* terraform: add Azure marketplace variable
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* config: add Azure marketplace variable
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* cli: use Terraform variables from config
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* terraform: pass down marketplace variable
* image: pad Azure images to 1GiB
* terraform: add version attribute to marketplace image
* semver: allow versions to be exported without prefix
* cli: boolean var to use marketplace images
* config: remove dive key
* dev-docs: add instructions on how to use marketplace images
* terraform: fix unit test
* terraform: only fetch image for non-marketplace images
* mpimage: refactor image selection
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* [remove] increase minor version for image build
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* terraform: ignore changes to source_image_reference on upgrade
* operator: add support for parsing Azure marketplace images
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* upgrade: fix imagefetcher call
* docs: add info about azure marketplace
* image: ensure more than 1GiB in size
* image: test to pad to 2GiB
* version: change back to v2.14.0-pre
* image: GPT-conformant image size padding
* [remove] increase version
* mpimage: inline prefix func
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* ci: add marketplace image e2e test
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
* [remove] register workflow
* ci: fix workflow name
* ci: only allow azure test
* cli: add marketplace image input to interface
* cli: fix argument passing
* version: roll back to v2.14.0
* ci: add force-flag support
* Update docs/docs/overview/license.md
* Update dev-docs/workflows/marketplace-images.md
Co-authored-by: Moritz Eckert <m1gh7ym0@gmail.com>
---------
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
Co-authored-by: Moritz Eckert <m1gh7ym0@gmail.com>
Co-authored-by: Thomas Tendyck <51411342+thomasten@users.noreply.github.com>
* Update CI to use different GCP project for e2e tests
* Update GCP image project service accounts
* Update default GCP bucket name for image builds
---------
Signed-off-by: Daniel Weiße <dw@edgeless.systems>
For reproducibility reasons, the final OS image does not ship the rpm database in sqlite format.
For supply chain security and license compliance reasons, we want to keep the rpm database of os images as a detached build artifact.
We now ship a reproducible, human readable manifest of installed rpms in the image under "/usr/share/constellation/packagemanifest" and upload the full rpm database as a build artifact (rpmdb.tar).
We don't want these options on other CSPs. This is temporary until AWS
fixed some background issues.
We need to set the option we want to set differently on each provider
once per provider as we need to keep some of the options we set with
higher priority.
Disabling SMT dynamically inside the image creates problems on AWS.
The problem should be fixed by disabling smt through the VMM.
By recommendation from AWS: add idle=poll.
This should improve our launch success rate while they investigate some
upstream issues.
* rename to attestationconfigapi + put client and fetcher inside pkg
* rename api/version to versionsapi and put fetcher + client inside pkg
* rename AttestationConfigAPIFetcher to Fetcher
* api: rename AttestationVersionRepo to Client
* api: move client into separate subpkg for
clearer import paths.
* api: rename configapi -> attestationconfig
* api: rename versionsapi -> versions
* api: rename sut to client
* api: split versionsapi client and make it public
* api: split versionapi fetcher and make it public
* config: move attestationversion type to config
* api: fix attestationconfig client test
Co-authored-by: Adrian Stobbe <stobbe.adrian@gmail.com>
Kernel cmdline embedded in UKIs had no null terminator before. With newer versions of mkosi, it is already null-terminated so we shouldn't null terminate it twice.
