WIP: URL blacklisting by policy.

chrome/browser/policy/url_blacklist_manager.cc

+// Copyright (c) 2011 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 "chrome/browser/policy/url_blacklist_manager.h"
+
+#include "base/string_number_conversions.h"
+#include "base/string_util.h"
+#include "base/values.h"
+#include "chrome/browser/net/url_fixer_upper.h"
+#include "chrome/browser/prefs/pref_service.h"
+#include "chrome/browser/profiles/profile.h"
+#include "chrome/common/chrome_notification_types.h"
+#include "chrome/common/pref_names.h"
+#include "content/browser/browser_thread.h"
+#include "content/common/notification_details.h"
+#include "content/common/notification_source.h"
+#include "googleurl/src/gurl.h"
+
+namespace policy {
+
+namespace {
+
+// Time to wait before starting an update of the blacklist. Scheduling another
+// update during this period will reset the timer.
+const int64 kUpdateDelayMs = 1000;
+
+// Maximum filters per policy. Filters over this index are ignored.
+const size_t kMaxFiltersPerPolicy = 100;
+
+typedef std::vector<std::string> StringVector;
+
+}  // namespace
+
+Blacklist::Blacklist() {
+}
+
+Blacklist::~Blacklist() {
+}
+
+void Blacklist::AddFilter(const std::string& filter, bool block) {
+  std::string scheme;
+  std::string host;
+  uint16 port;
+  std::string path;
+  SchemeFlag flag;
+  bool match_subdomains = true;
+
+  if (!FilterToComponents(filter, &scheme, &host, &port, &path)) {
+    LOG(WARNING) << "Invalid filter, ignoring: " << filter;
+    return;
+  }
+
+  if (!SchemeToFlag(scheme, &flag)) {
+    LOG(WARNING) << "Unsupported scheme in filter, ignoring filter: " << filter;
+    return;
+  }
+
+  // Special syntax to disable subdomain matching.
+  if (!host.empty() && host[0] == '.') {
+    host.erase(0, 1);
+    match_subdomains = false;
+  }
+
+  // Try to find an existing PathFilter with the same path prefix, port and
+  // match_subdomains value.
+  PathFilterList& list = host_filters_[host];
+  PathFilterList::iterator it;
+  for (it = list.begin(); it != list.end(); ++it) {
+    if (it->port == port && it->match_subdomains == match_subdomains &&
+        it->path_prefix == path)
+      break;
+  }
+  PathFilter* f;

[Mattias Nissler (ping if slow), 2011/08/31 14:58:09]

use a better name instead of f, maybe filter?

[Joao da Silva, 2011/09/01 12:47:36]

Done. |filter| is taken, renamed to |path_filter|.

+  if (it == list.end()) {
+    list.push_back(PathFilter(path, port, match_subdomains));
+    f = &list.back();
+  } else {
+    f = &(*it);
+  }
+
+  if (block)
+    f->blocked_schemes |= flag;
+  else
+    f->allowed_schemes |= flag;
+}
+
+void Blacklist::Block(const std::string& filter) {
+  AddFilter(filter, true);
+}
+
+void Blacklist::Allow(const std::string& filter) {
+  AddFilter(filter, false);
+}
+
+bool Blacklist::IsURLBlocked(const GURL& url) const {
+  SchemeFlag flag;
+  if (!SchemeToFlag(url.scheme(), &flag)) {
+    // Not a scheme that can be filtered.
+    return false;
+  }
+
+  std::string host(url.host());
+  int int_port = url.EffectiveIntPort();
+  const uint16 port = int_port > 0 ? int_port : 0;
+  const std::string& path = url.path();
+
+  // The first iteration through the loop will be an exact host match.
+  // Subsequent iterations are subdomain matches, and some filters don't apply
+  // to those.
+  bool is_matching_subdomains = false;
+  const bool host_is_ip = url.HostIsIPAddress();
+  for (;;) {
+    HostFilterTable::const_iterator host_filter = host_filters_.find(host);
+    if (host_filter != host_filters_.end()) {
+      const PathFilterList& v = host_filter->second;
+      size_t longest_length = 0;
+      bool is_blocked = false;
+      bool has_match = false;
+      bool has_exact_host_match = false;
+      for (PathFilterList::const_iterator it = v.begin(); it != v.end(); ++it) {
+        // Filters that apply to an exact hostname only take precedence over
+        // filters that can apply to subdomains too.
+        // E.g. ".google.com" filters take priority over "google.com".
+        if (has_exact_host_match && it->match_subdomains)
+          continue;
+
+        // Skip if filter doesn't apply to subdomains, and this is a subdomain.
+        if (is_matching_subdomains && !it->match_subdomains)
+          continue;
+
+        if (it->port != 0 && it->port != port)
+          continue;
+
+        // If this match can't be longer than the current match, skip it.
+        // For same size matches, the first rule to match takes precedence.
+        // If this is an exact host match, it can be actually shorter than
+        // a previous, non-exact match.
+        if ((has_match && it->path_prefix.length() <= longest_length) &&
+            (has_exact_host_match || it->match_subdomains)) {
+          continue;
+        }
+
+        // Skip if the filter's |path_prefix| is not a prefix of |path|.
+        if (path.compare(0, it->path_prefix.length(), it->path_prefix) != 0)
+          continue;
+
+        // Check if there is a blocked or allowed bit set for the scheme.
+        if ((it->allowed_schemes & flag) || (it->blocked_schemes & flag)) {
+          // This is the best match so far.
+          has_match = true;
+          has_exact_host_match = !it->match_subdomains;
+          longest_length = it->path_prefix.length();
+          // If both blocked and allowed bits are set, allowed takes precedence.
+          is_blocked = !(it->allowed_schemes & flag);
+        }
+      }
+      // If a match was found, return its decision.
+      if (has_match)
+        return is_blocked;
+    }

[Mattias Nissler (ping if slow), 2011/08/31 14:58:09]

newline

[Joao da Silva, 2011/09/01 12:47:36]

Done.

+    // Quit after trying the empty string (corresponding to host '*').
+    // Also skip subdomain matching for IP addresses.
+    if (host.empty() || host_is_ip)
+      break;

[Mattias Nissler (ping if slow), 2011/08/31 14:58:09]

newline

[Joao da Silva, 2011/09/01 12:47:36]

Done.

+    // No match found for this host. Try a subdomain match, by removing the
+    // leftmost subdomain from the hostname.
+    is_matching_subdomains = true;
+    size_t i = host.find('.');
+    if (i != std::string::npos)
+      ++i;
+    host.erase(0, i);
+  }
+
+  // Default is to allow.
+  return false;
+}
+
+// static
+bool Blacklist::SchemeToFlag(const std::string& scheme, SchemeFlag* flag) {
+  if (scheme.empty())
+    *flag = SCHEME_ALL;
+  else if (scheme == "http")
+    *flag = SCHEME_HTTP;
+  else if (scheme == "https")
+    *flag = SCHEME_HTTPS;
+  else if (scheme == "ftp")
+    *flag = SCHEME_FTP;
+  else
+    return false;
+  return true;
+}
+
+// static
+bool Blacklist::FilterToComponents(const std::string& filter,
+                                   std::string* scheme,
+                                   std::string* host,
+                                   uint16* port,
+                                   std::string* path) {
+  url_parse::Parsed parsed;
+  URLFixerUpper::SegmentURL(filter, &parsed);
+
+  if (!parsed.host.is_nonempty())
+    return false;
+
+  if (parsed.scheme.is_nonempty())
+    scheme->assign(filter, parsed.scheme.begin, parsed.scheme.len);
+  else
+    scheme->clear();
+
+  host->assign(filter, parsed.host.begin, parsed.host.len);
+  // Special '*' host, matches all hosts.
+  if (*host == "*")
+    host->clear();
+
+  if (parsed.port.is_nonempty()) {
+    int int_port;
+    if (!base::StringToInt(filter.substr(parsed.port.begin, parsed.port.len),
+                           &int_port)) {
+      return false;
+    }
+    if (int_port <= 0 || int_port > kuint16max)
+      return false;
+    *port = int_port;
+  } else {
+    // Match any port.
+    *port = 0;
+  }
+
+  if (parsed.path.is_nonempty())
+    path->assign(filter, parsed.path.begin, parsed.path.len);
+  else
+    path->clear();
+
+  return true;
+}
+
+namespace {

[Mattias Nissler (ping if slow), 2011/08/31 14:58:09]

move anonymous namespace to top of file.

[Joao da Silva, 2011/09/01 12:47:36]

Done.

+
+StringVector* ListValueToStringVector(const base::ListValue* list) {
+  StringVector* vector = new StringVector;
+
+  if (!list)
+    return vector;
+
+  vector->reserve(list->GetSize());
+  std::string s;
+  for (base::ListValue::const_iterator it = list->begin();
+       it != list->end() && vector->size() < kMaxFiltersPerPolicy; ++it) {
+    if ((*it)->GetAsString(&s))
+      vector->push_back(s);
+  }
+
+  return vector;
+}
+
+// A task that owns the Blacklist, and passes it to the URLBlacklistManager
+// on the IO thread, if the URLBlacklistManager still exists.
+class SetBlacklistTask : public Task {
+ public:
+  SetBlacklistTask(
+      base::WeakPtr<URLBlacklistManager> url_blacklist_manager,
+      Blacklist* blacklist)
+      : url_blacklist_manager_(url_blacklist_manager),
+        blacklist_(blacklist) {
+  }
+
+  virtual void Run() OVERRIDE {
+    DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+    if (url_blacklist_manager_)
+      url_blacklist_manager_->SetBlacklist(blacklist_.release());
+  }
+
+ private:
+  base::WeakPtr<URLBlacklistManager> url_blacklist_manager_;
+  scoped_ptr<Blacklist> blacklist_;
+  DISALLOW_COPY_AND_ASSIGN(SetBlacklistTask);
+};
+
+// A task that builds the blacklist on the FILE thread, and then posts another
+// task to pass it to the URLBlacklistManager on the IO thread.
+class BuildBlacklistTask : public Task {
+ public:
+  BuildBlacklistTask(
+      base::WeakPtr<URLBlacklistManager> url_blacklist_manager,
+      StringVector* blacklist,
+      StringVector* whitelist)
+      : url_blacklist_manager_(url_blacklist_manager),
+        blacklist_(blacklist),
+        whitelist_(whitelist) {
+  }
+
+  virtual void Run() OVERRIDE {
+    DCHECK(BrowserThread::CurrentlyOn(BrowserThread::FILE));
+
+    Blacklist* blacklist = NULL;
+
+    if (!blacklist_->empty()) {
+      blacklist = new Blacklist;
+      for (StringVector::iterator it = blacklist_->begin();
+           it != blacklist_->end(); ++it) {
+        blacklist->Block(*it);
+      }
+      for (StringVector::iterator it = whitelist_->begin();
+           it != whitelist_->end(); ++it) {
+        blacklist->Allow(*it);
+      }
+    }
+
+    BrowserThread::PostTask(
+        BrowserThread::IO, FROM_HERE,
+        new SetBlacklistTask(url_blacklist_manager_, blacklist));
+  }
+
+ private:
+  base::WeakPtr<URLBlacklistManager> url_blacklist_manager_;
+  scoped_ptr<StringVector> blacklist_;
+  scoped_ptr<StringVector> whitelist_;
+  DISALLOW_COPY_AND_ASSIGN(BuildBlacklistTask);
+};
+
+}  // namespace
+
+URLBlacklistManager::URLBlacklistManager(Profile* profile)
+    : ALLOW_THIS_IN_INITIALIZER_LIST(ui_method_factory_(this)),
+      ALLOW_THIS_IN_INITIALIZER_LIST(io_weak_ptr_factory_(this)),
+      pref_service_(profile->GetPrefs()) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  // This has to be created on the UI thread, but is only posted later, after
+  // the initialization on the IO thread is complete.
+  // Posting a task to invoke InitializeOnIOThread from here isn't safe, since
+  // we can't get weak ptrs for the IO thread yet.
+  initialize_on_ui_task_ = ui_method_factory_.NewRunnableMethod(
+      &URLBlacklistManager::InitializeOnUIThread);
+}
+
+void URLBlacklistManager::InitializeOnIOThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  DCHECK(initialize_on_ui_task_ != NULL);
+  io_weak_ptr_factory_.DetachFromThread();
+  weak_ptr_ = io_weak_ptr_factory_.GetWeakPtr();
+
+  // It is now safe to resume initialization on the UI thread, since it is now
+  // possible to get weak refs to self to use on the IO thread.
+  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE, initialize_on_ui_task_);
+}
+
+void URLBlacklistManager::InitializeOnUIThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  DCHECK(initialize_on_ui_task_ != NULL);
+  initialize_on_ui_task_ = NULL;
+
+  pref_change_registrar_.Init(pref_service_);
+  pref_change_registrar_.Add(prefs::kUrlBlacklist, this);
+  pref_change_registrar_.Add(prefs::kUrlWhitelist, this);
+
+  // Start enforcing the policies without a delay when they are present at
+  // startup.
+  if (pref_service_->HasPrefPath(prefs::kUrlBlacklist))
+    Update();
+}
+
+void URLBlacklistManager::ShutdownOnUIThread() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  // Cancel any pending updates, and stop listening for pref change updates.
+  // This also cancels a pending InitializeOnUIThread, if it hasn't executed
+  // yet.
+  ui_method_factory_.RevokeAll();
+  pref_change_registrar_.RemoveAll();
+}
+
+URLBlacklistManager::~URLBlacklistManager() {
+  // Cancel any weak ptrs on the IO thread.
+  io_weak_ptr_factory_.InvalidateWeakPtrs();
+}
+
+void URLBlacklistManager::Observe(int type,
+                                  const NotificationSource& source,
+                                  const NotificationDetails& details) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  DCHECK(type == chrome::NOTIFICATION_PREF_CHANGED);
+  PrefService* prefs = Source<PrefService>(source).ptr();
+  DCHECK(prefs == pref_service_);
+  std::string* pref_name = Details<std::string>(details).ptr();
+  DCHECK(*pref_name == prefs::kUrlBlacklist ||
+         *pref_name == prefs::kUrlWhitelist);
+  ScheduleUpdate();
+}
+
+void URLBlacklistManager::ScheduleUpdate() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  // Cancel pending updates, if any.
+  ui_method_factory_.RevokeAll();
+  PostUpdateTask(
+      ui_method_factory_.NewRunnableMethod(&URLBlacklistManager::Update));
+}
+
+void URLBlacklistManager::PostUpdateTask(Task* task) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+  // This is overriden in tests to post the task without the delay.
+  MessageLoop::current()->PostDelayedTask(FROM_HERE, task, kUpdateDelayMs);
+}
+
+void URLBlacklistManager::Update() {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
+
+  // The preferences can only be read on the UI thread.
+  StringVector* blacklist = ListValueToStringVector(
+      pref_service_->GetList(prefs::kUrlBlacklist));
+  StringVector* whitelist = ListValueToStringVector(
+      pref_service_->GetList(prefs::kUrlWhitelist));
+
+  BrowserThread::PostTask(
+      BrowserThread::FILE, FROM_HERE,
+      new BuildBlacklistTask(io_weak_ptr_factory_.GetWeakPtr(),
+                             blacklist, whitelist));
+}
+
+void URLBlacklistManager::SetBlacklist(Blacklist* blacklist) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  blacklist_.reset(blacklist);
+}
+
+bool URLBlacklistManager::IsURLBlocked(const GURL& url) const {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+
+  if (!blacklist_.get())
+    return false;
+
+  return blacklist_->IsURLBlocked(url);
+}
+
+// static
+void URLBlacklistManager::RegisterPrefs(PrefService* pref_service) {
+  pref_service->RegisterListPref(prefs::kUrlBlacklist,
+                                 PrefService::UNSYNCABLE_PREF);
+  pref_service->RegisterListPref(prefs::kUrlWhitelist,
+                                 PrefService::UNSYNCABLE_PREF);
+}
+
+}  // namespace policy