Customizations being nil isn't the same as an empty object. The
edge-installer types require customizations equalling nil.
This partially reverts bb486bcccb.
Writing the key to the user home directory has no effect for ostree
commits. Instead we write them using a fist-boot service.
In certain situations (e.g., when building an upgrade commit against an
existing parent), the user's home directory might not exist during the
build and the user key creation fails in the users stage, so let's
remove it entirely to avoid the issue.
Specifying a parent commit ID without a URL isn't allowed now. The
upgrade commit is built without a parent. The OS upgrade should still
work without it.
Getting the ID of the first commit is not necessary now.
Shorter and clearer error types.
- InvalidParameterError is now RefError. It is returned when a ref (the
new ref or the parent ref) is invalid.
- The ResolveRefError is not renamed. It is returned when a ref can't
be resolved (e.g., ref not found at URL).
New error type ParameterComboError. It is returned when a parent ref is
specified but no URL, which is an invalid parameter combination.
All parameters can now be specified at the same time.
See https://github.com/osbuild/osbuild-composer/issues/2131 for a
description of how the parameters are meant to interact.
In brief, the only invalid parameter combination is specifying a Parent
without a URL.
All other error conditions are for malformed URLs, invalid characters in
a ref, or error responses from the server.
Use the ostree package error types to keep the existing distinction
between Ref- and URL-related errors.
Introduce a new error condition for a general InvalidOSTreeParams
failure.
Errors implement the error interface, so we can return them as generic
errors. Callers can test the type only if they want to determine the
failure type.
The error types that can be returned from each function are documented
in the docstring.
No need to pass the entire image type. We just need the default ref.
This removes the distro package dependency from the ostree package,
which we will need so distro can use the ostree types and functions.
Adding three more combinations that weren't covered by previous tests:
- Supplying ref, parent, and URL: should result in an error
- Supplying ref and parent: OK
- Supplying parent, and URL: same as first case (ref gets default value
from image type)
Added default OSTreeRef() to test image type to cover the cases where
the ref isn't specified but affects the validation.
Separated and commented the test cases.
This was causing:
Feb 21 06:08:06 ip-10-30-18-43.us-east-1.aws.redhat.com osbuild-worker[20798]: time="2022-02-21T06:08:06Z" level=warning msg="Error reporting job result: error setting job status: 404 — Token not found (IMAGE-BUILDER-WORKER-5)"
Signed-off-by: Ondřej Budai <ondrej@budai.cz>
The new `GenImageKernelOptions` will retrun all needed kernel options
that are derived from the partition table. The only case - for now -
is if the partition table contains a LUKS decice, where the uuid of it
needs to be passed to the on the kernel command line so that it will
get unlocked.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
Use the new `GenImage{Prepare,Finish}Stages` in all liveImagePipelines
methods. This should allow us to handle partition tables with LUKS2 and
LVM2 transparently. Since none of the existing layouts contain any of
those, nothing should currently change.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
Add new helper functions that will return the necessary stages to
prepare a disk image, i.e. take care of creating the raw image
file, partitioning it, creating the devices on it (LUKS2, LVM2),
and formatting all the file systems: `GenImageStagesPrepare`.
Additionally, some partition layouts require some post-processing,
e.g. LVM2 where the volume group needs to be renamed "offline".
For this the `GenImageFinishStages` was added.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
This should be called at the end of the pipeline that creates an
image, to add stages that are needed to finish up the image.
Currently the only stage that will be added is the `lvm2.metadata`
stage in the case the partition layout contains LVM2.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
It is ok to not pass the creation time to osbuild, i.e. it can be
empty. If so, don't validate it.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
Instead of just passing in a single device, pass in a `Devices`
type, i.e. a map of devices, since the LVM can be located on a
LUKS2 container which is located inside a partition so multiple
devices are needed to access it.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
The only thing we needed from the passed in device was the filename,
so just pass in that directly.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
Verify that we get the necessary "luks2.format" and "lvm2.create"
stages for a partition table that has LVM2 on top of LUKS2.
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
Add a few test partition tables with well known configurations, like
LVM, LVM on top of LVM and Btrfs in addition to the plain layout that
is currently used in most partition tables. Add very basic checks,
such as that `NewPartitionTable` works and we have approximately the
correct size.
Co-Authored-By: Christian Kellner <christian@kellner.me>
Partition.Payload now supports every type of disk.Entity which enables
creating PartitionTables with LUKS, LVM, and Btrfs. \o/
Co-Authored-By: Achilleas Koutsou <achilleas@koutsou.net>
