| OLD | NEW |
| (Empty) |
| 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 // Filter performs filtering on data streams. Sample usage: | |
| 6 // | |
| 7 // IStream* pre_filter_source; | |
| 8 // ... | |
| 9 // Filter* filter = Filter::Factory(filter_type, size); | |
| 10 // int pre_filter_data_len = filter->stream_buffer_size(); | |
| 11 // pre_filter_source->read(filter->stream_buffer(), pre_filter_data_len); | |
| 12 // | |
| 13 // filter->FlushStreamBuffer(pre_filter_data_len); | |
| 14 // | |
| 15 // char post_filter_buf[kBufferSize]; | |
| 16 // int post_filter_data_len = kBufferSize; | |
| 17 // filter->ReadFilteredData(post_filter_buf, &post_filter_data_len); | |
| 18 // | |
| 19 // To filter a data stream, the caller first gets filter's stream_buffer_ | |
| 20 // through its accessor and fills in stream_buffer_ with pre-filter data, next | |
| 21 // calls FlushStreamBuffer to notify Filter, then calls ReadFilteredData | |
| 22 // repeatedly to get all the filtered data. After all data have been filtered | |
| 23 // and read out, the caller may fill in stream_buffer_ again. This | |
| 24 // WriteBuffer-Flush-Read cycle is repeated until reaching the end of data | |
| 25 // stream. | |
| 26 // | |
| 27 // A return of FILTER_OK from ReadData() means that more data is | |
| 28 // available to a future ReadData() call and data may not be written | |
| 29 // into stream_buffer(). A return of FILTER_NEED_MORE_DATA from ReadData() | |
| 30 // indicates that no data will be forthcoming from the filter until | |
| 31 // it receives more input data, and that the buffer at | |
| 32 // stream_buffer() may be written to. | |
| 33 // | |
| 34 // The filter being complete (no more data to provide) may be indicated | |
| 35 // by either returning FILTER_DONE or by returning FILTER_OK and indicating | |
| 36 // zero bytes output; consumers understand both those signals. Consumers | |
| 37 // are responsible for not calling ReadData() on a filter after one of these | |
| 38 // signals have been returned. Note that some filters may never signal that | |
| 39 // they are done (e.g. a pass-through filter will always | |
| 40 // say FILTER_NEED_MORE_DATA), so the consumer will also need to | |
| 41 // recognize the state of |no_more_input_data_available && | |
| 42 // filter->stream_data_len() == 0| as FILTER_DONE. | |
| 43 // | |
| 44 // The lifetime of a Filter instance is completely controlled by its caller. | |
| 45 | |
| 46 #ifndef NET_FILTER_FILTER_H__ | |
| 47 #define NET_FILTER_FILTER_H__ | |
| 48 | |
| 49 #include <string> | |
| 50 #include <vector> | |
| 51 | |
| 52 #include "base/basictypes.h" | |
| 53 #include "base/gtest_prod_util.h" | |
| 54 #include "base/memory/ref_counted.h" | |
| 55 #include "base/memory/scoped_ptr.h" | |
| 56 #include "base/time/time.h" | |
| 57 #include "net/base/net_export.h" | |
| 58 #include "net/base/sdch_manager.h" | |
| 59 | |
| 60 class GURL; | |
| 61 | |
| 62 namespace net { | |
| 63 | |
| 64 class BoundNetLog; | |
| 65 class IOBuffer; | |
| 66 class URLRequestContext; | |
| 67 | |
| 68 //------------------------------------------------------------------------------ | |
| 69 // Define an interface class that allows access to contextual information | |
| 70 // supplied by the owner of this filter. In the case where there are a chain of | |
| 71 // filters, there is only one owner of all the chained filters, and that context | |
| 72 // is passed to the constructor of all those filters. To be clear, the context | |
| 73 // does NOT reflect the position in a chain, or the fact that there are prior | |
| 74 // or later filters in a chain. | |
| 75 // | |
| 76 // TODO(rdsmith): FilterContext is a grab-bag of methods which may or may | |
| 77 // not be relevant for any particular filter, and it's getting worse over | |
| 78 // time. In addition, it only supports two filters, SDCH and gzip. | |
| 79 // It would make more sense to implement FilterContext as a | |
| 80 // base::SupportsUserData structure to which filter-specific information | |
| 81 // could be added by whatever the ultimate consumer of the filter chain is, | |
| 82 // and a particular filter (if included) could access that information. | |
| 83 class NET_EXPORT_PRIVATE FilterContext { | |
| 84 public: | |
| 85 // Enum to control what histograms are emitted near end-of-life of this | |
| 86 // instance. | |
| 87 enum StatisticSelector { | |
| 88 SDCH_DECODE, | |
| 89 SDCH_PASSTHROUGH, | |
| 90 SDCH_EXPERIMENT_DECODE, | |
| 91 SDCH_EXPERIMENT_HOLDBACK, | |
| 92 }; | |
| 93 | |
| 94 virtual ~FilterContext(); | |
| 95 | |
| 96 // What mime type was specified in the header for this data? | |
| 97 // Only makes senses for some types of contexts, and returns false | |
| 98 // when not applicable. | |
| 99 virtual bool GetMimeType(std::string* mime_type) const = 0; | |
| 100 | |
| 101 // What URL was used to access this data? | |
| 102 // Return false if gurl is not present. | |
| 103 virtual bool GetURL(GURL* gurl) const = 0; | |
| 104 | |
| 105 // What Content-Disposition header came with this data? | |
| 106 // Return false if no header was present. | |
| 107 virtual bool GetContentDisposition(std::string* disposition) const = 0; | |
| 108 | |
| 109 // When was this data requested from a server? | |
| 110 virtual base::Time GetRequestTime() const = 0; | |
| 111 | |
| 112 // Is data supplied from cache, or fresh across the net? | |
| 113 virtual bool IsCachedContent() const = 0; | |
| 114 | |
| 115 // Is this a download? | |
| 116 virtual bool IsDownload() const = 0; | |
| 117 | |
| 118 // Was this data flagged as a response to a request with an SDCH dictionary? | |
| 119 virtual SdchManager::DictionarySet* SdchDictionariesAdvertised() const = 0; | |
| 120 | |
| 121 // How many bytes were read from the net or cache so far (and potentially | |
| 122 // pushed into a filter for processing)? | |
| 123 virtual int64 GetByteReadCount() const = 0; | |
| 124 | |
| 125 // What response code was received with the associated network transaction? | |
| 126 // For example: 200 is ok. 4xx are error codes. etc. | |
| 127 virtual int GetResponseCode() const = 0; | |
| 128 | |
| 129 // The URLRequestContext associated with the request. | |
| 130 virtual const URLRequestContext* GetURLRequestContext() const = 0; | |
| 131 | |
| 132 // The following method forces the context to emit a specific set of | |
| 133 // statistics as selected by the argument. | |
| 134 virtual void RecordPacketStats(StatisticSelector statistic) const = 0; | |
| 135 | |
| 136 // The BoundNetLog of the associated request. | |
| 137 virtual const BoundNetLog& GetNetLog() const = 0; | |
| 138 }; | |
| 139 | |
| 140 //------------------------------------------------------------------------------ | |
| 141 class NET_EXPORT_PRIVATE Filter { | |
| 142 public: | |
| 143 // Return values of function ReadFilteredData. | |
| 144 enum FilterStatus { | |
| 145 // Read filtered data successfully | |
| 146 FILTER_OK, | |
| 147 // Read filtered data successfully, and the data in the buffer has been | |
| 148 // consumed by the filter, but more data is needed in order to continue | |
| 149 // filtering. At this point, the caller is free to reuse the filter | |
| 150 // buffer to provide more data. | |
| 151 FILTER_NEED_MORE_DATA, | |
| 152 // Read filtered data successfully, and filter reaches the end of the data | |
| 153 // stream. | |
| 154 FILTER_DONE, | |
| 155 // There is an error during filtering. | |
| 156 FILTER_ERROR | |
| 157 }; | |
| 158 | |
| 159 // Specifies type of filters that can be created. | |
| 160 enum FilterType { | |
| 161 FILTER_TYPE_DEFLATE, | |
| 162 FILTER_TYPE_GZIP, | |
| 163 FILTER_TYPE_GZIP_HELPING_SDCH, // Gzip possible, but pass through allowed. | |
| 164 FILTER_TYPE_SDCH, | |
| 165 FILTER_TYPE_SDCH_POSSIBLE, // Sdch possible, but pass through allowed. | |
| 166 FILTER_TYPE_UNSUPPORTED, | |
| 167 }; | |
| 168 | |
| 169 virtual ~Filter(); | |
| 170 | |
| 171 // Creates a Filter object. | |
| 172 // Parameters: Filter_types specifies the type of filter created; | |
| 173 // filter_context allows filters to acquire additional details needed for | |
| 174 // construction and operation, such as a specification of requisite input | |
| 175 // buffer size. | |
| 176 // If success, the function returns the pointer to the Filter object created. | |
| 177 // If failed or a filter is not needed, the function returns NULL. | |
| 178 // | |
| 179 // Note: filter_types is an array of filter types (content encoding types as | |
| 180 // provided in an HTTP header), which will be chained together serially to do | |
| 181 // successive filtering of data. The types in the vector are ordered based on | |
| 182 // encoding order, and the filters are chained to operate in the reverse | |
| 183 // (decoding) order. For example, types[0] = FILTER_TYPE_SDCH, | |
| 184 // types[1] = FILTER_TYPE_GZIP will cause data to first be gunzip filtered, | |
| 185 // and the resulting output from that filter will be sdch decoded. | |
| 186 static Filter* Factory(const std::vector<FilterType>& filter_types, | |
| 187 const FilterContext& filter_context); | |
| 188 | |
| 189 // A simpler version of Factory() which creates a single, unchained | |
| 190 // Filter of type FILTER_TYPE_GZIP, or NULL if the filter could not be | |
| 191 // initialized. | |
| 192 static Filter* GZipFactory(); | |
| 193 | |
| 194 // External call to obtain data from this filter chain. If ther is no | |
| 195 // next_filter_, then it obtains data from this specific filter. | |
| 196 FilterStatus ReadData(char* dest_buffer, int* dest_len); | |
| 197 | |
| 198 // Returns a pointer to the stream_buffer_. | |
| 199 IOBuffer* stream_buffer() const { return stream_buffer_.get(); } | |
| 200 | |
| 201 // Returns the maximum size of stream_buffer_ in number of chars. | |
| 202 int stream_buffer_size() const { return stream_buffer_size_; } | |
| 203 | |
| 204 // Returns the total number of chars remaining in stream_buffer_ to be | |
| 205 // filtered. | |
| 206 // | |
| 207 // If the function returns 0 then all data has been filtered, and the caller | |
| 208 // is safe to copy new data into stream_buffer_. | |
| 209 int stream_data_len() const { return stream_data_len_; } | |
| 210 | |
| 211 // Flushes stream_buffer_ for next round of filtering. After copying data to | |
| 212 // stream_buffer_, the caller should call this function to notify Filter to | |
| 213 // start filtering. Then after this function is called, the caller can get | |
| 214 // post-filtered data using ReadFilteredData. The caller must not write to | |
| 215 // stream_buffer_ and call this function again before stream_buffer_ is | |
| 216 // emptied out by ReadFilteredData. | |
| 217 // | |
| 218 // The input stream_data_len is the length (in number of chars) of valid | |
| 219 // data in stream_buffer_. It can not be greater than stream_buffer_size_. | |
| 220 // The function returns true if success, and false otherwise. | |
| 221 bool FlushStreamBuffer(int stream_data_len); | |
| 222 | |
| 223 // Translate the text of a filter name (from Content-Encoding header) into a | |
| 224 // FilterType. | |
| 225 static FilterType ConvertEncodingToType(const std::string& filter_type); | |
| 226 | |
| 227 // Given a array of encoding_types, try to do some error recovery adjustment | |
| 228 // to the list. This includes handling known bugs in the Apache server (where | |
| 229 // redundant gzip encoding is specified), as well as issues regarding SDCH | |
| 230 // encoding, where various proxies and anti-virus products modify or strip the | |
| 231 // encodings. These fixups require context, which includes whether this | |
| 232 // response was made to an SDCH request (i.e., an available dictionary was | |
| 233 // advertised in the GET), as well as the mime type of the content. | |
| 234 static void FixupEncodingTypes(const FilterContext& filter_context, | |
| 235 std::vector<FilterType>* encoding_types); | |
| 236 | |
| 237 // Returns a string describing the FilterTypes implemented by this filter. | |
| 238 std::string OrderedFilterList() const; | |
| 239 | |
| 240 protected: | |
| 241 friend class GZipUnitTest; | |
| 242 friend class SdchFilterChainingTest; | |
| 243 FRIEND_TEST_ALL_PREFIXES(FilterTest, ThreeFilterChain); | |
| 244 | |
| 245 explicit Filter(FilterType type_id); | |
| 246 | |
| 247 // Filters the data stored in stream_buffer_ and writes the output into the | |
| 248 // dest_buffer passed in. | |
| 249 // | |
| 250 // Upon entry, *dest_len is the total size (in number of chars) of the | |
| 251 // destination buffer. Upon exit, *dest_len is the actual number of chars | |
| 252 // written into the destination buffer. | |
| 253 // | |
| 254 // This function will fail if there is no pre-filter data in the | |
| 255 // stream_buffer_. On the other hand, *dest_len can be 0 upon successful | |
| 256 // return. For example, a decoding filter may process some pre-filter data | |
| 257 // but not produce output yet. | |
| 258 virtual FilterStatus ReadFilteredData(char* dest_buffer, int* dest_len) = 0; | |
| 259 | |
| 260 // Copy pre-filter data directly to destination buffer without decoding. | |
| 261 FilterStatus CopyOut(char* dest_buffer, int* dest_len); | |
| 262 | |
| 263 FilterStatus last_status() const { return last_status_; } | |
| 264 | |
| 265 // Buffer to hold the data to be filtered (the input queue). | |
| 266 scoped_refptr<IOBuffer> stream_buffer_; | |
| 267 | |
| 268 // Maximum size of stream_buffer_ in number of chars. | |
| 269 int stream_buffer_size_; | |
| 270 | |
| 271 // Pointer to the next data in stream_buffer_ to be filtered. | |
| 272 char* next_stream_data_; | |
| 273 | |
| 274 // Total number of remaining chars in stream_buffer_ to be filtered. | |
| 275 int stream_data_len_; | |
| 276 | |
| 277 private: | |
| 278 // Allocates and initializes stream_buffer_ and stream_buffer_size_. | |
| 279 void InitBuffer(int size); | |
| 280 | |
| 281 // A factory helper for creating filters for within a chain of potentially | |
| 282 // multiple encodings. If a chain of filters is created, then this may be | |
| 283 // called multiple times during the filter creation process. In most simple | |
| 284 // cases, this is only called once. Returns NULL and cleans up (deleting | |
| 285 // filter_list) if a new filter can't be constructed. | |
| 286 static Filter* PrependNewFilter(FilterType type_id, | |
| 287 const FilterContext& filter_context, | |
| 288 int buffer_size, | |
| 289 Filter* filter_list); | |
| 290 | |
| 291 // Helper methods for PrependNewFilter. If initialization is successful, | |
| 292 // they return a fully initialized Filter. Otherwise, return NULL. | |
| 293 static Filter* InitGZipFilter(FilterType type_id, int buffer_size); | |
| 294 static Filter* InitSdchFilter(FilterType type_id, | |
| 295 const FilterContext& filter_context, | |
| 296 int buffer_size); | |
| 297 | |
| 298 // Helper function to empty our output into the next filter's input. | |
| 299 void PushDataIntoNextFilter(); | |
| 300 | |
| 301 // Constructs a filter with an internal buffer of the given size. | |
| 302 // Only meant to be called by unit tests that need to control the buffer size. | |
| 303 static Filter* FactoryForTests(const std::vector<FilterType>& filter_types, | |
| 304 const FilterContext& filter_context, | |
| 305 int buffer_size); | |
| 306 | |
| 307 // An optional filter to process output from this filter. | |
| 308 scoped_ptr<Filter> next_filter_; | |
| 309 | |
| 310 // Remember what status or local filter last returned so we can better handle | |
| 311 // chained filters. | |
| 312 FilterStatus last_status_; | |
| 313 | |
| 314 // The filter type this filter was constructed from. | |
| 315 FilterType type_id_; | |
| 316 | |
| 317 DISALLOW_COPY_AND_ASSIGN(Filter); | |
| 318 }; | |
| 319 | |
| 320 } // namespace net | |
| 321 | |
| 322 #endif // NET_FILTER_FILTER_H__ | |
| OLD | NEW |