Add git recipe versions that track the latest git versions of u-boot and
the various OP-TEE recipes. This, in combination with the previously
existing trusted firmware a and m recipes, allows for using the latest
code in platform development and testing (as part of CI).
For CI usage, a KAS yml file has been created to allow for those recipes
to be used, and an entry for fvp-base has been added to the gitlab CI
yml file.
NOTE: the wildcard for corstone1000 u-boot PREFERRED_VERSION was causing
it to pick-up the newest version (and failing to apply the patches).
The wildcard is unnecessary, since it is using a layer supplied package.
So, remove it and everyone is happy.
Signed-off-by: Jon Mason <jon.mason@arm.com>
This job takes a few minutes and isn't useful unless it's being ran for
master, or is being actively worked on.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Reduce the number of tests being run in CI to reduce the amount of time
it takes to complete, while providing the same code coverage. Internal
CI runs went from 2.5h to 1.5h.
Signed-off-by: Jon Mason <jon.mason@arm.com>
We currently archive all of the task logs on every job, but this ends up
being between 2-10MB per job and our pipelines have ~100 jobs.
To save space and time, change the log collection to only happen if the
job fails, and explicitly expire them after a week.
This reverts meta-arm 60abe46, but in two years we've not really needed
successful logs, and they can be easily toggled back on if needed in a
branch.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Print all of the environment variables in the update-repos task for
introspection, instead of a subset.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Having local repo caches is a little fiddly to manage, and by definition
we're running CI inside GitLab which supports mirroring repositories
automatically.
As these mirrors are always available and update automatically, make
Kas reference directories opt-in and instead expect that the site is
either fine with full fetches, or is using KAS_PREMIRRORS.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The update-repos job can "fail with warnings" if the reference repository
fetch fails. This is intentionally a warning as the CI may have set
KAS_PREMIRRORS and a stale cache is fine.
However, by default artifacts are only saved on successful jobs, so if
this happens the lockfile.yml isn't saved. Ensure the artifacts are
always saved so the rest of the pipeline is successful.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add FVP support to sgi575 and run a boot test as part of CI. Networking
is not currently working and seems to require an older version of edk2
to boot the kernel. Also, the unique files for grub and wks do not seem
to be necessary.
Signed-off-by: Jon Mason <jon.mason@arm.com>
If update-repos fails with status 128 then that means it failed to fetch
the remote repositories. This should result in a warning not a failure
but flock was just returning status 1.
Save the exit code and if it returns 128 continue but exit with it
later, so the lockfile generation still occurs but the job doesn't fail.
Also, only call the update-repos script if KAS_REPO_REF_DIR has been set.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
This stops the job being stuck if the runners will only take jobs that
have been tagged.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
I accidentally created two variables sections, resulting in our build
jobs running on very limited containers.
Signed-off-by: Ross Burton <ross.burton@arm.com>
We were only setting the k8s CPU request in .build jobs not .setup. This
was intentional initially so that only the build jobs get more resources,
but some of the non-.build jobs are resource-heavy. For example, the
pending-updates job has to parse the entire metadata from scratch, and
that sometimes takes longer than usual when we only have two cores to
use.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Integrating the binary Arm GCC toolchain into OE is quite complicated
because the binary release and oe-core's toolchain are arranged slightly
differently, which makes it quite fragile.
As it's obviously a binary release we cannot patch it to fix issues.
Also it has some fairly sizable limitations: for example the kernel
headers are old (from linux 4.19) and the locale packaging is different
so locale package dependencies don't work.
The main historic users of the external toolchain no longer use it, so
remove it. The recipes will remain in the LTS branches for users who
are using it currently, but will not be part of the next release.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Acked-by: Romain Naour <romain.naour@smile.fr>
Acked-by: Sumit Garg <sumit.garg@linaro.org>
Acked-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
xen-image-minimal now requires systemd. Add poky-altcfg (which has
systemd amongst other things) as an includes in the xen.yml file to work
around this. Also, xen requires openssh instead of dropbear. So,
override that entry.
Signed-off-by: Jon Mason <jon.mason@arm.com>
qemuarm64-secureboot is using systemd for uefi-secureboot, which has
warnings with musl (and fails to compile with clang and musl). So,
modify the matrix to keep the coverage of everything else but musl.
Signed-off-by: Jon Mason <jon.mason@arm.com>
Encapsulate all UEFI Secure Boot required settings in one Kas
configuration file.
Introduce SBSIGN_KEYS_DIR variable where UEFI keys will be generated
to sign UEFI binaries.
Introduce uefi-secureboot machine feature, which is being used to
conditionally set the proper UEFI settings in recipes.
Replace Grub bootloader with systemd-boot, which it makes easier to
enable Secure Boot.
Advantages using systemd as Init Manager:
- Extending secure boot to userspace is a lot easier with systemd than
with sysvinit where custom scripts will need to be written for all use
cases.
- systemd supports dm-verity and TPM devices for encryption usecases out
of the box. Enabling them is a lot easier than writing custom scripts
for sysvinit.
- systemd also supports EUFI signing the UKI binaries which merge kernel,
command line and initrd which helps in bringing secure boot towards
rootfs.
- systemd offers a modular structure with unit files that are more
predictable and easier to manage than the complex and varied scripts
used by SysVinit. This modularity allows for better control and
customization of the boot process, which is beneficial in Secure Boot
environments.
- Add CI settings to build and test UEFI Secure Boot.
Add one test to verify Secure Boot using OE Testing infraestructure:
$ kas build ci/qemuarm64-secureboot.yml:ci/meta-secure-core.yml:ci/uefi-secureboot.yml:ci/testimage.yml
...
RESULTS - uefi_secureboot.UEFI_SB_TestSuite.test_uefi_secureboot: PASSED (0.62s)
...
SUMMARY:
core-image-base () - Ran 73 tests in 28.281s
core-image-base - OK - All required tests passed (successes=19, skipped=54, failures=0, errors=0)
Signed-off-by: Javier Tia <javier.tia@linaro.org>
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org> [yml file include fix]
Signed-off-by: Jon Mason <jon.mason@arm.com>
Using resulttool we can transform the oeqa JSON reports into JUnit XML,
which GitLab can display in pipelines and merge requests.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
We had two instances of the same job, so consolidate them into one.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Instead of always using KAS_WORK_DIR/build to refer to the build tree,
on the assumption that is where the build tree is, export KAS_BUILD_DIR
and use that variable instead.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add the changes necessary to get edk2 booting and testimage passing on
fvp-base. All that is really necessary is adding the dtb to the too
partition.
Signed-off-by: Jon Mason <jon.mason@arm.com>
The Kas 4.4 image includes the websocket module, needed to use the
public hashserv/sstate.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Added the Corstone-1000 FVP platform to the ACS test build as well as
adding the arm-systemready-firmware variant to the Corstone-1000 FVP
build.
Signed-off-by: Drew Reed <Drew.Reed@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
When trying to replicate a build locally, having the exact list of Kas
files that was used is very useful.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
These BSPs are now obsolete.
Users of generic-arm64 should use genericarm64 from meta-yocto-bsp.
Users of qemu-generic-arm64 should use sbsa-ref from meta-arm-bsp.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
This is mostly based on the existing qemu-generic-arm64 machine, but by
not being based on the genericarm64 and instead being specifically a
machine to run on the qemu sbsa-ref machine we get to tune differently.
Specifically, this configures sbsa-ref to be a Neoverse N2 (v9), and the
tune is set to match. Another notable difference to qemu-generic-arm64
is that the kernel configuration is at present defconfig. We may wish
to change this in the future to be the same fragmented configuration as
genericarm64.
We have to ignore two testimage parselogs failures: one from NUMA which
will be fixed in a future EDK2 release, and one from efifb where we
should be using the bochsdrm driver instead (further investigation is
needed)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add the new genericarm64 in meta-yocto-bsp to the CI.
This new BSP is heavily based on the meta-arm generic-arm64 machine, but
with an all-new fragmented kernel configuration.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add configuration settings to TF-A, OP-TEE and TS SPs needed to get TS
built and run on the fvp-base machine.
Signed-off-by: Gyorgy Szing <gyorgy.szing@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
To reduce build and test times in CI, move the dev kernel outside the
standard matrix. This results in it still being built and tested for
the platform, but only with gcc/glibc (and not against clang and musl).
This greatly reduces the number of permutations that need to be
verified.
Signed-off-by: Jon Mason <jon.mason@arm.com>
Setting an API key means we get higher rate limits. Because keys are
private, the key must be set in the environment of the runner.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Extend jobs-to-kas so the first argument is still the GitLab job name,
but allow further arguments to specify extra Kas files to use in
addition.
Then add a variable EXTRA_KAS_FILES to the CI configuration that
defaults to the empty string and pass this to jobs-to-kas.
This lets specific pipeline runs add extra Kas files, for example to use
experimental branches or enable extra features without touching the CI
directly.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add CI builds for Arm SystemReady Firmware within the fvp-base CI job and a new
Arm SystemReady IR ACS build job. Add the CI kas config for each of these
builds.
The ACS build can be controlled by the ACS_TEST GitLab variable to specify
whether or not to run the testimage. If this variable is not set, the
testimage step will not run. The job tag can be controlled by the ACS_TAG GitLab
variable.
Signed-off-by: Debbie Martin <Debbie.Martin@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Now that there's a FVP for aarch64, we don't need to pin the CI pipeline
to x86-64.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
To allow running the TF-A TFTF tests we need to ensure the images for
N1SDP and Corstone-1000 MPS3 boards build
Signed-off-by: Drew Reed <Drew.Reed@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Just execute all tests with the meta-arm tag, instead of hardcoding the
list of tests.
Also run two tests in parallel as there's no reason to limit it to one.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The Kas binary is identical, but the container has been rebuilt using
Debian 12 (Bookworm).
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The Kas container in version 4 onwards is based on Debian 12, which
forbids pip from installing files into /usr or ~/.local/.
We want to install the arbitrary dependencies for the documentation
build, so these should be installed in a venv.
The kas container doesn't currently install python3-venv, so we have to
install that manually (patch sent upstream).
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
