we should add unit test for func MaxQemuVCPUS in qemu_amd64_test.go
Signed-off-by: Penny Zheng <penny.zheng@arm.com>
Signed-off-by: Wei Chen <Wei.Chen@arm.com>
on aarch64, we support different gic interrupt controllers.
The maximum number of vCPUs depends on the GIC version, or on how
many redistributors we can fit into the memory map.
Fixes: #584
Signed-off-by: Penny Zheng <penny.zheng@arm.com>
Signed-off-by: Wei Chen <Wei.Chen@arm.com>
RemoveContainer is called right after SignalProcess(SIGKILL), the container
process might be still running and container Destroy() will fail, thus it's better
to wait on this process exited before to issue RemoveContainer.
Fixes: #690
Signed-off-by: fupan <lifupan@gmail.com>
Fixes#635
When container rootfs is block based in devicemapper use case, we can re-use
sandbox device manager to manage rootfs block plug/unplug, we don't detailed
description of block in container state file, instead we only need a Block index
referencing sandbox device.
Remove `HotpluggedDrive` and `RootfsPCIAddr` from state file because it's not
necessary any more.
Signed-off-by: Wei Zhang <zhangwei555@huawei.com>
For vm factory, we also need netns to be set otherwise we fail to
create new VMs in `s.network.run`.
Fixes: #681
Signed-off-by: Peng Tao <bergwolf@gmail.com>
Fixes#635
`ContainerPath` isn't meaningful in sandbox's `devices.json`, because
device is sandbox level which can be referenced by many containers, and
`ContainerPath` is per container which means nothing to a sandbox level
resource.
Signed-off-by: Wei Zhang <zhangwei555@huawei.com>
Fixes#635
Remove `Hotplugged bool` field from device and add two new fields
instead:
* `RefCount`: how many references to this device. One device can be
referenced(`NewDevice()`) many times by same/different container(s),
two devices are regarded identical if they have same hostPath
* `AttachCount`: how many times this device has been attached. A device
can only be hotplugged once to the qemu, every new Attach command will
add the AttachCount, and real `Detach` will be done only when
`AttachCount == 0`
Signed-off-by: Wei Zhang <zhangwei555@huawei.com>
We need to update the network docs to reflect CNM
and CNI are handled much the same way. Start off
by removing the incorrect CNI docs first.
Fixes#678
Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
Fix nasty bug which resulted in `kata-env` showing
`VMContainerCapable = true` even on amd64 systems without virtualisation
support (thankfully `kata-check` still showed the correct results).
Added arch-specific tests to avoid any possibility of regression.
Fixes#660.
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
golang version 1.8.3 is old and the runtime cannot even be built with
it now it seems.
Since it is no longer considered a stable version [1], move to the
oldest official stable version (version 1.9.7).
Fixes#642.
[1] - https://golang.org/dl/
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
Now that we support multiple branches, we changed how
static-checks.sh compares branches. We now need a
variable called $target_branch to make the correct
comparison when testing a PR.
Fixes: #663.
Signed-off-by: Salvador Fuentes <salvador.fuentes@intel.com>
Add a heuristic to ensure the amount of memory allocated to the
hypervisor is bigger than the size of the image.
This catches simple configuration issues where `default_memory=` is set
to a smaller value than the size of either the `image=` or `initrd=`
files.
If the configured image type is `initrd`, fail but only warn in the
logs for `image` as although it seems a highly unlikely scenario, it is
permitted.
Update tests to ensure that created resources have `>0` bytes.
Fixes#636.
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
The `TestMinimalRuntimeConfig` should not be using the real
resource files that might be installed on a system so make temporary
files instead to better control the test.
Split out `TestMinimalRuntimeConfigWithVsock` to reduce cyclomatic
complexity (along with dropping the config file delete at the end - not
required as the entire test-specific directory gets auto-deleted).
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
govmm has ExecuteBlockdevAdd() function and ExecuteBlockdevDel() function
doesn't compatible with qemu 2.8,because blockdev-add and x-blockdev-del usages
are different between qemu 2.7 and qemu 2.8
shortlog:
ce070d1 govmm: modify govmm to be compatible with qemu 2.8
0286ff9 qemu/qmp: support hotplug a nic whose qdisc is mq
8515ae4 qmp: Remind users that you must first call ExecuteQMPCapabilities()
21504d3 qemu/qmp: Add netdev_add with chardev support
ed34f61 Add some negative test cases for qmp.go
17cacc7 Add negative test cases for qemu.go
fixes: #637
Signed-off-by: flyflypeng <jiangpengfei9@huawei.com>
If the sandbox has been initialized with a factory, this means the
caller should be in charge of adding any network to the VM, and
virtcontainers library cannot make any assumptions about adding
the default underlying network.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
The CLI being the implementation of the OCI specification, and the
hooks being OCI specific, it makes sense to move the handling of any
OCI hooks to the CLI level. This changes allows the Kata API to
become OCI agnostic.
Fixes#599
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
As we want to call the OCI hook from the CLI, we need a way for the
CLI to figure out what is the network namespace used by the sandbox.
This is needed particularly because virtcontainers creates the netns
if none was provided.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This commit moves the network namespace creation out of virtcontainers
in order to anticipate the move of the OCI hooks to the CLI through a
follow up commit.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Since we removed the CNI implementation and that we agreed the network
should only be handled in a single way from virtcontainers, this patch
logically replace the "CNM" naming with "Default".
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Update the agent vendoring for the new `NewAgentClient()` API - the
agent client code will enable gRPC tracing when passed a context
containing an opentracing span.
Agent client code changes:
6d26d61 client: Add context parameter and enable tracing support
Fixes#640.
Signed-off-by: James O. D. Hunt <james.o.hunt@intel.com>
The storage implementation of filesystem was in the same file where
the resource storage interface was declared. It's more proper to
separate implementations and interface into different files.
Fixes#633
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This noop implementation of resourceStorage will allow for easier
unit testing of some sandbox functions.
Fixes#632
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
The specific agent implementation kata_agent was defining a very
useful generic function that is now moved to the global file
network.go.
Fixes#629
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
