Rename the class and ensure that all code related to bad wifi
avoidance has names that reflect its purpose. This will allow us
to reuse most of the the code for other purposes in future CLs.
Test: runtest frameworks-net passes
Bug: 34630278
Change-Id: Ida190e5a79e7a12f4eec07e1a3e4d9d138e83bcd
testApfFilterRa is failing with probabiliy 1/10 ~ 1/15 on the following
assert: assertDrop(program, packet, lifetime/6), for lifetime values
that are multiple of 6, where 6 is the hardcoded fraction of RA lifetime
to filter in ApfFilter.java.
When the lifetime is not a multiple of 6, the remainder of 1 to 5
seconds gives enough margin so that when the APF program is simulated
the faked lifetime of the program is less than lifetime/6 away and the
packet is dropped.
However for lifetimes which are exact multiples of 6, this margin is
always 0s and that result in nondeterminism in the result. This is
consistent with the obervation that the only failed assert was for a
lifetime of 300s, the only multiple of 6.
This can be observed by detecting the age limit at which the filter
stops dropping packet oscillating between lifetime/6 and lifetime/6 + 1
for lifetimes which are multiple of 6.
This patch fixes the flakyness by freezing the flow of time in tests so
that the expected filter age threshold is consistent and stable.
Test: no failure observed in 1000 runs.
Bug: 32561414
Change-Id: I5251d047039f34b82ce8a5d20ae46563e1e0cce8
This patch replaces System.currentTimeMillis() with
SystemClock.elapsedRealTime() to make RA lifetime computation more
resilient to various external events inducing jumps in
currentTimeMillis().
Test: ApfTest passes.
(cherry picked from commit 305af8e98a)
Change-Id: Idbde700025fecfecefb8162d66b94194a87829d5
This patch adds a getUint8 getter for ByteBuffers and changes ApfFilter
to make uses of getUint8/16/32 everywhere.
The return types of getUint16 is also changed from long to int, which
will expand gracefully to long as an unsigned int as it is guaranteed to
be positive after getUint16.
Test: ApfTest passes
(cherry picked from commit 995dd94673)
Change-Id: Idde3c9d03d39fbdf6f9b84d398f3fe8ea371483d
Test: as follows
- built
- flashed
- booted
- runtest frameworks-wifi passes (though no test covers this code yet)
Bug: 34210527
Change-Id: I2d44445982dfb3221f7b6394daf0479c1fa8a6e1
Test: as follows
- built
- flashed
- booted
- runtest frameworks-wifi passes (though no test covers this code yet)
Bug: 34210527
Change-Id: I2d44445982dfb3221f7b6394daf0479c1fa8a6e1
During IpManager startup, anything sending messages to the state machine
must not begin doing so until after the state machine has been started.
Reorder the constructor accordingly.
During shutdown, AvoidBadWifiTracker needs to unregister the registered
BroadcastReceiver and might as well also unregister the ContentObserver.
Test: backport from internal
Bug: 33388922
Change-Id: I58e07f7ccddaab160c153bcfb69fd45f50bb8710
During IpManager startup, anything sending messages to the state machine
must not begin doing so until after the state machine has been started.
Reorder the constructor accordingly.
During shutdown, AvoidBadWifiTracker needs to unregister the registered
BroadcastReceiver and might as well also unregister the ContentObserver.
Bug: 33388922
Change-Id: I58e07f7ccddaab160c153bcfb69fd45f50bb8710
Per https://tools.ietf.org/html/rfc7844#section-3.7
When using the anonymity profile, DHCP clients SHOULD NOT send the
Host Name option.
Once the net.hostname property is removed and ensured unset via CTS
this option will no longer be sent.
Test: as follows
- build, flashed
- connected to v4-only and dualstack networks, verified v4 works
- android.net.dhcp.DhcpPacketTest passes
- setprop net.hostname "" and verified no hostname option
Bug: 30979015
Change-Id: Ieab3f04ea38e176d9147bb788742a5397762c2e4
Test: new unit test + $ runtest franeworks-net
Bug: 32833400
(cherry picked from commit f98182ef5e)
Change-Id: I4c46304b9dc8105123fc02a29f99dbc835248eb0
Test: new unit test + $ runtest franeworks-net
Bug: 32833400
(cherry picked from commit f98182ef5e)
Change-Id: Ifaf6e778c811c7d865c790a293b1fce3f43cad1c
This patch replaces System.currentTimeMillis() with
SystemClock.elapsedRealTime() to make RA lifetime computation more
resilient to various external events inducing jumps in
currentTimeMillis().
Test: ApfTest passes.
(cherry picked from commit 305af8e98a)
Change-Id: If19011fc0c905948f2e42b975cfcc5f8672a95fb
This patch removes the synchronization around the private variable
mRunning inside of IpReachabilityMonitor and instead qualifeis the field
as volatile.
Synchronization is not needed for reads/writes on native fields or
object references because they are already guaranteed to be atomic.
Synchronization here was used for enforcing memory visibility across
concurrent threads indirectly through monitor acquire/release.
The volatile keyword achieves this in a more explicit way.
Also, this patch changes the way that probeAll() copies the
IpReachabilityMonitor's mIpWatchList by temporary holding mIpWatchList
keys into an ArrayList instead of a more expensive HashSet. Since Java
HashSet are just degenerated HashMaps, and that key iteration order is
based on key hash, the iteration order over this temporary collection
will be consistent for the same mIpWatchList.
Test: refactoring CL. Existing unit tests still pass.
(cherry picked from commit b0f1186c03)
Change-Id: I48d2b4d837a459150cd431b400ec01b87b48c014
This patch adds a getUint8 getter for ByteBuffers and changes ApfFilter
to make uses of getUint8/16/32 everywhere.
The return types of getUint16 is also changed from long to int, which
will expand gracefully to long as an unsigned int as it is guaranteed to
be positive after getUint16.
Test: ApfTest passes
(cherry picked from commit 995dd94673)
Change-Id: I606ebc5aedfcacde400d27cc6bc37145769b122c
If construction of a new IpReachabilityMonitor throws an IAE then
log it and immediately call onProvisioningFailure().
Test: runtest frameworks-wifi
passes, except for selectQualifiedNetworkDoesNotChooseDeletedEphemeral()
which fails with an NPE for unrelated reasons.
Bug: 31038971
Bug: 31742703
Change-Id: Ie91b8bdd509d06ad54d062bf446e74c092eb096c
(cherry picked from commit e452660466)
If construction of a new IpReachabilityMonitor throws an IAE then
log it and immediately call onProvisioningFailure().
Test: runtest frameworks-wifi
passes, except for selectQualifiedNetworkDoesNotChooseDeletedEphemeral()
which fails with an NPE for unrelated reasons.
Bug: 31038971
Bug: 31742703
Change-Id: Ie91b8bdd509d06ad54d062bf446e74c092eb096c
This patch adds an explicit check in the DHCP packet parser for
rejecting packets without a magic cookie, instead of relying on the
top-level try-catch-all in the parser.
This allows to add to DHCP error metrics this specific error.
It also allows to add two poor man's fuzzing tests that tries to find
additional gaps in the DHCP packet parser by
- trying to parse all subslices of a valid offer packet.
- trying to parse random byte arrays.
Test: covered by previously introduced malformed DHCP packet unit tests
+ additional fuzzing tests.
Bug: 31850211
Change-Id: If53c9ba9df78d7604ec018c9d67c237ae59c4833
This patch adds a try catch all to DHCP packet parsing so that
DhcpClient does not choke on malformed packets, brinding down with it
the whole framework.
Test: added new unit tests catching the issue fixed in this patch.
Bug: 31850211
Change-Id: I3c50a149fed6b2cbc4f40bb4f0e5bb2b56859b44
Additionally, add this utility class to IpManager for compatibility
verification. A follow-on CL will make use of IpManager's local
AvoidBadWifiTracker.
Bug: 31827713
Change-Id: If8c56c3f8076d6a5157ea180e361bbdadc2bc1dd
When IPv4 is provisioned on an interface with Apf capabilities,
ApfFilter will only keep track of the raw ipv4 address, with no
information about the subnet or prefix length.
This patch adds the missing prefix length information to ApfFilter. This
allows to calculate the subnet broadcast ipv4 address for more
precise ipv4 broadcast filtering when the multicast lock is not held.
Bug: 30231088
Change-Id: Iebaec040703647c4ced30bb585be173e97a1fae5
This patch refines the Apf IPv4 filter for interfaces with Apf
capabilities to drop packets when:
- the multicast lock is not held
- the packet is an IPv4 packet
- the packet is not a DHCP packet addressed to us
- the packet is L2 broadcast, or IPv4 multicast, or IPv4 broadcast
- caveat: subnet broadcast address is not checked.
This allows to drop IPv4 broadcast packets whose MAC destination
address is not the L2 broadcast address but the current address of the
interface. Such packets can be received on network that rewrite L2
addresses and can put significant pressure on battery by waking up the
phone unnecessarily.
Bug: 30231088
Change-Id: I8b1785fc5ceadaa1f2881765983e502135dcbc46
This patch improves the test coverage of ApfFilter for IPv4 broadcast and
multicast traffic.
Bug: 30231088
Change-Id: I2e1d5d5a7cbae17f3d5978a4cf5f10666b5c6104
