Home | History | Annotate | Download | only in base
      1 // Copyright (c) 2006-2008 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 // GZipFilter applies gzip and deflate content encoding/decoding to a data
      6 // stream. As specified by HTTP 1.1, with gzip encoding the content is
      7 // wrapped with a gzip header, and with deflate encoding the content is in
      8 // a raw, headerless DEFLATE stream.
      9 //
     10 // Internally GZipFilter uses zlib inflate to do decoding.
     11 //
     12 // GZipFilter is a subclass of Filter. See the latter's header file filter.h
     13 // for sample usage.
     14 
     15 #ifndef NET_BASE_GZIP_FILTER_H__
     16 #define NET_BASE_GZIP_FILTER_H__
     17 
     18 #include "base/scoped_ptr.h"
     19 #include "net/base/filter.h"
     20 
     21 class GZipHeader;
     22 typedef struct z_stream_s z_stream;
     23 
     24 class GZipFilter : public Filter {
     25  public:
     26   explicit GZipFilter(const FilterContext& filter_context);
     27 
     28   virtual ~GZipFilter();
     29 
     30   // Initializes filter decoding mode and internal control blocks.
     31   // Parameter filter_type specifies the type of filter, which corresponds to
     32   // either gzip or deflate decoding. The function returns true if success and
     33   // false otherwise.
     34   // The filter can only be initialized once.
     35   bool InitDecoding(Filter::FilterType filter_type);
     36 
     37   // Decodes the pre-filter data and writes the output into the dest_buffer
     38   // passed in.
     39   // The function returns FilterStatus. See filter.h for its description.
     40   //
     41   // Upon entry, *dest_len is the total size (in number of chars) of the
     42   // destination buffer. Upon exit, *dest_len is the actual number of chars
     43   // written into the destination buffer.
     44   //
     45   // This function will fail if there is no pre-filter data in the
     46   // stream_buffer_. On the other hand, *dest_len can be 0 upon successful
     47   // return. For example, the internal zlib may process some pre-filter data
     48   // but not produce output yet.
     49   virtual FilterStatus ReadFilteredData(char* dest_buffer, int* dest_len);
     50 
     51  private:
     52   enum DecodingStatus {
     53     DECODING_UNINITIALIZED,
     54     DECODING_IN_PROGRESS,
     55     DECODING_DONE,
     56     DECODING_ERROR
     57   };
     58 
     59   enum DecodingMode {
     60     DECODE_MODE_GZIP,
     61     DECODE_MODE_DEFLATE,
     62     DECODE_MODE_UNKNOWN
     63   };
     64 
     65   enum GZipCheckHeaderState {
     66     GZIP_CHECK_HEADER_IN_PROGRESS,
     67     GZIP_GET_COMPLETE_HEADER,
     68     GZIP_GET_INVALID_HEADER
     69   };
     70 
     71   static const int kGZipFooterSize = 8;
     72 
     73   // Parses and verifies the GZip header.
     74   // Upon exit, the function updates gzip_header_status_ accordingly.
     75   //
     76   // The function returns Filter::FILTER_OK if it gets a complete header and
     77   // there are more data in the pre-filter buffer.
     78   // The function returns Filter::FILTER_NEED_MORE_DATA if it parses all data
     79   // in the pre-filter buffer, either getting a complete header or a partial
     80   // header. The caller needs to check gzip_header_status_ and call this
     81   // function again for partial header.
     82   // The function returns Filter::FILTER_ERROR if error occurs.
     83   FilterStatus CheckGZipHeader();
     84 
     85   // Internal function to decode the pre-filter data and writes the output into
     86   // the dest_buffer passed in.
     87   //
     88   // This is the internal version of ReadFilteredData. See the latter's
     89   // comments for the use of function.
     90   FilterStatus DoInflate(char* dest_buffer, int* dest_len);
     91 
     92   // Inserts a zlib header to the data stream before calling zlib inflate.
     93   // This is used to work around server bugs. See more comments at the place
     94   // it is called in gzip_filter.cc.
     95   // The function returns true on success and false otherwise.
     96   bool InsertZlibHeader();
     97 
     98   // Skip the 8 byte GZip footer after z_stream_end
     99   void SkipGZipFooter();
    100 
    101   // Tracks the status of decoding.
    102   // This variable is initialized by InitDecoding and updated only by
    103   // ReadFilteredData.
    104   DecodingStatus decoding_status_;
    105 
    106   // Indicates the type of content decoding the GZipFilter is performing.
    107   // This variable is set only once by InitDecoding.
    108   DecodingMode decoding_mode_;
    109 
    110   // Used to parse the gzip header in gzip stream.
    111   // It is used when the decoding_mode_ is DECODE_MODE_GZIP.
    112   scoped_ptr<GZipHeader> gzip_header_;
    113 
    114   // Tracks the progress of parsing gzip header.
    115   // This variable is maintained by gzip_header_.
    116   GZipCheckHeaderState gzip_header_status_;
    117 
    118   // A flag used by InsertZlibHeader to record whether we've successfully added
    119   // a zlib header to this stream.
    120   bool zlib_header_added_;
    121 
    122   // Tracks how many bytes of gzip footer have been received.
    123   int gzip_footer_bytes_;
    124 
    125   // The control block of zlib which actually does the decoding.
    126   // This data structure is initialized by InitDecoding and updated only by
    127   // DoInflate, with InsertZlibHeader being the exception as a workaround.
    128   scoped_ptr<z_stream> zlib_stream_;
    129 
    130   // For robustness, when we see the solo sdch filter, we chain in a gzip filter
    131   // in front of it, with this flag to indicate that the gzip decoding might not
    132   // be needed.  This handles a strange case where "Content-Encoding: sdch,gzip"
    133   // is reduced by an errant proxy to "Content-Encoding: sdch", while the
    134   // content is indeed really gzipped result of sdch :-/.
    135   // If this flag is set, then we will revert to being a pass through filter if
    136   // we don't get a valid gzip header.
    137   bool possible_sdch_pass_through_;
    138 
    139   DISALLOW_EVIL_CONSTRUCTORS(GZipFilter);
    140 };
    141 
    142 #endif  // NET_BASE_GZIP_FILTER_H__
    143