Age | Commit message (Collapse) | Author | Files | Lines |
|
Use the new qemu for dual-screen capability.
Signed-off-by: Jan-Simon Moeller <jsmoeller@linuxfoundation.org>
Change-Id: I97a4d303bbc452902a4b555d00f05b0685e0e2de
|
|
It is not systemd-networkd that does bite use here, it is the
systemd-network-generator.service that will interfere with
our settings. This is triggered by the ip=dhcp setting on the
kernel cmdline and even the removed wired.network file will
not prevent his from interfering and taking the network
interface down and up for another dhcp run.
Question is ... WHY ?!: The interface IS already configured.
Connmand can be configured to ignore it and does actually ignore
or at least not take it down.
Signed-off-by: Jan-Simon Moeller <jsmoeller@linuxfoundation.org>
Change-Id: I2872ddfbc97e26ae84fcc3640a9129c28691557c
|
|
Signed-off-by: Jan-Simon Moeller <jsmoeller@linuxfoundation.org>
Change-Id: I21ec3f2772a2a997fa3d370712d1ce49bb46cd24
|
|
Using a more recent qemu seems to fix problem with screenshot.
The easiest way is to use docker to "host" a more recent debian with
recent qemu.
Change-Id: I7a72de4c6a4dceed7ec967713d7fee3d3523e120
Bug-AGL: SPEC-3764
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
qemu machines does not have a network card by default.
So let's add one.
Change-Id: I6ea5d5625ee5c24b5ee49652f1aa838580a6f952
Bug-AGL: SPEC-3631
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
We need to add a virtual display device for testing homescreen.
Let's add a vnc to do this.
But we need to avoid VNC to open a network port, so we will use an unix
socket.
For handling the fact that two job could run in parallel, each unix
patch need to be unique. We will generate an unique path with gerrit id,
branch, and a random int.
The vnc is added for all qemu since it seems to not hurt ARM/ARM64.
We also add a virtio display device on x86_64 since apps are tested only
on x86_64.
Change-Id: I22412df75cad16f151f73d0d93b0e38ae386ef4d
Bug-AGL: SPEC-3597
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
On 2020.02, qemu jobs fail to mount the rootfs.
This is due to LAVA ignore now compression for ramdisk and the AGL image
is in xz.
But AGL qemu jobs in fact does not use a ramdisk but a virtio disk.
Renaming the ramdisk definition entry to rootvd handle this issue.
Furthermore, let's introduce a new rootfs type rootvd.
Change-Id: I71ef6dee0d859c6aa0b29f1d47ac14dd4db900d8
Bug-AGL: SPEC-3210
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
Specifying the kernel type in "actions: boot:" is deprecated.
It should now be set in "deploy: kernel:".
So this patch set unconditonaly the type: in the deploy section (before
it was only for ramdisk method) and remove the type: from the deprecated
place.
But due to a bug in LAVA, NBD jobs rely on type: to be in the deprecated
place until 2020.02 (2020.02 has the fix).
So this patch must be applied only after LAVA 2020.02 was deployed.
Change-Id: Ida30092d2abe14b05d5e37f651f91ab33876e4f5
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
By default LAVA set the test device on IDE.
Stretch version of qemu doesnt care about this setting and was always
using virtio.
But buster qemu care now about this and fail with:
machine type does not support if=ide,bus=0,unit=0
This patch adds the necessary to restore the use of virtio.
Change-Id: If34194d82b6104b9b5be7118a5194bd3b61fc090
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
LAVA expect machine, memory and cpu to be passed as argument in context and not
via extra args.
This patchs introduce qemu_cpu, qemu_memory and qemu_machine as jinja
variables.
Change-Id: I8893943947a046f90be388febccfc29c7923d5b4
Bug-AGL: SPEC-2376
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
This patch adds suppport for the upsquare.
Upsquare use a GRUB boot method, so we needed to add a new
generic-grub-tftp boot template.
Since all x86_64 use the same yocto generated build name, we add a new
upsquare machine, but tweaked a bit utils/agljobtemplate.py to keep
intel-corei7-64 everywhere (except for the LAVA device-type).
Bug-AGL: SPEC-2075
Change-Id: I15c850fada53efbcda4fd58527d7b7414865f89d
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
|
|
Since iPXE based boards can use NBD boot[1], now in releng-scripts,
we also add support to generate LAVA jobs for iPXE based boards.
With this support, Intel-UP2, one of the community reference boards, can
do validation tests with the generated jobs easily.
[1] https://git.lavasoftware.org/lava/lava/merge_requests/64/diffs
Change-Id: I9adce67d7e9207181675e82bf22a653c73682eb8
Signed-off-by: Liu Wenlong <liuwl.fnst@cn.fujitsu.com>
|
|
Since the following commit:
https://github.com/Linaro/lava-dispatcher/commit/b8e7dc1317e263884c2e10bcf4ff822857a545a5
LAVA jobs need to have a set_port action.
This patch permit to generate a job with such action
Change-Id: I34152d146f3deb6839c320da86a0b75187a93070
Signed-off-by: Corentin LABBE <clabbe@baylibre.com>
Tested-on: m3ulcb
|
|
This change relaxes the login match a little bit by removing the username.
This will assist cases where we see not-recognized logins as well as
tests with different usernames (non-root).
Change-Id: Ied0ee9930f7804eb9883649495306e16964a75e6
Signed-off-by: Jan-Simon Möller <jsmoeller@linuxfoundation.org>
|
|
Depending on the boot type (ramdisk or nbd), specify a boot method.
This change helps defining the correct boot type automatically from
the boot protocol (ramdisk or nbd).
Tested with m3ulcb (nbd) and raspberrypi3 (ramdisk).
Change-Id: I84845c2fa94b32cd53c91a5330b25505e7b60930
Signed-off-by: Loys Ollivier <lollivier@baylibre.com>
|
|
This reverts commit 364641197bfe95512555abc36b7b62ac48166aaf.
Unfortunately, NBD jobs do not work with the new kernel type, so we can't remove the old boot type (at least until NBD is fixed.)
Change-Id: If9a64135ef75771a90655fdc4c6429d9b494fcf0
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
Boot type is deprecated in lava v2.
Replaced by kernel_type and define it from the kernel image name.
Change-Id: I6bcb326b5fb8b0230a9d519485741209bfab8ef2
Signed-off-by: Loys Ollivier <lollivier@baylibre.com>
|
|
Added new command line arguments to specifiy build artifacts names:
- "--rootfs-img": for the rootfs file name
- "--kernel-img": for the kernel name
- "--dtb-img": for the dtb file name
- "--modules-img": for the modules file name
Change-Id: I665dee2b326aeaaf35148345f0501e49950ba9aa
Signed-off-by: Loys Ollivier <lollivier@baylibre.com>
|
|
When testing with basic ramdisk support, we need to be able to detect
a simple root prompt.
Change-Id: I73302e2b7bed1d6e499fbdf3918ad3ed8dfdd03a
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
|
|
NFS is untested and not used at the moment.
Remove it for now to avoid unnecessary issues
Change-Id: Ide7f0663f3c31cde14dcda3c6e6694e22165bf49
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
|
|
Without transfer_overlay, LAVA will use libguestfs to insert the overlay
into the rootfs. This will cause problems with rootfs that have security
enabled, so use transfer_overlay to load the overlay after the rootfs
has been booted.
Tested with QEMU devices in lab-baylibre.
Change-Id: Ib27e5e862c64e8b302fa486f113fed4fdc6f5e83
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|
|
Change-Id: I5d545d3531f4c4190453724738076bd4eddfc4d6
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Loys Ollivier <lollivier@baylibre.com>
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
|