No functional change yet, since startReadPacketFilter() has no callers
at this time. In the future, this new hook will be used to take
periodic snapshots of the APF memory (for instance, when the device
wakes up).
Design note: WifiStateMachine grabs the APF data synchronously
from another thread, but then the data snapshot is delivered to IpClient
via an asynchronous reply, following the same pattern used by other
commands. This means that there's no (practical) way for IpClient to
read the APF data just before replacing the APF program.
Even with this limitation, it's still possible to reliably decode packet
counters and compute deltas relative to the last snapshot, provided that
the address range isn't cleared when installing a new APF filter.
Bug: 73804303
Test: Manual - called the new code and inspected 'dumpsys wifi' output.
Change-Id: Ia0923d71cf3ee4128fb1c381557316300adac1a3
This required advancing the IpManager to IpClient refactoring
(Bluetooth used IpManager and friends). Most importantly, the
Bluetooth code used WaitForProvisioningCallback, so this is
moved into IpClient proper now. Also: some more renaming
cleanup.
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes
Bug: 62476366
Bug: 72663849
Merged-In: I5e5b2c59ad3ea9ad574e8e683bcab3b09cfe0791
Merged-In: I89b45310181d952129fb2294db63654da9b4057f
Change-Id: I9eee2a9d5c31cdd48a054a7edc3403584efb1864
(cherry picked from commit 8d1fe54be1)
This change comprises the following parts:
[1] android.net.dns.ResolvUtil, containing methods that encapsulate the
use of the high bit in netids used in DNS resolution contexts.
[2] Updates to captive portal apps to call the ResolvUtil method that
enables DNS-over-TLS bypass for the captive portal app process.
Test: as follows
- builds
- flashes
- boots
- runtest frameworks-net passes
Bug: 64133961
Bug: 72345192
Merged-In: I0994b53d24ed25a2eb9e65429c61cf6fa87c7513
Merged-In: I4c49e23d8caa4d485df1c1d2f135a7282d439c0b
Change-Id: I2072c1f68d6978fa0d7e9d8693135a2c51bb0f87
(cherry picked from commit 2140529d9b)
Some WiFi hotspots send excessive multicast IPv6 ping requests which
cause considerable power drain. Thus, drop all multicast ICMPv6 ping
requests while the device is in doze mode.
Bug: 66930077
Test: runtest -x tests/net/java/android/net/apf/ApfTest.java
Test: manual - compared generated APF bytecode using apf_disassembler
Test: manual - force idle mode while sending multicast pings to device
Change-Id: I25dbf99985cf4413b4d0e3920d281fe6d663197a
(cherry picked from commit e2c70f8a60)
Instead of requiring an *exact* version, we now require a minimum
version. This is necessary if we wish to upgrade the APF interpreter in
vendor firmware independently of the framework code. Adding new opcodes
without changing the semantics of the old ones is always possible and
usually easy, so we can at least have a backwards-compatibility horizon
wider than a single version.
Bug: 36221302
Bug: 73804303
Change-Id: I4f1ed661afb2ede193188eb8086fbf2a8b45b33d
Test: Change extracted from aosp/647079 and tested there
(cherry picked from commit d6bf7ffea7)
Addresses a long-standing TODO. Now, when calling IpClient's
startProvisioning(), the interface has to be available (i.e.
InterfaceParams#getByName() must return non-null).
Also:
- add a test
- refactor for testability
- delete some constructors no longer used
- properly handle passed-in null IpClient.Callback
- some more IpManager -> IpClient renaming
- permit recording metrics before starting a provisioning
attempt (logging immediate errors) without Log.wtf().
Test: as follows
- built
- flashed
- booted
- runtest frameworks/opt/net/wifi/tests/wifitests/runtests.sh passes
- runtest frameworks-net passes
- basic WiFi IpClient connections works fine
Bug: 62476366
Bug: 73487570
Merged-In: I68e5e24122dc31e730cdbe8d75e33847e6332da4
Merged-In: Ifd27f5d908947cd7b4e1b8d54f9fa87e43ebb11b
Merged-In: Ief3c8e1652f69af0276fe35946ae1bf6e6b1b57e
Change-Id: Ic83ad2a65637277dcb273feb27b2d1bb7a11eb2b
(cherry picked from commit b152cd0aa4)
Add a simple CountDownLatch and a public method that callers can use
to block until IpClient has cleanly shutdown the state machine.
In cases where IpClients are created and destroyed dynamically for
the same interface name, this can be used to make sure only one
IpClient at a time is touching the given interface.
Test: as follows
- built
- flashed
- booted
- OTG ethernet works (i.e. client mode)
- removing and reinserting the ethernet dongle continues
to show the <-> Ethernet sysui icon and basic network
connectivity works
- removing and reinserting the cable (link up/down events)
also shows the <-> Ethernet sysui icon and basic network
connectivity works
Bug: 62476366
Change-Id: If4bffd54b7ebc088ec07cac10251e451f8161b6c
Add @GuardedBy for simple functions that require a single lock
and are named XYZLocked.
Derived by errorprone.
Bug: 73000847
Test: m
Test: m javac-check-framework RUN_ERROR_PRONE=true
Change-Id: Id1d9fbe5018250e3b9d1466fc5f670d5ad902fa3
Due to an issue resolving the boot classpath, the
KeepalivePacketData structure cannot be referenced
by frameworks/opt/telephony while it is in services.
-Move KeepalivePacketData to android.net
-Also, relocate IpUtils without changing the package
name.
Bug: 38350389
Test: compilation
Merged-In: If5fc63e9ad8b9b2d4c2fee47ff4bab2ab190a05a
Change-Id: If5fc63e9ad8b9b2d4c2fee47ff4bab2ab190a05a
(cherry picked from commit 41002e3080)
Due to an issue resolving the boot classpath, the
KeepalivePacketData structure cannot be referenced
by frameworks/opt/telephony while it is in services.
-Move KeepalivePacketData to android.net
-Also, relocate IpUtils without changing the package
name.
Bug: 38350389
Test: compilation
Change-Id: If5fc63e9ad8b9b2d4c2fee47ff4bab2ab190a05a
(cherry picked from commit bd4093be7cab51d6ad031854324134a9c8637c2c)
Also audit all constants, make some private, annotate some
with @VisibleForTesting.
Test: runtest framework && cts
Change-Id: Iaf5ea7abd36fd8d544dcc84654f6cb529196d654
NetworkInterface throws Exceptions every time you look at it askance.
Try to make something we instantiate fully, once, and pass it around.
Partial MacAddress-ification as well, for good measure.
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes
- basic IPv6 tethering works
Bug: 32163131
Bug: 62476366
Change-Id: I16c145fddb4e76387370338d792a438eed886d7f
This reverts commit 94209ab768.
Reason for revert: should not have auto-submitted prior to more extensive wifi team testing.
Change-Id: Ie81b10473caf34971226948038bc20dc4fa6a1ae
Additionally, no longer try to transition from within a State's
enter() method (this can encounter Log.wtf()s). Introduce some
CMD_JUMP_* commands and use deferMessage().
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes (except for IpConnectivityMetricsTest failures)
- manual changing from DHCP to static configurations works:
2017-12-11T19:06:19.082 - INVOKE onLinkPropertiesChange({{InterfaceName: wlan0 LinkAddresses: [] Routes: [] DnsAddresses: [] Domains: null MTU: 0}})
Bug: 69800563
Bug: 70394432
Change-Id: Ice249a48b66806c0270ec3f11dd2e8e387d4e29b
IpReachabilityMonitor (and IpNeighborMonitor) are only accessed from
the IpManager StateMachine's thread. Consequently lots of locking can
now be removed.
Additionally:
- rename BlockingSocketReader to PacketReader
- incorporate IpReachabilityMonitor output in dump()
Test: as follows
- runtest frameworks-net passes
- "adb shell ip neigh change <address> dev wlan0 nud failed"
triggers wifi to disconnect
Bug: 62476366
Bug: 67013397
Change-Id: I18aca29ae0019a72a7e559c2832e0d9b0e33d81e
This patch introduces a helper ApfConfiguration class for passing filter
parameters to the constructor of ApfFilter. The benefits are:
- less parameter boilerplate across IpManager, ApfFilter, ApfTest
- easier to add more parameters in the future (useful for hardware
counters options and further gservices configurable options)
There is no functional changes otherwise.
Test: runtest frameworks-net
Change-Id: I47231c21df80bc36b6d4e15369748017b50b8da1
Callers of IpClient can tell it via the Builder whether a stable or
a random link-layer identifier is used. IpClient will make the
appropriate IP-layer adjustments (for now this means RFC 7217
addresses or not).
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes
- ./frameworks/opt/net/wifi/tests/wifitests/runtests.sh runs as well
as possible on AOSP (no new errors)
Bug: 62476366
Change-Id: Ic9d52ea4f2c6c3588852ed41c81f8664d5cc298a
This change switches the signal used for provisioning rechecks from:
ACTION_SIM_STATE_CHANGED
to:
ACTION_CARRIER_CONFIG_CHANGED
Additionally:
- reexamine carrier config overrides during provisioning re-checks
(this just was never happening before).
- refactor shared code out to VersionedBroadcastListener
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net
Bug: 63400667
Bug: 67755969
Change-Id: Ib9d222eb7ca0e0dd988a1bd97ab32059189ada2c
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes
- both wifi and ethernet connectivity works
Bug: 62476366
Change-Id: I3b56e75d545594b88849a621e4ae09f2e1d45c69
This is a pure refactoring with no a behavior change other than
that these calls now go through android.system.Os, which immediately
delegates to Libcore.os.
This is a first step towards separating framework (via
android.system.Os) vs. libcore (via Libcore.os) clients of these
low level APIs. Separating these is a prerequisite towards moving
the API parts of android.system into framework, and moving the
rest into a different package in libcore.
Test: Treehugger
Bug: 67901714
Change-Id: Ifd8349ec5416e5693f40dba48fdf2bef651b7d81
Merged-In: Ifd8349ec5416e5693f40dba48fdf2bef651b7d81
Test: Pulled out of a client app ; the client app is behaving
Test: identically to how it was behaving before. Wrote unit
Test: tests in frameworks-net.
Change-Id: I397137748a95c65cee2e1e1b243a1a260e83a6f7
Convert the simplistic blocking read in a separate thread model to
the MessageQueue OnFileDescriptorEventListener model, albeit still
on a separate thread.
Test: as follows
- built
- flashed
- booted
- "runtest frameworks-net" passes
- basic IpManager functions still work,
including ConnectivityPacketTracker
Bug: 62476366
Bug: 67013397
Change-Id: I1f4a6707eba402338947fe3f5392a26660f05714
Test: as follows
- built
- flashed
- booted
- runtest frameworks-net passes
(except for UidRange and APF tests that are missing libvndk)
- regular WiFi STA mode attach works for IPv4-only, dualstack,
and IPv6-only networks
- tethering works with dualstack upstream and downstream
Bug: 62476366
Change-Id: Id807b3091b0b2935b01aeb3644245034312b4183
Add a configurable black for protocols which packets are to be dropped
by APF. This enales an OEM to configure the APF to filter out un-needed
packets from reaching the host
Bug: 62415182
Test: runtest frameworks-net
Merged-In: I86335a0f854d5e83a2b2767978cd69b2cc25c2f8
Signed-off-by: Ahmed ElArabawy <arabawy@google.com>
(cherry picked from commit caffbe1277)
Change-Id: I15a758d18464d73d24ba8364a867904bb3cd3b34
This patch changes IpManager to take into account static provisioning
information specified in the InitialConfiguration for IPv6 static
configuration.
When a valid InitialConfiguration with IPv6 content is specified,
IpManager will do the following things:
- at start(), it will push the IPv6 addresses in the config to netd
- it will observe all addresses be notified via Netlink
- when all addresses are there, it will patch in the associated IPv6
routes in the config, so that they get passed to ConnectivityService
through the usual mechanism
The logic triggering onProvisioningSuccess is also changed to take into
account InitialConfiguration: when all addresses and all routes in the
config are seen the provisioning is successful.
Bug: 62988545
Test: runtest frameworks-net, with newly added tests
Change-Id: I77ed7c576c4b198de7a4726be70c78b74689e98b
in b/37702435 battery was drained due to reception of
Ethernet frames with Eth-type < 0x600 (non ehternet II frames)
This change adds an XML file to drop/allow these frames
Dropping helps tp avoid acquiring a wakelock for the
reception of these packets, while allowing them may be needed
for some protocols
The Change allow the OEM to override the system settings
Bug: 37702435
Test: runtest frameworks-net
Merged-In: Idb6c215896ef230a8136ff4ecefc90ccdc5c1c6d
Merged-In: I16346bb98a0b9259f7c3f51e318c472cfc1bb492
Signed-off-by: Ahmed ElArabawy <arabawy@google.com>
(cherry pick from commit 4dd7c58fc8)
Change-Id: Ie466baf1e6395d6c28245e41179232ea8e523845
