1// Copyright 2014 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_FILTER_GZIP_FILTER_H_
16#define NET_FILTER_GZIP_FILTER_H_
17
18#include "base/basictypes.h"
19#include "base/memory/scoped_ptr.h"
20#include "net/filter/filter.h"
21
22typedef struct z_stream_s z_stream;
23
24namespace net {
25
26class GZipHeader;
27
28class GZipFilter : public Filter {
29 public:
30  virtual ~GZipFilter();
31
32  // Initializes filter decoding mode and internal control blocks.
33  // Parameter filter_type specifies the type of filter, which corresponds to
34  // either gzip or deflate decoding. The function returns true if success and
35  // false otherwise.
36  // The filter can only be initialized once.
37  bool InitDecoding(Filter::FilterType filter_type);
38
39  // Decodes the pre-filter data and writes the output into the dest_buffer
40  // passed in.
41  // The function returns FilterStatus. See filter.h for its description.
42  //
43  // Upon entry, *dest_len is the total size (in number of chars) of the
44  // destination buffer. Upon exit, *dest_len is the actual number of chars
45  // written into the destination buffer.
46  //
47  // This function will fail if there is no pre-filter data in the
48  // stream_buffer_. On the other hand, *dest_len can be 0 upon successful
49  // return. For example, the internal zlib may process some pre-filter data
50  // but not produce output yet.
51  virtual FilterStatus ReadFilteredData(char* dest_buffer,
52                                        int* dest_len) OVERRIDE;
53
54 private:
55  enum DecodingStatus {
56    DECODING_UNINITIALIZED,
57    DECODING_IN_PROGRESS,
58    DECODING_DONE,
59    DECODING_ERROR
60  };
61
62  enum DecodingMode {
63    DECODE_MODE_GZIP,
64    DECODE_MODE_DEFLATE,
65    DECODE_MODE_UNKNOWN
66  };
67
68  enum GZipCheckHeaderState {
69    GZIP_CHECK_HEADER_IN_PROGRESS,
70    GZIP_GET_COMPLETE_HEADER,
71    GZIP_GET_INVALID_HEADER
72  };
73
74  static const int kGZipFooterSize = 8;
75
76  // Only to be instantiated by Filter::Factory.
77  GZipFilter();
78  friend class Filter;
79
80  // Parses and verifies the GZip header.
81  // Upon exit, the function updates gzip_header_status_ accordingly.
82  //
83  // The function returns Filter::FILTER_OK if it gets a complete header and
84  // there are more data in the pre-filter buffer.
85  // The function returns Filter::FILTER_NEED_MORE_DATA if it parses all data
86  // in the pre-filter buffer, either getting a complete header or a partial
87  // header. The caller needs to check gzip_header_status_ and call this
88  // function again for partial header.
89  // The function returns Filter::FILTER_ERROR if error occurs.
90  FilterStatus CheckGZipHeader();
91
92  // Internal function to decode the pre-filter data and writes the output into
93  // the dest_buffer passed in.
94  //
95  // This is the internal version of ReadFilteredData. See the latter's
96  // comments for the use of function.
97  FilterStatus DoInflate(char* dest_buffer, int* dest_len);
98
99  // Inserts a zlib header to the data stream before calling zlib inflate.
100  // This is used to work around server bugs. See more comments at the place
101  // it is called in gzip_filter.cc.
102  // The function returns true on success and false otherwise.
103  bool InsertZlibHeader();
104
105  // Skip the 8 byte GZip footer after z_stream_end
106  void SkipGZipFooter();
107
108  // Tracks the status of decoding.
109  // This variable is initialized by InitDecoding and updated only by
110  // ReadFilteredData.
111  DecodingStatus decoding_status_;
112
113  // Indicates the type of content decoding the GZipFilter is performing.
114  // This variable is set only once by InitDecoding.
115  DecodingMode decoding_mode_;
116
117  // Used to parse the gzip header in gzip stream.
118  // It is used when the decoding_mode_ is DECODE_MODE_GZIP.
119  scoped_ptr<GZipHeader> gzip_header_;
120
121  // Tracks the progress of parsing gzip header.
122  // This variable is maintained by gzip_header_.
123  GZipCheckHeaderState gzip_header_status_;
124
125  // A flag used by InsertZlibHeader to record whether we've successfully added
126  // a zlib header to this stream.
127  bool zlib_header_added_;
128
129  // Tracks how many bytes of gzip footer have been received.
130  int gzip_footer_bytes_;
131
132  // The control block of zlib which actually does the decoding.
133  // This data structure is initialized by InitDecoding and updated only by
134  // DoInflate, with InsertZlibHeader being the exception as a workaround.
135  scoped_ptr<z_stream> zlib_stream_;
136
137  // For robustness, when we see the solo sdch filter, we chain in a gzip filter
138  // in front of it, with this flag to indicate that the gzip decoding might not
139  // be needed.  This handles a strange case where "Content-Encoding: sdch,gzip"
140  // is reduced by an errant proxy to "Content-Encoding: sdch", while the
141  // content is indeed really gzipped result of sdch :-/.
142  // If this flag is set, then we will revert to being a pass through filter if
143  // we don't get a valid gzip header.
144  bool possible_sdch_pass_through_;
145
146  DISALLOW_COPY_AND_ASSIGN(GZipFilter);
147};
148
149}  // namespace net
150
151#endif  // NET_FILTER_GZIP_FILTER_H__
152