mirror of
https://github.com/NVIDIA/OpenShell.git
synced 2026-10-19 05:50:41 +08:00
pull-request/3521
10
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Evan Lezar
|
e9271cb313 |
feat(gateway): support runtime image env overrides (#3504)
Closes #3502 Resolve trusted OCI runtime images with driver TOML, process environment, and compiled-default precedence across Docker, Podman, and Kubernetes. Signed-off-by: Evan Lezar <elezar@nvidia.com> |
||
|
Oliver Calder
|
6e865df349 |
feat(snap): ship the standalone prover binary in the snap (#3717)
Signed-off-by: Oliver Calder <oliver.calder@canonical.com> Signed-off-by: Evan Lezar <elezar@nvidia.com> |
||
|
Oliver Calder
|
1ad4e428a6 |
fix(snap): simplify snap hooks (#3988)
* fix(snap): simplify snap hooks
The `post-refresh` hook runs after initial snap installation as well, so
there is no need to call the `install` hook from within the
`post-refresh` hook; instead, the logic can simply be moved into the
`post-refresh` hook directly, and the `install` hook removed.
Also, the existing `install` hook logic looked for an insecure
configuration, and if found, replaced the entire configuration file with
a minimal default in the current format. But OpenShell does that default
behavior without any config file, so we may as well simply remove the
configuration file entirely to keep up-to-date with the current default
behavior. Let OpenShell create a configuration file if it needs to,
rather than auto-create one via the packaging scripts.
Signed-off-by: Oliver Calder <oliver.calder@canonical.com>
* fix(snap): remove the connect-plug-docker hook
The `openshell:docker` is auto-connected to the system `:docker` slot,
so there should not be a need to separately restart the gateway service
when the interface is connected.
For locally-built test snaps which were not published to the store, the
autoconnection is not made, but when the snap is installed, the gateway
will attempt to start anyway and fail to find any available compute
driver, so quickly restart until it hits the systemd start-limit, after
which systemd prevents the service from being started again. If a user
tries to manually connect their locally-built `openshell` snap to the
`:docker` slot, then the `connect-plug-docker` hook runs and triggers a
restart of the gateway, which will usually fail because the start limit
has already been hit. An error in the hook will thus cause the interface
connection to be undone, which is undesirable.
Thus, we can remove this hook entirely, and instead allow interface
connections to succeed as intended. The user still needs to manually
restart the gateway service after making a manual connection (as was the
case previously) and probably needs to `systemctl reset-failed` first,
but at least connection will succeed beforehand so they can proceed with
these steps.
Signed-off-by: Oliver Calder <oliver.calder@canonical.com>
* fix(snap): set refresh-mode: endure again, with manual restart
Return to the previous behavior before commit
|
||
|
Drew Newberry
|
a67567e583 |
fix(snap): require mTLS for the snap gateway (#3726)
* fix(snap): require mTLS for the snap gateway Replace the installer opt-in with an authenticated snap gateway. The wrapper no longer forces plaintext, so the gateway serves TLS from the bundle it already generates in $SNAP_COMMON/tls. The install hook writes a config that enables mTLS user auth instead of unauthenticated access, and a new post-refresh hook migrates the exact legacy default on existing installs. install.sh waits for the gateway, detects whether it serves TLS, copies the client bundle into the target user's snap state directory, and registers the gateway over HTTPS. Older plaintext snap revisions still register over HTTP with a warning. The release canary asserts mTLS auth and HTTPS registration. Signed-off-by: Drew Newberry <anewberry@nvidia.com> * fix(snap): pass config preflight and detect the mTLS gateway reliably An explicit [openshell.gateway.mtls_auth] table fails config preflight, which validates mTLS auth before the local TLS bundle supplies the client CA. Write a default that pins the Docker driver instead; with the wrapper's TLS bundle the gateway requires client certificates and enables mTLS user auth automatically, as the native packages do. The mTLS gateway rejects TLS handshakes without a client certificate, and it still answers plaintext loopback HTTP for sandbox service routing, so the installer could misdetect it as a legacy plaintext gateway. Probe HTTPS with the root-owned client bundle, and treat a gateway as legacy only when a plaintext gRPC Health call succeeds. Signed-off-by: Drew Newberry <anewberry@nvidia.com> * chore(snap): simplify install hook comment Signed-off-by: Drew Newberry <anewberry@nvidia.com> * fix(snap): migrate insecure gateway configs on refresh Signed-off-by: Drew Newberry <anewberry@nvidia.com> * refactor(snap): simplify mTLS detection and config migration Detect the mTLS snap from the installed revision's post-refresh hook instead of probing plaintext gRPC, and drop the scheme global. Remove the installer's pre-hook config fallback, which is dead now that every channel ships the install hook and which wrote the insecure default. Give the install hook a single write path with a simple backup name, and shorten the manual client certificate steps. Signed-off-by: Drew Newberry <anewberry@nvidia.com> * fix(snap): stop keeping a copy of replaced insecure configs Signed-off-by: Drew Newberry <anewberry@nvidia.com> * feat(install): make the snap an opt-in install method Stop selecting the OpenShell snap just because the snap command exists. Linux installs default to the Debian or RPM package; OPENSHELL_INSTALL_METHOD=snap (or deb, rpm) selects the package explicitly. Hosts that already have the OpenShell snap keep refreshing it rather than gaining a second gateway on the same port. The release canary and snap repro script opt in explicitly. Signed-off-by: Drew Newberry <anewberry@nvidia.com> * fix(snap): let the gateway auto-detect its compute driver Signed-off-by: Drew Newberry <anewberry@nvidia.com> * fix(snap): restart the gateway after refresh Published revisions use refresh-mode: endure, and snapd honors the old revision's setting during a refresh, so the plaintext gateway kept running with the migrated config unused until a manual restart. Restart the gateway from the post-refresh hook so the mTLS config takes effect immediately. Signed-off-by: Drew Newberry <anewberry@nvidia.com> * docs(snap): drop refresh notes from the snap description Signed-off-by: Drew Newberry <anewberry@nvidia.com> * docs(snap): trim snap refresh notes from installation docs Signed-off-by: Drew Newberry <anewberry@nvidia.com> --------- Signed-off-by: Drew Newberry <anewberry@nvidia.com> |
||
|
Oliver Calder
|
6c864ec9ab |
fix(install): avoid installing incompatible docker snap (#3666)
* fix(install): avoid installing incompatible docker snap The work to land RFC-0012 added new restrictions when interacting with Docker by setting `NoNewPrivs`. This prevents the `docker` snap from transitioning its AppArmor profile from `snap.docker.dockerd` to `docker-default` when it tries to launch a container. Thus, the `docker` snap is currently incompatible with OpenShell. This commit prevents `install.sh` from installing the `docker` snap before installing the `openshell` snap, and instead requires the user to install a non-snap Docker daemon before proceeding with installing the snap. Systems without the `snap` command are unaffected, since they install native packages without checking for the presence of Docker or other compute providers. Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(snap): only snapd 2.76 for openshell snap since store installs work Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fixup! fix(install): avoid installing incompatible docker snap Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fixup! fix(snap): only snapd 2.76 for openshell snap since store installs work Signed-off-by: Oliver Calder <oliver.calder@canonical.com> --------- Signed-off-by: Oliver Calder <oliver.calder@canonical.com> |
||
|
Oliver Calder
|
e60098d748 |
fix(snap): install openshell snap via install.sh when snap available (#3656)
* fix(snap): update stale snap docs and tests Previously, the `openshell` snap required the `docker` snap. Now, it works with any Docker daemon running on the system. Furthermore, the `snap-declaration` assertion on the `openshell` snap when installed from the Snap Store causes the `openshell` snap to always connect to the system `:docker` slot, rather than a slot provided by the `docker` snap. This commit updates the documentation, including the `description` field in `snapcraft.yaml`, to ensure that all information is correct and up-to-date. Additionally, some tests connected the `openshell:docker` plug to the `docker` snap's `docker:docker-daemon` slot, which is inconsistent with how the `openshell` snap operates when installed from the store. Update those tests to connect to the system `:docker` slot as well. Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(snap): require snapd 2.76 for openshell snap Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(nix): align snap gateway reproducer timeout with release-canary Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(snap): require snapd 2.77 for openshell snap Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fixup! fix(snap): require snapd 2.77 for openshell snap Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * feat(snap): install openshell snap via install.sh when snap available Change `install.sh` to install the `openshell` snap by default when snapd is installed on the host. This installs the snap from the `latest/stable` channel, which should match the most up-to-date release tag on github. If `OPENSHELL_VERSION=dev` is set for `install.sh`, then it will install the `openshell` snap from the `latest/edge` channel, which matches the latest dev release available on github. The `openshell` snap currently requires Docker in order to function. If a Docker daemon is already installed on the system, it will be used by the `openshell` snap. Otherwise, `install.sh` will install the `docker` snap first, wait for the Docker daemon to be ready, and then install the `openshell` snap. Also, update the `release-canary.yml` to split the `ubuntu-snap` job into `ubuntu-snap-system-docker` and `ubuntu-snap-provisions-docker`, which test the two aforementioned scenarios. Previously, `ubuntu-snap` manually installed a given snap artifact as built from CI, but with these new jobs, it instead uses `install.sh` to install the published `openshell` snap from the `latest/edge` track, thus matching the behavior of the other release canary jobs. Make corresponding changes to the `nix` guest reproducer, documentation, and tests. Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(snap): configure local gateway authentication The `openshell` snap runs the gateway as a systemd system service, which runs as root. Thus, the mTLS certs are generated by root and stored in a root-owned directory to which non-root users do not have access. For this reason, the snap's `openshell-gateway-wrapper` script sets `OPENSHELL_DISABLE_TLS=true`. This commit ensures that the `openshell` snap's gateway allows unauthenticated local access by writing a default `gateway.toml` configuration file during the install hook, which runs after the snap is first installed but before services are started. The config file contains sets `allow_unauthenticated_users = true`. Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fix(install): configure snap gateway authentication Recently, a new install hook was added which writes a default config file for the `openshell` snap to allow unauthenticated local access to the gateway. This is because the gateway service runs as root and the mTLS certificates are not accessible to non-root users. However, the `install.sh` script installs the `openshell` snap from the snap store, and the published version may not yet have that new install hook. Or, the user may already have the snap installed, in which case the install hook does not run. In either case, we need `install.sh` to ensure that the config file is written to set the gateway auth to `allow_unauthenticated_users = true`, and then restart the openshell gateway service. Signed-off-by: Oliver Calder <oliver.calder@canonical.com> * fixup! feat(snap): install openshell snap via install.sh when snap available Signed-off-by: Oliver Calder <oliver.calder@canonical.com> --------- Signed-off-by: Oliver Calder <oliver.calder@canonical.com> |
||
|
Drew Newberry
|
d6f3e1f5f2 |
fix(packaging): keep SPDX comments out of Debian control (#3483)
Signed-off-by: Drew Newberry <anewberry@nvidia.com> |
||
|
Johnny Greco
|
58b5f8f976 |
feat(prover): add standalone policy boundary checker (#3289)
* feat(prover): add standalone policy maximum checker Signed-off-by: Johnny Greco <jogreco@nvidia.com> * refactor(prover): simplify check scope schema Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): align containment and cancellation with runtime Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): stabilize containment checks in CI Signed-off-by: Johnny Greco <jogreco@nvidia.com> * test(prover): avoid solver in fast-path guard test Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): align string containment with runtime Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): reject ambiguous z3 string escapes Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): align containment with runtime boundaries Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover-cli): harden cancellation and invalid input Signed-off-by: Johnny Greco <jogreco@nvidia.com> * feat(packaging): install policy prover with OpenShell Signed-off-by: Johnny Greco <jogreco@nvidia.com> * docs(prover): clarify installation and check results Signed-off-by: Johnny Greco <jogreco@nvidia.com> * docs(build): describe prover distribution directly Signed-off-by: Johnny Greco <jogreco@nvidia.com> * refactor(prover): rename maximum policy to boundary Signed-off-by: Johnny Greco <jogreco@nvidia.com> * refactor(prover): localize fail-closed validation Signed-off-by: Johnny Greco <jogreco@nvidia.com> * test(prover): cover fail-closed CLI surfaces Signed-off-by: Johnny Greco <jogreco@nvidia.com> * test(prover): allow CI load for REST solver proof Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): use canonical policy schema for containment Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): preserve uncertainty for runtime binary globs Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): bound policy validation work Signed-off-by: Johnny Greco <jogreco@nvidia.com> * docs(prover): document validation resource limits Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(prover): make containment API extensible Signed-off-by: Johnny Greco <jogreco@nvidia.com> * docs(prover): define containment API contract Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(ci): integrate prover with consolidated builds Signed-off-by: Johnny Greco <jogreco@nvidia.com> * fix(ci): declare release packaging dependency Signed-off-by: Johnny Greco <jogreco@nvidia.com> --------- Signed-off-by: Johnny Greco <jogreco@nvidia.com> |
||
|
Jesse JaggarsandDrew Newberry
|
02b664bb0d |
refactor(config): normalize and enforce gateway schema v2 (#2814)
* refactor(config): normalize compute driver field names Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * refactor(config): introduce canonical gateway fields Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * refactor(config): enforce gateway schema version 2 Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): preserve compute driver runtime guarantees Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): address schema v2 review regressions Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): complete schema v2 migration safeguards Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(config): expand schema v2 regression coverage Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(config): add schema v2 parity manifest Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): correct parity manifest inventory Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * docs(config): record schema v2 intentional changes Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * docs(config): disposition schema v2 parity gaps Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): add dual schema parity harness Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): establish compute lifecycle parity baseline Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): preserve gateway option compatibility Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): record gateway option parity Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * docs(config): close gateway-wide parity gaps Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(podman): apply configured pids limit Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): validate Podman option parity Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): add Kubernetes option parity harness Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): record Kubernetes option parity Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): disposition VM parity lanes Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): add external driver parity lane Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(e2e): preserve external driver pull policy Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): attest parity artifacts and launches Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): require clean parity build sources Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): bind parity runtime artifacts Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(e2e): use isolated supervisor tags Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(e2e): qualify parity image tags Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(e2e): serve parity supervisor locally Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): isolate parity podman services Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): harden parity evidence provenance Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): pin parity sandbox artifacts Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): attest parity runtime inputs Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): bind parity runtime evidence Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): record compute boundary parity Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(e2e): disposition cross-cutting parity lanes Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(packaging): preflight gateway config upgrades Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): preserve rebase integration guarantees Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(ci): isolate temporary git signing config Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): update remaining schema v2 consumers Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(ci): provide e2fs tools to VM tests Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): align preflight with gateway startup Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(vm): preserve rootfs tar configuration Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * chore(config): adopt duration unit constructors Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(packaging): preflight RPM gateway config Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(config): address driver review findings Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(e2e): require fresh semantic parity evidence Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * fix(docker): update tests for renamed sandbox label Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> * test(gateway): preserve selective driver coverage after rebase Signed-off-by: Drew Newberry <anewberry@nvidia.com> --------- Signed-off-by: Jesse Jaggars <jjaggars@redhat.com> Signed-off-by: Drew Newberry <anewberry@nvidia.com> Co-authored-by: Drew Newberry <anewberry@nvidia.com> |
||
|
Taylor Mutch
|
7cea9d9b2b |
fix(gateway): align package TLS bootstrap path (#1601)
* fix(gateway): align package TLS bootstrap path Closes #1593 Default package-managed gateway services to a stable local TLS directory and use that same value for certificate generation and runtime startup. Signed-off-by: Taylor Mutch <taylormutch@gmail.com> * test(packaging): validate package asset paths exist Signed-off-by: Taylor Mutch <taylormutch@gmail.com> * ci(e2e): pin mise in kubernetes job Signed-off-by: Taylor Mutch <taylormutch@gmail.com> --------- Signed-off-by: Taylor Mutch <taylormutch@gmail.com> |