The Android Open Source Project | 9066cfe | 2009-03-03 19:31:44 -0800 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2008 The Android Open Source Project |
| 3 | * |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | * you may not use this file except in compliance with the License. |
| 6 | * You may obtain a copy of the License at |
| 7 | * |
| 8 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | * |
| 10 | * Unless required by applicable law or agreed to in writing, software |
| 11 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | * See the License for the specific language governing permissions and |
| 14 | * limitations under the License. |
| 15 | */ |
| 16 | |
| 17 | package com.android.server; |
| 18 | |
| 19 | import com.android.server.am.ActivityManagerService; |
| 20 | |
| 21 | import android.app.AlarmManager; |
| 22 | import android.app.PendingIntent; |
| 23 | import android.content.BroadcastReceiver; |
| 24 | import android.content.ContentResolver; |
| 25 | import android.content.Context; |
| 26 | import android.content.Intent; |
| 27 | import android.content.IntentFilter; |
| 28 | import android.os.Debug; |
| 29 | import android.os.Handler; |
| 30 | import android.os.Message; |
| 31 | import android.os.Process; |
| 32 | import android.os.SystemClock; |
| 33 | import android.os.SystemProperties; |
| 34 | import android.provider.Settings; |
| 35 | import android.util.Config; |
| 36 | import android.util.EventLog; |
| 37 | import android.util.Log; |
| 38 | |
| 39 | import java.io.IOException; |
| 40 | import java.util.ArrayList; |
| 41 | import java.util.Calendar; |
| 42 | |
| 43 | /** This class calls its monitor every minute. Killing this process if they don't return **/ |
| 44 | public class Watchdog extends Thread { |
| 45 | static final String TAG = "Watchdog"; |
| 46 | static final boolean localLOGV = false || Config.LOGV; |
| 47 | |
| 48 | // Set this to true to use debug default values. |
| 49 | static final boolean DB = false; |
| 50 | |
| 51 | static final int MONITOR = 2718; |
| 52 | static final int GLOBAL_PSS = 2719; |
| 53 | |
| 54 | static final int TIME_TO_WAIT = DB ? 15*1000 : 60*1000; |
| 55 | static final int EVENT_LOG_TAG = 2802; |
| 56 | static final int EVENT_LOG_PROC_PSS_TAG = 2803; |
| 57 | static final int EVENT_LOG_SOFT_RESET_TAG = 2804; |
| 58 | static final int EVENT_LOG_HARD_RESET_TAG = 2805; |
| 59 | static final int EVENT_LOG_PSS_STATS_TAG = 2806; |
| 60 | static final int EVENT_LOG_PROC_STATS_TAG = 2807; |
| 61 | static final int EVENT_LOG_SCHEDULED_REBOOT_TAG = 2808; |
| 62 | static final int EVENT_LOG_MEMINFO_TAG = 2809; |
| 63 | static final int EVENT_LOG_VMSTAT_TAG = 2810; |
| 64 | static final int EVENT_LOG_REQUESTED_REBOOT_TAG = 2811; |
| 65 | |
| 66 | static final int MEMCHECK_DEFAULT_INTERVAL = DB ? 30 : 30*60; // 30 minutes |
| 67 | static final int MEMCHECK_DEFAULT_LOG_REALTIME_INTERVAL = DB ? 60 : 2*60*60; // 2 hours |
| 68 | static final int MEMCHECK_DEFAULT_SYSTEM_SOFT_THRESHOLD = (DB ? 10:16)*1024*1024; // 16MB |
| 69 | static final int MEMCHECK_DEFAULT_SYSTEM_HARD_THRESHOLD = (DB ? 14:20)*1024*1024; // 20MB |
| 70 | static final int MEMCHECK_DEFAULT_PHONE_SOFT_THRESHOLD = (DB ? 4:8)*1024*1024; // 8MB |
| 71 | static final int MEMCHECK_DEFAULT_PHONE_HARD_THRESHOLD = (DB ? 8:12)*1024*1024; // 12MB |
| 72 | |
| 73 | static final int MEMCHECK_DEFAULT_EXEC_START_TIME = 1*60*60; // 1:00am |
| 74 | static final int MEMCHECK_DEFAULT_EXEC_END_TIME = 5*60*60; // 5:00am |
| 75 | static final int MEMCHECK_DEFAULT_MIN_SCREEN_OFF = DB ? 1*60 : 5*60; // 5 minutes |
| 76 | static final int MEMCHECK_DEFAULT_MIN_ALARM = DB ? 1*60 : 3*60; // 3 minutes |
| 77 | static final int MEMCHECK_DEFAULT_RECHECK_INTERVAL = DB ? 1*60 : 5*60; // 5 minutes |
| 78 | |
| 79 | static final int REBOOT_DEFAULT_INTERVAL = DB ? 1 : 0; // never force reboot |
| 80 | static final int REBOOT_DEFAULT_START_TIME = 3*60*60; // 3:00am |
| 81 | static final int REBOOT_DEFAULT_WINDOW = 60*60; // within 1 hour |
| 82 | |
| 83 | static final String CHECKUP_ACTION = "com.android.service.Watchdog.CHECKUP"; |
| 84 | static final String REBOOT_ACTION = "com.android.service.Watchdog.REBOOT"; |
| 85 | |
| 86 | static Watchdog sWatchdog; |
| 87 | |
| 88 | /* This handler will be used to post message back onto the main thread */ |
| 89 | final Handler mHandler; |
| 90 | final Runnable mGlobalPssCollected; |
| 91 | final ArrayList<Monitor> mMonitors = new ArrayList<Monitor>(); |
| 92 | ContentResolver mResolver; |
| 93 | BatteryService mBattery; |
| 94 | PowerManagerService mPower; |
| 95 | AlarmManagerService mAlarm; |
| 96 | ActivityManagerService mActivity; |
| 97 | boolean mCompleted; |
| 98 | boolean mForceKillSystem; |
| 99 | Monitor mCurrentMonitor; |
| 100 | |
| 101 | PssRequestor mPhoneReq; |
| 102 | int mPhonePid; |
| 103 | int mPhonePss; |
| 104 | |
| 105 | long mLastMemCheckTime = -(MEMCHECK_DEFAULT_INTERVAL*1000); |
| 106 | boolean mHavePss; |
| 107 | long mLastMemCheckRealtime = -(MEMCHECK_DEFAULT_LOG_REALTIME_INTERVAL*1000); |
| 108 | boolean mHaveGlobalPss; |
| 109 | final MemMonitor mSystemMemMonitor = new MemMonitor("system", |
| 110 | Settings.Gservices.MEMCHECK_SYSTEM_ENABLED, |
| 111 | Settings.Gservices.MEMCHECK_SYSTEM_SOFT_THRESHOLD, |
| 112 | MEMCHECK_DEFAULT_SYSTEM_SOFT_THRESHOLD, |
| 113 | Settings.Gservices.MEMCHECK_SYSTEM_HARD_THRESHOLD, |
| 114 | MEMCHECK_DEFAULT_SYSTEM_HARD_THRESHOLD); |
| 115 | final MemMonitor mPhoneMemMonitor = new MemMonitor("com.android.phone", |
| 116 | Settings.Gservices.MEMCHECK_PHONE_ENABLED, |
| 117 | Settings.Gservices.MEMCHECK_PHONE_SOFT_THRESHOLD, |
| 118 | MEMCHECK_DEFAULT_PHONE_SOFT_THRESHOLD, |
| 119 | Settings.Gservices.MEMCHECK_PHONE_HARD_THRESHOLD, |
| 120 | MEMCHECK_DEFAULT_PHONE_HARD_THRESHOLD); |
| 121 | |
| 122 | final Calendar mCalendar = Calendar.getInstance(); |
| 123 | long mMemcheckLastTime; |
| 124 | long mMemcheckExecStartTime; |
| 125 | long mMemcheckExecEndTime; |
| 126 | int mMinScreenOff = MEMCHECK_DEFAULT_MIN_SCREEN_OFF; |
| 127 | int mMinAlarm = MEMCHECK_DEFAULT_MIN_ALARM; |
| 128 | boolean mNeedScheduledCheck; |
| 129 | PendingIntent mCheckupIntent; |
| 130 | PendingIntent mRebootIntent; |
| 131 | |
| 132 | long mBootTime; |
| 133 | int mRebootInterval; |
| 134 | |
| 135 | boolean mReqRebootNoWait; // should wait for one interval before reboot? |
| 136 | int mReqRebootInterval = -1; // >= 0 if a reboot has been requested |
| 137 | int mReqRebootStartTime = -1; // >= 0 if a specific start time has been requested |
| 138 | int mReqRebootWindow = -1; // >= 0 if a specific window has been requested |
| 139 | int mReqMinScreenOff = -1; // >= 0 if a specific screen off time has been requested |
| 140 | int mReqMinNextAlarm = -1; // >= 0 if specific time to next alarm has been requested |
| 141 | int mReqRecheckInterval= -1; // >= 0 if a specific recheck interval has been requested |
| 142 | |
| 143 | /** |
| 144 | * This class monitors the memory in a particular process. |
| 145 | */ |
| 146 | final class MemMonitor { |
| 147 | final String mProcessName; |
| 148 | final String mEnabledSetting; |
| 149 | final String mSoftSetting; |
| 150 | final String mHardSetting; |
| 151 | |
| 152 | int mSoftThreshold; |
| 153 | int mHardThreshold; |
| 154 | boolean mEnabled; |
| 155 | long mLastPss; |
| 156 | |
| 157 | static final int STATE_OK = 0; |
| 158 | static final int STATE_SOFT = 1; |
| 159 | static final int STATE_HARD = 2; |
| 160 | int mState; |
| 161 | |
| 162 | MemMonitor(String processName, String enabledSetting, |
| 163 | String softSetting, int defSoftThreshold, |
| 164 | String hardSetting, int defHardThreshold) { |
| 165 | mProcessName = processName; |
| 166 | mEnabledSetting = enabledSetting; |
| 167 | mSoftSetting = softSetting; |
| 168 | mHardSetting = hardSetting; |
| 169 | mSoftThreshold = defSoftThreshold; |
| 170 | mHardThreshold = defHardThreshold; |
| 171 | } |
| 172 | |
| 173 | void retrieveSettings(ContentResolver resolver) { |
| 174 | mSoftThreshold = Settings.Gservices.getInt( |
| 175 | resolver, mSoftSetting, mSoftThreshold); |
| 176 | mHardThreshold = Settings.Gservices.getInt( |
| 177 | resolver, mHardSetting, mHardThreshold); |
| 178 | mEnabled = Settings.Gservices.getInt( |
| 179 | resolver, mEnabledSetting, 0) != 0; |
| 180 | } |
| 181 | |
| 182 | boolean checkLocked(long curTime, int pid, int pss) { |
| 183 | mLastPss = pss; |
| 184 | if (mLastPss < mSoftThreshold) { |
| 185 | mState = STATE_OK; |
| 186 | } else if (mLastPss < mHardThreshold) { |
| 187 | mState = STATE_SOFT; |
| 188 | } else { |
| 189 | mState = STATE_HARD; |
| 190 | } |
| 191 | EventLog.writeEvent(EVENT_LOG_PROC_PSS_TAG, mProcessName, pid, mLastPss); |
| 192 | |
| 193 | if (mState == STATE_OK) { |
| 194 | // Memory is good, don't recover. |
| 195 | return false; |
| 196 | } |
| 197 | |
| 198 | if (mState == STATE_HARD) { |
| 199 | // Memory is really bad, kill right now. |
| 200 | EventLog.writeEvent(EVENT_LOG_HARD_RESET_TAG, mProcessName, pid, |
| 201 | mHardThreshold, mLastPss); |
| 202 | return mEnabled; |
| 203 | } |
| 204 | |
| 205 | // It is time to schedule a reset... |
| 206 | // Check if we are currently within the time to kill processes due |
| 207 | // to memory use. |
| 208 | computeMemcheckTimesLocked(curTime); |
| 209 | String skipReason = null; |
| 210 | if (curTime < mMemcheckExecStartTime || curTime > mMemcheckExecEndTime) { |
| 211 | skipReason = "time"; |
| 212 | } else { |
| 213 | skipReason = shouldWeBeBrutalLocked(curTime); |
| 214 | } |
| 215 | EventLog.writeEvent(EVENT_LOG_SOFT_RESET_TAG, mProcessName, pid, |
| 216 | mSoftThreshold, mLastPss, skipReason != null ? skipReason : ""); |
| 217 | if (skipReason != null) { |
| 218 | mNeedScheduledCheck = true; |
| 219 | return false; |
| 220 | } |
| 221 | return mEnabled; |
| 222 | } |
| 223 | |
| 224 | void clear() { |
| 225 | mLastPss = 0; |
| 226 | mState = STATE_OK; |
| 227 | } |
| 228 | } |
| 229 | |
| 230 | /** |
| 231 | * Used for scheduling monitor callbacks and checking memory usage. |
| 232 | */ |
| 233 | final class HeartbeatHandler extends Handler { |
| 234 | @Override |
| 235 | public void handleMessage(Message msg) { |
| 236 | switch (msg.what) { |
| 237 | case GLOBAL_PSS: { |
| 238 | if (mHaveGlobalPss) { |
| 239 | // During the last pass we collected pss information, so |
| 240 | // now it is time to report it. |
| 241 | mHaveGlobalPss = false; |
| 242 | if (localLOGV) Log.v(TAG, "Received global pss, logging."); |
| 243 | logGlobalMemory(); |
| 244 | } |
| 245 | } break; |
| 246 | |
| 247 | case MONITOR: { |
| 248 | if (mHavePss) { |
| 249 | // During the last pass we collected pss information, so |
| 250 | // now it is time to report it. |
| 251 | mHavePss = false; |
| 252 | if (localLOGV) Log.v(TAG, "Have pss, checking memory."); |
| 253 | checkMemory(); |
| 254 | } |
| 255 | |
| 256 | if (mHaveGlobalPss) { |
| 257 | // During the last pass we collected pss information, so |
| 258 | // now it is time to report it. |
| 259 | mHaveGlobalPss = false; |
| 260 | if (localLOGV) Log.v(TAG, "Have global pss, logging."); |
| 261 | logGlobalMemory(); |
| 262 | } |
| 263 | |
| 264 | long now = SystemClock.uptimeMillis(); |
| 265 | |
| 266 | // See if we should force a reboot. |
| 267 | int rebootInterval = mReqRebootInterval >= 0 |
| 268 | ? mReqRebootInterval : Settings.Gservices.getInt( |
| 269 | mResolver, Settings.Gservices.REBOOT_INTERVAL, |
| 270 | REBOOT_DEFAULT_INTERVAL); |
| 271 | if (mRebootInterval != rebootInterval) { |
| 272 | mRebootInterval = rebootInterval; |
| 273 | // We have been running long enough that a reboot can |
| 274 | // be considered... |
| 275 | checkReboot(false); |
| 276 | } |
| 277 | |
| 278 | // See if we should check memory conditions. |
| 279 | long memCheckInterval = Settings.Gservices.getLong( |
| 280 | mResolver, Settings.Gservices.MEMCHECK_INTERVAL, |
| 281 | MEMCHECK_DEFAULT_INTERVAL) * 1000; |
| 282 | if ((mLastMemCheckTime+memCheckInterval) < now) { |
| 283 | // It is now time to collect pss information. This |
| 284 | // is async so we won't report it now. And to keep |
| 285 | // things simple, we will assume that everyone has |
| 286 | // reported back by the next MONITOR message. |
| 287 | mLastMemCheckTime = now; |
| 288 | if (localLOGV) Log.v(TAG, "Collecting memory usage."); |
| 289 | collectMemory(); |
| 290 | mHavePss = true; |
| 291 | |
| 292 | long memCheckRealtimeInterval = Settings.Gservices.getLong( |
| 293 | mResolver, Settings.Gservices.MEMCHECK_LOG_REALTIME_INTERVAL, |
| 294 | MEMCHECK_DEFAULT_LOG_REALTIME_INTERVAL) * 1000; |
| 295 | long realtimeNow = SystemClock.elapsedRealtime(); |
| 296 | if ((mLastMemCheckRealtime+memCheckRealtimeInterval) < realtimeNow) { |
| 297 | mLastMemCheckRealtime = realtimeNow; |
| 298 | if (localLOGV) Log.v(TAG, "Collecting global memory usage."); |
| 299 | collectGlobalMemory(); |
| 300 | mHaveGlobalPss = true; |
| 301 | } |
| 302 | } |
| 303 | |
| 304 | final int size = mMonitors.size(); |
| 305 | for (int i = 0 ; i < size ; i++) { |
| 306 | mCurrentMonitor = mMonitors.get(i); |
| 307 | mCurrentMonitor.monitor(); |
| 308 | } |
| 309 | |
| 310 | synchronized (Watchdog.this) { |
| 311 | mCompleted = true; |
| 312 | mCurrentMonitor = null; |
| 313 | } |
| 314 | } break; |
| 315 | } |
| 316 | } |
| 317 | } |
| 318 | |
| 319 | final class GlobalPssCollected implements Runnable { |
| 320 | public void run() { |
| 321 | mHandler.sendEmptyMessage(GLOBAL_PSS); |
| 322 | } |
| 323 | } |
| 324 | |
| 325 | final class CheckupReceiver extends BroadcastReceiver { |
| 326 | @Override |
| 327 | public void onReceive(Context c, Intent intent) { |
| 328 | if (localLOGV) Log.v(TAG, "Alarm went off, checking memory."); |
| 329 | checkMemory(); |
| 330 | } |
| 331 | } |
| 332 | |
| 333 | final class RebootReceiver extends BroadcastReceiver { |
| 334 | @Override |
| 335 | public void onReceive(Context c, Intent intent) { |
| 336 | if (localLOGV) Log.v(TAG, "Alarm went off, checking reboot."); |
| 337 | checkReboot(true); |
| 338 | } |
| 339 | } |
| 340 | |
| 341 | final class RebootRequestReceiver extends BroadcastReceiver { |
| 342 | @Override |
| 343 | public void onReceive(Context c, Intent intent) { |
| 344 | mReqRebootNoWait = intent.getIntExtra("nowait", 0) != 0; |
| 345 | mReqRebootInterval = intent.getIntExtra("interval", -1); |
| 346 | mReqRebootStartTime = intent.getIntExtra("startTime", -1); |
| 347 | mReqRebootWindow = intent.getIntExtra("window", -1); |
| 348 | mReqMinScreenOff = intent.getIntExtra("minScreenOff", -1); |
| 349 | mReqMinNextAlarm = intent.getIntExtra("minNextAlarm", -1); |
| 350 | mReqRecheckInterval = intent.getIntExtra("recheckInterval", -1); |
| 351 | EventLog.writeEvent(EVENT_LOG_REQUESTED_REBOOT_TAG, |
| 352 | mReqRebootNoWait ? 1 : 0, mReqRebootInterval, |
| 353 | mReqRecheckInterval, mReqRebootStartTime, |
| 354 | mReqRebootWindow, mReqMinScreenOff, mReqMinNextAlarm); |
| 355 | checkReboot(true); |
| 356 | } |
| 357 | } |
| 358 | |
| 359 | public interface Monitor { |
| 360 | void monitor(); |
| 361 | } |
| 362 | |
| 363 | public interface PssRequestor { |
| 364 | void requestPss(); |
| 365 | } |
| 366 | |
| 367 | public class PssStats { |
| 368 | public int mEmptyPss; |
| 369 | public int mEmptyCount; |
| 370 | public int mBackgroundPss; |
| 371 | public int mBackgroundCount; |
| 372 | public int mServicePss; |
| 373 | public int mServiceCount; |
| 374 | public int mVisiblePss; |
| 375 | public int mVisibleCount; |
| 376 | public int mForegroundPss; |
| 377 | public int mForegroundCount; |
| 378 | |
| 379 | public int mNoPssCount; |
| 380 | |
| 381 | public int mProcDeaths[] = new int[10]; |
| 382 | } |
| 383 | |
| 384 | public static Watchdog getInstance() { |
| 385 | if (sWatchdog == null) { |
| 386 | sWatchdog = new Watchdog(); |
| 387 | } |
| 388 | |
| 389 | return sWatchdog; |
| 390 | } |
| 391 | |
| 392 | private Watchdog() { |
| 393 | super("watchdog"); |
| 394 | mHandler = new HeartbeatHandler(); |
| 395 | mGlobalPssCollected = new GlobalPssCollected(); |
| 396 | } |
| 397 | |
| 398 | public void init(Context context, BatteryService battery, |
| 399 | PowerManagerService power, AlarmManagerService alarm, |
| 400 | ActivityManagerService activity) { |
| 401 | mResolver = context.getContentResolver(); |
| 402 | mBattery = battery; |
| 403 | mPower = power; |
| 404 | mAlarm = alarm; |
| 405 | mActivity = activity; |
| 406 | |
| 407 | context.registerReceiver(new CheckupReceiver(), |
| 408 | new IntentFilter(CHECKUP_ACTION)); |
| 409 | mCheckupIntent = PendingIntent.getBroadcast(context, |
| 410 | 0, new Intent(CHECKUP_ACTION), 0); |
| 411 | |
| 412 | context.registerReceiver(new RebootReceiver(), |
| 413 | new IntentFilter(REBOOT_ACTION)); |
| 414 | mRebootIntent = PendingIntent.getBroadcast(context, |
| 415 | 0, new Intent(REBOOT_ACTION), 0); |
| 416 | |
| 417 | context.registerReceiver(new RebootRequestReceiver(), |
| 418 | new IntentFilter(Intent.ACTION_REBOOT), |
| 419 | android.Manifest.permission.REBOOT, null); |
| 420 | |
| 421 | mBootTime = System.currentTimeMillis(); |
| 422 | } |
| 423 | |
| 424 | public void processStarted(PssRequestor req, String name, int pid) { |
| 425 | synchronized (this) { |
| 426 | if ("com.android.phone".equals(name)) { |
| 427 | mPhoneReq = req; |
| 428 | mPhonePid = pid; |
| 429 | mPhonePss = 0; |
| 430 | } |
| 431 | } |
| 432 | } |
| 433 | |
| 434 | public void reportPss(PssRequestor req, String name, int pss) { |
| 435 | synchronized (this) { |
| 436 | if (mPhoneReq == req) { |
| 437 | mPhonePss = pss; |
| 438 | } |
| 439 | } |
| 440 | } |
| 441 | |
| 442 | public void addMonitor(Monitor monitor) { |
| 443 | synchronized (this) { |
| 444 | if (isAlive()) { |
| 445 | throw new RuntimeException("Monitors can't be added while the Watchdog is running"); |
| 446 | } |
| 447 | mMonitors.add(monitor); |
| 448 | } |
| 449 | } |
| 450 | |
| 451 | /** |
| 452 | * Retrieve memory usage information from specific processes being |
| 453 | * monitored. This is an async operation, so must be done before doing |
| 454 | * memory checks. |
| 455 | */ |
| 456 | void collectMemory() { |
| 457 | synchronized (this) { |
| 458 | if (mPhoneReq != null) { |
| 459 | mPhoneReq.requestPss(); |
| 460 | } |
| 461 | } |
| 462 | } |
| 463 | |
| 464 | /** |
| 465 | * Retrieve memory usage over all application processes. This is an |
| 466 | * async operation, so must be done before doing memory checks. |
| 467 | */ |
| 468 | void collectGlobalMemory() { |
| 469 | mActivity.requestPss(mGlobalPssCollected); |
| 470 | } |
| 471 | |
| 472 | /** |
| 473 | * Check memory usage in the system, scheduling kills/reboots as needed. |
| 474 | * This always runs on the mHandler thread. |
| 475 | */ |
| 476 | void checkMemory() { |
| 477 | boolean needScheduledCheck; |
| 478 | long curTime; |
| 479 | long nextTime = 0; |
| 480 | |
| 481 | long recheckInterval = Settings.Gservices.getLong( |
| 482 | mResolver, Settings.Gservices.MEMCHECK_RECHECK_INTERVAL, |
| 483 | MEMCHECK_DEFAULT_RECHECK_INTERVAL) * 1000; |
| 484 | |
| 485 | mSystemMemMonitor.retrieveSettings(mResolver); |
| 486 | mPhoneMemMonitor.retrieveSettings(mResolver); |
| 487 | retrieveBrutalityAmount(); |
| 488 | |
| 489 | synchronized (this) { |
| 490 | curTime = System.currentTimeMillis(); |
| 491 | mNeedScheduledCheck = false; |
| 492 | |
| 493 | // How is the system doing? |
| 494 | if (mSystemMemMonitor.checkLocked(curTime, Process.myPid(), |
| 495 | (int)Process.getPss(Process.myPid()))) { |
| 496 | // Not good! Time to suicide. |
| 497 | mForceKillSystem = true; |
| 498 | notifyAll(); |
| 499 | return; |
| 500 | } |
| 501 | |
| 502 | // How is the phone process doing? |
| 503 | if (mPhoneReq != null) { |
| 504 | if (mPhoneMemMonitor.checkLocked(curTime, mPhonePid, |
| 505 | mPhonePss)) { |
| 506 | // Just kill the phone process and let it restart. |
Dianne Hackborn | 723738c | 2009-06-25 19:48:04 -0700 | [diff] [blame] | 507 | Log.i(TAG, "Watchdog is killing the phone process"); |
The Android Open Source Project | 9066cfe | 2009-03-03 19:31:44 -0800 | [diff] [blame] | 508 | Process.killProcess(mPhonePid); |
| 509 | } |
| 510 | } else { |
| 511 | mPhoneMemMonitor.clear(); |
| 512 | } |
| 513 | |
| 514 | needScheduledCheck = mNeedScheduledCheck; |
| 515 | if (needScheduledCheck) { |
| 516 | // Something is going bad, but now is not a good time to |
| 517 | // tear things down... schedule an alarm to check again soon. |
| 518 | nextTime = curTime + recheckInterval; |
| 519 | if (nextTime < mMemcheckExecStartTime) { |
| 520 | nextTime = mMemcheckExecStartTime; |
| 521 | } else if (nextTime >= mMemcheckExecEndTime){ |
| 522 | // Need to check during next exec time... so that needs |
| 523 | // to be computed. |
| 524 | if (localLOGV) Log.v(TAG, "Computing next time range"); |
| 525 | computeMemcheckTimesLocked(nextTime); |
| 526 | nextTime = mMemcheckExecStartTime; |
| 527 | } |
| 528 | |
| 529 | if (localLOGV) { |
| 530 | mCalendar.setTimeInMillis(nextTime); |
| 531 | Log.v(TAG, "Next Alarm Time: " + mCalendar); |
| 532 | } |
| 533 | } |
| 534 | } |
| 535 | |
| 536 | if (needScheduledCheck) { |
| 537 | if (localLOGV) Log.v(TAG, "Scheduling next memcheck alarm for " |
| 538 | + ((nextTime-curTime)/1000/60) + "m from now"); |
| 539 | mAlarm.remove(mCheckupIntent); |
| 540 | mAlarm.set(AlarmManager.RTC_WAKEUP, nextTime, mCheckupIntent); |
| 541 | } else { |
| 542 | if (localLOGV) Log.v(TAG, "No need to schedule a memcheck alarm!"); |
| 543 | mAlarm.remove(mCheckupIntent); |
| 544 | } |
| 545 | } |
| 546 | |
| 547 | final PssStats mPssStats = new PssStats(); |
| 548 | final String[] mMemInfoFields = new String[] { |
| 549 | "MemFree:", "Buffers:", "Cached:", |
| 550 | "Active:", "Inactive:", |
| 551 | "AnonPages:", "Mapped:", "Slab:", |
| 552 | "SReclaimable:", "SUnreclaim:", "PageTables:" }; |
| 553 | final long[] mMemInfoSizes = new long[mMemInfoFields.length]; |
| 554 | final String[] mVMStatFields = new String[] { |
| 555 | "pgfree ", "pgactivate ", "pgdeactivate ", |
| 556 | "pgfault ", "pgmajfault " }; |
| 557 | final long[] mVMStatSizes = new long[mVMStatFields.length]; |
| 558 | final long[] mPrevVMStatSizes = new long[mVMStatFields.length]; |
| 559 | long mLastLogGlobalMemoryTime; |
| 560 | |
| 561 | void logGlobalMemory() { |
| 562 | PssStats stats = mPssStats; |
| 563 | mActivity.collectPss(stats); |
| 564 | EventLog.writeEvent(EVENT_LOG_PSS_STATS_TAG, |
| 565 | stats.mEmptyPss, stats.mEmptyCount, |
| 566 | stats.mBackgroundPss, stats.mBackgroundCount, |
| 567 | stats.mServicePss, stats.mServiceCount, |
| 568 | stats.mVisiblePss, stats.mVisibleCount, |
| 569 | stats.mForegroundPss, stats.mForegroundCount, |
| 570 | stats.mNoPssCount); |
| 571 | EventLog.writeEvent(EVENT_LOG_PROC_STATS_TAG, |
| 572 | stats.mProcDeaths[0], stats.mProcDeaths[1], stats.mProcDeaths[2], |
| 573 | stats.mProcDeaths[3], stats.mProcDeaths[4]); |
| 574 | Process.readProcLines("/proc/meminfo", mMemInfoFields, mMemInfoSizes); |
| 575 | for (int i=0; i<mMemInfoSizes.length; i++) { |
| 576 | mMemInfoSizes[i] *= 1024; |
| 577 | } |
| 578 | EventLog.writeEvent(EVENT_LOG_MEMINFO_TAG, |
| 579 | (int)mMemInfoSizes[0], (int)mMemInfoSizes[1], (int)mMemInfoSizes[2], |
| 580 | (int)mMemInfoSizes[3], (int)mMemInfoSizes[4], |
| 581 | (int)mMemInfoSizes[5], (int)mMemInfoSizes[6], (int)mMemInfoSizes[7], |
| 582 | (int)mMemInfoSizes[8], (int)mMemInfoSizes[9], (int)mMemInfoSizes[10]); |
| 583 | long now = SystemClock.uptimeMillis(); |
| 584 | long dur = now - mLastLogGlobalMemoryTime; |
| 585 | mLastLogGlobalMemoryTime = now; |
| 586 | Process.readProcLines("/proc/vmstat", mVMStatFields, mVMStatSizes); |
| 587 | for (int i=0; i<mVMStatSizes.length; i++) { |
| 588 | long v = mVMStatSizes[i]; |
| 589 | mVMStatSizes[i] -= mPrevVMStatSizes[i]; |
| 590 | mPrevVMStatSizes[i] = v; |
| 591 | } |
| 592 | EventLog.writeEvent(EVENT_LOG_VMSTAT_TAG, dur, |
| 593 | (int)mVMStatSizes[0], (int)mVMStatSizes[1], (int)mVMStatSizes[2], |
| 594 | (int)mVMStatSizes[3], (int)mVMStatSizes[4]); |
| 595 | } |
| 596 | |
| 597 | void checkReboot(boolean fromAlarm) { |
| 598 | int rebootInterval = mReqRebootInterval >= 0 ? mReqRebootInterval |
| 599 | : Settings.Gservices.getInt( |
| 600 | mResolver, Settings.Gservices.REBOOT_INTERVAL, |
| 601 | REBOOT_DEFAULT_INTERVAL); |
| 602 | mRebootInterval = rebootInterval; |
| 603 | if (rebootInterval <= 0) { |
| 604 | // No reboot interval requested. |
| 605 | if (localLOGV) Log.v(TAG, "No need to schedule a reboot alarm!"); |
| 606 | mAlarm.remove(mRebootIntent); |
| 607 | return; |
| 608 | } |
| 609 | |
| 610 | long rebootStartTime = mReqRebootStartTime >= 0 ? mReqRebootStartTime |
| 611 | : Settings.Gservices.getLong( |
| 612 | mResolver, Settings.Gservices.REBOOT_START_TIME, |
| 613 | REBOOT_DEFAULT_START_TIME); |
| 614 | long rebootWindowMillis = (mReqRebootWindow >= 0 ? mReqRebootWindow |
| 615 | : Settings.Gservices.getLong( |
| 616 | mResolver, Settings.Gservices.REBOOT_WINDOW, |
| 617 | REBOOT_DEFAULT_WINDOW)) * 1000; |
| 618 | long recheckInterval = (mReqRecheckInterval >= 0 ? mReqRecheckInterval |
| 619 | : Settings.Gservices.getLong( |
| 620 | mResolver, Settings.Gservices.MEMCHECK_RECHECK_INTERVAL, |
| 621 | MEMCHECK_DEFAULT_RECHECK_INTERVAL)) * 1000; |
| 622 | |
| 623 | retrieveBrutalityAmount(); |
| 624 | |
| 625 | long realStartTime; |
| 626 | long now; |
| 627 | |
| 628 | synchronized (this) { |
| 629 | now = System.currentTimeMillis(); |
| 630 | realStartTime = computeCalendarTime(mCalendar, now, |
| 631 | rebootStartTime); |
| 632 | |
| 633 | long rebootIntervalMillis = rebootInterval*24*60*60*1000; |
| 634 | if (DB || mReqRebootNoWait || |
| 635 | (now-mBootTime) >= (rebootIntervalMillis-rebootWindowMillis)) { |
| 636 | if (fromAlarm && rebootWindowMillis <= 0) { |
| 637 | // No reboot window -- just immediately reboot. |
| 638 | EventLog.writeEvent(EVENT_LOG_SCHEDULED_REBOOT_TAG, now, |
| 639 | (int)rebootIntervalMillis, (int)rebootStartTime*1000, |
| 640 | (int)rebootWindowMillis, ""); |
| 641 | rebootSystem("Checkin scheduled forced"); |
| 642 | return; |
| 643 | } |
| 644 | |
| 645 | // Are we within the reboot window? |
| 646 | if (now < realStartTime) { |
| 647 | // Schedule alarm for next check interval. |
| 648 | realStartTime = computeCalendarTime(mCalendar, |
| 649 | now, rebootStartTime); |
| 650 | } else if (now < (realStartTime+rebootWindowMillis)) { |
| 651 | String doit = shouldWeBeBrutalLocked(now); |
| 652 | EventLog.writeEvent(EVENT_LOG_SCHEDULED_REBOOT_TAG, now, |
| 653 | (int)rebootInterval, (int)rebootStartTime*1000, |
| 654 | (int)rebootWindowMillis, doit != null ? doit : ""); |
| 655 | if (doit == null) { |
| 656 | rebootSystem("Checked scheduled range"); |
| 657 | return; |
| 658 | } |
| 659 | |
| 660 | // Schedule next alarm either within the window or in the |
| 661 | // next interval. |
| 662 | if ((now+recheckInterval) >= (realStartTime+rebootWindowMillis)) { |
| 663 | realStartTime = computeCalendarTime(mCalendar, |
| 664 | now + rebootIntervalMillis, rebootStartTime); |
| 665 | } else { |
| 666 | realStartTime = now + recheckInterval; |
| 667 | } |
| 668 | } else { |
| 669 | // Schedule alarm for next check interval. |
| 670 | realStartTime = computeCalendarTime(mCalendar, |
| 671 | now + rebootIntervalMillis, rebootStartTime); |
| 672 | } |
| 673 | } |
| 674 | } |
| 675 | |
| 676 | if (localLOGV) Log.v(TAG, "Scheduling next reboot alarm for " |
| 677 | + ((realStartTime-now)/1000/60) + "m from now"); |
| 678 | mAlarm.remove(mRebootIntent); |
| 679 | mAlarm.set(AlarmManager.RTC_WAKEUP, realStartTime, mRebootIntent); |
| 680 | } |
| 681 | |
| 682 | /** |
| 683 | * Perform a full reboot of the system. |
| 684 | */ |
| 685 | void rebootSystem(String reason) { |
| 686 | Log.i(TAG, "Rebooting system because: " + reason); |
| 687 | try { |
| 688 | android.os.Power.reboot(reason); |
| 689 | } catch (IOException e) { |
| 690 | Log.e(TAG, "Reboot failed!", e); |
| 691 | } |
| 692 | } |
| 693 | |
| 694 | /** |
| 695 | * Load the current Gservices settings for when |
| 696 | * {@link #shouldWeBeBrutalLocked} will allow the brutality to happen. |
| 697 | * Must not be called with the lock held. |
| 698 | */ |
| 699 | void retrieveBrutalityAmount() { |
| 700 | mMinScreenOff = (mReqMinScreenOff >= 0 ? mReqMinScreenOff |
| 701 | : Settings.Gservices.getInt( |
| 702 | mResolver, Settings.Gservices.MEMCHECK_MIN_SCREEN_OFF, |
| 703 | MEMCHECK_DEFAULT_MIN_SCREEN_OFF)) * 1000; |
| 704 | mMinAlarm = (mReqMinNextAlarm >= 0 ? mReqMinNextAlarm |
| 705 | : Settings.Gservices.getInt( |
| 706 | mResolver, Settings.Gservices.MEMCHECK_MIN_ALARM, |
| 707 | MEMCHECK_DEFAULT_MIN_ALARM)) * 1000; |
| 708 | } |
| 709 | |
| 710 | /** |
| 711 | * Determine whether it is a good time to kill, crash, or otherwise |
| 712 | * plunder the current situation for the overall long-term benefit of |
| 713 | * the world. |
| 714 | * |
| 715 | * @param curTime The current system time. |
| 716 | * @return Returns null if this is a good time, else a String with the |
| 717 | * text of why it is not a good time. |
| 718 | */ |
| 719 | String shouldWeBeBrutalLocked(long curTime) { |
| 720 | if (mBattery == null || !mBattery.isPowered()) { |
| 721 | return "battery"; |
| 722 | } |
| 723 | |
| 724 | if (mMinScreenOff >= 0 && (mPower == null || |
| 725 | mPower.timeSinceScreenOn() < mMinScreenOff)) { |
| 726 | return "screen"; |
| 727 | } |
| 728 | |
| 729 | if (mMinAlarm >= 0 && (mAlarm == null || |
| 730 | mAlarm.timeToNextAlarm() < mMinAlarm)) { |
| 731 | return "alarm"; |
| 732 | } |
| 733 | |
| 734 | return null; |
| 735 | } |
| 736 | |
| 737 | /** |
| 738 | * Compute the times during which we next would like to perform process |
| 739 | * restarts. |
| 740 | * |
| 741 | * @param curTime The current system time. |
| 742 | */ |
| 743 | void computeMemcheckTimesLocked(long curTime) { |
| 744 | if (mMemcheckLastTime == curTime) { |
| 745 | return; |
| 746 | } |
| 747 | |
| 748 | mMemcheckLastTime = curTime; |
| 749 | |
| 750 | long memcheckExecStartTime = Settings.Gservices.getLong( |
| 751 | mResolver, Settings.Gservices.MEMCHECK_EXEC_START_TIME, |
| 752 | MEMCHECK_DEFAULT_EXEC_START_TIME); |
| 753 | long memcheckExecEndTime = Settings.Gservices.getLong( |
| 754 | mResolver, Settings.Gservices.MEMCHECK_EXEC_END_TIME, |
| 755 | MEMCHECK_DEFAULT_EXEC_END_TIME); |
| 756 | |
| 757 | mMemcheckExecEndTime = computeCalendarTime(mCalendar, curTime, |
| 758 | memcheckExecEndTime); |
| 759 | if (mMemcheckExecEndTime < curTime) { |
| 760 | memcheckExecStartTime += 24*60*60; |
| 761 | memcheckExecEndTime += 24*60*60; |
| 762 | mMemcheckExecEndTime = computeCalendarTime(mCalendar, curTime, |
| 763 | memcheckExecEndTime); |
| 764 | } |
| 765 | mMemcheckExecStartTime = computeCalendarTime(mCalendar, curTime, |
| 766 | memcheckExecStartTime); |
| 767 | |
| 768 | if (localLOGV) { |
| 769 | mCalendar.setTimeInMillis(curTime); |
| 770 | Log.v(TAG, "Current Time: " + mCalendar); |
| 771 | mCalendar.setTimeInMillis(mMemcheckExecStartTime); |
| 772 | Log.v(TAG, "Start Check Time: " + mCalendar); |
| 773 | mCalendar.setTimeInMillis(mMemcheckExecEndTime); |
| 774 | Log.v(TAG, "End Check Time: " + mCalendar); |
| 775 | } |
| 776 | } |
| 777 | |
| 778 | static long computeCalendarTime(Calendar c, long curTime, |
| 779 | long secondsSinceMidnight) { |
| 780 | |
| 781 | // start with now |
| 782 | c.setTimeInMillis(curTime); |
| 783 | |
| 784 | int val = (int)secondsSinceMidnight / (60*60); |
| 785 | c.set(Calendar.HOUR_OF_DAY, val); |
| 786 | secondsSinceMidnight -= val * (60*60); |
| 787 | val = (int)secondsSinceMidnight / 60; |
| 788 | c.set(Calendar.MINUTE, val); |
| 789 | c.set(Calendar.SECOND, (int)secondsSinceMidnight - (val*60)); |
| 790 | c.set(Calendar.MILLISECOND, 0); |
| 791 | |
| 792 | long newTime = c.getTimeInMillis(); |
| 793 | if (newTime < curTime) { |
| 794 | // The given time (in seconds since midnight) has already passed for today, so advance |
| 795 | // by one day (due to daylight savings, etc., the delta may differ from 24 hours). |
| 796 | c.add(Calendar.DAY_OF_MONTH, 1); |
| 797 | newTime = c.getTimeInMillis(); |
| 798 | } |
| 799 | |
| 800 | return newTime; |
| 801 | } |
| 802 | |
| 803 | @Override |
| 804 | public void run() { |
| 805 | while (true) { |
| 806 | mCompleted = false; |
| 807 | mHandler.sendEmptyMessage(MONITOR); |
| 808 | |
| 809 | synchronized (this) { |
| 810 | long timeout = TIME_TO_WAIT; |
| 811 | |
| 812 | // NOTE: We use uptimeMillis() here because we do not want to increment the time we |
| 813 | // wait while asleep. If the device is asleep then the thing that we are waiting |
| 814 | // to timeout on is asleep as well and won't have a chance to run. Causing a false |
| 815 | // positive on when to kill things. |
| 816 | long start = SystemClock.uptimeMillis(); |
| 817 | do { |
| 818 | try { |
| 819 | wait(timeout); |
| 820 | } catch (InterruptedException e) { |
| 821 | if (SystemProperties.getBoolean("ro.secure", false)) { |
| 822 | // If this is a secure build, just log the error. |
| 823 | Log.e("WatchDog", "Woof! Woof! Interrupter!"); |
| 824 | } else { |
| 825 | throw new AssertionError("Someone interrupted the watchdog"); |
| 826 | } |
| 827 | } |
| 828 | timeout = TIME_TO_WAIT - (SystemClock.uptimeMillis() - start); |
| 829 | } while (timeout > 0 && !mForceKillSystem); |
| 830 | |
| 831 | if (mCompleted && !mForceKillSystem) { |
| 832 | // The monitors have returned. |
| 833 | continue; |
| 834 | } |
| 835 | } |
| 836 | |
| 837 | // If we got here, that means that the system is most likely hung. |
| 838 | // First send a SIGQUIT so that we can see where it was hung. Then |
| 839 | // kill this process so that the system will restart. |
| 840 | String name = (mCurrentMonitor != null) ? mCurrentMonitor.getClass().getName() : "null"; |
| 841 | EventLog.writeEvent(EVENT_LOG_TAG, name); |
| 842 | Process.sendSignal(Process.myPid(), Process.SIGNAL_QUIT); |
| 843 | |
| 844 | // Wait a bit longer before killing so we can make sure that the stacks are captured. |
| 845 | try { |
| 846 | Thread.sleep(10*1000); |
| 847 | } catch (InterruptedException e) { |
| 848 | } |
| 849 | |
| 850 | // Only kill the process if the debugger is not attached. |
| 851 | if (!Debug.isDebuggerConnected()) { |
Dianne Hackborn | 723738c | 2009-06-25 19:48:04 -0700 | [diff] [blame] | 852 | Log.i(TAG, "Watchdog is killing the system process"); |
The Android Open Source Project | 9066cfe | 2009-03-03 19:31:44 -0800 | [diff] [blame] | 853 | Process.killProcess(Process.myPid()); |
| 854 | } |
| 855 | } |
| 856 | } |
| 857 | } |