Added an api which apps can query to find out about their own usage
events. This is useful in understanding the apps historical bucket
changes.
Test: atest android.app.usage.cts.UsageStatsTest#testQueryEventsForSelf
Fixes: 71906213
Change-Id: Ie6144cfe515fd0748a317835918a9d8cf1aea007
UsageStats.getAppStandbyBuckets and setAppStandbyBuckets deal with large
maps of packages to buckets which may exceed the IPC transaction size
limit. Using ParceledListSlice to send and receive data across binder.
Test: atest android.app.usage.cts.UsageStatsTest#testGetAppStandbyBuckets
And manually:
adb shell am get-standby-bucket --user 0
adb shell am set-standby-bucket <package1> active <package2> rare ...
Locally, added protectionLevel development to permission
CHANGE_APP_IDLE_STATE and tested setAppStandbyBuckets with a test app.
Bug: 70890569
Change-Id: Ic1fb0465137d1b830d3ad1fbf3ac8698a409c623
On calling dumpsys usagestats, the last 24 hours of events will be
dumped rather than the events since the arbitrary time point
(also increase timestamp granularity to seconds)
Fixes: 73657997
Test: manual (adb shell dumpsys usagestats)
Change-Id: I92f1f17b4119d2023ed9b81e0610c19f62503bf0
Fixes overlapping of predictions, strong usage events
and mild usage events which have forced durations.
Having separate timeouts for ACTIVE and WORKING_SET
and moving between them when necessary prevents
getting stuck in the wrong state (higher) for longer
than necessary.
Bug: 73294677
Test: atest FrameworksServicesTests:AppStandbyControllerTests
Change-Id: I35530e62cffe2c86945b5da64a41704f807708ce
Expose some of the new events as SystemApi.
Make some of the timeouts configurable in AppStandbyController.
Make NOTIFICATION_SEEN event upgrade app to WORKING_SET for
12 hours. This is not perfect though and will require further
tweaking as it may result in the app becoming elevated higher
for much longer than necessary.
Change-Id: I62401cfabddf51b6f80b9bba8a358285b8cf9a51
Fixes: 72741441
Fixes: 72067231
Fixes: 72537465
Fixes: 72536347
Test: atest CtsUsageStatsTestCases:UsageStatsTests
Add @GuardedBy for simple functions that require a single lock
and are named XYZLocked.
Derived by errorprone.
Bug: 73000847
Test: m
Test: m javac-check-framework RUN_ERROR_PRONE=true
Change-Id: Id1d9fbe5018250e3b9d1466fc5f670d5ad902fa3
Relax the permission requirement for UsageStats
API querying inactive state of apps across users.
Bug: 72953020
Fixes: 72942083
Test: Manual -- create a secondary user
Change-Id: I9c5e12f718d9f424c99aa72dd70470c4b083c13f
After an OTA, all system apps were being
pushed into EXEMPTED for 4 hours because
of a race with the boot phase resulting in
it appearing as if app standby was disabled.
Another bug in updating the state out of EXEMPTED
was preventing checkIdleStates() from fixing
the issue soon after.
Bug: 72835804
Test: Manual:
Manually delete /data/system/usagestats/version
Reboot
Verify that correct apps are EXEMPTED
Automated:
atest FrameworksServicesTests:AppStandbyControllerTests
Change-Id: Ib53f0c45e5c2e2456442f6782ad5ca9b9d0c3d72
Push standby bucket changes into UsageStats as events.
Don't use the event for aggregation.
Bug: 63527785
Test: atest CtsUsageStatsTestCases:UsageStatsTest
Change-Id: Icd152b21935e3ef73aaa1a62f09574e51ce921bd
JobScheduler can store the last job run time
for an app and use it later to help with
throttling.
Also, allow reporting a usage for a temporary
period. This will put the app in ACTIVE for a
minimum specified period during which timeouts
and predictions will not be able to reduce the
bucket level.
Bug: 71536897
Test: atest AppIdleHistoryTests
atest AppStandbyControllerTests
Change-Id: I2985f7f3766671f15ea641972346c69875f17946
This is how we tell CTS if the device has reserved blocks set aside
for system critical services.
Test: builds, boots
Bug: 62024591
Change-Id: Ib970554235b49346f9e9df7d3d1646beb168cd92
Mirrors the design of TimeUnit and ChronoUnit which many developers
are already familiar with, making it easy to pick up and use.
Yes, this is an enum.
Bug: 70915728
Test: bit FrameworksCoreTests:android.util.DataUnitTest
Change-Id: Id0cfdac5c81ed89c3c9ece23c964acba4a4f8471
For JobScheduler, DeviceIdle and AppStandby constants, allow
using a more compact format than milliseconds,
which are a PITA to calculate.
So instead of 18640000000... whatever, you can
use PT2H (for 2 hours), or P2D (for 2 days), etc.
Uses Duration.parse() to do the parsing. See Duration
for format.
Test: adb shell settings put global app_standby_constants
screen_thresholds=0/PT2H/PT12H/P2D
Fixes: 71554131
Change-Id: I5141854ec7df6de266725a67f1f3e2a6e0b4c1c1
DUMP OF SERVICE Usagestats was empty in bugreports
because an unexpected -a argument was being passed.
adb shell pm dump <package> had a similar issue with --packages
Change-Id: Ic67c723b8edc74ec7c9c48582b08896ca3430a4a
Fix: 70777338
Test: adb shell pm dump <package> and bugreports now contain usage stats
Disable app standby features if app_standby_enabled
is set to 0. This allows for UI and experiments to
control the feature.
Bug: 70655630
Test: adb shell settings put global app_standby_enabled 0
adb shell dumpsys usagestats
adb shell am get-standby-bucket <packagename>
Change-Id: Id6c62b078e52040767100f2997832cc586bb0806
Don't allow EXEMPT to be overridden
Differentiate between shell and other callers who set
buckets.
Don't allow forced to be modified by bucketeer.
Don't allow bucketeer to modify NEVER bucket.
Fix a locking issue in listeners list. Don't use
the primary lock when calling out to listeners.
Fixes: 70622791
Fixes: 70622338
Test: atest FrameworksServicesTests:AppIdleHistoryTests
atest FrameworksServicesTests:AppStandbyControllerTests
atest CtsAppUsageHostTestCases
Change-Id: I22309478b947d6461235f5a77e08c7be86309e4b
This is different from SDK_INT, which can change due to OTA
software updates.
Test: cts-tradefed run commandAndExit cts-dev -m CtsOsTestCases -t android.os.cts.BuildTest
Bug: 69270887
Change-Id: Ic3f4b22b25dc08f3fbdbdf7622c953323447ea01
SystemApi to pass a map of packagename to bucket mapping.
Reduces the number of IPC calls to be made by bucketing
ML app.
Bug: 63527785
Test: atest CtsAppUsageHostTestCases
Change-Id: I4172fd30cc310092a1f182ae26267de9148009b7
Ignore predicted bucket if it's been too long since the
last update, and use internal timeout logic to set
standby buckets.
Keep track of last predicted time and persist it.
Bug: 70219058
Test: atest FrameworksServicesTests:AppStandbyControllerTests
Change-Id: Ib2992201d204b92fdd4a0a00754d81e7bbfc292f
Instrumentation package should be immediately marked as active so
that the jobs aren't throttled.
Also, make sure special apps' buckets are correctly initialized.
Fixes: 68900608
Fixes: 69799327
Fixes: 69793822
Fixes: 69788489
Test: atest CtsNetSecPolicyUsesCleartextTrafficTrueTestCases
atest FrameworksServicesTests:AppIdleHistoryTests
atest FrameworksServicesTests:AppStandbyControllerTests
atest CtsNetSecPolicyUsesCleartextTrafficUnspecifiedTestCases
atest CtsNetSecConfigDownloadManagerTestCases
atest CtsNetSecPolicyUsesCleartextTrafficFalseTestCases
cts-tradefed run cts-dev -m CtsAppUsageHostTestCases
cts-tradefed run cts-dev -m CtsHostsideNetworkTests
gts-tradefed run gts-dev -m GtsUsageStatsTestCases
Change-Id: Ic880b6f5ed1fdb7912613e882ccf7b2d4e388159
Only check for INTERACT_ACROSS_USERS in getAppStandbyBucket()
Bug: 70126098
Test: gts-tradefed run gts-dev -m GtsUsageStatsTestCases
Change-Id: I6c3f69ff56d8513f84c7d91162567d17c06980c3
Allow apps to query their own standby bucket.
Allow privileged apps with PACKAGE_USAGE_STATS permission to
query apps.
Allow privileged apps with CHANGE_APP_IDLE_STATE to set the
standby state for apps, but not for themselves.
Removed AppStandby class and moved constants into UsageStatsManager.
Bug: 63527785
Test: cts-tradefed run cts-dev -m CtsAppUsageHostTestCases
Change-Id: I3c1c20f6ecb6d54e248233696039286b243d663c
We need over 8 days of UsageEvents in order to capture weekly
patterns in our app usage models
Change-Id: I8e1149b82cfbaedae4d19bbbfaeffa989bfd54e7
Fix: 69244223
Test: It builds
Switch to using the new bucket thresholds and
remove the previous idle thresholds.
Bug: 63527785
Test: atest AppStandbyControllerTests
atest AppIdleHistoryTests
Change-Id: Ica3fd10fe172faa5b4e712b08273b77357fdae14
Start off with slightly more aggressive timeouts to
switch between buckets. 12Hrs, 24Hrs and 48Hrs.
Bug: 63527785
Test: atest AppStandbyControllerTests
Change-Id: I4aa0cc38ef91538192eda0f48dd3840217aa49c4
On a notification becoming visible, NOTIFICATION_SEEN event is
dispatched. This will only bump up the app's bucket to WORKING_SET and
no higher. Same goes for sync adapters associated with a content
provider. Only move them to WORKING_SET.
Updated NotificationManagerService to report event to usagestats on
visibility changes.
Bug: 63527785
Test: runtest -x
frameworks/base/services/tests/servicestests/src/com/android/server/usage/AppStandbyControllerTests.java
atest NotificationManagerServiceTest
Change-Id: I5b132e1fc1f70e2126473b43b9b1979fbc523b85
And inform listeners when the bucket changes, not just when
going in and out of RARE bucket.
Avoid redundant callbacks when informing listeners.
Bug: 63527785
Test: runtest -x
frameworks/base/services/tests/servicestests/src/com/android/server/usage/AppStandbyControllerTests.java
Change-Id: Icd98d59f597147fbf8ea4bf44edf4b3b3d5c8e14
Everything is now moved out of the pm command except for
the various install commands. I am going to hold of on
those since they require doing some resolution with the
current implementations in the package manager to make
sure they match and behave identically to the implementations
currently in the pm command. But other than that, everything
in pm is now just redirecting over to "cmd package".
Also fix up some of the dumpsys output of a few other sevices
when asking to print the data for a particular package, so
the "pm dump" command gives a little more sane result.
Test: manual
Change-Id: I139e06e560203b72243d7eea9543c2240db0f8f8
Manage the standby bucket in AppStandbyController
Default implementation of bucketing based on simple timeout:
ACTIVE, if recently used
12 hrs to move to WORKING_SET
2 days to move to FREQUENT
7 days to move to RARE
(subject to change)
RARE bucket equates to the old "idle" or "inactive" state for
an app.
Bug: 63527785
Test: AppStandbyControllerTests.java
Change-Id: I970d7afcdf47c31a9413da8fd4852066a13676a2
Move the App Standby code out of UsageStatsService into
an AppStandbyController class. This doesn't change any
behavior. Also, it doesn't change the entry points, which
are still in UsageStatsManager. The binder APIs are still
in UsageStatsService and are forwarded to
AppStandbyController.
This change is in preparation for switching to app standby
buckets instead of a boolean state.
Bug: 63527785
Test: adb shell dumpsys usagestats
adb shell am set-inactive <packagename> true
// Ensure state changes are persisted
Change-Id: Ic94357bdc891d1defae37bb29d533b6153a09967
