Home | History | Annotate | Download | only in filter 
      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 #include "net/filter/filter.h"
      6 
      7 #include "base/files/file_path.h"
      8 #include "base/strings/string_util.h"
      9 #include "net/base/filename_util_unsafe.h"
     10 #include "net/base/io_buffer.h"
     11 #include "net/base/mime_util.h"
     12 #include "net/filter/gzip_filter.h"
     13 #include "net/filter/sdch_filter.h"
     14 #include "net/url_request/url_request_context.h"
     15 #include "url/gurl.h"
     16 
     17 namespace {
     18 
     19 // Filter types (using canonical lower case only):
     20 const char kDeflate[]      = "deflate";
     21 const char kGZip[]         = "gzip";
     22 const char kXGZip[]        = "x-gzip";
     23 const char kSdch[]         = "sdch";
     24 // compress and x-compress are currently not supported.  If we decide to support
     25 // them, we'll need the same mime type compatibility hack we have for gzip.  For
     26 // more information, see Firefox's nsHttpChannel::ProcessNormal.
     27 
     28 // Mime types:
     29 const char kApplicationXGzip[]     = "application/x-gzip";
     30 const char kApplicationGzip[]      = "application/gzip";
     31 const char kApplicationXGunzip[]   = "application/x-gunzip";
     32 const char kTextHtml[]             = "text/html";
     33 
     34 // Buffer size allocated when de-compressing data.
     35 const int kFilterBufSize = 32 * 1024;
     36 
     37 }  // namespace
     38 
     39 namespace net {
     40 
     41 FilterContext::~FilterContext() {
     42 }
     43 
     44 Filter::~Filter() {}
     45 
     46 // static
     47 Filter* Filter::Factory(const std::vector<FilterType>& filter_types,
     48                         const FilterContext& filter_context) {
     49   if (filter_types.empty())
     50     return NULL;
     51 
     52   Filter* filter_list = NULL;  // Linked list of filters.
     53   for (size_t i = 0; i < filter_types.size(); i++) {
     54     filter_list = PrependNewFilter(filter_types[i], filter_context,
     55                                    kFilterBufSize, filter_list);
     56     if (!filter_list)
     57       return NULL;
     58   }
     59   return filter_list;
     60 }
     61 
     62 // static
     63 Filter* Filter::GZipFactory() {
     64   return InitGZipFilter(FILTER_TYPE_GZIP, kFilterBufSize);
     65 }
     66 
     67 // static
     68 Filter* Filter::FactoryForTests(const std::vector<FilterType>& filter_types,
     69                                 const FilterContext& filter_context,
     70                                 int buffer_size) {
     71   if (filter_types.empty())
     72     return NULL;
     73 
     74   Filter* filter_list = NULL;  // Linked list of filters.
     75   for (size_t i = 0; i < filter_types.size(); i++) {
     76     filter_list = PrependNewFilter(filter_types[i], filter_context,
     77                                    buffer_size, filter_list);
     78     if (!filter_list)
     79       return NULL;
     80   }
     81   return filter_list;
     82 }
     83 
     84 Filter::FilterStatus Filter::ReadData(char* dest_buffer, int* dest_len) {
     85   const int dest_buffer_capacity = *dest_len;
     86   if (last_status_ == FILTER_ERROR)
     87     return last_status_;
     88   if (!next_filter_.get())
     89     return last_status_ = ReadFilteredData(dest_buffer, dest_len);
     90   if (last_status_ == FILTER_NEED_MORE_DATA && !stream_data_len())
     91     return next_filter_->ReadData(dest_buffer, dest_len);
     92 
     93   do {
     94     if (next_filter_->last_status() == FILTER_NEED_MORE_DATA) {
     95       PushDataIntoNextFilter();
     96       if (FILTER_ERROR == last_status_)
     97         return FILTER_ERROR;
     98     }
     99     *dest_len = dest_buffer_capacity;  // Reset the input/output parameter.
    100     next_filter_->ReadData(dest_buffer, dest_len);
    101     if (FILTER_NEED_MORE_DATA == last_status_)
    102         return next_filter_->last_status();
    103 
    104     // In the case where this filter has data internally, and is indicating such
    105     // with a last_status_ of FILTER_OK, but at the same time the next filter in
    106     // the chain indicated it FILTER_NEED_MORE_DATA, we have to be cautious
    107     // about confusing the caller.  The API confusion can appear if we return
    108     // FILTER_OK (suggesting we have more data in aggregate), but yet we don't
    109     // populate our output buffer.  When that is the case, we need to
    110     // alternately call our filter element, and the next_filter element until we
    111     // get out of this state (by pumping data into the next filter until it
    112     // outputs data, or it runs out of data and reports that it NEED_MORE_DATA.)
    113   } while (FILTER_OK == last_status_ &&
    114            FILTER_NEED_MORE_DATA == next_filter_->last_status() &&
    115            0 == *dest_len);
    116 
    117   if (next_filter_->last_status() == FILTER_ERROR)
    118     return FILTER_ERROR;
    119   return FILTER_OK;
    120 }
    121 
    122 bool Filter::FlushStreamBuffer(int stream_data_len) {
    123   DCHECK_LE(stream_data_len, stream_buffer_size_);
    124   if (stream_data_len <= 0 || stream_data_len > stream_buffer_size_)
    125     return false;
    126 
    127   DCHECK(stream_buffer());
    128   // Bail out if there is more data in the stream buffer to be filtered.
    129   if (!stream_buffer() || stream_data_len_)
    130     return false;
    131 
    132   next_stream_data_ = stream_buffer()->data();
    133   stream_data_len_ = stream_data_len;
    134   return true;
    135 }
    136 
    137 // static
    138 Filter::FilterType Filter::ConvertEncodingToType(
    139     const std::string& filter_type) {
    140   FilterType type_id;
    141   if (LowerCaseEqualsASCII(filter_type, kDeflate)) {
    142     type_id = FILTER_TYPE_DEFLATE;
    143   } else if (LowerCaseEqualsASCII(filter_type, kGZip) ||
    144              LowerCaseEqualsASCII(filter_type, kXGZip)) {
    145     type_id = FILTER_TYPE_GZIP;
    146   } else if (LowerCaseEqualsASCII(filter_type, kSdch)) {
    147     type_id = FILTER_TYPE_SDCH;
    148   } else {
    149     // Note we also consider "identity" and "uncompressed" UNSUPPORTED as
    150     // filter should be disabled in such cases.
    151     type_id = FILTER_TYPE_UNSUPPORTED;
    152   }
    153   return type_id;
    154 }
    155 
    156 // static
    157 void Filter::FixupEncodingTypes(
    158     const FilterContext& filter_context,
    159     std::vector<FilterType>* encoding_types) {
    160   std::string mime_type;
    161   bool success = filter_context.GetMimeType(&mime_type);
    162   DCHECK(success || mime_type.empty());
    163 
    164   if ((1 == encoding_types->size()) &&
    165       (FILTER_TYPE_GZIP == encoding_types->front())) {
    166     if (LowerCaseEqualsASCII(mime_type, kApplicationXGzip) ||
    167         LowerCaseEqualsASCII(mime_type, kApplicationGzip) ||
    168         LowerCaseEqualsASCII(mime_type, kApplicationXGunzip))
    169       // The server has told us that it sent us gziped content with a gzip
    170       // content encoding.  Sadly, Apache mistakenly sets these headers for all
    171       // .gz files.  We match Firefox's nsHttpChannel::ProcessNormal and ignore
    172       // the Content-Encoding here.
    173       encoding_types->clear();
    174 
    175     GURL url;
    176     std::string disposition;
    177     success = filter_context.GetURL(&url);
    178     DCHECK(success);
    179     filter_context.GetContentDisposition(&disposition);
    180     // Don't supply a MIME type here, since that may cause disk IO.
    181     base::FilePath::StringType extension =
    182         GenerateFileExtensionUnsafe(url, disposition, "UTF-8", "", "", "");
    183 
    184     if (filter_context.IsDownload()) {
    185       // We don't want to decompress gzipped files when the user explicitly
    186       // asks to download them.
    187       // For the case of svgz files, we use the extension to distinguish
    188       // between svgz files and svg files compressed with gzip by the server.
    189       // When viewing a .svgz file, we need to uncompress it, but we don't
    190       // want to do that when downloading.
    191       // See Firefox's nonDecodableExtensions in nsExternalHelperAppService.cpp
    192       if (EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) ||
    193           LowerCaseEqualsASCII(extension, ".tgz") ||
    194           LowerCaseEqualsASCII(extension, ".svgz"))
    195         encoding_types->clear();
    196     } else {
    197       // When the user does not explicitly ask to download a file, if we get a
    198       // supported mime type, then we attempt to decompress in order to view it.
    199       // However, if it's not a supported mime type, then we will attempt to
    200       // download it, and in that case, don't decompress .gz/.tgz files.
    201       if ((EndsWith(extension, FILE_PATH_LITERAL(".gz"), false) ||
    202            LowerCaseEqualsASCII(extension, ".tgz")) &&
    203           !IsSupportedMimeType(mime_type))
    204         encoding_types->clear();
    205     }
    206   }
    207 
    208   // If the request was for SDCH content, then we might need additional fixups.
    209   if (!filter_context.IsSdchResponse()) {
    210     // It was not an SDCH request, so we'll just record stats.
    211     if (1 < encoding_types->size()) {
    212       // Multiple filters were intended to only be used for SDCH (thus far!)
    213       SdchManager::SdchErrorRecovery(
    214           SdchManager::MULTIENCODING_FOR_NON_SDCH_REQUEST);
    215     }
    216     if ((1 == encoding_types->size()) &&
    217         (FILTER_TYPE_SDCH == encoding_types->front())) {
    218         SdchManager::SdchErrorRecovery(
    219             SdchManager::SDCH_CONTENT_ENCODE_FOR_NON_SDCH_REQUEST);
    220     }
    221     return;
    222   }
    223 
    224   // The request was tagged as an SDCH request, which means the server supplied
    225   // a dictionary, and we advertised it in the request.  Some proxies will do
    226   // very strange things to the request, or the response, so we have to handle
    227   // them gracefully.
    228 
    229   // If content encoding included SDCH, then everything is "relatively" fine.
    230   if (!encoding_types->empty() &&
    231       (FILTER_TYPE_SDCH == encoding_types->front())) {
    232     // Some proxies (found currently in Argentina) strip the Content-Encoding
    233     // text from "sdch,gzip" to a mere "sdch" without modifying the compressed
    234     // payload.   To handle this gracefully, we simulate the "probably" deleted
    235     // ",gzip" by appending a tentative gzip decode, which will default to a
    236     // no-op pass through filter if it doesn't get gzip headers where expected.
    237     if (1 == encoding_types->size()) {
    238       encoding_types->push_back(FILTER_TYPE_GZIP_HELPING_SDCH);
    239       SdchManager::SdchErrorRecovery(
    240           SdchManager::OPTIONAL_GUNZIP_ENCODING_ADDED);
    241     }
    242     return;
    243   }
    244 
    245   // There are now several cases to handle for an SDCH request.  Foremost, if
    246   // the outbound request was stripped so as not to advertise support for
    247   // encodings, we might get back content with no encoding, or (for example)
    248   // just gzip.  We have to be sure that any changes we make allow for such
    249   // minimal coding to work.  That issue is why we use TENTATIVE filters if we
    250   // add any, as those filters sniff the content, and act as pass-through
    251   // filters if headers are not found.
    252 
    253   // If the outbound GET is not modified, then the server will generally try to
    254   // send us SDCH encoded content.  As that content returns, there are several
    255   // corruptions of the header "content-encoding" that proxies may perform (and
    256   // have been detected in the wild).  We already dealt with the a honest
    257   // content encoding of "sdch,gzip" being corrupted into "sdch" with on change
    258   // of the actual content.  Another common corruption is to either disscard
    259   // the accurate content encoding, or to replace it with gzip only (again, with
    260   // no change in actual content). The last observed corruption it to actually
    261   // change the content, such as by re-gzipping it, and that may happen along
    262   // with corruption of the stated content encoding (wow!).
    263 
    264   // The one unresolved failure mode comes when we advertise a dictionary, and
    265   // the server tries to *send* a gzipped file (not gzip encode content), and
    266   // then we could do a gzip decode :-(. Since SDCH is only (currently)
    267   // supported server side on paths that only send HTML content, this mode has
    268   // never surfaced in the wild (and is unlikely to).
    269   // We will gather a lot of stats as we perform the fixups
    270   if (StartsWithASCII(mime_type, kTextHtml, false)) {
    271     // Suspicious case: Advertised dictionary, but server didn't use sdch, and
    272     // we're HTML tagged.
    273     if (encoding_types->empty()) {
    274       SdchManager::SdchErrorRecovery(
    275           SdchManager::ADDED_CONTENT_ENCODING);
    276     } else if (1 == encoding_types->size()) {
    277       SdchManager::SdchErrorRecovery(
    278           SdchManager::FIXED_CONTENT_ENCODING);
    279     } else {
    280       SdchManager::SdchErrorRecovery(
    281           SdchManager::FIXED_CONTENT_ENCODINGS);
    282     }
    283   } else {
    284     // Remarkable case!?!  We advertised an SDCH dictionary, content-encoding
    285     // was not marked for SDCH processing: Why did the server suggest an SDCH
    286     // dictionary in the first place??.  Also, the content isn't
    287     // tagged as HTML, despite the fact that SDCH encoding is mostly likely for
    288     // HTML: Did some anti-virus system strip this tag (sometimes they strip
    289     // accept-encoding headers on the request)??  Does the content encoding not
    290     // start with "text/html" for some other reason??  We'll report this as a
    291     // fixup to a binary file, but it probably really is text/html (some how).
    292     if (encoding_types->empty()) {
    293       SdchManager::SdchErrorRecovery(
    294           SdchManager::BINARY_ADDED_CONTENT_ENCODING);
    295     } else if (1 == encoding_types->size()) {
    296       SdchManager::SdchErrorRecovery(
    297           SdchManager::BINARY_FIXED_CONTENT_ENCODING);
    298     } else {
    299       SdchManager::SdchErrorRecovery(
    300           SdchManager::BINARY_FIXED_CONTENT_ENCODINGS);
    301     }
    302   }
    303 
    304   // Leave the existing encoding type to be processed first, and add our
    305   // tentative decodings to be done afterwards.  Vodaphone UK reportedyl will
    306   // perform a second layer of gzip encoding atop the server's sdch,gzip
    307   // encoding, and then claim that the content encoding is a mere gzip.  As a
    308   // result we'll need (in that case) to do the gunzip, plus our tentative
    309   // gunzip and tentative SDCH decoding.
    310   // This approach nicely handles the empty() list as well, and should work with
    311   // other (as yet undiscovered) proxies the choose to re-compressed with some
    312   // other encoding (such as bzip2, etc.).
    313   encoding_types->insert(encoding_types->begin(),
    314                          FILTER_TYPE_GZIP_HELPING_SDCH);
    315   encoding_types->insert(encoding_types->begin(), FILTER_TYPE_SDCH_POSSIBLE);
    316   return;
    317 }
    318 
    319 Filter::Filter()
    320     : stream_buffer_(NULL),
    321       stream_buffer_size_(0),
    322       next_stream_data_(NULL),
    323       stream_data_len_(0),
    324       last_status_(FILTER_NEED_MORE_DATA) {}
    325 
    326 Filter::FilterStatus Filter::CopyOut(char* dest_buffer, int* dest_len) {
    327   int out_len;
    328   int input_len = *dest_len;
    329   *dest_len = 0;
    330 
    331   if (0 == stream_data_len_)
    332     return Filter::FILTER_NEED_MORE_DATA;
    333 
    334   out_len = std::min(input_len, stream_data_len_);
    335   memcpy(dest_buffer, next_stream_data_, out_len);
    336   *dest_len += out_len;
    337   stream_data_len_ -= out_len;
    338   if (0 == stream_data_len_) {
    339     next_stream_data_ = NULL;
    340     return Filter::FILTER_NEED_MORE_DATA;
    341   } else {
    342     next_stream_data_ += out_len;
    343     return Filter::FILTER_OK;
    344   }
    345 }
    346 
    347 // static
    348 Filter* Filter::InitGZipFilter(FilterType type_id, int buffer_size) {
    349   scoped_ptr<GZipFilter> gz_filter(new GZipFilter());
    350   gz_filter->InitBuffer(buffer_size);
    351   return gz_filter->InitDecoding(type_id) ? gz_filter.release() : NULL;
    352 }
    353 
    354 // static
    355 Filter* Filter::InitSdchFilter(FilterType type_id,
    356                                const FilterContext& filter_context,
    357                                int buffer_size) {
    358   scoped_ptr<SdchFilter> sdch_filter(new SdchFilter(filter_context));
    359   sdch_filter->InitBuffer(buffer_size);
    360   return sdch_filter->InitDecoding(type_id) ? sdch_filter.release() : NULL;
    361 }
    362 
    363 // static
    364 Filter* Filter::PrependNewFilter(FilterType type_id,
    365                                  const FilterContext& filter_context,
    366                                  int buffer_size,
    367                                  Filter* filter_list) {
    368   scoped_ptr<Filter> first_filter;  // Soon to be start of chain.
    369   switch (type_id) {
    370     case FILTER_TYPE_GZIP_HELPING_SDCH:
    371     case FILTER_TYPE_DEFLATE:
    372     case FILTER_TYPE_GZIP:
    373       first_filter.reset(InitGZipFilter(type_id, buffer_size));
    374       break;
    375     case FILTER_TYPE_SDCH:
    376     case FILTER_TYPE_SDCH_POSSIBLE:
    377       if (filter_context.GetURLRequestContext()->sdch_manager() &&
    378           SdchManager::sdch_enabled()) {
    379         first_filter.reset(
    380             InitSdchFilter(type_id, filter_context, buffer_size));
    381       }
    382       break;
    383     default:
    384       break;
    385   }
    386 
    387   if (!first_filter.get())
    388     return NULL;
    389 
    390   first_filter->next_filter_.reset(filter_list);
    391   return first_filter.release();
    392 }
    393 
    394 void Filter::InitBuffer(int buffer_size) {
    395   DCHECK(!stream_buffer());
    396   DCHECK_GT(buffer_size, 0);
    397   stream_buffer_ = new IOBuffer(buffer_size);
    398   stream_buffer_size_ = buffer_size;
    399 }
    400 
    401 void Filter::PushDataIntoNextFilter() {
    402   IOBuffer* next_buffer = next_filter_->stream_buffer();
    403   int next_size = next_filter_->stream_buffer_size();
    404   last_status_ = ReadFilteredData(next_buffer->data(), &next_size);
    405   if (FILTER_ERROR != last_status_)
    406     next_filter_->FlushStreamBuffer(next_size);
    407 }
    408 
    409 }  // namespace net
    410