First experimental implementation of the Clear-Site-Data header

content/browser/browsing_data/clear_site_data_throttle.cc

+// Copyright 2016 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 "content/browser/browsing_data/clear_site_data_throttle.h"
+
+#include "base/json/json_reader.h"
+#include "base/json/json_string_value_serializer.h"
+#include "base/strings/stringprintf.h"
+#include "base/values.h"
+#include "content/browser/frame_host/navigation_handle_impl.h"
+#include "content/public/browser/browser_context.h"
+#include "content/public/browser/content_browser_client.h"
+#include "content/public/browser/navigation_handle.h"
+#include "content/public/browser/web_contents.h"
+#include "content/public/common/content_client.h"
+#include "content/public/common/origin_util.h"
+#include "net/http/http_response_headers.h"
+#include "url/gurl.h"
+#include "url/origin.h"
+
+namespace content {
+
+namespace {
+
+static const char* kClearSiteDataHeader = "Clear-Site-Data";
+
+static const char* kTypesKey = "types";
+
+// Pretty-printed log output.
+static const char* kConsoleMessageFormat =
+    "Clear-Site-Data header on '%s': %s";
+static const char* kClearingOneType = "Clearing %s.";
+static const char* kClearingTwoTypes = "Clearing %s and %s.";
+static const char* kClearingThreeTypes = "Clearing %s, %s and %s.";

[Mike West, 2016/06/20 07:57:37]

Nit: Oxford comma, plz.

[msramek, 2016/07/15 16:47:39]

Done.

+
+}  // namespace
+
+// static
+std::unique_ptr<NavigationThrottle>
+    ClearSiteDataThrottle::CreateThrottleFor(NavigationHandle* handle) {
+  return std::unique_ptr<NavigationThrottle>(new ClearSiteDataThrottle(handle));
+}
+
+ClearSiteDataThrottle::~ClearSiteDataThrottle() {}
+
+ClearSiteDataThrottle::ThrottleCheckResult
+    ClearSiteDataThrottle::WillStartRequest() {
+  current_url_ = navigation_handle()->GetURL();
+  return PROCEED;
+}
+
+ClearSiteDataThrottle::ThrottleCheckResult
+    ClearSiteDataThrottle::WillRedirectRequest() {
+  // We are processing a redirect from url1 to url2. GetResponseHeaders()
+  // contains headers from url1, but GetURL() is already equal to url2. Handle
+  // the headers before updating the URL, so that |current_url_| corresponds
+  // to the URL that sent the headers.
+  HandleHeader();
+  current_url_ = navigation_handle()->GetURL();
+
+  return PROCEED;
+}
+
+ClearSiteDataThrottle::ThrottleCheckResult
+    ClearSiteDataThrottle::WillProcessResponse() {
+  HandleHeader();
+
+  // Now that RenderFrameHost is ready, output the console messages.
+  for (const ConsoleMessage& message : messages_) {
+    navigation_handle()->GetRenderFrameHost()->AddMessageToConsole(
+        message.level,
+        base::StringPrintf(
+            kConsoleMessageFormat, navigation_handle()->GetURL().spec().c_str(),
+            message.text.c_str()));
+  }
+
+  return PROCEED;
+}
+
+ClearSiteDataThrottle::ClearSiteDataThrottle(NavigationHandle* handle)
+    : NavigationThrottle(handle) {}
+
+void ClearSiteDataThrottle::HandleHeader() {
+  NavigationHandleImpl* handle =
+      static_cast<NavigationHandleImpl*>(navigation_handle());
+
+  // Ignore file URL navigations, as those have no response headers.
+  if (current_url_.SchemeIsFile())
+    return;
+
+  // Only accept the header on secure origins.
+  if (!IsOriginSecure(current_url_)) {
+    ConsoleLog(&messages_,
+               "Not supported for insecure origins.",
+               CONSOLE_MESSAGE_LEVEL_ERROR);
+    return;
+  }
+
+  // Extract the instances of the header and parse them.
+  size_t iter = 0;
+  std::string header_name;
+  std::string header_value;
+  while (handle->GetResponseHeaders()->EnumerateHeaderLines(
+      &iter, &header_name, &header_value)) {
+    if (header_name != kClearSiteDataHeader)
+      continue;
+
+    bool clear_cookies;
+    bool clear_storage;
+    bool clear_cache;
+
+    if (!ParseHeader(header_value, &clear_cookies, &clear_storage, &clear_cache,
+                     &messages_)) {
+      continue;
+    }
+
+    // If the header is valid, clear the data for this browser context
+    // and origin.
+    BrowserContext* browser_context =
+        handle->GetWebContents()->GetBrowserContext();
+    url::Origin origin(current_url_);
+
+    GetContentClient()->browser()->ClearSiteData(
+        browser_context, origin, clear_cookies, clear_storage, clear_cache);
+  }
+}
+
+bool ClearSiteDataThrottle::ParseHeader(
+    const std::string& header,
+    bool* clear_cookies, bool* clear_storage, bool* clear_cache,
+    std::vector<ConsoleMessage>* messages) {
+  std::unique_ptr<base::Value> parsed_header =
+      base::JSONReader::Read(header);
+
+  if (!parsed_header) {
+    ConsoleLog(messages,
+               base::StringPrintf("%s is not a valid JSON.", header.c_str()),
+               CONSOLE_MESSAGE_LEVEL_ERROR);
+    return false;
+  }
+
+  if (!parsed_header->GetAsDictionary(nullptr)) {
+    ConsoleLog(messages,
+               base::StringPrintf("%s is not a dictionary.", header.c_str()),
+               CONSOLE_MESSAGE_LEVEL_ERROR);
+    return false;
+  }

[Mike West, 2016/06/20 07:57:38]

It doesn't seem worthwhile to distinguish between this error case and the
following; I'd suggest just dropping this `if`, and changing the error message
on line 155 to something like "%s is not a JavaScript object with a 'types'
field."

[msramek, 2016/07/15 16:47:39]

Done. But I would still call it a JSON dictionary rather than a JavaScript
object. They're not the same thing (e.g. JSON doesn't allow trailing commas
while JS does), and we're really using a JSON parser here.

+
+  const base::ListValue* types;
+  if (!static_cast<base::DictionaryValue*>(parsed_header.get())
+          ->GetListWithoutPathExpansion(kTypesKey, &types)) {
+    ConsoleLog(messages,
+               base::StringPrintf(
+                   "No 'types' field present in %s.", header.c_str()),
+               CONSOLE_MESSAGE_LEVEL_ERROR);
+    return false;
+  }
+
+  *clear_cookies = false;
+  *clear_storage = false;
+  *clear_cache = false;
+
+  std::vector<std::string> type_names;
+  for (const std::unique_ptr<base::Value>& value : *types) {
+    std::string type;
+    value->GetAsString(&type);
+
+    bool* datatype = nullptr;
+
+    if (type == "cookies") {
+      datatype = clear_cookies;
+    } else if (type == "storage") {
+      datatype = clear_storage;
+    } else if (type == "cache") {
+      datatype = clear_cache;
+    } else {
+      std::string serialized_type;
+      JSONStringValueSerializer serializer(&serialized_type);
+      serializer.Serialize(*value);

[Mike West, 2016/06/20 07:57:38]

Nit: It doesn't look like you use |serialized_type| or |serializer| for
anything.

[msramek, 2016/07/15 16:47:39]

Done. Thanks for catching. If |value| is e.g. another dictionary, GetAsString()
above returns false, so |type| will be empty. In that case, I need to serialize
it to output it below.

+      ConsoleLog(messages,
+                 base::StringPrintf("Invalid type: '%s'.", type.c_str()),
+                 CONSOLE_MESSAGE_LEVEL_ERROR);
+      continue;
+    }
+
+    // Each data type should only be

[Mike West, 2016/06/20 07:57:37]

Nit: Should only be... ?

[msramek, 2016/07/15 16:47:39]

Correct. Every data type should exist, and nothing more! :)

( But really - it was supposed to say "...processed once". So that I don't have
to deal with duplicates. )

+    DCHECK(datatype != nullptr);

[Mike West, 2016/06/20 07:57:37]

Nit: `DCHECK(datatype)`

[msramek, 2016/07/15 16:47:39]

Done.

+    if (*datatype)
+      continue;
+
+    *datatype = true;
+    type_names.push_back(type);
+  }
+
+  if (!*clear_cookies && !*clear_storage && !*clear_cache) {
+    ConsoleLog(messages,
+               base::StringPrintf(
+                   "No valid types specified in %s.", header.c_str()),
+               CONSOLE_MESSAGE_LEVEL_ERROR);
+    return false;
+  }
+
+  // Pretty-print which types are to be cleared.
+  std::string output;
+  switch (type_names.size()) {
+    case 1:
+      output = base::StringPrintf(kClearingOneType, type_names[0].c_str());
+      break;
+    case 2:
+      output = base::StringPrintf(
+          kClearingTwoTypes, type_names[0].c_str(), type_names[1].c_str());
+      break;
+    case 3:
+      output = base::StringPrintf(
+          kClearingThreeTypes,
+          type_names[0].c_str(), type_names[1].c_str(), type_names[2].c_str());
+      break;
+    default:
+      NOTREACHED();
+  }
+  ConsoleLog(messages, output, CONSOLE_MESSAGE_LEVEL_LOG);

[Mike West, 2016/06/20 07:57:37]

Nit: Would you mind pulling this out into a static helper in the anonymous
namespace above?

[msramek, 2016/07/15 16:47:39]

Done. That also requires ConsoleMessage to be public.

+
+  return true;
+}
+
+void ClearSiteDataThrottle::ConsoleLog(
+    std::vector<ConsoleMessage>* messages,
+    const std::string& text, ConsoleMessageLevel level) {
+  messages->push_back({text, level});
+}
+
+}  // namespace content