Before this change, ZygoteProcess.preloadPackageForAbi returned
as soon as the command was written to the zygote socket and not
after the preload completed. This meant that there was a small
window of time before the server side of the socket polled its FDs
where a second command could be written to the zygote socket. This
would lead to only one of the commands being processed and the
other being dropped. The client side of that socket would then wait
forever for a response and bring down the system once the watchdog
timeout was hit.
Example failure case :
--------------
system_server：send command(preloadPackage)
system_server：send command(fork)
zygote：poll & process command(preloadPackage) // the fork command is dropped.
Example of normal operation :
------------------
system_server：send command(preloadPackage)
zygote：poll & process command(preloadPackage)
system_server：send command(fork)
zygote：poll & process command(fork)
This change makes preloadPackageForAbi synchronous, which ensures
that each POLLIN event corresponds to precisely one command.
Bug: 62886909
Bug: 13618569
Test: Manual
Contributed-By: yuqianyu@huawei.com
(cherry-picked from commit 24a3306c32)
Change-Id: I83faf974c9a70a6ab18323f692c1981784e4c56a
Bug: 29875093
Test: Run dumpsys meminfo -a, verify SwapPss adds up and is non-zero
for dalvik and native.
Change-Id: I79d0b6a59bf5f4e73f75f0b9540ec0fcc9e23b02
Add a new flag in the DevicePolicyManager so that we can Use
EuiccManager#eraseSubscriptions(PendingIntent) to erase all the carrier data
from eUICC chip if the user choose to "ERASE" from the Android device manager.
Bug: 37277944
Test: E2E
Change-Id: Ia78090a00d956c645725be4fd591e02ded8ec467
There are a few AIDL-generated stubs that perform "destroy"
transactions and call "this.destroy()". Previously, this
resolved to the "destroy()" method from their interface on
ART while resolving to Binder.destroy() on the RI, though
this is a subject of a bug report against the JDK,
http://bugs.java.com/bugdatabase/view_bug.do?bug_id=8021581 .
Resolving to the private Binder.destroy, inaccessible to
the Stub, would lead to throwing IllegalAccessErrror.
As we're changing the method lookup to be closer to RI, see
https://android-review.googlesource.com/413119 ,
we should stay clear of the problematic cases whether we
make a decision to follow the RI precisely or not. Therefore
we rename the Binder.destroy() to Binder.destroyBinder().
Test: Nexus 6P boots.
Bug: 62855082
Change-Id: I43baf76b6f3c681d93b411cecf2bc00fccafecac
Causes too many GCs and related slowdowns.
Verified that assistant launch from holding down home button is now
faster than N.
Test: make and flash
Bug: 62769566
Change-Id: Ib0c1f7a45831b241d3376d1e56db3c6937913b1b
Wrong description of the "tid" parameter used by getThreadPriority():
the function does not "change" the priority of the thread associated to
"tid".
Defined also the meaning of setting tid parameter to 0.
Test: None
Change-Id: I8552f3b110eaab1aefa51477e82d3e295547be34
Signed-off-by: Alessio Balsini <alessio.balsini@gmail.com>
...can't hide themselves
Tune the policies for when we tell about apps running in the
background after their services have stopped.
- If it ran while the screen was on, the time we require for it
to be running is much shorter (a couple seconds) as well as the
time we tell about it having run (with another tunable for the
minimum time we tell about this).
- If it has only run while the screen is off and stops a sufficient
amount of time before the screen goes on (currently a second) then
we will not show anything when the screen goes on.
- If it stops when the screen turns on, we will make sure the user
sees about it for a short period of time (currently 5 seconds).
Also includes some improved debug output about handler message
queues.
Test: manual
Change-Id: Iab438410d7182b2dfe4f9c1cce7069b26b34834c
We normally prevent apps from allocating into the "reserved" cache
space, but this change makes an exception for an active camera app,
since the user is probably trying to capture an important memory.
This change only lets the active camera app clear up to half of the
reserved space, since we don't want to completely destroy the
experience of all other apps.
Test: manual app before/during/after active camera session
Bug: 38267830
Change-Id: Ie9e63884fb2638ca881e10b894629eea84601648
It has been discovered that for background values of
AggregatedWakelock, Sync, Job, and partial wakelocks,
the value of getTotalTimeLocked is wrong and often very
negative. getTotalDurationMsLocked, which should provide the exact same
value in all of these cases (since background data is never pooled),
does not have this problem. So while the source of the bug is sought
out, we should use getTotalDurationMsLocked instead of getTotalTimeLocked for
these data.
Bug: 62352334
Test: cts-tradefed run cts-dev -m CtsDumpsysHostTestCases -t android.dumpsys.cts.BatteryStatsDumpsysTest
Change-Id: I78e84368615578483ab8e9e5f0ee1d067491be08
This forces everyone to go through sdcardfs, instead of letting them
around the back door.
Test: builds, boots
Bug: 38231314, 27992761
Change-Id: I97b24d25599c7f86f9b535689e2f4ecf68261dac
We've seen some @SystemApi methods protected with non-system
permissions, so give Doclava the platform AndroidManifest.xml so it
can parse the actual permission protection levels to look for APIs
that are letting in non-system apps.
Also document more @SystemApi permissions.
This is purely a docs change; no logic changes are being made.
Test: make -j32 update-api
Bug: 62263906
Change-Id: Ie0f0a5fb0033817bcc95060f2183a52ae4ae7b06
