Refactor BaseFile class to support sparse files

content/browser/download/base_file.h

 // 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.
 
 #ifndef CONTENT_BROWSER_DOWNLOAD_BASE_FILE_H_
 #define CONTENT_BROWSER_DOWNLOAD_BASE_FILE_H_
 
 #include <stddef.h>
 #include <stdint.h>
 
@@ skipping 43 matching lines @@
   //     location as determined by ContentBrowserClient.
   //
   // |file|: The base::File handle to use. If specified, BaseFile will not open
   //     a file and will use this handle. The file should be opened for both
   //     read and write. Only makes sense if |full_path| is non-empty since it
   //     implies that the caller already knows the path to the file.  There's no
   //     perfect way to come up with a canonical path for a file. So BaseFile
   //     will not attempt to determine the |full_path|.
   //
   // |bytes_so_far|: If a file is provided (via |full_path| or |file|), then
-  //     this argument specifies the size of the file to expect. It is legal for
-  //     the file to be larger, in which case the file will be truncated down to
-  //     this size. However, if the file is shorter, then the operation will
+  //     this argument specifies the amount of data that has been written to
+  //     the file. If |is_sparse_file| is false, this value should be the size
+  //     of the file to expect. It is legal for the file to be larger, in which
+  //     case the file will be truncated down to this size if |is_sparse_file|
+  //     is false. However, if the file is shorter, then the operation will
   //     fail with DOWNLOAD_INTERRUPT_REASON_FILE_TOO_SHORT.
+
   //
-  // |hash_so_far|: If |bytes_so_far| is non-zero, this specifies the SHA-256
-  //     hash of the first |bytes_so_far| bytes of the target file. If
-  //     specified, BaseFile will read the first |bytes_so_far| of the target
-  //     file in order to calculate the hash and verify that the file matches.
-  //     If there's a mismatch, then the operation fails with
-  //     DOWNLOAD_INTERRUPT_REASON_FILE_HASH_MISMATCH. Not used if |hash_state|
-  //     is also specified.
+  // |hash_so_far|: If |bytes_so_far| is non-zero and |is_sparse_file| is
+  //     false, this specifies the SHA-256 hash of the first |bytes_so_far|
+  //     bytes of the target file. If specified, BaseFile will read the first
+  //     |bytes_so_far| of the target file in order to calculate the hash and
+  //     verify that the file matches. If there's a mismatch, then the operation
+  //     fails with DOWNLOAD_INTERRUPT_REASON_FILE_HASH_MISMATCH. Not used if
+  //     |hash_state| is also specified.
   //
   // |hash_state|: The partial hash object to use. Only meaningful if there's a
   //     preexisting target file and it is non-empty (i.e. bytes_so_far is
   //     non-zero). If specified, BaseFile will assume that the bytes up to
   //     |bytes_so_far| has been accurately hashed into |hash_state| and will
-  //     ignore |hash_so_far|.
+  //     ignore |hash_so_far|. Not used if |is_sparse_file| is true.
+  //
+  // |is_sparse_file|: Specifies whether the file is a sparse file. If so, it is
+  //     possible that a write can happen at an offset that is larger than the
+  //     file size, thus creating holes in it.
   DownloadInterruptReason Initialize(
       const base::FilePath& full_path,
       const base::FilePath& default_directory,
       base::File file,
       int64_t bytes_so_far,
       const std::string& hash_so_far,
-      std::unique_ptr<crypto::SecureHash> hash_state);
+      std::unique_ptr<crypto::SecureHash> hash_state,
+      bool is_sparse_file);
+
+  // Write a new chunk of data to the file. Returns a DownloadInterruptReason
+  // indicating the result of the operation. Works only if |is_sparse_file| is
+  // false.
+  DownloadInterruptReason AppendDataToFile(const char* data, size_t data_len);
 
   // Write a new chunk of data to the file. Returns a DownloadInterruptReason
   // indicating the result of the operation.
-  DownloadInterruptReason AppendDataToFile(const char* data, size_t data_len);
+  DownloadInterruptReason WriteDataToFile(
+      int64_t offset,
+      const char* data,
+      size_t data_len);
 
   // Rename the download file. Returns a DownloadInterruptReason indicating the
   // result of the operation. A return code of NONE indicates that the rename
   // was successful. After a failure, the full_path() and in_progress() can be
   // used to determine the last known filename and whether the file is available
   // for writing or retrying the rename. Call Finish() to obtain the last known
   // hash state.
   DownloadInterruptReason Rename(const base::FilePath& full_path);
 
   // Mark the file as detached. Up until this method is called, BaseFile assumes
   // ownership of the file and hence will delete the file if the BaseFile object
   // is destroyed. Calling Detach() causes BaseFile to assume that it no longer
   // owns the file. Detach() can be called at any time. Close() must still be
   // called to close the file if it is open.
   void Detach();
 
   // Abort the download and automatically close and delete the file.
   void Cancel();
 
   // Indicate that the download has finished. No new data will be received.
   // Returns the SecureHash object representing the state of the hash function
-  // at the end of the operation.
+  // at the end of the operation. If |is_sparse_file_| is true, calling this
+  // will cause |secure_hash_| to get calculated.
   std::unique_ptr<crypto::SecureHash> Finish();
 
   // Informs the OS that this file came from the internet. Returns a
   // DownloadInterruptReason indicating the result of the operation.
   //
   // |client_guid|: The client GUID which will be used to identify the caller to
   //     the system AV scanning function.
   //
   // |source_url| / |referrer_url|: Source and referrer for the network request
   //     that originated this download. Will be used to annotate source
   //     information and also to determine the relative danger level of the
   //     file.
   DownloadInterruptReason AnnotateWithSourceInformation(
       const std::string& client_guid,
       const GURL& source_url,
       const GURL& referrer_url);
 
   // Returns the last known path to the download file. Can be empty if there's
   // no file.
   const base::FilePath& full_path() const { return full_path_; }
 
   // Returns true if the file is open. If true, the file can be written to or
   // renamed.
   bool in_progress() const { return file_.IsValid(); }
 
-  // Returns the number of bytes in the file pointed to by full_path().
+  // Returns the number of bytes that has been written so far. If
+  // |is_sparse_file_| is false, this should always be equal to the file size.
+  // If |is_sparse_file_| is true, this should not be larger than the file size
+  // as the file may contain holes in it.
   int64_t bytes_so_far() const { return bytes_so_far_; }
 
   std::string DebugString() const;
 
  private:
   friend class BaseFileTest;
   FRIEND_TEST_ALL_PREFIXES(BaseFileTest, IsEmptyHash);
 
   // Creates and opens the file_ if it is invalid.
   //
-  // If |hash_so_far| is not empty, then it must match the SHA-256 hash of the
-  // first |bytes_so_far_| bytes of |file_|. If there's a hash mismatch, Open()
-  // fails with DOWNLOAD_INTERRUPT_REASON_FILE_HASH_MISMATCH.
+  // If |is_sparse_file_| is false and |hash_so_far| is not empty, then it must
+  // match the SHA-256 hash of the first |bytes_so_far_| bytes of |file_|. If
+  // there's a hash mismatch, Open() fails with
+  // DOWNLOAD_INTERRUPT_REASON_FILE_HASH_MISMATCH.
   //
   // If the opened file is shorter than |bytes_so_far_| bytes, then Open() fails
-  // with DOWNLOAD_INTERRUPT_REASON_FILE_TOO_SHORT. If the opened file is
-  // longer, then the file is truncated to |bytes_so_far_|.
+  // with DOWNLOAD_INTERRUPT_REASON_FILE_TOO_SHORT. If the opened file is longer
+  // and |is_sparse_file_| is false, then the file is truncated to
+  // |bytes_so_far_|.
   //
   // Open() can fail for other reasons as well. In that case, it returns a
   // relevant interrupt reason. Unless Open() return
   // DOWNLOAD_INTERRUPT_REASON_NONE, it should be assumed that |file_| is not
   // valid.
   DownloadInterruptReason Open(const std::string& hash_so_far);
 
   // Closes and resets file_.
   void Close();
 
@@ skipping 10 matching lines @@
   int64_t CurrentSpeedAtTime(base::TimeTicks current_time) const;
 
   // Verifies that:
   // * Size of the file represented by |file_| is at least |bytes_so_far_|.
   //
   // * If |hash_to_expect| is not empty, then the result of hashing the first
   //   |bytes_so_far_| bytes of |file_| matches |hash_to_expect|.
   //
   // If the result is REASON_NONE, then on return |secure_hash_| is valid and
   // is ready to hash bytes from offset |bytes_so_far_| + 1.
+  // If |is_sparse_file_| is true, this function is only called when Finish()
+  // is called.
   DownloadInterruptReason CalculatePartialHash(
       const std::string& hash_to_expect);
 
   // Log a TYPE_DOWNLOAD_FILE_ERROR NetLog event with |error| and passes error
   // on through, converting to a |DownloadInterruptReason|.
   DownloadInterruptReason LogNetError(const char* operation, net::Error error);
 
   // Log the system error in |os_error| and converts it into a
   // |DownloadInterruptReason|.
   DownloadInterruptReason LogSystemError(const char* operation,
@@ skipping 17 matching lines @@
   // Used to calculate hash for the file when calculate_hash_ is set.
   std::unique_ptr<crypto::SecureHash> secure_hash_;
 
   // Start time for calculating speed.
   base::TimeTicks start_tick_;
 
   // Indicates that this class no longer owns the associated file, and so
   // won't delete it on destruction.
   bool detached_ = false;
 
+  // Whether the file is sparse.
+  // TODO(qinmin): pass the slice information to this class so that we can
+  // verify that writes are not overlapping.
+  bool is_sparse_file_ = false;
+
   net::NetLogWithSource net_log_;
 
   DISALLOW_COPY_AND_ASSIGN(BaseFile);
 };
 
 }  // namespace content
 
 #endif  // CONTENT_BROWSER_DOWNLOAD_BASE_FILE_H_