• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1Docker CI
2=========
3
4For llvmpipe and swrast CI, we run tests in a container containing
5VK-GL-CTS, on the shared GitLab runners provided by `freedesktop
6<http://freedesktop.org>`_
7
8Software architecture
9---------------------
10
11The Docker containers are rebuilt from the debian-install.sh script
12when DEBIAN\_TAG is changed in .gitlab-ci.yml, and
13debian-test-install.sh when DEBIAN\_ARM64\_TAG is changed in
14.gitlab-ci.yml.  The resulting images are around 500MB, and are
15expected to change approximately weekly (though an individual
16developer working on them may produce many more images while trying to
17come up with a working MR!).
18
19gitlab-runner is a client that polls gitlab.freedesktop.org for
20available jobs, with no inbound networking requirements.  Jobs can
21have tags, so we can have DUT-specific jobs that only run on runners
22with that tag marked in the GitLab UI.
23
24Since dEQP takes a long time to run, we mark the job as "parallel" at
25some level, which spawns multiple jobs from one definition, and then
26deqp-runner.sh takes the corresponding fraction of the test list for
27that job.
28
29To reduce dEQP runtime (or avoid tests with unreliable results), a
30deqp-runner.sh invocation can provide a list of tests to skip.  If
31your driver is not yet conformant, you can pass a list of expected
32failures, and the job will only fail on tests that aren't listed (look
33at the job's log for which specific tests failed).
34
35DUT requirements
36----------------
37
38In addition to the general :ref:`CI-farm-expectations`, using
39Docker requires:
40
41* DUTs must have a stable kernel and GPU reset (if applicable).
42
43If the system goes down during a test run, that job will eventually
44time out and fail (default 1 hour).  However, if the kernel can't
45reliably reset the GPU on failure, bugs in one MR may leak into
46spurious failures in another MR.  This would be an unacceptable impact
47on Mesa developers working on other drivers.
48
49* DUTs must be able to run Docker
50
51The Mesa gitlab-runner based test architecture is built around Docker,
52so that we can cache the Debian package installation and CTS build
53step across multiple test runs.  Since the images are large and change
54approximately weekly, the DUTs also need to be running some script to
55prune stale Docker images periodically in order to not run out of disk
56space as we rev those containers (perhaps `this script
57<https://gitlab.com/gitlab-org/gitlab-runner/issues/2980#note_169233611>`_).
58
59Note that Docker doesn't allow containers to be stored on NFS, and
60doesn't allow multiple Docker daemons to interact with the same
61network block device, so you will probably need some sort of physical
62storage on your DUTs.
63
64* DUTs must be public
65
66By including your device in .gitlab-ci.yml, you're effectively letting
67anyone on the internet run code on your device.  Docker containers may
68provide some limited protection, but how much you trust that and what
69you do to mitigate hostile access is up to you.
70
71* DUTs must expose the dri device nodes to the containers.
72
73Obviously, to get access to the HW, we need to pass the render node
74through.  This is done by adding ``devices = ["/dev/dri"]`` to the
75``runners.docker`` section of /etc/gitlab-runner/config.toml.
76