Currently, BroadcastInterceptingContext always runs broadcast
receivers on the thread that called sendBroadcast. This means:
1. Receivers might run on the wrong thread, making the test less
realistic.
2. If any receiver checks what thread it's running on, then
either the check needs to be modified or deleted, or the test
must call sendBroadcast on the thread that the receiver
expects to run on. The latter is impossible when there is more
than one receiver that needs to run on more than one thread.
This CL adds a setUseRegisteredHandlers method that allows tests
to say that they want each receiver to run on the Handler
specified at registration time.
This CL also enables the new mode for ConnectivityServiceTest,
and resolves a TODO to re-enable a disabled thread check.
The new mode cannot be enabled by default because it would break
most of the tests.
All the below tests pass except for the car bluetooth test
testAutoConnectAdapterOff_returnsImmediately which was already failing.
Bug: 173331190
Test: atest CarInputServiceTest BluetoothProfileDeviceManagerTest
Test: atest NetworkPolicyManagerServiceTest
Test: atest ConnectivityServiceTest
Test: atest TetheringTest TetheringNotificationUpdaterTest
Test: atest VersionedBroadcastListenerTest EntitlementManagerTest
Test: atest TetheringConfigurationTest
Test: atest CachedDeviceStateServiceTest
Test: atest EmergencyAffordanceServiceTest
Change-Id: I3303bb14516f07a55d82a16b59c111ab3f8b0389
1. Remove the legacy network type. This is only used for logging.
- Replace the logcat logging with the display transport
- Remove the EventLogTags logging, since it's likely not
actually used by anyone.
2. Remove code that checks for NetworkInfo objects in state
FAILED, since LockdownVpnTracker can never have received any
such NetworkInfo from ConnectivityService since lollipop.
Bug: 173331190
Test: passes existing tests in ConnectivityServiceTest
Change-Id: I66ed71e51ba18b95862f3a0a5df2775eecea501e
This information originates in non-mainline system server code
and needs to be passed to the connectivity mainline code.
Bug: 173331190
Test: already covered by CTS tests
Change-Id: Ic612b6a51f7ec13e2213c8754312cf716130c876
* changes:
Check registering system default callback needs NETWORK_SETTINGS.
Move VPN code from ConnectivityService to VpnManagerService.
Add a skeleton VpnManagerService, and start it on boot.
Convert LockdownVpnTracker to NetworkCallbacks.
Minor fixes to VpnTransportInfo.
Main implementation of ConnectivityService.setOemNetworkPreference. This
covers the main requirements of this method including listener
functionality.
Bug: 176495594
Bug: 177101287
Bug: 176494815
Test: atest FrameworksNetTests
atest NetworkStackTests
atest FrameworksNetIntegrationTests
atest NetworkStackIntegrationTests
atest CtsNetTestCasesLatestSdk
Change-Id: I8d318ab07785e52dd84d6261fdea8f318dce9bc5
ConnectivityService itself does not depend on mVpns or the Vpn
class any more. Most of this CL is simply moving code from one
class to another:
- Move the AIDL calls from IConnectivityManager to IVpnManager.
- Move the implementation from ConnectivityService to
the new VpnManagerService.
- Move the APIs from ConnectivityManager to VpnManager, but
temporarily maintain some shims in ConnectivityManager for the
moved calls so that existing callers do not have to be modified
in this CL.
- Update VpnService to call IVpnManager instead of
IConnectivityManager.
- Move the code that registers the VpnManager service from
ConnectivityFrameworkInitializer to SystemServiceRegistry.
Bug: 173331190
Test: atest HostsideVpnTests FrameworksNetTests CtsNetTestCases
Change-Id: I4911e2144df721a94fa00da9edf0dc372a7091c2
This adds a lot of unused code but this should make it easier to
review subsequent CLs.
Bug: 173331190
Test: builds, boots, "dumpsys vpnmanager" succeeds
Change-Id: Ied379654a0c3ab6242d3125661fe30f322395059
This will allow moving LockdownVpnTracker from the connectivity
to the VPN code. This requires moderate refactoring since it's
pretty tightly coupled to both.
In this CL:
1. Add an @hide API to tell ConnectivityService that legacy
lockdown VPN is enabled. I chose not to use the existing
setVpnRequiredForUids API because that method has specific
semantics and because it will be required long term since
it's used by non-legacy VPN types.
2. Instead of updating LockdownVpnTracker inline from the
ConnectivityService handler thread, have it listen to
NetworkCallbacks. This introduces an extra thread hop, but
most of the interactions between the lockdown VPN and CS were
via NetworkAgent, which is asynchronous anyway.
3. Add code to LegacyTypeTracker to send the extra
CONNECTIVITY_ACTION broadcast for the underlying network type
that is sent after the VPN connects. In order to do this, make
Make LockdownVpnTracker specify its underlying network
(via setUnderlyingNetworks) when it connects.
4. Reimplement LockdownVpnTracker#augmentNetworkInfo based on
information that is available in ConnectivityService.
5. Remove the code in LockdownVpnTracker that counted errors.
I think this code has not worked since lollipop, because
ConnectivityService never sees NetworkInfo objects in state
FAILED. This is because ConnectivityService only hears about
NetworkInfo objects via NetworkAgents, and LegacyVpnRunner
only registers its NetworkAgent when the connection succeeds.
Bug: 173331190
Test: passes existing tests in ConnectivityServiceTest
Change-Id: I66d18512882efd468ee0ecec61f28786a195b357
* changes:
[Mainline] android.system package APIs migration
Use public android.system.Int64Ref instead of Int32Ref which is hidden (CorePlatformApi) type of libcore
Use data that is already available in ConnectivityService
instead.
The behaviour of the new implementation is slightly different
from Q and R code when the permission check fails.
- The old code would throw a SecurityException if an app that
was not an active VPN called the method, and would return
INVALID_UID if the connection belonged to a UID that was not
subject to the VPN.
- The new code returns INVALID_UID in both cases.
This does not seem like a compatibility problem. The only case in
which the code throws SecurityException is if the app is not a
current VPN app, but the app already knows whether it is or not.
The docs don't mention that the method SecurityException, either.
Bug: 173331190
Test: atest FrameworksNetTests
Test: atest HostsideVpnTests
Change-Id: If3d031e74df33b5c97e12ebf02272faac6769d50
This currently stores the VPN type and a session name, but can be
extended in the future.
Bug: 173331190
Test: added coverage in VpnTest
Test: added coverage in ConnectivityServiceTest
Test: added coverage in NetworkAgentTest
Change-Id: I450858a9fa332c8d896dbdb4c14337d5ec23677f
This method allows internal components to track the system
default network. This differs from registerDefaultNetworkCallback
because that method sends callbacks for the default network for
the UID that called it. This may not be the system default
network, for example, when a VPN is up and applies to the UID.
Bug: 173331190
Test: new unit tests in ConnectivityServiceTest
Test: new unit tests in ConnectivityManagerTest
Test: new CTS tests in ConnectivityServiceTest
Test: new CTS tests in HostsideVpnTests in other CL in this topic
Change-Id: Id02748a2183f71b71ff2a53a580466b9dcecaa93
* changes:
[FUI07] Stop making NetworkState with NetworkInfo from external callers
[FUI06] Stop using NetworkInfo in NetworkState from external callers
