xref: /aosp_15_r20/external/cronet/net/base/data_url.cc (revision 6777b5387eb2ff775bb5750e3f5d96f37fb7352b)
1 // Copyright 2012 The Chromium Authors
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4 
5 // NOTE: based loosely on mozilla's nsDataChannel.cpp
6 
7 #include "net/base/data_url.h"
8 
9 #include <string>
10 #include <string_view>
11 
12 #include "base/base64.h"
13 #include "base/feature_list.h"
14 #include "base/features.h"
15 #include "base/ranges/algorithm.h"
16 #include "base/strings/escape.h"
17 #include "base/strings/string_split.h"
18 #include "base/strings/string_util.h"
19 #include "net/base/mime_util.h"
20 #include "net/http/http_response_headers.h"
21 #include "net/http/http_util.h"
22 #include "url/gurl.h"
23 
24 namespace net {
25 namespace {
26 
27 // https://infra.spec.whatwg.org/#ascii-whitespace, which is referenced by
28 // https://infra.spec.whatwg.org/#forgiving-base64, does not include \v in the
29 // set of ASCII whitespace characters the way Unicode does.
IsBase64Whitespace(char c)30 bool IsBase64Whitespace(char c) {
31   return c != '\v' && base::IsAsciiWhitespace(c);
32 }
33 
34 // A data URL is ready for decode if it:
35 //   - Doesn't need any extra padding.
36 //   - Does not have any escaped characters.
37 //   - Does not have any whitespace.
IsDataURLReadyForDecode(std::string_view body)38 bool IsDataURLReadyForDecode(std::string_view body) {
39   return (body.length() % 4) == 0 && base::ranges::none_of(body, [](char c) {
40            return c == '%' || IsBase64Whitespace(c);
41          });
42 }
43 
44 }  // namespace
45 
Parse(const GURL & url,std::string * mime_type,std::string * charset,std::string * data)46 bool DataURL::Parse(const GURL& url,
47                     std::string* mime_type,
48                     std::string* charset,
49                     std::string* data) {
50   if (!url.is_valid() || !url.has_scheme())
51     return false;
52 
53   DCHECK(mime_type->empty());
54   DCHECK(charset->empty());
55   DCHECK(!data || data->empty());
56 
57   std::string_view content;
58   std::string content_string;
59   if (base::FeatureList::IsEnabled(base::features::kOptimizeDataUrls)) {
60     // Avoid copying the URL content which can be expensive for large URLs.
61     content = url.GetContentPiece();
62   } else {
63     content_string = url.GetContent();
64     content = content_string;
65   }
66 
67   std::string_view::const_iterator comma = base::ranges::find(content, ',');
68   if (comma == content.end())
69     return false;
70 
71   std::vector<std::string_view> meta_data =
72       base::SplitStringPiece(base::MakeStringPiece(content.begin(), comma), ";",
73                              base::TRIM_WHITESPACE, base::SPLIT_WANT_ALL);
74 
75   // These are moved to |mime_type| and |charset| on success.
76   std::string mime_type_value;
77   std::string charset_value;
78   auto iter = meta_data.cbegin();
79   if (iter != meta_data.cend()) {
80     mime_type_value = base::ToLowerASCII(*iter);
81     ++iter;
82   }
83 
84   static constexpr std::string_view kBase64Tag("base64");
85   static constexpr std::string_view kCharsetTag("charset=");
86 
87   bool base64_encoded = false;
88   for (; iter != meta_data.cend(); ++iter) {
89     if (!base64_encoded &&
90         base::EqualsCaseInsensitiveASCII(*iter, kBase64Tag)) {
91       base64_encoded = true;
92     } else if (charset_value.empty() &&
93                base::StartsWith(*iter, kCharsetTag,
94                                 base::CompareCase::INSENSITIVE_ASCII)) {
95       charset_value = std::string(iter->substr(kCharsetTag.size()));
96       // The grammar for charset is not specially defined in RFC2045 and
97       // RFC2397. It just needs to be a token.
98       if (!HttpUtil::IsToken(charset_value))
99         return false;
100     }
101   }
102 
103   if (mime_type_value.empty()) {
104     // Fallback to the default if nothing specified in the mediatype part as
105     // specified in RFC2045. As specified in RFC2397, we use |charset| even if
106     // |mime_type| is empty.
107     mime_type_value = "text/plain";
108     if (charset_value.empty())
109       charset_value = "US-ASCII";
110   } else if (!ParseMimeTypeWithoutParameter(mime_type_value, nullptr,
111                                             nullptr)) {
112     // Fallback to the default as recommended in RFC2045 when the mediatype
113     // value is invalid. For this case, we don't respect |charset| but force it
114     // set to "US-ASCII".
115     mime_type_value = "text/plain";
116     charset_value = "US-ASCII";
117   }
118 
119   // The caller may not be interested in receiving the data.
120   if (data) {
121     // Preserve spaces if dealing with text or xml input, same as mozilla:
122     //   https://bugzilla.mozilla.org/show_bug.cgi?id=138052
123     // but strip them otherwise:
124     //   https://bugzilla.mozilla.org/show_bug.cgi?id=37200
125     // (Spaces in a data URL should be escaped, which is handled below, so any
126     // spaces now are wrong. People expect to be able to enter them in the URL
127     // bar for text, and it can't hurt, so we allow it.)
128     //
129     // TODO(mmenke): Is removing all spaces reasonable? GURL removes trailing
130     // spaces itself, anyways. Should we just trim leading spaces instead?
131     // Allowing random intermediary spaces seems unnecessary.
132 
133     auto raw_body = base::MakeStringPiece(comma + 1, content.end());
134 
135     // For base64, we may have url-escaped whitespace which is not part
136     // of the data, and should be stripped. Otherwise, the escaped whitespace
137     // could be part of the payload, so don't strip it.
138     if (base64_encoded) {
139       // If the data URL is well formed, we can decode it immediately.
140       if (base::FeatureList::IsEnabled(base::features::kOptimizeDataUrls) &&
141           IsDataURLReadyForDecode(raw_body)) {
142         if (!base::Base64Decode(raw_body, data))
143           return false;
144       } else {
145         std::string unescaped_body = base::UnescapeBinaryURLComponent(raw_body);
146         if (!base::Base64Decode(unescaped_body, data,
147                                 base::Base64DecodePolicy::kForgiving))
148           return false;
149       }
150     } else {
151       // Strip whitespace for non-text MIME types.
152       std::string temp;
153       if (!(mime_type_value.compare(0, 5, "text/") == 0 ||
154             mime_type_value.find("xml") != std::string::npos)) {
155         temp = std::string(raw_body);
156         std::erase_if(temp, base::IsAsciiWhitespace<char>);
157         raw_body = temp;
158       }
159 
160       *data = base::UnescapeBinaryURLComponent(raw_body);
161     }
162   }
163 
164   *mime_type = std::move(mime_type_value);
165   *charset = std::move(charset_value);
166   return true;
167 }
168 
BuildResponse(const GURL & url,std::string_view method,std::string * mime_type,std::string * charset,std::string * data,scoped_refptr<HttpResponseHeaders> * headers)169 Error DataURL::BuildResponse(const GURL& url,
170                              std::string_view method,
171                              std::string* mime_type,
172                              std::string* charset,
173                              std::string* data,
174                              scoped_refptr<HttpResponseHeaders>* headers) {
175   DCHECK(data);
176   DCHECK(!*headers);
177 
178   if (!DataURL::Parse(url, mime_type, charset, data))
179     return ERR_INVALID_URL;
180 
181   // |mime_type| set by DataURL::Parse() is guaranteed to be in
182   //     token "/" token
183   // form. |charset| can be an empty string.
184   DCHECK(!mime_type->empty());
185 
186   // "charset" in the Content-Type header is specified explicitly to follow
187   // the "token" ABNF in the HTTP spec. When the DataURL::Parse() call is
188   // successful, it's guaranteed that the string in |charset| follows the
189   // "token" ABNF.
190   std::string content_type = *mime_type;
191   if (!charset->empty())
192     content_type.append(";charset=" + *charset);
193   // The terminal double CRLF isn't needed by TryToCreate().
194   *headers = HttpResponseHeaders::TryToCreate(
195       "HTTP/1.1 200 OK\r\n"
196       "Content-Type:" +
197       content_type);
198   // Above line should always succeed - TryToCreate() only fails when there are
199   // nulls in the string, and DataURL::Parse() can't return nulls in anything
200   // but the |data| argument.
201   DCHECK(*headers);
202 
203   if (base::EqualsCaseInsensitiveASCII(method, "HEAD"))
204     data->clear();
205 
206   return OK;
207 }
208 
209 }  // namespace net
210