Since the rate limiter would be shared by cloud-hypervisor
and firecracker etc, thus move it from clh's config to
hypervisor config crate which would be shared by other vmm.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
Given that Rust-based VMMs like cloud-hypervisor, Firecracker, and
Dragonball naturally offer user-level block I/O rate limiting, I/O
throttling has been implemented to leverage this capability for these
VMMs. This PR specifically introduces support for cloud-hypervisor.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
Support for the share-rw=true parameter has been added. While this
parameter is essential for maintaining data consistency across multiple
QEMU instances sharing a backend disk image, its implementation also
serves to standardize parameters with the block device hotplug
functionality in kata-runtime/qemu.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
Docker tests have been broken for a while and should be removed if we
cannot maintain those.
For now, though, let's limit it to run only with one hypervisor and
avoid wasting resources for no reason.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
devmapper tests have been failing for a while. It's been breaking on the
kata-deploy deployment, which is most likely related to Disk Pressure.
Removing files was not enough to get the tests to run, so we'll just run
those with QEMU as a way to test fixes. Once we get the test working,
we can re-enable the other VMMs, but for now let's just not waste
resources for no reason.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
add_allow_all_policy_to_yaml now also sets the initdata annotation. So don't overwrite the
initdata annotation that was previously set by create_coco_pod_yaml_with_annotations.
Signed-off-by: Saul Paredes <saulparedes@microsoft.com>
Delete annotation from OCI spec and sandbox config. This is done after the optional initdata annotation value has been read.
Signed-off-by: Saul Paredes <saulparedes@microsoft.com>
Bump cri-o to 1.34.0 to try and remediate security advisories
CVE-2025-0750 and CVE-2025-4437.
Note: Running
```
go get github.com/cri-o/cri-o@v1.34.0
```
seems to bump a lot of other go modules, hence the size of the
vendor diff
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
In google.golang.org/grpc v1.72.0, `DialContext`, is deprecated, so
switch to use `NewClient` instead.
`grpc.WithBlock()` is deprecated and not recommend, so remove this
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
In `github.com/prometheus/common v0.62.0` expfmt.FmtText
is deprecated, so replace with `expfmt.NewFormat(expfmt.TypeTextPlain)`.
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
In `github.com/prometheus/common v0.62.0` expfmt.FmtText
is deprecated, so replace with `expfmt.NewFormat(expfmt.TypeTextPlain)`.
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
This is following Steve's suggestion, based on what's been done on
cloud-api-adaptor.
The reason we're doing it here is because we've seen pods being evicted
due to disk pressure.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
I've hit this when using a machine with slow internet connection, which
took ages to download the kata-cleanup image, and then helm timed out in
the middle of the cleanup, leading to the cleanup job being restarted
and then bailing with an error as the runtimeclasses that kata-deploy
tries to delete were already deleted.
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
Since the cloud hypervisor's resize vCPU is an asynchronous operation,
it's possible that the previous resize operation hasn't completed when
the request is sent, causing the current call to return an error.
Therefore, several retries can be performed to avoid this error.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
A minor release of QEMU is out, so update to it for fixes and features.
QEMU changelog: https://wiki.qemu.org/ChangeLog/10.1
Notes:
* AVX support is not an option to be enabled / disabled anymore.
* Passt requires Glibc 2.40.+, which means a dependency on Ubuntu 25.04
or newer, thus we're disabling it.
Signed-off-by: Alex Tibbles <alex@bleg.org>
Although versions of slab prior to 0.4.10, don't have a security
vulnearability, we can bump them all to keep things in sync
Signed-off-by: stevenhorsman <steven@uk.ibm.com>
```
Experimental fw_cfg Device Support
This feature enables passing configuration data and files, such as VM
boot configurations (kernel, kernel cmdline, e820 memory map, and ACPI
tables), from the host to the guest. (#7117)
Experimental ivshmem Device Support
Support for inter-VM shared memory has been added. For more information,
please refer to the ivshmem documentation. (#6703)
Firmware Boot Support on riscv64
In addition to direct kernel boot, firmware boot support has been added
on riscv64 hosts. (#7249)
Increased vCPU Limit on x86_64/kvm
The maximum number of supported vCPUs on x86_64 hosts using KVM has been
raised from 254 to 8192. (#7299)
Improved Block Performance with Small Block Sizes
Performance for virtio-blk with small block sizes (16KB and below)
is enhanced via submitting async IO requests in batches. (#7146)
Faster VM Pause Operation
The VM pause operation now is significantly faster particularly for VMs
with a large number of vCPUs. (#7290)
Updated Documentation on Windows Guest Support
Our Windows documentation now includes instructions to run Windows 11
guests, in addition to Windows Server guests. (#7218)
Policy on AI Generated Code
We will decline any contributions known to contain contents generated or
derived from using Large Language Models (LLMs). Details can be found
in our contributing documentation. (#7162)
Removed SGX Support
The SGX support has been removed, as announced in the deprecation notice two
release cycles ago. (#7093)
Notable Bug Fixes
Seccomp filter fixes with glibc v2.42 (#7327)
Various fixes related to (#7331, #7334, #7335)
```
From https://github.com/cloud-hypervisor/cloud-hypervisor/releases/tag/v48.0
Signed-off-by: Fabiano Fidêncio <fabiano@fidencio.org>
Exclude 'cgroup' namespace from namespace checks during `allow_linux`
validation. This complements the existing exclusion of the 'network'
namespace.
As runtime-rs has specific cgroup namespace configurations, and excluding it from
policy validation ensures parity between runtime-rs and runtime-go implementations.
This allows focusing validation on critical namespaces like PID, IPC, and MNT, while
avoiding potential policy mismatches due to another cgroup namespace management by
the runtime-rs.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
Add `normalize_namespace_type()` function to map "mount"
(case-insensitive) to "mnt" while keeping other values unchanged.
This ensures namespace comparisons treat "mount" and "mnt" as
equivalent.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
- Use set comparison to ignore ordering differences when matching
capabilities.
- Add normalization to strip "CAP_" prefix to support both CAP_XXX and
XXX formats.
This makes capability matching more robust against different ordering
and naming formats.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
As prvious configure with overlayfs is incorrect, which causes the agent
policy validation failure. And it's also different with runtime-go's
configuration. In this patch, we'll correct its fstype with overlay and
align with runtime on this matter.
Signed-off-by: Alex Lyn <alex.lyn@antgroup.com>
When hot-removing a block device, the kernel must first unmount the
device and then destroy it on the VM. Therefore, a
prepare_remove_block_device procedure must be added to wait for the
kernel to unmount the device before destroying it on the VM.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
When hot-removing a block device, the kernel must first remove the
device and then destroy it on the VM. Therefore, a
prepare_remove_block_device procedure must be added to wait for the
kernel to unmount the device before destroying it on the VM.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
Since Dragonball's MMIO bus only supports legacy interrupts, while
the PCI bus supports MSIX interrupts, to improve performance for block
devices, virtio-blk devices are set to PCI bus mode by default.
We had tested the virtio-blk's performance using the fio with the
following commands:
fio -filename=./test -direct=1 -iodepth 32 -thread -rw=randrw
-rwmixread=50 -ioengine=libaio -bs=4k -size=10G -numjobs=4
-group_reporting -name=mytest
When used the legacy interrupt, the io test is as below:
read : io=20485MB, bw=195162KB/s, iops=48790, runt=107485msec
write: io=20475MB, bw=195061KB/s, iops=48765, runt=107485msec
Once switched to msix innterrupt, the io test is as below:
read : io=20485MB, bw=260862KB/s, iops=65215, runt= 80414msec
write: io=20475MB, bw=260727KB/s, iops=65181, runt= 80414msec
We can get 34% performance improvement.
Signed-off-by: Fupan Li <fupan.lfp@antgroup.com>
