corstone500 was the only user of this and now uses 5.4, so this can be
removed.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Upgrade from 5.3.18 to to 5.4, a maintained stable release.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
trusted-services and psa-arch-tests both use FetchContent to download
more source during configure, and they can't easily be seeded with the
correct sources in advance (unlike TF-M).
Until this is fixed upstream, allow network use during do_configure.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
TF-M for corstone1000 uses libmetal and open-amp, downloaded during
configure by FetchContent. This is bad form, so add these sources and
license statements to the recipe so the fetch doesn't need to happen.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
TF-M uses FetchContent to download further sources at configure time, but
this is bad form as the sources and licenses are not tracked or archived.
This should now be blocked by the network isolation but as this doesn't
work in some environments (such as Docker containers) we can tell cmake
to never fetch to be sure that all of the sources used are declared in
SRC_URI.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Consolidate the CS1K kas files into a common file, and respect the
platform policy by setting the DISTRO to poky-tiny. Also add the
CS1K-specific recipes psa-arch-tests and ffa-debugfs-mod to the build.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Instead of adding perf to the target: list, add it to
CORE_IMAGE_EXTRA_INSTALL. This means that machine configurations which
need to change the target don't need to also build perf explicitly.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
It was used by SGI575, but it is now tracking the latest
release in oe-core.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
This recipe simply repacks existing artefacts, so there's no need to
depend on the toolchain.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The default kernel is now 5.15, so remove the preferred version
assignment.
The mailbox patches went upstream in a different form, so we can drop
the patches and drop in a customised Devicetree file required to use the
new drivers.
In the future we can possibly drop the devicetree patch if it is provided
by firmware, but for now this is known to work.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Instead of pinning the kernel to 5.7, use the latest default version.
There are no patches, so this shouldn't be a problem.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
KMACHINE defaults to MACHINE, so remove the redundant assignments.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The CS1K kernel used overrides to empty the kernel-image package, so
the initramfs didn't contain a kernel image.
The initramfs contained a kernel via a bad dependency in ffa-debugfs-mod,
but now that has been worked around we can remove the CS1K workaround.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
CS1K wants to put the ffa-debugfs kernel module into the initramfs, but
this pulls in the kernel image too because kernel modules depend on the
kernel image (via kernel-module-split.bbclass). Arguably this should be
controllable somehow, as it's not mandatory to have an image on the same
file system as the modules: initramfs or separate /boot partitions being
notable cases.
Until this is resolved upstream[1] we can work around it by removing the
RDEPENDS on the kernel from the kernel module. As the package is
generated during do_package this can't be a simple RDEPENDS:remove, but
has to be another package split function which runs after the module
splitting function.
[1] https://lore.kernel.org/openembedded-core/20220209173036.3823144-1-ross.burton@arm.com/T/#t
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
There's no need for ffa-debugfs-mod to RPROVIDE
kernel-module-arm-ffa-user, as this is provided by a sub-package of this
recipe already.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Move TC to an intermediate SHA based on 2.6. This SHA includes
patches for:
* Support for 8GB DRAM
* Support for GPU
* Removal of kernel command-line (not needed, kept in U-Boot)
Also removed patch that fixes build problems as the problem has
been solved in the updated SHA.
Signed-off-by: Anders Dellien <anders.dellien@arm.com>
Change-Id: Ic8c4ef35e8e7ed25711ff0813abab6b6c8564ef3
Signed-off-by: Jon Mason <jon.mason@arm.com>
Update the zephyr testcases to be the ones that build and pass tests
(where testing possible). Also, add comment to qemu-cortex-a53 about
not passing tests.
Signed-off-by: Jon Mason <jon.mason@arm.com>
When multiple threads are allowed xtest will fail and
leave the system in a bad state after repeated runs.
Signed-off-by: Ben Horgan <ben.horgan@arm.com>
Change-Id: I16f07df1a362540560975deaa5a291a68c332bfb
Signed-off-by: Jon Mason <jon.mason@arm.com>
Enable coresight components in the config and also port the patches
which were upstreamed in later version of the linux kernel.
Change-Id: I27983abd5f2945328f7465cc1b2af4f8e848b69b
Signed-off-by: Davidson K <davidson.kumaresan@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
If we don't specify a tag name GitLab uses the 'latest' tag, which for
Kas is moved whenever an image build is made.
Instead explicitly use the latest-release tag, which is only updated
when a release is made.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
imgtool from mcuboot uses python3-cryptography-native, and the latest
python3-cryptography explicitly loads the legacy provider, which is a
separate shared object in OpenSSL 3. The search path for providers is
hard-coded into the library so the wrong path is searched and the module
is not found.
Set OPENSSL_MODULES so the right path, so that the legacy module is
found. In the future we may be able to be removed this if the explict
use of legacy algorithms is removed
(https://github.com/pyca/cryptography/issues/6809).
This also means we can remove the downgrades of python3-crytography that
were being carried in meta-arm.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Enable memory-resident bitbake by setting the server timeout to 60s.
For most builds this doesn't have an impact as bitbake is ran once, but
this reduces the time for the pending-upgrades report from 10 minutes to
3 minutes, as it doesn't need to constantly restart bitbake.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add links from the patch names to the patches in cgit for ease of
inspection.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
As this can get called lots but the data doesn't change, cache
the responses. Also return a pathlib.Path, as this is 2022.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Instead of simply adding a tag saying "Patched", include the patch count
and make a judgement call on whether the patches are "safe" or not. For
example, patches which are entirely backports or inappropriate are
"safe", pending and denied patches are not.
Signed-off-by: Ross Burton <ross.burton@arm.com>
yocto-check-layers is very greedy when searching for layers, and will
find the test layers in Bitbake if given the chance, for example:
bitbake/lib/layerindexlib/tests/testdata/layer4
This layer has the collection name openembedded-layer and is only
compatible with Sumo. The selection of layer from collection name is
not deterministic, so it's possible that this layer is selected which
then fails the check as it isn't compatible.
Solve this by restricting the dependency layers in meta-arm to just
meta-arm/meta-*, so it doesn't recurse into meta-arm/work/poky/bitbake.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
As of meta-zephyr dde88ba the layer is structured different, split into
meta-zephyr-core and meta-zephyr-bsp.
As we define our own machines for use with Zephyr, we can just use
meta-zephyr-core.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Don't use SRCREV in the PV, but use SRCPV as this truncates the SHA.
Also bump to 1.2+git, as the psa-arch-tests project does tag releases.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
corstone1000 currently has a git snapshot of optee-os, so that the
version number is managable use SRCPV instead of SRCREV.
Also fix whitespace in SRC_URI.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
corstone1000 currently has a git snapshot of TF-M, so that the version
number is managable use SRCPV instead of SRCREV.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
It's not recommended to use ${S} in SRC_URI as S contains references
to the version which results in a circular list of variable lookups.
destsuffix is relative to WORKDIR and defaults to git/, so use relative
paths with the same result.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Corstone 700 is end-of-life, so remove it from meta-arm master.
It will remain in the stable branches for existing users, but new users
are encouraged to use Corstone 1000 instead.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Since 10e60cc the terminal_map doesn't exist, this piece of code wasn't
updated.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
If wait_for_login() times out then it raises an exception instead of
passing back return values, so the bootlog is never assigned.
We always want a boot log, so initialise it outside of wait_for_login and
pass it in.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Instead of patching a file we patched, integrate the changes directly
into the first patch.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Ross Burton
Anders Dellien
Jon Mason
Ben Horgan
Davidson K
Arunachalam Ganapathy