(this is required to make ipv6 tethering work with at least S.LSI modems)
Test: TreeHugger
Bug: 207057951
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: Ic178db928ec7f74f69d7d4739b3b8439ff026625
Note: due to the release version of the Connectivity/Tethering mainline
module being built from sc-mainline-prod, this won't actually take effect
until system/bpf bpfloader at version 0.6+ is merged in to that tree.
This doesn't really matter, since currently things default to v0.0+.
But there is no mainline module updatable pre-v0.2 supported OS
anyway. BpfLoader v0.2 is what shipped in Android S Beta 4 through
Android S Final.
Before S there simply was no bpfloader support for mainline updatable
ebpf code, while S Beta 3 and earlier shipped v0.0 which is badly
incompatible with even the current version of the mainline module.
Additionally v0.0 doesn't even parse this field, while v0.1 which
does was very short lived [~3 days] and can thus be utterly ignored.
As such this change is effectively a no-op, and even post merge
of bpfloader v0.6+ into sc-mainline-prod will still be effectively
a no-op.
So why do it? I want to explicitly document that these programs are S+,
so that I can change the default in the future to be T+.
Test: TreeHugger
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I7e5d0124700c7045abe16b1f3b504c9e88054ff2
Move tethering util files from android.net.util into
com.android.networkstack.tethering.util. The goal is move all of
tethering internal files into its own namespace
com.android.networkstack.tethering.util.
Bug: 205088391
Test: atest TetheringTests
atest CtsTetheringTest
atest TetheringPrivilegedTests
Change-Id: I6559fb4f873b3cad5b210b10e49df1b6c6914a70
With the network selection rewrite in S, rematching a single request can
now easily be done; this can be used as an optimization in
handleRegisterNetworkRequests to avoid rematching all requests when
registering a new one.
This can be disabled by a flag that is unset by default,
REMATCH_ALL_REQUESTS_ON_REGISTER.
Test: atest ConnectivityServiceTest
Change-Id: If76f79b41ac88863974f7025624667134bea2570
Define interfaces that match the signature of the existing
EthernetManager.TetheredInterfaceRequest and TetheredInterfaceCallback
classes and make EthernetManager.TetheredInterfaceRequest and
TetheredInterfaceCallback implement/subinterface these interfaces. The
new bluetooth API could also implement these interfaces to make API surface
consistent.
Test: TH would test the existing tests that use the subclass.
Bug: 190438212
Change-Id: I093972c111cb1d921076782492716d5a046be8fc
To avoid much code diverge between aosp and its downstream, we need to
separate the build target which build with latest stable sdk in release
branch and build with unfreeze sdk in dev branch. So we could use new
API with APIshim in dev branch without impacting release branch.
BYPASS_INCLUSIVE_LANGUAGE_REASON=the wording is the file name of
existing file.
Bug: 204265450
Test: m
Change-Id: I949fd83b1eed7461e0b381193b190cf15080bc1b
After separat BpfMap to a library, its package name is jarjar to
com.android.network.stack.util*. Also jarjar BpfMap of
TetheringPrivilegedTests to fix crash in jni registered.
Bug: 179733303
Test: atest TetheringPrivilegedTests
Change-Id: Ia9bc819a44fe26bf2ff7a46351ea60f2eeb02933
Now that S has gone to AOSP, this dependency is no longer
needed. It is, on the contrary, harmful as it prevents us
from adding shims that depend on T APIs in AOSP.
Test: builds
Change-Id: I64f0f394e826e58a9b7e772043bb6008428892e6
There is pollingConnector thread which start polling connector if
TetheringManager is created earlier than TetheringService started(during
device boot up). TetheringManager won't be GCed if pollingConnector
thread do not finish its task yet.
Bug: 177265744
Bug: 191798390
Bug: 187972579
Test: atest TetheringServiceTest
Change-Id: Id8c7d10c5172e1d5de460c5311ff9c20261facef
TetheringCallbackInteranl is inner class which explicitly reference
TetheringManager object. This causes TetheringManager can't be GC. Using
static nested class which has its own lifecycle and weak reference
TetheringManager object.
Still have a leak inside Tethering that TetheringCallbackInternal is
never unregistered. Currently it rely on binder died to remove the
reference, which usually happen in kill process. If process keep alive,
the TetheringCallbackInternal would not be freed even TetheringManager is
gone. Will have follow CL to fix this.
Bug: 177265744
Bug: 191798390
Bug: 187972579
Test: 1. lunch Settings with ON/OFF tethering, dump java heap.
2. close Settings and restart Settings again, dump java heap.
3. Compare java heap between step 1 and step 2.
Change-Id: I0e2a21b7988115098a033a581cd98da8bffe2791
StateMachine was in a custom filegroup in base.
It's now built in stand-alone library in modules-utils.
Bug: 198418216
Tag: #refactor
Test: Build
Merged-In: I7499fad6c4c5076e2bd98f0d9f91c5f243fb1ed2
Change-Id: I7499fad6c4c5076e2bd98f0d9f91c5f243fb1ed2
TetheringService call different Settings API to check write settings
permission between R and S.
Bug: 182211575
Test: install S module to R device and both run mts tests build with S
and R.
Change-Id: I3e7f6525e9776992bd96b17c132b749b12285bbd
Merged-In: I3e7f6525e9776992bd96b17c132b749b12285bbd
This is a no-op CL which add TetheringTester and move runDhcp related
logic to it.
Test: atest EthernetTetheringTest
Change-Id: Ib1c5647b2bd5a1b27c976450d3aa265aff8f5b70
- Rename the conntrack destination port check function
- Use CollectionUtils.contains to check the denied ports
- Add tests for the streams with ftp and pptp tcp port are
not able to be offloaded
Bug: 195914327
Test: atest TetheringCoverageTests
Change-Id: I7e2591bea1f6db46271efb0c30970fb8d4efe1e4
The correct return code to keep on processing any further TC
attached programs is 'TC_ACT_PIPE' and not 'TC_ACT_OK' (which
is terminal).
Without this the ipv6 tether offload program causes termination
of processing and the ipv6 clatd offload program never actually
handles any packets (while tethering is active).
This results in lack of bpf xlat64 offloading for tethered ipv4
traffic on an ipv6-only (cellular) network.
This in turn means incoming TCP packets get GRO'ed, do not get
bpf offloaded, and get delivered to the clat daemon, which
due to them being bigger than the mtu (due to gro) cannot
handle them and discards them.
This results in poor performance, since tcp falls back to 1 mss/mtu
sized packet per rtt.
Tested via tethering a linux laptop on an ipv6-only cellular connection
and downloading the linux kernel from kernel.org via 'wget -6' and 'wget -4'.
Before:
IPv6: over 2MB/s, observed:
5805 packets, including 4 sackOK
IPv4: under 1MB/s, observed:
9300 packets, including 8 sackOK, 387 sack 1, 501 sack 2, 2310 sack 3
After:
IPv6: over 7MB/s, observed:
16702 packets, including 4 sackOK
IPv4: over 9MB/s, observed:
32755 packets, including 2 sackOK
Test: builds, TreeHugger, see above
Bug: 195624908
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I623dacb5a37dc689cea34499c3906c11fcaf946c
This test fails on devices where physical Ethernet interfaces are
available but marked restricted, like cuttlefish.
Bug: 197462993
Test: test-only change
Change-Id: I15c991b2e43e2d5e823dcdcfbd74adfd9b2f6f08
