Currently when debug layer apk is not installed, GraphicsEnvironment
logs the message and proceeds, which results in null pointer access
crash when later trying to access primaryAbi. Avoid null access by an
early return statement.
Bug: N/A
Test: specify not installed debug layer apk and no crash
Change-Id: I4f3b86b0a8cfff4fc608b3a59ebf0872db0465ff
Prior to this change, we would iterate over _all_ UIDs in
the system, select the one for SystemServer and then iterate over
_all_ threads in _all_ processes running under that UID.
This iteration involves reading many directories in /proc and
many files in those directories.
In the new implementation, we only read time_in_state for the
threads in the system server process itself, identifying them by
the PID. This optimization reduces the cost 5:1 to 8:1 in terms
of CPU time, number of files read and amount of RAM used.
Bug: 167994570
Bug: 163063901
Test: atest FrameworksCoreTests:com.android.internal.os.SystemServerCpuThreadReaderTest
Test: atest FrameworksCoreTests:com.android.internal.os.BatteryStatsTests
Test: atest FrameworksCoreTests:com.android.internal.os.KernelSingleProcessCpuThreadReaderTest
Change-Id: Iaa2eabc9f39999b2867bbffbac93b282cde74165
Introduce angle.egl.features to assist ANGLE debugging. The input
contains lists for each opt-in apps. The format is each list of features
is separated by a comma, each feature in each list is separated by a
colon, e.g. feature1:feature2:feature3,feature1:feature3:feature5
Bug: b/169156341
Test: use adb shell command to verify
Test: atest android.gputools.cts.CtsRootlessGpuDebugHostTest
Test: atest CtsAngleIntegrationHostTestCases
Change-Id: I752047df1c2567b4ebf900c3acb4a965e35dd6c2
Added 1 method:
boolean isAmbientDisplaySuppressedForTokenByApp(String token, int appUid)
Context:
The Settings team is working on adding a string to the AOD settings page when ambient display is suppressed by Bedtime mode: b/168790245.
The current PowerManager#isAmbientDisplaySuppressedForToken(String token) method will only work if the calling app is also the app that suppressed the ambient display. That is, if Digital Wellbeing called suppressAmbientDisplay("winddown"), then isAmbientDisplaySuppressedForToken("winddown") will only return true for Digital Wellbeing. It will return false for Settings app
This CL adds a hidden API PowerManager#isAmbientDisplaySuppressedForTokenByApp(String token, int appUid) that will return true if the given app is suppressing ambient display with the given token. Settings can then call isAmbientDisplaySuppressedForTokenByApp("winddown", digitalWellbeingUid) to get whether ambient display is suppressed by Bedtime mode.
Test: atest FrameworksServicesTests:PowerManagerServiceTest
Bug: 169241595
Change-Id: I41aecc4ff0ab159d67e62193af27362a96c3174c
Log more information to assist debugging without having to rebuild the
system image. Adjust ANGLE related properties.
Bug: b/169156341
Test: atest android.gputools.cts.CtsRootlessGpuDebugHostTest
Test: atest CtsAngleIntegrationHostTestCases
Change-Id: I30b0056d42d30ccadc72fd642068c3aa66592a5c
Corresponding to change in IServiceManager API, for finding out what
instances of a VINTF service are installed.
Bug: 168715768
Test: manual
Change-Id: Ic54b0cfd1bd42e45dc66858dbd614db7ba22083f
In a situation when ION heap size is 0 dumpsys meminfo will not report
ION memory usage. Fix this by differentiating 0-size ION heap or pool
from the cases when /sys/kernel/ion/* do not exist or not readable.
Bug: 168942451
Test: emulate 0-size ION heap and run dumpsys meminfo
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Change-Id: I66624c9e3c73a59462cce8830a36323677db0029
Incremental Serivce periodically polls loading progress and sends to
Package Manager Service. Package Manager provides APIs for other
interested parties to listen to the loading progress.
BUG: 165841827
Test: unit test
Change-Id: I44b9e17c2240b9efe53bc09fc728b6671f1f7dfe
To speed up recurring ApplicationInfo creation, this caches
the the appInfoFlags/appInfoPrivateFlags at parse time, since
they're immutable at that point. This saves having to calculate
it hundreds of times for all the packages in getInstalledPackages
or similar.
This also saves the base app data directories for the system user,
or user 0, interning the strings since they're shared across
packages with the same volume UUID. A better solution would
cache these paths during boot scan and re-use the strings directly
rather than re-building and relying on interning, but there was
no good mechanism for that.
This decreases string append computation time and compromises
between persistent memory and performance.
This also compacts boolean fields into a bitset, decreasing
the amount persisted in both the object memory and the Parcelable
output.
In combination these changes result in a net decrease in memory
usage, although the difference is neglible, on the order of
~4KB for 181 packages or ~22B a package. The increase in speed is
roughly 2/3rds saved off the total time of generateWithComponents.
An important note is that hideAsParsed/hideAsFinal are now
required to be called because they now calculate the derived
fields to be cached.
Bug: 153656459
Test: atest CrossProfileAppsServiceImplRoboTest
Test: atest PackageManagerComponentLabelIconOverrideTest
Test: atest PackageParserTest
Test: atest UserSystemPackageInstallerTest
Test: atest DexoptUtilsTest
Test: atest com.android.server.pm.parsing
Test: atest PackageInfoUserFieldsTest
Test: atest com.android.server.pm.ScanTests
Test: atest android.content.pm.cts.PackageManagerTest
Change-Id: I977edb9dec720893ccb1ce5b9df33733c408d3c1
Apps could use Runtime.exec() to spawn child process and framework
will have no idea about its lifecycle. Now track those processes
whenever we find them - currently during the cpu stats sampling
they could be spotted. If it's consuming too much CPU while its
parent app process are also in the background, kill it.
By default we allow up to 32 such processes; the process with the
worst oom adj score of their parents will be killed if there are
too many of them.
Bug: 160548789
Bug: 157089413
Test: atest AppChildProcessTest
Test: atest CtsAppTestCases:ActivityManagerTest
Change-Id: I95b20a1f36ccd43d09027201d14872305b169d8d
After consulting with the ART team, we learned that sending them
detailed advisory native allocation sizes isn't as useful when the
overall lifecycle of the object is tightly managed, which is the
case with Parcel objects. (Because Parcel users use explicit
obtain() and recycle() methods, any variable-sized native
allocations have already been freed by the time a Parcel instance
is considered for GC.)
The Parcel benchmarks referenced below are showing a uniform ~3%
performance improvement across 1, 4, and 16 thread cases. Note that
this is in addition to the improvements recently made with the shift
to a linked-list pooling design.
Bug: 165032569
Test: ./frameworks/base/libs/hwui/tests/scripts/prep_generic.sh little && atest CorePerfTests:android.os.ParcelObtainPerfTest
Change-Id: Id0ce9b3bff1d0ffb426a9f105c7a54eb00060f85
