| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Because every aktualizr package was depending on aktualizr-configs, all
of the default configs were getting installed to /usr/lib/sota, despite
that by default none of them were ever used on the device. That
dependency has been removed.
The provisioning recipes still install the recipes they need to the
default location that aktualizr actually uses (/usr/lib/sota/conf.d). If
you want to use anything else, a specific recipe should be created to do
so.
This change rendered the aktualizr-host-tools package useless, so it has
been removed.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
| |
And it's a newer version at that!
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
| |
The version in poky already supports a native build.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Previous initramfs fit specific workaround was removed as part of commit
6eecf1593d, causing ostree to provide an init argument as part of the
kernel command line arguments when it is not really needed.
Bring back the workaround by simply generating an empty file in case
kernel image type is fit and initramfs is also used by the target.
Signed-off-by: Ricardo Salveti <ricardo@foundries.io>
|
|
|
|
| |
Signed-off-by: Dominik Wild <dominik.wild@glutz.com>
|
|
|
|
| |
Signed-off-by: Anatoliy Odukha <aodukha@gmail.com>
|
|\
| |
| | |
Feat/thud/2020.9
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* failing "which repo" causes the run.buildinfo_manifest task to fail
before it even reaches the test for empty repotool variable:
dash -x some-image/1.0-r2/temp/run.buildinfo_manifest.80233
+ export systemd_system_unitdir=/usr/lib/systemd/system
+ export systemd_unitdir=/usr/lib/systemd
+ export systemd_user_unitdir=/usr/lib/systemd/user
+ buildinfo_manifest
+ which repo
+ repotool=
+ bb_sh_exit_handler
+ ret=1
+ [ 1 != 0 ]
+ echo WARNING: exit code 1 from a shell command.
WARNING: exit code 1 from a shell command.
+ exit 1
causing nasty long python exception from do_image task
with this fix, it just shows an warning again:
dash -x some-image/1.0-r2/temp/run.buildinfo_manifest.80233
+ export systemd_user_unitdir=/usr/lib/systemd/user
+ buildinfo_manifest
+ which repo
+ true
+ repotool=
+ [ -n ]
+ bbwarn Android repo tool not found; manifest not copied.
+ [ -p some-image/1.0-r2/temp/fifo.80233 ]
+ echo WARNING: Android repo tool not found; manifest not copied.
WARNING: Android repo tool not found; manifest not copied.
+ ret=0
+ trap 0
+ exit 0
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Now we have changed to inherit python3native to provide python3 binary
for repo, then we need explicitly call python3 to execute repo, or else
it will firstly try to locate python which will fail if python is not
in HOSTTOOLS.
This fixes a following warning:
| Android repo tool failed to run; manifest not copied
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* nfs-home-mount.service is installed by do_install_append_sota, so only when sota is enabled
but it was always added to SYSTEMD_SERVICE_ even when sota wasn't enabled resulting in
do_patch failure:
nfs-utils-2.5.1-r0 do_package: SYSTEMD_SERVICE_nfs-utils value nfs-home-mount.service does not exist
* use sota override everywhere
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
nfs-mountd.service and nfs-statd.service rely on /var/lib/nfs to be
present in rootfs, or else they would run into problems for some files
missing in it.
But a ostree/ota rootfs does not have that directory (removed during
do_image), to fix that, we now copy them to /var/local/lib/nfs and
mount it to /var/lib/nfs at runtime, which is done in
nfs-home-mount.service.
nfs-mountd.service and nfs-statd.service need run after
nfs-home-mount.service.
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
|/
|
|
|
|
|
| |
The functionality was (presumably unintentionally) removed, but the
option was still present.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|\
| |
| | |
Feat/thud/2020.8
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Following the changes in ostree's deployment of the kernel, initramfs
and devicetree in /lib/modules/$kver, the deployment method of the
device tree also changed. Instead of picking the first device tree it
finds at a given location, ostree looks at a file named devicetree, next
to kernel and initramfs in /lib/modules/$kver.
This commit modifies ostree-kernel-initramfs to deploy the devicetree
from the sota-defined variable OSTREE_DEVICETREE. It will pick the
first one from the list of device trees that OSTREE_DEVICETREE defines,
and copy it to /lib/modules/$kver. Note that since OSTREE_DEVICETREE
equals to KERNEL_DEVICETREE when it isn't explicitly defined, it could
indeed be a list of device trees.
Signed-off-by: antznin <agodard@witekio.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When setting intertask dependencies, we should use PN instead of
IMAGE_BASENAME to refer to a image recipe, since PN is generated from
recipe file name, while IMAGE_BASENAME is a variable that could be
changed, it's not guaranteed to always equal to PN.
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* otherwise it fails with useless error:
ostree-kernel-initramfs/0.0.1-r0/temp/run.do_install.3011' failed with exit code 1:
cp: -r not specified; omitting directory 'tmp-glibc/deploy/images/qemux86/'
WARNING: exit code 1 from a shell command.
because of
cp ${DEPLOY_DIR_IMAGE}/${OSTREE_KERNEL} $kerneldir/vmlinuz
in do_install will try to copy whole ${DEPLOY_DIR_IMAGE}/ when
${OSTREE_KERNEL} is empty
as reported in:
https://github.com/advancedtelematic/meta-updater/pull/740#issuecomment-651952735
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
| |
| |
| |
| | |
Signed-off-by: antznin <agodard@witekio.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
INITRAMFS_IMAGE
* when INITRAMFS_IMAGE isn't defined at all, the "d.getVar('INITRAMFS_IMAGE') == ''" part
doesn't do anything useful, because the unexpanded version of this ends in
do_install[depends] variable and breaks parsing:
meta-updater/recipes-sota/ostree-kernel-initramfs/ostree-kernel-initramfs_0.0.1.bb:do_install[depends], dependency ''][d.getVar('INITRAMFS_IMAGE') in 'virtual/kernel:do_deploy ${@['${INITRAMFS_IMAGE}:do_image_complete', ''][d.getVar('INITRAMFS_IMAGE') == '']} virtual/fakeroot-native:do_populate_sysroot' does not contain exactly one ':' character.
Task 'depends' should be specified in the form 'packagename:task'
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We did the work a while ago to make aktualizr-info work for Secondaries,
but until now we weren't putting the tool into the secondary-image we
use for testing. Now it's there. Actually, it's in every image that
inherits from sota.bbclass, which is probably a good thing.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We package kernel image, devicetrees, initramfs and install them to
/usr/lib/modules/${KERNEL_VERSION}, which is the preferred location
according to ostree's new implementation, this could simplify the
deployment.
Reference:
https://github.com/ostreedev/ostree/commit/3ab0d5e6644885440bac6abd17b6d2637df5435f
To let initramfs-ostree-image be able to be depended by
ostree-kernel-initramfs, it must inherit nopackages to avoid a annoying
QA warning like the follows:
| WARNING: initramfs-ostree-image-0.0.1-r0 do_package: Manifest ...initramfs-ostree-image.packagedata
| not found in colibri_imx6 armv7ahf-neon-imx armv7at2hf-neon-imx armv7at2hf-neon armv7ahf-neon armv7at2hf-vfp
| armv7ahf-vfp armv6thf-vfp armv6hf-vfp armv5tehf-vfp armv5ehf-vfp armv5thf-vfp armv5hf-vfp allarch
| x86_64_x86_64-nativesdk (variant '')?
Also we need define OSTREE_KERNEL, OSTREE_DEPLOY_DEVICETREE and
OSTREE_DEVICETREE in sota.bbclass so they could be accessed in other
recipes as well as in image recipes.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The name 'virtual' is a recipe specific concept but not a package
specific one, it's confusing that a package provides 'virtual/' names.
Let's drop 'virtual/' from network-configuration, to keep consistent
with yocto naming styles.
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| | |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|/
|
|
|
|
|
|
|
|
|
| |
When setting `OSTREE_DEPLOY_DEVICETREE` to 1, it will by default deploy
all the device tree blobs present in `KERNEL_DEVICETREE`. Adding
`OSTREE_DEVICETREE` would allow specifying a specific device tree blob
(or several dtbs), thus resulting in only the specified ones in the
image. This is particularely useful because ostree selects the first
device tree it finds in /boot, and discards the remaining ones.
Signed-off-by: antznin <agodard@witekio.com>
|
|\
| |
| | |
2020.7 on thud
|
| |
| |
| |
| | |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
| |
| |
| |
| |
| |
| | |
And garage-sign to 0.7.1-4
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The OSTree commit command allows to add metadata to the commit. This
might be customized in a distro layer for distribution specific needs.
Allow to pass extra arguments using EXTRA_OSTREE_COMMIT variable
(using a variable named similar to EXTRA_OEMAKE used to pass extra make
arguments).
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
| |
| |
| | |
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Use named reference when deploying the device tree. This makes sure that
this reference will end up in the deployments .origin file, which will
be visible by the user in ostree admin status. This reference will also
be used when running ostree admin upgrade. This is not really required
when using aktualizr, but can be useful during test/debugging when using
pure OSTree updates. Make sure to make ${OSTREE_REPO} accessible via
http and add a remote on the device called ${OSTREE_OSNAME} pointing to
the http exposed archive OSTree. With that ostree admin upgrade should
work.
Note: We could use the name reference in ostree pull-local already, but
this is potentially racy if multiple builds are committing to the same
branch. Use the OSTree commit hash to get the actual commit and recreate
a local reference to this commit. This makes absolutely sure that we use
the same OSTree commit this bitbake execution committed during the
do_image_ostreecommit task.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Relying on a OSTree branch reference has been problematic in the past
and addressed by adding more attributes to it in commit 202a8c70ba8c
("image_types_ostree: Add a unique ref to fix simultaneous bitbaking.").
However, depening on what kind of OpenEmbedded builds are running in
parallel, even more attributes would need to be taken into account.
Instead of relying on a reference, store the exact ostree commit hash
in a manifest file and reuse it in the do_image_ota deploy task. This
guarantees that the correct reference gets picked even when two builds
with the exact same machine/image name run in parallel.
Note: This gets rid of the second branch name again. If the branch name
with image name is preferred, the variable OSTREE_BRANCHNAME can be
used:
OSTREE_BRANCHNAME = "${SOTA_HARDWARE_ID}-${IMAGE_BASENAME}"
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
| |
| |
| |
| |
| |
| | |
To avoid generating a qemuboot.conf for qemu machines, it's useless for
a initramfs image.
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The current sanity check are too strict for some boolean variables,
introduce sota_check_boolean_variable to allow a boolean value to be
set like yes/y/true/t/1 or no/n/false/f/0.
Also change to use oe.types.boolean to check their values.
Signed-off-by: Ming Liu <ming.liu@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The home directory currently are commited to the OSTree, presumably
to then use it for the deployment. However, we do have access to the
original rootfs in the OSTree deployment tasks (do_image_ota) hence
transferring the files "via OSTree" is not necessary. We do already
carry over some files from the original OE rootfs to /var/sota.
Follow this approach for /var/local and /home as well.
The home will still be stored in the sysroot as documented in
https://ostree.readthedocs.io/en/latest/manual/adapting-existing/.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of copying the files to be commited to the ostree just use a
hardlink tree. This improves performance and wasts less diskspace.
When using this method the root directory has already the correct
permission bits set. Also get rid of the unnecessary sync. This halfs
the execution time of the do_image_ostree tasks in my measurments.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The current code clears tmp and then creates a symlink inside it to
/sysroot/tmp:
tmp
└── tmp -> sysroot/tmp
This is likely a mistake and the root tmp should have pointed to
sysroot/tmp. However, since /tmp is mounted as a tmpfs anyways, we can
get rid of all this logic.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|/
|
|
|
|
|
|
| |
The environment variable OSTREE_BOOT_PARTITION is only used when using
GRUB. Move the export into the if statement. Also add a comment why
manually adding /boot/loader{.0} directory is necessary.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|\
| |
| | |
Feat/thud/2020.6
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| |
| |
| |
| | |
This allows you to see how much bandwidth was used with `systemd status
aktualizr`.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|/
|
|
|
|
| |
Now also works with zsh < 5.5
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
|\
| |
| | |
aktualizr: Bump to 2020.5 and get garage-sign from the new AWS bucket.
|
|/
|
|
|
|
|
| |
Also bump the garage-sign version. New version includes fixes for
supporting targets > 2 GB.
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
|\
| |
| | |
Feat/thud/2020.4
|
| |
| |
| |
| |
| |
| | |
Was changed recently
Signed-off-by: Laurent Bonnans <laurent.bonnans@here.com>
|
| |
| |
| |
| | |
Signed-off-by: Patrick Vacek <patrickvacek@gmail.com>
|
| |
| |
| |
| | |
Signed-off-by: Mykhaylo Sul <ext-mykhaylo.sul@here.com>
|