Home | History | Annotate | Download | only in url_request
      1 // Copyright (c) 2011 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 "net/url_request/url_request_job.h"
      6 
      7 #include "base/compiler_specific.h"
      8 #include "base/message_loop.h"
      9 #include "base/string_number_conversions.h"
     10 #include "base/string_util.h"
     11 #include "net/base/auth.h"
     12 #include "net/base/host_port_pair.h"
     13 #include "net/base/io_buffer.h"
     14 #include "net/base/load_states.h"
     15 #include "net/base/net_errors.h"
     16 #include "net/base/network_delegate.h"
     17 #include "net/http/http_response_headers.h"
     18 #include "net/url_request/url_request.h"
     19 #include "net/url_request/url_request_context.h"
     20 #include "net/url_request/url_request_job_tracker.h"
     21 
     22 namespace net {
     23 
     24 URLRequestJob::URLRequestJob(URLRequest* request)
     25     : request_(request),
     26       done_(false),
     27       prefilter_bytes_read_(0),
     28       postfilter_bytes_read_(0),
     29       filter_input_byte_count_(0),
     30       filter_needs_more_output_space_(false),
     31       filtered_read_buffer_len_(0),
     32       has_handled_response_(false),
     33       expected_content_size_(-1),
     34       deferred_redirect_status_code_(-1),
     35       ALLOW_THIS_IN_INITIALIZER_LIST(method_factory_(this)) {
     36   g_url_request_job_tracker.AddNewJob(this);
     37 }
     38 
     39 void URLRequestJob::SetUpload(UploadData* upload) {
     40 }
     41 
     42 void URLRequestJob::SetExtraRequestHeaders(
     43     const HttpRequestHeaders& headers) {
     44 }
     45 
     46 void URLRequestJob::Kill() {
     47   // Make sure the request is notified that we are done.  We assume that the
     48   // request took care of setting its error status before calling Kill.
     49   if (request_)
     50     NotifyCanceled();
     51 }
     52 
     53 void URLRequestJob::DetachRequest() {
     54   request_ = NULL;
     55 }
     56 
     57 // This function calls ReadData to get stream data. If a filter exists, passes
     58 // the data to the attached filter. Then returns the output from filter back to
     59 // the caller.
     60 bool URLRequestJob::Read(IOBuffer* buf, int buf_size, int *bytes_read) {
     61   bool rv = false;
     62 
     63   DCHECK_LT(buf_size, 1000000);  // sanity check
     64   DCHECK(buf);
     65   DCHECK(bytes_read);
     66   DCHECK(filtered_read_buffer_ == NULL);
     67   DCHECK_EQ(0, filtered_read_buffer_len_);
     68 
     69   *bytes_read = 0;
     70 
     71   // Skip Filter if not present
     72   if (!filter_.get()) {
     73     rv = ReadRawDataHelper(buf, buf_size, bytes_read);
     74   } else {
     75     // Save the caller's buffers while we do IO
     76     // in the filter's buffers.
     77     filtered_read_buffer_ = buf;
     78     filtered_read_buffer_len_ = buf_size;
     79 
     80     if (ReadFilteredData(bytes_read)) {
     81       rv = true;   // we have data to return
     82     } else {
     83       rv = false;  // error, or a new IO is pending
     84     }
     85   }
     86   if (rv && *bytes_read == 0)
     87     NotifyDone(URLRequestStatus());
     88   return rv;
     89 }
     90 
     91 void URLRequestJob::StopCaching() {
     92   // Nothing to do here.
     93 }
     94 
     95 LoadState URLRequestJob::GetLoadState() const {
     96   return LOAD_STATE_IDLE;
     97 }
     98 
     99 uint64 URLRequestJob::GetUploadProgress() const {
    100   return 0;
    101 }
    102 
    103 bool URLRequestJob::GetCharset(std::string* charset) {
    104   return false;
    105 }
    106 
    107 void URLRequestJob::GetResponseInfo(HttpResponseInfo* info) {
    108 }
    109 
    110 bool URLRequestJob::GetResponseCookies(std::vector<std::string>* cookies) {
    111   return false;
    112 }
    113 
    114 Filter* URLRequestJob::SetupFilter() const {
    115   return NULL;
    116 }
    117 
    118 bool URLRequestJob::IsRedirectResponse(GURL* location,
    119                                        int* http_status_code) {
    120   // For non-HTTP jobs, headers will be null.
    121   HttpResponseHeaders* headers = request_->response_headers();
    122   if (!headers)
    123     return false;
    124 
    125   std::string value;
    126   if (!headers->IsRedirect(&value))
    127     return false;
    128 
    129   *location = request_->url().Resolve(value);
    130   *http_status_code = headers->response_code();
    131   return true;
    132 }
    133 
    134 bool URLRequestJob::IsSafeRedirect(const GURL& location) {
    135   return true;
    136 }
    137 
    138 bool URLRequestJob::NeedsAuth() {
    139   return false;
    140 }
    141 
    142 void URLRequestJob::GetAuthChallengeInfo(
    143     scoped_refptr<AuthChallengeInfo>* auth_info) {
    144   // This will only be called if NeedsAuth() returns true, in which
    145   // case the derived class should implement this!
    146   NOTREACHED();
    147 }
    148 
    149 void URLRequestJob::SetAuth(const string16& username,
    150                             const string16& password) {
    151   // This will only be called if NeedsAuth() returns true, in which
    152   // case the derived class should implement this!
    153   NOTREACHED();
    154 }
    155 
    156 void URLRequestJob::CancelAuth() {
    157   // This will only be called if NeedsAuth() returns true, in which
    158   // case the derived class should implement this!
    159   NOTREACHED();
    160 }
    161 
    162 void URLRequestJob::ContinueWithCertificate(
    163     X509Certificate* client_cert) {
    164   // The derived class should implement this!
    165   NOTREACHED();
    166 }
    167 
    168 void URLRequestJob::ContinueDespiteLastError() {
    169   // Implementations should know how to recover from errors they generate.
    170   // If this code was reached, we are trying to recover from an error that
    171   // we don't know how to recover from.
    172   NOTREACHED();
    173 }
    174 
    175 void URLRequestJob::FollowDeferredRedirect() {
    176   DCHECK(deferred_redirect_status_code_ != -1);
    177 
    178   // NOTE: deferred_redirect_url_ may be invalid, and attempting to redirect to
    179   // such an URL will fail inside FollowRedirect.  The DCHECK above asserts
    180   // that we called OnReceivedRedirect.
    181 
    182   // It is also possible that FollowRedirect will drop the last reference to
    183   // this job, so we need to reset our members before calling it.
    184 
    185   GURL redirect_url = deferred_redirect_url_;
    186   int redirect_status_code = deferred_redirect_status_code_;
    187 
    188   deferred_redirect_url_ = GURL();
    189   deferred_redirect_status_code_ = -1;
    190 
    191   FollowRedirect(redirect_url, redirect_status_code);
    192 }
    193 
    194 bool URLRequestJob::GetMimeType(std::string* mime_type) const {
    195   return false;
    196 }
    197 
    198 int URLRequestJob::GetResponseCode() const {
    199   return -1;
    200 }
    201 
    202 HostPortPair URLRequestJob::GetSocketAddress() const {
    203   return HostPortPair();
    204 }
    205 
    206 URLRequestJob::~URLRequestJob() {
    207   g_url_request_job_tracker.RemoveJob(this);
    208 }
    209 
    210 void URLRequestJob::NotifyHeadersComplete() {
    211   if (!request_ || !request_->delegate())
    212     return;  // The request was destroyed, so there is no more work to do.
    213 
    214   if (has_handled_response_)
    215     return;
    216 
    217   DCHECK(!request_->status().is_io_pending());
    218 
    219   // Initialize to the current time, and let the subclass optionally override
    220   // the time stamps if it has that information.  The default request_time is
    221   // set by URLRequest before it calls our Start method.
    222   request_->response_info_.response_time = base::Time::Now();
    223   GetResponseInfo(&request_->response_info_);
    224 
    225   // When notifying the delegate, the delegate can release the request
    226   // (and thus release 'this').  After calling to the delgate, we must
    227   // check the request pointer to see if it still exists, and return
    228   // immediately if it has been destroyed.  self_preservation ensures our
    229   // survival until we can get out of this method.
    230   scoped_refptr<URLRequestJob> self_preservation(this);
    231 
    232   GURL new_location;
    233   int http_status_code;
    234   if (IsRedirectResponse(&new_location, &http_status_code)) {
    235     const GURL& url = request_->url();
    236 
    237     // Move the reference fragment of the old location to the new one if the
    238     // new one has none. This duplicates mozilla's behavior.
    239     if (url.is_valid() && url.has_ref() && !new_location.has_ref()) {
    240       GURL::Replacements replacements;
    241       // Reference the |ref| directly out of the original URL to avoid a
    242       // malloc.
    243       replacements.SetRef(url.spec().data(),
    244                           url.parsed_for_possibly_invalid_spec().ref);
    245       new_location = new_location.ReplaceComponents(replacements);
    246     }
    247 
    248     bool defer_redirect = false;
    249     request_->ReceivedRedirect(new_location, &defer_redirect);
    250 
    251     // Ensure that the request wasn't detached or destroyed in ReceivedRedirect
    252     if (!request_ || !request_->delegate())
    253       return;
    254 
    255     // If we were not cancelled, then maybe follow the redirect.
    256     if (request_->status().is_success()) {
    257       if (defer_redirect) {
    258         deferred_redirect_url_ = new_location;
    259         deferred_redirect_status_code_ = http_status_code;
    260       } else {
    261         FollowRedirect(new_location, http_status_code);
    262       }
    263       return;
    264     }
    265   } else if (NeedsAuth()) {
    266     scoped_refptr<AuthChallengeInfo> auth_info;
    267     GetAuthChallengeInfo(&auth_info);
    268     // Need to check for a NULL auth_info because the server may have failed
    269     // to send a challenge with the 401 response.
    270     if (auth_info) {
    271       request_->delegate()->OnAuthRequired(request_, auth_info);
    272       // Wait for SetAuth or CancelAuth to be called.
    273       return;
    274     }
    275   }
    276 
    277   has_handled_response_ = true;
    278   if (request_->status().is_success())
    279     filter_.reset(SetupFilter());
    280 
    281   if (!filter_.get()) {
    282     std::string content_length;
    283     request_->GetResponseHeaderByName("content-length", &content_length);
    284     if (!content_length.empty())
    285       base::StringToInt64(content_length, &expected_content_size_);
    286   }
    287 
    288   request_->ResponseStarted();
    289 }
    290 
    291 void URLRequestJob::NotifyReadComplete(int bytes_read) {
    292   if (!request_ || !request_->delegate())
    293     return;  // The request was destroyed, so there is no more work to do.
    294 
    295   // TODO(darin): Bug 1004233. Re-enable this test once all of the chrome
    296   // unit_tests have been fixed to not trip this.
    297   //DCHECK(!request_->status().is_io_pending());
    298 
    299   // The headers should be complete before reads complete
    300   DCHECK(has_handled_response_);
    301 
    302   OnRawReadComplete(bytes_read);
    303 
    304   // Don't notify if we had an error.
    305   if (!request_->status().is_success())
    306     return;
    307 
    308   // When notifying the delegate, the delegate can release the request
    309   // (and thus release 'this').  After calling to the delgate, we must
    310   // check the request pointer to see if it still exists, and return
    311   // immediately if it has been destroyed.  self_preservation ensures our
    312   // survival until we can get out of this method.
    313   scoped_refptr<URLRequestJob> self_preservation(this);
    314 
    315   prefilter_bytes_read_ += bytes_read;
    316   if (filter_.get()) {
    317     // Tell the filter that it has more data
    318     FilteredDataRead(bytes_read);
    319 
    320     // Filter the data.
    321     int filter_bytes_read = 0;
    322     if (ReadFilteredData(&filter_bytes_read)) {
    323       postfilter_bytes_read_ += filter_bytes_read;
    324       if (request_->context() && request_->context()->network_delegate()) {
    325         request_->context()->network_delegate()->NotifyReadCompleted(
    326             request_, filter_bytes_read);
    327       }
    328       request_->delegate()->OnReadCompleted(request_, filter_bytes_read);
    329     }
    330   } else {
    331     postfilter_bytes_read_ += bytes_read;
    332     if (request_->context() && request_->context()->network_delegate()) {
    333       request_->context()->network_delegate()->NotifyReadCompleted(
    334           request_, bytes_read);
    335     }
    336     request_->delegate()->OnReadCompleted(request_, bytes_read);
    337   }
    338 }
    339 
    340 void URLRequestJob::NotifyStartError(const URLRequestStatus &status) {
    341   DCHECK(!has_handled_response_);
    342   has_handled_response_ = true;
    343   if (request_) {
    344     request_->set_status(status);
    345     request_->ResponseStarted();
    346   }
    347 }
    348 
    349 void URLRequestJob::NotifyDone(const URLRequestStatus &status) {
    350   DCHECK(!done_) << "Job sending done notification twice";
    351   if (done_)
    352     return;
    353   done_ = true;
    354 
    355   // Unless there was an error, we should have at least tried to handle
    356   // the response before getting here.
    357   DCHECK(has_handled_response_ || !status.is_success());
    358 
    359   // As with NotifyReadComplete, we need to take care to notice if we were
    360   // destroyed during a delegate callback.
    361   if (request_) {
    362     request_->set_is_pending(false);
    363     // With async IO, it's quite possible to have a few outstanding
    364     // requests.  We could receive a request to Cancel, followed shortly
    365     // by a successful IO.  For tracking the status(), once there is
    366     // an error, we do not change the status back to success.  To
    367     // enforce this, only set the status if the job is so far
    368     // successful.
    369     if (request_->status().is_success())
    370       request_->set_status(status);
    371   }
    372 
    373   g_url_request_job_tracker.OnJobDone(this, status);
    374 
    375   // Complete this notification later.  This prevents us from re-entering the
    376   // delegate if we're done because of a synchronous call.
    377   MessageLoop::current()->PostTask(
    378       FROM_HERE,
    379       method_factory_.NewRunnableMethod(&URLRequestJob::CompleteNotifyDone));
    380 }
    381 
    382 void URLRequestJob::CompleteNotifyDone() {
    383   // Check if we should notify the delegate that we're done because of an error.
    384   if (request_ &&
    385       !request_->status().is_success() &&
    386       request_->delegate()) {
    387     // We report the error differently depending on whether we've called
    388     // OnResponseStarted yet.
    389     if (has_handled_response_) {
    390       // We signal the error by calling OnReadComplete with a bytes_read of -1.
    391       if (request_->context() && request_->context()->network_delegate())
    392         request_->context()->network_delegate()->NotifyReadCompleted(
    393             request_, -1);
    394       request_->delegate()->OnReadCompleted(request_, -1);
    395     } else {
    396       has_handled_response_ = true;
    397       request_->ResponseStarted();
    398     }
    399   }
    400 }
    401 
    402 void URLRequestJob::NotifyCanceled() {
    403   if (!done_) {
    404     NotifyDone(URLRequestStatus(URLRequestStatus::CANCELED,
    405                                 ERR_ABORTED));
    406   }
    407 }
    408 
    409 void URLRequestJob::NotifyRestartRequired() {
    410   DCHECK(!has_handled_response_);
    411   if (GetStatus().status() != URLRequestStatus::CANCELED)
    412     request_->Restart();
    413 }
    414 
    415 bool URLRequestJob::ReadRawData(IOBuffer* buf, int buf_size,
    416                                 int *bytes_read) {
    417   DCHECK(bytes_read);
    418   *bytes_read = 0;
    419   NotifyDone(URLRequestStatus());
    420   return false;
    421 }
    422 
    423 void URLRequestJob::FilteredDataRead(int bytes_read) {
    424   DCHECK(filter_.get());  // don't add data if there is no filter
    425   filter_->FlushStreamBuffer(bytes_read);
    426 }
    427 
    428 bool URLRequestJob::ReadFilteredData(int* bytes_read) {
    429   DCHECK(filter_.get());  // don't add data if there is no filter
    430   DCHECK(filtered_read_buffer_ != NULL);  // we need to have a buffer to fill
    431   DCHECK_GT(filtered_read_buffer_len_, 0);  // sanity check
    432   DCHECK_LT(filtered_read_buffer_len_, 1000000);  // sanity check
    433   DCHECK(raw_read_buffer_ == NULL);  // there should be no raw read buffer yet
    434 
    435   bool rv = false;
    436   *bytes_read = 0;
    437 
    438   if (is_done())
    439     return true;
    440 
    441   if (!filter_needs_more_output_space_ && !filter_->stream_data_len()) {
    442     // We don't have any raw data to work with, so
    443     // read from the socket.
    444     int filtered_data_read;
    445     if (ReadRawDataForFilter(&filtered_data_read)) {
    446       if (filtered_data_read > 0) {
    447         filter_->FlushStreamBuffer(filtered_data_read);  // Give data to filter.
    448       } else {
    449         return true;  // EOF
    450       }
    451     } else {
    452       return false;  // IO Pending (or error)
    453     }
    454   }
    455 
    456   if ((filter_->stream_data_len() || filter_needs_more_output_space_)
    457       && !is_done()) {
    458     // Get filtered data.
    459     int filtered_data_len = filtered_read_buffer_len_;
    460     Filter::FilterStatus status;
    461     int output_buffer_size = filtered_data_len;
    462     status = filter_->ReadData(filtered_read_buffer_->data(),
    463                                &filtered_data_len);
    464 
    465     if (filter_needs_more_output_space_ && 0 == filtered_data_len) {
    466       // filter_needs_more_output_space_ was mistaken... there are no more bytes
    467       // and we should have at least tried to fill up the filter's input buffer.
    468       // Correct the state, and try again.
    469       filter_needs_more_output_space_ = false;
    470       return ReadFilteredData(bytes_read);
    471     }
    472 
    473     switch (status) {
    474       case Filter::FILTER_DONE: {
    475         filter_needs_more_output_space_ = false;
    476         *bytes_read = filtered_data_len;
    477         rv = true;
    478         break;
    479       }
    480       case Filter::FILTER_NEED_MORE_DATA: {
    481         filter_needs_more_output_space_ =
    482             (filtered_data_len == output_buffer_size);
    483         // We have finished filtering all data currently in the buffer.
    484         // There might be some space left in the output buffer. One can
    485         // consider reading more data from the stream to feed the filter
    486         // and filling up the output buffer. This leads to more complicated
    487         // buffer management and data notification mechanisms.
    488         // We can revisit this issue if there is a real perf need.
    489         if (filtered_data_len > 0) {
    490           *bytes_read = filtered_data_len;
    491           rv = true;
    492         } else {
    493           // Read again since we haven't received enough data yet (e.g., we may
    494           // not have a complete gzip header yet)
    495           rv = ReadFilteredData(bytes_read);
    496         }
    497         break;
    498       }
    499       case Filter::FILTER_OK: {
    500         filter_needs_more_output_space_ =
    501             (filtered_data_len == output_buffer_size);
    502         *bytes_read = filtered_data_len;
    503         rv = true;
    504         break;
    505       }
    506       case Filter::FILTER_ERROR: {
    507         filter_needs_more_output_space_ = false;
    508         NotifyDone(URLRequestStatus(URLRequestStatus::FAILED,
    509                    ERR_CONTENT_DECODING_FAILED));
    510         rv = false;
    511         break;
    512       }
    513       default: {
    514         NOTREACHED();
    515         filter_needs_more_output_space_ = false;
    516         rv = false;
    517         break;
    518       }
    519     }
    520   } else {
    521     // we are done, or there is no data left.
    522     rv = true;
    523   }
    524 
    525   if (rv) {
    526     // When we successfully finished a read, we no longer need to
    527     // save the caller's buffers. Release our reference.
    528     filtered_read_buffer_ = NULL;
    529     filtered_read_buffer_len_ = 0;
    530   }
    531   return rv;
    532 }
    533 
    534 const URLRequestStatus URLRequestJob::GetStatus() {
    535   if (request_)
    536     return request_->status();
    537   // If the request is gone, we must be cancelled.
    538   return URLRequestStatus(URLRequestStatus::CANCELED,
    539                           ERR_ABORTED);
    540 }
    541 
    542 void URLRequestJob::SetStatus(const URLRequestStatus &status) {
    543   if (request_)
    544     request_->set_status(status);
    545 }
    546 
    547 bool URLRequestJob::ReadRawDataForFilter(int* bytes_read) {
    548   bool rv = false;
    549 
    550   DCHECK(bytes_read);
    551   DCHECK(filter_.get());
    552 
    553   *bytes_read = 0;
    554 
    555   // Get more pre-filtered data if needed.
    556   // TODO(mbelshe): is it possible that the filter needs *MORE* data
    557   //    when there is some data already in the buffer?
    558   if (!filter_->stream_data_len() && !is_done()) {
    559     IOBuffer* stream_buffer = filter_->stream_buffer();
    560     int stream_buffer_size = filter_->stream_buffer_size();
    561     rv = ReadRawDataHelper(stream_buffer, stream_buffer_size, bytes_read);
    562   }
    563   return rv;
    564 }
    565 
    566 bool URLRequestJob::ReadRawDataHelper(IOBuffer* buf, int buf_size,
    567                                       int* bytes_read) {
    568   DCHECK(!request_->status().is_io_pending());
    569   DCHECK(raw_read_buffer_ == NULL);
    570 
    571   // Keep a pointer to the read buffer, so we have access to it in the
    572   // OnRawReadComplete() callback in the event that the read completes
    573   // asynchronously.
    574   raw_read_buffer_ = buf;
    575   bool rv = ReadRawData(buf, buf_size, bytes_read);
    576 
    577   if (!request_->status().is_io_pending()) {
    578     // If the read completes synchronously, either success or failure,
    579     // invoke the OnRawReadComplete callback so we can account for the
    580     // completed read.
    581     OnRawReadComplete(*bytes_read);
    582   }
    583   return rv;
    584 }
    585 
    586 void URLRequestJob::FollowRedirect(const GURL& location, int http_status_code) {
    587   g_url_request_job_tracker.OnJobRedirect(this, location, http_status_code);
    588 
    589   int rv = request_->Redirect(location, http_status_code);
    590   if (rv != OK)
    591     NotifyDone(URLRequestStatus(URLRequestStatus::FAILED, rv));
    592 }
    593 
    594 void URLRequestJob::OnRawReadComplete(int bytes_read) {
    595   DCHECK(raw_read_buffer_);
    596   if (bytes_read > 0) {
    597     RecordBytesRead(bytes_read);
    598   }
    599   raw_read_buffer_ = NULL;
    600 }
    601 
    602 void URLRequestJob::RecordBytesRead(int bytes_read) {
    603   filter_input_byte_count_ += bytes_read;
    604   UpdatePacketReadTimes();  // Facilitate stats recording if it is active.
    605   g_url_request_job_tracker.OnBytesRead(this, raw_read_buffer_->data(),
    606                                         bytes_read);
    607 }
    608 
    609 bool URLRequestJob::FilterHasData() {
    610     return filter_.get() && filter_->stream_data_len();
    611 }
    612 
    613 void URLRequestJob::UpdatePacketReadTimes() {
    614 }
    615 
    616 }  // namespace net
    617