android.display being in the foreground cpuset group is an issue. As
seen on M/S, during heavily CPU load it is not given core 3 even though
it might be free and causes jank. This patch adds the thread to the
top-app group to ensure it is placed on all cores during scheduling
decisions.
Doing this required a couple of changes:
- new API to set per-thread cpusets
- changes to DisplayManagerService to set the thread to top-app group
- changes to SystemServer to set the policy toward the end, as doing it
during start of the DisplayManagerService was in issue (issue being
SystemServer calls setSystemProcess.. -> setProcessGroup which overrides
the group settings for threads in the system server process, including
android.display)
Bug: 36631902
Test: Boot and make sure android.display thread is in the top-app group
Change-Id: Icc394ea0ffcf159d11728ad38de114234a29d20f
Signed-off-by: Joel Fernandes <joelaf@google.com>
StatFs.restat() and the StatFs constructor can throw
IllegalArgumentException. This was not previously documented;
not all callers took this into account, for example:
http://r.android.com/251290
This CL adds documentation to those methods. It also adds
comments to two of the callers.
Separately from this CL, we may in addition consider adding
new API StatFs.checkedRestat() and StatFs.checkedCreate()
or similar that throw IOException; we cannot change the
existing constructor and method since they are public.
Test: Checked that "make" still completed successfully.
Change-Id: I6a0b3cb7718939408937c61de7c3b000b948fa59
Instead of trying to be clever by poking at underlying flash part
sizes, rely on the fact that device storage printed on retail
packaging is a power-of-two value.
For a typical device with a 23GiB data partition, this will return
a value of "32GB" which matches the retail packaging.
Test: builds, boots
Bug: 34827187
Change-Id: Ib4cf7f637dffc9238252e1fedcd86dc8b5cf656d
Added background times and counts for an app's scheduled job usage.
Changes DualTimer to be a subclass of Timer so that background timers
can be easily accessed from the main timer (which is convenient for
jobs, whose ArrayMap of Timers is directly accessed outside of
BatteryStatsImpl).
Bug: 35669746
Test: runtest -x
frameworks/base/core/tests/coretests/src/com/android/internal/os/BatteryStatsTests.java
Change-Id: Ic1d85db34346ebda94ed39f134fc0794a5877815
(cherry picked from commit d253f537a0)
The wifi scan actual time is desired (in the batterystats.proto) to be
in msec instead of usec. It makes more sense to therefore output the
value in msec rather than convert it when it gets read into the proto.
Bug: 35677312
Test: runtest -x
frameworks/base/core/tests/coretests/src/com/android/internal/os/BatteryStatsTests.java
Change-Id: I22fce69b6faecd00f79837369344e04b426a1c33
It's confusing, but f_bsize is not the value you're looking for; the
real block size is f_frsize. Fix all those bugs.
Also, the vast majority of clients are interested in the usable
disk space, not including reserved space.
Test: builds, boots
Bug: 36840579
Change-Id: Ib1470389afd49c14cab62282ec1e978ebb2c4791
Added background times and counts for an app's scheduled job usage.
Changes DualTimer to be a subclass of Timer so that background timers
can be easily accessed from the main timer (which is convenient for
jobs, whose ArrayMap of Timers is directly accessed outside of
BatteryStatsImpl).
Bug: 35669746
Test: runtest -x
frameworks/base/core/tests/coretests/src/com/android/internal/os/BatteryStatsTests.java
Change-Id: Ic1d85db34346ebda94ed39f134fc0794a5877815
Yum!
Also needed to have a Context.revokeUriPermission() variant that is sane,
so reasonable CTS tests can be written.
Test: new ClipDataJobTest added.
Change-Id: Ia3135ea788a6e32c971bae7dab3a844d0ef4139c
Make it clear that now this restriction applies across the whole device,
and is settable by profile owner.
Bug: 31000521
Test: None; just javadoc update
Change-Id: If82d8f4167cf1b77d6e10554f1d96b2d3a2a54a0
The old .equals code uses Objects.deepEquals,
which does not handle for, e.g. ArrayList<byte[]>.
Create our own recursive HidlSupport.deepEquals
and deepHashCode method to handle these edge cases.
Bug: 36454147
Test: hidl_test_java
Change-Id: I56dc48985486b561263c260cf43aa90f30d581d0
* changes:
Compile secondary dex files according to REASON_BACKGROUND_DEXOPT
Register secondary dex files for JIT profiling
Move DexLoadReporter out of LoadedApk
Set pm.BackgroundDexOptService as the source of true
Add missing return in DexManager
Update package use info when the app data is updated
Previously callback methods of ProxyFileDescriptorCallback were invoked
on a background thread prepared in the framework. So all methods were
invoked and processed synchronously. This was problem because if it took
time to fetch bytes of one file, operations for other files were also
blocked.
The CL changes ProxyFileDescriptorCallback methods to be invoked on
Handler passed by apps. Now application can prepare a Handler per file
so that one file does not block others.
Bug: 35229514
Test: cts-tradefed run cts -m CtsOsTestCases -t android.os.storage.cts.StorageManagerTest#testOpenProxyFileDescriptor_async
Change-Id: Ibadc4aad4c0373a3da586459a8f775e40288c895
(cherry picked from commit 4f156065c8)
Test: boot, and check that profiles get recorded for secondary dex files
Bug: 32871170
Bug: 26719109
(cherry picked from commit f5a7bfc8d5)
Merged-In: I2de23ef44eee3f1783ae698821f1c6d88c66c9a6
Change-Id: Id9bbd630b8485dc17eeef846295458df5cfe446a
The old .equals code uses Objects.deepEquals,
which does not handle for, e.g. ArrayList<byte[]>.
Create our own recursive HidlSupport.deepEquals
and deepHashCode method to handle these edge cases.
Bug: 36454147
Test: hidl_test_java
Change-Id: I56dc48985486b561263c260cf43aa90f30d581d0
Support loading a WebView package which specifies the name of a "donor"
that provides missing files. This allows a preinstalled stub WebView to
function by loading its code and assets from the preinstalled Monochrome
implementation, as long as the versions are close enough that the
manifest contents are compatible, which should be fine since
preinstalled versions will match.
To do this, we replace the stub's code paths in AppplicationInfo with
the donor's, so that all Java and native code and resources are loaded
from the donor APK at runtime instead of from the (mostly empty) stub.
To get the ClassLoader with the modified path cached as if it was the
regular path, we introduce a new "cacheKey" parameter in
ApplicationLoaders.
Bug: 21643067
Test: build "new" stub WebView upstream in chromium and test loading
Change-Id: I08cc9122b1c9def3e1206974f3e0e8973cca3419
It's extremely difficult to test storage related logic on devices
that don't have physical SD card slots. So to support better
debugging and testing, add a new "virtual disk" feature which mounts
a 512MB file through loop device.
Also move ParcelFileDescriptor.open() over to using Os.open() so
that it gets StrictMode treatment.
Bug: 34903607
Test: builds, boots, virtual disk works
Change-Id: I072a3a412cfcc8a2a3472919b7273a1ed794fd98
1. Listen to calling package binder death stopping the scanning on that.
2. Don't restart scanning when a request with the same values was made.
Bug: 30932767
Test: 1.:
- Using a test app start scanning and kill the app.
- In debug mode ensure that DeviceDiscoveryService#stopScan gets triggered
2.:
- Start scanning and rotate the device while device chooser is visible
- Ensure no visible loss of state is happening
Change-Id: If126a2c963da90172a956fbd88e6b3dcd7ac8b01
They were added for a CL that later stopped using them, so there
is no current need for them in public API.
Bug: 35812784
Change-Id: I2a9a30bee65894235f2cc81ab1bd5a37c9296c02
Test: build succeeds
When Keystore attempts to gather the application attestation
ID for an app that runs not as the primary user/owner of the
device, the PackageManager would deny package manager access
to the required PackageInfo on the grounds that the request
came from a different user
(violates android.permission.INTERACT_ACROSS_USERS)
This patch adds an additional check to
KeyAttestationApplicationIdProviderService, that verifies the
caller is indeed Keystore. Then it drops the caller context
and retrieves the requested PackageInfo, before restoring
the calling context.
Bug: 35719178
Test: install APK [1] and run as secondary user
[1] https://drive.google.com/file/d/0BzV-JgYFLSDIRGY1WENncmFMVW8/view?ts=58d3f3e6
Change-Id: I91f93a56d10498e1ee3bdb983c701033594c3e57
