The GitHub hosted runners for ARM64 do not provide virtualisation
support, thus we're just skipping the tests as those would check whether
or not the system is "VMContainerCapable".
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
To support virtio-net-ccw for s390x, add CCW devices to the Endpoint
interface. Add respective fields and functions to implementing structs.
Device paths may be empty. PciPath resolves this by being a list that
may be empty, but this design does not map to CcwDevice. Use a pointer
instead.
Signed-off-by: Jakob Naucke <jakob.naucke@ibm.com>
Field is being used for both PCI and CCW devices. Name it devicePath
to avoid confusion when the device isn't a PCI device.
Signed-off-by: Jakob Naucke <jakob.naucke@ibm.com>
Adding devices by CDI annotation can fail for a variety of reasons. If
that happens, it's helpful to know the root cause of the issue (CDI spec
missing, malformatted, requested device not present, etc.).
This commit adds the root cause of the CDI device addition to the errors
reported back to the caller. Since this error is bubbled up all the way
back to the shimv2 task.Create handler, it will be visible in Kubernetes
logs and enable fixing the root cause.
Signed-off-by: Markus Rudy <mr@edgeless.systems>
We need a proper ID otherwise QEMU sometimes fails with invalid ID.
Use the same pattern as with the old VFIO implementation.
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
For each IOMMUFD device create an object and assign
it to the device, we need additional information that
is populated now correctly to decide if we run the old VFIO
or new VFIO backend.
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
With the create_container_timeout the dial_timeout is lest important.
Add the custom timeout for GPUs in create_container_timeout
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
use upstream qemu in snp and nvidia snp configs.
load ovmf with bios flag on qemu cmdline instead of file.
Fixes: #10750
Signed-Off-By: Ryan Savino <ryan.savino@amd.com>
snp standard attestation with the upstream kernel and qemu do not support extended attestation with certs.
Fixes: #10750
Signed-Off-By: Ryan Savino <ryan.savino@amd.com>
This commit introduces a new gRPC device type, `vfio-ap-cold`, to support
VFIO-AP coldplug. This enables the VM guest to handle passthrough devices
differently from VFIO-AP hotplug.
With this new type, the guest no longer needs to wait for events (e.g., device
addition) because the device already exists at the time the device type is checked.
Signed-off-by: Hyounggyu Choi <Hyounggyu.Choi@ibm.com>
Even though ociSpec.Linux.Devices is preserved when vfio_mode is VFIO,
it has not been updated correctly for coldplug scenarios. This happens
because the device info passed to the agent via CreateContainerRequest
is dropped by the Kata runtime.
This commit ensures that the device info is added to the sandbox's
device manager when vfio_mode is VFIO and coldPlugVFIO is true
(e.g., vfio-ap-cold), allowing ociSpec.Linux.Devices to be properly
updated with the device information before the container is created on
the guest.
Signed-off-by: Hyounggyu Choi <Hyounggyu.Choi@ibm.com>
Bumps the go_modules group with 1 update in the /src/runtime directory: [golang.org/x/net](https://github.com/golang/net).
Bumps the go_modules group with 1 update in the /src/tools/csi-kata-directvolume directory: [golang.org/x/net](https://github.com/golang/net).
Bumps the go_modules group with 1 update in the /tools/testing/kata-webhook directory: [golang.org/x/net](https://github.com/golang/net).
Updates `golang.org/x/net` from 0.25.0 to 0.33.0
- [Commits](https://github.com/golang/net/compare/v0.25.0...v0.33.0)
Updates `golang.org/x/net` from 0.23.0 to 0.33.0
- [Commits](https://github.com/golang/net/compare/v0.25.0...v0.33.0)
Updates `golang.org/x/net` from 0.23.0 to 0.33.0
- [Commits](https://github.com/golang/net/compare/v0.25.0...v0.33.0)
---
updated-dependencies:
- dependency-name: golang.org/x/net
dependency-type: indirect
dependency-group: go_modules
- dependency-name: golang.org/x/net
dependency-type: direct:production
dependency-group: go_modules
- dependency-name: golang.org/x/net
dependency-type: indirect
dependency-group: go_modules
...
Signed-off-by: dependabot[bot] <support@github.com>
With newer kernels we have a new backend for VFIO
called IOMMUFD this is a departure from VFIO IOMMU Groups
since it has only one device associated with an IOMMUFD entry.
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
Skip logging empty lines of text from the Guest console output, if
there are any such lines.
Without this change, the Guest console log from CLH + /dev/pts/0 has
twice as many lines of text. Half of these lines are empty.
Fixes: #10737
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
Since
be93fd5372,
which is included in QEMU since version 9.2.0, the options for the
`device_add` QMP command need to be typed correctly.
This makes it so that instead of `"on"`, the value is set to `true`,
matching QEMU's expectations.
This has been tested on QEMU 9.2.0 and QEMU 9.1.2, so before and after
the change.
The compatibility with incorrectly typed options for the `device_add`
command is deprecated since version 6.2.0 [^1].
[^1]: https://qemu-project.gitlab.io/qemu/about/deprecated.html#incorrectly-typed-device-add-arguments-since-6-2
Signed-off-by: Moritz Sanft <58110325+msanft@users.noreply.github.com>
This commit sets memory config `shared` to false in cloud hypervisor
when creating vm with shared_fs=None && hugePages = false.
Currently in runtime/virtcontainers/clh.go,the memory config shared is by default set to true.
As per the CLH memory document,
(a) shared=true is needed in case like when using virtio_fs since virtiofs daemon runs as separate process than clh.
(b) for shared_fs=none + hugespages=false, shared=false can be set to use private anonymous memory for guest (with no file backing).
(c) Another memory config thp (use transparent huge pages) is always enabled by default.
As per documentation, (b) + (c) can be used in combination.
However, with the current CLH implementation, the above combination cannot be used since shared=true is always set.
Fixes#10547
Signed-off-by: Sumedh Alok Sharma <sumsharma@microsoft.com>
Bump some actions that are significantly out-of-date
and out of sync with the versions used in other workflows
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
With full debug logging enabled there might be around 1,500 redials
so log just ~15 of these redials to avoid flooding the log.
Signed-off-by: Dan Mihai <dmihai@microsoft.com>
This reverts commit f15e16b692, as we
don't have to do this since we're relying on the
`static_sandbox_resource_mgmt` feature, which gives us the correct
amount of memory and CPUs to be allocated.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
The qemu-coco-dev runtime class should be as close as possible to what
the TEEs runtime classes are doing, and this was one of the options that
ended up overlooked till now.
Shout out to Dan Mihai for noticing that!
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
Add GPU annotations for remote hypervisor to help
with the right instance selection based on number of GPUs
and model
Signed-off-by: Pradipta Banerjee <pradipta.banerjee@gmail.com>
We may decide to add this later on, but for now this is only targetting
TEEs and the confidential image / initrd.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
Return of proper error to the initiator is not guaranteed.
Method StopVM could kill shim process together with VM pieces.
Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
When QEMU is terminated by signal 15, it deletes the PidFile.
Upon detecting that QEMU has exited, the shim executes the stopVM function.
If the PidFile is not found, the PID is set to 0.
Subsequently, the shim executes `kill -9 0`, which terminates the current process group.
This prevents any further logic from being executed, resulting in resources not being cleaned up.
Signed-off-by: wangyaqi54 <wangyaqi54@jd.com>
When using network adapters that support SR-IOV, a VFIO device can be
plugged into a guest VM and claimed as a network interface. This can
significantly enhance network performance.
Fixes: #9758
Signed-off-by: Lei Huang <leih@nvidia.com>
As we don't have any CI, nor maintainer to keep ACRN code around, we
better have it removed than give users the expectation that it should or
would work at some point.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
kata-shim was not reporting `inactive_file` in memory stat.
This memory is deducted by containerd when calculating the size of container working set, as it can be paged out by the operating
system under memory pressure. Without reporting `inactive_file`, containerd will over report container memory usage.
[Here](https://github.com/containerd/containerd/blob/v1.7.22/pkg/cri/server/container_stats_list_linux.go#L117) is where containerd
deducts `inactive_file` from memory usage.
Note that kata-shim correctly reports `total_inactive_file` for cgroup v1, but this was not implemented for cgroup v2.
This commit:
- Adds code in kata-shim to report "inactive_file" memory for cgroup v2
- Implements reporting of all available cgroup v2 memory stats to containerd
- Uses defensive coding to avoid assuming existence of any memory.stat fields
The list of available cgroup v2 memory stats defined by containerd can be found
[here](https://pkg.go.dev/github.com/containerd/cgroups/v2/stats#MemoryStat).
Fixes#10280
Signed-off-by: Alex Man <alexman@stripe.com>
It will panic when users do GPU vfio passthrough with cdi in runtime.
The root cause is that CustomSpec.Annotations is nil when new element
added.
To address this issue, initialization is introduced when it's nil.
Fixes#10266
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
I know right now we're always passing a value for that, but this doesn't
really have to be set unless attestation is used. Thus, let's also omit
it in case it's empty.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
This is a quick and simple pre-req for supporting initData, which will
take advantage of the mrconfigid in the TDX case.
While already adding mrconfigid, which is hardcoded empty right now,
let's do the same for mrowner and mrownerconfig, and leave it prepared
for future expansions.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
The reason we're relying on yet another function to do so is because the
TDX object will be used in its qom / qapi json format.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
This error is specific to SNP platforms, so let's make sure we only
error this out when an SNP platform is used.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
...or by using a binary with additional suffix.
This allows having multiple versions of nydus-overlayfs installed on the
host, telling nydus-snapshotter which one to use while still detecting
Nydus is used.
Signed-off-by: Paul Meyer <49727155+katexochen@users.noreply.github.com>
PhysicalEndpoint unbinds its VF interface and rebinds it as a VFIO device,
then cold-plugs the VFIO device into the guest kernel.
When `cold_plug_vfio` is set to "no-port", cold-plugging the VFIO device
will fail.
This change checks if `cold_plug_vfio` is enabled before creating PhysicalEndpoint
to avoid unnecessary VFIO rebind operations.
Fixes: #10162
Signed-off-by: Lei Huang <leih@nvidia.com>
This patch re-generates the client code for Cloud Hypervisor v41.0.
Note: The client code of cloud-hypervisor's OpenAPI is automatically
generated by openapi-generator.
Fixes: #10203
Signed-off-by: Bo Chen <chen.bo@intel.com>
This commit allows `cdh_api_timeout` to be configured from the configuration file.
The configuration is commented out with specifying a default value (50s) because
the default value is configured in the agent.
Signed-off-by: Hyounggyu Choi <Hyounggyu.Choi@ibm.com>
Improved error handling to provide clearer feedback on request failures.
For example:
Improve createcontainer request timeout error message from
"Error: failed to create containerd task: failed to create shim task:context deadline exceed"
to "Error: failed to create containerd task: failed to create shim task: CreateContainerRequest timed out: context deadline exceed".
Fixes: #10173 -- part II
Signed-off-by: ChengyuZhu6 <chengyu.zhu@intel.com>
Our code for handling images being pulled inside the guest relies on a
containerType ("sandbox" or "container") being set as part of the
container annotations, which is done by the CRI Engine being used, and
depending on the used CRI Engine we check for a specfic annotation
related to the image-name, which is then passed to the agent.
However, when running kata-containers without kubernetes, specifically
when using `nerdctl`, none of those annotations are set at all.
One thing that we can do to allow folks to use `nerdctl`, however, is to
take advantage of the `--label` flag, and document on our side that
users must pass `io.kubernetes.cri.image-name=$image_name` as part of
the label.
By doing this, and changing our "fallback" so we can always look for
such annotation, we ensure that nerdctl will work when using the nydus
snapshotter, with kata-containers, to perform image pulling inside the
pod sandbox / guest.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
This parameter has been deprecated for a long time and QEMU 9.1.0 finally removes it.
Fixes: kata-containers#10112
Signed-off-by: Tom Dohrmann <erbse.13@gmx.de>
Similar to HotAttach, the HotDetach method signature for network
endoints needs to be changed as well to allow for the method to make
use of device manager to manage the hot unplug of physical network
devices.
Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
Enable physical network interfaces to be hotplugged.
For this, we need to change the signature of the HotAttach method
to make use of Sandbox instead of Hypervisor. Similar approach was
followed for Attach method, but this change was overlooked for
HotAttach.
The signature change is required in order to make use of
device manager and receiver for physical network
enpoints.
Fixes: #8405
Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
The current implementation for device binding using driver bind/unbind
and new_id fails in the scenario when the physical device is not bound
to a driver before assigning it to vfio.
There exists and updated mechanism to accomplish the same that does not
have the same issue as above.
The driver_override field for a device allows us to specify the driver for a device
rather than relying on the bound driver to provide a positive match of the
device. It also has other advantages referenced here:
https://patchwork.kernel.org/project/linux-pci/patch/1396372540.476.160.camel@ul30vt.home/
So use the updated driver_override mechanism for binding/unbinding a
physical device/virtual function to vfio-pci.
Signed-off-by: liangxianlong <liang.xianlong@zte.com.cn>
Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
Sets SharedFS config to NoSharedFS for remote hypervisor in order to start the file watcher which syncs files from the host to the guest VMs.
Signed-off-by: Silenio Quarti <silenio_quarti@ca.ibm.com>
The DAN feature has already been implemented in kata-runtime-rs, and
this commit brings the same capability to the Go kata-runtime.
Fixes: #9758
Signed-off-by: Lei Huang <leih@nvidia.com>
Do not install the packages librados-dev and librbd-dev as they are not needed for building static qemu.
Add machine option cap-ail-mode-3=off while creating the VM to qemu cmdline.
Fixes: #9893
Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
This patch re-generates the client code for Cloud Hypervisor v40.0.
Note: The client code of cloud-hypervisor's OpenAPI is automatically
generated by openapi-generator.
Fixes: #9929
Signed-off-by: Bo Chen <chen.bo@intel.com>
Move the `sandbox.agent.setPolicy` call out of the remoteHypervisor
if, block, so we can use the policy implementation on peer pods
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
Commit 'ca02c9f5124e' implements the vhost-user-blk reconnection functionality,
However, it has missed assigning VhostUserDeviceReconnect when new the QEMU
HypervisorConfig, resulting in VhostUserDeviceReconnect always set to default value 0.
Real change is this line, most of changes caused by go format,
return vc.HypervisorConfig{
// ...
VhostUserDeviceReconnect: h.VhostUserDeviceReconnect,
}, nil
Fixes: #9848
Signed-off-by: markyangcc <mmdou3@163.com>
This is a counterpart of commit abf52420a4 for the qemu-coco-dev
configuration. By allowing default_vcpu and default_memory annotations
users can fine-tune the VM based on the size of the container
image to avoid issues related with pulling large images in the guest.
Signed-off-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Just like the TEE configurations (sev, snp, tdx) we want to have the
qemu-coco-dev using shared_fs=none.
Fixes: #9676
Signed-off-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
While running with a remote hypervisor, whenever kata-monitor tries to access
metrics from the shim, the shim does a "panic" and no metric can be gathered.
The function GetVirtioFsPid() is called on metrics gathering, and had a call
to "panic()". Since there is no virtiofs process for remote hypervisor, the
right implementation is to return nil. The caller expects that, and will skip
metrics gathering for virtiofs.
Fixes: #9826
Signed-off-by: Julien Ropé <jrope@redhat.com>
This patch re-generates the client code for Cloud Hypervisor v39.0.
Note: The client code of cloud-hypervisor's OpenAPI is automatically
generated by openapi-generator.
Fixes: #8694, #9574
Signed-off-by: Bo Chen <chen.bo@intel.com>
When the total number of files observed is greater than limit, return (-1, err).
When the returned err is not nil, the func countFiles should return -1.
Fixes:#9780
Signed-off-by: gaohuatao <gaohuatao@bytedance.com>
This is required to provide the hashes of kernel, initrd and cmdline
needed during the attestation of the coco.
Fixes: #9150
Signed-off-by: Niteesh Dubey <niteesh@us.ibm.com>
Since yq frequently updates, let's upgrade to a version from February to
bypass potential issues with versions 4.41-4.43 for now. We can always
upgrade to the newest version if necessary.
Fixes#9354
Depends-on:github.com/kata-containers/tests#5818
Signed-off-by: Beraldo Leal <bleal@redhat.com>
golang.mk is not ready to deal with non GOPATH installs. This is
breaking test on s390x.
Since previous steps here are installing go and yq our way, we could
skip this aditional check. A full refactor to golang.mk would be needed
to work with different paths.
Signed-off-by: Beraldo Leal <bleal@redhat.com>
We need to remove the device from the tracking map, a container
restart will increment the bus index and we will get out of root-ports
and crash the machine.
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
We need special handling for pod_sandbox, pod_container and
single_container how and when to inject CDI devices
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
In Kubernetes we still do not have proper VM sizing
at sandbox creation level. This KEP tries to mitigates
that: kubernetes/enhancements#4113 but this can take
some time until Kube and containerd or other runtimes
have those changes rolled out.
Before we used a static config of VFIO ports, and we
introduced CDI support which needs a patched contianerd.
We want to eliminate the patched continerd in the GPU case
as well.
Fixes: #8860
Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
For the TD attestation to work the connection to QGS on the host is needed.
By default QGS runs on vsock port 4050, but can be modified by the host owner.
Format of the qemu object follows the SocketAddress structure, so it needs to be provided in the JSON format, as in the example below:
-object '{"qom-type":"tdx-guest","id":"tdx","quote-generation-socket":{"type":"vsock","cid":"2","port":"4050"}}'
Fixes: #9497
Signed-off-by: Jakub Ledworowski <jakub.ledworowski@intel.com>
Remove `rand.Seed` call to resolve the following failure:
```
rand.Seed is deprecated: As of Go 1.20 there is no reason to call Seed with a random value.
```
The go rand.Seed docs: https://pkg.go.dev/math/rand@go1.20#Seed
back this up and states:
> If Seed is not called, the generator is seeded randomly at program startup.
so I believe we can just delete the call.
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
For now, let's allow the users to set the default_cpu and default_memory
when using TDX, as they may hit issues related to the size of the
container image that must be pulled and unpacked inside the guest,
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
- Make due to us bumping the golang version used in our CI
but `make vendor` fails without the go version in the runtime go.mod
being increased, so update this and run go mod tidy
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
We shouldn't be using 9p, at all, with TEEs, as off right now we have no
way to ensure the channels are encrypted. The way to work this around
for now is using guest pull, either with containerd + nydus snapshotter
or with CRI-O; or even tardev snapshotter for pulling on the host (which
is the approach used by MSFT).
This is only done for TDX for now, leaving the generic, AMD, and IBM
related stuff for the folks working on those to switch and debug
possible issues on their environment.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
ResizeMemory for Cloud Hypervisor is missing a check for the new
requested memory being greater than the max hotplug size after
alignment. Add the check, and since an earlier check for this
setsrequested memory to the max hotplug size, do the same in the
post-alignment check.
Fixes#9640
Signed-off-by: Chelsea Mafrica <chelsea.e.mafrica@intel.com>
By default, when a container is created with the `--privileged` flag,
all devices in `/dev` from the host are mounted into the guest. If
there is a block device(e.g. `/dev/dm`) followed by a generic
device(e.g. `/dev/null`)，two identical block devices(`/dev/dm`)
would be requested to the kata agent causing the agent to exit with error:
> Conflicting device updates for /dev/dm-2
As the generic device type does not hit any cases defined in `switch`，
the variable `kataDevice` which is defined outside of the loop is still
the value of the previous block device rather than `nil`. Defining `kataDevice`
in the loop fixes this bug.
Signed-off-by: cncal <flycalvin@qq.com>
This was needed when we were using an old (and not maintained anymore)
host stack. Considering what we have as part of the distros, Today,
this can simply be dropped, as I cannot find any reference of this one
being needed in any up-to-date documentation.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
This reverts commit b7cccfa019.
The `private=on` bit has never made its way upstream, and was removed
from the latest iteration that we're using. With that in mind, let's
revert its usage in the code.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
This reverts commit 582b5b6b19.
The `private=on` bit has never made its way upstream, and was removed
from the latest iteration that we're using. With that in mind, let's
revert its addition, and later on its usage in the code.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Let's add the PLACEHOLDER_FOR_DISTRO_{QEMU,OVMF}_WITH_TDX_SUPPORT
variables instead of actually setting a path, so we can easily replace
those as part of our deployment scripts.
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Everytime I create contianer on arm64 machine, containerd/kata logs a redundant warning
as follows:
``` shell
time="2024-05-07" level=warning msg="<nil>" arch=arm64 name=containerd-shim-v2
pid=xxx sandbox=fdd1f05 source=virtcontainers/hypervisor
```
I added an error statement so that the error would be logged when it occurs.
Signed-off-by: cncal <flycalvin@qq.com>
