Since 492b2c56, we create user and group upfront, now using 1000:1000 as
IDs. This can cause unexpected glitches when using the container without
kas-container in environments where older version already created files
with the previously used IDs. In order to stay compatible, switch the
default IDs back to 30000:30000.
This will not affect any user of kas-container.
Reported-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
As we no longer create the builder user at runtime, placing data into
/etc/skel at runtime is semantically not correct anymore. Instead, we
bind mount host paths below /var/kas/userdata. By that, we now place
the data into a directory which is fully handled by us.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch fixes a regression introduced in 492b2c5. As the builder user
is no longer created in the entrypoint, the data from the home skeleton
is also not copied anymore. This breaks the ssh config (including
known_hosts) when using kas-container with --ssh-dir, as the ssh dir is
mounted into the skeleton, but not copied to the builders home.
With this patch, we now explicitly copy the .ssh folder into the builder
users home, in case it is mounted from the host.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
With the added auto-caching logic of the referenced repos in 1c2c859,
kas-container has to mount this directory in read-write mode.
Otherwise, initial clones will fail.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Neither git nor hg currently provide a production-ready replacement for
weak SHA-1 commit IDs. Furthermore, kas mixes commit IDs and symbolic
commit names in refspec. This permits attackers who gained control over
a repository that kas fetches from to present manipulated content
without kas noticing this.
Aditya Sirish A Yelgundhalli recently reported one potential attack
scenario, using branches that shadow commit IDs. While trying to
mitigate this particular case, it became clear that there is no simple
solutions with the given tools and interfaces.
For now, warn prominently that only trusted sources should be used.
There are extensions planned to address the issue at its root, likely by
introducing content checksums.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
The user environment may lack /usr/sbin while certain podman
configuration will need, e.g., iptables for the setup. This can cause
Error: plugin type="bridge" failed (add): cni plugin bridge failed: failed to locate iptables: exec: "iptables": executable file not found in $PATH
Resolved that by appending /usr/sbin to the PATH in privileged podman
mode.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Dockerfile and container-entrypoint were missing that header.
Furthermore, the leading comment in Dockerfile got out-of-date, and we
should rather add section marker for the kas and kas-isar targets.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This is for the sake of the gitlab-ci runner which does not properly
aligns the ownership of the repo it checks out with the UID:GID of
our builder user. Reason not yet understood and hard to debug (logging
of the runner is incomplete).
Work around this issue by disabling safe.directory checks in case the
container is called without kas-container as wrapper (means, when it is
called without setting "--user=root"). This preserves git's checks for
the common interactive case, the more critical one.
Reported-by: Ross Burton <ross@burtonini.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Already create the builder user/group during container image build and
only align the IDs in the entrypoint if started with a non-zero USER_ID.
The primary gain is code simplification because this removes some
dynamics from the entrypoint.
As this refactoring avoids that gitlab-ci runners start the container as
root, it was also supposed to resolve the mismatch between the owner of
the checked-out repo and builder user. Unfortunately, this does not work
yet, and the reason is still unclear.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Leave the isar-only commands commented-out in the container-entrypoint
and simply remove the comment when building kas-isar. This is simpler,
more readable and also more robust against changes of the entrypoint
file.
While at it, avoid a separate layer for modifying container-entrypoint.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
The pattern
mkdir test_commands
rm -rf test_commands
makes no sense. If there were test_commands before, mkdir would have
failed. Simply build the path and copy the source content in.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This file shall describe the vulnerability disclosure process for kas
and the security context in which kas should be seen. Reporting
vulnerabilities via github has already been activated.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
For a long time already (c9326cc1ed), the documentation suggests a
different ordering of entries generated by *_conf_header than we
actually create. Finally fix this.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Was lost in the refactoring of a6b18abc8a.
Signed-off-by: Hannah Kiekens <hannah.kiekens@mind.be>
[Jan: refer to the causing commit]
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Since a6b18abc8a, we only have a single Dockerfile and a --target
instead.
We actually also need to touch the unmodified proxy line, wrapping it
around, so that doc8 remains happy.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
The current detection mechanism assumes that if the docker command
is available the engine behind is also docker. In fact nowadays a
lot of people use /usr/bin/docker as an alias for podman. In that
case the script expects a docker engine and misses to use
podman-specific settings.
To fix this, run "docker -v" and check if the output indicates a
podman engine or a real docker engine. Otherwise error out as the
enginge is not supported.
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a test for cloning with KAS_REPO_REF_DIR.
It explicitly tests the case that two repos with a single upstream URI
are added and also tests mercurial.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
[Jan: ensure Python 3.5 compatibility]
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a workaround for python versions < 3.8.
There, the dir created by TemporaryDirectory must still exist when
leaving the context manager.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch reworks the logic when setting KAS_REPO_REF_DIR.
When this variable is set, a two-staged clone is used:
First, a bare-clone (or similar) is created in the ref-dir, according to
the naming scheme. This clone is executed in a way that is both
reentrant as well as race-free across multiple instances of KAS working
on the same dir. Internally we clone into a tmpdir below the refdir and
rename on success to guarantee the atomicity of the operation on POSIX
compliant filesystems.
Second, the clone in the KAS_WORK_DIR is executed against the local
copy. After that, the origin url is redirected to the upstream url.
By that, the KAS_REPO_REF_DIR directory can be cached across builds
which significantly speedsup clone times against large repos.
In case the requested refspec is already in the cache (very likely in CI
builds), no direct access to the upstream repo is required. This logic
is crucial for CI systems in China, where e.g. access to github is
blocked from time to time.
The clone-from-local logic is currently only implemented for the git plugin
as HG misses the caching logic. Repo implementations that do not support this
logic can simply opt-out for the first stage by returning 'true'.
The existing user-facing logic of KAS_REPO_REF_DIR is not affected.
However, internally we no longer clone via --reference as this still
requires access to the remote repo, even if the requested commit is
already in the local copy.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch is a preparation for the cloning via reference logic.
Instead of handling the initial-clone specially, we just run the full
fetch_async logic. This is required as clones via a local mirror have a
different remote url which needs to be rewritten.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Only Ubuntu 20.04 is still supported with 3.5 and 3.6, so pin to it.
Expand the test to 3.10 and 3.11 at this chance. Quoting is needed now
to avoid that 3.10 becomes 3.1.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a test for the --resolve-env option in the dump plugin.
Both cases resolve and non-resolve are tested, each for yaml and json
output format.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds the option --resolve-env to the dump plugin.
When enabled, all variables in the 'env' section of the kas config file
are set to the captured value (at time of executing the dump plugin).
This helps to debug build issues on CI runners by precisely capturing
the relevant environment.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch extends the test of the dump plugin.
In addition, we test if relative refspecs are expanded and if the
generated output can be used as input to kas again.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a flag --resolve-refs to the dump plugin.
Once enabled, all relative refspecs (e.g. branch or tag names) are
resolved and replaced by their exact revision (before patches are
applied).
When re-running kas on the flattened and expanded config file, the build
is executed against exactly the versions of the dependencies that where
used when dumping the configuration. This helps to keep track of build
versions for future use (e.g. reproducible builds) or for
version-tracking tools like renovate.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a programmatic attribute 'revision' to the repo class.
This attribute contains the exact revision at the time of the checkout
of a repository. By that, we can avoid the ambiguity of refspecs
containing tags or branch names. Internally, the revision is not yet
used but just made available for future downstream users (e.g. plugins).
Note, that the revision has to be re-queried on each access, as the
Config class re-instantiates the repos for each consumer.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds support to use the dump plugin in kas-container.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a basic test of the dump plugin. We check if the configuration
is flattened and external references are included.
In addition, we check if no other files are referenced anymore.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch adds a new plugin 'dump' which resolves all kas config
includes and combines them in a single file. The format of the output
file can either be JSON or YAML. In case of YAML, the generated config
fulfills the kas-config schema and can be used as input to kas again.
In addition, the generated files can be used by external tools to
analyse the dependencies and versions of the project.
While the generated configuration is semantically identical to the
input config, we do not guarantee binary compatibility as especially
YAML provides multiple ways to serialize strings and null values.
Not giving this guarantee makes it easier to evolve and maintain the plugin.
The plugin itself extends the checkout plugin, as referenced repos have
to be checked-out first to resolve cross-repo references. This also
requires, that the declared refspec of all referenced repos can actually
be found and checked out.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
The kas-container trace / debug infrastructure has to write to stderr to
not interfer with kas output to stdout. Note, that the kas debug
messages already write to stderr. By that, the kas-container script can
be used in pipes similar to a direct invocation of kas.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch forwards the DISTRO_APT_PREMIRRORS environment variable into
the container when using the kas-container script. This variable is
commonly used to externally set mirrors for debian apt, e.g. in a CI
environment.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
These additional test cases check if the env section
can:
- export variables with their default value
- add variables with 'None' assigned only to
BB_ENV_PASSTHROUGH_ADDITIONS or to the deprecated
BB_ENV_EXTRAWHITE
- bitbake: make sure that variables actually make it into bitbake
The BB_ENV_EXTRAWHITE variable is still present, so a test case has also
been added for this.
Signed-off-by: Benedikt Niedermayr <benedikt.niedermayr@siemens.com>
[Jan: ensure python 3.5 compatibility, adjust copyright header]
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Until now all environment variables in the kas config's
env section require a default value.
In some cases we want to rely on the weak assignment
of bitbake instead of relying on the default value of the
variable in the kas config.
This means if we assign the 'None' value or nothing to the variable
in the kas config then it is only added to BB_ENV_EXTRAWHITE or
BB_ENV_PASSTHROUGH_ADDITIONS without setting it as environment
variable.
Signed-off-by: Benedikt Niedermayr <benedikt.niedermayr@siemens.com>
[Jan: fixed overlong line]
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch forwards environment variables to control the concurrency and
parallelism of the build to bitbake.
With that, the user can reduce the concurrency in case of running short
on memory. Further, by reducing the concurrency on very large and NUMA
machines, the build sometimes can be significantly speed up.
Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This patch prevents deployment of the pip cache folder that is only
needed during installation. Which is also done with the apt,
where at the end of the install operation we delete the
cache and all temporary folders.
The reducion of the image size is insignificant:
| builder@1373d0b95f44:/build$ sudo du -sh /root/.cache/pip
| 1.6M /root/.cache/pip
Signed-off-by: Jose Quaresma <jose.quaresma@foundries.io>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
The install package list are actually taking 1:1 from their documentation,
so there some packages that can already installed by other downstream layers.
This will not change any image sizes on all the layers in use.
Signed-off-by: Jose Quaresma <jose.quaresma@foundries.io>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Container engines per default log all container output into journald
(Podman) or a json file (Docker).
A build with kas-container/bitbake containes lots of useless build
output like progress lines, etc.
In case of Podman, where journald is used, this means that all this is
spammed into the journald buffer and possible also send to ttys or even
serial ttys. This might make sense for containers that output only
status information about running services, but for a build process, this
is mostly just spamming and causes important information to be lost.
Since the stdout/stderr output of the container is used for
kas-container ordinarily, the background logging from the container
image can be deactivated.
Therefor add the `--log-driver=none` parameter to all container
runtimes command lines.
Signed-off-by: Claudius Heine <ch@denx.de>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
This moves poky to 4.0.3 and Isar to the next revision of the day. The
latter is needed as we have no release with sbuild support yet.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
