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.
5 #include "net/filter/filter.h"
7 #include "base/files/file_path.h"
8 #include "base/strings/string_util.h"
9 #include "net/base/io_buffer.h"
10 #include "net/base/mime_util.h"
11 #include "net/filter/gzip_filter.h"
12 #include "net/filter/sdch_filter.h"
16 // Filter types (using canonical lower case only):
17 const char kDeflate[] = "deflate";
18 const char kGZip[] = "gzip";
19 const char kXGZip[] = "x-gzip";
20 const char kSdch[] = "sdch";
21 // compress and x-compress are currently not supported. If we decide to support
22 // them, we'll need the same mime type compatibility hack we have for gzip. For
23 // more information, see Firefox's nsHttpChannel::ProcessNormal.
26 const char kApplicationXGzip[] = "application/x-gzip";
27 const char kApplicationGzip[] = "application/gzip";
28 const char kApplicationXGunzip[] = "application/x-gunzip";
29 const char kTextHtml[] = "text/html";
31 // Buffer size allocated when de-compressing data.
32 const int kFilterBufSize = 32 * 1024;
38 FilterContext::~FilterContext() {
44 Filter* Filter::Factory(const std::vector<FilterType>& filter_types,
45 const FilterContext& filter_context) {
46 if (filter_types.empty())
49 Filter* filter_list = NULL; // Linked list of filters.
50 for (size_t i = 0; i < filter_types.size(); i++) {
51 filter_list = PrependNewFilter(filter_types[i], filter_context,
52 kFilterBufSize, filter_list);
60 Filter* Filter::GZipFactory() {
61 return InitGZipFilter(FILTER_TYPE_GZIP, kFilterBufSize);
65 Filter* Filter::FactoryForTests(const std::vector<FilterType>& filter_types,
66 const FilterContext& filter_context,
68 if (filter_types.empty())
71 Filter* filter_list = NULL; // Linked list of filters.
72 for (size_t i = 0; i < filter_types.size(); i++) {
73 filter_list = PrependNewFilter(filter_types[i], filter_context,
74 buffer_size, filter_list);
81 Filter::FilterStatus Filter::ReadData(char* dest_buffer, int* dest_len) {
82 const int dest_buffer_capacity = *dest_len;
83 if (last_status_ == FILTER_ERROR)
85 if (!next_filter_.get())
86 return last_status_ = ReadFilteredData(dest_buffer, dest_len);
87 if (last_status_ == FILTER_NEED_MORE_DATA && !stream_data_len())
88 return next_filter_->ReadData(dest_buffer, dest_len);
91 if (next_filter_->last_status() == FILTER_NEED_MORE_DATA) {
92 PushDataIntoNextFilter();
93 if (FILTER_ERROR == last_status_)
96 *dest_len = dest_buffer_capacity; // Reset the input/output parameter.
97 next_filter_->ReadData(dest_buffer, dest_len);
98 if (FILTER_NEED_MORE_DATA == last_status_)
99 return next_filter_->last_status();
101 // In the case where this filter has data internally, and is indicating such
102 // with a last_status_ of FILTER_OK, but at the same time the next filter in
103 // the chain indicated it FILTER_NEED_MORE_DATA, we have to be cautious
104 // about confusing the caller. The API confusion can appear if we return
105 // FILTER_OK (suggesting we have more data in aggregate), but yet we don't
106 // populate our output buffer. When that is the case, we need to
107 // alternately call our filter element, and the next_filter element until we
108 // get out of this state (by pumping data into the next filter until it
109 // outputs data, or it runs out of data and reports that it NEED_MORE_DATA.)
110 } while (FILTER_OK == last_status_ &&
111 FILTER_NEED_MORE_DATA == next_filter_->last_status() &&
114 if (next_filter_->last_status() == FILTER_ERROR)
119 bool Filter::FlushStreamBuffer(int stream_data_len) {
120 DCHECK_LE(stream_data_len, stream_buffer_size_);
121 if (stream_data_len <= 0 || stream_data_len > stream_buffer_size_)
124 DCHECK(stream_buffer());
125 // Bail out if there is more data in the stream buffer to be filtered.
126 if (!stream_buffer() || stream_data_len_)
129 next_stream_data_ = stream_buffer()->data();
130 stream_data_len_ = stream_data_len;
135 Filter::FilterType Filter::ConvertEncodingToType(
136 const std::string& filter_type) {
138 if (LowerCaseEqualsASCII(filter_type, kDeflate)) {
139 type_id = FILTER_TYPE_DEFLATE;
140 } else if (LowerCaseEqualsASCII(filter_type, kGZip) ||
141 LowerCaseEqualsASCII(filter_type, kXGZip)) {
142 type_id = FILTER_TYPE_GZIP;
143 } else if (LowerCaseEqualsASCII(filter_type, kSdch)) {
144 type_id = FILTER_TYPE_SDCH;
146 // Note we also consider "identity" and "uncompressed" UNSUPPORTED as
147 // filter should be disabled in such cases.
148 type_id = FILTER_TYPE_UNSUPPORTED;
154 void Filter::FixupEncodingTypes(
155 const FilterContext& filter_context,
156 std::vector<FilterType>* encoding_types) {
157 std::string mime_type;
158 bool success = filter_context.GetMimeType(&mime_type);
159 DCHECK(success || mime_type.empty());
161 if ((1 == encoding_types->size()) &&
162 (FILTER_TYPE_GZIP == encoding_types->front())) {
163 if (LowerCaseEqualsASCII(mime_type, kApplicationXGzip) ||
164 LowerCaseEqualsASCII(mime_type, kApplicationGzip) ||
165 LowerCaseEqualsASCII(mime_type, kApplicationXGunzip))
166 // The server has told us that it sent us gziped content with a gzip
167 // content encoding. Sadly, Apache mistakenly sets these headers for all
168 // .gz files. We match Firefox's nsHttpChannel::ProcessNormal and ignore
169 // the Content-Encoding here.
170 encoding_types->clear();
173 success = filter_context.GetURL(&url);
175 base::FilePath filename =
176 base::FilePath().AppendASCII(url.ExtractFileName());
177 base::FilePath::StringType extension = filename.Extension();
179 if (filter_context.IsDownload()) {
180 // We don't want to decompress gzipped files when the user explicitly
181 // asks to download them.
182 // For the case of svgz files, we use the extension to distinguish
183 // between svgz files and svg files compressed with gzip by the server.
184 // When viewing a .svgz file, we need to uncompress it, but we don't
185 // want to do that when downloading.
186 // See Firefox's nonDecodableExtensions in nsExternalHelperAppService.cpp
187 if (EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) ||
188 LowerCaseEqualsASCII(extension, ".tgz") ||
189 LowerCaseEqualsASCII(extension, ".svgz"))
190 encoding_types->clear();
192 // When the user does not explicitly ask to download a file, if we get a
193 // supported mime type, then we attempt to decompress in order to view it.
194 // However, if it's not a supported mime type, then we will attempt to
195 // download it, and in that case, don't decompress .gz/.tgz files.
196 if ((EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) ||
197 LowerCaseEqualsASCII(extension, ".tgz")) &&
198 !IsSupportedMimeType(mime_type))
199 encoding_types->clear();
203 // If the request was for SDCH content, then we might need additional fixups.
204 if (!filter_context.IsSdchResponse()) {
205 // It was not an SDCH request, so we'll just record stats.
206 if (1 < encoding_types->size()) {
207 // Multiple filters were intended to only be used for SDCH (thus far!)
208 SdchManager::SdchErrorRecovery(
209 SdchManager::MULTIENCODING_FOR_NON_SDCH_REQUEST);
211 if ((1 == encoding_types->size()) &&
212 (FILTER_TYPE_SDCH == encoding_types->front())) {
213 SdchManager::SdchErrorRecovery(
214 SdchManager::SDCH_CONTENT_ENCODE_FOR_NON_SDCH_REQUEST);
219 // The request was tagged as an SDCH request, which means the server supplied
220 // a dictionary, and we advertised it in the request. Some proxies will do
221 // very strange things to the request, or the response, so we have to handle
224 // If content encoding included SDCH, then everything is "relatively" fine.
225 if (!encoding_types->empty() &&
226 (FILTER_TYPE_SDCH == encoding_types->front())) {
227 // Some proxies (found currently in Argentina) strip the Content-Encoding
228 // text from "sdch,gzip" to a mere "sdch" without modifying the compressed
229 // payload. To handle this gracefully, we simulate the "probably" deleted
230 // ",gzip" by appending a tentative gzip decode, which will default to a
231 // no-op pass through filter if it doesn't get gzip headers where expected.
232 if (1 == encoding_types->size()) {
233 encoding_types->push_back(FILTER_TYPE_GZIP_HELPING_SDCH);
234 SdchManager::SdchErrorRecovery(
235 SdchManager::OPTIONAL_GUNZIP_ENCODING_ADDED);
240 // There are now several cases to handle for an SDCH request. Foremost, if
241 // the outbound request was stripped so as not to advertise support for
242 // encodings, we might get back content with no encoding, or (for example)
243 // just gzip. We have to be sure that any changes we make allow for such
244 // minimal coding to work. That issue is why we use TENTATIVE filters if we
245 // add any, as those filters sniff the content, and act as pass-through
246 // filters if headers are not found.
248 // If the outbound GET is not modified, then the server will generally try to
249 // send us SDCH encoded content. As that content returns, there are several
250 // corruptions of the header "content-encoding" that proxies may perform (and
251 // have been detected in the wild). We already dealt with the a honest
252 // content encoding of "sdch,gzip" being corrupted into "sdch" with on change
253 // of the actual content. Another common corruption is to either disscard
254 // the accurate content encoding, or to replace it with gzip only (again, with
255 // no change in actual content). The last observed corruption it to actually
256 // change the content, such as by re-gzipping it, and that may happen along
257 // with corruption of the stated content encoding (wow!).
259 // The one unresolved failure mode comes when we advertise a dictionary, and
260 // the server tries to *send* a gzipped file (not gzip encode content), and
261 // then we could do a gzip decode :-(. Since SDCH is only (currently)
262 // supported server side on paths that only send HTML content, this mode has
263 // never surfaced in the wild (and is unlikely to).
264 // We will gather a lot of stats as we perform the fixups
265 if (StartsWithASCII(mime_type, kTextHtml, false)) {
266 // Suspicious case: Advertised dictionary, but server didn't use sdch, and
267 // we're HTML tagged.
268 if (encoding_types->empty()) {
269 SdchManager::SdchErrorRecovery(
270 SdchManager::ADDED_CONTENT_ENCODING);
271 } else if (1 == encoding_types->size()) {
272 SdchManager::SdchErrorRecovery(
273 SdchManager::FIXED_CONTENT_ENCODING);
275 SdchManager::SdchErrorRecovery(
276 SdchManager::FIXED_CONTENT_ENCODINGS);
279 // Remarkable case!?! We advertised an SDCH dictionary, content-encoding
280 // was not marked for SDCH processing: Why did the server suggest an SDCH
281 // dictionary in the first place??. Also, the content isn't
282 // tagged as HTML, despite the fact that SDCH encoding is mostly likely for
283 // HTML: Did some anti-virus system strip this tag (sometimes they strip
284 // accept-encoding headers on the request)?? Does the content encoding not
285 // start with "text/html" for some other reason?? We'll report this as a
286 // fixup to a binary file, but it probably really is text/html (some how).
287 if (encoding_types->empty()) {
288 SdchManager::SdchErrorRecovery(
289 SdchManager::BINARY_ADDED_CONTENT_ENCODING);
290 } else if (1 == encoding_types->size()) {
291 SdchManager::SdchErrorRecovery(
292 SdchManager::BINARY_FIXED_CONTENT_ENCODING);
294 SdchManager::SdchErrorRecovery(
295 SdchManager::BINARY_FIXED_CONTENT_ENCODINGS);
299 // Leave the existing encoding type to be processed first, and add our
300 // tentative decodings to be done afterwards. Vodaphone UK reportedyl will
301 // perform a second layer of gzip encoding atop the server's sdch,gzip
302 // encoding, and then claim that the content encoding is a mere gzip. As a
303 // result we'll need (in that case) to do the gunzip, plus our tentative
304 // gunzip and tentative SDCH decoding.
305 // This approach nicely handles the empty() list as well, and should work with
306 // other (as yet undiscovered) proxies the choose to re-compressed with some
307 // other encoding (such as bzip2, etc.).
308 encoding_types->insert(encoding_types->begin(),
309 FILTER_TYPE_GZIP_HELPING_SDCH);
310 encoding_types->insert(encoding_types->begin(), FILTER_TYPE_SDCH_POSSIBLE);
315 : stream_buffer_(NULL),
316 stream_buffer_size_(0),
317 next_stream_data_(NULL),
319 last_status_(FILTER_NEED_MORE_DATA) {}
321 Filter::FilterStatus Filter::CopyOut(char* dest_buffer, int* dest_len) {
323 int input_len = *dest_len;
326 if (0 == stream_data_len_)
327 return Filter::FILTER_NEED_MORE_DATA;
329 out_len = std::min(input_len, stream_data_len_);
330 memcpy(dest_buffer, next_stream_data_, out_len);
331 *dest_len += out_len;
332 stream_data_len_ -= out_len;
333 if (0 == stream_data_len_) {
334 next_stream_data_ = NULL;
335 return Filter::FILTER_NEED_MORE_DATA;
337 next_stream_data_ += out_len;
338 return Filter::FILTER_OK;
343 Filter* Filter::InitGZipFilter(FilterType type_id, int buffer_size) {
344 scoped_ptr<GZipFilter> gz_filter(new GZipFilter());
345 gz_filter->InitBuffer(buffer_size);
346 return gz_filter->InitDecoding(type_id) ? gz_filter.release() : NULL;
350 Filter* Filter::InitSdchFilter(FilterType type_id,
351 const FilterContext& filter_context,
353 scoped_ptr<SdchFilter> sdch_filter(new SdchFilter(filter_context));
354 sdch_filter->InitBuffer(buffer_size);
355 return sdch_filter->InitDecoding(type_id) ? sdch_filter.release() : NULL;
359 Filter* Filter::PrependNewFilter(FilterType type_id,
360 const FilterContext& filter_context,
362 Filter* filter_list) {
363 scoped_ptr<Filter> first_filter; // Soon to be start of chain.
365 case FILTER_TYPE_GZIP_HELPING_SDCH:
366 case FILTER_TYPE_DEFLATE:
367 case FILTER_TYPE_GZIP:
368 first_filter.reset(InitGZipFilter(type_id, buffer_size));
370 case FILTER_TYPE_SDCH:
371 case FILTER_TYPE_SDCH_POSSIBLE:
372 if (SdchManager::Global() && SdchManager::sdch_enabled()) {
374 InitSdchFilter(type_id, filter_context, buffer_size));
381 if (!first_filter.get())
384 first_filter->next_filter_.reset(filter_list);
385 return first_filter.release();
388 void Filter::InitBuffer(int buffer_size) {
389 DCHECK(!stream_buffer());
390 DCHECK_GT(buffer_size, 0);
391 stream_buffer_ = new IOBuffer(buffer_size);
392 stream_buffer_size_ = buffer_size;
395 void Filter::PushDataIntoNextFilter() {
396 IOBuffer* next_buffer = next_filter_->stream_buffer();
397 int next_size = next_filter_->stream_buffer_size();
398 last_status_ = ReadFilteredData(next_buffer->data(), &next_size);
399 if (FILTER_ERROR != last_status_)
400 next_filter_->FlushStreamBuffer(next_size);