_name_lowercase &&
329         value == old_header_value)
330       continue;
331 
332     new_raw_headers.append(old_header_name);
333     new_raw_headers.push_back(':');
334     new_raw_headers.push_back(' ');
335     new_raw_headers.append(old_header_value);
336     new_raw_headers.push_back('\0');
337   }
338   new_raw_headers.push_back('\0');
339 
340   // Make this object hold the new data.
341   raw_headers_.clear();
342   parsed_.clear();
343   Parse(new_raw_headers);
344 }
345 
AddHeader(const std::string & header)346 void HttpResponseHeaders::AddHeader(const std::string& header) {
347   CheckDoesNotHaveEmbededNulls(header);
348   DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 2]);
349   DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 1]);
350   // Don't copy the last null.
351   std::string new_raw_headers(raw_headers_, 0, raw_headers_.size() - 1);
352   new_raw_headers.append(header);
353   new_raw_headers.push_back('\0');
354   new_raw_headers.push_back('\0');
355 
356   // Make this object hold the new data.
357   raw_headers_.clear();
358   parsed_.clear();
359   Parse(new_raw_headers);
360 }
361 
ReplaceStatusLine(const std::string & new_status)362 void HttpResponseHeaders::ReplaceStatusLine(const std::string& new_status) {
363   CheckDoesNotHaveEmbededNulls(new_status);
364   // Copy up to the null byte.  This just copies the status line.
365   std::string new_raw_headers(new_status);
366   new_raw_headers.push_back('\0');
367 
368   HeaderSet empty_to_remove;
369   MergeWithHeaders(new_raw_headers, empty_to_remove);
370 }
371 
UpdateWithNewRange(const HttpByteRange & byte_range,int64 resource_size,bool replace_status_line)372 void HttpResponseHeaders::UpdateWithNewRange(
373     const HttpByteRange& byte_range,
374     int64 resource_size,
375     bool replace_status_line) {
376   DCHECK(byte_range.IsValid());
377   DCHECK(byte_range.HasFirstBytePosition());
378   DCHECK(byte_range.HasLastBytePosition());
379 
380   const char kLengthHeader[] = "Content-Length";
381   const char kRangeHeader[] = "Content-Range";
382 
383   RemoveHeader(kLengthHeader);
384   RemoveHeader(kRangeHeader);
385 
386   int64 start = byte_range.first_byte_position();
387   int64 end = byte_range.last_byte_position();
388   int64 range_len = end - start + 1;
389 
390   if (replace_status_line)
391     ReplaceStatusLine("HTTP/1.1 206 Partial Content");
392 
393   AddHeader(base::StringPrintf("%s: bytes %" PRId64 "-%" PRId64 "/%" PRId64,
394                                kRangeHeader, start, end, resource_size));
395   AddHeader(base::StringPrintf("%s: %" PRId64, kLengthHeader, range_len));
396 }
397 
Parse(const std::string & raw_input)398 void HttpResponseHeaders::Parse(const std::string& raw_input) {
399   raw_headers_.reserve(raw_input.size());
400 
401   // ParseStatusLine adds a normalized status line to raw_headers_
402   std::string::const_iterator line_begin = raw_input.begin();
403   std::string::const_iterator line_end =
404       std::find(line_begin, raw_input.end(), '\0');
405   // has_headers = true, if there is any data following the status line.
406   // Used by ParseStatusLine() to decide if a HTTP/0.9 is really a HTTP/1.0.
407   bool has_headers = (line_end != raw_input.end() &&
408                       (line_end + 1) != raw_input.end() &&
409                       *(line_end + 1) != '\0');
410   ParseStatusLine(line_begin, line_end, has_headers);
411   raw_headers_.push_back('\0');  // Terminate status line with a null.
412 
413   if (line_end == raw_input.end()) {
414     raw_headers_.push_back('\0');  // Ensure the headers end with a double null.
415 
416     DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 2]);
417     DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 1]);
418     return;
419   }
420 
421   // Including a terminating null byte.
422   size_t status_line_len = raw_headers_.size();
423 
424   // Now, we add the rest of the raw headers to raw_headers_, and begin parsing
425   // it (to populate our parsed_ vector).
426   raw_headers_.append(line_end + 1, raw_input.end());
427 
428   // Ensure the headers end with a double null.
429   while (raw_headers_.size() < 2 ||
430          raw_headers_[raw_headers_.size() - 2] != '\0' ||
431          raw_headers_[raw_headers_.size() - 1] != '\0') {
432     raw_headers_.push_back('\0');
433   }
434 
435   // Adjust to point at the null byte following the status line
436   line_end = raw_headers_.begin() + status_line_len - 1;
437 
438   HttpUtil::HeadersIterator headers(line_end + 1, raw_headers_.end(),
439                                     std::string(1, '\0'));
440   while (headers.GetNext()) {
441     AddHeader(headers.name_begin(),
442               headers.name_end(),
443               headers.values_begin(),
444               headers.values_end());
445   }
446 
447   DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 2]);
448   DCHECK_EQ('\0', raw_headers_[raw_headers_.size() - 1]);
449 }
450 
451 // Append all of our headers to the final output string.
GetNormalizedHeaders(std::string * output) const452 void HttpResponseHeaders::GetNormalizedHeaders(std::string* output) const {
453   // copy up to the null byte.  this just copies the status line.
454   output->assign(raw_headers_.c_str());
455 
456   // headers may appear multiple times (not necessarily in succession) in the
457   // header data, so we build a map from header name to generated header lines.
458   // to preserve the order of the original headers, the actual values are kept
459   // in a separate list.  finally, the list of headers is flattened to form
460   // the normalized block of headers.
461   //
462   // NOTE: We take special care to preserve the whitespace around any commas
463   // that may occur in the original response headers.  Because our consumer may
464   // be a web app, we cannot be certain of the semantics of commas despite the
465   // fact that RFC 2616 says that they should be regarded as value separators.
466   //
467   typedef base::hash_map<std::string, size_t> HeadersMap;
468   HeadersMap headers_map;
469   HeadersMap::iterator iter = headers_map.end();
470 
471   std::vector<std::string> headers;
472 
473   for (size_t i = 0; i < parsed_.size(); ++i) {
474     DCHECK(!parsed_[i].is_continuation());
475 
476     std::string name(parsed_[i].name_begin, parsed_[i].name_end);
477     std::string lower_name = StringToLowerASCII(name);
478 
479     iter = headers_map.find(lower_name);
480     if (iter == headers_map.end()) {
481       iter = headers_map.insert(
482           HeadersMap::value_type(lower_name, headers.size())).first;
483       headers.push_back(name + ": ");
484     } else {
485       headers[iter->second].append(", ");
486     }
487 
488     std::string::const_iterator value_begin = parsed_[i].value_begin;
489     std::string::const_iterator value_end = parsed_[i].value_end;
490     while (++i < parsed_.size() && parsed_[i].is_continuation())
491       value_end = parsed_[i].value_end;
492     --i;
493 
494     headers[iter->second].append(value_begin, value_end);
495   }
496 
497   for (size_t i = 0; i < headers.size(); ++i) {
498     output->push_back('\n');
499     output->append(headers[i]);
500   }
501 
502   output->push_back('\n');
503 }
504 
GetNormalizedHeader(const std::string & name,std::string * value) const505 bool HttpResponseHeaders::GetNormalizedHeader(const std::string& name,
506                                               std::string* value) const {
507   // If you hit this assertion, please use EnumerateHeader instead!
508   DCHECK(!HttpUtil::IsNonCoalescingHeader(name));
509 
510   value->clear();
511 
512   bool found = false;
513   size_t i = 0;
514   while (i < parsed_.size()) {
515     i = FindHeader(i, name);
516     if (i == std::string::npos)
517       break;
518 
519     found = true;
520 
521     if (!value->empty())
522       value->append(", ");
523 
524     std::string::const_iterator value_begin = parsed_[i].value_begin;
525     std::string::const_iterator value_end = parsed_[i].value_end;
526     while (++i < parsed_.size() && parsed_[i].is_continuation())
527       value_end = parsed_[i].value_end;
528     value->append(value_begin, value_end);
529   }
530 
531   return found;
532 }
533 
GetStatusLine() const534 std::string HttpResponseHeaders::GetStatusLine() const {
535   // copy up to the null byte.
536   return std::string(raw_headers_.c_str());
537 }
538 
GetStatusText() const539 std::string HttpResponseHeaders::GetStatusText() const {
540   // GetStatusLine() is already normalized, so it has the format:
541   // <http_version> SP <response_code> SP <status_text>
542   std::string status_text = GetStatusLine();
543   std::string::const_iterator begin = status_text.begin();
544   std::string::const_iterator end = status_text.end();
545   for (int i = 0; i < 2; ++i)
546     begin = std::find(begin, end, ' ') + 1;
547   return std::string(begin, end);
548 }
549 
EnumerateHeaderLines(void ** iter,std::string * name,std::string * value) const550 bool HttpResponseHeaders::EnumerateHeaderLines(void** iter,
551                                                std::string* name,
552                                                std::string* value) const {
553   size_t i = reinterpret_cast<size_t>(*iter);
554   if (i == parsed_.size())
555     return false;
556 
557   DCHECK(!parsed_[i].is_continuation());
558 
559   name->assign(parsed_[i].name_begin, parsed_[i].name_end);
560 
561   std::string::const_iterator value_begin = parsed_[i].value_begin;
562   std::string::const_iterator value_end = parsed_[i].value_end;
563   while (++i < parsed_.size() && parsed_[i].is_continuation())
564     value_end = parsed_[i].value_end;
565 
566   value->assign(value_begin, value_end);
567 
568   *iter = reinterpret_cast<void*>(i);
569   return true;
570 }
571 
EnumerateHeader(void ** iter,const base::StringPiece & name,std::string * value) const572 bool HttpResponseHeaders::EnumerateHeader(void** iter,
573                                           const base::StringPiece& name,
574                                           std::string* value) const {
575   size_t i;
576   if (!iter || !*iter) {
577     i = FindHeader(0, name);
578   } else {
579     i = reinterpret_cast<size_t>(*iter);
580     if (i >= parsed_.size()) {
581       i = std::string::npos;
582     } else if (!parsed_[i].is_continuation()) {
583       i = FindHeader(i, name);
584     }
585   }
586 
587   if (i == std::string::npos) {
588     value->clear();
589     return false;
590   }
591 
592   if (iter)
593     *iter = reinterpret_cast<void*>(i + 1);
594   value->assign(parsed_[i].value_begin, parsed_[i].value_end);
595   return true;
596 }
597 
HasHeaderValue(const base::StringPiece & name,const base::StringPiece & value) const598 bool HttpResponseHeaders::HasHeaderValue(const base::StringPiece& name,
599                                          const base::StringPiece& value) const {
600   // The value has to be an exact match.  This is important since
601   // 'cache-control: no-cache' != 'cache-control: no-cache="foo"'
602   void* iter = NULL;
603   std::string temp;
604   while (EnumerateHeader(&iter, name, &temp)) {
605     if (value.size() == temp.size() &&
606         std::equal(temp.begin(), temp.end(), value.begin(),
607                    base::CaseInsensitiveCompare<char>()))
608       return true;
609   }
610   return false;
611 }
612 
HasHeader(const base::StringPiece & name) const613 bool HttpResponseHeaders::HasHeader(const base::StringPiece& name) const {
614   return FindHeader(0, name) != std::string::npos;
615 }
616 
HttpResponseHeaders()617 HttpResponseHeaders::HttpResponseHeaders() : response_code_(-1) {
618 }
619 
~HttpResponseHeaders()620 HttpResponseHeaders::~HttpResponseHeaders() {
621 }
622 
623 // Note: this implementation implicitly assumes that line_end points at a valid
624 // sentinel character (such as '\0').
625 // static
ParseVersion(std::string::const_iterator line_begin,std::string::const_iterator line_end)626 HttpVersion HttpResponseHeaders::ParseVersion(
627     std::string::const_iterator line_begin,
628     std::string::const_iterator line_end) {
629   std::string::const_iterator p = line_begin;
630 
631   // RFC2616 sec 3.1: HTTP-Version   = "HTTP" "/" 1*DIGIT "." 1*DIGIT
632   // TODO: (1*DIGIT apparently means one or more digits, but we only handle 1).
633   // TODO: handle leading zeros, which is allowed by the rfc1616 sec 3.1.
634 
635   if ((line_end - p < 4) || !LowerCaseEqualsASCII(p, p + 4, "http")) {
636     DVLOG(1) << "missing status line";
637     return HttpVersion();
638   }
639 
640   p += 4;
641 
642   if (p >= line_end || *p != '/') {
643     DVLOG(1) << "missing version";
644     return HttpVersion();
645   }
646 
647   std::string::const_iterator dot = std::find(p, line_end, '.');
648   if (dot == line_end) {
649     DVLOG(1) << "malformed version";
650     return HttpVersion();
651   }
652 
653   ++p;  // from / to first digit.
654   ++dot;  // from . to second digit.
655 
656   if (!(*p >= '0' && *p <= '9' && *dot >= '0' && *dot <= '9')) {
657     DVLOG(1) << "malformed version number";
658     return HttpVersion();
659   }
660 
661   uint16 major = *p - '0';
662   uint16 minor = *dot - '0';
663 
664   return HttpVersion(major, minor);
665 }
666 
667 // Note: this implementation implicitly assumes that line_end points at a valid
668 // sentinel character (such as '\0').
ParseStatusLine(std::string::const_iterator line_begin,std::string::const_iterator line_end,bool has_headers)669 void HttpResponseHeaders::ParseStatusLine(
670     std::string::const_iterator line_begin,
671     std::string::const_iterator line_end,
672     bool has_headers) {
673   // Extract the version number
674   parsed_http_version_ = ParseVersion(line_begin, line_end);
675 
676   // Clamp the version number to one of: {0.9, 1.0, 1.1}
677   if (parsed_http_version_ == HttpVersion(0, 9) && !has_headers) {
678     http_version_ = HttpVersion(0, 9);
679     raw_headers_ = "HTTP/0.9";
680   } else if (parsed_http_version_ >= HttpVersion(1, 1)) {
681     http_version_ = HttpVersion(1, 1);
682     raw_headers_ = "HTTP/1.1";
683   } else {
684     // Treat everything else like HTTP 1.0
685     http_version_ = HttpVersion(1, 0);
686     raw_headers_ = "HTTP/1.0";
687   }
688   if (parsed_http_version_ != http_version_) {
689     DVLOG(1) << "assuming HTTP/" << http_version_.major_value() << "."
690              << http_version_.minor_value();
691   }
692 
693   // TODO(eroman): this doesn't make sense if ParseVersion failed.
694   std::string::const_iterator p = std::find(line_begin, line_end, ' ');
695 
696   if (p == line_end) {
697     DVLOG(1) << "missing response status; assuming 200 OK";
698     raw_headers_.append(" 200 OK");
699     response_code_ = 200;
700     return;
701   }
702 
703   // Skip whitespace.
704   while (*p == ' ')
705     ++p;
706 
707   std::string::const_iterator code = p;
708   while (*p >= '0' && *p <= '9')
709     ++p;
710 
711   if (p == code) {
712     DVLOG(1) << "missing response status number; assuming 200";
713     raw_headers_.append(" 200 OK");
714     response_code_ = 200;
715     return;
716   }
717   raw_headers_.push_back(' ');
718   raw_headers_.append(code, p);
719   raw_headers_.push_back(' ');
720   base::StringToInt(StringPiece(code, p), &response_code_);
721 
722   // Skip whitespace.
723   while (*p == ' ')
724     ++p;
725 
726   // Trim trailing whitespace.
727   while (line_end > p && line_end[-1] == ' ')
728     --line_end;
729 
730   if (p == line_end) {
731     DVLOG(1) << "missing response status text; assuming OK";
732     // Not super critical what we put here. Just use "OK"
733     // even if it isn't descriptive of response_code_.
734     raw_headers_.append("OK");
735   } else {
736     raw_headers_.append(p, line_end);
737   }
738 }
739 
FindHeader(size_t from,const base::StringPiece & search) const740 size_t HttpResponseHeaders::FindHeader(size_t from,
741                                        const base::StringPiece& search) const {
742   for (size_t i = from; i < parsed_.size(); ++i) {
743     if (parsed_[i].is_continuation())
744       continue;
745     const std::string::const_iterator& name_begin = parsed_[i].name_begin;
746     const std::string::const_iterator& name_end = parsed_[i].name_end;
747     if (static_cast<size_t>(name_end - name_begin) == search.size() &&
748         std::equal(name_begin, name_end, search.begin(),
749                    base::CaseInsensitiveCompare<char>()))
750       return i;
751   }
752 
753   return std::string::npos;
754 }
755 
AddHeader(std::string::const_iterator name_begin,std::string::const_iterator name_end,std::string::const_iterator values_begin,std::string::const_iterator values_end)756 void HttpResponseHeaders::AddHeader(std::string::const_iterator name_begin,
757                                     std::string::const_iterator name_end,
758                                     std::string::const_iterator values_begin,
759                                     std::string::const_iterator values_end) {
760   // If the header can be coalesced, then we should split it up.
761   if (values_begin == values_end ||
762       HttpUtil::IsNonCoalescingHeader(name_begin, name_end)) {
763     AddToParsed(name_begin, name_end, values_begin, values_end);
764   } else {
765     HttpUtil::ValuesIterator it(values_begin, values_end, ',');
766     while (it.GetNext()) {
767       AddToParsed(name_begin, name_end, it.value_begin(), it.value_end());
768       // clobber these so that subsequent values are treated as continuations
769       name_begin = name_end = raw_headers_.end();
770     }
771   }
772 }
773 
AddToParsed(std::string::const_iterator name_begin,std::string::const_iterator name_end,std::string::const_iterator value_begin,std::string::const_iterator value_end)774 void HttpResponseHeaders::AddToParsed(std::string::const_iterator name_begin,
775                                       std::string::const_iterator name_end,
776                                       std::string::const_iterator value_begin,
777                                       std::string::const_iterator value_end) {
778   ParsedHeader header;
779   header.name_begin = name_begin;
780   header.name_end = name_end;
781   header.value_begin = value_begin;
782   header.value_end = value_end;
783   parsed_.push_back(header);
784 }
785 
AddNonCacheableHeaders(HeaderSet * result) const786 void HttpResponseHeaders::AddNonCacheableHeaders(HeaderSet* result) const {
787   // Add server specified transients.  Any 'cache-control: no-cache="foo,bar"'
788   // headers present in the response specify additional headers that we should
789   // not store in the cache.
790   const char kCacheControl[] = "cache-control";
791   const char kPrefix[] = "no-cache=\"";
792   const size_t kPrefixLen = sizeof(kPrefix) - 1;
793 
794   std::string value;
795   void* iter = NULL;
796   while (EnumerateHeader(&iter, kCacheControl, &value)) {
797     // If the value is smaller than the prefix and a terminal quote, skip
798     // it.
799     if (value.size() <= kPrefixLen ||
800         value.compare(0, kPrefixLen, kPrefix) != 0) {
801       continue;
802     }
803     // if it doesn't end with a quote, then treat as malformed
804     if (value[value.size()-1] != '\"')
805       continue;
806 
807     // process the value as a comma-separated list of items. Each
808     // item can be wrapped by linear white space.
809     std::string::const_iterator item = value.begin() + kPrefixLen;
810     std::string::const_iterator end = value.end() - 1;
811     while (item != end) {
812       // Find the comma to compute the length of the current item,
813       // and the position of the next one.
814       std::string::const_iterator item_next = std::find(item, end, ',');
815       std::string::const_iterator item_end = end;
816       if (item_next != end) {
817         // Skip over comma for next position.
818         item_end = item_next;
819         item_next++;
820       }
821       // trim off leading and trailing whitespace in this item.
822       HttpUtil::TrimLWS(&item, &item_end);
823 
824       // assuming the header is not empty, lowercase and insert into set
825       if (item_end > item) {
826         std::string name(&*item, item_end - item);
827         StringToLowerASCII(&name);
828         result->insert(name);
829       }
830 
831       // Continue to next item.
832       item = item_next;
833     }
834   }
835 }
836 
AddHopByHopHeaders(HeaderSet * result)837 void HttpResponseHeaders::AddHopByHopHeaders(HeaderSet* result) {
838   for (size_t i = 0; i < arraysize(kHopByHopResponseHeaders); ++i)
839     result->insert(std::string(kHopByHopResponseHeaders[i]));
840 }
841 
AddCookieHeaders(HeaderSet * result)842 void HttpResponseHeaders::AddCookieHeaders(HeaderSet* result) {
843   for (size_t i = 0; i < arraysize(kCookieResponseHeaders); ++i)
844     result->insert(std::string(kCookieResponseHeaders[i]));
845 }
846 
AddChallengeHeaders(HeaderSet * result)847 void HttpResponseHeaders::AddChallengeHeaders(HeaderSet* result) {
848   for (size_t i = 0; i < arraysize(kChallengeResponseHeaders); ++i)
849     result->insert(std::string(kChallengeResponseHeaders[i]));
850 }
851 
AddHopContentRangeHeaders(HeaderSet * result)852 void HttpResponseHeaders::AddHopContentRangeHeaders(HeaderSet* result) {
853   result->insert(kContentRange);
854 }
855 
AddSecurityStateHeaders(HeaderSet * result)856 void HttpResponseHeaders::AddSecurityStateHeaders(HeaderSet* result) {
857   for (size_t i = 0; i < arraysize(kSecurityStateHeaders); ++i)
858     result->insert(std::string(kSecurityStateHeaders[i]));
859 }
860 
GetMimeTypeAndCharset(std::string * mime_type,std::string * charset) const861 void HttpResponseHeaders::GetMimeTypeAndCharset(std::string* mime_type,
862                                                 std::string* charset) const {
863   mime_type->clear();
864   charset->clear();
865 
866   std::string name = "content-type";
867   std::string value;
868 
869   bool had_charset = false;
870 
871   void* iter = NULL;
872   while (EnumerateHeader(&iter, name, &value))
873     HttpUtil::ParseContentType(value, mime_type, charset, &had_charset, NULL);
874 }
875 
GetMimeType(std::string * mime_type) const876 bool HttpResponseHeaders::GetMimeType(std::string* mime_type) const {
877   std::string unused;
878   GetMimeTypeAndCharset(mime_type, &unused);
879   return !mime_type->empty();
880 }
881 
GetCharset(std::string * charset) const882 bool HttpResponseHeaders::GetCharset(std::string* charset) const {
883   std::string unused;
884   GetMimeTypeAndCharset(&unused, charset);
885   return !charset->empty();
886 }
887 
IsRedirect(std::string * location) const888 bool HttpResponseHeaders::IsRedirect(std::string* location) const {
889   if (!IsRedirectResponseCode(response_code_))
890     return false;
891 
892   // If we lack a Location header, then we can't treat this as a redirect.
893   // We assume that the first non-empty location value is the target URL that
894   // we want to follow.  TODO(darin): Is this consistent with other browsers?
895   size_t i = std::string::npos;
896   do {
897     i = FindHeader(++i, "location");
898     if (i == std::string::npos)
899       return false;
900     // If the location value is empty, then it doesn't count.
901   } while (parsed_[i].value_begin == parsed_[i].value_end);
902 
903   if (location) {
904     // Escape any non-ASCII characters to preserve them.  The server should
905     // only be returning ASCII here, but for compat we need to do this.
906     *location = EscapeNonASCII(
907         std::string(parsed_[i].value_begin, parsed_[i].value_end));
908   }
909 
910   return true;
911 }
912 
913 // static
IsRedirectResponseCode(int response_code)914 bool HttpResponseHeaders::IsRedirectResponseCode(int response_code) {
915   // Users probably want to see 300 (multiple choice) pages, so we don't count
916   // them as redirects that need to be followed.
917   return (response_code == 301 ||
918           response_code == 302 ||
919           response_code == 303 ||
920           response_code == 307 ||
921           response_code == 308);
922 }
923 
924 // From RFC 2616 section 13.2.4:
925 //
926 // The calculation to determine if a response has expired is quite simple:
927 //
928 //   response_is_fresh = (freshness_lifetime > current_age)
929 //
930 // Of course, there are other factors that can force a response to always be
931 // validated or re-fetched.
932 //
RequiresValidation(const Time & request_time,const Time & response_time,const Time & current_time) const933 bool HttpResponseHeaders::RequiresValidation(const Time& request_time,
934                                              const Time& response_time,
935                                              const Time& current_time) const {
936   TimeDelta lifetime =
937       GetFreshnessLifetime(response_time);
938   if (lifetime == TimeDelta())
939     return true;
940 
941   return lifetime <= GetCurrentAge(request_time, response_time, current_time);
942 }
943 
944 // From RFC 2616 section 13.2.4:
945 //
946 // The max-age directive takes priority over Expires, so if max-age is present
947 // in a response, the calculation is simply:
948 //
949 //   freshness_lifetime = max_age_value
950 //
951 // Otherwise, if Expires is present in the response, the calculation is:
952 //
953 //   freshness_lifetime = expires_value - date_value
954 //
955 // Note that neither of these calculations is vulnerable to clock skew, since
956 // all of the information comes from the origin server.
957 //
958 // Also, if the response does have a Last-Modified time, the heuristic
959 // expiration value SHOULD be no more than some fraction of the interval since
960 // that time. A typical setting of this fraction might be 10%:
961 //
962 //   freshness_lifetime = (date_value - last_modified_value) * 0.10
963 //
GetFreshnessLifetime(const Time & response_time) const964 TimeDelta HttpResponseHeaders::GetFreshnessLifetime(
965     const Time& response_time) const {
966   // Check for headers that force a response to never be fresh.  For backwards
967   // compat, we treat "Pragma: no-cache" as a synonym for "Cache-Control:
968   // no-cache" even though RFC 2616 does not specify it.
969   if (HasHeaderValue("cache-control", "no-cache") ||
970       HasHeaderValue("cache-control", "no-store") ||
971       HasHeaderValue("pragma", "no-cache") ||
972       HasHeaderValue("vary", "*"))  // see RFC 2616 section 13.6
973     return TimeDelta();  // not fresh
974 
975   // NOTE: "Cache-Control: max-age" overrides Expires, so we only check the
976   // Expires header after checking for max-age in GetFreshnessLifetime.  This
977   // is important since "Expires: <date in the past>" means not fresh, but
978   // it should not trump a max-age value.
979 
980   TimeDelta max_age_value;
981   if (GetMaxAgeValue(&max_age_value))
982     return max_age_value;
983 
984   // If there is no Date header, then assume that the server response was
985   // generated at the time when we received the response.
986   Time date_value;
987   if (!GetDateValue(&date_value))
988     date_value = response_time;
989 
990   Time expires_value;
991   if (GetExpiresValue(&expires_value)) {
992     // The expires value can be a date in the past!
993     if (expires_value > date_value)
994       return expires_value - date_value;
995 
996     return TimeDelta();  // not fresh
997   }
998 
999   // From RFC 2616 section 13.4:
1000   //
1001   //   A response received with a status code of 200, 203, 206, 300, 301 or 410
1002   //   MAY be stored by a cache and used in reply to a subsequent request,
1003   //   subject to the expiration mechanism, unless a cache-control directive
1004   //   prohibits caching.
1005   //   ...
1006   //   A response received with any other status code (e.g. status codes 302
1007   //   and 307) MUST NOT be returned in a reply to a subsequent request unless
1008   //   there are cache-control directives or another header(s) that explicitly
1009   //   allow it.
1010   //
1011   // From RFC 2616 section 14.9.4:
1012   //
1013   //   When the must-revalidate directive is present in a response received by
1014   //   a cache, that cache MUST NOT use the entry after it becomes stale to
1015   //   respond to a subsequent request without first revalidating it with the
1016   //   origin server. (I.e., the cache MUST do an end-to-end revalidation every
1017   //   time, if, based solely on the origin server's Expires or max-age value,
1018   //   the cached response is stale.)
1019   //
1020   // https://datatracker.ietf.org/doc/draft-reschke-http-status-308/ is an
1021   // experimental RFC that adds 308 permanent redirect as well, for which "any
1022   // future references ... SHOULD use one of the returned URIs."
1023   if ((response_code_ == 200 || response_code_ == 203 ||
1024        response_code_ == 206) &&
1025       !HasHeaderValue("cache-control", "must-revalidate")) {
1026     // TODO(darin): Implement a smarter heuristic.
1027     Time last_modified_value;
1028     if (GetLastModifiedValue(&last_modified_value)) {
1029       // The last-modified value can be a date in the past!
1030       if (last_modified_value <= date_value)
1031         return (date_value - last_modified_value) / 10;
1032     }
1033   }
1034 
1035   // These responses are implicitly fresh (unless otherwise overruled):
1036   if (response_code_ == 300 || response_code_ == 301 || response_code_ == 308 ||
1037       response_code_ == 410) {
1038     return TimeDelta::Max();
1039   }
1040 
1041   return TimeDelta();  // not fresh
1042 }
1043 
1044 // From RFC 2616 section 13.2.3:
1045 //
1046 // Summary of age calculation algorithm, when a cache receives a response:
1047 //
1048 //   /*
1049 //    * age_value
1050 //    *      is the value of Age: header received by the cache with
1051 //    *              this response.
1052 //    * date_value
1053 //    *      is the value of the origin server's Date: header
1054 //    * request_time
1055 //    *      is the (local) time when the cache made the request
1056 //    *              that resulted in this cached response
1057 //    * response_time
1058 //    *      is the (local) time when the cache received the
1059 //    *              response
1060 //    * now
1061 //    *      is the current (local) time
1062 //    */
1063 //   apparent_age = max(0, response_time - date_value);
1064 //   corrected_received_age = max(apparent_age, age_value);
1065 //   response_delay = response_time - request_time;
1066 //   corrected_initial_age = corrected_received_age + response_delay;
1067 //   resident_time = now - response_time;
1068 //   current_age   = corrected_initial_age + resident_time;
1069 //
GetCurrentAge(const Time & request_time,const Time & response_time,const Time & current_time) const1070 TimeDelta HttpResponseHeaders::GetCurrentAge(const Time& request_time,
1071                                              const Time& response_time,
1072                                              const Time& current_time) const {
1073   // If there is no Date header, then assume that the server response was
1074   // generated at the time when we received the response.
1075   Time date_value;
1076   if (!GetDateValue(&date_value))
1077     date_value = response_time;
1078 
1079   // If there is no Age header, then assume age is zero.  GetAgeValue does not
1080   // modify its out param if the value does not exist.
1081   TimeDelta age_value;
1082   GetAgeValue(&age_value);
1083 
1084   TimeDelta apparent_age = std::max(TimeDelta(), response_time - date_value);
1085   TimeDelta corrected_received_age = std::max(apparent_age, age_value);
1086   TimeDelta response_delay = response_time - request_time;
1087   TimeDelta corrected_initial_age = corrected_received_age + response_delay;
1088   TimeDelta resident_time = current_time - response_time;
1089   TimeDelta current_age = corrected_initial_age + resident_time;
1090 
1091   return current_age;
1092 }
1093 
GetMaxAgeValue(TimeDelta * result) const1094 bool HttpResponseHeaders::GetMaxAgeValue(TimeDelta* result) const {
1095   std::string name = "cache-control";
1096   std::string value;
1097 
1098   const char kMaxAgePrefix[] = "max-age=";
1099   const size_t kMaxAgePrefixLen = arraysize(kMaxAgePrefix) - 1;
1100 
1101   void* iter = NULL;
1102   while (EnumerateHeader(&iter, name, &value)) {
1103     if (value.size() > kMaxAgePrefixLen) {
1104       if (LowerCaseEqualsASCII(value.begin(),
1105                                value.begin() + kMaxAgePrefixLen,
1106                                kMaxAgePrefix)) {
1107         int64 seconds;
1108         base::StringToInt64(StringPiece(value.begin() + kMaxAgePrefixLen,
1109                                         value.end()),
1110                             &seconds);
1111         *result = TimeDelta::FromSeconds(seconds);
1112         return true;
1113       }
1114     }
1115   }
1116 
1117   return false;
1118 }
1119 
GetAgeValue(TimeDelta * result) const1120 bool HttpResponseHeaders::GetAgeValue(TimeDelta* result) const {
1121   std::string value;
1122   if (!EnumerateHeader(NULL, "Age", &value))
1123     return false;
1124 
1125   int64 seconds;
1126   base::StringToInt64(value, &seconds);
1127   *result = TimeDelta::FromSeconds(seconds);
1128   return true;
1129 }
1130 
GetDateValue(Time * result) const1131 bool HttpResponseHeaders::GetDateValue(Time* result) const {
1132   return GetTimeValuedHeader("Date", result);
1133 }
1134 
GetLastModifiedValue(Time * result) const1135 bool HttpResponseHeaders::GetLastModifiedValue(Time* result) const {
1136   return GetTimeValuedHeader("Last-Modified", result);
1137 }
1138 
GetExpiresValue(Time * result) const1139 bool HttpResponseHeaders::GetExpiresValue(Time* result) const {
1140   return GetTimeValuedHeader("Expires", result);
1141 }
1142 
GetTimeValuedHeader(const std::string & name,Time * result) const1143 bool HttpResponseHeaders::GetTimeValuedHeader(const std::string& name,
1144                                               Time* result) const {
1145   std::string value;
1146   if (!EnumerateHeader(NULL, name, &value))
1147     return false;
1148 
1149   // When parsing HTTP dates it's beneficial to default to GMT because:
1150   // 1. RFC2616 3.3.1 says times should always be specified in GMT
1151   // 2. Only counter-example incorrectly appended "UTC" (crbug.com/153759)
1152   // 3. When adjusting cookie expiration times for clock skew
1153   //    (crbug.com/135131) this better matches our cookie expiration
1154   //    time parser which ignores timezone specifiers and assumes GMT.
1155   // 4. This is exactly what Firefox does.
1156   // TODO(pauljensen): The ideal solution would be to return false if the
1157   // timezone could not be understood so as to avoid makeing other calculations
1158   // based on an incorrect time.  This would require modifying the time
1159   // library or duplicating the code. (http://crbug.com/158327)
1160   return Time::FromUTCString(value.c_str(), result);
1161 }
1162 
IsKeepAlive() const1163 bool HttpResponseHeaders::IsKeepAlive() const {
1164   if (http_version_ < HttpVersion(1, 0))
1165     return false;
1166 
1167   // NOTE: It is perhaps risky to assume that a Proxy-Connection header is
1168   // meaningful when we don't know that this response was from a proxy, but
1169   // Mozilla also does this, so we'll do the same.
1170   std::string connection_val;
1171   if (!EnumerateHeader(NULL, "connection", &connection_val))
1172     EnumerateHeader(NULL, "proxy-connection", &connection_val);
1173 
1174   bool keep_alive;
1175 
1176   if (http_version_ == HttpVersion(1, 0)) {
1177     // HTTP/1.0 responses default to NOT keep-alive
1178     keep_alive = LowerCaseEqualsASCII(connection_val, "keep-alive");
1179   } else {
1180     // HTTP/1.1 responses default to keep-alive
1181     keep_alive = !LowerCaseEqualsASCII(connection_val, "close");
1182   }
1183 
1184   return keep_alive;
1185 }
1186 
HasStrongValidators() const1187 bool HttpResponseHeaders::HasStrongValidators() const {
1188   std::string etag_header;
1189   EnumerateHeader(NULL, "etag", &etag_header);
1190   std::string last_modified_header;
1191   EnumerateHeader(NULL, "Last-Modified", &last_modified_header);
1192   std::string date_header;
1193   EnumerateHeader(NULL, "Date", &date_header);
1194   return HttpUtil::HasStrongValidators(GetHttpVersion(),
1195                                        etag_header,
1196                                        last_modified_header,
1197                                        date_header);
1198 }
1199 
1200 // From RFC 2616:
1201 // Content-Length = "Content-Length" ":" 1*DIGIT
GetContentLength() const1202 int64 HttpResponseHeaders::GetContentLength() const {
1203   return GetInt64HeaderValue("content-length");
1204 }
1205 
GetInt64HeaderValue(const std::string & header) const1206 int64 HttpResponseHeaders::GetInt64HeaderValue(
1207     const std::string& header) const {
1208   void* iter = NULL;
1209   std::string content_length_val;
1210   if (!EnumerateHeader(&iter, header, &content_length_val))
1211     return -1;
1212 
1213   if (content_length_val.empty())
1214     return -1;
1215 
1216   if (content_length_val[0] == '+')
1217     return -1;
1218 
1219   int64 result;
1220   bool ok = base::StringToInt64(content_length_val, &result);
1221   if (!ok || result < 0)
1222     return -1;
1223 
1224   return result;
1225 }
1226 
1227 // From RFC 2616 14.16:
1228 // content-range-spec =
1229 //     bytes-unit SP byte-range-resp-spec "/" ( instance-length | "*" )
1230 // byte-range-resp-spec = (first-byte-pos "-" last-byte-pos) | "*"
1231 // instance-length = 1*DIGIT
1232 // bytes-unit = "bytes"
GetContentRange(int64 * first_byte_position,int64 * last_byte_position,int64 * instance_length) const1233 bool HttpResponseHeaders::GetContentRange(int64* first_byte_position,
1234                                           int64* last_byte_position,
1235                                           int64* instance_length) const {
1236   void* iter = NULL;
1237   std::string content_range_spec;
1238   *first_byte_position = *last_byte_position = *instance_length = -1;
1239   if (!EnumerateHeader(&iter, kContentRange, &content_range_spec))
1240     return false;
1241 
1242   // If the header value is empty, we have an invalid header.
1243   if (content_range_spec.empty())
1244     return false;
1245 
1246   size_t space_position = content_range_spec.find(' ');
1247   if (space_position == std::string::npos)
1248     return false;
1249 
1250   // Invalid header if it doesn't contain "bytes-unit".
1251   std::string::const_iterator content_range_spec_begin =
1252       content_range_spec.begin();
1253   std::string::const_iterator content_range_spec_end =
1254       content_range_spec.begin() + space_position;
1255   HttpUtil::TrimLWS(&content_range_spec_begin, &content_range_spec_end);
1256   if (!LowerCaseEqualsASCII(content_range_spec_begin,
1257                             content_range_spec_end,
1258                             "bytes")) {
1259     return false;
1260   }
1261 
1262   size_t slash_position = content_range_spec.find('/', space_position + 1);
1263   if (slash_position == std::string::npos)
1264     return false;
1265 
1266   // Obtain the part behind the space and before slash.
1267   std::string::const_iterator byte_range_resp_spec_begin =
1268       content_range_spec.begin() + space_position + 1;
1269   std::string::const_iterator byte_range_resp_spec_end =
1270       content_range_spec.begin() + slash_position;
1271   HttpUtil::TrimLWS(&byte_range_resp_spec_begin, &byte_range_resp_spec_end);
1272 
1273   // Parse the byte-range-resp-spec part.
1274   std::string byte_range_resp_spec(byte_range_resp_spec_begin,
1275                                    byte_range_resp_spec_end);
1276   // If byte-range-resp-spec != "*".
1277   if (!LowerCaseEqualsASCII(byte_range_resp_spec, "*")) {
1278     size_t minus_position = byte_range_resp_spec.find('-');
1279     if (minus_position != std::string::npos) {
1280       // Obtain first-byte-pos.
1281       std::string::const_iterator first_byte_pos_begin =
1282           byte_range_resp_spec.begin();
1283       std::string::const_iterator first_byte_pos_end =
1284           byte_range_resp_spec.begin() + minus_position;
1285       HttpUtil::TrimLWS(&first_byte_pos_begin, &first_byte_pos_end);
1286 
1287       bool ok = base::StringToInt64(StringPiece(first_byte_pos_begin,
1288                                                 first_byte_pos_end),
1289                                     first_byte_position);
1290 
1291       // Obtain last-byte-pos.
1292       std::string::const_iterator last_byte_pos_begin =
1293           byte_range_resp_spec.begin() + minus_position + 1;
1294       std::string::const_iterator last_byte_pos_end =
1295           byte_range_resp_spec.end();
1296       HttpUtil::TrimLWS(&last_byte_pos_begin, &last_byte_pos_end);
1297 
1298       ok &= base::StringToInt64(StringPiece(last_byte_pos_begin,
1299                                             last_byte_pos_end),
1300                                 last_byte_position);
1301       if (!ok) {
1302         *first_byte_position = *last_byte_position = -1;
1303         return false;
1304       }
1305       if (*first_byte_position < 0 || *last_byte_position < 0 ||
1306           *first_byte_position > *last_byte_position)
1307         return false;
1308     } else {
1309       return false;
1310     }
1311   }
1312 
1313   // Parse the instance-length part.
1314   // If instance-length == "*".
1315   std::string::const_iterator instance_length_begin =
1316       content_range_spec.begin() + slash_position + 1;
1317   std::string::const_iterator instance_length_end =
1318       content_range_spec.end();
1319   HttpUtil::TrimLWS(&instance_length_begin, &instance_length_end);
1320 
1321   if (LowerCaseEqualsASCII(instance_length_begin, instance_length_end, "*")) {
1322     return false;
1323   } else if (!base::StringToInt64(StringPiece(instance_length_begin,
1324                                               instance_length_end),
1325                                   instance_length)) {
1326     *instance_length = -1;
1327     return false;
1328   }
1329 
1330   // We have all the values; let's verify that they make sense for a 206
1331   // response.
1332   if (*first_byte_position < 0 || *last_byte_position < 0 ||
1333       *instance_length < 0 || *instance_length - 1 < *last_byte_position)
1334     return false;
1335 
1336   return true;
1337 }
1338 
NetLogCallback(NetLog::LogLevel log_level) const1339 base::Value* HttpResponseHeaders::NetLogCallback(
1340     NetLog::LogLevel log_level) const {
1341   base::DictionaryValue* dict = new base::DictionaryValue();
1342   base::ListValue* headers = new base::ListValue();
1343   headers->Append(new base::StringValue(GetStatusLine()));
1344   void* iterator = NULL;
1345   std::string name;
1346   std::string value;
1347   while (EnumerateHeaderLines(&iterator, &name, &value)) {
1348     std::string log_value = ElideHeaderValueForNetLog(log_level, name, value);
1349     headers->Append(
1350       new base::StringValue(
1351           base::StringPrintf("%s: %s", name.c_str(), log_value.c_str())));
1352   }
1353   dict->Set("headers", headers);
1354   return dict;
1355 }
1356 
1357 // static
FromNetLogParam(const base::Value * event_param,scoped_refptr<HttpResponseHeaders> * http_response_headers)1358 bool HttpResponseHeaders::FromNetLogParam(
1359     const base::Value* event_param,
1360     scoped_refptr<HttpResponseHeaders>* http_response_headers) {
1361   *http_response_headers = NULL;
1362 
1363   const base::DictionaryValue* dict = NULL;
1364   const base::ListValue* header_list = NULL;
1365 
1366   if (!event_param ||
1367       !event_param->GetAsDictionary(&dict) ||
1368       !dict->GetList("headers", &header_list)) {
1369     return false;
1370   }
1371 
1372   std::string raw_headers;
1373   for (base::ListValue::const_iterator it = header_list->begin();
1374        it != header_list->end();
1375        ++it) {
1376     std::string header_line;
1377     if (!(*it)->GetAsString(&header_line))
1378       return false;
1379 
1380     raw_headers.append(header_line);
1381     raw_headers.push_back('\0');
1382   }
1383   raw_headers.push_back('\0');
1384   *http_response_headers = new HttpResponseHeaders(raw_headers);
1385   return true;
1386 }
1387 
IsChunkEncoded() const1388 bool HttpResponseHeaders::IsChunkEncoded() const {
1389   // Ignore spurious chunked responses from HTTP/1.0 servers and proxies.
1390   return GetHttpVersion() >= HttpVersion(1, 1) &&
1391       HasHeaderValue("Transfer-Encoding", "chunked");
1392 }
1393 
1394 }  // namespace net
1395