Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(389)

Unified Diff: remoting/host/me2me_preference_pane.mm

Issue 10171020: Preference Pane for chromoting on Mac (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/src
Patch Set: Created 8 years, 8 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
Index: remoting/host/me2me_preference_pane.mm
diff --git a/remoting/host/me2me_preference_pane.mm b/remoting/host/me2me_preference_pane.mm
new file mode 100644
index 0000000000000000000000000000000000000000..13f7b27d119c88d0d0b86c1bd2255acb4dec9c9e
--- /dev/null
+++ b/remoting/host/me2me_preference_pane.mm
@@ -0,0 +1,360 @@
+// 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.
+
+#import <Cocoa/Cocoa.h>
+#include <launch.h>
+#import <PreferencePanes/PreferencePanes.h>
+#import <SecurityInterface/SFAuthorizationView.h>
+
+#include "base/eintr_wrapper.h"
+#include "base/file_path.h"
+#include "base/file_util.h"
+#include "base/logging.h"
+#include "base/mac/authorization_util.h"
+#include "base/mac/foundation_util.h"
+#include "base/mac/launchd.h"
+#include "base/mac/mac_logging.h"
+#include "base/mac/scoped_launch_data.h"
+#include "base/memory/scoped_ptr.h"
+#include "base/stringprintf.h"
+#include "base/sys_string_conversions.h"
+#include "remoting/host/host_config.h"
+#include "remoting/host/json_host_config.h"
+#include "remoting/protocol/me2me_host_authenticator_factory.h"
+
+namespace {
+// The name of the Remoting Host service that is registered with launchd.
+#define kServiceName "org.chromium.chromoting"
+#define kConfigDir "/Library/PrivilegedHelperTools/"
+
+// This helper script is executed as root. It is passed a command-line option
+// (--enable or --disable), which causes it to create or remove a trigger file.
+// The trigger file (defined in the service's plist file) informs launchd
+// whether the Host service should be running. Creating the trigger file causes
+// launchd to immediately start the service. Deleting the trigger file has no
+// immediate effect, but it prevents the service from being restarted if it
+// becomes stopped.
+const char kHelperTool[] = kConfigDir kServiceName ".me2me.sh";
garykac 2012/04/27 17:14:09 At some point, it would be nice to have all these
Lambros 2012/04/28 01:54:19 Agreed, though I'd prefer to wait until this PrefP
+
+bool GetTemporaryConfigFilePath(FilePath* path) {
+ if (!file_util::GetTempDir(path))
+ return false;
+ *path = path->Append(kServiceName ".json");
+ return true;
+}
Jamie 2012/04/27 18:21:04 I don't think this is a problem, because we presen
Lambros 2012/04/28 01:54:19 Mac OS temp directories are private per-user, so w
+
+bool IsPinValid(const std::string& pin, const std::string& host_id,
+ const std::string& host_secret_hash) {
+ remoting::protocol::SharedSecretHash hash;
+ if (!hash.Parse(host_secret_hash)) {
+ NSLog(@"Invalid host_secret_hash.");
+ return false;
+ }
+ std::string result =
+ remoting::protocol::AuthenticationMethod::ApplyHashFunction(
+ hash.hash_function, host_id, pin);
+ return result == hash.value;
+}
+
+} // namespace
+
+@interface Me2MePreferencePane : NSPreferencePane {
+ IBOutlet NSTextField* status_message_;
+ IBOutlet NSButton* disable_button_;
+ IBOutlet NSTextField* pin_instruction_message_;
+ IBOutlet NSTextField* email_;
+ IBOutlet NSTextField* pin_;
+ IBOutlet NSButton* apply_button_;
+ IBOutlet SFAuthorizationView* authorization_view_;
Jamie 2012/04/27 18:21:04 This is probably my ObjC ignorance, but why no @sy
dcaiafa 2012/04/27 19:25:00 Because these are instance variables, not properti
+
+ // Holds the new proposed configuration if a temporary config file is
+ // present.
+ scoped_ptr<remoting::JsonHostConfig> config_;
+
+ NSTimer* timer_;
+
+ // Flags to control the UI state. These are computed in the update...Status
+ // methods. The updateUI method updates all the controls according to the
+ // values of these flags and the data stored in config_.
+ BOOL is_service_running_;
+ BOOL is_pane_unlocked_;
+ // True if a new proposed config file has been created.
+ BOOL have_new_config_;
+}
+
+- (void)mainViewDidLoad;
+- (void)willSelect;
+- (void)willUnselect;
+- (IBAction)onDisable:(id)sender;
+- (IBAction)onApply:(id)sender;
+- (void)onNewConfigFile:(NSNotification *)notification;
Jamie 2012/04/27 18:21:04 Nit: No space before *, here and elsewhere.
Lambros 2012/04/28 01:54:19 Done.
+- (void)onTimer:(NSTimer*)timer;
Jamie 2012/04/27 18:21:04 onTimer is a rather generic name. Can you name it
Lambros 2012/04/28 01:54:19 Done.
+- (void)authorizationViewDidAuthorize:(SFAuthorizationView *)view;
+- (void)authorizationViewDidDeauthorize:(SFAuthorizationView *)view;
+- (void)updateServiceStatus;
+- (void)updateAuthorizationStatus;
+- (void)updateConfigStatus;
+- (void)updateUI;
+
+// Save the new config to the system, and either start the service or inform
+// the currently-running service of the new config.
+- (void)applyNewServiceConfig;
+
+- (BOOL)runHelperAsRootWithCommand:(const char*)command
+ InputData:(const std::string&)input_data;
Jamie 2012/04/27 18:21:04 Nit: s/InputData/inputData/
Lambros 2012/04/28 01:54:19 Done.
+@end
+
+@implementation Me2MePreferencePane
+
+- (void)mainViewDidLoad {
+ [authorization_view_ setDelegate:self];
+ [authorization_view_ setString:kAuthorizationRightExecute];
+ [authorization_view_ setAutoupdate:YES];
+}
+
+- (void)willSelect {
+ have_new_config_ = NO;
+
+ NSDistributedNotificationCenter* center =
+ [NSDistributedNotificationCenter defaultCenter];
+ [center addObserver:self
dcaiafa 2012/04/27 17:06:00 You have to release this observer.
Lambros 2012/04/28 01:54:19 Done.
+ selector:@selector(onNewConfigFile:)
+ name:@kServiceName
+ object:nil];
+
+ timer_ = [[NSTimer scheduledTimerWithTimeInterval:2.0
+ target:self
+ selector:@selector(onTimer:)
+ userInfo:nil
+ repeats:YES] retain];
+ [self updateServiceStatus];
+ [self updateAuthorizationStatus];
+ [self updateConfigStatus];
Jamie 2012/04/27 18:21:04 Why do this here? Don't you only expect a new conf
Lambros 2012/04/28 01:54:19 Yes, it's to cope with the not-already-running cas
+ [self updateUI];
+}
+
+- (void)willUnselect {
+ [timer_ invalidate];
+ [timer_ release];
aharper 2012/04/30 21:23:32 nil timer_
Lambros 2012/05/02 19:05:28 Done.
+}
+
+- (void)onApply:(id)sender {
+ if (!have_new_config_) {
+ // It shouldn't be possible to hit the button if there is no config to
+ // apply, but check anyway just in case it happens somehow.
+ return;
+ }
+
+ // Ensure the authorization token is up-to-date before using it.
+ [self updateAuthorizationStatus];
+ [self updateUI];
+
+ std::string pin = base::SysNSStringToUTF8([pin_ stringValue]);
+ std::string host_id, host_secret_hash;
+ if (!config_->GetString(remoting::kHostIdConfigPath, &host_id) ||
+ !config_->GetString(remoting::kHostSecretHashConfigPath,
+ &host_secret_hash)) {
+ LOG(ERROR) << "Failed to get config data for PIN verification";
+ return;
aharper 2012/04/30 21:23:32 Should this be an error alert?
Lambros 2012/05/02 19:05:28 Done.
+ }
+ if (!IsPinValid(pin, host_id, host_secret_hash)) {
aharper 2012/04/30 21:23:32 I'm not seeing how this validation ties the email
Lambros 2012/05/02 19:05:28 Almost. The host registration was done using the
Wez 2012/05/03 00:13:44 Alex's point is that we're displaying the email ad
+ NSAlert* alert = [[NSAlert alloc] init];
+ [alert setMessageText:@"Incorrect PIN entered"];
+ [alert setAlertStyle:NSWarningAlertStyle];
+ [alert runModal];
aharper 2012/04/30 21:23:32 This runs application modal, which is probably ina
Lambros 2012/05/02 19:05:28 Done.
+ [alert release];
+ return;
+ }
+
+ [self applyNewServiceConfig];
+ [self updateUI];
+}
+
+- (void)onDisable:(id)sender {
+ // Ensure the authorization token is up-to-date before using it.
+ [self updateAuthorizationStatus];
+ [self updateUI];
+
+ if (![self runHelperAsRootWithCommand:"--disable"
+ InputData:""]) {
+ NSLog(@"Failed to run the helper tool");
aharper 2012/04/30 21:23:32 Should this be an error alert?
Lambros 2012/05/02 19:05:28 Done.
+ }
+
+ // Stop the launchd job. This cannot easily be done by the helper tool,
+ // since the launchd job runs in the current user's context.
+ base::mac::ScopedLaunchData response(
+ base::mac::MessageForJob(kServiceName, LAUNCH_KEY_STOPJOB));
+ if (!response) {
+ LOG(ERROR) << "Failed to send message to launchd";
Jamie 2012/04/27 18:21:04 Is logging an error sufficient here (and similarly
Lambros 2012/04/28 01:54:19 I expect this particular error would hardly ever h
+ return;
+ }
+
+ // Got a response, so check if launchd sent a non-zero error code, otherwise
+ // assume the command was successful.
Jamie 2012/04/27 18:21:04 The word "assume" makes me nervous. Why do we need
Lambros 2012/04/28 01:54:19 The thing that could go wrong (theoretically) is,
+ if (launch_data_get_type(response.get()) == LAUNCH_DATA_ERRNO) {
+ int error = launch_data_get_errno(response.get());
+ if (error) {
+ LOG(ERROR) << "launchd returned error " << error;
+ }
+ }
+}
+
+- (void)onNewConfigFile:(NSNotification *)notification {
+ [self updateConfigStatus];
+ [self updateUI];
+}
+
+- (void)onTimer:(NSTimer*)timer {
+ BOOL was_running = is_service_running_;
+ [self updateServiceStatus];
+ if (was_running != is_service_running_)
+ [self updateUI];
+}
+
+- (void)authorizationViewDidAuthorize:(SFAuthorizationView *)view {
+ [self updateAuthorizationStatus];
+ [self updateUI];
+}
+
+- (void)authorizationViewDidDeauthorize:(SFAuthorizationView *)view {
+ [self updateAuthorizationStatus];
+ [self updateUI];
+}
+
+- (void)updateServiceStatus {
+ pid_t job_pid = base::mac::PIDForJob(kServiceName);
+ is_service_running_ = (job_pid > 0);
+}
+
+- (void)updateAuthorizationStatus {
+ is_pane_unlocked_ = [authorization_view_ updateStatus:authorization_view_];
+}
+
+- (void)updateConfigStatus {
+ FilePath file;
+ if (!GetTemporaryConfigFilePath(&file)) {
+ NSLog(@"Failed to get path of configuration data.");
Jamie 2012/04/27 18:21:04 You're mixing LOG with NSLog; please stick to one
Lambros 2012/04/28 01:54:19 Done.
+ return;
+ }
+ if (!file_util::PathExists(file)) {
+ return;
Jamie 2012/04/27 18:21:04 No log here? If this is called even when there's n
Lambros 2012/04/28 01:54:19 This gets called whenever the pref-pane is opened.
+ }
+ config_.reset(new remoting::JsonHostConfig(file));
+ if (config_->Read()) {
+ have_new_config_ = YES;
+ file_util::Delete(file, false);
+ } else {
+ NSLog(@"Error reading configuration data from %s", file.value().c_str());
+ have_new_config_ = NO;
Jamie 2012/04/27 18:21:04 This should be set at the top of the function, sin
Lambros 2012/04/28 01:54:19 I've documented the intent of this function, clari
+ }
+}
+
+- (void)updateUI {
+ if (is_service_running_) {
+ [status_message_ setStringValue:@"Me2Me is enabled"];
+ } else {
+ [status_message_ setStringValue:@"Me2Me is disabled"];
+ }
+
+ std::string email;
+ if (config_.get()) {
+ if (!config_->GetString(remoting::kXmppLoginConfigPath, &email)) {
+ NSLog(@"Failed to get config item '%s'", remoting::kXmppLoginConfigPath);
+ }
+ }
+ [email_ setStringValue:base::SysUTF8ToNSString(email)];
+
+ [disable_button_ setEnabled:(is_pane_unlocked_ && is_service_running_)];
+ [pin_instruction_message_ setEnabled:have_new_config_];
+ [email_ setEnabled:have_new_config_];
+ [pin_ setEnabled:have_new_config_];
+ [apply_button_ setEnabled:(is_pane_unlocked_ && have_new_config_)];
+}
+
+- (void)applyNewServiceConfig {
+ [self updateServiceStatus];
+ std::string serialized_config = config_->GetSerializedData();
+ const char* command = is_service_running_ ? "--save-config" : "--enable";
Jamie 2012/04/27 18:21:04 It doesn't need to be part of this CL, but I think
Lambros 2012/04/28 01:54:19 This would be nicer. Unfortunately, I don't think
+ if (![self runHelperAsRootWithCommand:command
+ InputData:serialized_config]) {
+ NSLog(@"Failed to run the helper tool");
+ return;
+ }
+ have_new_config_ = NO;
Jamie 2012/04/27 18:21:04 Shouldn't we clear this earlier? If we get an erro
Lambros 2012/04/28 01:54:19 Well, the user could push the Apply button again :
+
+ // If the service was previously running, send a signal to cause it to reload
+ // its configuration.
Jamie 2012/04/27 18:21:04 This also feels like it would be better done by th
Lambros 2012/04/28 01:54:19 Same problem as before (although if we could someh
+ if (is_service_running_) {
+ pid_t job_pid = base::mac::PIDForJob(kServiceName);
+ if (job_pid > 0) {
+ kill(job_pid, SIGHUP);
+ } else {
+ NSLog(@"Failed to obtain PID of service %s", kServiceName);
+ }
+ }
+}
+
+- (BOOL)runHelperAsRootWithCommand:(const char*)command
+ InputData:(const std::string&)input_data {
Jamie 2012/04/27 18:21:04 Nit: s/InputData/inputData/ and align colons.
Lambros 2012/04/28 01:54:19 Done.
+ if (!file_util::VerifyPathControlledByAdmin(FilePath(kHelperTool))) {
aharper 2012/04/30 21:23:32 This method name made me nervous, because the task
Lambros 2012/05/02 19:05:28 This check was added in response to a comment made
Lambros 2012/05/03 01:47:04 Dropped, and added TODO comment.
+ LOG(ERROR) << "Security check failed for: " << kHelperTool;
+ return NO;
+ }
+
+ AuthorizationRef authorization =
+ [[authorization_view_ authorization] authorizationRef];
+ if (!authorization) {
+ LOG(ERROR) << "Failed to obtain authorizationRef";
+ return NO;
+ }
+
+ const char* arguments[] = { command, NULL };
+ FILE* pipe = NULL;
+ pid_t pid;
+ OSStatus status = base::mac::ExecuteWithPrivilegesAndGetPID(
+ authorization,
+ kHelperTool,
+ kAuthorizationFlagDefaults,
+ arguments,
+ &pipe,
+ &pid);
+ if (status != errAuthorizationSuccess) {
+ OSSTATUS_LOG(ERROR, status) << "AuthorizationExecuteWithPrivileges";
Jamie 2012/04/27 18:21:04 Yet another type of log? Any reason this needs to
Lambros 2012/04/28 01:54:19 I've ditched NSLog, so I think we're OK now.
+ return NO;
+ }
+ if (pid == -1) {
+ LOG(ERROR) << "Failed to get child PID";
+ return NO;
+ }
+
+ DCHECK(pipe);
+ if (!input_data.empty()) {
+ size_t bytes_written = fwrite(input_data.data(), sizeof(char),
+ input_data.size(), pipe);
+ // According to the fwrite manpage, a partial count is returned only if a
+ // write error has occurred.
+ if (bytes_written != input_data.size()) {
+ LOG(ERROR) << "Failed to write data to child process";
Jamie 2012/04/27 18:21:04 Close and return NO?
Lambros 2012/04/28 01:54:19 Sadly, no, because we still want to call waitpid t
+ }
+ // Need to close, since the child waits for EOF on its stdin.
+ if (fclose(pipe) != 0) {
dcaiafa 2012/04/27 17:06:00 Shouldn't you close the pipe even if input_data.em
Jamie 2012/04/27 18:21:04 +1
Lambros 2012/04/28 01:54:19 Ooh, well spotted! That would leak the FILE* if no
+ PLOG(ERROR) << "fclose";
Jamie 2012/04/27 18:21:04 Return NO?
Lambros 2012/04/28 01:54:19 Done.
+ }
+ }
+
+ int exit_status;
+ pid_t wait_result = HANDLE_EINTR(waitpid(pid, &exit_status, 0));
+ if (wait_result != pid) {
+ PLOG(ERROR) << "waitpid";
+ return NO;
+ }
+ if (WIFEXITED(exit_status) && WEXITSTATUS(exit_status) == 0) {
+ return YES;
+ } else {
+ LOG(ERROR) << kHelperTool << " failed with exit status " << exit_status;
+ return NO;
+ }
+}
+
+@end

Powered by Google App Engine
This is Rietveld 408576698