Age | Commit message (Collapse) | Author |
|
Similarly to the U-Boot template, if extra_kernel_args is defined then
add it to the job context for Depthcharge.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
syslog-style printk message from the kernel are available when booting
with "console_msg_format=syslog" on the kernel command-line. This
makes it much easier for log parsers to get the log-level information
from the kernel boot logs.
Automatically add this command-line argument to all u-boot and qemu
jobs.
NOTE: requires jinja2 "do" extension, which is now added when creating
jinja2 environment
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
Replace the igt test plan with igt_drm_kms as it only covers the
DRM/KMS part, and use igt-parser.sh from the buster-igt rootfs. Also
drop core_get_client_auth as it has been merged with core_auth.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Call all the tests from the standard igt installation path
/usr/libexec/igt-gpu-tools as this is where they are installed in
buster-igt using meson.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
On lab-baylibre, we have many slaves running on potato.
While they are still powerful enough for running lot of jobs in
parallel, they lack a bit on IO.
So while running many deploy in parallel, they hit sometimes the deploy timeout.
This patch increases the deploy timeout, 2 minutes is too short.
|
|
Add a "baseline" test plan intended to ultimately replace "boot"
tests. This starts by using bootrr to verify that all the expected
device drivers have been intialised correctly, to strike a good
compromise between runtime and test coverage.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Add barebox variant to the test plans.
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
|
|
Replace the core_prop_blob test case with kms_prop_blob since it has
been renamed.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Rename the v4l2 test plan as v4l2-compliance to make it clearer what
it is about. There may be other v4l2 related test plans in the future
with a different name.
* rename the v4l2 template files
* update test-configs.yaml accordingly
* update kernel-arch-complete.sh and lava-boot-v2.sh accordingly
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Update build.py to create and push a build_environment variable to
distinguish between builds with different docker images.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Pass the v4l2 driver name to v4l2-parser.sh using a v4l2_driver
template paramter, and set the test plan name dynamically with
plan_name to allow variants (v4l2-vivid etc...).
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Add a v4l2 template variant to run it on QEMU. This can be done with
the "vivid" virtual video driver enabled in the kernel build.
* add generic-qemu-v4l2-template.jinja2 to enable v4l2 on qemu
* add image_arg block in generic-qemu-boot-template.jinja2 to make it
possible to override the QEMU command line in test plans
* add the vivid.no_error_inj=1 to the QEMU kernel command line when
running the v4l2 test plan. This is required with the vivid driver,
otherwise v4l2-compliance aborts with this message:
The vivid driver has error injection enabled which will cause
the compliance test to fail. Load the vivid module with the
no_error_inj=1 module option to disable error injection.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
In order to be able to test with both 32-bit and 64-bit user-space on
x86, use i386 and x86_64 arch names explicitly. The main difference
is that test plans using Debian will now be running with a 64-bit
user-space on x86_64 platforms instead of 32-bit. There is no 64-bit
Buildroot file system image available for the time being.
* update test-configs.yaml with new arch maps for each file system
type to deal with both i386 and x86_64
* update the arch value for x86 devices so they are now either i386 or
x86_64
* update defconfig filters in test-configs.yaml as the defconfig for
these archs is now "defconfig" rather than i386_defconfig and
x86_64_defconfig
* update qemu template for x86_64
* use the same x86 Docker images to build kernels for either i386 or
x86_64 as the native toolchains are identical for x86
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
This reverts commit a08d0afc7523e478873eeae43cbe03999754ba49.
As all the output of v4l2-compliance is now going through
v4l2-parser.sh and not the basic LAVA results parser on the serial
console, there is no need to redirect it via /dev/kmsg.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
* add tests overlay with /usr/bin/v4l2-parser.sh to parse the output
of v4l2-compliance and call lava-test-case for each test case with
valid test case IDs (no spaces...)
* update v4l2.jinja2 to call this script instead of parsing the test
cases inline
* enable streaming test cases with v4l2-compliance -s
* print the v4l2 driver and device names so they appear in the log
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Make all v4l2 messages go through /dev/kmsg to avoid collisions with
kernel messages causing LAVA parsing errors.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
The v4l2-compliance test suite is not a test case, so only rely on the
LAVA results parsing to collect test case results. This removes the
misleading "v4l2-compliance" test case from the results.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Define some test sets for "networking" and "filesystem" in the simple
test plan, and add one test case "whoami" outside test sets. This is
primarily to check that test sets are parsed correctly in the backend
as sub-groups and used properly in email reports or the frontend.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Add QEMU variant of the simple test plan and enable it in
test-configs.yaml.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Follow the same format as other test plans to avoid duplicating the
test steps for each variant by including a common simple.jinja2 file.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
All the information in the templates .ini files is now either
contained in lib.test_configs or deprecated with LAVA v2. So these
files can be safely removed now.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
In order to be consistent with other test plans, rename boot templates
with ramdisk-boot-template and fix templates that extend it
accordingly.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Now that LAVA v1 has been deprecated, all the JSON templates can be
removed. LAVA v2 uses only the Jinja2 templates.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Add two more tests to check the presence of all the usb devices
after suspending the system using rtcwake with the modes 'mem' and
'freeze'.
Signed-off-by: Ana Guerrero Lopez <ana.guerrero@collabora.com>
|
|
Use a feature from lava to use /dev/kmsg for LAVA test signals.
This change requires a LAVA instance with this patch included
in LAVA releases after 2018.5
https://review.linaro.org/#/c/25377/
Signed-off-by: Ana Guerrero Lopez <ana.guerrero@collabora.com>
|
|
Initial test plan that checks for /dev/cros_ec
This allows to know if the driver probe has run successfully and the
kernel has been able to establish communication with the embedded controller.
Signed-off-by: Ana Guerrero Lopez <ana.guerrero@collabora.com>
|
|
Initial test plan running some of the binaries from the test suite.
This test aims to cover the DRM kernel-user interface and is not specific
to Intel graphics.
This test plan uses /dev/kmsg for LAVA test signals, this feature requires
a LAVA instance with this patch https://review.linaro.org/#/c/25377/
Signed-off-by: Ana Guerrero Lopez <ana.guerrero@collabora.com>
|
|
Remove workaround with dmesg and use new feature from lava
to use /dev/kmsg for LAVA test signals.
This change requires a LAVA instance with this patch
https://review.linaro.org/#/c/25377/
|
|
Initial test plan that runs v4l2-compliance in the system and parses
the output.
Signed-off-by: Ana Guerrero Lopez <ana.guerrero@collabora.com>
|
|
|
|
When using a debian initramfs, if the boot falls through to the
ramdisk, the shell prompt is "(initramfs)". Add that to the list
of expected prompts.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
For now, it just checks that lsusb's output is non-empty.
Most devices will have at least one USB device, and those that don't can
have any kind of non-intrusive USB device plugged in.
This will test for now that the driver for the host USB controller
probed correctly and the very basics are working.
|
|
Add a start for a test plan that checks that system power states work,
along with RTC wake ups.
v2: add test plan name to file name of templates
v3: extract common fragment of test definition to its own file
reduce test timeout from 10 to 5 minutes
repeat 10 times each suspend/resume cycle
silence KERN_INFO as it confuses LAVA on wakeup
v4: extend boot/generic-depthcharge-tftp-ramdisk-template.jinja2
|
|
Add some command line options that are needed to generate LAVA jobs
with alternative callbacks.
* add --callback-url to override the default callback base URL with an
arbitrary one, the KernelCI backend API URL is used by default
* add --callback-dataset to choose the dataset sent in the callback
payload, the default being 'all'
* add --calback-type to choose whether to use the bare 'custom' URL or
append the extra 'kernelci' path and query needed by the KernelCI
(defaults to the latter so this remains backwards compatible)
* move code that deals with the callback template variables into a
separate function
* update kernel-ci-base.jinja2 to set the url and dataset accordingly
when these options are used
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
In order to be able to run any test beyond plain boot to a prompt on
QEMU, i.e. with some LAVA "test" steps, the guestfs interface needs to
be set to virtio. Set this for all QEMU boot jobs so that test
templates can also inherit from it.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Add generic templates for plain boot and boot-nfs jobs for devices
using Coreboot/Depthcharge (typically Chrome OS devices). This
involves removing the kernel image type attribute for Depthcharge as
it needs a FIT image rather than the legacy U-Boot image conversion to
uImage.
Please note that these devices need to be manually updated with a
Depthcharge custom build with the debug CLI config enabled and the
tftpboot command available.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
|
|
|
|
|
|
The rpi3 has been blacklisted to as many old kernels as I could
find, and the console setting is now overriden using context.
V2 blacklisting was not working, it was being checked against the
tree name not git describe.
And do not boot test kselftest config.
|
|
to try and help baylibre lab
|
|
jobs-from-api.py:
For the expected defconfig counts to work, we need to fetch all results
including the fails and then filter them out before building jobs
device-map:
Remove NFS boots from apq8016-sbc as they just don't work
templates:
Cleanup indentation for future includes/extends
|
|
QEMU templates don't have indentation for deploy or boot action, so
remove extra indentation from deploy and boot actions for grub/pxe/uboot
also.
This allows any additional actions which might be appended (e.g test
actions) to have the same indent level for QEMU and grub/ipxe/uboot
jobs.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
|
|
Add NFS boot types
Add boot-be templates for BIG_ENDIAN defconfigs
|
|
Add the SoC type as platform.mach taken from the device_map to the
LAVA v2 jobs definition. This can then be used in the KernelCI
backend as the SoC name.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>
|
|
Only generate kselftest jobs for kernels that have been built with all
the kselftest defconfig options.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
|
|
Add a platform.name to the metadata
|
|
|