OLD | NEW |
1 // Copyright 2014 The Chromium Authors. All rights reserved. | 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 | 2 // Use of this source code is governed by a BSD-style license that can be |
3 // found in the LICENSE file. | 3 // found in the LICENSE file. |
4 | 4 |
5 // The basic usage of the Filter interface is described in the comment at | 5 // The basic usage of the Filter interface is described in the comment at |
6 // the beginning of filter.h. If Filter::Factory is passed a vector of | 6 // the beginning of filter.h. If Filter::Factory is passed a vector of |
7 // size greater than 1, that interface is implemented by a series of filters | 7 // size greater than 1, that interface is implemented by a series of filters |
8 // connected in a chain. In such a case the first filter | 8 // connected in a chain. In such a case the first filter |
9 // in the chain proxies calls to ReadData() so that its return values | 9 // in the chain proxies calls to ReadData() so that its return values |
10 // apply to the entire chain. | 10 // apply to the entire chain. |
11 // | 11 // |
12 // In a filter chain, the data flows from first filter (held by the | 12 // In a filter chain, the data flows from first filter (held by the |
13 // caller) down the chain. When ReadData() is called on any filter | 13 // caller) down the chain. When ReadData() is called on any filter |
14 // except for the last filter, it proxies the call down the chain, | 14 // except for the last filter, it proxies the call down the chain, |
15 // filling in the input buffers of subsequent filters if needed (== | 15 // filling in the input buffers of subsequent filters if needed (== |
16 // that filter's last_status() value is FILTER_NEED_MORE_DATA) and | 16 // that filter's last_status() value is FILTER_NEED_MORE_DATA) and |
17 // available (== the current filter has data it can output). The last | 17 // available (== the current filter has data it can output). The last |
18 // Filter will then output data if possible, and return | 18 // Filter will then output data if possible, and return |
19 // FILTER_NEED_MORE_DATA if not. Because the indirection pushes | 19 // FILTER_NEED_MORE_DATA if not. Because the indirection pushes |
20 // data along the filter chain at each level if it's available and the | 20 // data along the filter chain at each level if it's available and the |
21 // next filter needs it, a return value of FILTER_NEED_MORE_DATA from the | 21 // next filter needs it, a return value of FILTER_NEED_MORE_DATA from the |
22 // final filter will apply to the entire chain. | 22 // final filter will apply to the entire chain. |
23 | 23 |
24 #include "net/filter/filter.h" | 24 #include "net/filter/filter.h" |
25 | 25 |
26 #include "base/files/file_path.h" | 26 #include "base/files/file_path.h" |
27 #include "base/strings/string_util.h" | 27 #include "base/strings/string_util.h" |
28 #include "net/base/filename_util_unsafe.h" | 28 #include "net/base/filename_util_unsafe.h" |
29 #include "net/base/io_buffer.h" | 29 #include "net/base/io_buffer.h" |
30 #include "net/base/mime_util.h" | 30 #include "net/base/mime_util.h" |
31 #include "net/base/sdch_net_log_params.h" | 31 #include "net/base/sdch_net_log_params.h" |
32 #include "net/filter/gzip_filter.h" | 32 #include "net/filter/gzip_filter.h" |
33 #include "net/filter/sdch_filter.h" | 33 #include "net/filter/sdch_filter.h" |
34 #include "net/url_request/url_request_context.h" | 34 #include "net/url_request/url_request_context.h" |
35 #include "url/gurl.h" | 35 #include "url/gurl.h" |
36 | 36 |
37 namespace net { | 37 namespace net { |
38 | 38 |
39 namespace { | 39 namespace { |
40 | 40 |
41 // Filter types (using canonical lower case only): | 41 // Filter types (using canonical lower case only): |
42 const char kDeflate[] = "deflate"; | 42 const char kDeflate[] = "deflate"; |
43 const char kGZip[] = "gzip"; | 43 const char kGZip[] = "gzip"; |
44 const char kXGZip[] = "x-gzip"; | 44 const char kXGZip[] = "x-gzip"; |
45 const char kSdch[] = "sdch"; | 45 const char kSdch[] = "sdch"; |
46 // compress and x-compress are currently not supported. If we decide to support | 46 // compress and x-compress are currently not supported. If we decide to support |
47 // them, we'll need the same mime type compatibility hack we have for gzip. For | 47 // them, we'll need the same mime type compatibility hack we have for gzip. For |
48 // more information, see Firefox's nsHttpChannel::ProcessNormal. | 48 // more information, see Firefox's nsHttpChannel::ProcessNormal. |
49 | 49 |
50 // Mime types: | 50 // Mime types: |
51 const char kApplicationXGzip[] = "application/x-gzip"; | 51 const char kApplicationXGzip[] = "application/x-gzip"; |
52 const char kApplicationGzip[] = "application/gzip"; | 52 const char kApplicationGzip[] = "application/gzip"; |
53 const char kApplicationXGunzip[] = "application/x-gunzip"; | 53 const char kApplicationXGunzip[] = "application/x-gunzip"; |
54 const char kTextHtml[] = "text/html"; | 54 const char kTextHtml[] = "text/html"; |
55 | 55 |
56 // Buffer size allocated when de-compressing data. | 56 // Buffer size allocated when de-compressing data. |
57 const int kFilterBufSize = 32 * 1024; | 57 const int kFilterBufSize = 32 * 1024; |
(...skipping 88 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
146 return FILTER_ERROR; | 146 return FILTER_ERROR; |
147 } | 147 } |
148 *dest_len = dest_buffer_capacity; // Reset the input/output parameter. | 148 *dest_len = dest_buffer_capacity; // Reset the input/output parameter. |
149 next_filter_->ReadData(dest_buffer, dest_len); | 149 next_filter_->ReadData(dest_buffer, dest_len); |
150 if (FILTER_NEED_MORE_DATA == last_status_) | 150 if (FILTER_NEED_MORE_DATA == last_status_) |
151 return next_filter_->last_status(); | 151 return next_filter_->last_status(); |
152 | 152 |
153 // In the case where this filter has data internally, and is indicating such | 153 // In the case where this filter has data internally, and is indicating such |
154 // with a last_status_ of FILTER_OK, but at the same time the next filter in | 154 // with a last_status_ of FILTER_OK, but at the same time the next filter in |
155 // the chain indicated it FILTER_NEED_MORE_DATA, we have to be cautious | 155 // the chain indicated it FILTER_NEED_MORE_DATA, we have to be cautious |
156 // about confusing the caller. The API confusion can appear if we return | 156 // about confusing the caller. The API confusion can appear if we return |
157 // FILTER_OK (suggesting we have more data in aggregate), but yet we don't | 157 // FILTER_OK (suggesting we have more data in aggregate), but yet we don't |
158 // populate our output buffer. When that is the case, we need to | 158 // populate our output buffer. When that is the case, we need to |
159 // alternately call our filter element, and the next_filter element until we | 159 // alternately call our filter element, and the next_filter element until we |
160 // get out of this state (by pumping data into the next filter until it | 160 // get out of this state (by pumping data into the next filter until it |
161 // outputs data, or it runs out of data and reports that it NEED_MORE_DATA.) | 161 // outputs data, or it runs out of data and reports that it NEED_MORE_DATA.) |
162 } while (FILTER_OK == last_status_ && | 162 } while (FILTER_OK == last_status_ && |
163 FILTER_NEED_MORE_DATA == next_filter_->last_status() && | 163 FILTER_NEED_MORE_DATA == next_filter_->last_status() && |
164 0 == *dest_len); | 164 0 == *dest_len); |
165 | 165 |
166 if (next_filter_->last_status() == FILTER_ERROR) | 166 if (next_filter_->last_status() == FILTER_ERROR) |
167 return FILTER_ERROR; | 167 return FILTER_ERROR; |
168 return FILTER_OK; | 168 return FILTER_OK; |
(...skipping 41 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
210 std::string mime_type; | 210 std::string mime_type; |
211 bool success = filter_context.GetMimeType(&mime_type); | 211 bool success = filter_context.GetMimeType(&mime_type); |
212 DCHECK(success || mime_type.empty()); | 212 DCHECK(success || mime_type.empty()); |
213 | 213 |
214 if ((1 == encoding_types->size()) && | 214 if ((1 == encoding_types->size()) && |
215 (FILTER_TYPE_GZIP == encoding_types->front())) { | 215 (FILTER_TYPE_GZIP == encoding_types->front())) { |
216 if (LowerCaseEqualsASCII(mime_type, kApplicationXGzip) || | 216 if (LowerCaseEqualsASCII(mime_type, kApplicationXGzip) || |
217 LowerCaseEqualsASCII(mime_type, kApplicationGzip) || | 217 LowerCaseEqualsASCII(mime_type, kApplicationGzip) || |
218 LowerCaseEqualsASCII(mime_type, kApplicationXGunzip)) | 218 LowerCaseEqualsASCII(mime_type, kApplicationXGunzip)) |
219 // The server has told us that it sent us gziped content with a gzip | 219 // The server has told us that it sent us gziped content with a gzip |
220 // content encoding. Sadly, Apache mistakenly sets these headers for all | 220 // content encoding. Sadly, Apache mistakenly sets these headers for all |
221 // .gz files. We match Firefox's nsHttpChannel::ProcessNormal and ignore | 221 // .gz files. We match Firefox's nsHttpChannel::ProcessNormal and ignore |
222 // the Content-Encoding here. | 222 // the Content-Encoding here. |
223 encoding_types->clear(); | 223 encoding_types->clear(); |
224 | 224 |
225 GURL url; | 225 GURL url; |
226 std::string disposition; | 226 std::string disposition; |
227 success = filter_context.GetURL(&url); | 227 success = filter_context.GetURL(&url); |
228 DCHECK(success); | 228 DCHECK(success); |
229 filter_context.GetContentDisposition(&disposition); | 229 filter_context.GetContentDisposition(&disposition); |
230 // Don't supply a MIME type here, since that may cause disk IO. | 230 // Don't supply a MIME type here, since that may cause disk IO. |
231 base::FilePath::StringType extension = | 231 base::FilePath::StringType extension = |
(...skipping 17 matching lines...) Expand all Loading... |
249 // However, if it's not a supported mime type, then we will attempt to | 249 // However, if it's not a supported mime type, then we will attempt to |
250 // download it, and in that case, don't decompress .gz/.tgz files. | 250 // download it, and in that case, don't decompress .gz/.tgz files. |
251 if ((EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) || | 251 if ((EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) || |
252 LowerCaseEqualsASCII(extension, ".tgz")) && | 252 LowerCaseEqualsASCII(extension, ".tgz")) && |
253 !IsSupportedMimeType(mime_type)) | 253 !IsSupportedMimeType(mime_type)) |
254 encoding_types->clear(); | 254 encoding_types->clear(); |
255 } | 255 } |
256 } | 256 } |
257 | 257 |
258 // If the request was for SDCH content, then we might need additional fixups. | 258 // If the request was for SDCH content, then we might need additional fixups. |
259 if (!filter_context.SdchResponseExpected()) { | 259 if (!filter_context.SdchDictionariesAdvertised()) { |
260 // It was not an SDCH request, so we'll just record stats. | 260 // It was not an SDCH request, so we'll just record stats. |
261 if (1 < encoding_types->size()) { | 261 if (1 < encoding_types->size()) { |
262 // Multiple filters were intended to only be used for SDCH (thus far!) | 262 // Multiple filters were intended to only be used for SDCH (thus far!) |
263 LogSdchProblem(filter_context, SDCH_MULTIENCODING_FOR_NON_SDCH_REQUEST); | 263 LogSdchProblem(filter_context, SDCH_MULTIENCODING_FOR_NON_SDCH_REQUEST); |
264 } | 264 } |
265 if ((1 == encoding_types->size()) && | 265 if ((1 == encoding_types->size()) && |
266 (FILTER_TYPE_SDCH == encoding_types->front())) { | 266 (FILTER_TYPE_SDCH == encoding_types->front())) { |
267 LogSdchProblem(filter_context, | 267 LogSdchProblem(filter_context, |
268 SDCH_SDCH_CONTENT_ENCODE_FOR_NON_SDCH_REQUEST); | 268 SDCH_SDCH_CONTENT_ENCODE_FOR_NON_SDCH_REQUEST); |
269 } | 269 } |
270 return; | 270 return; |
271 } | 271 } |
272 | 272 |
273 // The request was tagged as an SDCH request, which means the server supplied | 273 // The request was tagged as an SDCH request, which means the server supplied |
274 // a dictionary, and we advertised it in the request. Some proxies will do | 274 // a dictionary, and we advertised it in the request. Some proxies will do |
275 // very strange things to the request, or the response, so we have to handle | 275 // very strange things to the request, or the response, so we have to handle |
276 // them gracefully. | 276 // them gracefully. |
277 | 277 |
278 // If content encoding included SDCH, then everything is "relatively" fine. | 278 // If content encoding included SDCH, then everything is "relatively" fine. |
279 if (!encoding_types->empty() && | 279 if (!encoding_types->empty() && |
280 (FILTER_TYPE_SDCH == encoding_types->front())) { | 280 (FILTER_TYPE_SDCH == encoding_types->front())) { |
281 // Some proxies (found currently in Argentina) strip the Content-Encoding | 281 // Some proxies (found currently in Argentina) strip the Content-Encoding |
282 // text from "sdch,gzip" to a mere "sdch" without modifying the compressed | 282 // text from "sdch,gzip" to a mere "sdch" without modifying the compressed |
283 // payload. To handle this gracefully, we simulate the "probably" deleted | 283 // payload. To handle this gracefully, we simulate the "probably" deleted |
284 // ",gzip" by appending a tentative gzip decode, which will default to a | 284 // ",gzip" by appending a tentative gzip decode, which will default to a |
285 // no-op pass through filter if it doesn't get gzip headers where expected. | 285 // no-op pass through filter if it doesn't get gzip headers where expected. |
286 if (1 == encoding_types->size()) { | 286 if (1 == encoding_types->size()) { |
287 encoding_types->push_back(FILTER_TYPE_GZIP_HELPING_SDCH); | 287 encoding_types->push_back(FILTER_TYPE_GZIP_HELPING_SDCH); |
288 LogSdchProblem(filter_context, SDCH_OPTIONAL_GUNZIP_ENCODING_ADDED); | 288 LogSdchProblem(filter_context, SDCH_OPTIONAL_GUNZIP_ENCODING_ADDED); |
289 } | 289 } |
290 return; | 290 return; |
291 } | 291 } |
292 | 292 |
293 // There are now several cases to handle for an SDCH request. Foremost, if | 293 // There are now several cases to handle for an SDCH request. Foremost, if |
294 // the outbound request was stripped so as not to advertise support for | 294 // the outbound request was stripped so as not to advertise support for |
295 // encodings, we might get back content with no encoding, or (for example) | 295 // encodings, we might get back content with no encoding, or (for example) |
296 // just gzip. We have to be sure that any changes we make allow for such | 296 // just gzip. We have to be sure that any changes we make allow for such |
297 // minimal coding to work. That issue is why we use TENTATIVE filters if we | 297 // minimal coding to work. That issue is why we use TENTATIVE filters if we |
298 // add any, as those filters sniff the content, and act as pass-through | 298 // add any, as those filters sniff the content, and act as pass-through |
299 // filters if headers are not found. | 299 // filters if headers are not found. |
300 | 300 |
301 // If the outbound GET is not modified, then the server will generally try to | 301 // If the outbound GET is not modified, then the server will generally try to |
302 // send us SDCH encoded content. As that content returns, there are several | 302 // send us SDCH encoded content. As that content returns, there are several |
303 // corruptions of the header "content-encoding" that proxies may perform (and | 303 // corruptions of the header "content-encoding" that proxies may perform (and |
304 // have been detected in the wild). We already dealt with the a honest | 304 // have been detected in the wild). We already dealt with the a honest |
305 // content encoding of "sdch,gzip" being corrupted into "sdch" with on change | 305 // content encoding of "sdch,gzip" being corrupted into "sdch" with on change |
306 // of the actual content. Another common corruption is to either disscard | 306 // of the actual content. Another common corruption is to either disscard |
307 // the accurate content encoding, or to replace it with gzip only (again, with | 307 // the accurate content encoding, or to replace it with gzip only (again, with |
308 // no change in actual content). The last observed corruption it to actually | 308 // no change in actual content). The last observed corruption it to actually |
309 // change the content, such as by re-gzipping it, and that may happen along | 309 // change the content, such as by re-gzipping it, and that may happen along |
310 // with corruption of the stated content encoding (wow!). | 310 // with corruption of the stated content encoding (wow!). |
311 | 311 |
312 // The one unresolved failure mode comes when we advertise a dictionary, and | 312 // The one unresolved failure mode comes when we advertise a dictionary, and |
313 // the server tries to *send* a gzipped file (not gzip encode content), and | 313 // the server tries to *send* a gzipped file (not gzip encode content), and |
314 // then we could do a gzip decode :-(. Since SDCH is only (currently) | 314 // then we could do a gzip decode :-(. Since SDCH is only (currently) |
315 // supported server side on paths that only send HTML content, this mode has | 315 // supported server side on paths that only send HTML content, this mode has |
316 // never surfaced in the wild (and is unlikely to). | 316 // never surfaced in the wild (and is unlikely to). |
317 // We will gather a lot of stats as we perform the fixups | 317 // We will gather a lot of stats as we perform the fixups |
318 if (StartsWithASCII(mime_type, kTextHtml, false)) { | 318 if (StartsWithASCII(mime_type, kTextHtml, false)) { |
319 // Suspicious case: Advertised dictionary, but server didn't use sdch, and | 319 // Suspicious case: Advertised dictionary, but server didn't use sdch, and |
320 // we're HTML tagged. | 320 // we're HTML tagged. |
321 if (encoding_types->empty()) { | 321 if (encoding_types->empty()) { |
322 LogSdchProblem(filter_context, SDCH_ADDED_CONTENT_ENCODING); | 322 LogSdchProblem(filter_context, SDCH_ADDED_CONTENT_ENCODING); |
323 } else if (1 == encoding_types->size()) { | 323 } else if (1 == encoding_types->size()) { |
324 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODING); | 324 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODING); |
325 } else { | 325 } else { |
326 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODINGS); | 326 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODINGS); |
327 } | 327 } |
328 } else { | 328 } else { |
329 // Remarkable case!?! We advertised an SDCH dictionary, content-encoding | 329 // Remarkable case!?! We advertised an SDCH dictionary, content-encoding |
330 // was not marked for SDCH processing: Why did the server suggest an SDCH | 330 // was not marked for SDCH processing: Why did the server suggest an SDCH |
331 // dictionary in the first place??. Also, the content isn't | 331 // dictionary in the first place??. Also, the content isn't |
332 // tagged as HTML, despite the fact that SDCH encoding is mostly likely for | 332 // tagged as HTML, despite the fact that SDCH encoding is mostly likely for |
333 // HTML: Did some anti-virus system strip this tag (sometimes they strip | 333 // HTML: Did some anti-virus system strip this tag (sometimes they strip |
334 // accept-encoding headers on the request)?? Does the content encoding not | 334 // accept-encoding headers on the request)?? Does the content encoding not |
335 // start with "text/html" for some other reason?? We'll report this as a | 335 // start with "text/html" for some other reason?? We'll report this as a |
336 // fixup to a binary file, but it probably really is text/html (some how). | 336 // fixup to a binary file, but it probably really is text/html (some how). |
337 if (encoding_types->empty()) { | 337 if (encoding_types->empty()) { |
338 LogSdchProblem(filter_context, SDCH_BINARY_ADDED_CONTENT_ENCODING); | 338 LogSdchProblem(filter_context, SDCH_BINARY_ADDED_CONTENT_ENCODING); |
339 } else if (1 == encoding_types->size()) { | 339 } else if (1 == encoding_types->size()) { |
340 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODING); | 340 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODING); |
341 } else { | 341 } else { |
342 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODINGS); | 342 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODINGS); |
343 } | 343 } |
344 } | 344 } |
345 | 345 |
346 // Leave the existing encoding type to be processed first, and add our | 346 // Leave the existing encoding type to be processed first, and add our |
347 // tentative decodings to be done afterwards. Vodaphone UK reportedyl will | 347 // tentative decodings to be done afterwards. Vodaphone UK reportedyl will |
348 // perform a second layer of gzip encoding atop the server's sdch,gzip | 348 // perform a second layer of gzip encoding atop the server's sdch,gzip |
349 // encoding, and then claim that the content encoding is a mere gzip. As a | 349 // encoding, and then claim that the content encoding is a mere gzip. As a |
350 // result we'll need (in that case) to do the gunzip, plus our tentative | 350 // result we'll need (in that case) to do the gunzip, plus our tentative |
351 // gunzip and tentative SDCH decoding. | 351 // gunzip and tentative SDCH decoding. |
352 // This approach nicely handles the empty() list as well, and should work with | 352 // This approach nicely handles the empty() list as well, and should work with |
353 // other (as yet undiscovered) proxies the choose to re-compressed with some | 353 // other (as yet undiscovered) proxies the choose to re-compressed with some |
354 // other encoding (such as bzip2, etc.). | 354 // other encoding (such as bzip2, etc.). |
355 encoding_types->insert(encoding_types->begin(), | 355 encoding_types->insert(encoding_types->begin(), |
356 FILTER_TYPE_GZIP_HELPING_SDCH); | 356 FILTER_TYPE_GZIP_HELPING_SDCH); |
357 encoding_types->insert(encoding_types->begin(), FILTER_TYPE_SDCH_POSSIBLE); | 357 encoding_types->insert(encoding_types->begin(), FILTER_TYPE_SDCH_POSSIBLE); |
358 return; | 358 return; |
359 } | 359 } |
(...skipping 92 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
452 | 452 |
453 void Filter::PushDataIntoNextFilter() { | 453 void Filter::PushDataIntoNextFilter() { |
454 IOBuffer* next_buffer = next_filter_->stream_buffer(); | 454 IOBuffer* next_buffer = next_filter_->stream_buffer(); |
455 int next_size = next_filter_->stream_buffer_size(); | 455 int next_size = next_filter_->stream_buffer_size(); |
456 last_status_ = ReadFilteredData(next_buffer->data(), &next_size); | 456 last_status_ = ReadFilteredData(next_buffer->data(), &next_size); |
457 if (FILTER_ERROR != last_status_) | 457 if (FILTER_ERROR != last_status_) |
458 next_filter_->FlushStreamBuffer(next_size); | 458 next_filter_->FlushStreamBuffer(next_size); |
459 } | 459 } |
460 | 460 |
461 } // namespace net | 461 } // namespace net |
OLD | NEW |