Skip to content

AGENTS.md

Latest commit

 

History

History
157 lines (118 loc) · 5.44 KB

AGENTS.md

File metadata and controls

157 lines (118 loc) · 5.44 KB
Raw

Netdev CI test instructions

This repository (NIPA) implements the netdev CI. Two distinct kinds of tests are relevant when working here:

  1. Netdev CI checks — the checks NIPA runs against kernel patches (builds, checkpatch, selftests, KUnit, etc.), which can be reproduced locally against a $linux tree (see the rest of this document).
  2. NIPA's own test suite — unit tests and lint checks for this repository's Python code (see below).

Netdev CI checks (kernel patches)

Use these instructions to reproduce the tests run by the netdev CI (NIPA) before submitting kernel patches whenever possible. The CI runs:

  • LLM reviews for each patch (Sashiko and Clashiko).
  • Static checks for each patch, including builds, documentation checks, checkpatch.pl, and Python, shell, and YAML checks.
  • Runtime tests in VMs, including kselftests and KUnit.

LLM reviews

Sashiko can be run locally with different models. Follow the Sashiko installation instructions: configure the model, then run:

sashiko init
sashiko review

Clashiko uses Sashiko with multiple models for cross-reviews. It is available at https://netdev-ai.bots.linux.dev/sashiko.

Static tests

Static tests are implemented by scripts in the tests directory. Run them against a patch series with ingest_mdir.py:

  1. Export the patches, for example with b4 send -o (...) or git format-patch -o (...).
  2. Check out a branch based on net-next (or net) in the Linux tree.
  3. Run the ingestion script from this repository.

Example:

## Export patches
$ cd $linux                          # your kernel source dir
$ b4 prep --set-prefixes 'net-next'  # or net
$ b4 send -o /tmp/my-series

## Create a branch on top of net-next (or net)
$ git switch -c test <base-branch>  # e.g. net-next or net

## Execute the tests; this can take a while
$ cd $nipa                           # the source dir of this repo
$ ./docker/build/run.sh --pull ./ingest_mdir.py \
        --mdir /tmp/my-series --tree $linux --result-dir out

Images can be built locally with ./docker/build.sh. When using ./docker/build/run.sh, --pull downloads the latest image. Docker is shown above, but Podman can be used instead.

Also, tests can be disabled with ./ingest_mdir.py --disable-test (...) depending on the modified code, e.g. no need to build the kernel when only modifying the selftests. Check ./ingest_mdir.py --list-tests.

KUnit

See the KUnit instructions. Run all KUnit tests on x86_64 directly in the Linux tree from the selftests container:

$ cd $linux
$ $nipa/docker/selftests/run.sh --pull \
        ./tools/testing/kunit/kunit.py run --alltests --arch=x86_64

Net selftests

The CI runs selftests in the latest Fedora environment inside lightweight VMs using virtme-ng. The recommended local workflow is to run vng, make, and qemu commands through the nipa-selftests container:

$ cd $linux
$ alias run="${nipa}/docker/selftests/run.sh"
$ docker pull ghcr.io/linux-netdev/nipa-selftests:latest

$ TARGETS=(net net/af_unix net/forwarding net/hsr net/mptcp net/netfilter \
    net/openvswitch net/ovpn net/packetdrill net/ppp net/rds net/tcp_ao nci)
$ for target in "${TARGETS[@]}"; do
    rm -f .config
    run vng --build --force --config "tools/testing/selftests/${target}/config"  # add '--config kernel/configs/debug.config' to validate with a debug kernel
    run make headers
    run make -C tools/testing/selftests TARGETS="${target}"
    run sudo vng --run . --user root -a mitigations=off --rw \
        --cpus 4 --network loop -- \
        make -C tools/testing/selftests TARGETS="${target}" run_tests
        # the output is TAP parsable to find failed and skipped tests
  done

Adapt TARGETS to the modified code. Driver-related targets can also be tested:

drivers/net drivers/net/bonding drivers/net/hw drivers/net/netconsole
drivers/net/netdevsim drivers/net/team drivers/net/virtio_net

Check for:

  • failed tests: not ok (...)
  • skipped ones: ok (...) # SKIP
  • any type of error in the tests
  • or reported by the kernel: Call Trace:, timeout, unexpected stop
  • with and without a debug kernel config, adding this to the vng --build command: --config kernel/configs/debug.config
  • with a debug kernel config, check for leaks: scan (echo scan > /sys/kernel/debug/kmemleak), wait 5 seconds, scan again, check for leaks (cat /sys/kernel/debug/kmemleak)

For additional details and troubleshooting tips, see the netdev selftests CI guide.

NIPA's own tests

Run before committing changes to this repository's Python code.

Unit tests (unittest-based, no pytest dependency required):

$ python3 -m unittest discover -s contest/hw/tests -p "test_*.py"
$ python3 -m unittest discover -s ml-agent/tests -p "test_*.py"

Syntax check used by .github/workflows/lint.yml (runs on every changed .py file):

$ python3 -m py_compile $(git ls-files '*.py')

This repository also follows the style defined in .style.yapf (not currently enforced in CI, but recommended before submitting changes):

$ pip install yapf
$ yapf --diff --style .style.yapf <changed_file>.py