In BlobURLRequestJob, open files asynchronously to avoid blocking the IO thread

webkit/blob/blob_url_request_job.cc

 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "webkit/blob/blob_url_request_job.h"
 
 #include "base/compiler_specific.h"
 #include "base/file_path.h"
 #include "base/file_util.h"
 #include "base/file_util_proxy.h"
 #include "base/message_loop.h"
 #include "base/message_loop_proxy.h"
 #include "base/string_number_conversions.h"
+#include "net/base/file_stream.h"
 #include "net/base/io_buffer.h"
 #include "net/base/net_errors.h"
 #include "net/http/http_request_headers.h"
 #include "net/http/http_response_headers.h"
 #include "net/http/http_response_info.h"
 #include "net/http/http_util.h"
 #include "net/url_request/url_request.h"
 #include "net/url_request/url_request_error_job.h"
 #include "net/url_request/url_request_status.h"
 
 namespace webkit_blob {
 
 static const int kHTTPOk = 200;
 static const int kHTTPPartialContent = 206;
 static const int kHTTPNotAllowed = 403;
 static const int kHTTPNotFound = 404;
 static const int kHTTPMethodNotAllow = 405;
 static const int kHTTPRequestedRangeNotSatisfiable = 416;
 static const int kHTTPInternalError = 500;
 
 static const char kHTTPOKText[] = "OK";
 static const char kHTTPPartialContentText[] = "Partial Content";
 static const char kHTTPNotAllowedText[] = "Not Allowed";
 static const char kHTTPNotFoundText[] = "Not Found";
 static const char kHTTPMethodNotAllowText[] = "Method Not Allowed";
 static const char kHTTPRequestedRangeNotSatisfiableText[] =
     "Requested Range Not Satisfiable";
 static const char kHTTPInternalErrorText[] = "Internal Server Error";
 
+static const int kFileOpenFlags = base::PLATFORM_FILE_OPEN |
+                                  base::PLATFORM_FILE_READ |
+                                  base::PLATFORM_FILE_ASYNC;
+
 BlobURLRequestJob::BlobURLRequestJob(
     net::URLRequest* request,
     BlobData* blob_data,
     base::MessageLoopProxy* file_thread_proxy)
     : net::URLRequestJob(request),
       callback_factory_(ALLOW_THIS_IN_INITIALIZER_LIST(this)),
       blob_data_(blob_data),
       file_thread_proxy_(file_thread_proxy),
       ALLOW_THIS_IN_INITIALIZER_LIST(
           io_callback_(this, &BlobURLRequestJob::DidRead)),
       item_index_(0),
       total_size_(0),
       current_item_offset_(0),
       remaining_bytes_(0),
       read_buf_offset_(0),
       read_buf_size_(0),
       read_buf_remaining_bytes_(0),
+      bytes_to_read_(0),
       error_(false),
       headers_set_(false),
       byte_range_set_(false),
       ALLOW_THIS_IN_INITIALIZER_LIST(method_factory_(this)) {
 }
 
 BlobURLRequestJob::~BlobURLRequestJob() {
+  // FileStream's destructor won't close it for us because we passed in our own
+  // file handle.
+  CloseStream();
 }
 
 void BlobURLRequestJob::Start() {
   // Continue asynchronously.
   MessageLoop::current()->PostTask(
       FROM_HERE,
       method_factory_.NewRunnableMethod(&BlobURLRequestJob::DidStart));
 }
 
 void BlobURLRequestJob::DidStart() {
   // We only support GET request per the spec.
   if (request()->method() != "GET") {
     NotifyFailure(net::ERR_METHOD_NOT_SUPPORTED);
     return;
   }
 
   // If the blob data is not present, bail out.
   if (!blob_data_) {
     NotifyFailure(net::ERR_FILE_NOT_FOUND);
     return;
   }
 
   CountSize();
 }
 
+void BlobURLRequestJob::CloseStream() {
+  if (stream_ != NULL) {
+    stream_->Close();
+    stream_.reset(NULL);
+  }
+}
+
 void BlobURLRequestJob::Kill() {
-  stream_.Close();
+  CloseStream();
 
   net::URLRequestJob::Kill();
   callback_factory_.RevokeAll();
   method_factory_.RevokeAll();
 }
 
 void BlobURLRequestJob::ResolveFile(const FilePath& file_path) {

[michaeln, 2011/03/11 06:03:47]

nit: maybe DCHECK(file_thread_proxy_)

[adamk, 2011/03/14 17:42:14]

Added that DCHECK to the constructor (that'll make it useful, as it will
actually include the errant caller in the stack trace).

-  // If the file thread proxy is provided, we can use it get the file info.
-  if (file_thread_proxy_) {
-    base::FileUtilProxy::GetFileInfo(
-          file_thread_proxy_,
-          file_path,
-          callback_factory_.NewCallback(&BlobURLRequestJob::DidResolve));
-    return;
-  }
-
-  // Otherwise, we use current thread, i.e. IO thread, as this is the case when
-  // we run the unittest or test shell.
-  // TODO(jianli): Consider using the proxy of current thread.
-  base::PlatformFileInfo file_info;
-  bool exists = file_util::GetFileInfo(file_path, &file_info);
-
-  // Continue asynchronously.
-  MessageLoop::current()->PostTask(
-      FROM_HERE,
-      method_factory_.NewRunnableMethod(
-          &BlobURLRequestJob::DidResolve,
-          exists ? base::PLATFORM_FILE_OK : base::PLATFORM_FILE_ERROR_NOT_FOUND,
-          file_info));
+  base::FileUtilProxy::GetFileInfo(
+        file_thread_proxy_,
+        file_path,
+        callback_factory_.NewCallback(&BlobURLRequestJob::DidResolve));
 }
 
 void BlobURLRequestJob::DidResolve(base::PlatformFileError rv,
                                    const base::PlatformFileInfo& file_info) {
   // If an error occured, bail out.
   if (rv == base::PLATFORM_FILE_ERROR_NOT_FOUND) {
     NotifyFailure(net::ERR_FILE_NOT_FOUND);
     return;
   } else if (rv != base::PLATFORM_FILE_OK) {
     NotifyFailure(net::ERR_FAILED);
@@ skipping 114 matching lines @@
   // Read until we encounter an error or could not get the data immediately.
   while (remaining_bytes_ > 0 && read_buf_remaining_bytes_ > 0) {
     if (!ReadItem())
       return false;
   }
 
   *bytes_read = ReadCompleted();
   return true;
 }
 
+int BlobURLRequestJob::ComputeBytesToRead() const {
+  int64 current_item_remaining_bytes =
+      item_length_list_[item_index_] - current_item_offset_;
+  int bytes_to_read = (read_buf_remaining_bytes_ > current_item_remaining_bytes)
+      ? static_cast<int>(current_item_remaining_bytes)
+      : read_buf_remaining_bytes_;
+  if (bytes_to_read > remaining_bytes_)
+    bytes_to_read = static_cast<int>(remaining_bytes_);
+  return bytes_to_read;
+}
+
 bool BlobURLRequestJob::ReadItem() {
   // Are we done with reading all the blob data?
   if (remaining_bytes_ == 0)
     return true;
 
   // If we get to the last item but still expect something to read, bail out
   // since something is wrong.
   if (item_index_ >= blob_data_->items().size()) {
     NotifyFailure(net::ERR_FAILED);
     return false;
   }
 
-  const BlobData::Item& item = blob_data_->items().at(item_index_);
-
   // Compute the bytes to read for current item.
-  int64 current_item_remaining_bytes =
-      item_length_list_[item_index_] - current_item_offset_;
-  int bytes_to_read = (read_buf_remaining_bytes_ > current_item_remaining_bytes)
-      ? static_cast<int>(current_item_remaining_bytes)
-      : read_buf_remaining_bytes_;
-  if (bytes_to_read > remaining_bytes_)
-    bytes_to_read = static_cast<int>(remaining_bytes_);
+  bytes_to_read_ = ComputeBytesToRead();
 
   // If nothing to read for current item, advance to next item.
-  if (bytes_to_read == 0) {
+  if (bytes_to_read_ == 0) {
     AdvanceItem();
     return ReadItem();
   }
 
   // Do the reading.
+  const BlobData::Item& item = blob_data_->items().at(item_index_);
   switch (item.type()) {
     case BlobData::TYPE_DATA:
-      return ReadBytes(item, bytes_to_read);
+      return ReadBytes(item);
     case BlobData::TYPE_FILE:
-      return ReadFile(item, bytes_to_read);
+      return DispatchReadFile(item);
     default:
       DCHECK(false);
       return false;
   }
 }
 
-bool BlobURLRequestJob::ReadBytes(const BlobData::Item& item,
-                                  int bytes_to_read) {
-  DCHECK(read_buf_remaining_bytes_ >= bytes_to_read);
+bool BlobURLRequestJob::ReadBytes(const BlobData::Item& item) {
+  DCHECK(read_buf_remaining_bytes_ >= bytes_to_read_);
 
   memcpy(read_buf_->data() + read_buf_offset_,
          &item.data().at(0) + item.offset() + current_item_offset_,
-         bytes_to_read);
+         bytes_to_read_);
 
-  AdvanceBytesRead(bytes_to_read);
+  AdvanceBytesRead(bytes_to_read_);
   return true;
 }
 
-bool BlobURLRequestJob::ReadFile(const BlobData::Item& item,
-                                 int bytes_to_read) {
-  DCHECK(read_buf_remaining_bytes_ >= bytes_to_read);
+bool BlobURLRequestJob::DispatchReadFile(const BlobData::Item& item) {
+  // If the stream already exists, keep reading from it.
+  if (stream_ != NULL)
+    return ReadFile(item);
 
-  // Open the file if not yet.
-  if (!stream_.IsOpen()) {
-    int rv = stream_.Open(item.file_path(), base::PLATFORM_FILE_OPEN |
-        base::PLATFORM_FILE_READ | base::PLATFORM_FILE_ASYNC);
-    if (rv != net::OK) {
-      NotifyFailure(net::ERR_FAILED);
-      return false;
-    }
+  base::FileUtilProxy::CreateOrOpen(
+      file_thread_proxy_, item.file_path(), kFileOpenFlags,
+      callback_factory_.NewCallback(&BlobURLRequestJob::DidOpen));
+  SetStatus(net::URLRequestStatus(net::URLRequestStatus::IO_PENDING, 0));
+  return false;
+}
 
-    // Seek the file if needed.
-    int64 offset = current_item_offset_ + static_cast<int64>(item.offset());
-    if (offset > 0) {
-      if (offset != stream_.Seek(net::FROM_BEGIN, offset)) {
-        NotifyFailure(net::ERR_FAILED);
-        return false;
-      }
-    }
+void BlobURLRequestJob::DidOpen(base::PlatformFileError rv,
+                                base::PassPlatformFile file,
+                                bool created) {
+  if (rv != base::PLATFORM_FILE_OK) {
+    NotifyFailure(net::ERR_FAILED);
+    return;
   }
 
+  DCHECK(!stream_.get());
+  stream_.reset(new net::FileStream(file.ReleaseValue(), kFileOpenFlags));
+
+  // Seek the file if needed.
+  const BlobData::Item& item = blob_data_->items().at(item_index_);
+  int64 offset = current_item_offset_ + static_cast<int64>(item.offset());
+  if (offset > 0 && offset != stream_->Seek(net::FROM_BEGIN, offset)) {
+    NotifyFailure(net::ERR_FAILED);
+    return;
+  }
+
+  ReadFile(item);
+}
+
+bool BlobURLRequestJob::ReadFile(const BlobData::Item& item) {
+  DCHECK(stream_.get());
+  DCHECK(stream_->IsOpen());
+  DCHECK(read_buf_remaining_bytes_ >= bytes_to_read_);
+
   // Start the asynchronous reading.
-  int rv = stream_.Read(read_buf_->data() + read_buf_offset_,
-                        bytes_to_read,
-                        &io_callback_);
+  int rv = stream_->Read(read_buf_->data() + read_buf_offset_,
+                         bytes_to_read_,
+                         &io_callback_);
 
   // If I/O pending error is returned, we just need to wait.
   if (rv == net::ERR_IO_PENDING) {
     SetStatus(net::URLRequestStatus(net::URLRequestStatus::IO_PENDING, 0));
     return false;
   }
 
   // For all other errors, bail out.
   if (rv < 0) {
     NotifyFailure(net::ERR_FAILED);
     return false;
   }
 
   // Otherwise, data is immediately available.
-  AdvanceBytesRead(rv);
+  if (GetStatus().is_io_pending())
+    DidRead(rv);

[michaeln, 2011/03/11 06:03:47]

ah... is this what your new large test is about... if this also fixes a bug with
reading large blobs maybe mention that in the CL description too.

[adamk, 2011/03/14 17:42:14]

Added something to the CL description.  There wasn't a bug in the original code,
but my initial version of this change _did_ have a bug.  The test is a good idea
to leave in the CL nonetheless (unless you think I should split it into another
change).

+  else
+    AdvanceBytesRead(rv);
+
   return true;
 }
 
 void BlobURLRequestJob::DidRead(int result) {
   if (result < 0) {
     NotifyFailure(net::ERR_FAILED);
     return;
   }
   SetStatus(net::URLRequestStatus());  // Clear the IO_PENDING status
 
   AdvanceBytesRead(result);
 
   // If the read buffer is completely filled, we're done.
   if (!read_buf_remaining_bytes_) {
     int bytes_read = ReadCompleted();
     NotifyReadComplete(bytes_read);
     return;
   }
 
   // Otherwise, continue the reading.
   int bytes_read = 0;
   if (ReadLoop(&bytes_read))
     NotifyReadComplete(bytes_read);
 }
 
 void BlobURLRequestJob::AdvanceItem() {
   // Close the stream if the current item is a file.
-  if (stream_.IsOpen())
-    stream_.Close();
+  CloseStream();
 
   // Advance to the next item.
   item_index_++;
   current_item_offset_ = 0;
 }
 
 void BlobURLRequestJob::AdvanceBytesRead(int result) {
   DCHECK_GT(result, 0);
 
   // Do we finish reading the current item?
@@ skipping 145 matching lines @@
         // We don't support multiple range requests in one single URL request,
         // because we need to do multipart encoding here.
         // TODO(jianli): Support multipart byte range requests.
         NotifyFailure(net::ERR_REQUEST_RANGE_NOT_SATISFIABLE);
       }
     }
   }
 }
 
 }  // namespace webkit_blob