1 /* 2 * libjingle 3 * Copyright 2004--2005, Google Inc. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright notice, 9 * this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright notice, 11 * this list of conditions and the following disclaimer in the documentation 12 * and/or other materials provided with the distribution. 13 * 3. The name of the author may not be used to endorse or promote products 14 * derived from this software without specific prior written permission. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED 17 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 18 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO 19 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 20 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 22 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 23 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 24 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 25 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 // Copyright 2005 Google Inc. All Rights Reserved. 29 // 30 31 32 #ifdef WIN32 33 #include "talk/base/win32.h" 34 #else // !WIN32 35 #define SEC_E_CERT_EXPIRED (-2146893016) 36 #endif // !WIN32 37 38 #include "talk/base/common.h" 39 #include "talk/base/httpbase.h" 40 #include "talk/base/logging.h" 41 #include "talk/base/socket.h" 42 #include "talk/base/stringutils.h" 43 #include "talk/base/thread.h" 44 45 namespace talk_base { 46 47 ////////////////////////////////////////////////////////////////////// 48 // Helpers 49 ////////////////////////////////////////////////////////////////////// 50 51 bool MatchHeader(const char* str, size_t len, HttpHeader header) { 52 const char* const header_str = ToString(header); 53 const size_t header_len = strlen(header_str); 54 return (len == header_len) && (_strnicmp(str, header_str, header_len) == 0); 55 } 56 57 enum { 58 MSG_READ 59 }; 60 61 ////////////////////////////////////////////////////////////////////// 62 // HttpParser 63 ////////////////////////////////////////////////////////////////////// 64 65 HttpParser::HttpParser() { 66 reset(); 67 } 68 69 HttpParser::~HttpParser() { 70 } 71 72 void 73 HttpParser::reset() { 74 state_ = ST_LEADER; 75 chunked_ = false; 76 data_size_ = SIZE_UNKNOWN; 77 } 78 79 HttpParser::ProcessResult 80 HttpParser::Process(const char* buffer, size_t len, size_t* processed, 81 HttpError* error) { 82 *processed = 0; 83 *error = HE_NONE; 84 85 if (state_ >= ST_COMPLETE) { 86 ASSERT(false); 87 return PR_COMPLETE; 88 } 89 90 while (true) { 91 if (state_ < ST_DATA) { 92 size_t pos = *processed; 93 while ((pos < len) && (buffer[pos] != '\n')) { 94 pos += 1; 95 } 96 if (pos >= len) { 97 break; // don't have a full header 98 } 99 const char* line = buffer + *processed; 100 size_t len = (pos - *processed); 101 *processed = pos + 1; 102 while ((len > 0) && isspace(static_cast<unsigned char>(line[len-1]))) { 103 len -= 1; 104 } 105 ProcessResult result = ProcessLine(line, len, error); 106 LOG(LS_VERBOSE) << "Processed line, result=" << result; 107 108 if (PR_CONTINUE != result) { 109 return result; 110 } 111 } else if (data_size_ == 0) { 112 if (chunked_) { 113 state_ = ST_CHUNKTERM; 114 } else { 115 return PR_COMPLETE; 116 } 117 } else { 118 size_t available = len - *processed; 119 if (available <= 0) { 120 break; // no more data 121 } 122 if ((data_size_ != SIZE_UNKNOWN) && (available > data_size_)) { 123 available = data_size_; 124 } 125 size_t read = 0; 126 ProcessResult result = ProcessData(buffer + *processed, available, read, 127 error); 128 LOG(LS_VERBOSE) << "Processed data, result: " << result << " read: " 129 << read << " err: " << error; 130 131 if (PR_CONTINUE != result) { 132 return result; 133 } 134 *processed += read; 135 if (data_size_ != SIZE_UNKNOWN) { 136 data_size_ -= read; 137 } 138 } 139 } 140 141 return PR_CONTINUE; 142 } 143 144 HttpParser::ProcessResult 145 HttpParser::ProcessLine(const char* line, size_t len, HttpError* error) { 146 LOG_F(LS_VERBOSE) << " state: " << state_ << " line: " 147 << std::string(line, len) << " len: " << len << " err: " 148 << error; 149 150 switch (state_) { 151 case ST_LEADER: 152 state_ = ST_HEADERS; 153 return ProcessLeader(line, len, error); 154 155 case ST_HEADERS: 156 if (len > 0) { 157 const char* value = strchrn(line, len, ':'); 158 if (!value) { 159 *error = HE_PROTOCOL; 160 return PR_COMPLETE; 161 } 162 size_t nlen = (value - line); 163 const char* eol = line + len; 164 do { 165 value += 1; 166 } while ((value < eol) && isspace(static_cast<unsigned char>(*value))); 167 size_t vlen = eol - value; 168 if (MatchHeader(line, nlen, HH_CONTENT_LENGTH)) { 169 // sscanf isn't safe with strings that aren't null-terminated, and there 170 // is no guarantee that |value| is. 171 // Create a local copy that is null-terminated. 172 std::string value_str(value, vlen); 173 unsigned int temp_size; 174 if (sscanf(value_str.c_str(), "%u", &temp_size) != 1) { 175 *error = HE_PROTOCOL; 176 return PR_COMPLETE; 177 } 178 data_size_ = static_cast<size_t>(temp_size); 179 } else if (MatchHeader(line, nlen, HH_TRANSFER_ENCODING)) { 180 if ((vlen == 7) && (_strnicmp(value, "chunked", 7) == 0)) { 181 chunked_ = true; 182 } else if ((vlen == 8) && (_strnicmp(value, "identity", 8) == 0)) { 183 chunked_ = false; 184 } else { 185 *error = HE_PROTOCOL; 186 return PR_COMPLETE; 187 } 188 } 189 return ProcessHeader(line, nlen, value, vlen, error); 190 } else { 191 state_ = chunked_ ? ST_CHUNKSIZE : ST_DATA; 192 return ProcessHeaderComplete(chunked_, data_size_, error); 193 } 194 break; 195 196 case ST_CHUNKSIZE: 197 if (len > 0) { 198 char* ptr = NULL; 199 data_size_ = strtoul(line, &ptr, 16); 200 if (ptr != line + len) { 201 *error = HE_PROTOCOL; 202 return PR_COMPLETE; 203 } 204 state_ = (data_size_ == 0) ? ST_TRAILERS : ST_DATA; 205 } else { 206 *error = HE_PROTOCOL; 207 return PR_COMPLETE; 208 } 209 break; 210 211 case ST_CHUNKTERM: 212 if (len > 0) { 213 *error = HE_PROTOCOL; 214 return PR_COMPLETE; 215 } else { 216 state_ = chunked_ ? ST_CHUNKSIZE : ST_DATA; 217 } 218 break; 219 220 case ST_TRAILERS: 221 if (len == 0) { 222 return PR_COMPLETE; 223 } 224 // *error = onHttpRecvTrailer(); 225 break; 226 227 default: 228 ASSERT(false); 229 break; 230 } 231 232 return PR_CONTINUE; 233 } 234 235 bool 236 HttpParser::is_valid_end_of_input() const { 237 return (state_ == ST_DATA) && (data_size_ == SIZE_UNKNOWN); 238 } 239 240 void 241 HttpParser::complete(HttpError error) { 242 if (state_ < ST_COMPLETE) { 243 state_ = ST_COMPLETE; 244 OnComplete(error); 245 } 246 } 247 248 ////////////////////////////////////////////////////////////////////// 249 // HttpBase::DocumentStream 250 ////////////////////////////////////////////////////////////////////// 251 252 class BlockingMemoryStream : public ExternalMemoryStream { 253 public: 254 BlockingMemoryStream(char* buffer, size_t size) 255 : ExternalMemoryStream(buffer, size) { } 256 257 virtual StreamResult DoReserve(size_t size, int* error) { 258 return (buffer_length_ >= size) ? SR_SUCCESS : SR_BLOCK; 259 } 260 }; 261 262 class HttpBase::DocumentStream : public StreamInterface { 263 public: 264 DocumentStream(HttpBase* base) : base_(base), error_(HE_DEFAULT) { } 265 266 virtual StreamState GetState() const { 267 if (NULL == base_) 268 return SS_CLOSED; 269 if (HM_RECV == base_->mode_) 270 return SS_OPEN; 271 return SS_OPENING; 272 } 273 274 virtual StreamResult Read(void* buffer, size_t buffer_len, 275 size_t* read, int* error) { 276 if (!base_) { 277 if (error) *error = error_; 278 return (HE_NONE == error_) ? SR_EOS : SR_ERROR; 279 } 280 281 if (HM_RECV != base_->mode_) { 282 return SR_BLOCK; 283 } 284 285 // DoReceiveLoop writes http document data to the StreamInterface* document 286 // member of HttpData. In this case, we want this data to be written 287 // directly to our buffer. To accomplish this, we wrap our buffer with a 288 // StreamInterface, and replace the existing document with our wrapper. 289 // When the method returns, we restore the old document. Ideally, we would 290 // pass our StreamInterface* to DoReceiveLoop, but due to the callbacks 291 // of HttpParser, we would still need to store the pointer temporarily. 292 scoped_ptr<StreamInterface> 293 stream(new BlockingMemoryStream(reinterpret_cast<char*>(buffer), 294 buffer_len)); 295 296 // Replace the existing document with our wrapped buffer. 297 base_->data_->document.swap(stream); 298 299 // Pump the I/O loop. DoReceiveLoop is guaranteed not to attempt to 300 // complete the I/O process, which means that our wrapper is not in danger 301 // of being deleted. To ensure this, DoReceiveLoop returns true when it 302 // wants complete to be called. We make sure to uninstall our wrapper 303 // before calling complete(). 304 HttpError http_error; 305 bool complete = base_->DoReceiveLoop(&http_error); 306 307 // Reinstall the original output document. 308 base_->data_->document.swap(stream); 309 310 // If we reach the end of the receive stream, we disconnect our stream 311 // adapter from the HttpBase, and further calls to read will either return 312 // EOS or ERROR, appropriately. Finally, we call complete(). 313 StreamResult result = SR_BLOCK; 314 if (complete) { 315 HttpBase* base = Disconnect(http_error); 316 if (error) *error = error_; 317 result = (HE_NONE == error_) ? SR_EOS : SR_ERROR; 318 base->complete(http_error); 319 } 320 321 // Even if we are complete, if some data was read we must return SUCCESS. 322 // Future Reads will return EOS or ERROR based on the error_ variable. 323 size_t position; 324 stream->GetPosition(&position); 325 if (position > 0) { 326 if (read) *read = position; 327 result = SR_SUCCESS; 328 } 329 return result; 330 } 331 332 virtual StreamResult Write(const void* data, size_t data_len, 333 size_t* written, int* error) { 334 if (error) *error = -1; 335 return SR_ERROR; 336 } 337 338 virtual void Close() { 339 if (base_) { 340 HttpBase* base = Disconnect(HE_NONE); 341 if (HM_RECV == base->mode_ && base->http_stream_) { 342 // Read I/O could have been stalled on the user of this DocumentStream, 343 // so restart the I/O process now that we've removed ourselves. 344 base->http_stream_->PostEvent(SE_READ, 0); 345 } 346 } 347 } 348 349 virtual bool GetAvailable(size_t* size) const { 350 if (!base_ || HM_RECV != base_->mode_) 351 return false; 352 size_t data_size = base_->GetDataRemaining(); 353 if (SIZE_UNKNOWN == data_size) 354 return false; 355 if (size) 356 *size = data_size; 357 return true; 358 } 359 360 HttpBase* Disconnect(HttpError error) { 361 ASSERT(NULL != base_); 362 ASSERT(NULL != base_->doc_stream_); 363 HttpBase* base = base_; 364 base_->doc_stream_ = NULL; 365 base_ = NULL; 366 error_ = error; 367 return base; 368 } 369 370 private: 371 HttpBase* base_; 372 HttpError error_; 373 }; 374 375 ////////////////////////////////////////////////////////////////////// 376 // HttpBase 377 ////////////////////////////////////////////////////////////////////// 378 379 HttpBase::HttpBase() : mode_(HM_NONE), data_(NULL), notify_(NULL), 380 http_stream_(NULL), doc_stream_(NULL) { 381 } 382 383 HttpBase::~HttpBase() { 384 ASSERT(HM_NONE == mode_); 385 } 386 387 bool 388 HttpBase::isConnected() const { 389 return (http_stream_ != NULL) && (http_stream_->GetState() == SS_OPEN); 390 } 391 392 bool 393 HttpBase::attach(StreamInterface* stream) { 394 if ((mode_ != HM_NONE) || (http_stream_ != NULL) || (stream == NULL)) { 395 ASSERT(false); 396 return false; 397 } 398 http_stream_ = stream; 399 http_stream_->SignalEvent.connect(this, &HttpBase::OnHttpStreamEvent); 400 mode_ = (http_stream_->GetState() == SS_OPENING) ? HM_CONNECT : HM_NONE; 401 return true; 402 } 403 404 StreamInterface* 405 HttpBase::detach() { 406 ASSERT(HM_NONE == mode_); 407 if (mode_ != HM_NONE) { 408 return NULL; 409 } 410 StreamInterface* stream = http_stream_; 411 http_stream_ = NULL; 412 if (stream) { 413 stream->SignalEvent.disconnect(this); 414 } 415 return stream; 416 } 417 418 void 419 HttpBase::send(HttpData* data) { 420 ASSERT(HM_NONE == mode_); 421 if (mode_ != HM_NONE) { 422 return; 423 } else if (!isConnected()) { 424 OnHttpStreamEvent(http_stream_, SE_CLOSE, HE_DISCONNECTED); 425 return; 426 } 427 428 mode_ = HM_SEND; 429 data_ = data; 430 len_ = 0; 431 ignore_data_ = chunk_data_ = false; 432 433 if (data_->document) { 434 data_->document->SignalEvent.connect(this, &HttpBase::OnDocumentEvent); 435 } 436 437 std::string encoding; 438 if (data_->hasHeader(HH_TRANSFER_ENCODING, &encoding) 439 && (encoding == "chunked")) { 440 chunk_data_ = true; 441 } 442 443 len_ = data_->formatLeader(buffer_, sizeof(buffer_)); 444 len_ += strcpyn(buffer_ + len_, sizeof(buffer_) - len_, "\r\n"); 445 446 header_ = data_->begin(); 447 if (header_ == data_->end()) { 448 // We must call this at least once, in the case where there are no headers. 449 queue_headers(); 450 } 451 452 flush_data(); 453 } 454 455 void 456 HttpBase::recv(HttpData* data) { 457 ASSERT(HM_NONE == mode_); 458 if (mode_ != HM_NONE) { 459 return; 460 } else if (!isConnected()) { 461 OnHttpStreamEvent(http_stream_, SE_CLOSE, HE_DISCONNECTED); 462 return; 463 } 464 465 mode_ = HM_RECV; 466 data_ = data; 467 len_ = 0; 468 ignore_data_ = chunk_data_ = false; 469 470 reset(); 471 if (doc_stream_) { 472 doc_stream_->SignalEvent(doc_stream_, SE_OPEN | SE_READ, 0); 473 } else { 474 read_and_process_data(); 475 } 476 } 477 478 void 479 HttpBase::abort(HttpError err) { 480 if (mode_ != HM_NONE) { 481 if (http_stream_ != NULL) { 482 http_stream_->Close(); 483 } 484 do_complete(err); 485 } 486 } 487 488 StreamInterface* HttpBase::GetDocumentStream() { 489 if (doc_stream_) 490 return NULL; 491 doc_stream_ = new DocumentStream(this); 492 return doc_stream_; 493 } 494 495 HttpError HttpBase::HandleStreamClose(int error) { 496 if (http_stream_ != NULL) { 497 http_stream_->Close(); 498 } 499 if (error == 0) { 500 if ((mode_ == HM_RECV) && is_valid_end_of_input()) { 501 return HE_NONE; 502 } else { 503 return HE_DISCONNECTED; 504 } 505 } else if (error == SOCKET_EACCES) { 506 return HE_AUTH; 507 } else if (error == SEC_E_CERT_EXPIRED) { 508 return HE_CERTIFICATE_EXPIRED; 509 } 510 LOG_F(LS_ERROR) << "(" << error << ")"; 511 return (HM_CONNECT == mode_) ? HE_CONNECT_FAILED : HE_SOCKET_ERROR; 512 } 513 514 bool HttpBase::DoReceiveLoop(HttpError* error) { 515 ASSERT(HM_RECV == mode_); 516 ASSERT(NULL != error); 517 518 // Do to the latency between receiving read notifications from 519 // pseudotcpchannel, we rely on repeated calls to read in order to acheive 520 // ideal throughput. The number of reads is limited to prevent starving 521 // the caller. 522 523 size_t loop_count = 0; 524 const size_t kMaxReadCount = 20; 525 bool process_requires_more_data = false; 526 do { 527 // The most frequent use of this function is response to new data available 528 // on http_stream_. Therefore, we optimize by attempting to read from the 529 // network first (as opposed to processing existing data first). 530 531 if (len_ < sizeof(buffer_)) { 532 // Attempt to buffer more data. 533 size_t read; 534 int read_error; 535 StreamResult read_result = http_stream_->Read(buffer_ + len_, 536 sizeof(buffer_) - len_, 537 &read, &read_error); 538 switch (read_result) { 539 case SR_SUCCESS: 540 ASSERT(len_ + read <= sizeof(buffer_)); 541 len_ += read; 542 break; 543 case SR_BLOCK: 544 if (process_requires_more_data) { 545 // We're can't make progress until more data is available. 546 return false; 547 } 548 // Attempt to process the data already in our buffer. 549 break; 550 case SR_EOS: 551 // Clean close, with no error. Fall through to HandleStreamClose. 552 read_error = 0; 553 case SR_ERROR: 554 *error = HandleStreamClose(read_error); 555 return true; 556 } 557 } else if (process_requires_more_data) { 558 // We have too much unprocessed data in our buffer. This should only 559 // occur when a single HTTP header is longer than the buffer size (32K). 560 // Anything longer than that is almost certainly an error. 561 *error = HE_OVERFLOW; 562 return true; 563 } 564 565 // Process data in our buffer. Process is not guaranteed to process all 566 // the buffered data. In particular, it will wait until a complete 567 // protocol element (such as http header, or chunk size) is available, 568 // before processing it in its entirety. Also, it is valid and sometimes 569 // necessary to call Process with an empty buffer, since the state machine 570 // may have interrupted state transitions to complete. 571 size_t processed; 572 ProcessResult process_result = Process(buffer_, len_, &processed, 573 error); 574 ASSERT(processed <= len_); 575 len_ -= processed; 576 memmove(buffer_, buffer_ + processed, len_); 577 switch (process_result) { 578 case PR_CONTINUE: 579 // We need more data to make progress. 580 process_requires_more_data = true; 581 break; 582 case PR_BLOCK: 583 // We're stalled on writing the processed data. 584 return false; 585 case PR_COMPLETE: 586 // *error already contains the correct code. 587 return true; 588 } 589 } while (++loop_count <= kMaxReadCount); 590 591 LOG_F(LS_WARNING) << "danger of starvation"; 592 return false; 593 } 594 595 void 596 HttpBase::read_and_process_data() { 597 HttpError error; 598 if (DoReceiveLoop(&error)) { 599 complete(error); 600 } 601 } 602 603 void 604 HttpBase::flush_data() { 605 ASSERT(HM_SEND == mode_); 606 607 // When send_required is true, no more buffering can occur without a network 608 // write. 609 bool send_required = (len_ >= sizeof(buffer_)); 610 611 while (true) { 612 ASSERT(len_ <= sizeof(buffer_)); 613 614 // HTTP is inherently sensitive to round trip latency, since a frequent use 615 // case is for small requests and responses to be sent back and forth, and 616 // the lack of pipelining forces a single request to take a minimum of the 617 // round trip time. As a result, it is to our benefit to pack as much data 618 // into each packet as possible. Thus, we defer network writes until we've 619 // buffered as much data as possible. 620 621 if (!send_required && (header_ != data_->end())) { 622 // First, attempt to queue more header data. 623 send_required = queue_headers(); 624 } 625 626 if (!send_required && data_->document) { 627 // Next, attempt to queue document data. 628 629 const size_t kChunkDigits = 8; 630 size_t offset, reserve; 631 if (chunk_data_) { 632 // Reserve characters at the start for X-byte hex value and \r\n 633 offset = len_ + kChunkDigits + 2; 634 // ... and 2 characters at the end for \r\n 635 reserve = offset + 2; 636 } else { 637 offset = len_; 638 reserve = offset; 639 } 640 641 if (reserve >= sizeof(buffer_)) { 642 send_required = true; 643 } else { 644 size_t read; 645 int error; 646 StreamResult result = data_->document->Read(buffer_ + offset, 647 sizeof(buffer_) - reserve, 648 &read, &error); 649 if (result == SR_SUCCESS) { 650 ASSERT(reserve + read <= sizeof(buffer_)); 651 if (chunk_data_) { 652 // Prepend the chunk length in hex. 653 // Note: sprintfn appends a null terminator, which is why we can't 654 // combine it with the line terminator. 655 sprintfn(buffer_ + len_, kChunkDigits + 1, "%.*x", 656 kChunkDigits, read); 657 // Add line terminator to the chunk length. 658 memcpy(buffer_ + len_ + kChunkDigits, "\r\n", 2); 659 // Add line terminator to the end of the chunk. 660 memcpy(buffer_ + offset + read, "\r\n", 2); 661 } 662 len_ = reserve + read; 663 } else if (result == SR_BLOCK) { 664 // Nothing to do but flush data to the network. 665 send_required = true; 666 } else if (result == SR_EOS) { 667 if (chunk_data_) { 668 // Append the empty chunk and empty trailers, then turn off 669 // chunking. 670 ASSERT(len_ + 5 <= sizeof(buffer_)); 671 memcpy(buffer_ + len_, "0\r\n\r\n", 5); 672 len_ += 5; 673 chunk_data_ = false; 674 } else if (0 == len_) { 675 // No more data to read, and no more data to write. 676 do_complete(); 677 return; 678 } 679 // Although we are done reading data, there is still data which needs 680 // to be flushed to the network. 681 send_required = true; 682 } else { 683 LOG_F(LS_ERROR) << "Read error: " << error; 684 do_complete(HE_STREAM); 685 return; 686 } 687 } 688 } 689 690 if (0 == len_) { 691 // No data currently available to send. 692 if (!data_->document) { 693 // If there is no source document, that means we're done. 694 do_complete(); 695 } 696 return; 697 } 698 699 size_t written; 700 int error; 701 StreamResult result = http_stream_->Write(buffer_, len_, &written, &error); 702 if (result == SR_SUCCESS) { 703 ASSERT(written <= len_); 704 len_ -= written; 705 memmove(buffer_, buffer_ + written, len_); 706 send_required = false; 707 } else if (result == SR_BLOCK) { 708 if (send_required) { 709 // Nothing more we can do until network is writeable. 710 return; 711 } 712 } else { 713 ASSERT(result == SR_ERROR); 714 LOG_F(LS_ERROR) << "error"; 715 OnHttpStreamEvent(http_stream_, SE_CLOSE, error); 716 return; 717 } 718 } 719 720 ASSERT(false); 721 } 722 723 bool 724 HttpBase::queue_headers() { 725 ASSERT(HM_SEND == mode_); 726 while (header_ != data_->end()) { 727 size_t len = sprintfn(buffer_ + len_, sizeof(buffer_) - len_, 728 "%.*s: %.*s\r\n", 729 header_->first.size(), header_->first.data(), 730 header_->second.size(), header_->second.data()); 731 if (len_ + len < sizeof(buffer_) - 3) { 732 len_ += len; 733 ++header_; 734 } else if (len_ == 0) { 735 LOG(WARNING) << "discarding header that is too long: " << header_->first; 736 ++header_; 737 } else { 738 // Not enough room for the next header, write to network first. 739 return true; 740 } 741 } 742 // End of headers 743 len_ += strcpyn(buffer_ + len_, sizeof(buffer_) - len_, "\r\n"); 744 return false; 745 } 746 747 void 748 HttpBase::do_complete(HttpError err) { 749 ASSERT(mode_ != HM_NONE); 750 HttpMode mode = mode_; 751 mode_ = HM_NONE; 752 if (data_ && data_->document) { 753 data_->document->SignalEvent.disconnect(this); 754 } 755 data_ = NULL; 756 if ((HM_RECV == mode) && doc_stream_) { 757 ASSERT(HE_NONE != err); // We should have Disconnected doc_stream_ already. 758 DocumentStream* ds = doc_stream_; 759 ds->Disconnect(err); 760 ds->SignalEvent(ds, SE_CLOSE, err); 761 } 762 if (notify_) { 763 notify_->onHttpComplete(mode, err); 764 } 765 } 766 767 // 768 // Stream Signals 769 // 770 771 void 772 HttpBase::OnHttpStreamEvent(StreamInterface* stream, int events, int error) { 773 ASSERT(stream == http_stream_); 774 if ((events & SE_OPEN) && (mode_ == HM_CONNECT)) { 775 do_complete(); 776 return; 777 } 778 779 if ((events & SE_WRITE) && (mode_ == HM_SEND)) { 780 flush_data(); 781 return; 782 } 783 784 if ((events & SE_READ) && (mode_ == HM_RECV)) { 785 if (doc_stream_) { 786 doc_stream_->SignalEvent(doc_stream_, SE_READ, 0); 787 } else { 788 read_and_process_data(); 789 } 790 return; 791 } 792 793 if ((events & SE_CLOSE) == 0) 794 return; 795 796 HttpError http_error = HandleStreamClose(error); 797 if (mode_ == HM_RECV) { 798 complete(http_error); 799 } else if (mode_ != HM_NONE) { 800 do_complete(http_error); 801 } else if (notify_) { 802 notify_->onHttpClosed(http_error); 803 } 804 } 805 806 void 807 HttpBase::OnDocumentEvent(StreamInterface* stream, int events, int error) { 808 ASSERT(stream == data_->document.get()); 809 if ((events & SE_WRITE) && (mode_ == HM_RECV)) { 810 read_and_process_data(); 811 return; 812 } 813 814 if ((events & SE_READ) && (mode_ == HM_SEND)) { 815 flush_data(); 816 return; 817 } 818 819 if (events & SE_CLOSE) { 820 LOG_F(LS_ERROR) << "Read error: " << error; 821 do_complete(HE_STREAM); 822 return; 823 } 824 } 825 826 // 827 // HttpParser Implementation 828 // 829 830 HttpParser::ProcessResult 831 HttpBase::ProcessLeader(const char* line, size_t len, HttpError* error) { 832 *error = data_->parseLeader(line, len); 833 return (HE_NONE == *error) ? PR_CONTINUE : PR_COMPLETE; 834 } 835 836 HttpParser::ProcessResult 837 HttpBase::ProcessHeader(const char* name, size_t nlen, const char* value, 838 size_t vlen, HttpError* error) { 839 std::string sname(name, nlen), svalue(value, vlen); 840 data_->addHeader(sname, svalue); 841 return PR_CONTINUE; 842 } 843 844 HttpParser::ProcessResult 845 HttpBase::ProcessHeaderComplete(bool chunked, size_t& data_size, 846 HttpError* error) { 847 StreamInterface* old_docstream = doc_stream_; 848 if (notify_) { 849 *error = notify_->onHttpHeaderComplete(chunked, data_size); 850 // The request must not be aborted as a result of this callback. 851 ASSERT(NULL != data_); 852 } 853 if ((HE_NONE == *error) && data_->document) { 854 data_->document->SignalEvent.connect(this, &HttpBase::OnDocumentEvent); 855 } 856 if (HE_NONE != *error) { 857 return PR_COMPLETE; 858 } 859 if (old_docstream != doc_stream_) { 860 // Break out of Process loop, since our I/O model just changed. 861 return PR_BLOCK; 862 } 863 return PR_CONTINUE; 864 } 865 866 HttpParser::ProcessResult 867 HttpBase::ProcessData(const char* data, size_t len, size_t& read, 868 HttpError* error) { 869 if (ignore_data_ || !data_->document) { 870 read = len; 871 return PR_CONTINUE; 872 } 873 int write_error = 0; 874 switch (data_->document->Write(data, len, &read, &write_error)) { 875 case SR_SUCCESS: 876 return PR_CONTINUE; 877 case SR_BLOCK: 878 return PR_BLOCK; 879 case SR_EOS: 880 LOG_F(LS_ERROR) << "Unexpected EOS"; 881 *error = HE_STREAM; 882 return PR_COMPLETE; 883 case SR_ERROR: 884 default: 885 LOG_F(LS_ERROR) << "Write error: " << write_error; 886 *error = HE_STREAM; 887 return PR_COMPLETE; 888 } 889 } 890 891 void 892 HttpBase::OnComplete(HttpError err) { 893 LOG_F(LS_VERBOSE); 894 do_complete(err); 895 } 896 897 } // namespace talk_base 898