The API surface is planned to move to the Connectivity module, meaning
that it will be maintained across multiple Android releases.
This is incompatible with TestApi semantics, which are designed to be
removable across releases. Effectively the APIs become System/module
API, so mark them as such considering that TestApi is not supported by
modules.
Test: m
Bug: 174436414
Change-Id: Icd32fcbb65e9a4bb2b67cb9da7c971281be0781a
* changes:
Address comments on aosp/1539753, aosp/1542487 and aosp/1547496.
Fix propagating underlying caps when a network disconnects.
Test for bugs with suspended VPN underlying networks.
CaptivePortalData only has 7 fields on R, not 8.
Bug: 177290955
Change-Id: I156738d07821fee43420d8af14c4bce36dd0e4a4
Test: atest CtsNetTestCasesLatestSdk:CaptivePortalDataTest on R
aosp/1513052, which generalized support for underlying networks,
broke default network switching when the network underlying a VPN
disconnects.
This is because it calls propagateUnderlyingNetworkCapabilities
in the middle of the bookkeeping operations needed when a
network is disconnected (specifically, after all satisified
requests are removed from the disconnecting network, but before
mDefaultNetworkNai is updated). This is completely incorrect
because propagateUnderlyingNetworkCapabilities can trigger a
network rematch, and running a rematch when the request data
structures are inconsistent is obviously wrong. See the test
changes in this CL for an example of the damage.
Fix this by moving propagateUnderlyingNetworkCapabilities to
before the bookeeping operations begin. It must be before
mDefaultNetworkNai is updated, because otherwise it will not know
that the default network is disconnecting, and it will not be
able to propagate capabilities to VPNs that set underlying
networks to null (i.e., to the default network). It must be
after the nai is removed from mNetworkForNetId because
otherwise it will think that the underlying network is still
connected.
Bug: 173331190
Test: accompanying unit test shows lots of bugs removed
Change-Id: Ibf376a6fa4b34d1c96f8506fa8abbb7595a8c272
This will be mainly used by VCN management service which will
need to hold the networks but preserve the backgrounded-ness
of the networks.
Test: android.net.ConnectivityManagerTest#testRequestType
Test: android.net.cts.ConnectivityManagerTest#testRequestBackgroundNetwork /
--rerun-until-failure 100
Test: ConnectivityServiceTest#testBackgroundNetworks
Test: m -j doc-comment-check-docs
Bug: 175662146
Change-Id: If9aaa87b7e71c2b695ac7b08858850e975b28bb6
This test flakes about 1% of the time on my device. This is due
to the test calling expectNetworkRejectNonSecureVpn() before the
handler thread has processed the call to setAlwaysOnVpnPackage.
Fix this by waiting for callbacks where it is possible to do so,
and by calling waitForIdle where it is not.
Test: test-only change
Change-Id: Ic5750d4fdf9e7e3654a2b6ade74da6bc5de18fb6
Code review comments have suggested that this test is too long
and difficult to understand. Split it into two tests and put some
of the common setup into setup methods and statics.
Bug: 173331190
Test: test-only change
Change-Id: I9fa888c940d7048f1ba6836a5706fbdb84b5f5c9
These fields have been recently audited, confirmed never to be
null, and annotated @NonNull. Ensure that they can never become
null by throwing exceptions in the codepaths that set them.
Also remove some null checks.
Test: atest FrameworksNetTests
Change-Id: I6ce5bb4d69a990f1c857c599b7e50e372352eb87
If IOException happens while trying to start keepalives sockets,
an invalid fd will be constructed. It will fail with
ERROR_INVALID_SOCKET if the user later calls start(). Current
design to construct the invalid fd use the hidden
ParcelFileDescriptor constructor which will not work for the
incoming ConnectivityService mainline. Thus, replace it with
the other formal API.
Bug: 170598012
Test: atest FrameworksNetTests
Change-Id: I57fd7ec2281c5e8c75481200bbde723ecf96982a
1. network cannot be null because it's final and the only caller
of the constructor passes in a fresh copy.
2. linkProperties cannot be null because:
- The only caller of the constructor passes in a fresh copy.
- The only assignment to nai.linkProperties is in
updateLinkProperties, and that dereferences newLp a few
lines before assigning it.
3. networkCapabilities cannot be null because:
- The only caller of the constructor passes in a fresh copy.
- There are no direct assignments to nai.networkCapabilities.
- All 4 assignments happen via getAndSetNetworkCapabilities:
- registerNetworkAgentInternal, updateCapabilities and
applyBackgroundChangeForRematch all pass in the return
value of mixInCapabilities, which is always non-null.
- updateNetworkInfo re-assigns the capabilities that were
already set in the object.
4. networkAgentConfig cannot be null because it's final and
the only caller of the NetworkAgentInfo constructor passes
in a fresh copy.
Address a TODO in ConnectivityService to remove a check for the
nai's capabilities being null.
Also fix a style nit pointed out by the linter.
Test: atest FrameworksNetTests
Change-Id: I2e265f1c60a73d6da0222dd6711bc53c32db06ec
