Avoid disk accesses on the wrong thread in URLRequestFileJob

net/url_request/url_request_file_job.cc

 // Copyright (c) 2012 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.
 
 // For loading files, we make use of overlapped i/o to ensure that reading from
 // the filesystem (e.g., a network filesystem) does not block the calling
 // thread.  An alternative approach would be to use a background thread or pool
 // of threads, but it seems better to leverage the operating system's ability
 // to do background file reads for us.
 //
@@ skipping 24 matching lines @@
 #include "net/base/net_errors.h"
 #include "net/base/net_util.h"
 #include "net/http/http_util.h"
 #include "net/url_request/url_request.h"
 #include "net/url_request/url_request_context.h"
 #include "net/url_request/url_request_error_job.h"
 #include "net/url_request/url_request_file_dir_job.h"
 
 namespace net {
 
-class URLRequestFileJob::AsyncResolver
-    : public base::RefCountedThreadSafe<URLRequestFileJob::AsyncResolver> {
- public:
-  explicit AsyncResolver(URLRequestFileJob* owner)
-      : owner_(owner), owner_loop_(MessageLoop::current()) {
-  }
+namespace {
 
-  void Resolve(const FilePath& file_path) {
-    base::PlatformFileInfo file_info;
-    bool exists = file_util::GetFileInfo(file_path, &file_info);
-    base::AutoLock locked(lock_);
-    if (owner_loop_) {
-      owner_loop_->PostTask(
-          FROM_HERE,
-          base::Bind(&AsyncResolver::ReturnResults, this, exists, file_info));
-    }
-  }
+void ResolveAndGetMimeType(const FilePath& file_path,
+                           bool* exists,
+                           base::PlatformFileInfo* file_info,
+                           bool* mime_type_result,
+                           std::string* mime_type) {
+  *exists = file_util::GetFileInfo(file_path, file_info);
+  // On Windows GetMimeTypeFromFile() goes to the registry. Thus it should be
+  // done in WorkerPool.
+  *mime_type_result = GetMimeTypeFromFile(file_path, mime_type);
+}
 
-  void Cancel() {
-    owner_ = NULL;
+}
 
-    base::AutoLock locked(lock_);
-    owner_loop_ = NULL;
-  }
-
- private:
-  friend class base::RefCountedThreadSafe<URLRequestFileJob::AsyncResolver>;
-
-  ~AsyncResolver() {}
-
-  void ReturnResults(bool exists, const base::PlatformFileInfo& file_info) {
-    if (owner_)
-      owner_->DidResolve(exists, file_info);
-  }
-
-  URLRequestFileJob* owner_;
-
-  base::Lock lock_;
-  MessageLoop* owner_loop_;
-};
 
 URLRequestFileJob::URLRequestFileJob(URLRequest* request,
                                      const FilePath& file_path)
     : URLRequestJob(request, request->context()->network_delegate()),
       file_path_(file_path),
       stream_(NULL),
       is_directory_(false),
-      remaining_bytes_(0) {
+      mime_type_result_(false),
+      file_size_(-1),
+      remaining_bytes_(0),
+      weak_ptr_factory_(ALLOW_THIS_IN_INITIALIZER_LIST(this)) {
 }
 
 // static
 URLRequestJob* URLRequestFileJob::Factory(URLRequest* request,
                                           const std::string& scheme) {
   FilePath file_path;
   const bool is_file = FileURLToFilePath(request->url(), &file_path);
 
   // Check file access permissions.
   if (!IsFileAccessAllowed(*request, file_path))
     return new URLRequestErrorJob(request, ERR_ACCESS_DENIED);
 
   // We need to decide whether to create URLRequestFileJob for file access or
   // URLRequestFileDirJob for directory access. To avoid accessing the
   // filesystem, we only look at the path string here.
   // The code in the URLRequestFileJob::Start() method discovers that a path,
   // which doesn't end with a slash, should really be treated as a directory,
   // and it then redirects to the URLRequestFileDirJob.
   if (is_file &&
       file_util::EndsWithSeparator(file_path) &&
       file_path.IsAbsolute())
     return new URLRequestFileDirJob(request, file_path);
 
   // Use a regular file request job for all non-directories (including invalid
   // file names).
   return new URLRequestFileJob(request, file_path);
 }
 
 void URLRequestFileJob::Start() {
-  DCHECK(!async_resolver_);
-  async_resolver_ = new AsyncResolver(this);
-  base::WorkerPool::PostTask(
+  bool* exists = new bool;
+  base::PlatformFileInfo* file_info = new base::PlatformFileInfo;

[eroman, 2012/07/12 19:13:16]

This feels a bit crowded now with the extra parameters.

I suggest packaging them into a single struct, and then just doing one
allocation. For instance:

struct ResolveAndGetMimeTypeResult {
  bool exists;
  base::PlatformFileInfo file_info;
  bool mime_type_result;
  std::string read_mime_type;
};

You could also use this struct to replace the following member variables:

  bool is_directory_;
  bool mime_type_result_;
  std::string mime_type_;
  int64 file_size_;

Since it would be sufficient to have:
  ResolveAndGetMimeTypeResult resolve_result_;


(Note, I'm not sure if PlatformFileInfo supports default assignment. If it
doesn't might want to have the struct simply encapsulate {is_directory,
file_size} instead).

+  bool* mime_type_result = new bool;
+  std::string* read_mime_type = new std::string;
+  base::WorkerPool::PostTaskAndReply(
       FROM_HERE,
-      base::Bind(&AsyncResolver::Resolve, async_resolver_.get(), file_path_),
+      base::Bind(&ResolveAndGetMimeType, file_path_,
+                 base::Unretained(exists),
+                 base::Unretained(file_info),
+                 base::Unretained(mime_type_result),
+                 base::Unretained(read_mime_type)),
+      base::Bind(&URLRequestFileJob::DidResolveAndMimeType,
+                 weak_ptr_factory_.GetWeakPtr(),
+                 base::Owned(exists),
+                 base::Owned(file_info),
+                 base::Owned(mime_type_result),
+                 base::Owned(read_mime_type)),
       true);
 }
 
 void URLRequestFileJob::Kill() {
-  // URL requests should not block on the disk!
-  //   http://code.google.com/p/chromium/issues/detail?id=59849
-  base::ThreadRestrictions::ScopedAllowIO allow_io;
-  stream_.CloseSync();
-
-  if (async_resolver_) {
-    async_resolver_->Cancel();
-    async_resolver_ = NULL;
-  }
+  stream_.CloseAndCancelAsync();
+  weak_ptr_factory_.InvalidateWeakPtrs();
 
   URLRequestJob::Kill();
 }
 
 bool URLRequestFileJob::ReadRawData(IOBuffer* dest, int dest_size,
                                     int *bytes_read) {
   DCHECK_NE(dest_size, 0);
   DCHECK(bytes_read);
   DCHECK_GE(remaining_bytes_, 0);
 
   if (remaining_bytes_ < dest_size)
     dest_size = static_cast<int>(remaining_bytes_);
 
   // If we should copy zero bytes because |remaining_bytes_| is zero, short
   // circuit here.
   if (!dest_size) {
     *bytes_read = 0;
     return true;
   }
 
   int rv = stream_.Read(dest, dest_size,
                         base::Bind(&URLRequestFileJob::DidRead,
-                                   base::Unretained(this)));
+                                   weak_ptr_factory_.GetWeakPtr()));
   if (rv >= 0) {
     // Data is immediately available.
     *bytes_read = rv;
     remaining_bytes_ -= rv;
     DCHECK_GE(remaining_bytes_, 0);
     return true;
   }
 
   // Otherwise, a read error occured.  We may just need to wait...
   if (rv == ERR_IO_PENDING) {
@@ skipping 41 matching lines @@
 #endif
 }
 
 Filter* URLRequestFileJob::SetupFilter() const {
   // Bug 9936 - .svgz files needs to be decompressed.
   return LowerCaseEqualsASCII(file_path_.Extension(), ".svgz")
       ? Filter::GZipFactory() : NULL;
 }
 
 bool URLRequestFileJob::GetMimeType(std::string* mime_type) const {
-  // URL requests should not block on the disk!  On Windows this goes to the
-  // registry.
-  //   http://code.google.com/p/chromium/issues/detail?id=59849
-  base::ThreadRestrictions::ScopedAllowIO allow_io;
   DCHECK(request_);
-  return GetMimeTypeFromFile(file_path_, mime_type);
+  *mime_type = mime_type_;
+  return mime_type_result_;
 }
 
 void URLRequestFileJob::SetExtraRequestHeaders(
     const HttpRequestHeaders& headers) {
   std::string range_header;
   if (headers.GetHeader(HttpRequestHeaders::kRange, &range_header)) {
     // We only care about "Range" header here.
     std::vector<HttpByteRange> ranges;
     if (HttpUtil::ParseRangeHeader(range_header, &ranges)) {
       if (ranges.size() == 1) {
@@ skipping 16 matching lines @@
   const URLRequestContext* context = request.context();
   if (!context)
     return false;
   const NetworkDelegate* delegate = context->network_delegate();
   if (delegate)
     return delegate->CanAccessFile(request, path);
   return false;
 }
 
 URLRequestFileJob::~URLRequestFileJob() {
-  DCHECK(!async_resolver_);
 }
 
-void URLRequestFileJob::DidResolve(
-    bool exists, const base::PlatformFileInfo& file_info) {
-  async_resolver_ = NULL;
+void URLRequestFileJob::DidResolveAndMimeType(
+      const bool* exists,
+      const base::PlatformFileInfo* file_info,
+      const bool* mime_type_result,
+      const std::string* read_mime_type) {
+  is_directory_ = file_info->is_directory;
+  file_size_ = file_info->size;
+  mime_type_result_ = *mime_type_result;
+  mime_type_ = *read_mime_type;
 
-  // We may have been orphaned...
-  if (!request_)
-    return;
-
-  is_directory_ = file_info.is_directory;
-
-  int rv = OK;
   // We use URLRequestFileJob to handle files as well as directories without
   // trailing slash.
   // If a directory does not exist, we return ERR_FILE_NOT_FOUND. Otherwise,
   // we will append trailing slash and redirect to FileDirJob.
   // A special case is "\" on Windows. We should resolve as invalid.
   // However, Windows resolves "\" to "C:\", thus reports it as existent.
   // So what happens is we append it with trailing slash and redirect it to
   // FileDirJob where it is resolved as invalid.
-  if (!exists) {
-    rv = ERR_FILE_NOT_FOUND;
-  } else if (!is_directory_) {
-    // URL requests should not block on the disk!
-    //   http://code.google.com/p/chromium/issues/detail?id=59849
-    base::ThreadRestrictions::ScopedAllowIO allow_io;
-
-    int flags = base::PLATFORM_FILE_OPEN |
-                base::PLATFORM_FILE_READ |
-                base::PLATFORM_FILE_ASYNC;
-    rv = stream_.OpenSync(file_path_, flags);
+  if (!(*exists)) {
+    DidOpen(ERR_FILE_NOT_FOUND);
+    return;
   }
-
-  if (rv != OK) {
-    NotifyDone(URLRequestStatus(URLRequestStatus::FAILED, rv));
+  if (is_directory_) {
+    DidOpen(OK);
     return;
   }
 
-  if (!byte_range_.ComputeBounds(file_info.size)) {
+  int flags = base::PLATFORM_FILE_OPEN |
+              base::PLATFORM_FILE_READ |
+              base::PLATFORM_FILE_ASYNC;
+  int rv = stream_.Open(file_path_, flags,
+                        base::Bind(&URLRequestFileJob::DidOpen,
+                                   weak_ptr_factory_.GetWeakPtr()));
+  if (rv != ERR_IO_PENDING)
+    DidOpen(rv);
+}
+
+void URLRequestFileJob::DidOpen(int result) {
+  if (result != OK) {
+    NotifyDone(URLRequestStatus(URLRequestStatus::FAILED, result));
+    return;
+  }
+
+  if (!byte_range_.ComputeBounds(file_size_)) {
     NotifyDone(URLRequestStatus(URLRequestStatus::FAILED,
                ERR_REQUEST_RANGE_NOT_SATISFIABLE));
     return;
   }
 
   remaining_bytes_ = byte_range_.last_byte_position() -
                      byte_range_.first_byte_position() + 1;
   DCHECK_GE(remaining_bytes_, 0);
 
-  // URL requests should not block on the disk!
-  //   http://code.google.com/p/chromium/issues/detail?id=59849
-  {
-    base::ThreadRestrictions::ScopedAllowIO allow_io;
-    // Do the seek at the beginning of the request.
-    if (remaining_bytes_ > 0 &&
-        byte_range_.first_byte_position() != 0 &&
-        byte_range_.first_byte_position() !=
-        stream_.SeekSync(FROM_BEGIN, byte_range_.first_byte_position())) {
-      NotifyDone(URLRequestStatus(URLRequestStatus::FAILED,
-                                  ERR_REQUEST_RANGE_NOT_SATISFIABLE));
-      return;
-    }
+  if (remaining_bytes_ > 0 && byte_range_.first_byte_position() != 0) {
+    int rv = stream_.Seek(FROM_BEGIN, byte_range_.first_byte_position(),
+                          base::Bind(&URLRequestFileJob::DidSeek,
+                                     weak_ptr_factory_.GetWeakPtr()));
+    if (rv != ERR_IO_PENDING)
+      DidSeek(-1);
+  } else {
+    DidSeek(byte_range_.first_byte_position());
+  }
+}
+
+void URLRequestFileJob::DidSeek(int64 result) {
+  if (result != byte_range_.first_byte_position()) {
+    NotifyDone(URLRequestStatus(URLRequestStatus::FAILED,
+                                ERR_REQUEST_RANGE_NOT_SATISFIABLE));
+    return;
   }
 
   set_expected_content_size(remaining_bytes_);
   NotifyHeadersComplete();
 }
 
 void URLRequestFileJob::DidRead(int result) {
   if (result > 0) {
     SetStatus(URLRequestStatus());  // Clear the IO_PENDING status
   } else if (result == 0) {
     NotifyDone(URLRequestStatus());
   } else {
     NotifyDone(URLRequestStatus(URLRequestStatus::FAILED, result));
   }
 
   remaining_bytes_ -= result;
   DCHECK_GE(remaining_bytes_, 0);
 
   NotifyReadComplete(result);
 }
 
 }  // namespace net