On a conventional (e.g. runc) container, passing in a VFIO group device,
/dev/vfio/NN, will result in the same VFIO group device being available
within the container.
With Kata, however, the VFIO device will be bound to the guest kernel's
driver (if it has one), possibly appearing as some other device (or a
network interface) within the guest.
This add a new `vfio_mode` option to alter this. If set to "vfio" it will
instruct the agent to remap VFIO devices to the VFIO driver within the
guest as well, meaning they will appear as VFIO devices within the
container.
Unlike a runc container, the VFIO devices will have different names to the
host, since the names correspond to the IOMMU groups of the guest and those
can't be remapped with namespaces.
For now we keep 'guest-kernel' as the value in the default configuration
files, to maintain current Kata behaviour. In future we should change this
to 'vfio' as the default. That will make Kata's default behaviour more
closely resemble OCI specified behaviour.
fixes#693
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently constrainGRPCSpec always removes VFIO devices from the OCI
container spec which will be used for the inner container. For
upcoming support for VFIO devices in DPDK usecases we'll need to not
do that.
As a preliminary to that, add an extra parameter to the function to
control whether or not it will remove the VFIO devices from the spec.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
"constraint" is a noun, "constrain" is the associated verb, which makes
more sense in this context.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In order to support DPDK workloads, we need to change the way VFIO devices
will be handled in Kata containers. However, the current method, although
it is not remotely OCI compliant has real uses. Therefore, introduce a new
runtime configuration field "vfio_mode" to control how VFIO devices will be
presented to the container.
We also add a new sandbox annotation -
io.katacontainers.config.runtime.vfio_mode - to override this on a
per-sandbox basis.
For now, the only allowed value is "guest-kernel" which refers to the
current behaviour where VFIO devices added to the container will be bound
to whatever driver in the VM kernel claims them.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This PR includes these optimize changes:
- Remove the dependency on the container engine.
The old code uses runc to generate config.json and
Docker to export rootfs, that will be heavy and need
additional dependency.
Using a fixed config for busybox image can avoid
the heavy processing above.
- Moved duplicate code to pkg/katatestutils package
Fixes: #2752
Signed-off-by: bin <bin@hyper.sh>
cri-containerd project has been merged into containerd repo, and
we should not reference it any more in code and docs.
This commit will use containerd package instead of cri-containerd
package.
Fixes: #2791
Signed-off-by: bin <bin@hyper.sh>
The tracing tags for api.go contain `"packages"` as a tag name,
whereas all other tags contain `"package"`.
Fixes: #2847
Signed-off-by: Christophe de Dinechin <dinechin@redhat.com>
Display a pseudo path to the sandbox socket in the output of
`kata-runtime env` for those hypervisors that use Hybrid VSOCK.
The path is not a real path since the command does not create a sandbox.
The output includes a `{ID}` tag which would be replaced with the real
sandbox ID (name) when the sandbox was created.
This feature is only useful for agent tracing with the trace forwarder
where the configured hypervisor uses Hybrid VSOCK.
Note that the features required a new `setConfig()` method to be added
to the `hypervisor` interface. This isn't normally needed as the
specified hypervisor configuration passed to `setConfig()` is also
passed to `createSandbox()`. However the new call is required by
`kata-runtime env` to display the correct socket path for Firecracker.
The new method isn't wholly redundant for the main code path though as
it's now used by each hypervisor's `createSandbox()` call.
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
The 'quiet' kernel parameter can avoid guest kernel logs while booting,
which can reduce boot time.
Fix: #2820
Signed-off-by: Bo Chen <chen.bo@intel.com>
We will need to have console output from the guest only for debugging
purposes. As a result, we can turn-off both the serial and
virtio-console devices by default for better boot time.
Fixes: #2820
Signed-off-by: Bo Chen <chen.bo@intel.com>
The variable for 'name' in config-settings.go.in was previously
hardcoded as "kata". In e7c42fb it was changed to the runtime name,
which is "kata-runtime". Add a variable to specify a syslog identifier
for consistency for tests and documentation that use it.
Fixes#2806
Signed-off-by: Chelsea Mafrica <chelsea.e.mafrica@intel.com>
Update the sandbox dir clean up logic to be more appropriate
Add different seeds for randInt() method
Fixes#2770
Signed-off-by: Feng Wang <feng.wang@databricks.com>
This patch adds an option "disable_seccomp" to the config
hypervisor.clh, from which users can disable the `seccomp`
feature from Cloud Hypervisor when needed (for debugging purposes).
Fixes: #2782
Signed-off-by: Bo Chen <chen.bo@intel.com>
This patch enables the `seccomp` feature from Cloud Hypervisor which
provides fine-grained allowed syscalls for each of its worker
threads. It brings important security benefits, while would increase
memory footprint.
Fixes: #2782
Signed-off-by: Bo Chen <chen.bo@intel.com>
Shim management server is running in a go routine, in test mode
this will cause the directory where the listen socket
file(/run/vc/sbs/777-77-77777777/shim-monitor.sock) in leak
after the tests finished.
Fixes: #2805
Signed-off-by: bin <bin@hyper.sh>
This commit does two chagnes:
- move code for managing temp users to rootless.go.
- use common function in qemu.go when shutdown the VM.
Fixes: #2759
Signed-off-by: bin <bin@hyper.sh>
Even CCA, which is the confidential compute archtecture, has not been
ready, add a empty implementation to avoid static check error.
Fixes: #2789
Signed-off-by: Jianyong Wu <jianyong.wu@arm.com>
Suggested-by: Fabiano Fidêncio <fidencio@redhat.com>
Exclude from lint checking for it is ultimately only used in
architecture-specific code.
Fixes: #2273
Signed-off-by: Jakob Naucke <jakob.naucke@ibm.com>
Bump containerd to v1.5.7 in order to bring in a fix for CVE-2021-41103,
"insufficiently restricted permissions ons plugins directories
(https://github.com/advisories/GHSA-c2h3-6mxw-7mvq)".
dependabot found a potential security vulnerability and raised a PR to
fix it. However, dependabot does not properly follows nor understands
the needed of our CIs (mainly related to formatting the PR and whatnot),
thus I'm re-raising it.
Fixes: #2796
Supersedes: #2787
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Our check for the IP family is working as long as we have either a
gateway or a destination IP. Some routes are missing both.
The RT netlink messages provide the IP family information for each
route, so we can carry that piece of information up to the guest. That
will allow for a more reliable route IP family determination.
Signed-off-by: Samuel Ortiz <s.ortiz@apple.com>
We need to be able to get the IP family from the netlink route meesages,
and the Route.Family field only got recently added to the netlink
package.
The update generates static check warnings about the call for
nethandler.Delete() being deprecated in favor of a Close() call instead.
So we include the s/Delete()/Close()/ change as part of this PR.
Signed-off-by: Samuel Ortiz <s.ortiz@apple.com>
Reduce the cloud-hypervisor log level from `Debug` to `Info` when hypervisor
debug is enabled. This is required since `Debug` level:
- Is overkill for debugging hypervisor failures.
- Effectively hides the output from the guest kernel and userland: CLH
generates so much output that the output from the guest gets "lost in
the noise" (experiments show that for each full CLH debug message, at most
1 _byte_ of guest output is displayed).
Fixes: #2726.
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
