When run with configuration `sandbox_cgroup_only=false`, we will call
`gen_overhead_path()` as the overhead path. The `cgroup-rs` will push
the path with the subsystem prefix by `PathBuf::push()`. When the path
has prefix “/” it will act as root path, such as
```
let mut path = PathBuf::from("/tmp");
path.push("/etc");
assert_eq!(path, PathBuf::from("/etc"));
```
So we shoud not set overhead path with prefix "/".
Fixes: #4687
Signed-off-by: Quanwei Zhou <quanweiZhou@linux.alibaba.com>
During use, there will be cases where the container is in the stop state
and get another stop. In this case, the second stop needs to be ignored.
Fixes: #4683
Signed-off-by: Quanwei Zhou <quanweiZhou@linux.alibaba.com>
Update dbs-xxx commit ID for aarch64 in runtime-rs/Cargo.toml file to add
dependencies for aarch64.
Fixes: #4676
Signed-off-by: xuejun-xj <jiyunxue@alibaba.linux.com>
Module anyhow::anyhow is only used on x86_64 architecture in
crates/hypervisor/src/device/vfio.rs file.
Fixes: #4676
Signed-off-by: xuejun-xj <jiyunxue@alibaba.linux.com>
As a built in VMM, Path, jailer path, ctlpath are not needed for
Dragonball. So we don't generate those value in Makefile.
Fixes: #4677
Signed-off-by: Zhongtao Hu <zhongtaohu.tim@linux.alibaba.com>
New nix is supporting UMOUNT_NOFOLLOW, upgrade nix
version to use this flag instead of the self-defined flag.
Fixes: #4670
Signed-off-by: liubin <liubin0329@gmail.com>
io_uring is a Linux API for asynchronous I/O introduced in qemu 5.0.
It is designed to better performance than older aio API.
We could leverage this in order to get better storage performance.
We should be adding liburing-dev to qemu build to leverage this feature.
However liburing-dev package is not available in ubuntu 20.04,
it is avaiable in 22.04.
Upgrading the ubuntu version in the dockerfile to 22.04 is causing
issues in the static qemu build related to libpmem.
So instead we are building liburing from source until those build issues
are solved.
Fixes: #4645
Signed-off-by: Archana Shinde <archana.m.shinde@intel.com>
When a container terminated, we should make sure there's no processes
left after destroying the container.
Before this commit, kata-agent depended on the kernel's pidns
to destroy all of the process in a container after the 1 process
exit in a container. This is true for those container using a
separated pidns, but for the case of shared pidns within the
sandbox, the container exit wouldn't trigger the pidns terminated,
and there would be some daemon process left in this container, this
wasn't expected.
Fixes: #4663
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
1. support annotation for runtime.name, hypervisor_name, agent_name.
2. fix parse memory from annotation
Signed-off-by: Zhongtao Hu <zhongtaohu.tim@linux.alibaba.com>
Change get_mounts to get paths from a borrowed argument rather than
calling get_paths a second time.
Fixes#3768
Signed-off-by: Derek Lee <derlee@redhat.com>
Currently $BUILD_DIR will be used to create a directory as:
/opt/kata/share/kata-qemu${BUILD_DIR}
It means that when passing a BUILD_DIR, like "foo", a name would be
built like /opt/kata/share/kata-qemufoo
We should, instead, be building it as /opt/kata/share/kata-qemu-foo.
Fixes: #4638
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Instead of always naming the binary as "-experimental", let's take
advantage of the $BUILD_SUFFIX that's already passed and correctly name
the binary according to it.
Fixes: #4638
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Passing the URL to be used to download the kernel tarball is useful in
various scenarios, mainly when doing a downstream build, thus let's add
this new option.
This new option also works around a known issue of the Dockerfile used
to build the kernel not having `yq` installed.
Fixes: #4629
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
There's no need to have the entire function for building SEV / TDX
duplicated.
Let's remove those functions and create a `get_tee_kernel()` which takes
the TEE as the argument.
Fixes: #4627
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
Fix the warning "unused import: `dbs_arch::gic::Error as GICError`" and
"unused import: `dbs_arch::gic::GICDevice`" in file src/vm/mod.rs when
compiling.
Fixes: #4544
Signed-off-by: xuejun-xj <jiyunxue@alibaba.linux.com>
Signed-off-by: jingshan <jingshan@linux.alibaba.com>
As string "com1", "com2" and "rtc" are used in two files
(device_manager/mod.rs and device_manager/legacy.rs), we use public
const variables COM1, COM2 and RTC to replace them respectively.
Fixes: #4544
Signed-off-by: xuejun-xj <jiyunxue@alibaba.linux.com>
Signed-off-by: jingshan <jingshan@linux.alibaba.com>
Implement generate_virtio_device_info() and
get_virtio_mmio_device_info() functions su support the mmio_device_info
member, which is used by FDT.
Fixes: #4544
Signed-off-by: xuejun-xj <jiyunxue@linux.alibaba.com>
Signed-off-by: jingshan <jingshan@linux.alibaba.com>
Instead of passing a `KATA_CONF_FILE` environament variable, let's rely
on the configured (in the container engine) config path, as both
containerd and CRI-O support it, and we're using this for both of them.
Fixes: #4608
Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
