Auto-generate policy for nginx-deployment pods, instead of hard-coding
the "allow all" policy.
Note that the `busybox_pod` - created using `kubectl run` - still
doesn't have an Init Data annotation, so it is using the default policy
built into the Kata Guest rootfs image file.
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
Auto-generate agent policy in k8s-liveness-probes.bats, instead of using
the non-confidential "allow all" policy.
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
Auto-generate the agent policy for pod-secret-env.yaml, using
"genpolicy -c inject_secret.yaml".
Support for passing Secret specification files as "-c" arguments of
genpolicy has been added when fixing #10033 with PR #10986.
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
This fixes that error everywhere by adding a `name:` field to all jobs that
were missing it. We keep the same name as the job ID to ensure no
disturbance to the required job names.
Signed-off-by: Aurélien Bombo <abombo@microsoft.com>
-o pipefail in particular ensures that exec_host() returns the right exit
code.
-u is also added for good measure. Note that $BATS_TEST_DIRNAME is set by
bats so we move its usage inside the function.
Signed-off-by: Aurélien Bombo <abombo@microsoft.com>
As a consequence of moving away from Advanced Security for Zizmor, it now
checks the entire codebase and will error out on this PR and future.
To be reverted once we address all Zizmor findings in a future PR.
Signed-off-by: Aurélien Bombo <abombo@microsoft.com>
The Hadolint warning DL3007 (pin the version explicitly) is no
longer applicable.
We have updated the base image to use a specific version
digest, which satisfies the linter's requirement for reproducible
builds. This commit removes the corresponding inline ignore comment.
Signed-off-by: Hyounggyu Choi <Hyounggyu.Choi@ibm.com>
This change enables to run the QEMU VMM using a non-root user when rootless flag is set true in the configuration.
Signed-off-by: stevenfryto <sunzitai_1832@bupt.edu.cn>
This commit introduces generic support for running the VMM in rootless mode in runtime-rs:
1.Detect whether the VMM is running in rootless mode.
2.Before starting the VMM process, create a non-root user and launch the VMM with that user’s UID and GID; also add the KVM user's group ID to the VMM process's supplementary groups so the VMM process can access /dev/kvm.
3.Add the setup of the rootless directory located in the dir /run/user/<uid> directory, and modify some path variables to be functions that return the path with the rootless directory prefix when running in rootless mode.
Fixes: #11414
Signed-off-by: stevenfryto <sunzitai_1832@bupt.edu.cn>
We recently hit the following error during build:
```
RUN ssh-keygen -t ed25519 -f /etc/ssh/ssh_host_ed25519_key -P ""
OpenSSL version mismatch. Built against 3050003f, you have 30500010
```
This happened because `alpine:latest` moved forward and the `ssh-keygen`
binary in the base image was compiled against a newer OpenSSL version
that is not available at runtime.
Pinning the base image to the stable release (3.20) avoids the mismatch
and ensures consistent builds.
Signed-off-by: Hyounggyu Choi <Hyounggyu.Choi@ibm.com>
This change fixes clean up logic when running tests
in a vm booted with qemu wrt to qmp.sock & console.sock
files, and no longer assumes any path for them.
Signed-off-by: Sumedh Alok Sharma <sumsharma@microsoft.com>
Change the default block driver to virtio-scsi.
Since the latest qemu's commit:
https://gitlab.com/qemu-project/qemu/-/commit/
984a32f17e8dab0dc3d2328c46cb3e0c0a472a73
brings a bug for virtio-blk-pci with io_uring mode at line:
https://gitlab.com/qemu-project/qemu/-/commit/
984a32f17e8dab0dc3d2328c46cb3e0c0a472a73#
ce8eeb01f8b84f8cb8d3c35684d473fe1ee670f9_345_352
In order to avoid this issue, change the default block driver
to virtio-scsi.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
As OCI Spec annotation has been updated with adding or remove items,
we should use the updated annotation as the passed argument.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
This commit removes the InitData annotation from the OCI Spec's
annotations.
Similar to the Policy annotation, InitData is now exclusively handled
and transmitted to the guest via the sandbox's init data mechanism.
Removing this redundant and potentially large annotation simplifies the
OCI Spec and streamlines the guest initialization process.
This change aligns the handling of InitData with existing practices
within runtime-go.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
The repro below would show this error in the logs (in debug mode only):
fatal runtime error: IO Safety violation: owned file descriptor already closed
The issue was that the `pseudo.slave` file descriptor was being owned by
multiple variables simultaneously. When any of those variables would go out
of scope, they would close the same file descriptor, which is undefined
behavior.
To fix this, we clone: we create a new file descriptOR that refers to the same
file descriptION as the original. When the cloned descriptor is closed, this
affect neither the original descriptor nor the description. Only when the last
descriptor is closed does the kernel cleans up the description.
Note that we purposely consume (not clone) the original descriptor with
`child_stdin` as `pseudo` is NOT dropped automatically.
Repro
-----
Prerequisites:
- Use Rust 1.80+.
- Build the agent in debug mode.
$ cat busybox.yaml
apiVersion: v1
kind: Pod
metadata:
name: busybox
spec:
containers:
- image: busybox:latest
name: busybox
runtimeClassName: kata
$ kubectl apply -f busyboox.yaml
pod/busybox created
$ kubectl exec -it busybox -- sh
error: Internal error occurred: Internal error occurred: error executing
command in container: failed to exec in container: failed to start exec
"e6c602352849647201860c1e1888d99ea3166512f1cc548b9d7f2533129508a9":
cannot enter container 76a499cbf747b9806689e51f6ba35e46d735064a3f176f9be034777e93a242d5,
with err ttrpc: closed
Fixes: #11054
Signed-off-by: Aurélien Bombo <abombo@microsoft.com>
Log how much time "kubectl get pods" and each test case are taking,
just in case that will reveal unusually slow test clusters, and/or
opportunities to improve tests.
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
In the previous commit we've added some code that broke `cargo fmt --
--check` without even noticing, as the code didn't go through the CI
process (due to it being a security advisory).
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
Trailing slash in DEFAULT_KATA_GUEST_SANDBOX_DIR caused double slashes
in mount_point (e.g. "/run/kata-containers/sandbox//shm"), which failed
OPA strict equality checks against policy mount_point. Removing it aligns
generated paths with policy and fixes CreateSandboxRequest denial.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
coco-guest-components tarball is used as is for both vanilla coco
rootfs and the nvidia enabled rootfs. nvidia-attester can be built
without nvml so make it globally enabled for coco-guest-components.
Signed-off-by: Mikko Ylinen <mikko.ylinen@intel.com>
`allow_interactive_exec` requires a sandbox-name annotation, however
this is only added for pods by genpolicy. Other pod-generating resources
have unpredictable sandbox names.
This patch instead uses a regex for the sandbox name in genpolicy, based
on the specified metadata and following Kubernetes' naming logic. The
generated regex is then used in the policy to correctly match the
sandbox name.
Fixes: #11823
Signed-off-by: Charlotte Hartmann Paludo <git@charlotteharludo.com>
Co-authored-by: Paul Meyer <katexochen0@gmail.com>
Co-authored-by: Markus Rudy <mr@edgeless.systems>
This commit addresses an issue where base64 output, when used with a
default configuration, would introduce newlines, causing decoding to
fail on the runtime.
The fix ensures base64 output is a single, continuous line using the -w0
flag. This guarantees the encoded string is a valid Base64 sequence,
preventing potential runtime errors caused by invalid characters.
Note that: When you use the base64 command without any parameters, it
typically automatically adds newlines to the output, usually every 76 chars.
In contrast, base64 -w0 explicitly tells the command not to add any
newlines (-w for wrap, and 0 for a width of zero), which results in a
continuous string with no whitespace.
This is a critical distinction because if you pass a Base64 string with
newlines to a runtime, it may be treated as an invalid string, causing
the decoding process to fail.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
