GitLab

Mylène Josserand authored Dec 17, 2020 and Guillaume Tucker committed Jan 21, 2021

Add the possibility to set the "failure_retry" parameters
into the device dictionnary in case we would like to
re-test all the failing jobs (with "infrastructure_error" or
"job_error_exception")
Signed-off-by: Mylène Josserand <mylene.josserand@collabora.com>

Rémi Duraffort authored Apr 23, 2020 and Guillaume Tucker committed Jan 21, 2021

This is folowing the drop of Boot root action.
Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>
Signed-off-by: Mylène Josserand <mylene.josserand@collabora.com>

Rémi Duraffort authored Apr 17, 2020 and Guillaume Tucker committed Jan 21, 2021

Classes that inherit from Deployment, Boot and LavaTest are in fact just
returning the root action to add to the pipeline.

Also make the pipeline construction easier to read.
Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>
Signed-off-by: Mylène Josserand <mylene.josserand@collabora.com>

For some reason yet unknown, calling `.current_job()` on a Device
object when its state is not "running" can cause some server errors.
This is likely to be due to some discrepancy in the database.  The
schedule.all_devices() XMLRPC call causes an HTTP502 when this
happens.  Work around this problem by only looking for the current job
of devices that are in the running state.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

In some cases, a device can be in the "running" state but still not
have any current job associated with it.  This causes some exceptions,
as the code tries to get a `.pk` attribute from a None object.  The
root cause of the problem is likely due to some glitch in the state
machine, where the device state changes before the new job has been
associated with it.  Work around this by printing a message with "(no
current job)" when this happens, to avoid an exception and to help
debug the root cause of the problem.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

The panda board used to require some special kernel command line
arguments before support was fully added to the device tree.  This is
not required any more with at least kernel 3.18 so it's safe to drop
those arguments now.  With older kernel such as 2.6.35 it's still
possible to specify some extra kernel arguments in the job definition.

Also, this more recent change in the device tree actually caused some
kernel alert message when using the old kernel command line arguments:

  b4778e787fe9 ("ARM: dts: omap4-panda-common: Add CMA pools and enable IPU & DSP")

Dropping them clears the error message.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Sjoerd Simons authored Sep 03, 2020 and Guillaume Tucker committed Jan 21, 2021

Suffix the snapshot version with .collabora to make it more clear this
is a collabora forked version
Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

Mylène Josserand authored Sep 18, 2020 and Guillaume Tucker committed Jan 08, 2021

Handle the timeout of depthcharge as an Infrastucture error
and not an job error.

Applied-Upstream: 2020.09, 1e21b3e1

Signed-off-by: Mylène Josserand <mylene.josserand@collabora.com>

Sjoerd Simons authored Sep 03, 2020 and Guillaume Tucker committed Jan 08, 2021

Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

For non-fit kernels the ramdisk is passed on the command line; If no
ramdisk is used (which is valid) set the ramdisk value to "" otherwise
the line will be dropped from the commands
Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

Add a template for the hp-11A-G6-EE-grunt x86_64 Chromebook device
type using Depthcharge.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Depthcharge now supports an extra optional argument with the tftpboot
command to download a separate ramdisk file.  This is useful for
architectures that don't have FIT support in Depthcharge such as x86.
Add a DEPTHCHARGE_RAMDISK for this purpose, to not conflict with the
global RAMDISK parameter which would also be set when using FIT format
and result in the ramdisk to be downloaded twice.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Depthcharge has support for FIT images on arm/arm64 but not x86.  To
deal with this, make the FIT image conversion optional to only use it
when applicable.  If the "mkimage_fit_arch" parameter is not defined
in the device type, the kernel image is left as-is and not converted
to FIT.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

The FIT parameter used in Depthcharge device templates assumes the
kernel image is of the FIT format.  While this is normally true for
arm/arm64, it's a misleading name for other architectures such as x86
which don't have FIT support in Depthcharge.  As such, rename it to
DEPTHCHARGE_KERNEL to also not conflict with the global KERNEL
parameter.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

The docstring for the strings.substitute() function states that a line
will be dropped if a value is None, but the code just checks if the
values evaluate to False.  As a result, values with an empty string
also cause the line to be dropped.  Fix that by testing for None
explicitly.

Fixes: 82746b05

 ("Drop u-boot commands if the placeholder is not replaced")
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Enable the vland block in the qemu.jinja2 base template to be able to
share the device_ip parameter among other things.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

This reverts commit ae2e910e

.

It seems like this change removes the possibility to set an
alternative favicon without editing the base template.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Rémi Duraffort authored Mar 25, 2020 and Guillaume Tucker committed Mar 27, 2020

When using more than one namespace, the connection is passed from one namespace
to the other and should not be terminated by qemu.
Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>

Enable running LXC containers alongside QEMU devices.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Add device type support for the mt8173-elm-hana Chromebook.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Enable running LXC containers alongside devices that use GRUB.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Sjoerd Simons authored Sep 28, 2018 and Guillaume Tucker committed Mar 27, 2020

If the server does not provide extra dispatcher configuration check for
a local file to include in the runs. Useful to e.g. configure the IP to
use for communicating with the duts.
Signed-off-by: Sjoerd Simons <sjoerd.simons@collabora.co.uk>

Denis Pynkin authored Mar 27, 2018 and Guillaume Tucker committed Mar 27, 2020

Allow to specify any path in `lava_test_results_dir` for deployments.

Originally guest image for Qemu is mounted to target mount point but in the same
time path inside the image skip only 1 directory in hierarchy doubling
the rest of the path in Qemu VM, for instance:
`/var/lib/lava-5/lib/lava-5/bin/...` instead of `/var/lib/lava-5/bin/...`.
This behavior causes the issues in `lava-test-shell` test:
 lava-test: # lava-test: # sh: 9: /var/lib/lava-4/bin/lava-test-runner: not found
Signed-off-by: Denis Pynkin <denis.pynkin@collabora.com>

Some Ubuntu-derived distros such as Apertis have a read-only root file
system, and it can't be re-mounted as rw.  So to be able to store LAVA
test files, move the directories to /var/lib/ instead as this has
write permissions.
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

This was suggested by Huawei to avoid some boot failures.

Change-Id: I93e146a4c24d69ea91c1bf1e02331234546306fc
Signed-off-by: Guillaume Tucker <guillaume.tucker@collabora.com>

Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>

docker: add a variable to add any extra arguments

See merge request lava/lava!1020

Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>

device-types: bcm2711-rpi-4-b: support booting rpi-sources

See merge request lava/lava!1019

docker: allow to skip calls to check_owners

See merge request lava/lava!1018

Fix nbd deploy

See merge request lava/lava!1017

On large systems, this can be really long.
Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>

The image type was not taken into account due to the missing PrepareKernelAction
Signed-off-by: Rémi Duraffort <remi.duraffort@linaro.org>

Add unit tests for filtering in REST API

Closes #359

See merge request lava/lava!1015

The current device-type of bcm2711-rpi-4-b works well with mainline
Linux.
But the rpi-sources does not boot with it.
First it miss the cma size.
Then without 8250.nr_uarts=1, the rpi-sources fail to open the console.
But with 8250.nr_uarts=1, ttyS1 become ttyS0.

Instead of hardcoding too many things, lets's permit to override extra_kernel_args.

Fixes #359

Support new permission management via REST API

Closes #330

See merge request lava/lava!1005

Fixes #330

Allow to add overlays to resources

Closes #320 and #307

See merge request lava/lava!1004

NXP IMX templates rework

See merge request lava/lava!1014