1// Copyright (c) 2012 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/proxy/proxy_script_fetcher_impl.h" 6 7#include "base/compiler_specific.h" 8#include "base/i18n/icu_string_conversions.h" 9#include "base/logging.h" 10#include "base/message_loop/message_loop.h" 11#include "base/strings/string_util.h" 12#include "net/base/data_url.h" 13#include "net/base/io_buffer.h" 14#include "net/base/load_flags.h" 15#include "net/base/net_errors.h" 16#include "net/cert/cert_status_flags.h" 17#include "net/http/http_response_headers.h" 18#include "net/url_request/url_request_context.h" 19 20// TODO(eroman): 21// - Support auth-prompts (http://crbug.com/77366) 22 23namespace net { 24 25namespace { 26 27// The maximum size (in bytes) allowed for a PAC script. Responses exceeding 28// this will fail with ERR_FILE_TOO_BIG. 29const int kDefaultMaxResponseBytes = 1048576; // 1 megabyte 30 31// The maximum duration (in milliseconds) allowed for fetching the PAC script. 32// Responses exceeding this will fail with ERR_TIMED_OUT. 33const int kDefaultMaxDurationMs = 300000; // 5 minutes 34 35// Returns true if |mime_type| is one of the known PAC mime type. 36bool IsPacMimeType(const std::string& mime_type) { 37 static const char * const kSupportedPacMimeTypes[] = { 38 "application/x-ns-proxy-autoconfig", 39 "application/x-javascript-config", 40 }; 41 for (size_t i = 0; i < arraysize(kSupportedPacMimeTypes); ++i) { 42 if (LowerCaseEqualsASCII(mime_type, kSupportedPacMimeTypes[i])) 43 return true; 44 } 45 return false; 46} 47 48// Converts |bytes| (which is encoded by |charset|) to UTF16, saving the resul 49// to |*utf16|. 50// If |charset| is empty, then we don't know what it was and guess. 51void ConvertResponseToUTF16(const std::string& charset, 52 const std::string& bytes, 53 base::string16* utf16) { 54 const char* codepage; 55 56 if (charset.empty()) { 57 // Assume ISO-8859-1 if no charset was specified. 58 codepage = base::kCodepageLatin1; 59 } else { 60 // Otherwise trust the charset that was provided. 61 codepage = charset.c_str(); 62 } 63 64 // We will be generous in the conversion -- if any characters lie 65 // outside of |charset| (i.e. invalid), then substitute them with 66 // U+FFFD rather than failing. 67 base::CodepageToUTF16(bytes, codepage, 68 base::OnStringConversionError::SUBSTITUTE, 69 utf16); 70} 71 72} // namespace 73 74ProxyScriptFetcherImpl::ProxyScriptFetcherImpl( 75 URLRequestContext* url_request_context) 76 : weak_factory_(this), 77 url_request_context_(url_request_context), 78 buf_(new IOBuffer(kBufSize)), 79 next_id_(0), 80 cur_request_id_(0), 81 result_code_(OK), 82 result_text_(NULL), 83 max_response_bytes_(kDefaultMaxResponseBytes), 84 max_duration_(base::TimeDelta::FromMilliseconds(kDefaultMaxDurationMs)) { 85 DCHECK(url_request_context); 86} 87 88ProxyScriptFetcherImpl::~ProxyScriptFetcherImpl() { 89 // The URLRequest's destructor will cancel the outstanding request, and 90 // ensure that the delegate (this) is not called again. 91} 92 93base::TimeDelta ProxyScriptFetcherImpl::SetTimeoutConstraint( 94 base::TimeDelta timeout) { 95 base::TimeDelta prev = max_duration_; 96 max_duration_ = timeout; 97 return prev; 98} 99 100size_t ProxyScriptFetcherImpl::SetSizeConstraint(size_t size_bytes) { 101 size_t prev = max_response_bytes_; 102 max_response_bytes_ = size_bytes; 103 return prev; 104} 105 106void ProxyScriptFetcherImpl::OnResponseCompleted(URLRequest* request) { 107 DCHECK_EQ(request, cur_request_.get()); 108 109 // Use |result_code_| as the request's error if we have already set it to 110 // something specific. 111 if (result_code_ == OK && !request->status().is_success()) 112 result_code_ = request->status().error(); 113 114 FetchCompleted(); 115} 116 117int ProxyScriptFetcherImpl::Fetch( 118 const GURL& url, base::string16* text, const CompletionCallback& callback) { 119 // It is invalid to call Fetch() while a request is already in progress. 120 DCHECK(!cur_request_.get()); 121 DCHECK(!callback.is_null()); 122 DCHECK(text); 123 124 // Handle base-64 encoded data-urls that contain custom PAC scripts. 125 if (url.SchemeIs("data")) { 126 std::string mime_type; 127 std::string charset; 128 std::string data; 129 if (!DataURL::Parse(url, &mime_type, &charset, &data)) 130 return ERR_FAILED; 131 132 ConvertResponseToUTF16(charset, data, text); 133 return OK; 134 } 135 136 cur_request_.reset(url_request_context_->CreateRequest(url, this)); 137 cur_request_->set_method("GET"); 138 139 // Make sure that the PAC script is downloaded using a direct connection, 140 // to avoid circular dependencies (fetching is a part of proxy resolution). 141 // Also disable the use of the disk cache. The cache is disabled so that if 142 // the user switches networks we don't potentially use the cached response 143 // from old network when we should in fact be re-fetching on the new network. 144 // If the PAC script is hosted on an HTTPS server we bypass revocation 145 // checking in order to avoid a circular dependency when attempting to fetch 146 // the OCSP response or CRL. We could make the revocation check go direct but 147 // the proxy might be the only way to the outside world. 148 cur_request_->set_load_flags(LOAD_BYPASS_PROXY | LOAD_DISABLE_CACHE | 149 LOAD_DISABLE_CERT_REVOCATION_CHECKING); 150 151 // Save the caller's info for notification on completion. 152 callback_ = callback; 153 result_text_ = text; 154 155 bytes_read_so_far_.clear(); 156 157 // Post a task to timeout this request if it takes too long. 158 cur_request_id_ = ++next_id_; 159 base::MessageLoop::current()->PostDelayedTask( 160 FROM_HERE, 161 base::Bind(&ProxyScriptFetcherImpl::OnTimeout, 162 weak_factory_.GetWeakPtr(), 163 cur_request_id_), 164 max_duration_); 165 166 // Start the request. 167 cur_request_->Start(); 168 return ERR_IO_PENDING; 169} 170 171void ProxyScriptFetcherImpl::Cancel() { 172 // ResetCurRequestState will free the URLRequest, which will cause 173 // cancellation. 174 ResetCurRequestState(); 175} 176 177URLRequestContext* ProxyScriptFetcherImpl::GetRequestContext() const { 178 return url_request_context_; 179} 180 181void ProxyScriptFetcherImpl::OnAuthRequired(URLRequest* request, 182 AuthChallengeInfo* auth_info) { 183 DCHECK_EQ(request, cur_request_.get()); 184 // TODO(eroman): http://crbug.com/77366 185 LOG(WARNING) << "Auth required to fetch PAC script, aborting."; 186 result_code_ = ERR_NOT_IMPLEMENTED; 187 request->CancelAuth(); 188} 189 190void ProxyScriptFetcherImpl::OnSSLCertificateError(URLRequest* request, 191 const SSLInfo& ssl_info, 192 bool fatal) { 193 DCHECK_EQ(request, cur_request_.get()); 194 // Revocation check failures are not fatal. 195 if (IsCertStatusMinorError(ssl_info.cert_status)) { 196 request->ContinueDespiteLastError(); 197 return; 198 } 199 LOG(WARNING) << "SSL certificate error when fetching PAC script, aborting."; 200 // Certificate errors are in same space as net errors. 201 result_code_ = MapCertStatusToNetError(ssl_info.cert_status); 202 request->Cancel(); 203} 204 205void ProxyScriptFetcherImpl::OnResponseStarted(URLRequest* request) { 206 DCHECK_EQ(request, cur_request_.get()); 207 208 if (!request->status().is_success()) { 209 OnResponseCompleted(request); 210 return; 211 } 212 213 // Require HTTP responses to have a success status code. 214 if (request->url().SchemeIs("http") || request->url().SchemeIs("https")) { 215 // NOTE about status codes: We are like Firefox 3 in this respect. 216 // {IE 7, Safari 3, Opera 9.5} do not care about the status code. 217 if (request->GetResponseCode() != 200) { 218 VLOG(1) << "Fetched PAC script had (bad) status line: " 219 << request->response_headers()->GetStatusLine(); 220 result_code_ = ERR_PAC_STATUS_NOT_OK; 221 request->Cancel(); 222 return; 223 } 224 225 // NOTE about mime types: We do not enforce mime types on PAC files. 226 // This is for compatibility with {IE 7, Firefox 3, Opera 9.5}. We will 227 // however log mismatches to help with debugging. 228 std::string mime_type; 229 cur_request_->GetMimeType(&mime_type); 230 if (!IsPacMimeType(mime_type)) { 231 VLOG(1) << "Fetched PAC script does not have a proper mime type: " 232 << mime_type; 233 } 234 } 235 236 ReadBody(request); 237} 238 239void ProxyScriptFetcherImpl::OnReadCompleted(URLRequest* request, 240 int num_bytes) { 241 DCHECK_EQ(request, cur_request_.get()); 242 if (ConsumeBytesRead(request, num_bytes)) { 243 // Keep reading. 244 ReadBody(request); 245 } 246} 247 248void ProxyScriptFetcherImpl::ReadBody(URLRequest* request) { 249 // Read as many bytes as are available synchronously. 250 while (true) { 251 int num_bytes; 252 if (!request->Read(buf_.get(), kBufSize, &num_bytes)) { 253 // Check whether the read failed synchronously. 254 if (!request->status().is_io_pending()) 255 OnResponseCompleted(request); 256 return; 257 } 258 if (!ConsumeBytesRead(request, num_bytes)) 259 return; 260 } 261} 262 263bool ProxyScriptFetcherImpl::ConsumeBytesRead(URLRequest* request, 264 int num_bytes) { 265 if (num_bytes <= 0) { 266 // Error while reading, or EOF. 267 OnResponseCompleted(request); 268 return false; 269 } 270 271 // Enforce maximum size bound. 272 if (num_bytes + bytes_read_so_far_.size() > 273 static_cast<size_t>(max_response_bytes_)) { 274 result_code_ = ERR_FILE_TOO_BIG; 275 request->Cancel(); 276 return false; 277 } 278 279 bytes_read_so_far_.append(buf_->data(), num_bytes); 280 return true; 281} 282 283void ProxyScriptFetcherImpl::FetchCompleted() { 284 if (result_code_ == OK) { 285 // The caller expects the response to be encoded as UTF16. 286 std::string charset; 287 cur_request_->GetCharset(&charset); 288 ConvertResponseToUTF16(charset, bytes_read_so_far_, result_text_); 289 } else { 290 // On error, the caller expects empty string for bytes. 291 result_text_->clear(); 292 } 293 294 int result_code = result_code_; 295 CompletionCallback callback = callback_; 296 297 ResetCurRequestState(); 298 299 callback.Run(result_code); 300} 301 302void ProxyScriptFetcherImpl::ResetCurRequestState() { 303 cur_request_.reset(); 304 cur_request_id_ = 0; 305 callback_.Reset(); 306 result_code_ = OK; 307 result_text_ = NULL; 308} 309 310void ProxyScriptFetcherImpl::OnTimeout(int id) { 311 // Timeout tasks may outlive the URLRequest they reference. Make sure it 312 // is still applicable. 313 if (cur_request_id_ != id) 314 return; 315 316 DCHECK(cur_request_.get()); 317 result_code_ = ERR_TIMED_OUT; 318 cur_request_->Cancel(); 319} 320 321} // namespace net 322