Currently, various codepaths in ConnectivityService disconnect
networks using NetworkAgentInfo#disconnect. This posts a message
to the NetworkAgent to disconnect and also posts a message to
ConnectivityService to call disconnectAndDestroyNetwork, which
performs all cleanup in ConnectivityService. These two messages
race and the order is non-deterministic.
Instead, always disconnect using disconnectAndDestroyNetwork,
and have disconnectAndDestroyNetwork post the message to the
agent to disconnect.
This fixes a bug where if wifi uses unregisterAfterReplacement
twice in quick succession, when the third agent connects it
doesn't work because the interface is still being used by the
second network.
Also remove the import of IllegalArgumentException to keep the
linter happy (java.lang.* never needs to be imported).
Bug: 286649301
Test: covered by existing tests
Change-Id: Ie01f5589d6839ac6db25f0ba98fc929fbb5b0a96
Now, MdnsServiceTypeClient is created using a network. However,
all tether interfaces use the same network (null), which
means they use the same client for the same service type. This is
not the intended behavior, as each interface should have its own
client. Therefore, MdnsServiceTypeClient creation should be
changed to use SocketKey, which includes both the network and
interface index. This will allow each interface to have its own
client.
Bug: 278018903
Test: atest FrameworksNetTests android.net.cts.NsdManagerTest
Change-Id: I34b7d983f00b67198befb5bf71fc511cf0dabae6
This is a part of preliminary work for sending a v6 NAT-T
keepalive packet in a v6 only network. This commit
release the constraint on constructing a v6 NAT-T keepalive
packet.
Bug: 196453719
Test: Manually connect to v6 only network and check if the
keepalive packets are sent with subsequent commit
Test: atest FrameworksNetTests with the updated test
Change-Id: I9dc96e840265c5dee6b9e3ef2b3901a0aeb4ff9e
Currently, all socket changes are identified using a Network
object. However, the Network object is null for all tethering
interface sockets, which means that the socket cannot be
identified in some places. Therefore, the Network object should
be replaced with a SocketKey object, which includes both the
network and interface index.
Bug: 278018903
Test: atest FrameworksNetTests android.net.cts.NsdManagerTest
Change-Id: Ib49981a4071ecab18c7cf3a8827d1459529492a9
Register a receiver of ACTION_DEFAULT_SUBSCRIPTION_CHANGED to update
the cached default subId.
Bug: 273451360
Test: atest FrameworksNetTests
Change-Id: I3a38af7615fde2811449585d4e659ffaf29c5670
Use AlarmManger ELAPSED_REALTIME_WAKEUP to trigger writing the metrics
to statsd on an interval of 24 hours.
Bug: 273451360
Test: statsd_testdrive
Test: atest FrameworksNetTests
Change-Id: I5e21de058336819189532beffe2b3239a3c69599
This allows VpnManagerServiceTest continues to work after
adding an attribution tag to VpnManagerService
Bug: 278943609
Test: atest FrameworksNetTests:android.net.connectivity.com.android.server.VpnManagerServiceTest
Change-Id: I4228a7fb308b8d01aee53f0734f4c5753908473a
Merged-In: I4228a7fb308b8d01aee53f0734f4c5753908473a
upTimeMillis does not include time spent in deep sleep. Use
elapsedRealtime instead.
The time spent in deep sleep is important to include since the
keepalives may still be ongoing during deep sleep and so that time
should be included in duration calculations.
Bug: 273451360
Fixes: 281639289
Test: atest FrameworksNetTests
Change-Id: I54350d49bdfec58cc00dea6d82cba01f5d387bb0
Store the number of requests and app uids on start keepalives. The
values are also stored in KeepaliveStats and can be recalculated from
KeepaliveStats when reset. Fill the DailyKeepaliveInfoReported proto
with these values.
Bug: 273451360
Test: atest FrameworksNetTests
Change-Id: Ifb9a0048896d31c8ea897cf14768e94dcfd3b010
Register a listener to get notified of SubscriptionInfo changes and
store all carrierIds of active subscriptions in a cache. The executor
for the listener callback runs on a different thread to the connectivity
thread but posts the SubscriptionInfo list to the connectivity thread
for caching.
Bug: 273451360
Test: atest FrameworksNetTests
Change-Id: I889d4da725ccda713367309c257622a0bf9939f3
Create a nested class KeepaliveStats to store and track the carrier info
and lifetime metrics for each keepalive where a keepalive is identified
by its network and slot number. Then aggregate the metrics using
(carrierId, transportTypes, intervalMs) as the key before writing the
metrics into the KeepaliveLifetimePerCarrier proto. Also aggregates the
metrics whenever a keepalive is stopped to free up the (network, slot).
Bug: 273451360
Test: atest FrameworksNetTests
Change-Id: I113ec68053710a7f41f64a9d6fa4dd83bd924225
Following CLs update LegacyNetworkActivityTracker to support multiple
network activities including non default network.
To avoid breaking default network activity APIs, this CL adds tests.
Following CL fix known issues of LegacyNetworkActivityTracker so this
CL also adds test for known issue behavior.
Bug: 279380356
Test: FrameworksNetTests
Change-Id: Ic180efe690807e940d0b412a38ebc5e6d22691c4
When a TCP keepalive start request is sent from ConnectivityService
to KeepaliveTracker, the keepalive is tracked on both
AutomaticOnOffKeepaliveTracker and KeepaliveTracker. The existing
design stops the keepalive inside KeepaliveTracker without
notifying AutomaticOnOffKeepaliveTracker for fd initiated stops.
This causes AutomaticOnOffKeepaliveTracker to lose track of these
fd initiated stops and cause a leak on the object.
The updated design sends the event to ConnectivityService handler
to handle the event down from ConnectivityService. This ensures
that each stakeholder class will get the stop event.
Bug: 283885097
Test: atest FrameworksNetTests
Test: atest CtsNetTestCases
Change-Id: I3c40d80694cd2c046f3a19ddb8f437878c98ab43
In case that the task is not canceled successfully, use session ID to
check if this task should continue to schedule more.
Bug: 285260665
Test: atest CtsNetTest FrameworksNetTests
Change-Id: Ic953ae781449a7ebf943999c8c8d3040a97641d9
When the cache flush bit is set on host address records in a response,
known records are cleared and replaced with the contents of the packet.
However even if the response contains the same records as before, this
caused it to be marked as modified, and response modified callbacks are
sent for every incoming packet.
Instead, ensure that the response is only marked as modified if a
previously unknown address record is added, or if a previously known
address record is removed.
The issue wasn't found by the existing testDecodeWithNoChange test,
because it used a service record for testhost2 instead of testhost1, so
the address records were ignored.
Bug: 285997766
Bug: 285084489
Test: atest CtsNetTest FrameworksNetTests
Change-Id: Ic0f19adf5d9bde7cdab766e49cf677b319a2219b
If the MdnsServiceTypeClient was removed when the
SocketCreationCallback#onAllSocketsDestroyed() was called, the
MdnsSocketClientBase#notifyNetworkUnrequested() will not be
called when a listener is unregistered from the
MdnsDiscoveryManager. This is because the serviceTypeClients will
be cleared after the socket is destroyed. However, these dead
listeners will be re-added to the MdnsServiceTypeClient when a
new socket is created. They will then continue to send callbacks
to the NsdService. Therefore, the notifyNetworkUnrequested()
should be moved to the beginning of the handleRegisterListener()
to ensure that the socket unregistration is properly notified.
Bug: 285997766
Test: atest FrameworksNetTests android.net.cts.NsdManagerTest
Change-Id: I31c766305018889f50a7c12c836174c340d01d7f
The receipt time in the MdnsResponse need to be updated every time new
packet is received. And then the record refreshing logic could calculate
remaining TTL correctly.
Bug: 285260665
Test: atest CtsNetTest FrameworksNetTests
Change-Id: Ib7a290ea0ea8e552c71c657696397e8114fcee52
When an mDNS request (discovery, advertising, resolving...) is
registered and gets assigned a socket on a wifi network, take the
multicast lock to ensure that it can reliably receive mDNS responses.
This is limited to when the application has importance
FOREGROUND_SERVICE or higher.
NsdManager is not documented to require usage of the multicast lock,
which has caused various reports about its reliability. Taking the lock
while the app is in the foreground should address the large majority of
cases, while limiting battery impact.
Going forward this should allow developers on U+ to not take the
lock themselves, allowing optimizations on devices supporting APF,
where instead of taking the lock APF would let through only select
mDNS packets.
Bug: 284389438
Test: atest
Change-Id: I1ce85220eac4a1529b6716d50727c1c462356846
Test: new test for this behavior in the preliminary change
Test: FrameworksNetTests NetworkStackTests
Fixes: 138810051
Fixes: 140610528
Change-Id: I95a979d232fb60ece2e33e972bf5d66d20357a1f
See system/netd/include/Fwmark.h which reserves the bottom 21 bits
of skb->mark for netid, etc... so out of necessity these have to be
in the top 11 bits.
In practice the only values that really make sense are either:
mark == mask == 0 (disabled)
or (mark == mask) being one of the top 11 available bits, for example:
mark == mask == 0x80000000 (enabled, using top-most bit)
(only a single bit makes sense, as this is really just a boolean signal,
and only the topmost is known to be used on any real devices)
Let's just force the use of 0x80000000 for ecosystem consistency.
Bug: 284334830
Test: TreeHugger
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I199e9b00de845b3747940426ea6644426ab72e87
* changes:
FinalizePausedKeepalive in handleStopAllKeepalives.
Add tests for pause, resume and stop keepalives.
Ensure autoKi is not stored when keepalive stops or is not started.
Add tests for NATT keepalives stopped internally in KeepaliveTracker.
aosp/2490881 updated to use InetDiagMessage.destroyLiveTcpSocket for all
devices.
But it is possible that netd socketDestory is modified in T- devices.
So this CL revert changes to keep using netd socketDestroy in T-
devices.
Test: atest FrameworksNetTests
Bug: 284253763
Change-Id: I9b61f10e975d2e38e9829a8c01d3af706e2518ef
If there is no request for all networks, the tethered/local only
interface changes will be ignored. However, these available
interfaces are not used for mDNS when a user requests a socket
with a null network because the interfaces are lost due to the
previous ignore. Therefore, the interface changes should be
retained and will be used for socket creations afterwards.
Bug: 284939720
Test: atest FrameworksNetTests android.net.cts.NsdManagerTest
Change-Id: If830eb53af26f21f497314477b131ce28468a971
mDNS is used by wifi p2p in many places. However,
MdnsSocketProvider does not monitor changes to the wifi p2p
connection to get the names of the available wifi p2p interfaces.
This prevents mDNS from registering or discovering services on
the wifi p2p interfaces. Therefore, listen to the wifi p2p change
intent to know the available interfaces and status changes that
can be used by mDNS.
Bug: 284263838
Test: atest FrameworksNetTests android.net.cts.NsdManagerTest
Change-Id: If03514f1286c0507e5862372272234dd07eb084d
As mentioned in RFC6762 7.1. The records only needed to be renewed if
at least half of the TTL passed. Usually A/AAAA records are included in
the response to the SRV record query, they are not refreshed individually.
Bug: 285261577
Test: atest CtsNetTest FrameworksNetTests
Change-Id: Ifd7140de0d733191256184c5481412e1822d279b
Turn on removeExpiredService feature by: 1) Remove the unnecessary
allowSearchOptionsToRemoveExpiredService flag. 2) Turn on the
removeExpiredService flag in the MdnsSearchOptions.
Bug: 285260665
Test: atest CtsNetTest FrameworksNetTests
Change-Id: Ib115b40e70b0f81a7877deb73af7d61e2e0c385f
Store the handler that should be running all the public methods and
use it by comparing its thread with the current thread.
Also refactor the tests to use and run on a test handler thread.
Bug: 273451360
Test: atest FrameworksNetTests
Change-Id: I053f1043a3b518728c4b30b94eec1d33fbfe39c0
In the case that a keepalive is paused and handleStopAllKeepalives is
called, there is no KeepaliveInfo for the paused keepalive and so no
onError callback will be called. The autoKi will also be cleaned up so
no callback will ever called that notifies the keepalive is stopped.
Bug: 281646074
Test: atest FrameworksNetTests
Change-Id: I8fcaa1f07746235326c7ae05d97e20fd27927fea
The tests added verify the following:
1. Stopping the keepalive will cleanup the autoKi.
2. Pausing the keepalive does not cleanup the autoKi and stopping
a paused keepalive will still call the onStopped callback.
3. A starting error by resuming the keepalive will cleanup the
autoKi.
4. handleStopAllKeepalives also does the appropriate callback for paused
keepalives.
5. Stopping a resumed keepalive stops the correct slot when a second
keepalive is started while the first is paused.
Bug: 281646074
Test: atest FrameworksNetTests
Change-Id: I8925fbe40323dc4584a111d0cf31de016525ef41
In the cases below, the keepalive is already cleaned up internally in
KeepaliveTracker, but are not be cleaned up in
AutomaticOnOffKeepaliveTracker. This means the mAutomaticOnOffKeepalives
list is storing stopped keepalives and the metrics will still conisder
the keepalive as active.
1. In KeepaliveInfo.start, return whether the keepalive is successfully
starting(i.e. mStartedState == STARTING). Check for this in
handleStartKeepalive and return early on failure, without storing the
autoKi. Also check in handleResumeKeepalive.
2. In handleEventSocketKeepalive, return whether the handleStopKeepalive
is called, indicating the keepalive is forced to stop and was not
already STOPPING. This should happen when the state is STARTING but
the network agent returned an unsuccessful event.
3. In handleCheckKeepalivesStillValid, move the checking logic to
AutomaticOnOffKeepaliveTracker so the handleStopKeepalive is called.
Bug: 281646074
Test: atest FrameworksNetTests
Change-Id: I7839464534baed43abbd40b321287132da25b978
MdnsSocketClient doesn't support requesting specific network. As a
result, when processing the response, all the MdnsServiceTypeClient
should be called.
Bug: 285064577
Test: atest CtsNetTestCases FrameworksNetTests
Change-Id: Id95f1e78d60063ccee7f5855a21fa6b5a48fbd3e
Add tests where NATT keepalives are stopped and cleaned up within
KeepaliveTracker in these scenarios:
1. startNattKeepalive is called with an invalid interval.
2. The state is STARTING and the network agent returns an error event.
3. The keepalive started successfully but becomes invalid and
handleCheckKeepalivesStillValid is called on the network.
Bug: 281646074
Test: atest FrameworksNetTests
Change-Id: Iadf9f06030cae20f36d950f520c8dcc4181924ad
In some use cases for the MdnsDiscoveryManager, providing looper as an
argument to the MdnsDiscoveryManager constructor is not feasible. Adding
a constructor without looper can help to match those use cases. If
constructor without looper is used. The shutDown() must be called as a
part of cleanup process.
Bug: 283914408
Test: atest FrameworksNetTests
Change-Id: I0ab04d67bae127628c6d67c4b8c109201612a54b
