Data stall detection using DNS event (7a70a7e0) · Commits · e / os / android_frameworks_base

core/java/android/net/metrics/NetworkEvent.java

+3 −0

Original line number	Original line	Diff line number	Diff line
	@@ -44,6 +44,8 @@ public final class NetworkEvent implements Parcelable {
	public static final int NETWORK_FIRST_VALIDATION_PORTAL_FOUND = 10;		public static final int NETWORK_FIRST_VALIDATION_PORTAL_FOUND = 10;
	public static final int NETWORK_REVALIDATION_PORTAL_FOUND = 11;		public static final int NETWORK_REVALIDATION_PORTAL_FOUND = 11;

			public static final int NETWORK_CONSECUTIVE_DNS_TIMEOUT_FOUND = 12;

	@IntDef(value = {		@IntDef(value = {
	NETWORK_CONNECTED,		NETWORK_CONNECTED,
	NETWORK_VALIDATED,		NETWORK_VALIDATED,
	@@ -56,6 +58,7 @@ public final class NetworkEvent implements Parcelable {
	NETWORK_REVALIDATION_SUCCESS,		NETWORK_REVALIDATION_SUCCESS,
	NETWORK_FIRST_VALIDATION_PORTAL_FOUND,		NETWORK_FIRST_VALIDATION_PORTAL_FOUND,
	NETWORK_REVALIDATION_PORTAL_FOUND,		NETWORK_REVALIDATION_PORTAL_FOUND,
			NETWORK_CONSECUTIVE_DNS_TIMEOUT_FOUND,
	})		})
	@Retention(RetentionPolicy.SOURCE)		@Retention(RetentionPolicy.SOURCE)
	public @interface EventType {}		public @interface EventType {}

core/java/android/provider/Settings.java

+35 −0

Original line number	Original line	Diff line number	Diff line
	@@ -10435,6 +10435,41 @@ public final class Settings {
	*/		*/
	public static final String CAPTIVE_PORTAL_USER_AGENT = "captive_portal_user_agent";		public static final String CAPTIVE_PORTAL_USER_AGENT = "captive_portal_user_agent";

			/**
			* The threshold value for the number of consecutive dns timeout events received to be a
			* signal of data stall. Set the value to 0 or less than 0 to disable. Note that the value
			* should be larger than 0 if the DNS data stall detection is enabled.
			*
			* @hide
			*/
			public static final String DATA_STALL_CONSECUTIVE_DNS_TIMEOUT_THRESHOLD =
			"data_stall_consecutive_dns_timeout_threshold";

			/**
			* The minimal time interval in milliseconds for data stall reevaluation.
			*
			* @hide
			*/
			public static final String DATA_STALL_MIN_EVALUATE_INTERVAL =
			"data_stall_min_evaluate_interval";

			/**
			* DNS timeouts older than this timeout (in milliseconds) are not considered for detecting
			* a data stall.
			*
			* @hide
			*/
			public static final String DATA_STALL_VALID_DNS_TIME_THRESHOLD =
			"data_stall_valid_dns_time_threshold";

			/**
			* Which data stall detection signal to use. Possible values are a union of the powers of 2
			* of DATA_STALL_EVALUATION_TYPE_*.
			*
			* @hide
			*/
			public static final String DATA_STALL_EVALUATION_TYPE = "data_stall_evaluation_type";

	/**		/**
	* Whether network service discovery is enabled.		* Whether network service discovery is enabled.
	*		*

core/tests/coretests/src/android/provider/SettingsBackupTest.java

+4 −0

Original line number	Original line	Diff line number	Diff line
	@@ -185,6 +185,10 @@ public class SettingsBackupTest {
	Settings.Global.DATA_ROAMING,		Settings.Global.DATA_ROAMING,
	Settings.Global.DATA_STALL_ALARM_AGGRESSIVE_DELAY_IN_MS,		Settings.Global.DATA_STALL_ALARM_AGGRESSIVE_DELAY_IN_MS,
	Settings.Global.DATA_STALL_ALARM_NON_AGGRESSIVE_DELAY_IN_MS,		Settings.Global.DATA_STALL_ALARM_NON_AGGRESSIVE_DELAY_IN_MS,
			Settings.Global.DATA_STALL_CONSECUTIVE_DNS_TIMEOUT_THRESHOLD,
			Settings.Global.DATA_STALL_EVALUATION_TYPE,
			Settings.Global.DATA_STALL_MIN_EVALUATE_INTERVAL,
			Settings.Global.DATA_STALL_VALID_DNS_TIME_THRESHOLD,
	Settings.Global.DEBUG_APP,		Settings.Global.DEBUG_APP,
	Settings.Global.DEBUG_VIEW_ATTRIBUTES,		Settings.Global.DEBUG_VIEW_ATTRIBUTES,
	Settings.Global.DEFAULT_DNS_SERVER,		Settings.Global.DEFAULT_DNS_SERVER,

services/core/java/com/android/server/ConnectivityService.java

+18 −0

Original line number	Original line	Diff line number	Diff line
	@@ -1659,6 +1659,24 @@ public class ConnectivityService extends IConnectivityManager.Stub
	loge("Error parsing ip address in validation event");		loge("Error parsing ip address in validation event");
	}		}
	}		}

			@Override
			public void onDnsEvent(int netId, int eventType, int returnCode, String hostname,
			String[] ipAddresses, int ipAddressesCount, long timestamp, int uid) {
			NetworkAgentInfo nai = getNetworkAgentInfoForNetId(netId);
			// Netd event only allow registrants from system. Each NetworkMonitor thread is under
			// the caller thread of registerNetworkAgent. Thus, it's not allowed to register netd
			// event callback for certain nai. e.g. cellular. Register here to pass to
			// NetworkMonitor instead.
			// TODO: Move the Dns Event to NetworkMonitor. Use Binder.clearCallingIdentity() in
			// registerNetworkAgent to have NetworkMonitor created with system process as design
			// expectation. Also, NetdEventListenerService only allow one callback from each
			// caller type. Need to re-factor NetdEventListenerService to allow multiple
			// NetworkMonitor registrants.
			if (nai != null && nai.satisfies(mDefaultRequest)) {
			nai.networkMonitor.sendMessage(NetworkMonitor.EVENT_DNS_NOTIFICATION, returnCode);
			}
			}
	};		};

	@VisibleForTesting		@VisibleForTesting

services/core/java/com/android/server/connectivity/NetworkMonitor.java

+187 −1

Original line number	Original line	Diff line number	Diff line
	@@ -72,6 +72,7 @@ import android.util.Log;
	import com.android.internal.annotations.VisibleForTesting;		import com.android.internal.annotations.VisibleForTesting;
	import com.android.internal.util.ArrayUtils;		import com.android.internal.util.ArrayUtils;
	import com.android.internal.util.Protocol;		import com.android.internal.util.Protocol;
			import com.android.internal.util.RingBufferIndices;
	import com.android.internal.util.State;		import com.android.internal.util.State;
	import com.android.internal.util.StateMachine;		import com.android.internal.util.StateMachine;
	import com.android.server.connectivity.DnsManager.PrivateDnsConfig;		import com.android.server.connectivity.DnsManager.PrivateDnsConfig;
	@@ -99,7 +100,7 @@ public class NetworkMonitor extends StateMachine {
	private static final String TAG = NetworkMonitor.class.getSimpleName();		private static final String TAG = NetworkMonitor.class.getSimpleName();
	private static final boolean DBG = true;		private static final boolean DBG = true;
	private static final boolean VDBG = false;		private static final boolean VDBG = false;
			private static final boolean VDBG_STALL = Log.isLoggable(TAG, Log.DEBUG);
	// Default configuration values for captive portal detection probes.		// Default configuration values for captive portal detection probes.
	// TODO: append a random length parameter to the default HTTPS url.		// TODO: append a random length parameter to the default HTTPS url.
	// TODO: randomize browser version ids in the default User-Agent String.		// TODO: randomize browser version ids in the default User-Agent String.
	@@ -116,6 +117,15 @@ public class NetworkMonitor extends StateMachine {
	private static final int SOCKET_TIMEOUT_MS = 10000;		private static final int SOCKET_TIMEOUT_MS = 10000;
	private static final int PROBE_TIMEOUT_MS = 3000;		private static final int PROBE_TIMEOUT_MS = 3000;

			// Default configuration values for data stall detection.
			private static final int DEFAULT_CONSECUTIVE_DNS_TIMEOUT_THRESHOLD = 5;
			private static final int DEFAULT_DATA_STALL_MIN_EVALUATE_TIME_MS = 60 * 1000;
			private static final int DEFAULT_DATA_STALL_VALID_DNS_TIME_THRESHOLD_MS = 30 * 60 * 1000;

			private static final int DATA_STALL_EVALUATION_TYPE_DNS = 1;
			private static final int DEFAULT_DATA_STALL_EVALUATION_TYPES =
			(1 << DATA_STALL_EVALUATION_TYPE_DNS);

	static enum EvaluationResult {		static enum EvaluationResult {
	VALIDATED(true),		VALIDATED(true),
	CAPTIVE_PORTAL(false);		CAPTIVE_PORTAL(false);
	@@ -233,6 +243,12 @@ public class NetworkMonitor extends StateMachine {
	*/		*/
	public static final int CMD_PROBE_COMPLETE = BASE + 16;		public static final int CMD_PROBE_COMPLETE = BASE + 16;

			/**
			* ConnectivityService notifies NetworkMonitor of DNS query responses event.
			* arg1 = returncode in OnDnsEvent which indicates the response code for the DNS query.
			*/
			public static final int EVENT_DNS_NOTIFICATION = BASE + 17;

	// Start mReevaluateDelayMs at this value and double.		// Start mReevaluateDelayMs at this value and double.
	private static final int INITIAL_REEVALUATE_DELAY_MS = 1000;		private static final int INITIAL_REEVALUATE_DELAY_MS = 1000;
	private static final int MAX_REEVALUATE_DELAY_MS = 10601000;		private static final int MAX_REEVALUATE_DELAY_MS = 10601000;
	@@ -314,6 +330,12 @@ public class NetworkMonitor extends StateMachine {
	private int mReevaluateDelayMs = INITIAL_REEVALUATE_DELAY_MS;		private int mReevaluateDelayMs = INITIAL_REEVALUATE_DELAY_MS;
	private int mEvaluateAttempts = 0;		private int mEvaluateAttempts = 0;
	private volatile int mProbeToken = 0;		private volatile int mProbeToken = 0;
			private final int mConsecutiveDnsTimeoutThreshold;
			private final int mDataStallMinEvaluateTime;
			private final int mDataStallValidDnsTimeThreshold;
			private final int mDataStallEvaluationType;
			private final DnsStallDetector mDnsStallDetector;
			private long mLastProbeTime;

	public NetworkMonitor(Context context, Handler handler, NetworkAgentInfo networkAgentInfo,		public NetworkMonitor(Context context, Handler handler, NetworkAgentInfo networkAgentInfo,
	NetworkRequest defaultRequest) {		NetworkRequest defaultRequest) {
	@@ -359,6 +381,12 @@ public class NetworkMonitor extends StateMachine {
	mCaptivePortalFallbackUrls = makeCaptivePortalFallbackUrls();		mCaptivePortalFallbackUrls = makeCaptivePortalFallbackUrls();
	mCaptivePortalFallbackSpecs = makeCaptivePortalFallbackProbeSpecs();		mCaptivePortalFallbackSpecs = makeCaptivePortalFallbackProbeSpecs();
	mRandom = deps.getRandom();		mRandom = deps.getRandom();
			// TODO: Evaluate to move data stall configuration to a specific class.
			mConsecutiveDnsTimeoutThreshold = getConsecutiveDnsTimeoutThreshold();
			mDnsStallDetector = new DnsStallDetector(mConsecutiveDnsTimeoutThreshold);
			mDataStallMinEvaluateTime = getDataStallMinEvaluateTime();
			mDataStallValidDnsTimeThreshold = getDataStallValidDnsTimeThreshold();
			mDataStallEvaluationType = getDataStallEvalutionType();

	start();		start();
	}		}
	@@ -507,6 +535,9 @@ public class NetworkMonitor extends StateMachine {
	sendMessage(CMD_EVALUATE_PRIVATE_DNS);		sendMessage(CMD_EVALUATE_PRIVATE_DNS);
	break;		break;
	}		}
			case EVENT_DNS_NOTIFICATION:
			mDnsStallDetector.accumulateConsecutiveDnsTimeoutCount(message.arg1);
			break;
	default:		default:
	break;		break;
	}		}
	@@ -537,6 +568,13 @@ public class NetworkMonitor extends StateMachine {
	case CMD_EVALUATE_PRIVATE_DNS:		case CMD_EVALUATE_PRIVATE_DNS:
	transitionTo(mEvaluatingPrivateDnsState);		transitionTo(mEvaluatingPrivateDnsState);
	break;		break;
			case EVENT_DNS_NOTIFICATION:
			mDnsStallDetector.accumulateConsecutiveDnsTimeoutCount(message.arg1);
			if (isDataStall()) {
			validationLog("Suspecting data stall, reevaluate");
			transitionTo(mEvaluatingState);
			}
			break;
	default:		default:
	return NOT_HANDLED;		return NOT_HANDLED;
	}		}
	@@ -856,6 +894,7 @@ public class NetworkMonitor extends StateMachine {

	final CaptivePortalProbeResult probeResult =		final CaptivePortalProbeResult probeResult =
	(CaptivePortalProbeResult) message.obj;		(CaptivePortalProbeResult) message.obj;
			mLastProbeTime = SystemClock.elapsedRealtime();
	if (probeResult.isSuccessful()) {		if (probeResult.isSuccessful()) {
	// Transit EvaluatingPrivateDnsState to get to Validated		// Transit EvaluatingPrivateDnsState to get to Validated
	// state (even if no Private DNS validation required).		// state (even if no Private DNS validation required).
	@@ -883,6 +922,7 @@ public class NetworkMonitor extends StateMachine {
	// Leave the event to EvaluatingState. Defer this message will result in reset		// Leave the event to EvaluatingState. Defer this message will result in reset
	// of mReevaluateDelayMs and mEvaluateAttempts.		// of mReevaluateDelayMs and mEvaluateAttempts.
	case CMD_NETWORK_DISCONNECTED:		case CMD_NETWORK_DISCONNECTED:
			case EVENT_DNS_NOTIFICATION:
	return NOT_HANDLED;		return NOT_HANDLED;
	default:		default:
	// TODO: Some events may able to handle in this state, instead of deferring to		// TODO: Some events may able to handle in this state, instead of deferring to
	@@ -947,6 +987,29 @@ public class NetworkMonitor extends StateMachine {
	Settings.Global.CAPTIVE_PORTAL_HTTPS_URL, DEFAULT_HTTPS_URL);		Settings.Global.CAPTIVE_PORTAL_HTTPS_URL, DEFAULT_HTTPS_URL);
	}		}

			private int getConsecutiveDnsTimeoutThreshold() {
			return mDependencies.getSetting(mContext,
			Settings.Global.DATA_STALL_CONSECUTIVE_DNS_TIMEOUT_THRESHOLD,
			DEFAULT_CONSECUTIVE_DNS_TIMEOUT_THRESHOLD);
			}

			private int getDataStallMinEvaluateTime() {
			return mDependencies.getSetting(mContext,
			Settings.Global.DATA_STALL_MIN_EVALUATE_INTERVAL,
			DEFAULT_DATA_STALL_MIN_EVALUATE_TIME_MS);
			}

			private int getDataStallValidDnsTimeThreshold() {
			return mDependencies.getSetting(mContext,
			Settings.Global.DATA_STALL_VALID_DNS_TIME_THRESHOLD,
			DEFAULT_DATA_STALL_VALID_DNS_TIME_THRESHOLD_MS);
			}

			private int getDataStallEvalutionType() {
			return mDependencies.getSetting(mContext, Settings.Global.DATA_STALL_EVALUATION_TYPE,
			DEFAULT_DATA_STALL_EVALUATION_TYPES);
			}

	// Static for direct access by ConnectivityService		// Static for direct access by ConnectivityService
	public static String getCaptivePortalServerHttpUrl(Context context) {		public static String getCaptivePortalServerHttpUrl(Context context) {
	return getCaptivePortalServerHttpUrl(Dependencies.DEFAULT, context);		return getCaptivePortalServerHttpUrl(Dependencies.DEFAULT, context);
	@@ -1462,4 +1525,127 @@ public class NetworkMonitor extends StateMachine {

	public static final Dependencies DEFAULT = new Dependencies();		public static final Dependencies DEFAULT = new Dependencies();
	}		}

			/**
			* Methods in this class perform no locking because all accesses are performed on the state
			* machine's thread. Need to consider the thread safety if it ever could be accessed outside the
			* state machine.
			*/
			@VisibleForTesting
			protected class DnsStallDetector {
			private static final int DEFAULT_DNS_LOG_SIZE = 50;
			private int mConsecutiveTimeoutCount = 0;
			private int mSize;
			final DnsResult[] mDnsEvents;
			final RingBufferIndices mResultIndices;

			DnsStallDetector(int size) {
			mSize = Math.max(DEFAULT_DNS_LOG_SIZE, size);
			mDnsEvents = new DnsResult[mSize];
			mResultIndices = new RingBufferIndices(mSize);
			}

			@VisibleForTesting
			protected void accumulateConsecutiveDnsTimeoutCount(int code) {
			final DnsResult result = new DnsResult(code);
			mDnsEvents[mResultIndices.add()] = result;
			if (result.isTimeout()) {
			mConsecutiveTimeoutCount++;
			} else {
			// Keep the event in mDnsEvents without clearing it so that there are logs to do the
			// simulation and analysis.
			mConsecutiveTimeoutCount = 0;
			}
			}

			private boolean isDataStallSuspected(int timeoutCountThreshold, int validTime) {
			if (timeoutCountThreshold <= 0) {
			Log.wtf(TAG, "Timeout count threshold should be larger than 0.");
			return false;
			}

			// Check if the consecutive timeout count reach the threshold or not.
			if (mConsecutiveTimeoutCount < timeoutCountThreshold) {
			return false;
			}

			// Check if the target dns event index is valid or not.
			final int firstConsecutiveTimeoutIndex =
			mResultIndices.indexOf(mResultIndices.size() - timeoutCountThreshold);

			// If the dns timeout events happened long time ago, the events are meaningless for
			// data stall evaluation. Thus, check if the first consecutive timeout dns event
			// considered in the evaluation happened in defined threshold time.
			final long now = SystemClock.elapsedRealtime();
			final long firstTimeoutTime = now - mDnsEvents[firstConsecutiveTimeoutIndex].mTimeStamp;
			return (firstTimeoutTime < validTime);
			}

			int getConsecutiveTimeoutCount() {
			return mConsecutiveTimeoutCount;
			}
			}

			private static class DnsResult {
			// TODO: Need to move the DNS return code definition to a specific class once unify DNS
			// response code is done.
			private static final int RETURN_CODE_DNS_TIMEOUT = 255;

			private final long mTimeStamp;
			private final int mReturnCode;

			DnsResult(int code) {
			mTimeStamp = SystemClock.elapsedRealtime();
			mReturnCode = code;
			}

			private boolean isTimeout() {
			return mReturnCode == RETURN_CODE_DNS_TIMEOUT;
			}
			}


			@VisibleForTesting
			protected DnsStallDetector getDnsStallDetector() {
			return mDnsStallDetector;
			}

			private boolean dataStallEvaluateTypeEnabled(int type) {
			return (mDataStallEvaluationType & (1 << type)) != 0;
			}

			@VisibleForTesting
			protected long getLastProbeTime() {
			return mLastProbeTime;
			}

			@VisibleForTesting
			protected boolean isDataStall() {
			boolean result = false;
			// Reevaluation will generate traffic. Thus, set a minimal reevaluation timer to limit the
			// possible traffic cost in metered network.
			if (mNetworkAgentInfo.networkCapabilities.isMetered()
			&& (SystemClock.elapsedRealtime() - getLastProbeTime()
			< mDataStallMinEvaluateTime)) {
			return false;
			}

			// Check dns signal. Suspect it may be a data stall if both :
			// 1. The number of consecutive DNS query timeouts > mConsecutiveDnsTimeoutThreshold.
			// 2. Those consecutive DNS queries happened in the last mValidDataStallDnsTimeThreshold ms.
			if (dataStallEvaluateTypeEnabled(DATA_STALL_EVALUATION_TYPE_DNS)) {
			if (mDnsStallDetector.isDataStallSuspected(mConsecutiveDnsTimeoutThreshold,
			mDataStallValidDnsTimeThreshold)) {
			result = true;
			logNetworkEvent(NetworkEvent.NETWORK_CONSECUTIVE_DNS_TIMEOUT_FOUND);
			}
			}

			if (VDBG_STALL) {
			log("isDataStall: result=" + result + ", consecutive dns timeout count="
			+ mDnsStallDetector.getConsecutiveTimeoutCount());
			}

			return result;
			}
	}		}