net/proxy_resolution/pac_file_fetcher_impl.cc

// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#include "net/proxy_resolution/pac_file_fetcher_impl.h"

#include "base/compiler_specific.h"
#include "base/location.h"
#include "base/logging.h"
#include "base/metrics/histogram_macros.h"
#include "base/single_thread_task_runner.h"
#include "base/strings/string_util.h"
#include "base/threading/thread_task_runner_handle.h"
#include "net/base/data_url.h"
#include "net/base/io_buffer.h"
#include "net/base/load_flags.h"
#include "net/base/net_errors.h"
#include "net/base/net_string_util.h"
#include "net/base/request_priority.h"
#include "net/cert/cert_status_flags.h"
#include "net/http/http_response_headers.h"
#include "net/url_request/redirect_info.h"
#include "net/url_request/url_request_context.h"

// TODO(eroman):
//   - Support auth-prompts (http://crbug.com/77366)

namespace net {

namespace {

// The maximum size (in bytes) allowed for a PAC script. Responses exceeding
// this will fail with ERR_FILE_TOO_BIG.
const int kDefaultMaxResponseBytes = 1048576;  // 1 megabyte

// The maximum duration (in milliseconds) allowed for fetching the PAC script.
// Responses exceeding this will fail with ERR_TIMED_OUT.
//
// This timeout applies to both scripts fetched in the course of WPAD, as well
// as explicitly configured ones.
//
// If the default timeout is too high, auto-detect can stall for a long time,
// and if it is too low then slow loading scripts may be skipped.
//
// 30 seconds is a compromise between those competing goals. This value also
// appears to match Microsoft Edge (based on testing).
constexpr base::TimeDelta kDefaultMaxDuration =
    base::TimeDelta::FromSeconds(30);

// Returns true if |mime_type| is one of the known PAC mime type.
bool IsPacMimeType(const std::string& mime_type) {
  static const char* const kSupportedPacMimeTypes[] = {
      "application/x-ns-proxy-autoconfig", "application/x-javascript-config",
  };
  for (size_t i = 0; i < arraysize(kSupportedPacMimeTypes); ++i) {
    if (base::LowerCaseEqualsASCII(mime_type, kSupportedPacMimeTypes[i]))
      return true;
  }
  return false;
}

// Converts |bytes| (which is encoded by |charset|) to UTF16, saving the resul
// to |*utf16|.
// If |charset| is empty, then we don't know what it was and guess.
void ConvertResponseToUTF16(const std::string& charset,
                            const std::string& bytes,
                            base::string16* utf16) {
  const char* codepage;

  if (charset.empty()) {
    // Assume ISO-8859-1 if no charset was specified.
    codepage = kCharsetLatin1;
  } else {
    // Otherwise trust the charset that was provided.
    codepage = charset.c_str();
  }

  // Be generous in the conversion -- if any characters lie outside of |charset|
  // (i.e. invalid), then substitute them with U+FFFD rather than failing.
  ConvertToUTF16WithSubstitutions(bytes, codepage, utf16);
}

}  // namespace

std::unique_ptr<PacFileFetcherImpl> PacFileFetcherImpl::Create(
    URLRequestContext* url_request_context) {
  return base::WrapUnique(new PacFileFetcherImpl(url_request_context, false));
}

std::unique_ptr<PacFileFetcherImpl>
PacFileFetcherImpl::CreateWithFileUrlSupport(
    URLRequestContext* url_request_context) {
  return base::WrapUnique(new PacFileFetcherImpl(url_request_context, true));
}

PacFileFetcherImpl::~PacFileFetcherImpl() {
  // The URLRequest's destructor will cancel the outstanding request, and
  // ensure that the delegate (this) is not called again.
}

base::TimeDelta PacFileFetcherImpl::SetTimeoutConstraint(
    base::TimeDelta timeout) {
  base::TimeDelta prev = max_duration_;
  max_duration_ = timeout;
  return prev;
}

size_t PacFileFetcherImpl::SetSizeConstraint(size_t size_bytes) {
  size_t prev = max_response_bytes_;
  max_response_bytes_ = size_bytes;
  return prev;
}

void PacFileFetcherImpl::OnResponseCompleted(URLRequest* request,
                                             int net_error) {
  DCHECK_EQ(request, cur_request_.get());

  // Use |result_code_| as the request's error if we have already set it to
  // something specific.
  if (result_code_ == OK && net_error != OK)
    result_code_ = net_error;

  FetchCompleted();
}

int PacFileFetcherImpl::Fetch(
    const GURL& url,
    base::string16* text,
    CompletionOnceCallback callback,
    const NetworkTrafficAnnotationTag traffic_annotation) {
  // It is invalid to call Fetch() while a request is already in progress.
  DCHECK(!cur_request_.get());
  DCHECK(!callback.is_null());
  DCHECK(text);

  if (!url_request_context_)
    return ERR_CONTEXT_SHUT_DOWN;

  if (!IsUrlSchemeAllowed(url))
    return ERR_DISALLOWED_URL_SCHEME;

  // Handle base-64 encoded data-urls that contain custom PAC scripts.
  if (url.SchemeIs("data")) {
    std::string mime_type;
    std::string charset;
    std::string data;
    if (!DataURL::Parse(url, &mime_type, &charset, &data))
      return ERR_FAILED;

    ConvertResponseToUTF16(charset, data, text);
    return OK;
  }

  DCHECK(fetch_start_time_.is_null());
  fetch_start_time_ = base::TimeTicks::Now();

  // Use highest priority, so if socket pools are being used for other types of
  // requests, PAC requests are aren't blocked on them.
  cur_request_ = url_request_context_->CreateRequest(url, MAXIMUM_PRIORITY,
                                                     this, traffic_annotation);
  cur_request_->set_is_pac_request(true);

  // Make sure that the PAC script is downloaded using a direct connection,
  // to avoid circular dependencies (fetching is a part of proxy resolution).
  // Also disable the use of the disk cache. The cache is disabled so that if
  // the user switches networks we don't potentially use the cached response
  // from old network when we should in fact be re-fetching on the new network.
  // If the PAC script is hosted on an HTTPS server we bypass revocation
  // checking in order to avoid a circular dependency when attempting to fetch
  // the OCSP response or CRL. We could make the revocation check go direct but
  // the proxy might be the only way to the outside world.  IGNORE_LIMITS is
  // used to avoid blocking proxy resolution on other network requests.
  cur_request_->SetLoadFlags(LOAD_BYPASS_PROXY | LOAD_DISABLE_CACHE |
                             LOAD_DISABLE_CERT_NETWORK_FETCHES |
                             LOAD_IGNORE_LIMITS);

  // Save the caller's info for notification on completion.
  callback_ = std::move(callback);
  result_text_ = text;

  bytes_read_so_far_.clear();

  // Post a task to timeout this request if it takes too long.
  cur_request_id_ = ++next_id_;

  base::ThreadTaskRunnerHandle::Get()->PostDelayedTask(
      FROM_HERE,
      base::Bind(&PacFileFetcherImpl::OnTimeout, weak_factory_.GetWeakPtr(),
                 cur_request_id_),
      max_duration_);

  // Start the request.
  cur_request_->Start();
  return ERR_IO_PENDING;
}

void PacFileFetcherImpl::Cancel() {
  // ResetCurRequestState will free the URLRequest, which will cause
  // cancellation.
  ResetCurRequestState();
}

URLRequestContext* PacFileFetcherImpl::GetRequestContext() const {
  return url_request_context_;
}

void PacFileFetcherImpl::OnShutdown() {
  url_request_context_ = nullptr;

  if (cur_request_) {
    result_code_ = ERR_CONTEXT_SHUT_DOWN;
    FetchCompleted();
  }
}

void PacFileFetcherImpl::OnReceivedRedirect(URLRequest* request,
                                            const RedirectInfo& redirect_info,
                                            bool* defer_redirect) {
  int error = OK;

  // Redirection to file:// is never OK. Ordinarily this is handled lower in the
  // stack (|FileProtocolHandler::IsSafeRedirectTarget|), but this is reachable
  // when built without file:// suppport. Return the same error for consistency.
  if (redirect_info.new_url.SchemeIsFile()) {
    error = ERR_UNSAFE_REDIRECT;
  } else if (!IsUrlSchemeAllowed(redirect_info.new_url)) {
    error = ERR_DISALLOWED_URL_SCHEME;
  }

  if (error != OK) {
    // Fail the redirect.
    request->CancelWithError(error);
    OnResponseCompleted(request, error);
  }
}

void PacFileFetcherImpl::OnAuthRequired(URLRequest* request,
                                        AuthChallengeInfo* auth_info) {
  DCHECK_EQ(request, cur_request_.get());
  // TODO(eroman): http://crbug.com/77366
  LOG(WARNING) << "Auth required to fetch PAC script, aborting.";
  result_code_ = ERR_NOT_IMPLEMENTED;
  request->CancelAuth();
}

void PacFileFetcherImpl::OnSSLCertificateError(URLRequest* request,
                                               const SSLInfo& ssl_info,
                                               bool fatal) {
  DCHECK_EQ(request, cur_request_.get());
  // Revocation check failures are not fatal.
  if (IsCertStatusMinorError(ssl_info.cert_status)) {
    request->ContinueDespiteLastError();
    return;
  }
  LOG(WARNING) << "SSL certificate error when fetching PAC script, aborting.";
  // Certificate errors are in same space as net errors.
  result_code_ = MapCertStatusToNetError(ssl_info.cert_status);
  request->Cancel();
}

void PacFileFetcherImpl::OnResponseStarted(URLRequest* request, int net_error) {
  DCHECK_EQ(request, cur_request_.get());
  DCHECK_NE(ERR_IO_PENDING, net_error);

  if (net_error != OK) {
    OnResponseCompleted(request, net_error);
    return;
  }

  // Require HTTP responses to have a success status code.
  if (request->url().SchemeIsHTTPOrHTTPS()) {
    // NOTE about status codes: We are like Firefox 3 in this respect.
    // {IE 7, Safari 3, Opera 9.5} do not care about the status code.
    if (request->GetResponseCode() != 200) {
      VLOG(1) << "Fetched PAC script had (bad) status line: "
              << request->response_headers()->GetStatusLine();
      result_code_ = ERR_PAC_STATUS_NOT_OK;
      request->Cancel();
      return;
    }

    // NOTE about mime types: We do not enforce mime types on PAC files.
    // This is for compatibility with {IE 7, Firefox 3, Opera 9.5}. We will
    // however log mismatches to help with debugging.
    std::string mime_type;
    cur_request_->GetMimeType(&mime_type);
    if (!IsPacMimeType(mime_type)) {
      VLOG(1) << "Fetched PAC script does not have a proper mime type: "
              << mime_type;
    }
  }

  ReadBody(request);
}

void PacFileFetcherImpl::OnReadCompleted(URLRequest* request, int num_bytes) {
  DCHECK_NE(ERR_IO_PENDING, num_bytes);

  DCHECK_EQ(request, cur_request_.get());
  if (ConsumeBytesRead(request, num_bytes)) {
    // Keep reading.
    ReadBody(request);
  }
}

PacFileFetcherImpl::PacFileFetcherImpl(URLRequestContext* url_request_context,
                                       bool allow_file_url)
    : url_request_context_(url_request_context),
      buf_(new IOBuffer(kBufSize)),
      next_id_(0),
      cur_request_id_(0),
      result_code_(OK),
      result_text_(NULL),
      max_response_bytes_(kDefaultMaxResponseBytes),
      max_duration_(kDefaultMaxDuration),
      allow_file_url_(allow_file_url),
      weak_factory_(this) {
  DCHECK(url_request_context);
}

bool PacFileFetcherImpl::IsUrlSchemeAllowed(const GURL& url) const {
  // Always allow http://, https://, data:, and ftp://.
  if (url.SchemeIsHTTPOrHTTPS() || url.SchemeIs("ftp") || url.SchemeIs("data"))
    return true;

  // Only permit file:// if |allow_file_url_| was set. file:// should not be
  // allowed for URLs that were auto-detected, or as the result of a server-side
  // redirect.
  if (url.SchemeIsFile())
    return allow_file_url_;

  // Disallow any other URL scheme.
  return false;
}

void PacFileFetcherImpl::ReadBody(URLRequest* request) {
  // Read as many bytes as are available synchronously.
  while (true) {
    int num_bytes = request->Read(buf_.get(), kBufSize);
    if (num_bytes == ERR_IO_PENDING)
      return;

    if (num_bytes < 0) {
      OnResponseCompleted(request, num_bytes);
      return;
    }

    if (!ConsumeBytesRead(request, num_bytes))
      return;
  }
}

bool PacFileFetcherImpl::ConsumeBytesRead(URLRequest* request, int num_bytes) {
  if (fetch_time_to_first_byte_.is_null())
    fetch_time_to_first_byte_ = base::TimeTicks::Now();

  if (num_bytes <= 0) {
    // Error while reading, or EOF.
    OnResponseCompleted(request, num_bytes);
    return false;
  }

  // Enforce maximum size bound.
  if (num_bytes + bytes_read_so_far_.size() >
      static_cast<size_t>(max_response_bytes_)) {
    result_code_ = ERR_FILE_TOO_BIG;
    request->Cancel();
    return false;
  }

  bytes_read_so_far_.append(buf_->data(), num_bytes);
  return true;
}

void PacFileFetcherImpl::FetchCompleted() {
  if (result_code_ == OK) {
    // Calculate duration of time for PAC file fetch to complete.
    DCHECK(!fetch_start_time_.is_null());
    DCHECK(!fetch_time_to_first_byte_.is_null());
    UMA_HISTOGRAM_MEDIUM_TIMES("Net.ProxyScriptFetcher.SuccessDuration",
                               base::TimeTicks::Now() - fetch_start_time_);
    UMA_HISTOGRAM_MEDIUM_TIMES("Net.ProxyScriptFetcher.FirstByteDuration",
                               fetch_time_to_first_byte_ - fetch_start_time_);

    // The caller expects the response to be encoded as UTF16.
    std::string charset;
    cur_request_->GetCharset(&charset);
    ConvertResponseToUTF16(charset, bytes_read_so_far_, result_text_);
  } else {
    // On error, the caller expects empty string for bytes.
    result_text_->clear();
  }

  int result_code = result_code_;
  CompletionOnceCallback callback = std::move(callback_);

  ResetCurRequestState();

  std::move(callback).Run(result_code);
}

void PacFileFetcherImpl::ResetCurRequestState() {
  cur_request_.reset();
  cur_request_id_ = 0;
  callback_.Reset();
  result_code_ = OK;
  result_text_ = NULL;
  fetch_start_time_ = base::TimeTicks();
  fetch_time_to_first_byte_ = base::TimeTicks();
}

void PacFileFetcherImpl::OnTimeout(int id) {
  // Timeout tasks may outlive the URLRequest they reference. Make sure it
  // is still applicable.
  if (cur_request_id_ != id)
    return;

  DCHECK(cur_request_.get());
  result_code_ = ERR_TIMED_OUT;
  FetchCompleted();
}

}  // namespace net