Introduce a 3-second delay in each packet send.
Before sending each packet, wait for 3 seconds to give the
firmware time to turn on the APF.
The firmware may have throughput calculation logic that
depends on a certain time window.
Flag: EXEMPT test only
Bug: 434614102
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Test: ./cts-tradefed run commandAndExit cts \
-m CtsConnectivityMultiDevicesTestCases \
-t ApfV6Test --shard-count 2
Change-Id: I0807562cd87cbfe90c90d48cf2d0e41673c153b0
Improve the APF test by changing the packet count assertion from
`== 1` to `> matched_pkt_count_before_test`.
This handles scenarios where the packet counter might already
have a non-zero value or increment due to other traffic, making
the test more reliable. The test now asserts that the specific
test packet caused the counter to increase.
Flag: EXEMPT test only
Bug: 419427388
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Change-Id: Id3c0de786e674214948e7030519fe25383c9958c
The test was observed to be flaky when sending a single packet.
Sending three packets increases the likelihood of the test's
assertion passing reliably, addressing potential timing issues
or packet loss.
Flag: EXEMPT test only
Bug: 419427388
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test --iteration 10
Change-Id: I42426d49a81871e6c8e275b1bd8cb0964290ce17
The `expect_with_retry` utility expects the lambda to return a
boolean value. However, the previous implementation wrapped the
comparison result `apf_utils.get_matched_packet_counts(...) == 1`
inside a set `{...}`.
This caused the lambda to return a set (e.g., `{True}` or `{False}`),
which is truthy if non-empty, but not the actual boolean result.
This commit removes the unnecessary set wrapper, ensuring the lambda
directly returns the boolean outcome of the packet count comparison
Flag: EXEMPT test only
Bug: 419427388
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Change-Id: Idd77d905e5be7933b3213df5cb772773286becfa
Adds `retry_interval_sec` to the packet count assertion in `ApfTestBase`.
This addresses intermittent test flakiness by allowing more time for the
server device to capture offloaded packets.
Flag: EXEMPT test only
Bug: 419427388
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test --iteration 10
Change-Id: Ia0ac83a7054c6a606f300546306fef1b8fccd701
Previously, IPv4 and IPv6 addresses were retrieved in `ApfTestBase.setup_class`,
which meant all APF tests assumed the presence of both.
This change refactors `ApfTestBase` to introduce dedicated methods:
- `get_and_expect_ipv4_addresses_exist()`
- `get_and_expect_ipv6_addresses_exist()`
These methods now handle the retrieval and assertion of IP addresses.
Individual test cases in `ApfV6Test` are updated to call these methods
only when the respective address type is required for the specific test
scenario.
This improves test robustness by ensuring tests only depend on the
necessary IP addresses, allowing for more precise test setup and failure
reporting. Additionally, the doze mode activation is reordered to occur
after initial network setup.
Flag: EXEMPT mainline
Bug: 419427388
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Change-Id: I340bcb77fe66cc9927dcd58e5899cdf9a4375daa
This commit introduces a verification step that runs before APF
offload tests.
It uses the `cmd network_stack apf <iface> config` shell command
to dump the list of enabled offloads. By checking this list first,
we can ensure that the specific offload under test is actually
enabled, avoiding misleading failures and improving test reliability.
Flag: EXEMPT mainline
Bug: 419427388
Test: ./cts-tradefed run commandAndExit cts -m CtsConnectivityMultiDevicesTestCases -t ApfV6Test --shard-count 2
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Change-Id: I65fecdcd4f2f9e9586da2684bf68cec215d0f359
Adds a new helper function `get_apf_config_from_cmd` to retrieve the
list of enabled APF offloads for a given network interface.
Flag: EXEMPT mainline
Bug: 419427388
Test: atest NetworkStaticLibHostPythonTests
Change-Id: I0acc9fcbf2e8bc6f4d8f309c2ec64c003c325c42
In mobly, it will root the device if the device is rootable.
If there is some operation needs root permission,
the test result will be inconsistent between user and userdebug build.
Flag: EXEMPT mainline
Bug: 419423721
Test: atest CtsConnectivityMultiDevicesTestCases
Change-Id: Ib962b565aeaf8f0e1737b700a16420277e088e6b
Use the following commands to apply pyformat
pyformat -i -r --force_quote_type single staticlibs/
pyformat -i -r --force_quote_type single tests/
Flag: EXEMPT mainline
Bug: 419423721
Test: TH
Change-Id: I0d3914ec652664ed4ec5d44afc25d97ff8bb4626
If client knows network name and passphrase, it can connect to the
server (Group owner) without user interaction. In this case, it needs
to use interface address of group owner, not device address
Bug: 408245757
Test: atest CtsConnectivityMultiDevicesTestCases
Flag: EXEMPT bug fix
Change-Id: I47bc8609e63564f4226a4ef71ea9e382f5f919fe
Signed-off-by: Rohit Kothari <r.kothari@samsung.corp-partner.google.com>
Rohit Kothariauthored and Paul Hucommited 2025-05-07 02:47:38 -07:00
Network name for Wi-Fi Direct group must starts with the prefix DIRECT-xy.
But CTS sometimes makes name with non-character in the position of “xy”.
To prevent issue, network name needs to be generated with the prefix DIRECT-xy.
Bug: 408245757
Test: atest CtsConnectivityMultiDevicesTestCases
Flag: EXEMPT bug fix
Change-Id: Ida07616b920b32dadd76f732ff2d87cc6ecc960d
Signed-off-by: Rohit Kothari <r.kothari@samsung.corp-partner.google.com>
Rohit Kothariauthored and Paul Hucommited 2025-05-07 00:34:24 -07:00
This commit removes the test_igmpv3_general_query_offload() test
logic. This test fails on user builds because they run on non-rootable
devices, preventing the necessary test operations.
Additionally, the APF counter check logic is removed as it was causing
test flakiness and instability.
Bug: 406570694
Test: TH
Flag: EXEMPT test only change
Change-Id: I9f2f32c49737721e537d578d6df41d4062d0684f
This commit enhances the stability of the APFv{4,6} multi-device tests by addressing several sources of flakiness and improving the test setup and validation.
Specifically, this commit:
* Increased sleep time in the test setup. This ensures APF is activated before each test, preventing unexpected behavior.
* Added sleep time in `ApfV4Test` to ensure APF remains enabled and is not disabled due to high throughput scenarios.
* Changed the validation order to first check received packets and then APF counters. This helps to more easily identify the root cause of test failures.
* Changed the Ethernet destination address from multicast to unicast in `ApfV6Test` to prevent packets from being filtered due to DTIM skip.
* Removed the `test_broadcast_arp_request_offload()` test, which was flaky due to DTIM skip and is adequately covered by the `test_unicast_arp_request_offload()` test.
Bug: 406570694
Flag: TEST_ONLY
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV4Test
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV6Test
Test: run cts -m CtsConnectivityMultiDevicesTestCases --skip-preconditions -t ApfV4Test --shard-count 2
Test: run cts -m CtsConnectivityMultiDevicesTestCases --skip-preconditions -t ApfV6Test --shard-count 2
Change-Id: I053509459078d0bd94e16a6bf7349b948f930e6c
Jimi Chenauthored and Yuyang Huangcommited 2025-04-02 04:31:01 -07:00
This reverts commit 372a15220a.
Reason for revert: this is Android B+ CTS and shouldn't be run with previous release, this can help to catch human error.
Change-Id: I867c63511d4f640e07b8317506291de284efcef3
CTS tests screen-on device in the beginning, but APF is not activated after screen-off in the test setup.
This change adds 3 secs delay to ensure APF activation before test execution.
Bug: 402667362
Test: ./cts-tradefed run cts-on-gsi -m CtsConnectivityMultiDevicesTestCases -t ApfV4Test --shard-count 2
Test: atest CtsConnectivityMultiDevicesTestCases:ApfV4Test
Change-Id: I067ce792b5f89e2f0051933982ebc734607fbf98
The 'ip link show' requires root permissions, causing errors.
This change replaces 'ip link show' with 'ip link'.
Bug: 401123634
Test: atest NetworkStaticLibHostPythonTests
Change-Id: I82b219ae9179f23e3233f797aeb4aa3740e6d3ec
Currently, upstream network validation is always required, even
for type None. This is unnecessary for APF and mDNS tests, as
they can be executed on local networks. Additionally, some
devices lack the ability to perform network validation when the
hotspot is enabled due to hardware limitations. Therefore, remove
the validation check for upstream type None.
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases:ConnectivityMultiDevicesTest
Bug: 335368434
Bug: 396551848
Change-Id: If2a96cf9d62e37fc27be6193f9ce9b9cfc8613df
Junyu Laiauthored and Paul Hucommited 2025-02-17 21:22:19 -08:00
Some tests may disable wifi when testing, for example,
tests which setup cellular default network.
Reconnect to wifi to prevent from flaky.
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases:ConnectivityMultiDevicesTest
Bug: N/A
Change-Id: Ic0e27efc0e82698618c651c0a78b000c0ffe2131
The `send_packet_and_expect_reply_received` start capurting packets,
and then send out request packets, then get expected packets counts to validate.
In the end, stop capturing for the specified interface.
Test: m connectivity_multi_devices_snippet && atest CtsConnectivityMultiDevicesTestCases
Change-Id: I275534a45ce0ed363fa8ab9d12272d25845aa598
This commit introduces checks for APF version, VSR API level,
and Android version to decide whether an APFv4 test should be
run, skipped, or fail:
* If Android version is less than V, skip the test
* If VSR API level is greater than or equal to 14 and APF
version is less than 4, fail the test
* Otherwise, if APF version is less than 4, skip the test
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases:ConnectivityMultiDevicesTest
Fix: 352132428
Change-Id: I023164fb59ad40f489abea2e0c45761efc39699b
This is a no-op refactoring, which is preparation for follow-up
changes to add more apf tests.
This includes:
1. Add a method to send packet and expect counter increases.
2. Inline the packet construction code in the test method.
Test: atest NetworkStaticLibHostPythonTests
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases
Bug: 350880989
Change-Id: Iadcfd97740033306af1b6a474004cbf0699ddd1b
This is a no-op refactoring, which is preparation for follow-up
changes to add more apf test classes.
This includes:
1. Move initialization code to a base class, which setup a hotspot
and a client connects to it, and enable doze mode to
activate the apf.
2. Check whether send raw packet is supported before testing.
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases
Test: atest NetworkStaticLibHostPythonTests
Bug: 350880989
Change-Id: I6376935883653182728e473e2f5533407d993d7b
This is a no-op refactoring, which is preparation for follow-up
changes to add more apfv4 tests.
This includes:
1. Add a entry point for multi-device tests to collect
test cases from different files.
2. Move multi-device initialization code to a base class.
3. Make different test classes inherit the base class.
Test: m connectivity_multi_devices_snippet && \
atest CtsConnectivityMultiDevicesTestCases
Bug: 350880989
Change-Id: I294c85d8e09633bbf02ba1543b9f2ea3e988e20d
Jimi Chen
commited 
Paul Hu
authored and
Android (Google) Code Review
commited 
Rohit Kothari
authored and
Yuyang Huang
commited 
Gerrit Code Review
commited 
Treehugger Robot
authored and
Maciej Żenczykowski
authored and
Junyu Lai
authored and