WTF will happen in RegularMaintenanceJobService due to
scheduling job and starting maintenance job are in different
processes. In AGO project, scheduling job use inprocess process.
But starting maintenance job use networkstack process. So
IpmemoryService objcet will be null when trying to do the
maintenance job.
Bug: 132763800
Test: Verify on both AGO and non-AGO projects that WFT will not
happen.
Test: atest NetworkStackTests
Change-Id: I545fc8384ef7b3d97edf2270fa033d8afac2669c
Merged-In: If7f61a8b5ddc8cdd25928cd35afa7fece2c8dde5
Merged-In: I8043c405b3c78d8dd1e71ec584d7a2fdaa2fd989
Merged-In: I545fc8384ef7b3d97edf2270fa033d8afac2669c
(cherry picked from commit 20abd7d4e6)
The in-process NetworkStack runs in the system process, so it does not
need to be marked persistent.
Doing so causes a spurious com.android.networkstack.inprocess process to
be running, wasting memory.
Bug: 131046856
Test: Flashed sailfish, networking works, no extra process started
Test: Flashed walleye, networking still works
Test: aapt dump xmltree NetworkStack.apk AndroidManifest.xml unchanged
Change-Id: I0eb896c825ffe2581950c0bdd2b31ebd7eae0982
This reverts commit 796afddd6c.
Reason for revert: breaks updatability because persistent processes are not updatable.
Bug: 131046856
Change-Id: I35ded113b31e742f13ad8523a10af3a83a111374
Merged-In: I35ded113b31e742f13ad8523a10af3a83a111374
(cherry picked from commit b2a5d85522)
The network stack should never be killed, because if it does,
recovering networking requires a reboot. Make the process
persistent to reduce the chances of it being killed by the OOM
killer.
Bug: 131046856
Test: builds, boots
Test: process oom_score_adj is -800
Test: process goes from "Persistent Service" to "Persistent" in dumpsys meminfo
Change-Id: Ic8ebf6356e23489f6cd984d680ff9f1b4cbbe52d
Merged-In: Ic8ebf6356e23489f6cd984d680ff9f1b4cbbe52d
(cherry picked from commit c1eedfcd12)
Implement regular maintenance of IpMemoryStoreService. Regular
maintenance is scheduled for when the device is idle with access
power and a minimum interval of one day.
Bug: 113554482
Test: atest NetworkStackTests
Change-Id: Id3985e30d12307fc2e9fcbe782caaf97a627cef3
Permissions defined in AndroidManifest_Base.xml would be inherited in
AndroidManifest_InProcess.xml and cause issues on svelte builds, since
privileged permissions are not whitelisted for
com.android.networkstack.inprocess.
The network stack running in the system_server process does not need
these permissions anyway since it has system_server permissions.
(clean cherry-pick from Ica63ce586e701b1f2415a97f63ba27cc25e4830a)
Test: flashed, booted crosshatch_svelte-eng, WiFi works
Test: flashed, booted blueline-eng, WiFi works
Fixes: 129485310
Merged-In: Ica63ce586e701b1f2415a97f63ba27cc25e4830a
Change-Id: Ib4b478fb39f43a2fd4221c7d1f1ea51de04e0fba
targetSdkVersion was set to 28 in aosp/916713, however this was broken
by the manifest refactor in aosp/923338: the targetSdk is not preserved
in the manifest merge.
Test: aapt dump badging shows targetSdkVersion 28
Change-Id: I7d06b141856e96ef011db32094bfc62a8cd1b350
Instead of having a library in the classpath where the NetworkStack is
created by reflection on devices using the network stack in-process,
just bind to a version of the network stack service that runs in the
system process.
If the in-process version is installed it will be used, otherwise the
module is used with its own process.
Bug: 127908503
Test: blueline boots and has functional WiFi
Test: atest FrameworksNetTests NetworkStackTests
Test: svelte build boots and has functional WiFi
Change-Id: I7722b173e5686e0dbb9cfddcfb34344b0f4135d9
