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>
This is in the Kas image from 3.2, so this is now obsolete.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Mostly taken from master, rationalise the variable definitions and stop
hardcoding /persist/.
Users of this CI will have to ensure that they've set CACHE_DIR if they
want anything actually cached.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
If the repository reference directory gets corrupted it's not easy to
wipe it, so add a variable CI_CLEAN_REPOS that if set in the pipeline
will clean the clones and re-fetch them.
Also, stop the fetch from detaching during the garbage collection, just
in case it was a long-running GC that got killed that caused the
corruption in the first place.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
When ccache is enabled trusted-firmware-a recipe fails with this
error message:
make: *** No rule to make target 'aarch64-poky-linux-gcc'. Stop.
ccache prefix CC variable with 'ccache' word before compiler. Because
there are no quotes assigned to CC, only 'ccache' is assigned. The
compiler becomes a make target, producing the build error.
Add single quotes to LD is a good measure to prevent this kind of error.
Signed-off-by: Javier Tia <javier.tia@linaro.org>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Some networks limit outgoing git: traffic, so use https:.
Fixes: 0cec3e5 ("arm/gem5/boot-wrapper-aarch64: Move main recipe to meta-arm")
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Fix the build with DISTRO_FEATURES containing "usrmerge":
make: *** No rule to make target '/.../optee-os/4.1.0/recipe-sysroot/lib/optee_armtz/bc50d971-d4c9-42c4-82cb-343fb7f37896.stripped.elf', needed by '/.../optee-os/4.1.0/build/core/early_ta_bc50d971-d4c9-42c4-82cb-343fb7f37896.c'. Stop.
Fixes: 6a105f47b9 ("optee-ftpm: Install artifacts into nonarch_base_libdir")
Signed-off-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
git.linaro.org/landing-teams/working no longer seems to serve snapshots
in tar.gz format, breaking the N1SDP build for kirkstone.
Using same tag as the existing snapshot, fetch the binaries with a
conventional git clone, as per master.
Signed-off-by: Adam Johnston <adam.johnston@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Linux 5.15.149 contains the following change[1]:
perf cs-etm: Bump minimum OpenCSD version to ensure a bugfix is present
Since commit d927ef5004ef ("perf cs-etm: Add exception level consistency
check"), the exception that was added to Perf will be triggered unless
the following bugfix from OpenCSD is present:
[...]
The opencsd requirement is now 1.2.1 onwards, but the kirkstone branch
has 1.2.0. The 1.2.1 release was just three bugfixes on top of 1.2.0 so
this should be a safe upgrade:
3fecfb0 opencsd: etm4x: Update etm4x / ete decoder to flush context to client immediately.
db97bbc opencsd: decode: Update decoder base to implement memory accessor invalidation.
a7b77aa opencsd: memaccess: Allow memory accessors to have cache invalidated by decoder.
[1] a7af1be5cf5507dd6c157b7a25453942f805db76
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
GCC 11.4 has improved code generation and needs the __aarch64_cas4_sync
intrinsic from libgcc, but one of the modules here doesn't link to
libgcc.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add basic support for running edk2 on qemuarm and qemuarm64. This
necessitated the need to add ACPI and EFI to the default kernel configs
for these machines.
Signed-off-by: Jon Mason <jon.mason@arm.com>
The poky commit 74f086529911bdaf07a8414d50de411e20739541 updates
the kernel to v5.15.103 which breaks the PCI quirk patch for the N1SDP.
Signed-off-by: Vikas Katariya <vikas.katariya@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
For some reason the kas 3.2.1 container fails:
No such file or directory: '/builds/engineering/yocto/meta-arm/ci/ci/base.yml'
Note the repeated /ci/, which is wrong.
Pin the kas container to 3.2 for now until this is resolved.
Signed-off-by: Ross Burton <ross.burton@arm.com>
A recent commit compressed the kernel image (to Image.gz) and
by default enabled an initramfs image. In the case for when
such that (initramfs) is not desirable, the deploy step of the
Juno firmware will still try to install the Image file, (not
Image.gz), so this fails:
ERROR: firmware-image-juno-1.0-r0 do_deploy: ExecutionError('/oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/temp/run.do_deploy.360477', 1, None, None)
ERROR: Logfile of failure stored in: /oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/temp/log.do_deploy.360477
Log data follows:
| DEBUG: Executing python function sstate_task_prefunc
| DEBUG: Python function sstate_task_prefunc finished
| DEBUG: Executing shell function do_deploy
| cp: cannot stat '/oe/build/tmp-glibc/deploy/images/juno/Image': No such file or directory
| WARNING: /oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/temp/run.do_deploy.360477:152 exit 1 from 'cp -L -f /oe/build/tmp-glibc/deploy/images/juno/Image /oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/image/juno-firmware-19.06/SOFTWARE/'
| WARNING: Backtrace (BB generated script):
| #1: do_deploy, /oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/temp/run.do_deploy.360477, line 152
| #2: main, /oe/build/tmp-glibc/work/juno-oe-linux/firmware-image-juno/1.0-r0/temp/run.do_deploy.360477, line 163
NOTE: recipe firmware-image-juno-1.0-r0: task do_deploy: Failed
ERROR: Task (../meta-arm/meta-arm-bsp/recipes-bsp/images/firmware-image-juno.bb:do_deploy) failed with exit code '1'
This updates the else case for when an initramfs image is not
in use so that the right kernel image is deployed, by using
the KERNEL_IMAGETYPE variable, to use either version of the
kernel image.
Signed-off-by: Daniel Díaz <daniel.diaz@linaro.org>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Change u-boot and machine config to default to booting a compressed
initramfs. This allows for easier testing. A compressed image is
needed as the image is too big for the storage, and the error notifying
of such is vague.
Signed-off-by: Jon Mason <jon.mason@arm.com>
The mailbox patches don't apply to 5.15.78 so until they can be rebased
and retested, pin the kernel to 5.15.72.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Add set -e so errors are fatal.
Allow HOST_ARCH and VER to be overridden by the environment, for testing.
Pull the tarball basename into a variable to reduce duplication.
Turn the wget call into a function to reduce duplication.
Drop the big-endian binaries as we never use those.
Signed-off-by: Ross Burton <ross.burton@arm.com>
As far as we know nobody is actually using the Arm Compiler recipe: 6.17
does a network operation on every call to check the license and this
fails with the network isolation that do_compile has in kirkstone, and
6.18 is behind a loginwall so we cannot download it in a recipe.
Unless we have actual users asking for a recipe, remove it from the layer
to avoid confusion.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
The recipe supports the use of both Arm's binary GCC (aka GNU Arm
Embedded Compiler, or gnu-rm) and binary Clang (aka Arm Compiler).
However, armcompiler was never tested and doesn't work: 6.17 does a
network operation on every call to check the license which fails with
the network isolation in do_compile tasks, and 6.18 is behind a
loginwall so we can't automatically fetch it in a recipe.
Simplify the recipe to hardcode the use of gnu-rm, and remove the clang
support.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
DEFAULT_TAG and CPU_REQUEST are being used to help with internal Gitlab
pipeline setups know which type of machines to run on, but has no value
outside of Arm Corp. Gitlab CI allows for variables to be overridden
by default. So, we can give it a default value of NULL/empty and have
everything work internally and externally by default.
Signed-off-by: Jon Mason <jon.mason@arm.com>
Tag all jobs with the DEFAULT_TAG variable so each instance can control
what tags the jobs have, whilst still explicitly tagging the jobs which
need specific tags (such as x86_64 for jobs which need to run x86-only
binaries)
Signed-off-by: Ross Burton <ross.burton@arm.com>
The Kas container needs to use the entrypoint as that is where the user
changes from root to a normal user.
Also set the KUBERNETES_CPU_REQUEST to the variable CPU_REQUEST as this
needs to be tuned per-deployment.
Signed-off-by: Ross Burton <ross.burton@arm.com>
The linux-yocto recipe in oe-core upgraded to 5.15.72, so rebase the
doorbell patches to apply correctly.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
To fit the kernel image into the allotted space, a compressed kernel
image is now needed. Use the Image.gz from the kernel build process
and change the relevant places to use the new image name. This also
necessitates adding an unzip command to u-boot to uncompress it to
memory (and the loadm is still needed to setup the efi mem boot device).
Also, the unzipped image is larger than before. So, increase the size
that loadm is copying.
This change shrinks the initramfs bundle size from 12MB to 4.8MB
Signed-off-by: Jon Mason <jon.mason@arm.com>
Signed-off-by: Abdellatif El Khlifi <abdellatif.elkhlifi@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
It is sometimes useful to be able to configure the behavior of FVPs
using environment variables, e.g. for licensing or plugins.
Add a new FVP option: FVP_ENV_PASSTHROUGH, which allows the Bitbake
variables to be passed to the environment to be specified explicitly (in
a similar way to BB_ENV_PASSTHROUGH). This ensures that:
* FVPs launched via runfvp have a reproducable environment
* FVPs launched via testimage (which run from an isolated Bitbake task)
can receive environment variables
Change the self-tests to use cwd instead of PATH to find the mock FVPs,
as the PATH environment variable is no longer passed through.
Issue-Id: SCM-4964
Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
Change-Id: Idf6ac6d41fda4cd5f950bc383c2fc1fa1acdf4e3
Signed-off-by: Jon Mason <jon.mason@arm.com>
The 1.5.x branch no longer exists on the tf-m git repo, but the SHA is
still present. Update the branch name to allow for this.
Signed-off-by: Jon Mason <jon.mason@arm.com>
Historically external-arm-toolchain recipe packaged all gcc headers from
${libdir}/gcc/${TARGET_SYS}/${BINV}/include - some would be picked up by
packages like gcc-sanitizers, libssp-dev. libquadmath-dev or libgomp-dev.
The rest would fall into catch-all libgcc-dev package.
Unfortunately, that could result in a conflict with a target gcc, which
also packages some of those files, like unwind.h or stddef.h, among others.
The conflict could be seen with this config:
EXTRA_IMAGE_FEATURES += "dev-pkgs tools-sdk"
TCMODE = "external-arm"
EXTERNAL_TOOLCHAIN = "/OE/toolchains/gcc-arm-11.2-2022.02-x86_64-aarch64-none-linux-gnu"
And the error message is:
Error: Transaction test error:
file /usr/lib/gcc/aarch64-poky-linux/11.2.1/include/stddef.h conflicts between attempted installs of libgcc-s-dev-11.2.1-r0.1.cortexa57 and gcc-arm+11.2-r2022.02.1.cortexa57
file /usr/lib/gcc/aarch64-poky-linux/11.2.1/include/unwind.h conflicts between attempted installs of libgcc-s-dev-11.2.1-r0.1.cortexa57 and gcc-arm+11.2-r2022.02.1.cortexa57
Modify external-arm-toolchain recipe according to how libgcc in OE-Core
handles those header files by removing and not packaging them:
https://git.openembedded.org/openembedded-core/tree/meta/recipes-devtools/gcc/libgcc-common.inc#n40
Also need to adjust gcc recipe to pick up unwind.h from EXTERNAL_TOOLCHAIN
location now, since libgcc-dev no longer carries it, and install it into
STAGING_LIBDIR_NATIVE, where OE-Core gcc-target.inc expects it from
gcc-cross:
https://git.openembedded.org/openembedded-core/tree/meta/recipes-devtools/gcc/gcc-target.inc#n164
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
In complex stacks, e.g. with many cores or many init scripts, the time
to Linux shell may be more than 10 minutes. Make the boot timeout
configurable using TEST_FVP_LINUX_BOOT_TIMEOUT, leaving the default
value at 10 minutes.
Issue-Id: SCM-4958
Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
Change-Id: Ie074acd4b4509d0230d1f77a2a527d497bb295ce
Signed-off-by: Jon Mason <jon.mason@arm.com>
Since upstream OP-TEE 3.16.0, CFLAGS32 and CFLAGS64 can be used to set
the sysroot option that previously required the LIBGCC_LOCATE_CFLAGS
patches.
Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
Signed-off-by: Jon Mason <jon.mason@arm.com>
Add documentation for how to use the OEQA framework to test targets in
meta-arm. Include instructions on using OEFVPTarget as well as the
OEFVPSerialTarget introduced by the recent refactor of runfvp.
Issue-Id: SCM-4954
Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
Change-Id: I146ec1c82214471fe9d18a999fd92efb38f652f9
Signed-off-by: Jon Mason <jon.mason@arm.com>
The runfvp refactor to enable OEFVPSerialTarget created FVP_CONSOLES
which maps the names used for serial ports in test cases to the names
used for serial ports in the FVP stdout.
Refactor the FVP_CONSOLE section -> FVP_CONSOLES, noting the the
'default' console is still used for the --console runfvp flag.
Issue-Id: SCM-4954
Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
Change-Id: Ieb13d74cfd425900f44b4b2e6d125393e7b456ad
Signed-off-by: Jon Mason <jon.mason@arm.com>
meta-gem5 needs the 5.4 kernel, which is only present in the
meta-arm-bsp layer. Add this as a dependency to resolve issues.
Signed-off-by: Jon Mason <jon.mason@arm.com>
This appears to be historical from when the toolchain was in meta-linaro.
It isn't needed anymore, there's one bbappend in meta-arm-toolchain for
grub which is part of oe-core, so will never be dangling.
This variable has a global effect, so leaving it in here has a negative
impact on users.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
lib32-optee-os do_compile failed since OPTEE_ARCH is not set correctly.
In file included from lib/libutee/include/link.h:9,
from lib/libutee/arch/arm/tcb.c:37:
lib/libutee/include/elf.h:67:2: error: #error Unknown architecture
67 | #error Unknown architecture
| ^~~~~
In file included from lib/libutee/arch/arm/tcb.c:37:
lib/libutee/include/link.h:13:9: error: unknown type name 'Elf_Addr'
13 | Elf_Addr dlpi_addr; /* module relocation base */
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Jon Mason <jon.mason@arm.com>
