Merge pull request #4921 from liubin/fix/2920-delete-vergen

runtime-rs: delete vergen dependency

.github/workflows/add-pr-sizing-label.yaml

@@ -33,6 +33,8 @@ jobs:
           GITHUB_TOKEN: ${{ secrets.KATA_GITHUB_ACTIONS_PR_SIZE_TOKEN }}
         run: |
           pr=${{ github.event.number }}
+          # Removing man-db, workflow kept failing, fixes: #4480
+          sudo apt -y remove --purge man-db
           sudo apt -y install diffstat patchutils
 
           pr-add-size-label.sh -p "$pr"

.github/workflows/commit-message-check.yaml

@@ -63,7 +63,8 @@ jobs:
         #   the entire commit message.
         #
         # - Body lines *can* be longer than the maximum if they start
-        #   with a non-alphabetic character.
+        #   with a non-alphabetic character or if there is no whitespace in
+        #   the line.
         #
         #   This allows stack traces, log files snippets, emails, long URLs,
         #   etc to be specified. Some of these naturally "work" as they start
@@ -74,8 +75,8 @@ jobs:
         #
         # - A SoB comment can be any length (as it is unreasonable to penalise
         #   people with long names/email addresses :)
-        pattern: '^.+(\n([a-zA-Z].{0,149}|[^a-zA-Z\n].*|Signed-off-by:.*|))+$'
-        error: 'Body line too long (max 72)'
+        pattern: '^.+(\n([a-zA-Z].{0,150}|[^a-zA-Z\n].*|[^\s\n]*|Signed-off-by:.*|))+$'
+        error: 'Body line too long (max 150)'
         post_error: ${{ env.error_msg }}
 
     - name: Check Fixes

.github/workflows/docs-url-alive-check.yaml

@@ -10,35 +10,32 @@ jobs:
         go-version: [1.17.x]
         os: [ubuntu-20.04]
     runs-on: ${{ matrix.os }}
+    # don't run this action on forks
+    if: github.repository_owner == 'kata-containers'
     env:
       target_branch: ${{ github.base_ref }}
     steps:
     - name: Install Go
-      if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
       uses: actions/setup-go@v2
       with:
         go-version: ${{ matrix.go-version }}
       env:
         GOPATH: ${{ runner.workspace }}/kata-containers
     - name: Set env
-      if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
       run: |
         echo "GOPATH=${{ github.workspace }}" >> $GITHUB_ENV
         echo "${{ github.workspace }}/bin" >> $GITHUB_PATH
     - name: Checkout code
-      if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
       uses: actions/checkout@v2
       with:
         fetch-depth: 0
         path: ./src/github.com/${{ github.repository }}
     - name: Setup
-      if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
       run: |
         cd ${GOPATH}/src/github.com/${{ github.repository }} && ./ci/setup.sh
       env:
         GOPATH: ${{ runner.workspace }}/kata-containers
     # docs url alive check
     - name: Docs URL Alive Check
-      if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
       run: |
         cd ${GOPATH}/src/github.com/${{ github.repository }} && make docs-url-alive-check

.github/workflows/kata-deploy-push.yaml

@@ -24,6 +24,7 @@ jobs:
           - firecracker
           - rootfs-image
           - rootfs-initrd
+          - virtiofsd
     steps:
       - uses: actions/checkout@v2
       - name: Install docker

.github/workflows/kata-deploy-test.yaml

@@ -1,4 +1,5 @@
 on:
+  workflow_dispatch: # this is used to trigger the workflow on non-main branches
   issue_comment:
     types: [created, edited]
 
@@ -47,6 +48,7 @@ jobs:
           - rootfs-image
           - rootfs-initrd
           - shim-v2
+          - virtiofsd
     steps:
       - name: get-PR-ref
         id: get-PR-ref

.github/workflows/release.yaml

@@ -1,8 +1,8 @@
-name: Publish Kata 2.x release artifacts
+name: Publish Kata release artifacts
 on:
   push:
     tags:
-     - '2.*'
+      - '[0-9]+.[0-9]+.[0-9]+*'
 
 jobs:
   build-asset:
@@ -17,6 +17,7 @@ jobs:
           - rootfs-image
           - rootfs-initrd
           - shim-v2
+          - virtiofsd
     steps:
       - uses: actions/checkout@v2
       - name: Install docker

.github/workflows/snap-release.yaml

@@ -1,8 +1,9 @@
-name: Release Kata 2.x in snapcraft store
+name: Release Kata in snapcraft store
 on:
   push:
     tags:
-      - '2.*'
+      - '[0-9]+.[0-9]+.[0-9]+*'
+
 jobs:
   release-snap:
     runs-on: ubuntu-20.04
@@ -19,6 +20,8 @@ jobs:
 
       - name: Build snap
         run: |
+          # Removing man-db, workflow kept failing, fixes: #4480
+          sudo apt -y remove --purge man-db
           sudo apt-get install -y git git-extras
           kata_url="https://github.com/kata-containers/kata-containers"
           latest_version=$(git ls-remote --tags ${kata_url}  | egrep -o "refs.*" | egrep -v "\-alpha|\-rc|{}" | egrep -o "[[:digit:]]+\.[[:digit:]]+\.[[:digit:]]+" | sort -V -r | head -1)
@@ -26,7 +29,7 @@ jobs:
           # Check semantic versioning format (x.y.z) and if the current tag is the latest tag
           if echo "${current_version}" | grep -q "^[[:digit:]]\+\.[[:digit:]]\+\.[[:digit:]]\+$" && echo -e "$latest_version\n$current_version" | sort -C -V; then
             # Current version is the latest version, build it
-            snapcraft -d snap --destructive-mode
+            snapcraft snap --debug --destructive-mode
           fi
 
       - name: Upload snap

.github/workflows/snap.yaml

@@ -24,4 +24,4 @@ jobs:
       - name: Build snap
         if: ${{ !contains(github.event.pull_request.labels.*.name, 'force-skip-ci') }}
         run: |
-          snapcraft -d snap --destructive-mode
+          snapcraft snap --debug --destructive-mode

.gitignore

@@ -10,4 +10,5 @@ src/agent/kata-agent.service
 src/agent/protocols/src/*.rs
 !src/agent/protocols/src/lib.rs
 build
+src/tools/log-parser/kata-log-parser
 

Makefile

@@ -6,8 +6,10 @@
 # List of available components
 COMPONENTS =
 
+COMPONENTS += libs
 COMPONENTS += agent
 COMPONENTS += runtime
+COMPONENTS += runtime-rs
 
 # List of available tools
 TOOLS =
@@ -15,16 +17,12 @@ TOOLS =
 TOOLS += agent-ctl
 TOOLS += trace-forwarder
 TOOLS += runk
+TOOLS += log-parser
 
 STANDARD_TARGETS = build check clean install test vendor
 
 default: all
 
-all: logging-crate-tests build
-
-logging-crate-tests:
-	make -C src/libs/logging
-
 include utils.mk
 include ./tools/packaging/kata-deploy/local-build/Makefile
 
@@ -48,7 +46,6 @@ docs-url-alive-check:
 	binary-tarball \
 	default \
 	install-binary-tarball \
-	logging-crate-tests \
 	static-checks \
 	docs-url-alive-check
 

README.md

@@ -71,6 +71,7 @@ See the [official documentation](docs) including:
 - [Developer guide](docs/Developer-Guide.md)
 - [Design documents](docs/design)
   - [Architecture overview](docs/design/architecture)
+  - [Architecture 3.0 overview](docs/design/architecture_3.0/)
 
 ## Configuration
 
@@ -117,6 +118,8 @@ The table below lists the core parts of the project:
 |-|-|-|
 | [runtime](src/runtime) | core | Main component run by a container manager and providing a containerd shimv2 runtime implementation. |
 | [agent](src/agent) | core | Management process running inside the virtual machine / POD that sets up the container environment. |
+| [libraries](src/libs) | core | Library crates shared by multiple Kata Container components or published to [`crates.io`](https://crates.io/index.html) |
+| [`dragonball`](src/dragonball) | core | An optional built-in VMM brings out-of-the-box Kata Containers experience with optimizations on container workloads |
 | [documentation](docs) | documentation | Documentation common to all components (such as design and install documentation). |
 | [libraries](src/libs) | core | Library crates shared by multiple Kata Container components or published to [`crates.io`](https://crates.io/index.html) |
 | [tests](https://github.com/kata-containers/tests) | tests | Excludes unit tests which live with the main code. |
@@ -140,7 +143,7 @@ The table below lists the remaining parts of the project:
 
 Kata Containers is now
 [available natively for most distributions](docs/install/README.md#packaged-installation-methods).
-However, packaging scripts and metadata are still used to generate snap and GitHub releases. See
+However, packaging scripts and metadata are still used to generate [snap](snap/local) and GitHub releases. See
 the [components](#components) section for further details.
 
 ## Glossary of Terms

VERSION

@@ -1 +1 @@
-2.5.0-alpha1
+3.0.0-alpha0

ci/darwin-test.sh

@@ -11,10 +11,10 @@ runtimedir=$cidir/../src/runtime
 
 build_working_packages() {
 	# working packages:
-	device_api=$runtimedir/virtcontainers/device/api
-	device_config=$runtimedir/virtcontainers/device/config
-	device_drivers=$runtimedir/virtcontainers/device/drivers
-	device_manager=$runtimedir/virtcontainers/device/manager
+	device_api=$runtimedir/pkg/device/api
+	device_config=$runtimedir/pkg/device/config
+	device_drivers=$runtimedir/pkg/device/drivers
+	device_manager=$runtimedir/pkg/device/manager
 	rc_pkg_dir=$runtimedir/pkg/resourcecontrol/
 	utils_pkg_dir=$runtimedir/virtcontainers/utils
 

ci/go-test.sh

@@ -1,12 +0,0 @@
-#!/usr/bin/env bash
-#
-# Copyright (c) 2020 Intel Corporation
-#
-# SPDX-License-Identifier: Apache-2.0
-
-set -e
-
-cidir=$(dirname "$0")
-source "${cidir}/lib.sh"
-
-run_go_test

ci/lib.sh

@@ -18,6 +18,13 @@ clone_tests_repo()
 {
 	if [ -d "$tests_repo_dir" ]; then
 		[ -n "${CI:-}" ] && return
+		# git config --global --add safe.directory will always append
+		# the target to .gitconfig without checking the existence of
+		# the target, so it's better to check it before adding the target repo.
+		local sd="$(git config --global --get safe.directory ${tests_repo_dir} || true)"
+		if [ -z "${sd}" ]; then
+			git config --global --add safe.directory ${tests_repo_dir}
+		fi
 		pushd "${tests_repo_dir}"
 		git checkout "${branch}"
 		git pull
@@ -39,12 +46,6 @@ run_static_checks()
 	bash "$tests_repo_dir/.ci/static-checks.sh" "$@"
 }
 
-run_go_test()
-{
-	clone_tests_repo
-	bash "$tests_repo_dir/.ci/go-test.sh"
-}
-
 run_docs_url_alive_check()
 {
 	clone_tests_repo

docs/Developer-Guide.md

@@ -116,7 +116,7 @@ detailed below.
 The Kata logs appear in the `containerd` log files, along with logs from `containerd` itself.
 
 For more information about `containerd` debug, please see the
-[`containerd` documentation](https://github.com/containerd/containerd/blob/master/docs/getting-started.md).
+[`containerd` documentation](https://github.com/containerd/containerd/blob/main/docs/getting-started.md).
 
 #### Enabling full `containerd` debug
 
@@ -425,7 +425,7 @@ To build utilizing the same options as Kata, you should make use of the `configu
 $ cd $your_qemu_directory
 $ $packaging_dir/scripts/configure-hypervisor.sh kata-qemu > kata.cfg
 $ eval ./configure "$(cat kata.cfg)"
-$ make -j $(nproc)
+$ make -j $(nproc --ignore=1)
 $ sudo -E make install
 ```
 
@@ -465,7 +465,7 @@ script and paste its output directly into a
 > [runtime](../src/runtime) repository.
 
 To perform analysis on Kata logs, use the
-[`kata-log-parser`](https://github.com/kata-containers/tests/tree/main/cmd/log-parser)
+[`kata-log-parser`](../src/tools/log-parser)
 tool, which can convert the logs into formats (e.g. JSON, TOML, XML, and YAML).
 
 See [Set up a debug console](#set-up-a-debug-console).
@@ -700,11 +700,11 @@ options to have the kernel boot messages logged into the system journal.
 For generic information on enabling debug in the configuration file, see the
 [Enable full debug](#enable-full-debug) section.
 
-The kernel boot messages will appear in the `containerd` or `CRI-O` log appropriately,
+The kernel boot messages will appear in the `kata` logs (and in the `containerd` or `CRI-O` log appropriately).
 such as:
 
 ```bash
-$ sudo journalctl -t containerd
+$ sudo journalctl -t kata
 -- Logs begin at Thu 2020-02-13 16:20:40 UTC, end at Thu 2020-02-13 16:30:23 UTC. --
 ...
 time="2020-09-15T14:56:23.095113803+08:00" level=debug msg="reading guest console" console-protocol=unix console-url=/run/vc/vm/ab9f633385d4987828d342e47554fc6442445b32039023eeddaa971c1bb56791/console.sock pid=107642 sandbox=ab9f633385d4987828d342e47554fc6442445b32039023eeddaa971c1bb56791 source=virtcontainers subsystem=sandbox vmconsole="[    0.395399] brd: module loaded"
@@ -714,3 +714,4 @@ time="2020-09-15T14:56:23.105268162+08:00" level=debug msg="reading guest consol
 time="2020-09-15T14:56:23.121121598+08:00" level=debug msg="reading guest console" console-protocol=unix console-url=/run/vc/vm/ab9f633385d4987828d342e47554fc6442445b32039023eeddaa971c1bb56791/console.sock pid=107642 sandbox=ab9f633385d4987828d342e47554fc6442445b32039023eeddaa971c1bb56791 source=virtcontainers subsystem=sandbox vmconsole="[    0.421324] memmap_init_zone_device initialised 32768 pages in 12ms"
 ...
 ```
+Refer to the [kata-log-parser documentation](../src/tools/log-parser/README.md) which is useful to fetch these.

docs/Limitations.md

@@ -60,17 +60,26 @@ This section lists items that might be possible to fix.
 ## OCI CLI commands
 
 ### Docker and Podman support
-Currently Kata Containers does not support Docker or Podman.
+Currently Kata Containers does not support Podman.
 
 See issue https://github.com/kata-containers/kata-containers/issues/722 for more information.
 
+Docker supports Kata Containers since 22.06:
+
+```bash
+$ sudo docker run --runtime io.containerd.kata.v2
+```
+
+Kata Containers works perfectly with containerd, we recommend to use
+containerd's Docker-style command line tool [`nerdctl`](https://github.com/containerd/nerdctl).
+
 ## Runtime commands
 
 ### checkpoint and restore
 
 The runtime does not provide `checkpoint` and `restore` commands. There
 are discussions about using VM save and restore to give us a
-`[criu](https://github.com/checkpoint-restore/criu)`-like functionality,
+[`criu`](https://github.com/checkpoint-restore/criu)-like functionality,
 which might provide a solution.
 
 Note that the OCI standard does not specify `checkpoint` and `restore`
@@ -93,6 +102,42 @@ All other configurations are supported and are working properly.
 
 ## Networking
 
+### Host network
+
+Host network (`nerdctl/docker run --net=host`or [Kubernetes `HostNetwork`](https://kubernetes.io/docs/reference/kubernetes-api/workload-resources/pod-v1/#hosts-namespaces)) is not supported.
+It is not possible to directly access the host networking configuration
+from within the VM.
+
+The `--net=host` option can still be used with `runc` containers and
+inter-mixed with running Kata Containers, thus enabling use of `--net=host`
+when necessary.
+
+It should be noted, currently passing the `--net=host` option into a
+Kata Container may result in the Kata Container networking setup
+modifying, re-configuring and therefore possibly breaking the host
+networking setup. Do not use `--net=host` with Kata Containers.
+
+### Support for joining an existing VM network
+
+Docker supports the ability for containers to join another containers
+namespace with the `docker run --net=containers` syntax. This allows
+multiple containers to share a common network namespace and the network
+interfaces placed in the network namespace. Kata Containers does not
+support network namespace sharing. If a Kata Container is setup to
+share the network namespace of a `runc` container, the runtime
+effectively takes over all the network interfaces assigned to the
+namespace and binds them to the VM. Consequently, the `runc` container loses
+its network connectivity.
+
+### docker run --link
+
+The runtime does not support the `docker run --link` command. This
+command is now deprecated by docker and we have no intention of adding support.
+Equivalent functionality can be achieved with the newer docker networking commands.
+
+See more documentation at
+[docs.docker.com](https://docs.docker.com/network/links/).
+
 ## Resource management
 
 Due to the way VMs differ in their CPU and memory allocation, and sharing

docs/Release-Process.md

@@ -4,11 +4,11 @@
 ## Requirements
 
 - [hub](https://github.com/github/hub)
-  * Using an [application token](https://github.com/settings/tokens) is required for hub.
+  * Using an [application token](https://github.com/settings/tokens) is required for hub (set to a GITHUB_TOKEN environment variable).
 
 - GitHub permissions to push tags and create releases in Kata repositories.
 
-- GPG configured to sign git tags. https://help.github.com/articles/generating-a-new-gpg-key/
+- GPG configured to sign git tags. https://docs.github.com/en/authentication/managing-commit-signature-verification/generating-a-new-gpg-key
 
 - You should configure your GitHub to use your ssh keys (to push to branches). See https://help.github.com/articles/adding-a-new-ssh-key-to-your-github-account/.
     * As an alternative, configure hub to push and fork with HTTPS, `git config --global hub.protocol https` (Not tested yet) *
@@ -48,7 +48,7 @@
 ### Merge all bump version Pull requests
 
   - The above step will create a GitHub pull request in the Kata projects. Trigger the CI using `/test` command on each bump Pull request.
-  - Trigger the test-kata-deploy workflow on the kata-containers repository bump Pull request using `/test_kata_deploy` (monitor under the "action" tab).
+  - Trigger the `test-kata-deploy` workflow which is under the `Actions` tab on the repository GitHub page (make sure to select the correct branch and validate it passes).
   - Check any failures and fix if needed.
   - Work with the Kata approvers to verify that the CI works and the pull requests are merged.
 

docs/Unit-Test-Advice.md

@@ -320,7 +320,7 @@ mod tests {
 
 ## Test user
 
-[Unit tests are run *twice*](https://github.com/kata-containers/tests/blob/main/.ci/go-test.sh):
+[Unit tests are run *twice*](../src/runtime/go-test.sh):
 
 - as the current user
 - as the `root` user (if different to the current user)

docs/code-pr-advice.md

@@ -79,7 +79,7 @@ a "`BUG: feature X not implemented see {bug-url}`" type error.
 - Don't use multiple log calls when a single log call could be used.
 
 - Use structured logging where possible to allow
-  [standard tooling](https://github.com/kata-containers/tests/tree/main/cmd/log-parser)
+  [standard tooling](../src/tools/log-parser)
   be able to extract the log fields.
 
 ### Names

docs/design/README.md

@@ -12,7 +12,7 @@ Kata Containers design documents:
 - [Metrics(Kata 2.0)](kata-2-0-metrics.md)
 - [Design for Kata Containers `Lazyload` ability with `nydus`](kata-nydus-design.md)
 - [Design for direct-assigned volume](direct-blk-device-assignment.md)
-
+- [Design for core-scheduling](core-scheduling.md)
 ---
 
 - [Design proposals](proposals)

docs/design/architecture/kubernetes.md

@@ -17,7 +17,7 @@ Kubelet instance is responsible for managing the lifecycle of pods
 within the nodes and eventually relies on a container runtime to
 handle execution. The Kubelet architecture decouples lifecycle
 management from container execution through a dedicated gRPC based
-[Container Runtime Interface (CRI)](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node/container-runtime-interface-v1.md).
+[Container Runtime Interface (CRI)](https://github.com/kubernetes/design-proposals-archive/blob/main/node/container-runtime-interface-v1.md).
 
 In other words, a Kubelet is a CRI client and expects a CRI
 implementation to handle the server side of the interface.

docs/design/architecture/storage.md

@@ -1,5 +1,17 @@
 # Storage
 
+## Limits
+
+Kata Containers is [compatible](README.md#compatibility) with existing
+standards and runtime. From the perspective of storage, this means no
+limits are placed on the amount of storage a container
+[workload](README.md#workload) may use.
+
+Since cgroups are not able to set limits on storage allocation, if you
+wish to constrain the amount of storage a container uses, consider
+using an existing facility such as `quota(1)` limits or
+[device mapper](#devicemapper) limits.
+
 ## virtio SCSI
 
 If a block-based graph driver is [configured](README.md#configuration),
@@ -20,7 +32,7 @@ For virtio-fs, the [runtime](README.md#runtime) starts one `virtiofsd` daemon
 ## Devicemapper
 
 The
-[devicemapper `snapshotter`](https://github.com/containerd/containerd/tree/master/snapshots/devmapper)
+[devicemapper `snapshotter`](https://github.com/containerd/containerd/tree/main/snapshots/devmapper)
 is a special case. The `snapshotter` uses dedicated block devices
 rather than formatted filesystems, and operates at the block level
 rather than the file level. This knowledge is used to directly use the

docs/design/architecture_3.0/README.md

@@ -0,0 +1,169 @@
+# Kata 3.0 Architecture
+## Overview
+In cloud-native scenarios, there is an increased demand for container startup speed, resource consumption, stability, and security, areas where the present Kata Containers runtime is challenged relative to other runtimes. To achieve this, we propose a solid, field-tested and secure Rust version of the kata-runtime.
+
+Also, we provide the following designs:
+
+- Turn key solution with builtin `Dragonball` Sandbox
+- Async I/O to reduce resource consumption
+- Extensible framework for multiple services, runtimes and hypervisors
+- Lifecycle management for sandbox and container associated resources
+
+### Rationale for choosing Rust
+
+We chose Rust because it is designed as a system language with a focus on efficiency.
+In contrast to Go, Rust makes a variety of design trade-offs in order to obtain
+good execution performance, with innovative techniques that, in contrast to C or
+C++, provide reasonable protection against common memory errors (buffer
+overflow, invalid pointers, range errors), error checking (ensuring errors are
+dealt with), thread safety, ownership of resources, and more.
+
+These benefits were verified in our project when the Kata Containers guest agent
+was rewritten in Rust. We notably saw a significant reduction in memory usage
+with the Rust-based implementation.
+
+
+## Design
+### Architecture
+![architecture](./images/architecture.png)
+### Built-in VMM
+#### Current Kata 2.x architecture
+![not_builtin_vmm](./images/not_built_in_vmm.png)
+As shown in the figure, runtime and VMM are separate processes. The runtime process forks the VMM process and interacts through the inter-process RPC. Typically, process interaction consumes more resources than peers within the process, and it will result in relatively low efficiency. At the same time, the cost of resource operation and maintenance should be considered. For example, when performing resource recovery under abnormal conditions, the exception of any process must be detected by others and activate the appropriate resource recovery process. If there are additional processes, the recovery becomes even more difficult.
+#### How To Support Built-in VMM
+We provide `Dragonball` Sandbox to enable built-in VMM by integrating VMM's function into the Rust library. We could perform VMM-related functionalities by using the library. Because runtime and VMM  are in the same process, there is a benefit in terms of message processing speed and API synchronization. It can also guarantee the consistency of the runtime and the VMM life cycle, reducing resource recovery and exception handling maintenance, as shown in the figure:
+![builtin_vmm](./images/built_in_vmm.png)
+### Async Support
+#### Why Need Async
+**Async is already in stable Rust and allows us to write async code**
+
+- Async provides significantly reduced CPU and memory overhead, especially for workloads with a large amount of IO-bound tasks
+- Async is zero-cost in Rust, which means that you only pay for what you use. Specifically, you can use async without heap allocations and dynamic dispatch, which greatly improves efficiency
+- For more (see [Why Async?](https://rust-lang.github.io/async-book/01_getting_started/02_why_async.html) and [The State of Asynchronous Rust](https://rust-lang.github.io/async-book/01_getting_started/03_state_of_async_rust.html)).
+
+**There may be several problems if implementing kata-runtime with Sync Rust**
+
+- Too many threads with a new TTRPC connection
+   - TTRPC threads: reaper thread(1) + listener thread(1) + client handler(2)
+- Add 3 I/O threads with a new container
+- In Sync mode, implementing a timeout mechanism is challenging. For example, in TTRPC API interaction, the timeout mechanism is difficult to align with Golang
+#### How To Support Async
+The kata-runtime is controlled by TOKIO_RUNTIME_WORKER_THREADS to run the OS thread, which is 2 threads by default. For TTRPC and container-related threads run in the `tokio` thread in a unified manner, and related dependencies need to be switched to Async, such as Timer, File, Netlink, etc. With the help of Async, we can easily support no-block I/O and timer. Currently, we only utilize Async for kata-runtime. The built-in VMM keeps the OS thread because it can ensure that the threads are controllable.
+
+**For N tokio worker threads and M containers**
+
+- Sync runtime(both OS thread and `tokio` task are OS thread but without `tokio` worker thread)  OS thread number:  4 + 12*M
+- Async runtime(only OS thread is OS thread) OS thread number: 2 + N
+```shell
+├─ main(OS thread)
+├─ async-logger(OS thread)
+└─ tokio worker(N * OS thread)
+  ├─ agent log forwarder(1 * tokio task)
+  ├─ health check thread(1 * tokio task)
+  ├─ TTRPC reaper thread(M * tokio task)
+  ├─ TTRPC listener thread(M * tokio task)
+  ├─ TTRPC client handler thread(7 * M * tokio task)
+  ├─ container stdin io thread(M * tokio task)
+  ├─ container stdin io thread(M * tokio task)
+  └─ container stdin io thread(M * tokio task)
+```
+### Extensible Framework
+The Kata 3.x runtime is designed with the extension of service, runtime, and hypervisor, combined with configuration to meet the needs of different scenarios. At present, the service provides a register mechanism to support multiple services. Services could interact with runtime through messages. In addition, the runtime handler handles messages from services. To meet the needs of a binary that supports multiple runtimes and hypervisors, the startup must obtain the runtime handler type and hypervisor type through configuration.
+
+![framework](./images/framework.png)
+### Resource Manager
+In our case, there will be a variety of resources, and every resource has several subtypes. Especially for `Virt-Container`, every subtype of resource has different operations. And there may be dependencies, such as the share-fs rootfs and the share-fs volume will use share-fs resources to share files to the VM. Currently, network and share-fs are regarded as sandbox resources, while rootfs, volume, and cgroup are regarded as container resources. Also, we abstract a common interface for each resource and use subclass operations to evaluate the differences between different subtypes.
+![resource manager](./images/resourceManager.png)
+
+## Roadmap
+
+- Stage 1 (June): provide basic features (current delivered)
+- Stage 2 (September): support common features
+- Stage 3: support full features
+
+| **Class**                  | **Sub-Class**       | **Development Stage** |
+| -------------------------- | ------------------- | --------------------- |
+| Service                    | task service        | Stage 1               |
+|                            | extend service      | Stage 3               |
+|                            | image service       | Stage 3               |
+| Runtime handler            | `Virt-Container`    | Stage 1               |
+|                            | `Wasm-Container`    | Stage 3               |
+|                            | `Linux-Container`   | Stage 3               |
+| Endpoint                   | VETH Endpoint       | Stage 1               |
+|                            | Physical Endpoint   | Stage 2               |
+|                            | Tap Endpoint        | Stage 2               |
+|                            | `Tuntap` Endpoint   | Stage 2               |
+|                            | `IPVlan` Endpoint   | Stage 3               |
+|                            | `MacVlan` Endpoint  | Stage 3               |
+|                            | MACVTAP Endpoint    | Stage 3               |
+|                            | `VhostUserEndpoint` | Stage 3               |
+| Network Interworking Model | Tc filter           | Stage 1               |
+|                            | `MacVtap`           | Stage 3               |
+| Storage                    | Virtio-fs           | Stage 1               |
+|                            | `nydus`             | Stage 2               |
+| Hypervisor                 | `Dragonball`        | Stage 1               |
+|                            | QEMU                | Stage 2               |
+|                            | ACRN                | Stage 3               |
+|                            | Cloud Hypervisor    | Stage 3               |
+|                            | Firecracker         | Stage 3               |
+
+## FAQ
+
+- Are the "service", "message dispatcher" and "runtime handler" all part of the single Kata 3.x runtime binary?
+
+  Yes. They are components in Kata 3.x runtime. And they will be packed into one binary.
+  1. Service is an interface, which is responsible for handling multiple services like task service, image service and etc. 
+  2. Message dispatcher, it is used to match multiple requests from the service module. 
+  3. Runtime handler is used to deal with the operation for sandbox and container. 
+- What is the name of the Kata 3.x runtime binary?
+  
+  Apparently we can't use `containerd-shim-v2-kata` because it's already used. We are facing the hardest issue of "naming" again. Any suggestions are welcomed.
+  Internally we use `containerd-shim-v2-rund`.
+
+- Is the Kata 3.x design compatible with the containerd shimv2 architecture? 
+  
+  Yes. It is designed to follow the functionality of go version kata.  And it implements the `containerd shim v2` interface/protocol.
+
+- How will users migrate to the Kata 3.x architecture?
+  
+  The migration plan will be provided before the Kata 3.x is merging into the main branch.
+
+- Is `Dragonball` limited to its own built-in VMM? Can the `Dragonball` system be configured to work using an external `Dragonball` VMM/hypervisor?
+
+  The `Dragonball` could work as an external hypervisor. However, stability and performance is challenging in this case. Built in VMM could optimise the container overhead, and it's easy to maintain stability.
+
+  `runD` is the `containerd-shim-v2` counterpart of `runC` and can run a pod/containers. `Dragonball` is a `microvm`/VMM that is designed to run container workloads. Instead of `microvm`/VMM, we sometimes refer to it as secure sandbox.
+
+- QEMU, Cloud Hypervisor and Firecracker support are planned, but how that would work. Are they working in separate process?
+ 
+  Yes. They are unable to work as built in VMM.
+
+- What is `upcall`?
+  
+    The `upcall` is used to hotplug CPU/memory/MMIO devices, and it solves two issues.
+    1. avoid dependency on PCI/ACPI
+    2. avoid dependency on `udevd` within guest and get deterministic results for hotplug operations. So `upcall` is an alternative to ACPI based CPU/memory/device hotplug. And we may cooperate with the community to add support for ACPI based CPU/memory/device hotplug if needed.
+   
+    `Dbs-upcall` is a `vsock-based` direct communication tool between VMM and guests. The server side of the `upcall` is a driver in guest kernel (kernel patches are needed for this feature) and it'll start to serve the requests once the kernel has started. And the client side is in VMM , it'll be a thread that communicates with VSOCK through `uds`. We have accomplished device hotplug / hot-unplug directly through `upcall` in order to avoid virtualization of ACPI  to minimize virtual machine's overhead. And there could be many other usage through this direct communication channel. It's already open source.
+   https://github.com/openanolis/dragonball-sandbox/tree/main/crates/dbs-upcall 
+
+- The URL below says the kernel patches work with 4.19, but do they also work with 5.15+ ?
+  
+  Forward compatibility should be achievable, we have ported it to 5.10 based kernel.
+
+- Are these patches platform-specific or would they work for any architecture that supports VSOCK?
+  
+  It's almost platform independent, but some message related to CPU hotplug are platform dependent.
+
+- Could the kernel driver be replaced with a userland daemon in the guest using loopback VSOCK?
+  
+  We need to create device nodes for hot-added CPU/memory/devices, so it's not easy for userspace daemon to do these tasks.
+
+- The fact that `upcall` allows communication between the VMM and the guest suggests that this architecture might be incompatible with https://github.com/confidential-containers where the VMM should have no knowledge of what happens inside the VM.
+  
+  1. `TDX` doesn't support CPU/memory hotplug yet.
+  2. For ACPI based device hotplug, it depends on ACPI `DSDT` table, and the guest kernel will execute `ASL` code to handle during handling those hotplug event. And it should be easier to audit VSOCK based communication than ACPI `ASL` methods.
+
+- What is the security boundary for the monolithic / "Built-in VMM" case?
+  
+  It has the security boundary of virtualization. More details will be provided in next stage.

docs/design/core-scheduling.md

@@ -0,0 +1,12 @@
+# Core scheduling
+
+Core scheduling is a Linux kernel feature that allows only trusted tasks to run concurrently on
+CPUs sharing compute resources (for example, hyper-threads on a core).
+
+Containerd versions >= 1.6.4 leverage this to treat all of the processes associated with a
+given pod or container to be a single group of trusted tasks. To indicate this should be carried
+out, containerd sets the `SCHED_CORE` environment variable for each shim it spawns. When this is
+set, the Kata Containers shim implementation uses the `prctl` syscall to create a new core scheduling
+domain for the shim process itself as well as future VMM processes it will start.
+
+For more details on the core scheduling feature, see the [Linux documentation](https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/core-scheduling.html).

docs/design/host-cgroups.md

@@ -12,7 +12,7 @@ The OCI [runtime specification][linux-config] provides guidance on where the con
   > [`cgroupsPath`][cgroupspath]: (string, OPTIONAL) path to the cgroups. It can be used to either control the cgroups
   > hierarchy for containers or to run a new process in an existing container
 
-Cgroups are hierarchical, and this can be seen with the following pod example:
+The cgroups are hierarchical, and this can be seen with the following pod example:
 
 - Pod 1: `cgroupsPath=/kubepods/pod1`
   - Container 1: `cgroupsPath=/kubepods/pod1/container1`
@@ -247,14 +247,14 @@ cgroup size and constraints accordingly.
 
 # Supported cgroups
 
-Kata Containers currently only supports cgroups `v1`. 
+Kata Containers currently supports cgroups `v1` and `v2`. 
 
 In the following sections each cgroup is described briefly.
 
-## Cgroups V1
+## cgroups v1
 
-`Cgroups V1` are under a [`tmpfs`][1] filesystem mounted at `/sys/fs/cgroup`, where each cgroup is
-mounted under a separate cgroup filesystem. A `Cgroups v1` hierarchy may look like the following
+`cgroups v1` are under a [`tmpfs`][1] filesystem mounted at `/sys/fs/cgroup`, where each cgroup is
+mounted under a separate cgroup filesystem. A `cgroups v1` hierarchy may look like the following
 diagram:
 
 ```
@@ -301,13 +301,12 @@ diagram:
 A process can join a cgroup by writing its process id (`pid`) to `cgroup.procs` file,
 or join a cgroup partially by writing the task (thread) id (`tid`) to the `tasks` file.
 
-Kata Containers only supports `v1`.
 To know more about `cgroups v1`, see [cgroupsv1(7)][2].
 
-## Cgroups V2
+## cgroups v2
 
-`Cgroups v2` are also known as unified cgroups, unlike `cgroups v1`, the cgroups are
-mounted under the same cgroup filesystem. A `Cgroups v2` hierarchy may look like the following
+`cgroups v2` are also known as unified cgroups, unlike `cgroups v1`, the cgroups are
+mounted under the same cgroup filesystem. A `cgroups v2` hierarchy may look like the following
 diagram:
 
 ```
@@ -354,8 +353,6 @@ Same as `cgroups v1`, a process can join the cgroup by writing its process id (`
 `cgroup.procs` file, or join a cgroup partially by writing the task (thread) id (`tid`) to
 `cgroup.threads` file.
 
-Kata Containers does not support cgroups `v2` on the host.
-
 ### Distro Support
 
 Many Linux distributions do not yet support `cgroups v2`, as it is quite a recent addition.

docs/how-to/README.md

@@ -5,7 +5,7 @@
 - [Run Kata containers with `crictl`](run-kata-with-crictl.md)
 - [Run Kata Containers with Kubernetes](run-kata-with-k8s.md)
 - [How to use Kata Containers and Containerd](containerd-kata.md)
-- [How to use Kata Containers and CRI (containerd) with Kubernetes](how-to-use-k8s-with-cri-containerd-and-kata.md)
+- [How to use Kata Containers and containerd with Kubernetes](how-to-use-k8s-with-containerd-and-kata.md)
 - [Kata Containers and service mesh for Kubernetes](service-mesh.md)
 - [How to import Kata Containers logs into Fluentd](how-to-import-kata-logs-with-fluentd.md)
 

docs/how-to/containerd-kata.md

@@ -40,7 +40,7 @@ use `RuntimeClass` instead of the deprecated annotations.
 ### Containerd Runtime V2 API: Shim V2 API
 
 The [`containerd-shim-kata-v2` (short as `shimv2` in this documentation)](../../src/runtime/cmd/containerd-shim-kata-v2/)
-implements the [Containerd Runtime V2 (Shim API)](https://github.com/containerd/containerd/tree/master/runtime/v2) for Kata.
+implements the [Containerd Runtime V2 (Shim API)](https://github.com/containerd/containerd/tree/main/runtime/v2) for Kata.
 With `shimv2`, Kubernetes can launch Pod and OCI-compatible containers with one shim per Pod. Prior to `shimv2`, `2N+1` 
 shims (i.e. a `containerd-shim` and a `kata-shim` for each container and the Pod sandbox itself) and no standalone `kata-proxy` 
 process were used, even with VSOCK not available.
@@ -72,7 +72,6 @@ $ command -v containerd
 
 ### Install CNI plugins
 
-> **Note:** You do not need to install CNI plugins if you do not want to use containerd with Kubernetes.
 > If you have installed Kubernetes with `kubeadm`, you might have already installed the CNI plugins.
 
 You can manually install CNI plugins as follows:
@@ -94,8 +93,8 @@ $ popd
 You can install the `cri-tools` from source code:
 
 ```bash
-$ go get github.com/kubernetes-incubator/cri-tools
-$ pushd $GOPATH/src/github.com/kubernetes-incubator/cri-tools
+$ go get github.com/kubernetes-sigs/cri-tools
+$ pushd $GOPATH/src/github.com/kubernetes-sigs/cri-tools
 $ make
 $ sudo -E make install
 $ popd
@@ -131,74 +130,42 @@ For
 
 The `RuntimeClass` is suggested.
 
-The following configuration includes three runtime classes:
+The following configuration includes two runtime classes:
 - `plugins.cri.containerd.runtimes.runc`: the runc, and it is the default runtime.
-- `plugins.cri.containerd.runtimes.kata`: The function in containerd (reference [the document here](https://github.com/containerd/containerd/tree/master/runtime/v2#binary-naming)) 
+- `plugins.cri.containerd.runtimes.kata`: The function in containerd (reference [the document here](https://github.com/containerd/containerd/tree/main/runtime/v2#binary-naming)) 
   where the dot-connected string `io.containerd.kata.v2` is translated to `containerd-shim-kata-v2` (i.e. the 
-  binary name of the Kata implementation of [Containerd Runtime V2 (Shim API)](https://github.com/containerd/containerd/tree/master/runtime/v2)).
-- `plugins.cri.containerd.runtimes.katacli`: the `containerd-shim-runc-v1` calls `kata-runtime`, which is the legacy process.
+  binary name of the Kata implementation of [Containerd Runtime V2 (Shim API)](https://github.com/containerd/containerd/tree/main/runtime/v2)).
 
 ```toml
     [plugins.cri.containerd]
       no_pivot = false
     [plugins.cri.containerd.runtimes]
-      [plugins.cri.containerd.runtimes.runc]
-         runtime_type = "io.containerd.runc.v1"
-         [plugins.cri.containerd.runtimes.runc.options]
-           NoPivotRoot = false
-           NoNewKeyring = false
-           ShimCgroup = ""
-           IoUid = 0
-           IoGid = 0
-           BinaryName = "runc"
-           Root = ""
-           CriuPath = ""
-           SystemdCgroup = false
+      [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
+         privileged_without_host_devices = false
+         runtime_type = "io.containerd.runc.v2"
+        [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
+            BinaryName = ""
+            CriuImagePath = ""
+            CriuPath = ""
+            CriuWorkPath = ""
+            IoGid = 0
       [plugins.cri.containerd.runtimes.kata]
          runtime_type = "io.containerd.kata.v2"
-      [plugins.cri.containerd.runtimes.katacli]
-         runtime_type = "io.containerd.runc.v1"
-         [plugins.cri.containerd.runtimes.katacli.options]
-           NoPivotRoot = false
-           NoNewKeyring = false
-           ShimCgroup = ""
-           IoUid = 0
-           IoGid = 0
-           BinaryName = "/usr/bin/kata-runtime"
-           Root = ""
-           CriuPath = ""
-           SystemdCgroup = false
-```
-
-From Containerd v1.2.4 and Kata v1.6.0, there is a new runtime option supported, which allows you to specify a specific Kata configuration file as follows:
-
-```toml
-      [plugins.cri.containerd.runtimes.kata]
-         runtime_type = "io.containerd.kata.v2"
-	 privileged_without_host_devices = true
-	 [plugins.cri.containerd.runtimes.kata.options]
-	   ConfigPath = "/etc/kata-containers/config.toml"
+         privileged_without_host_devices = true
+         pod_annotations = ["io.katacontainers.*"]
+         container_annotations = ["io.katacontainers.*"]
+         [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.kata.options]
+            ConfigPath = "/opt/kata/share/defaults/kata-containers/configuration.toml"
 ```
 
 `privileged_without_host_devices` tells containerd that a privileged Kata container should not have direct access to all host devices. If unset, containerd will pass all host devices to Kata container, which may cause security issues.
 
+`pod_annotations` is the list of pod annotations passed to both the pod sandbox as well as container through the OCI config.
+
+`container_annotations` is the list of container annotations passed through to the OCI config of the containers.
+
 This `ConfigPath` option is optional. If you do not specify it, shimv2 first tries to get the configuration file from the environment variable `KATA_CONF_FILE`. If neither are set, shimv2 will use the default Kata configuration file paths (`/etc/kata-containers/configuration.toml` and `/usr/share/defaults/kata-containers/configuration.toml`).
 
-If you use Containerd older than v1.2.4 or a version of Kata older than v1.6.0  and also want to specify a configuration file, you can use the following workaround, since the shimv2 accepts an environment variable, `KATA_CONF_FILE` for the configuration file path. Then, you can create a
-shell script with the following:
-
-```bash
-#!/usr/bin/env bash
-KATA_CONF_FILE=/etc/kata-containers/firecracker.toml containerd-shim-kata-v2 $@
-```
-
-Name it as `/usr/local/bin/containerd-shim-katafc-v2` and reference it in the configuration of containerd:
-
-```toml
-      [plugins.cri.containerd.runtimes.kata-firecracker]
-         runtime_type = "io.containerd.katafc.v2"
-```
-
 #### Kata Containers as the runtime for untrusted workload
 
 For cases without `RuntimeClass` support, we can use the legacy annotation method to support using Kata Containers 
@@ -218,28 +185,8 @@ and then, run an untrusted workload with Kata Containers:
       runtime_type = "io.containerd.kata.v2"
 ```
 
-For the earlier versions of Kata Containers and containerd that do not support Runtime V2 (Shim API), you can use the following alternative configuration:
-
-```toml
-    [plugins.cri.containerd]
-  
-    # "plugins.cri.containerd.default_runtime" is the runtime to use in containerd.
-    [plugins.cri.containerd.default_runtime]
-      # runtime_type is the runtime type to use in containerd e.g. io.containerd.runtime.v1.linux
-      runtime_type = "io.containerd.runtime.v1.linux"
-
-    # "plugins.cri.containerd.untrusted_workload_runtime" is a runtime to run untrusted workloads on it.
-    [plugins.cri.containerd.untrusted_workload_runtime]
-      # runtime_type is the runtime type to use in containerd e.g. io.containerd.runtime.v1.linux
-      runtime_type = "io.containerd.runtime.v1.linux"
-
-      # runtime_engine is the name of the runtime engine used by containerd.
-      runtime_engine = "/usr/bin/kata-runtime"
-```
-
 You can find more information on the [Containerd config documentation](https://github.com/containerd/cri/blob/master/docs/config.md)
 
-
 #### Kata Containers as the default runtime
 
 If you want to set Kata Containers as the only runtime in the deployment, you can simply configure as follows:
@@ -250,15 +197,6 @@ If you want to set Kata Containers as the only runtime in the deployment, you ca
       runtime_type = "io.containerd.kata.v2"
 ```
 
-Alternatively, for the earlier versions of Kata Containers and containerd that do not support Runtime V2 (Shim API), you can use the following alternative configuration:
-
-```toml
-    [plugins.cri.containerd]
-    [plugins.cri.containerd.default_runtime]
-      runtime_type = "io.containerd.runtime.v1.linux"
-      runtime_engine = "/usr/bin/kata-runtime"
-```
-
 ### Configuration for `cri-tools`
 
 > **Note:** If you skipped the [Install `cri-tools`](#install-cri-tools) section, you can skip this section too.
@@ -312,10 +250,12 @@ To run a container with Kata Containers through the containerd command line, you
 
 ```bash
 $ sudo ctr image pull docker.io/library/busybox:latest
-$ sudo ctr run --runtime io.containerd.run.kata.v2 -t --rm docker.io/library/busybox:latest hello sh
+$ sudo ctr run --cni --runtime io.containerd.run.kata.v2 -t --rm docker.io/library/busybox:latest hello sh
 ```
 
 This launches a BusyBox container named `hello`, and it will be removed by `--rm` after it quits.
+The `--cni` flag enables CNI networking for the container. Without this flag, a container with just a
+loopback interface is created.
 
 ### Launch Pods with `crictl` command line
 

docs/how-to/data/kata-monitor-daemonset.yml

@@ -45,6 +45,9 @@ spec:
         - name: containerdsocket
           mountPath: /run/containerd/containerd.sock
           readOnly: true
+        - name: sbs
+          mountPath: /run/vc/sbs/
+          readOnly: true
       terminationGracePeriodSeconds: 30
       volumes:
       - name: containerdtask
@@ -53,3 +56,6 @@ spec:
       - name: containerdsocket
         hostPath:
           path: /run/containerd/containerd.sock
+      - name: sbs
+        hostPath:
+          path: /run/vc/sbs/

docs/how-to/how-to-import-kata-logs-with-fluentd.md

@@ -68,7 +68,7 @@ the Kata logs import to the EFK stack.
 > stack they are able to utilise in order to modify and test as necessary.
 
 Minikube by default
-[configures](https://github.com/kubernetes/minikube/blob/master/deploy/iso/minikube-iso/board/coreos/minikube/rootfs-overlay/etc/systemd/journald.conf)
+[configures](https://github.com/kubernetes/minikube/blob/master/deploy/iso/minikube-iso/board/minikube/x86_64/rootfs-overlay/etc/systemd/journald.conf)
  the `systemd-journald` with the
 [`Storage=volatile`](https://www.freedesktop.org/software/systemd/man/journald.conf.html) option,
 which results in the journal being stored in `/run/log/journal`. Unfortunately, the Minikube EFK
@@ -163,7 +163,7 @@ sub-filter on, for instance, the `SYSLOG_IDENTIFIER` to differentiate the Kata c
 on the `PRIORITY` to filter out critical issues etc.
 
 Kata generates a significant amount of Kata specific information, which can be seen as
-[`logfmt`](https://github.com/kata-containers/tests/tree/main/cmd/log-parser#logfile-requirements).
+[`logfmt`](../../src/tools/log-parser/README.md#logfile-requirements).
 data contained in the `MESSAGE` field. Imported as-is, there is no easy way to filter on that data
 in Kibana:
 

docs/how-to/how-to-set-prometheus-in-k8s.md

@@ -19,7 +19,7 @@ Also you should ensure that `kubectl` working correctly.
 > **Note**: More information about Kubernetes integrations:
 >   - [Run Kata Containers with Kubernetes](run-kata-with-k8s.md)
 >   - [How to use Kata Containers and Containerd](containerd-kata.md)
->   - [How to use Kata Containers and CRI (containerd plugin) with Kubernetes](how-to-use-k8s-with-cri-containerd-and-kata.md)
+>   - [How to use Kata Containers and containerd with Kubernetes](how-to-use-k8s-with-containerd-and-kata.md)
 
 ## Configure Prometheus
 

docs/how-to/how-to-set-sandbox-config-kata.md

@@ -91,6 +91,7 @@ There are several kinds of Kata configurations and they are listed below.
 | `io.katacontainers.config.hypervisor.virtio_fs_daemon` | string | virtio-fs `vhost-user` daemon path |
 | `io.katacontainers.config.hypervisor.virtio_fs_extra_args` | string | extra options passed to `virtiofs` daemon |
 | `io.katacontainers.config.hypervisor.enable_guest_swap` | `boolean` | enable swap in the guest |
+| `io.katacontainers.config.hypervisor.use_legacy_serial` | `boolean` | uses legacy serial device for guest's console (QEMU) |
 
 ## Container Options
 | Key | Value Type | Comments |
@@ -172,7 +173,7 @@ kind: Pod
 metadata:
   name: pod2
   annotations:
-    io.katacontainers.config.runtime.disable_guest_seccomp: false
+    io.katacontainers.config.runtime.disable_guest_seccomp: "false"
 spec:
   runtimeClassName: kata
   containers:

docs/how-to/how-to-use-k8s-with-containerd-and-kata.md

@@ -1,15 +1,15 @@
-# How to use Kata Containers and CRI (containerd plugin) with Kubernetes
+# How to use Kata Containers and containerd with Kubernetes
 
 This document describes how to set up a single-machine Kubernetes (k8s) cluster.
 
 The Kubernetes cluster will use the
-[CRI containerd](https://github.com/containerd/containerd/) and
-[Kata Containers](https://katacontainers.io) to launch untrusted workloads.
+[containerd](https://github.com/containerd/containerd/) and
+[Kata Containers](https://katacontainers.io) to launch workloads.
 
 ## Requirements
 
 - Kubernetes, Kubelet, `kubeadm`
-- containerd with `cri` plug-in
+- containerd
 - Kata Containers
 
 > **Note:** For information about the supported versions of these components,
@@ -149,7 +149,7 @@ $ sudo -E kubectl taint nodes --all node-role.kubernetes.io/master-
 
 ## Create runtime class for Kata Containers
 
-By default, all pods are created with the default runtime configured in CRI containerd plugin.
+By default, all pods are created with the default runtime configured in containerd.
 From Kubernetes v1.12, users can use [`RuntimeClass`](https://kubernetes.io/docs/concepts/containers/runtime-class/#runtime-class) to specify a different runtime for Pods.
 
 ```bash
@@ -166,7 +166,7 @@ $ sudo -E kubectl apply -f runtime.yaml
 
 ## Run pod in Kata Containers
 
-If a pod has the `runtimeClassName` set to `kata`, the CRI plugin runs the pod with the
+If a pod has the `runtimeClassName` set to `kata`, the CRI runs the pod with the
 [Kata Containers runtime](../../src/runtime/README.md).
 
 - Create an pod configuration that using Kata Containers runtime

docs/how-to/privileged.md

@@ -31,7 +31,7 @@ See below example config:
   [plugins.cri]
     [plugins.cri.containerd]
        [plugins.cri.containerd.runtimes.runc]
-         runtime_type = "io.containerd.runc.v1"
+         runtime_type = "io.containerd.runc.v2"
          privileged_without_host_devices = false
        [plugins.cri.containerd.runtimes.kata]
          runtime_type = "io.containerd.kata.v2"
@@ -40,7 +40,7 @@ See below example config:
            ConfigPath = "/opt/kata/share/defaults/kata-containers/configuration.toml"
 ```
 
- - [Kata Containers with Containerd and CRI documentation](how-to-use-k8s-with-cri-containerd-and-kata.md)
+ - [How to use Kata Containers and containerd with Kubernetes](how-to-use-k8s-with-containerd-and-kata.md)
  - [Containerd CRI config documentation](https://github.com/containerd/containerd/blob/main/docs/cri/config.md)
 
 #### CRI-O

docs/how-to/run-kata-with-k8s.md

@@ -15,7 +15,7 @@ After choosing one CRI implementation, you must make the appropriate configurati
 to ensure it integrates with Kata Containers.
 
 Kata Containers 1.5 introduced the `shimv2` for containerd 1.2.0, reducing the components
-required to spawn pods and containers, and this is the preferred way to run Kata Containers with Kubernetes ([as documented here](../how-to/how-to-use-k8s-with-cri-containerd-and-kata.md#configure-containerd-to-use-kata-containers)).
+required to spawn pods and containers, and this is the preferred way to run Kata Containers with Kubernetes ([as documented here](../how-to/how-to-use-k8s-with-containerd-and-kata.md#configure-containerd-to-use-kata-containers)).
 
 An equivalent shim implementation for CRI-O is planned.
 
@@ -57,7 +57,7 @@ content shown below:
 
 To customize containerd to select Kata Containers runtime, follow our
 "Configure containerd to use Kata Containers" internal documentation
-[here](../how-to/how-to-use-k8s-with-cri-containerd-and-kata.md#configure-containerd-to-use-kata-containers).
+[here](../how-to/how-to-use-k8s-with-containerd-and-kata.md#configure-containerd-to-use-kata-containers).
 
 ## Install Kubernetes
 
@@ -85,7 +85,7 @@ Environment="KUBELET_EXTRA_ARGS=--container-runtime=remote --runtime-request-tim
 Environment="KUBELET_EXTRA_ARGS=--container-runtime=remote --runtime-request-timeout=15m --container-runtime-endpoint=unix:///run/containerd/containerd.sock"
 ```
 For more information about containerd see the "Configure Kubelet to use containerd"
-documentation [here](../how-to/how-to-use-k8s-with-cri-containerd-and-kata.md#configure-kubelet-to-use-containerd).
+documentation [here](../how-to/how-to-use-k8s-with-containerd-and-kata.md#configure-kubelet-to-use-containerd).
 
 ## Run a Kubernetes pod with Kata Containers
 
@@ -99,7 +99,18 @@ $ sudo systemctl restart kubelet
 $ sudo kubeadm init --ignore-preflight-errors=all --cri-socket /var/run/crio/crio.sock --pod-network-cidr=10.244.0.0/16
 
 # If using containerd
-$ sudo kubeadm init --ignore-preflight-errors=all --cri-socket /run/containerd/containerd.sock --pod-network-cidr=10.244.0.0/16
+$ cat <<EOF | tee kubeadm-config.yaml
+apiVersion: kubeadm.k8s.io/v1beta3
+kind: InitConfiguration
+nodeRegistration:
+  criSocket: "/run/containerd/containerd.sock"
+---
+kind: KubeletConfiguration
+apiVersion: kubelet.config.k8s.io/v1beta1
+cgroupDriver: cgroupfs
+podCIDR: "10.244.0.0/16"
+EOF
+$ sudo kubeadm init --ignore-preflight-errors=all --config kubeadm-config.yaml
 
 $ export KUBECONFIG=/etc/kubernetes/admin.conf
 ```

docs/hypervisors.md

@@ -33,6 +33,7 @@ are available, their default values and how each setting can be used.
 [Cloud Hypervisor] | rust | `aarch64`, `x86_64` | Type 2 ([KVM]) | `configuration-clh.toml` |
 [Firecracker] | rust | `aarch64`, `x86_64` | Type 2 ([KVM]) | `configuration-fc.toml` |
 [QEMU] | C | all | Type 2 ([KVM]) | `configuration-qemu.toml` |
+[`Dragonball`] | rust | `aarch64`, `x86_64` | Type 2 ([KVM]) | `configuration-dragonball.toml` |
 
 ## Determine currently configured hypervisor
 
@@ -52,6 +53,7 @@ the hypervisors:
 [Cloud Hypervisor] | Low latency, small memory footprint, small attack surface | Minimal | | excellent | excellent | High performance modern cloud workloads | |
 [Firecracker] | Very slimline | Extremely minimal | Doesn't support all device types | excellent | excellent | Serverless / FaaS | |
 [QEMU] | Lots of features | Lots | | good | good | Good option for most users | | All users |
+[`Dragonball`] | Built-in VMM,  low CPU and memory overhead| Minimal | | excellent | excellent | Optimized for most container workloads | `out-of-the-box` Kata Containers experience |
 
 For further details, see the [Virtualization in Kata Containers](design/virtualization.md) document and the official documentation for each hypervisor.
 
@@ -60,3 +62,4 @@ For further details, see the [Virtualization in Kata Containers](design/virtuali
 [Firecracker]: https://github.com/firecracker-microvm/firecracker
 [KVM]: https://en.wikipedia.org/wiki/Kernel-based_Virtual_Machine
 [QEMU]: http://www.qemu-project.org
+[`Dragonball`]: https://github.com/openanolis/dragonball-sandbox

docs/install/README.md

@@ -79,3 +79,6 @@ versions. This is not recommended for normal users.
 * [upgrading document](../Upgrading.md)
 * [developer guide](../Developer-Guide.md)
 * [runtime documentation](../../src/runtime/README.md)
+
+## Kata Containers 3.0 rust runtime installation
+* [installation guide](../install/kata-containers-3.0-rust-runtime-installation-guide.md)

docs/install/container-manager/containerd/containerd-install.md

@@ -19,12 +19,6 @@
 > - If you decide to proceed and install a Kata Containers release, you can
 >   still check for the latest version of Kata Containers by running
 >   `kata-runtime check --only-list-releases`.
->
-> - These instructions will not work for Fedora 31 and higher since those
->   distribution versions only support cgroups version 2 by default. However,
->   Kata Containers currently requires cgroups version 1 (on the host side). See
->   https://github.com/kata-containers/kata-containers/issues/927 for further
->   details.
 
 ## Install Kata Containers
 

docs/install/kata-containers-3.0-rust-runtime-installation-guide.md

@@ -0,0 +1,101 @@
+# Kata Containers 3.0 rust runtime installation
+The following is an overview of the different installation methods available. 
+
+## Prerequisites
+
+Kata Containers 3.0 rust runtime requires nested virtualization or bare metal. Check 
+[hardware requirements](/src/runtime/README.md#hardware-requirements) to see if your system is capable of running Kata 
+Containers.
+
+### Platform support
+
+Kata Containers 3.0 rust runtime currently runs on 64-bit systems supporting the following
+architectures:
+
+> **Notes:**
+> For other architectures, see https://github.com/kata-containers/kata-containers/issues/4320
+
+| Architecture | Virtualization technology |
+|-|-|
+| `x86_64`| [Intel](https://www.intel.com) VT-x |
+| `aarch64` ("`arm64`")| [ARM](https://www.arm.com) Hyp |
+
+## Packaged installation methods
+
+| Installation method                                  | Description                                                                                  | Automatic updates | Use case                                                                                      | Availability
+|------------------------------------------------------|----------------------------------------------------------------------------------------------|-------------------|-----------------------------------------------------------------------------------------------|----------- |
+| [Using kata-deploy](#kata-deploy-installation)       | The preferred way to deploy the Kata Containers distributed binaries on a Kubernetes cluster | **No!**           | Best way to give it a try on kata-containers on an already up and running Kubernetes cluster. | No |
+| [Using official distro packages](#official-packages) | Kata packages provided by Linux distributions official repositories                          | yes               | Recommended for most users. | No |                                                                   
+| [Using snap](#snap-installation)                     | Easy to install                                                                              | yes               | Good alternative to official distro packages.                                                 | No |
+| [Automatic](#automatic-installation)                 | Run a single command to install a full system                                                | **No!**           | For those wanting the latest release quickly.                                                 | No |
+| [Manual](#manual-installation)                       | Follow a guide step-by-step to install a working system                                      | **No!**           | For those who want the latest release with more control.                                      | No |
+| [Build from source](#build-from-source-installation) | Build the software components manually                                                       | **No!**           | Power users and developers only.  | Yes |              
+
+### Kata Deploy Installation
+`ToDo`
+### Official packages
+`ToDo`
+### Snap Installation
+`ToDo`
+### Automatic Installation
+`ToDo`
+### Manual Installation
+`ToDo`
+
+## Build from source installation
+
+### Rust Environment Set Up
+
+* Download `Rustup` and install  `Rust`
+    > **Notes:**
+    > Rust version 1.58 is needed
+
+    Example for `x86_64`
+    ```
+    $ curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
+    $ source $HOME/.cargo/env
+    $ rustup install 1.58
+    $ rustup default 1.58-x86_64-unknown-linux-gnu
+    ```
+
+* Musl support for fully static binary
+    
+    Example for `x86_64`
+    ```
+    $ rustup target add x86_64-unknown-linux-musl
+    ```
+* [Musl `libc`](http://musl.libc.org/) install
+
+    Example for musl 1.2.3
+    ```
+    $ curl -O https://git.musl-libc.org/cgit/musl/snapshot/musl-1.2.3.tar.gz
+    $ tar vxf musl-1.2.3.tar.gz
+    $ cd musl-1.2.3/
+    $ ./configure --prefix=/usr/local/
+    $ make && sudo make install
+    ```
+
+
+### Install Kata 3.0 Rust Runtime Shim
+
+```
+$ git clone https://github.com/kata-containers/kata-containers.git
+$ cd kata-containers/src/runtime-rs
+$ make && sudo make install
+```
+After running the command above, the default config file `configuration.toml` will be installed under `/usr/share/defaults/kata-containers/`,  the binary file `containerd-shim-kata-v2` will be installed under `/user/local/bin` .
+
+### Build Kata Containers Kernel
+Follow the [Kernel installation guide](/tools/packaging/kernel/README.md).
+
+### Build Kata Rootfs
+Follow the [Rootfs installation guide](../../tools/osbuilder/rootfs-builder/README.md).
+
+### Build Kata Image
+Follow the [Image installation guide](../../tools/osbuilder/image-builder/README.md).
+
+### Install Containerd
+
+Follow the [Containerd installation guide](container-manager/containerd/containerd-install.md).
+
+

docs/use-cases/NVIDIA-GPU-passthrough-and-Kata.md

@@ -2,20 +2,20 @@
 
 An NVIDIA GPU device can be passed to a Kata Containers container using GPU
 passthrough (NVIDIA GPU pass-through mode) as well as GPU mediated passthrough
-(NVIDIA vGPU mode).
+(NVIDIA `vGPU` mode).
 
 NVIDIA GPU pass-through mode, an entire physical GPU is directly assigned to one
 VM, bypassing the NVIDIA Virtual GPU Manager. In this mode of operation, the GPU
 is accessed exclusively by the NVIDIA driver running in the VM to which it is
 assigned. The GPU is not shared among VMs.
 
-NVIDIA Virtual GPU (vGPU) enables multiple virtual machines (VMs) to have
+NVIDIA Virtual GPU (`vGPU`) enables multiple virtual machines (VMs) to have
 simultaneous, direct access to a single physical GPU, using the same NVIDIA
 graphics drivers that are deployed on non-virtualized operating systems. By
-doing this, NVIDIA vGPU provides VMs with unparalleled graphics performance,
+doing this, NVIDIA `vGPU` provides VMs with unparalleled graphics performance,
 compute performance, and application compatibility, together with the
 cost-effectiveness and scalability brought about by sharing a GPU among multiple
-workloads. A vGPU can be either time-sliced or Multi-Instance GPU (MIG)-backed
+workloads. A `vGPU` can be either time-sliced or Multi-Instance GPU (MIG)-backed
 with [MIG-slices](https://docs.nvidia.com/datacenter/tesla/mig-user-guide/).
 
 | Technology | Description | Behavior | Detail |
@@ -46,14 +46,14 @@ $ lspci -s d0:00.0 -vv | grep Region
 For large BARs devices, MMIO mapping above 4G address space should be `enabled`
 in the PCI configuration of the BIOS.
 
-Some hardware vendors use different name in BIOS, such as:
+Some hardware vendors use a different name in BIOS, such as:
 
 - Above 4G Decoding
 - Memory Hole for PCI MMIO
 - Memory Mapped I/O above 4GB
 
 If one is using a GPU based on the Ampere architecture and later additionally
-SR-IOV needs to be enabled for the vGPU use-case.
+SR-IOV needs to be enabled for the `vGPU` use-case.
 
 The following steps outline the workflow for using an NVIDIA GPU with Kata.
 
@@ -154,7 +154,7 @@ $ ./build-kernel.sh -v 5.15.23 -g nvidia build
 $ sudo -E ./build-kernel.sh -v 5.15.23 -g nvidia install
 ```
 
-To build NVIDIA Driver in Kata container, `linux-headers` is required.
+To build NVIDIA Driver in Kata container, `linux-headers` are required.
 This is a way to generate deb packages for `linux-headers`:
 
 > **Note**:
@@ -177,7 +177,7 @@ kernel = "/usr/share/kata-containers/vmlinuz-nvidia-gpu.container"
 
 Use the following steps to pass an NVIDIA GPU device in pass-through mode with Kata:
 
-1. Find the Bus-Device-Function (BDF) for GPU device on host:
+1. Find the Bus-Device-Function (BDF) for the GPU device on the host:
 
    ```sh
    $ sudo lspci -nn -D | grep -i nvidia
@@ -219,7 +219,7 @@ Use the following steps to pass an NVIDIA GPU device in pass-through mode with K
    crw-rw-rw- 1 root    root     10, 196 Mar 18 02:27 vfio
    ```
 
-4. Start a Kata container with GPU device:
+4. Start a Kata container with the GPU device:
 
    ```sh
    # You may need to `modprobe vhost-vsock` if you get
@@ -246,9 +246,228 @@ Use the following steps to pass an NVIDIA GPU device in pass-through mode with K
 ## NVIDIA vGPU mode with Kata Containers
 
 NVIDIA vGPU is a licensed product on all supported GPU boards. A software license
-is required to enable all vGPU features within the guest VM.
+is required to enable all vGPU features within the guest VM. NVIDIA vGPU manager
+needs to be installed on the host to configure GPUs in vGPU mode. See [NVIDIA Virtual GPU Software Documentation v14.0 through 14.1](https://docs.nvidia.com/grid/14.0/) for more details.
 
-> **TODO**: Will follow up with instructions
+### NVIDIA vGPU time-sliced
+
+In the time-sliced mode, the GPU is not partitioned and the workload uses the
+whole GPU and shares access to the GPU engines. Processes are scheduled in
+series. The best effort scheduler is the default one and can be exchanged by
+other scheduling policies see the documentation above how to do that.
+
+Beware if you had `MIG` enabled before to disable `MIG` on the GPU if you want
+to use `time-sliced` `vGPU`.
+
+```sh
+$ sudo nvidia-smi -mig 0
+```
+
+Enable the virtual functions for the physical GPU in the `sysfs` file system.
+
+```sh
+$ sudo /usr/lib/nvidia/sriov-manage -e 0000:41:00.0
+```
+
+Get the `BDF` of the available virtual function on the GPU, and choose one for the
+following steps.
+
+```sh
+$ cd /sys/bus/pci/devices/0000:41:00.0/
+$ ls -l |  grep virtfn
+```
+
+#### List all available vGPU instances
+
+The following shell snippet will walk the `sysfs` and only print instances
+that are available, that can be created.
+
+```sh
+# The 00.0 is often the PF of the device the VFs will have the funciont in the
+# BDF incremented by some values so e.g. the very first VF is 0000:41:00.4
+
+cd /sys/bus/pci/devices/0000:41:00.0/
+
+for vf in $(ls -d virtfn*)
+do
+        BDF=$(basename $(readlink -f $vf))
+        for md in $(ls -d $vf/mdev_supported_types/*)
+        do
+                AVAIL=$(cat $md/available_instances)
+                NAME=$(cat $md/name)
+                DIR=$(basename $md)
+
+                if [ $AVAIL -gt 0 ]; then
+                        echo "| BDF          | INSTANCES | NAME           | DIR        |"
+                        echo "+--------------+-----------+----------------+------------+"
+                        printf "| %12s |%10d |%15s | %10s |\n\n" "$BDF" "$AVAIL" "$NAME" "$DIR"
+                fi
+
+        done
+done
+```
+
+If there are available instances you get something like this (for the first VF),
+beware that the output is highly dependent on the GPU you have, if there is no
+output check again if `MIG` is really disabled.
+
+```sh
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 |  GRID A100D-4C | nvidia-692 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 |  GRID A100D-8C | nvidia-693 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 | GRID A100D-10C | nvidia-694 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 | GRID A100D-16C | nvidia-695 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 | GRID A100D-20C | nvidia-696 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 | GRID A100D-40C | nvidia-697 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 | GRID A100D-80C | nvidia-698 |
+
+```
+
+Change to the `mdev_supported_types` directory for the virtual function on which
+you want to create the `vGPU`. Taking the first output as an example:
+
+```sh
+$ cd virtfn0/mdev_supported_types/nvidia-692
+$ UUIDGEN=$(uuidgen)
+$ sudo bash -c "echo $UUIDGEN > create"
+```
+
+Confirm that the `vGPU` was created. You should see the `UUID` pointing to a
+subdirectory of the `sysfs` space.
+
+```sh
+$ ls -l /sys/bus/mdev/devices/
+```
+
+Get the `IOMMU` group number and verify there is a `VFIO` device created to use
+with Kata.
+
+```sh
+$ ls -l /sys/bus/mdev/devices/*/
+$ ls -l /dev/vfio
+```
+
+Use the `VFIO` device created in the same way as in the pass-through use-case.
+Beware that the guest needs the NVIDIA guest drivers, so one would need to build
+a new guest `OS` image.
+
+### NVIDIA vGPU MIG-backed
+
+We're not going into detail what `MIG` is but briefly it is a technology to
+partition the hardware into independent instances with guaranteed quality of
+service. For more details see [NVIDIA Multi-Instance GPU User Guide](https://docs.nvidia.com/datacenter/tesla/mig-user-guide/).
+
+First enable `MIG` mode for a GPU, depending on the platform you're running
+a reboot would be necessary. Some platforms support GPU reset.
+
+```sh
+$ sudo nvidia-smi -mig 1
+```
+
+If the platform supports a GPU reset one can run, otherwise you will get a
+warning to reboot the server.
+
+```sh
+$ sudo nvidia-smi --gpu-reset
+```
+
+The driver per default provides a number of profiles that users can opt-in when
+configuring the MIG feature.
+
+```sh
+$ sudo nvidia-smi mig -lgip
++-----------------------------------------------------------------------------+
+| GPU instance profiles:                                                      |
+| GPU   Name             ID    Instances   Memory     P2P    SM    DEC   ENC  |
+|                              Free/Total   GiB              CE    JPEG  OFA  |
+|=============================================================================|
+|   0  MIG 1g.10gb       19     7/7        9.50       No     14     0     0   |
+|                                                             1     0     0   |
++-----------------------------------------------------------------------------+
+|   0  MIG 1g.10gb+me    20     1/1        9.50       No     14     1     0   |
+|                                                             1     1     1   |
++-----------------------------------------------------------------------------+
+|   0  MIG 2g.20gb       14     3/3        19.50      No     28     1     0   |
+|                                                             2     0     0   |
++-----------------------------------------------------------------------------+
+                              ...
+```
+
+Create the GPU instances that correspond to the `vGPU` types of the `MIG-backed`
+`vGPUs` that you will create [NVIDIA A100 PCIe 80GB Virtual GPU Types](https://docs.nvidia.com/grid/13.0/grid-vgpu-user-guide/index.html#vgpu-types-nvidia-a100-pcie-80gb).
+
+```sh
+# MIG 1g.10gb --> vGPU A100D-1-10C
+$ sudo nvidia-smi mig -cgi 19
+```
+
+List the GPU instances and get the GPU instance id to create the compute
+instance.
+
+```sh
+$ sudo nvidia-smi mig -lgi # list the created GPU instances
+$ sudo nvidia-smi mig -cci -gi 9 # each GPU instance can have several compute
+                                 # instances. Instance -> Workload
+```
+
+Verify that the compute instances were created within the GPU instance
+
+```sh
+$ nvidia-smi
+                              ... snip ...
++-----------------------------------------------------------------------------+
+| MIG devices:                                                                |
++------------------+----------------------+-----------+-----------------------+
+| GPU  GI  CI  MIG |         Memory-Usage |        Vol|         Shared        |
+|      ID  ID  Dev |           BAR1-Usage | SM     Unc| CE  ENC  DEC  OFA  JPG|
+|                  |                      |        ECC|                       |
+|==================+======================+===========+=======================|
+|  0    9   0   0  |      0MiB /  9728MiB | 14      0 |  1   0    0    0    0 |
+|                  |      0MiB /  4095MiB |           |                       |
++------------------+----------------------+-----------+-----------------------+
+                              ... snip ...
+```
+
+We can use the [snippet](#list-all-available-vgpu-instances) from before to list
+the available `vGPU` instances, this time `MIG-backed`.
+
+```sh
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.4 |         1 |GRID A100D-1-10C | nvidia-699 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:00.5 |         1 |GRID A100D-1-10C | nvidia-699 |
+
+| BDF          | INSTANCES | NAME           | DIR        |
++--------------+-----------+----------------+------------+
+| 0000:41:01.6 |         1 |GRID A100D-1-10C | nvidia-699 |
+                       ... snip ...
+```
+
+Repeat the steps after the [snippet](#list-all-available-vgpu-instances) listing
+to create the corresponding `mdev` device and use the guest `OS` created in the
+previous section with `time-sliced` `vGPUs`.
 
 ## Install NVIDIA Driver + Toolkit in Kata Containers Guest OS
 
@@ -263,7 +482,7 @@ export EXTRA_PKGS="gcc make curl gnupg"
 ```
 
 Having the `$ROOTFS_DIR` exported in the previous step we can now install all the
-need parts in the guest OS. In this case we have an Ubuntu based rootfs.
+needed parts in the guest OS. In this case, we have an Ubuntu based rootfs.
 
 First off all mount the special filesystems into the rootfs
 
@@ -281,9 +500,9 @@ Now we can enter `chroot`
 $ sudo chroot ${ROOTFS_DIR}
 ```
 
-Inside the rootfs one is going to install the drivers and toolkit to enable easy
-creation of GPU containers with Kata. We can also use this rootfs for any other
-container not specifically only for GPUs.
+Inside the rootfs one is going to install the drivers and toolkit to enable the
+easy creation of GPU containers with Kata. We can also use this rootfs for any
+other container not specifically only for GPUs.
 
 As a prerequisite install the copied kernel development packages
 
@@ -304,6 +523,7 @@ $ ./NVIDIA-Linux-x86_64-510.54.run -x
 $ cd NVIDIA-Linux-x86_64-510.54
 $ ./nvidia-installer -k 5.15.23-nvidia-gpu
 ```
+
 Having the drivers installed we need to install the toolkit which will take care
 of providing the right bits into the container.
 
@@ -325,7 +545,7 @@ Create the hook execution file for Kata:
 /usr/bin/nvidia-container-toolkit -debug $@
 ```
 
-As a last step one can do some cleanup of files or package caches. Build the
+As the last step one can do some cleanup of files or package caches. Build the
 rootfs and configure it for use with Kata according to the development guide.
 
 Enable the `guest_hook_path` in Kata's `configuration.toml`
@@ -334,7 +554,7 @@ Enable the `guest_hook_path` in Kata's `configuration.toml`
 guest_hook_path = "/usr/share/oci/hooks"
 ```
 
-One has build a NVIDIA rootfs, kernel and now we can run any GPU container
+One has built a NVIDIA rootfs, kernel and now we can run any GPU container
 without installing the drivers into the container. Check NVIDIA device status
 with `nvidia-smi`
 
@@ -362,7 +582,7 @@ Fri Mar 18 10:36:59 2022
 +-----------------------------------------------------------------------------+
 ```
 
-As a last step one can remove the additional packages and files that were added
+As the last step one can remove the additional packages and files that were added
 to the `$ROOTFS_DIR` to keep it as small as possible.
 
 ## References

docs/use-cases/using-Intel-QAT-and-kata.md

@@ -279,8 +279,8 @@ $ export KERNEL_EXTRAVERSION=$(awk '/^EXTRAVERSION =/{print $NF}' $GOPATH/$LINUX
 $ export KERNEL_ROOTFS_DIR=${KERNEL_MAJOR_VERSION}.${KERNEL_PATHLEVEL}.${KERNEL_SUBLEVEL}${KERNEL_EXTRAVERSION}
 $ cd $QAT_SRC
 $ KERNEL_SOURCE_ROOT=$GOPATH/$LINUX_VER ./configure --enable-icp-sriov=guest
-$ sudo -E make all -j$(nproc)
-$ sudo -E make INSTALL_MOD_PATH=$ROOTFS_DIR qat-driver-install -j$(nproc)
+$ sudo -E make all -j $($(nproc ${CI:+--ignore 1}))
+$ sudo -E make INSTALL_MOD_PATH=$ROOTFS_DIR qat-driver-install -j $($(nproc ${CI:+--ignore 1}))
 ```
 
 The `usdm_drv` module also needs to be copied into the rootfs modules path and
@@ -312,7 +312,7 @@ working properly with the Kata Containers VM.
 
 ### Build OpenSSL Intel® QAT engine container
 
-Use the OpenSSL Intel® QAT [Dockerfile](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/master/demo/openssl-qat-engine) 
+Use the OpenSSL Intel® QAT [Dockerfile](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/demo/openssl-qat-engine) 
 to build a container image with an optimized OpenSSL engine for 
 Intel® QAT. Using `docker build` with the Kata Containers runtime can sometimes
 have issues. Therefore, make sure that `runc` is the default Docker container 
@@ -444,7 +444,7 @@ $ sudo docker save -o openssl-qat-engine.tar openssl-qat-engine:latest
 $ sudo ctr -n=k8s.io images import openssl-qat-engine.tar
 ```
 
-The [Intel® QAT Plugin](https://github.com/intel/intel-device-plugins-for-kubernetes/blob/master/cmd/qat_plugin/README.md)
+The [Intel® QAT Plugin](https://github.com/intel/intel-device-plugins-for-kubernetes/blob/main/cmd/qat_plugin/README.md)
 needs to be started so that the virtual functions can be discovered and
 used by Kubernetes. 
 

docs/use-cases/using-Intel-SGX-and-kata.md

@@ -18,16 +18,13 @@ CONFIG_X86_SGX_KVM=y
 
 * Kubernetes cluster configured with:
    * [`kata-deploy`](../../tools/packaging/kata-deploy) based Kata Containers installation
-   * [Intel SGX Kubernetes device plugin](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/cmd/sgx_plugin#deploying-with-pre-built-images)
+   * [Intel SGX Kubernetes device plugin](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/cmd/sgx_plugin#deploying-with-pre-built-images) and associated components including [operator](https://github.com/intel/intel-device-plugins-for-kubernetes/blob/main/cmd/operator/README.md) and dependencies
 
 > Note: Kata Containers supports creating VM sandboxes with Intel® SGX enabled
 > using [cloud-hypervisor](https://github.com/cloud-hypervisor/cloud-hypervisor/) and [QEMU](https://www.qemu.org/) VMMs only.
 
 ### Kata Containers Configuration
 
-Before running a Kata Container make sure that your version of `crio` or `containerd`
-supports annotations.
-
 For `containerd` check in `/etc/containerd/config.toml` that the list of `pod_annotations` passed
 to the `sandbox` are: `["io.katacontainers.*", "sgx.intel.com/epc"]`.
 
@@ -99,4 +96,4 @@ because socket passthrough is not supported. An alternative is to deploy the `ae
 container.
 * Projects like [Gramine Shielded Containers (GSC)](https://gramine-gsc.readthedocs.io/en/latest/) are
 also known to work. For GSC specifically, the Kata guest kernel needs to have the `CONFIG_NUMA=y`
-enabled and at least one CPU online when running the GSC container.
+enabled and at least one CPU online when running the GSC container. The Kata Containers guest kernel currently has `CONFIG_NUMA=y` enabled by default.

snap/local/README.md

@@ -22,21 +22,35 @@ $ sudo snap install kata-containers --classic
 
 ## Build and install snap image
 
-Run next command at the root directory of the packaging repository.
+Run the command below which will use the packaging Makefile to build the snap image:
 
 ```sh
-$ make snap
+$ make -C tools/packaging snap
 ```
 
+> **Warning:**
+>
+> By default, `snapcraft` will create a clean virtual machine
+> environment to build the snap in using the `multipass` tool.
+>
+> However, `multipass` is silently disabled when `--destructive-mode` is
+> used.
+>
+> Since building the Kata Containers package currently requires
+> `--destructive-mode`, the snap will be built using the host
+> environment. To avoid parts of the build auto-detecting additional
+> features to enable (for example for QEMU), we recommend that you
+> only run the snap build in a minimal host environment.
+
 To install the resulting snap image, snap must be put in [classic mode][3] and the
-security confinement must be disabled (*--classic*). Also since the resulting snap
-has not been signed the verification of signature must be omitted (*--dangerous*).
+security confinement must be disabled (`--classic`). Also since the resulting snap
+has not been signed the verification of signature must be omitted (`--dangerous`).
 
 ```sh
-$ sudo snap install --classic --dangerous kata-containers_[VERSION]_[ARCH].snap
+$ sudo snap install --classic --dangerous "kata-containers_${version}_${arch}.snap"
 ```
 
-Replace `VERSION` with the current version of Kata Containers and `ARCH` with
+Replace `${version}` with the current version of Kata Containers and `${arch}` with
 the system architecture.
 
 ## Configure Kata Containers
@@ -76,12 +90,12 @@ then a new configuration file can be [created](#configure-kata-containers)
 and [configured][7].
 
 [1]: https://docs.snapcraft.io/snaps/intro
-[2]: ../docs/design/architecture/README.md#root-filesystem-image
+[2]: ../../docs/design/architecture/README.md#root-filesystem-image
 [3]: https://docs.snapcraft.io/reference/confinement#classic
-[4]: https://github.com/kata-containers/runtime#configuration
+[4]: https://github.com/kata-containers/kata-containers/tree/main/src/runtime#configuration
 [5]: https://docs.docker.com/engine/reference/commandline/dockerd
-[6]: ../docs/install/docker/ubuntu-docker-install.md
-[7]: ../docs/Developer-Guide.md#configure-to-use-initrd-or-rootfs-image
+[6]: ../../docs/install/docker/ubuntu-docker-install.md
+[7]: ../../docs/Developer-Guide.md#configure-to-use-initrd-or-rootfs-image
 [8]: https://snapcraft.io/kata-containers
-[9]: ../docs/Developer-Guide.md#run-kata-containers-with-docker
-[10]: ../docs/Developer-Guide.md#run-kata-containers-with-kubernetes
+[9]: ../../docs/Developer-Guide.md#run-kata-containers-with-docker
+[10]: ../../docs/Developer-Guide.md#run-kata-containers-with-kubernetes

snap/local/snap-common.sh

@@ -0,0 +1,114 @@
+#!/usr/bin/env bash
+#
+# Copyright (c) 2022 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Description: Idempotent script to be sourced by all parts in a
+#   snapcraft config file.
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+# XXX: Bash-specific code. zsh doesn't support this option and that *does*
+# matter if this script is run sourced... since it'll be using zsh! ;)
+[ -n "$BASH_VERSION" ] && set -o errtrace
+
+[ -n "${DEBUG:-}" ] && set -o xtrace
+
+die()
+{
+	echo >&2 "ERROR: $0: $*"
+}
+
+[ -n "${SNAPCRAFT_STAGE:-}" ] ||\
+	die "must be sourced from a snapcraft config file"
+
+snap_yq_version=3.4.1
+
+snap_common_install_yq()
+{
+      export yq="${SNAPCRAFT_STAGE}/bin/yq"
+
+      local yq_pkg
+      yq_pkg="github.com/mikefarah/yq"
+
+      local yq_url
+      yq_url="https://${yq_pkg}/releases/download/${snap_yq_version}/yq_${goos}_${goarch}"
+      curl -o "${yq}" -L "${yq_url}"
+      chmod +x "${yq}"
+}
+
+# Function that should be called for each snap "part" in
+# snapcraft.yaml.
+snap_common_main()
+{
+	# Architecture
+	arch="$(uname -m)"
+
+	case "${arch}" in
+		aarch64)
+			goarch="arm64"
+			qemu_arch="${arch}"
+			;;
+
+		ppc64le)
+			goarch="ppc64le"
+			qemu_arch="ppc64"
+			;;
+
+		s390x)
+			goarch="${arch}"
+			qemu_arch="${arch}"
+			;;
+
+		x86_64)
+			goarch="amd64"
+			qemu_arch="${arch}"
+			;;
+
+		*) die "unsupported architecture: ${arch}" ;;
+	esac
+
+	dpkg_arch=$(dpkg --print-architecture)
+
+	# golang
+	#
+	# We need the O/S name in golang format, but since we don't
+	# know if the godeps part has run, we don't know if golang is
+	# available yet, hence fall back to a standard system command.
+	goos="$(go env GOOS &>/dev/null || true)"
+	[ -z "$goos" ] && goos=$(uname -s|tr '[A-Z]' '[a-z]')
+
+	export GOROOT="${SNAPCRAFT_STAGE}"
+	export GOPATH="${GOROOT}/gopath"
+	export GO111MODULE="auto"
+
+	mkdir -p "${GOPATH}/bin"
+	export PATH="${GOPATH}/bin:${PATH}"
+
+	# Proxy
+	export http_proxy="${http_proxy:-}"
+	export https_proxy="${https_proxy:-}"
+
+	# Binaries
+	mkdir -p "${SNAPCRAFT_STAGE}/bin"
+
+	export PATH="$PATH:${SNAPCRAFT_STAGE}/bin"
+
+	# YAML query tool
+	export yq="${SNAPCRAFT_STAGE}/bin/yq"
+
+	# Kata paths
+	export kata_dir=$(printf "%s/src/github.com/%s/%s" \
+		"${GOPATH}" \
+		"${SNAPCRAFT_PROJECT_NAME}" \
+		"${SNAPCRAFT_PROJECT_NAME}")
+
+	export versions_file="${kata_dir}/versions.yaml"
+
+	[ -n "${yq:-}" ] && [ -x "${yq:-}" ] || snap_common_install_yq
+}
+
+snap_common_main

snap/snapcraft.yaml

@@ -1,4 +1,5 @@
 name: kata-containers
+website: https://github.com/kata-containers/kata-containers
 summary: Build lightweight VMs that seamlessly plug into the containers ecosystem
 description: |
   Kata Containers is an open source project and community working to build a
@@ -18,20 +19,18 @@ parts:
       - git
       - git-extras
     override-pull: |
-      version="9999"
-      kata_url="https://github.com/kata-containers/kata-containers"
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
 
-      if echo "${GITHUB_REF}" | grep -q -E "^refs/tags"; then
-        version=$(echo ${GITHUB_REF} | cut -d/ -f3)
+      version="9999"
+
+      if echo "${GITHUB_REF:-}" | grep -q -E "^refs/tags"; then
+        version=$(echo ${GITHUB_REF:-} | cut -d/ -f3)
         git checkout ${version}
       fi
 
       snapcraftctl set-grade "stable"
       snapcraftctl set-version "${version}"
 
-      # setup GOPATH - this repo dir should be there
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
       mkdir -p $(dirname ${kata_dir})
       ln -sf $(realpath "${SNAPCRAFT_STAGE}/..") ${kata_dir}
 
@@ -43,31 +42,46 @@ parts:
     build-packages:
       - curl
     override-build: |
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
+
       # put everything in stage
-      cd ${SNAPCRAFT_STAGE}
+      cd "${SNAPCRAFT_STAGE}"
 
-      yq_path="./yq"
-      yq_pkg="github.com/mikefarah/yq"
-      goos="linux"
-      case "$(uname -m)" in
-        aarch64) goarch="arm64";;
-        ppc64le) goarch="ppc64le";;
-        x86_64) goarch="amd64";;
-        s390x) goarch="s390x";;
-        *) echo "unsupported architecture: $(uname -m)"; exit 1;;
-      esac
-
-      yq_version=3.4.1
-      yq_url="https://${yq_pkg}/releases/download/${yq_version}/yq_${goos}_${goarch}"
-      curl -o "${yq_path}" -L "${yq_url}"
-      chmod +x "${yq_path}"
-
-      kata_dir=gopath/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
-      version="$(${yq_path} r ${kata_dir}/versions.yaml languages.golang.meta.newest-version)"
+      version="$(${yq} r ${kata_dir}/versions.yaml languages.golang.meta.newest-version)"
       tarfile="go${version}.${goos}-${goarch}.tar.gz"
       curl -LO https://golang.org/dl/${tarfile}
       tar -xf ${tarfile} --strip-components=1
 
+  rustdeps:
+    after: [metadata]
+    plugin: nil
+    prime:
+      - -*
+    build-packages:
+      - curl
+    override-build: |
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
+
+      # put everything in stage
+      cd "${SNAPCRAFT_STAGE}"
+
+      version="$(${yq} r ${kata_dir}/versions.yaml languages.rust.meta.newest-version)"
+      if ! command -v rustup > /dev/null; then
+        curl https://sh.rustup.rs -sSf | sh -s -- -y --default-toolchain ${version}
+      fi
+
+      export PATH=${PATH}:${HOME}/.cargo/bin
+      rustup toolchain install ${version}
+      rustup default ${version}
+      if [ "${arch}" == "ppc64le" ] || [ "${arch}" == "s390x" ] ; then
+        [ "${arch}" == "ppc64le" ] && arch="powerpc64le"
+        rustup target add ${arch}-unknown-linux-gnu
+      else
+        rustup target add ${arch}-unknown-linux-musl
+        $([ "$(whoami)" != "root" ] && echo sudo) ln -sf /usr/bin/g++ /bin/musl-g++
+      fi
+      rustup component add rustfmt
+
   image:
     after: [godeps, qemu, kernel]
     plugin: nil
@@ -80,28 +94,17 @@ parts:
       - uidmap
       - gnupg2
     override-build: |
-      [ "$(uname -m)" = "ppc64le" ] || [ "$(uname -m)" = "s390x" ] && sudo apt-get --no-install-recommends install -y protobuf-compiler
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
 
-      yq=${SNAPCRAFT_STAGE}/yq
+      [ "${arch}" = "ppc64le" ] || [ "${arch}" = "s390x" ] && sudo apt-get --no-install-recommends install -y protobuf-compiler
 
-      # set GOPATH
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
-
-      export GOROOT=${SNAPCRAFT_STAGE}
-      export PATH="${GOROOT}/bin:${PATH}"
-      export GO111MODULE="auto"
-
-      http_proxy=${http_proxy:-""}
-      https_proxy=${https_proxy:-""}
       if [ -n "$http_proxy" ]; then
         echo "Setting proxy $http_proxy"
-        sudo -E systemctl set-environment http_proxy=$http_proxy || true
-        sudo -E systemctl set-environment https_proxy=$https_proxy || true
+        sudo -E systemctl set-environment http_proxy="$http_proxy" || true
+        sudo -E systemctl set-environment https_proxy="$https_proxy" || true
       fi
 
       # Copy yq binary. It's used in the container
-      mkdir -p "${GOPATH}/bin/"
       cp -a "${yq}" "${GOPATH}/bin/"
 
       echo "Unmasking docker service"
@@ -112,63 +115,54 @@ parts:
       echo "Starting docker"
       sudo -E systemctl start docker || true
 
-      cd ${kata_dir}/tools/osbuilder
+      cd "${kata_dir}/tools/osbuilder"
 
       # build image
       export AGENT_INIT=yes
       export USE_DOCKER=1
       export DEBUG=1
-      arch="$(uname -m)"
       initrd_distro=$(${yq} r -X ${kata_dir}/versions.yaml assets.initrd.architecture.${arch}.name)
       image_distro=$(${yq} r -X ${kata_dir}/versions.yaml assets.image.architecture.${arch}.name)
       case "$arch" in
         x86_64)
           # In some build systems it's impossible to build a rootfs image, try with the initrd image
-          sudo -E PATH=$PATH make image DISTRO=${image_distro} || sudo -E PATH=$PATH make initrd DISTRO=${initrd_distro}
+          sudo -E PATH=$PATH make image DISTRO="${image_distro}" || sudo -E PATH="$PATH" make initrd DISTRO="${initrd_distro}"
         ;;
 
         aarch64|ppc64le|s390x)
-          sudo -E PATH=$PATH make initrd DISTRO=${initrd_distro}
+          sudo -E PATH="$PATH" make initrd DISTRO="${initrd_distro}"
         ;;
 
-        *) echo "unsupported architecture: $(uname -m)"; exit 1;;
+        *) die "unsupported architecture: ${arch}" ;;
       esac
 
       # Install image
-      kata_image_dir=${SNAPCRAFT_PART_INSTALL}/usr/share/kata-containers
-      mkdir -p ${kata_image_dir}
-      cp kata-containers*.img ${kata_image_dir}
+      kata_image_dir="${SNAPCRAFT_PART_INSTALL}/usr/share/kata-containers"
+      mkdir -p "${kata_image_dir}"
+      cp kata-containers*.img "${kata_image_dir}"
 
   runtime:
     after: [godeps, image, cloud-hypervisor]
     plugin: nil
     build-attributes: [no-patchelf]
     override-build: |
-      # set GOPATH
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      export GOROOT=${SNAPCRAFT_STAGE}
-      export PATH="${GOROOT}/bin:${PATH}"
-      export GO111MODULE="auto"
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
 
-      cd ${kata_dir}/src/runtime
+      cd "${kata_dir}/src/runtime"
 
-      # setup arch
-      arch=$(uname -m)
-      if [ ${arch} = "ppc64le" ]; then
-        arch="ppc64"
-      fi
+      qemu_cmd="qemu-system-${qemu_arch}"
 
       # build and install runtime
       make \
-        PREFIX=/snap/${SNAPCRAFT_PROJECT_NAME}/current/usr \
+        PREFIX="/snap/${SNAPCRAFT_PROJECT_NAME}/current/usr" \
         SKIP_GO_VERSION_CHECK=1 \
-        QEMUCMD=qemu-system-$arch
+        QEMUCMD="${qemu_cmd}"
+
       make install \
         PREFIX=/usr \
-        DESTDIR=${SNAPCRAFT_PART_INSTALL} \
+        DESTDIR="${SNAPCRAFT_PART_INSTALL}" \
         SKIP_GO_VERSION_CHECK=1 \
-        QEMUCMD=qemu-system-$arch
+        QEMUCMD="${qemu_cmd}"
 
       if [ ! -f ${SNAPCRAFT_PART_INSTALL}/../../image/install/usr/share/kata-containers/kata-containers.img ]; then
         sed -i -e "s|^image =.*|initrd = \"/snap/${SNAPCRAFT_PROJECT_NAME}/current/usr/share/kata-containers/kata-containers-initrd.img\"|" \
@@ -185,44 +179,37 @@ parts:
       - bison
       - flex
     override-build: |
-      yq=${SNAPCRAFT_STAGE}/yq
-      export PATH="${PATH}:${SNAPCRAFT_STAGE}"
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
-      versions_file="${kata_dir}/versions.yaml"
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
+
       kernel_version="$(${yq} r $versions_file assets.kernel.version)"
       #Remove extra 'v'
-      kernel_version=${kernel_version#v}
+      kernel_version="${kernel_version#v}"
 
-      [ "$(uname -m)" = "s390x" ] && sudo apt-get --no-install-recommends install -y libssl-dev
+      [ "${arch}" = "s390x" ] && sudo apt-get --no-install-recommends install -y libssl-dev
 
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      export GO111MODULE="auto"
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
-
-      cd ${kata_dir}/tools/packaging/kernel
+      cd "${kata_dir}/tools/packaging/kernel"
       kernel_dir_prefix="kata-linux-"
 
       # Setup and build kernel
-      ./build-kernel.sh -v ${kernel_version} -d setup
+      ./build-kernel.sh -v "${kernel_version}" -d setup
       cd ${kernel_dir_prefix}*
-      make -j $(($(nproc)-1)) EXTRAVERSION=".container"
+      make -j $(nproc ${CI:+--ignore 1}) EXTRAVERSION=".container"
 
-      kernel_suffix=${kernel_version}.container
-      kata_kernel_dir=${SNAPCRAFT_PART_INSTALL}/usr/share/kata-containers
-      mkdir -p ${kata_kernel_dir}
+      kernel_suffix="${kernel_version}.container"
+      kata_kernel_dir="${SNAPCRAFT_PART_INSTALL}/usr/share/kata-containers"
+      mkdir -p "${kata_kernel_dir}"
 
       # Install bz kernel
-      make install INSTALL_PATH=${kata_kernel_dir} EXTRAVERSION=".container" || true
-      vmlinuz_name=vmlinuz-${kernel_suffix}
-      ln -sf ${vmlinuz_name} ${kata_kernel_dir}/vmlinuz.container
+      make install INSTALL_PATH="${kata_kernel_dir}" EXTRAVERSION=".container" || true
+      vmlinuz_name="vmlinuz-${kernel_suffix}"
+      ln -sf "${vmlinuz_name}" "${kata_kernel_dir}/vmlinuz.container"
 
       # Install raw kernel
-      vmlinux_path=vmlinux
-      [ "$(uname -m)" = "s390x" ] && vmlinux_path=arch/s390/boot/compressed/vmlinux
-      vmlinux_name=vmlinux-${kernel_suffix}
-      cp ${vmlinux_path} ${kata_kernel_dir}/${vmlinux_name}
-      ln -sf ${vmlinux_name} ${kata_kernel_dir}/vmlinux.container
+      vmlinux_path="vmlinux"
+      [ "${arch}" = "s390x" ] && vmlinux_path="arch/s390/boot/compressed/vmlinux"
+      vmlinux_name="vmlinux-${kernel_suffix}"
+      cp "${vmlinux_path}" "${kata_kernel_dir}/${vmlinux_name}"
+      ln -sf "${vmlinux_name}" "${kata_kernel_dir}/vmlinux.container"
 
   qemu:
     plugin: make
@@ -249,12 +236,8 @@ parts:
       - libselinux1-dev
       - ninja-build
     override-build: |
-      yq=${SNAPCRAFT_STAGE}/yq
-      export GOPATH=${SNAPCRAFT_STAGE}/gopath
-      export GO111MODULE="auto"
-      kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
 
-      versions_file="${kata_dir}/versions.yaml"
       branch="$(${yq} r ${versions_file} assets.hypervisor.qemu.version)"
       url="$(${yq} r ${versions_file} assets.hypervisor.qemu.url)"
       commit=""
@@ -262,11 +245,11 @@ parts:
       patches_version_dir="${kata_dir}/tools/packaging/qemu/patches/tag_patches/${branch}"
 
       # download source
-      qemu_dir=${SNAPCRAFT_STAGE}/qemu
+      qemu_dir="${SNAPCRAFT_STAGE}/qemu"
       rm -rf "${qemu_dir}"
       git clone --depth 1 --branch ${branch} --single-branch ${url} "${qemu_dir}"
-      cd ${qemu_dir}
-      [ -z "${commit}" ] || git checkout ${commit}
+      cd "${qemu_dir}"
+      [ -z "${commit}" ] || git checkout "${commit}"
 
       [ -n "$(ls -A ui/keycodemapdb)" ] || git clone --depth 1 https://github.com/qemu/keycodemapdb ui/keycodemapdb/
       [ -n "$(ls -A capstone)" ] || git clone --depth 1 https://github.com/qemu/capstone capstone
@@ -277,10 +260,10 @@ parts:
       ${kata_dir}/tools/packaging/scripts/apply_patches.sh "${patches_version_dir}"
 
       # Only x86_64 supports libpmem
-      [ "$(uname -m)" = "x86_64" ] && sudo apt-get --no-install-recommends install -y apt-utils ca-certificates libpmem-dev
+      [ "${arch}" = "x86_64" ] && sudo apt-get --no-install-recommends install -y apt-utils ca-certificates libpmem-dev
 
-      configure_hypervisor=${kata_dir}/tools/packaging/scripts/configure-hypervisor.sh
-      chmod +x ${configure_hypervisor}
+      configure_hypervisor="${kata_dir}/tools/packaging/scripts/configure-hypervisor.sh"
+      chmod +x "${configure_hypervisor}"
       # static build. The --prefix, --libdir, --libexecdir, --datadir arguments are
       # based on PREFIX and set by configure-hypervisor.sh
       echo "$(PREFIX=/snap/${SNAPCRAFT_PROJECT_NAME}/current/usr ${configure_hypervisor} -s kata-qemu) \
@@ -290,17 +273,17 @@ parts:
       # Copy QEMU configurations (Kconfigs)
       case "${branch}" in
       "v5.1.0")
-        cp -a ${kata_dir}/tools/packaging/qemu/default-configs/* default-configs
+        cp -a "${kata_dir}"/tools/packaging/qemu/default-configs/* default-configs
         ;;
 
       *)
-        cp -a ${kata_dir}/tools/packaging/qemu/default-configs/* configs/devices/
+        cp -a "${kata_dir}"/tools/packaging/qemu/default-configs/* configs/devices/
         ;;
       esac
 
       # build and install
-      make -j $(($(nproc)-1))
-      make install DESTDIR=${SNAPCRAFT_PART_INSTALL}
+      make -j $(nproc ${CI:+--ignore 1})
+      make install DESTDIR="${SNAPCRAFT_PART_INSTALL}"
     prime:
       - -snap/
       - -usr/bin/qemu-ga
@@ -316,26 +299,67 @@ parts:
       # Hack: move qemu to /
       "snap/kata-containers/current/": "./"
 
+  virtiofsd:
+    plugin: nil
+    after: [godeps, rustdeps]
+    override-build: |
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
+
+      # Currently, powerpc makes use of the QEMU's C implementation.
+      # The other platforms make use of the new rust virtiofsd.
+      #
+      # See "tools/packaging/scripts/configure-hypervisor.sh".
+      if [ "${arch}" == "ppc64le" ]
+      then
+          echo "INFO: Building QEMU's C version of virtiofsd"
+          # Handled by the 'qemu' part, so nothing more to do here.
+          exit 0
+      else
+          echo "INFO: Building rust version of virtiofsd"
+      fi
+
+      cd "${kata_dir}"
+
+      export PATH=${PATH}:${HOME}/.cargo/bin
+      # Download the rust implementation of virtiofsd
+      tools/packaging/static-build/virtiofsd/build-static-virtiofsd.sh
+      sudo install \
+        --owner='root' \
+        --group='root' \
+        --mode=0755 \
+        -D \
+        --target-directory="${SNAPCRAFT_PART_INSTALL}/usr/libexec/" \
+        virtiofsd/virtiofsd
+
   cloud-hypervisor:
     plugin: nil
     after: [godeps]
     override-build: |
-      arch=$(uname -m)
-      if [ "{$arch}" == "aarch64" ] || [ "${arch}" == "x64_64" ]; then 
+      source "${SNAPCRAFT_PROJECT_DIR}/snap/local/snap-common.sh"
+
+      if [ "${arch}" == "aarch64" ] || [ "${arch}" == "x86_64" ]; then
           sudo apt-get -y update
           sudo apt-get -y install ca-certificates curl gnupg lsb-release
-          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --batch --yes --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
-          echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
+          curl -fsSL https://download.docker.com/linux/ubuntu/gpg |\
+              sudo gpg --batch --yes --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
+          distro_codename=$(lsb_release -cs)
+          echo "deb [arch=${dpkg_arch} signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu ${distro_codename} stable" |\
+              sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
           sudo apt-get -y update
           sudo apt-get -y install docker-ce docker-ce-cli containerd.io
           sudo systemctl start docker.socket
 
-          export GOPATH=${SNAPCRAFT_STAGE}/gopath
-          kata_dir=${GOPATH}/src/github.com/${SNAPCRAFT_PROJECT_NAME}/${SNAPCRAFT_PROJECT_NAME}
-          cd ${kata_dir}
+          cd "${SNAPCRAFT_PROJECT_DIR}"
           sudo -E NO_TTY=true make cloud-hypervisor-tarball
-          tar xvJpf build/kata-static-cloud-hypervisor.tar.xz -C /tmp/
-          install -D /tmp/opt/kata/bin/cloud-hypervisor ${SNAPCRAFT_PART_INSTALL}/usr/bin/cloud-hypervisor
+
+          tarfile="${SNAPCRAFT_PROJECT_DIR}/tools/packaging/kata-deploy/local-build/build/kata-static-cloud-hypervisor.tar.xz"
+          tmpdir=$(mktemp -d)
+
+          tar -xvJpf "${tarfile}" -C "${tmpdir}"
+
+          install -D "${tmpdir}/opt/kata/bin/cloud-hypervisor" "${SNAPCRAFT_PART_INSTALL}/usr/bin/cloud-hypervisor"
+
+          rm -rf "${tmpdir}"
       fi
 
 apps:

src/agent/Cargo.toml

@@ -7,19 +7,20 @@ edition = "2018"
 [dependencies]
 oci = { path = "../libs/oci" }
 rustjail = { path = "rustjail" }
-protocols = { path = "../libs/protocols" }
+protocols = { path = "../libs/protocols", features = ["async"] }
 lazy_static = "1.3.0"
-ttrpc = { version = "0.5.0", features = ["async", "protobuf-codec"], default-features = false }
-protobuf = "=2.14.0"
+ttrpc = { version = "0.6.0", features = ["async"], default-features = false }
+protobuf = "2.27.0"
 libc = "0.2.58"
-nix = "0.23.0"
+nix = "0.24.1"
 capctl = "0.2.0"
 serde_json = "1.0.39"
 scan_fmt = "0.2.3"
 scopeguard = "1.0.0"
 thiserror = "1.0.26"
-regex = "1.5.4"
+regex = "1.5.5"
 serial_test = "0.5.1"
+kata-sys-util = { path = "../libs/kata-sys-util" }
 sysinfo = "0.23.0"
 
 # Async helpers
@@ -32,7 +33,7 @@ tokio = { version = "1.14.0", features = ["full"] }
 tokio-vsock = "0.3.1"
 
 netlink-sys = { version = "0.7.0", features = ["tokio_socket",]}
-rtnetlink = "0.8.0"
+rtnetlink = "0.11.0"
 netlink-packet-utils = "0.4.1"
 ipnetwork = "0.17.0"
 
@@ -81,8 +82,3 @@ standard-oci-runtime = ["rustjail/standard-oci-runtime"]
 [[bin]]
 name = "kata-agent"
 path = "src/main.rs"
-
-[[bin]]
-name = "oci-kata-agent"
-path = "src/main.rs"
-required-features = ["standard-oci-runtime"]

src/agent/Makefile

@@ -107,20 +107,17 @@ endef
 ##TARGET default: build code
 default: $(TARGET) show-header
 
-$(TARGET): $(GENERATED_CODE) logging-crate-tests $(TARGET_PATH)
-
-logging-crate-tests:
-	make -C $(CWD)/../libs/logging
+$(TARGET): $(GENERATED_CODE) $(TARGET_PATH)
 
 $(TARGET_PATH): show-summary
-	@RUSTFLAGS="$(EXTRA_RUSTFLAGS) --deny warnings" cargo build --target $(TRIPLE) --$(BUILD_TYPE) $(EXTRA_RUSTFEATURES)
+	@RUSTFLAGS="$(EXTRA_RUSTFLAGS) --deny warnings" cargo build --target $(TRIPLE) $(if $(findstring release,$(BUILD_TYPE)),--release) $(EXTRA_RUSTFEATURES)
 
 $(GENERATED_FILES): %: %.in
 	@sed $(foreach r,$(GENERATED_REPLACEMENTS),-e 's|@$r@|$($r)|g') "$<" > "$@"
 
 ##TARGET optimize: optimized  build
 optimize: show-summary show-header
-	@RUSTFLAGS="-C link-arg=-s $(EXTRA_RUSTFLAGS) --deny warnings" cargo build --target $(TRIPLE) --$(BUILD_TYPE) $(EXTRA_RUSTFEATURES)
+	@RUSTFLAGS="-C link-arg=-s $(EXTRA_RUSTFLAGS) --deny warnings" cargo build --target $(TRIPLE) $(if $(findstring release,$(BUILD_TYPE)),--release) $(EXTRA_RUSTFEATURES)
 
 ##TARGET install: install agent
 install: install-services
@@ -203,7 +200,6 @@ codecov-html: check_tarpaulin
 
 .PHONY: \
 	help \
-	logging-crate-tests \
 	optimize \
 	show-header \
 	show-summary \

src/agent/rustjail/Cargo.toml

@@ -16,18 +16,18 @@ scopeguard = "1.0.0"
 capctl = "0.2.0"
 lazy_static = "1.3.0"
 libc = "0.2.58"
-protobuf = "=2.14.0"
+protobuf = "2.27.0"
 slog = "2.5.2"
 slog-scope = "4.1.2"
 scan_fmt = "0.2.6"
-regex = "1.5.4"
+regex = "1.5.5"
 path-absolutize = "1.2.0"
 anyhow = "1.0.32"
 cgroups = { package = "cgroups-rs", version = "0.2.8" }
 rlimit = "0.5.3"
 cfg-if = "0.1.0"
 
-tokio = { version = "1.2.0", features = ["sync", "io-util", "process", "time", "macros"] }
+tokio = { version = "1.2.0", features = ["sync", "io-util", "process", "time", "macros", "rt"] }
 futures = "0.3.17"
 async-trait = "0.1.31"
 inotify = "0.9.2"

src/agent/rustjail/src/cgroups/fs/mod.rs

@@ -911,9 +911,8 @@ pub fn get_paths() -> Result<HashMap<String, String>> {
     Ok(m)
 }
 
-pub fn get_mounts() -> Result<HashMap<String, String>> {
+pub fn get_mounts(paths: &HashMap<String, String>) -> Result<HashMap<String, String>> {
     let mut m = HashMap::new();
-    let paths = get_paths()?;
 
     for l in fs::read_to_string(MOUNTS)?.lines() {
         let p: Vec<&str> = l.splitn(2, " - ").collect();
@@ -951,7 +950,7 @@ impl Manager {
         let mut m = HashMap::new();
 
         let paths = get_paths()?;
-        let mounts = get_mounts()?;
+        let mounts = get_mounts(&paths)?;
 
         for key in paths.keys() {
             let mnt = mounts.get(key);

src/agent/rustjail/src/console.rs

@@ -58,10 +58,7 @@ pub fn setup_master_console(socket_fd: RawFd) -> Result<()> {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::skip_if_not_root;
-    use std::fs::File;
     use std::os::unix::net::UnixListener;
-    use std::path::PathBuf;
     use tempfile::{self, tempdir};
 
     const CONSOLE_SOCKET: &str = "console-socket";

src/agent/rustjail/src/container.rs

@@ -42,7 +42,7 @@ use nix::pty;
 use nix::sched::{self, CloneFlags};
 use nix::sys::signal::{self, Signal};
 use nix::sys::stat::{self, Mode};
-use nix::unistd::{self, fork, ForkResult, Gid, Pid, Uid};
+use nix::unistd::{self, fork, ForkResult, Gid, Pid, Uid, User};
 use std::os::unix::fs::MetadataExt;
 use std::os::unix::io::AsRawFd;
 
@@ -64,8 +64,6 @@ use rlimit::{setrlimit, Resource, Rlim};
 use tokio::io::AsyncBufReadExt;
 use tokio::sync::Mutex;
 
-use crate::utils;
-
 pub const EXEC_FIFO_FILENAME: &str = "exec.fifo";
 
 const INIT: &str = "INIT";
@@ -78,9 +76,6 @@ const HOME_ENV_KEY: &str = "HOME";
 const PIDNS_FD: &str = "PIDNS_FD";
 const CONSOLE_SOCKET_FD: &str = "CONSOLE_SOCKET_FD";
 
-#[cfg(feature = "standard-oci-runtime")]
-const OCI_AGENT_BINARY: &str = "oci-kata-agent";
-
 #[derive(Debug)]
 pub struct ContainerStatus {
     pre_status: ContainerState,
@@ -227,7 +222,7 @@ pub trait BaseContainer {
     async fn start(&mut self, p: Process) -> Result<()>;
     async fn run(&mut self, p: Process) -> Result<()>;
     async fn destroy(&mut self) -> Result<()>;
-    fn exec(&mut self) -> Result<()>;
+    async fn exec(&mut self) -> Result<()>;
 }
 
 // LinuxContainer protected by Mutex
@@ -592,14 +587,20 @@ fn do_init_child(cwfd: RawFd) -> Result<()> {
 
     // only change stdio devices owner when user
     // isn't root.
-    if guser.uid != 0 {
-        set_stdio_permissions(guser.uid)?;
+    if !uid.is_root() {
+        set_stdio_permissions(uid)?;
     }
 
     setid(uid, gid)?;
 
     if !guser.additional_gids.is_empty() {
-        setgroups(guser.additional_gids.as_slice()).map_err(|e| {
+        let gids: Vec<Gid> = guser
+            .additional_gids
+            .iter()
+            .map(|gid| Gid::from_raw(*gid))
+            .collect();
+
+        unistd::setgroups(&gids).map_err(|e| {
             let _ = write_sync(
                 cwfd,
                 SYNC_FAILED,
@@ -639,11 +640,6 @@ fn do_init_child(cwfd: RawFd) -> Result<()> {
         capabilities::drop_privileges(cfd_log, c)?;
     }
 
-    if init {
-        // notify parent to run poststart hooks
-        write_sync(cwfd, SYNC_SUCCESS, "")?;
-    }
-
     let args = oci_process.args.to_vec();
     let env = oci_process.env.to_vec();
 
@@ -665,12 +661,17 @@ fn do_init_child(cwfd: RawFd) -> Result<()> {
         }
     }
 
-    // set the "HOME" env getting from "/etc/passwd", if
-    // there's no uid entry in /etc/passwd, set "/" as the
-    // home env.
     if env::var_os(HOME_ENV_KEY).is_none() {
-        let home_dir = utils::home_dir(guser.uid).unwrap_or_else(|_| String::from("/"));
-        env::set_var(HOME_ENV_KEY, home_dir);
+        // try to set "HOME" env by uid
+        if let Ok(Some(user)) = User::from_uid(Uid::from_raw(guser.uid)) {
+            if let Ok(user_home_dir) = user.dir.into_os_string().into_string() {
+                env::set_var(HOME_ENV_KEY, user_home_dir);
+            }
+        }
+        // set default home dir as "/" if "HOME" env is still empty
+        if env::var_os(HOME_ENV_KEY).is_none() {
+            env::set_var(HOME_ENV_KEY, String::from("/"));
+        }
     }
 
     let exec_file = Path::new(&args[0]);
@@ -730,7 +731,7 @@ fn do_init_child(cwfd: RawFd) -> Result<()> {
 // within the container to the specified user.
 // The ownership needs to match because it is created outside of
 // the container and needs to be localized.
-fn set_stdio_permissions(uid: libc::uid_t) -> Result<()> {
+fn set_stdio_permissions(uid: Uid) -> Result<()> {
     let meta = fs::metadata("/dev/null")?;
     let fds = [
         std::io::stdin().as_raw_fd(),
@@ -745,19 +746,13 @@ fn set_stdio_permissions(uid: libc::uid_t) -> Result<()> {
             continue;
         }
 
-        // According to the POSIX specification, -1 is used to indicate that owner and group
-        // are not to be changed.  Since uid_t and gid_t are unsigned types, we have to wrap
-        // around to get -1.
-        let gid = 0u32.wrapping_sub(1);
-
         // We only change the uid owner (as it is possible for the mount to
         // prefer a different gid, and there's no reason for us to change it).
         // The reason why we don't just leave the default uid=X mount setup is
         // that users expect to be able to actually use their console. Without
         // this code, you couldn't effectively run as a non-root user inside a
         // container and also have a console set up.
-        let res = unsafe { libc::fchown(*fd, uid, gid) };
-        Errno::result(res).map_err(|e| anyhow!(e).context("set stdio permissions failed"))?;
+        unistd::fchown(*fd, Some(uid), None).with_context(|| "set stdio permissions failed")?;
     }
 
     Ok(())
@@ -951,15 +946,7 @@ impl BaseContainer for LinuxContainer {
             let _ = unistd::close(pid);
         });
 
-        cfg_if::cfg_if! {
-            if #[cfg(feature = "standard-oci-runtime")] {
-                let exec_path = PathBuf::from(OCI_AGENT_BINARY);
-            }
-            else {
-                let exec_path = std::env::current_exe()?;
-            }
-        }
-
+        let exec_path = std::env::current_exe()?;
         let mut child = std::process::Command::new(exec_path);
 
         #[allow(unused_mut)]
@@ -1062,7 +1049,7 @@ impl BaseContainer for LinuxContainer {
         self.start(p).await?;
 
         if init {
-            self.exec()?;
+            self.exec().await?;
             self.status.transition(ContainerState::Running);
         }
 
@@ -1074,7 +1061,19 @@ impl BaseContainer for LinuxContainer {
         let st = self.oci_state()?;
 
         for pid in self.processes.keys() {
-            signal::kill(Pid::from_raw(*pid), Some(Signal::SIGKILL))?;
+            match signal::kill(Pid::from_raw(*pid), Some(Signal::SIGKILL)) {
+                Err(Errno::ESRCH) => {
+                    info!(
+                        self.logger,
+                        "kill encounters ESRCH, pid: {}, container: {}",
+                        pid,
+                        self.id.clone()
+                    );
+                    continue;
+                }
+                Err(err) => return Err(anyhow!(err)),
+                Ok(_) => continue,
+            }
         }
 
         if spec.hooks.is_some() {
@@ -1093,12 +1092,22 @@ impl BaseContainer for LinuxContainer {
         fs::remove_dir_all(&self.root)?;
 
         if let Some(cgm) = self.cgroup_manager.as_mut() {
+            // Kill all of the processes created in this container to prevent
+            // the leak of some daemon process when this container shared pidns
+            // with the sandbox.
+            let pids = cgm.get_pids().context("get cgroup pids")?;
+            for i in pids {
+                if let Err(e) = signal::kill(Pid::from_raw(i), Signal::SIGKILL) {
+                    warn!(self.logger, "kill the process {} error: {:?}", i, e);
+                }
+            }
+
             cgm.destroy().context("destroy cgroups")?;
         }
         Ok(())
     }
 
-    fn exec(&mut self) -> Result<()> {
+    async fn exec(&mut self) -> Result<()> {
         let fifo = format!("{}/{}", &self.root, EXEC_FIFO_FILENAME);
         let fd = fcntl::open(fifo.as_str(), OFlag::O_WRONLY, Mode::from_bits_truncate(0))?;
         let data: &[u8] = &[0];
@@ -1110,6 +1119,26 @@ impl BaseContainer for LinuxContainer {
             .as_secs();
 
         self.status.transition(ContainerState::Running);
+
+        let spec = self
+            .config
+            .spec
+            .as_ref()
+            .ok_or_else(|| anyhow!("OCI spec was not found"))?;
+        let st = self.oci_state()?;
+
+        // run poststart hook
+        if spec.hooks.is_some() {
+            info!(self.logger, "poststart hook");
+            let hooks = spec
+                .hooks
+                .as_ref()
+                .ok_or_else(|| anyhow!("OCI hooks were not found"))?;
+            for h in hooks.poststart.iter() {
+                execute_hook(&self.logger, h, &st).await?;
+            }
+        }
+
         unistd::close(fd)?;
 
         Ok(())
@@ -1152,7 +1181,7 @@ fn do_exec(args: &[String]) -> ! {
     unreachable!()
 }
 
-fn update_namespaces(logger: &Logger, spec: &mut Spec, init_pid: RawFd) -> Result<()> {
+pub fn update_namespaces(logger: &Logger, spec: &mut Spec, init_pid: RawFd) -> Result<()> {
     info!(logger, "updating namespaces");
     let linux = spec
         .linux
@@ -1331,20 +1360,6 @@ async fn join_namespaces(
         // notify child run prestart hooks completed
         info!(logger, "notify child run prestart hook completed!");
         write_async(pipe_w, SYNC_SUCCESS, "").await?;
-
-        info!(logger, "notify child parent ready to run poststart hook!");
-        // wait to run poststart hook
-        read_async(pipe_r).await?;
-        info!(logger, "get ready to run poststart hook!");
-
-        // run poststart hook
-        if spec.hooks.is_some() {
-            info!(logger, "poststart hook");
-            let hooks = spec.hooks.as_ref().unwrap();
-            for h in hooks.poststart.iter() {
-                execute_hook(&logger, h, st).await?;
-            }
-        }
     }
 
     info!(logger, "wait for child process ready to run exec");
@@ -1473,12 +1488,6 @@ impl LinuxContainer {
     }
 }
 
-fn setgroups(grps: &[libc::gid_t]) -> Result<()> {
-    let ret = unsafe { libc::setgroups(grps.len(), grps.as_ptr() as *const libc::gid_t) };
-    Errno::result(ret).map(drop)?;
-    Ok(())
-}
-
 use std::fs::OpenOptions;
 use std::io::Write;
 
@@ -1648,6 +1657,7 @@ mod tests {
     use super::*;
     use crate::process::Process;
     use crate::skip_if_not_root;
+    use nix::unistd::Uid;
     use std::fs;
     use std::os::unix::fs::MetadataExt;
     use std::os::unix::io::AsRawFd;
@@ -1793,7 +1803,7 @@ mod tests {
         let old_uid = meta.uid();
 
         let uid = 1000;
-        set_stdio_permissions(uid).unwrap();
+        set_stdio_permissions(Uid::from_raw(uid)).unwrap();
 
         let meta = fs::metadata("/dev/stdin").unwrap();
         assert_eq!(meta.uid(), uid);
@@ -1805,7 +1815,7 @@ mod tests {
         assert_eq!(meta.uid(), uid);
 
         // restore the uid
-        set_stdio_permissions(old_uid).unwrap();
+        set_stdio_permissions(Uid::from_raw(old_uid)).unwrap();
     }
 
     #[test]
@@ -2086,9 +2096,10 @@ mod tests {
         assert!(ret.is_ok(), "Expecting Ok, Got {:?}", ret);
     }
 
-    #[test]
-    fn test_linuxcontainer_exec() {
-        let ret = new_linux_container_and_then(|mut c: LinuxContainer| c.exec());
+    #[tokio::test]
+    async fn test_linuxcontainer_exec() {
+        let (c, _dir) = new_linux_container();
+        let ret = c.unwrap().exec().await;
         assert!(ret.is_err(), "Expecting Err, Got {:?}", ret);
     }
 

src/agent/rustjail/src/lib.rs

@@ -41,7 +41,6 @@ pub mod seccomp;
 pub mod specconv;
 pub mod sync;
 pub mod sync_with_async;
-pub mod utils;
 pub mod validator;
 
 use std::collections::HashMap;

src/agent/rustjail/src/mount.rs

@@ -780,18 +780,31 @@ fn mount_from(
             Path::new(&dest).parent().unwrap()
         };
 
-        let _ = fs::create_dir_all(&dir).map_err(|e| {
+        fs::create_dir_all(&dir).map_err(|e| {
             log_child!(
                 cfd_log,
                 "create dir {}: {}",
                 dir.to_str().unwrap(),
                 e.to_string()
-            )
-        });
+            );
+            e
+        })?;
 
         // make sure file exists so we can bind over it
         if !src.is_dir() {
-            let _ = OpenOptions::new().create(true).write(true).open(&dest);
+            let _ = OpenOptions::new()
+                .create(true)
+                .write(true)
+                .open(&dest)
+                .map_err(|e| {
+                    log_child!(
+                        cfd_log,
+                        "open/create dest error. {}: {:?}",
+                        dest.as_str(),
+                        e
+                    );
+                    e
+                })?;
         }
         src.to_str().unwrap().to_string()
     } else {
@@ -804,8 +817,10 @@ fn mount_from(
         }
     };
 
-    let _ = stat::stat(dest.as_str())
-        .map_err(|e| log_child!(cfd_log, "dest stat error. {}: {:?}", dest.as_str(), e));
+    let _ = stat::stat(dest.as_str()).map_err(|e| {
+        log_child!(cfd_log, "dest stat error. {}: {:?}", dest.as_str(), e);
+        e
+    })?;
 
     mount(
         Some(src.as_str()),

src/agent/rustjail/src/process.rs

@@ -5,7 +5,7 @@
 
 use libc::pid_t;
 use std::fs::File;
-use std::os::unix::io::RawFd;
+use std::os::unix::io::{AsRawFd, RawFd};
 use tokio::sync::mpsc::Sender;
 
 use nix::errno::Errno;
@@ -28,7 +28,6 @@ macro_rules! close_process_stream {
     ($self: ident, $stream:ident, $stream_type: ident) => {
         if $self.$stream.is_some() {
             $self.close_stream(StreamType::$stream_type);
-            let _ = unistd::close($self.$stream.unwrap());
             $self.$stream = None;
         }
     };
@@ -137,19 +136,25 @@ impl Process {
         info!(logger, "before create console socket!");
 
         if !p.tty {
-            info!(logger, "created console socket!");
+            if cfg!(feature = "standard-oci-runtime") {
+                p.stdin = Some(std::io::stdin().as_raw_fd());
+                p.stdout = Some(std::io::stdout().as_raw_fd());
+                p.stderr = Some(std::io::stderr().as_raw_fd());
+            } else {
+                info!(logger, "created console socket!");
 
-            let (stdin, pstdin) = unistd::pipe2(OFlag::O_CLOEXEC)?;
-            p.parent_stdin = Some(pstdin);
-            p.stdin = Some(stdin);
+                let (stdin, pstdin) = unistd::pipe2(OFlag::O_CLOEXEC)?;
+                p.parent_stdin = Some(pstdin);
+                p.stdin = Some(stdin);
 
-            let (pstdout, stdout) = create_extended_pipe(OFlag::O_CLOEXEC, pipe_size)?;
-            p.parent_stdout = Some(pstdout);
-            p.stdout = Some(stdout);
+                let (pstdout, stdout) = create_extended_pipe(OFlag::O_CLOEXEC, pipe_size)?;
+                p.parent_stdout = Some(pstdout);
+                p.stdout = Some(stdout);
 
-            let (pstderr, stderr) = create_extended_pipe(OFlag::O_CLOEXEC, pipe_size)?;
-            p.parent_stderr = Some(pstderr);
-            p.stderr = Some(stderr);
+                let (pstderr, stderr) = create_extended_pipe(OFlag::O_CLOEXEC, pipe_size)?;
+                p.parent_stderr = Some(pstderr);
+                p.stderr = Some(stderr);
+            }
         }
         Ok(p)
     }
@@ -219,7 +224,7 @@ impl Process {
         Some(writer)
     }
 
-    pub fn close_stream(&mut self, stream_type: StreamType) {
+    fn close_stream(&mut self, stream_type: StreamType) {
         let _ = self.readers.remove(&stream_type);
         let _ = self.writers.remove(&stream_type);
     }
@@ -284,5 +289,11 @@ mod tests {
         // group of the calling process.
         process.pid = 0;
         assert!(process.signal(libc::SIGCONT).is_ok());
+
+        if cfg!(feature = "standard-oci-runtime") {
+            assert_eq!(process.stdin.unwrap(), std::io::stdin().as_raw_fd());
+            assert_eq!(process.stdout.unwrap(), std::io::stdout().as_raw_fd());
+            assert_eq!(process.stderr.unwrap(), std::io::stderr().as_raw_fd());
+        }
     }
 }

src/agent/rustjail/src/utils.rs

@@ -1,120 +0,0 @@
-// Copyright (c) 2021 Ant Group
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-use anyhow::{anyhow, Context, Result};
-use libc::gid_t;
-use libc::uid_t;
-use std::fs::File;
-use std::io::{BufRead, BufReader};
-
-const PASSWD_FILE: &str = "/etc/passwd";
-
-// An entry from /etc/passwd
-#[derive(Debug, PartialEq, PartialOrd)]
-pub struct PasswdEntry {
-    // username
-    pub name: String,
-    // user password
-    pub passwd: String,
-    // user id
-    pub uid: uid_t,
-    // group id
-    pub gid: gid_t,
-    // user Information
-    pub gecos: String,
-    // home directory
-    pub dir: String,
-    // User's Shell
-    pub shell: String,
-}
-
-// get an entry for a given `uid` from `/etc/passwd`
-fn get_entry_by_uid(uid: uid_t, path: &str) -> Result<PasswdEntry> {
-    let file = File::open(path).with_context(|| format!("open file {}", path))?;
-    let mut reader = BufReader::new(file);
-
-    let mut line = String::new();
-    loop {
-        line.clear();
-        match reader.read_line(&mut line) {
-            Ok(0) => return Err(anyhow!(format!("file {} is empty", path))),
-            Ok(_) => (),
-            Err(e) => {
-                return Err(anyhow!(format!(
-                    "failed to read file {} with {:?}",
-                    path, e
-                )))
-            }
-        }
-
-        if line.starts_with('#') {
-            continue;
-        }
-
-        let parts: Vec<&str> = line.split(':').map(|part| part.trim()).collect();
-        if parts.len() != 7 {
-            continue;
-        }
-
-        match parts[2].parse() {
-            Err(_e) => continue,
-            Ok(new_uid) => {
-                if uid != new_uid {
-                    continue;
-                }
-
-                let entry = PasswdEntry {
-                    name: parts[0].to_string(),
-                    passwd: parts[1].to_string(),
-                    uid: new_uid,
-                    gid: parts[3].parse().unwrap_or(0),
-                    gecos: parts[4].to_string(),
-                    dir: parts[5].to_string(),
-                    shell: parts[6].to_string(),
-                };
-
-                return Ok(entry);
-            }
-        }
-    }
-}
-
-pub fn home_dir(uid: uid_t) -> Result<String> {
-    get_entry_by_uid(uid, PASSWD_FILE).map(|entry| entry.dir)
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use std::io::Write;
-    use tempfile::Builder;
-
-    #[test]
-    fn test_get_entry_by_uid() {
-        let tmpdir = Builder::new().tempdir().unwrap();
-        let tmpdir_path = tmpdir.path().to_str().unwrap();
-        let temp_passwd = format!("{}/passwd", tmpdir_path);
-
-        let mut tempf = File::create(temp_passwd.as_str()).unwrap();
-        let passwd_entries = "root:x:0:0:root:/root0:/bin/bash
-root:x:1:0:root:/root1:/bin/bash
-#root:x:1:0:root:/rootx:/bin/bash
-root:x:2:0:root:/root2:/bin/bash
-root:x:3:0:root:/root3
-root:x:3:0:root:/root3:/bin/bash";
-        writeln!(tempf, "{}", passwd_entries).unwrap();
-
-        let entry = get_entry_by_uid(0, temp_passwd.as_str()).unwrap();
-        assert_eq!(entry.dir.as_str(), "/root0");
-
-        let entry = get_entry_by_uid(1, temp_passwd.as_str()).unwrap();
-        assert_eq!(entry.dir.as_str(), "/root1");
-
-        let entry = get_entry_by_uid(2, temp_passwd.as_str()).unwrap();
-        assert_eq!(entry.dir.as_str(), "/root2");
-
-        let entry = get_entry_by_uid(3, temp_passwd.as_str()).unwrap();
-        assert_eq!(entry.dir.as_str(), "/root3");
-    }
-}

src/agent/src/console.rs

@@ -9,7 +9,7 @@ use anyhow::{anyhow, Result};
 use nix::fcntl::{self, FcntlArg, FdFlag, OFlag};
 use nix::libc::{STDERR_FILENO, STDIN_FILENO, STDOUT_FILENO};
 use nix::pty::{openpty, OpenptyResult};
-use nix::sys::socket::{self, AddressFamily, SockAddr, SockFlag, SockType};
+use nix::sys::socket::{self, AddressFamily, SockFlag, SockType, VsockAddr};
 use nix::sys::stat::Mode;
 use nix::sys::wait;
 use nix::unistd::{self, close, dup2, fork, setsid, ForkResult, Pid};
@@ -67,7 +67,7 @@ pub async fn debug_console_handler(
             SockFlag::SOCK_CLOEXEC,
             None,
         )?;
-        let addr = SockAddr::new_vsock(libc::VMADDR_CID_ANY, port);
+        let addr = VsockAddr::new(libc::VMADDR_CID_ANY, port);
         socket::bind(listenfd, &addr)?;
         socket::listen(listenfd, 1)?;
 

src/agent/src/main.rs

@@ -22,12 +22,11 @@ extern crate slog;
 use anyhow::{anyhow, Context, Result};
 use clap::{AppSettings, Parser};
 use nix::fcntl::OFlag;
-use nix::sys::socket::{self, AddressFamily, SockAddr, SockFlag, SockType};
+use nix::sys::socket::{self, AddressFamily, SockFlag, SockType, VsockAddr};
 use nix::unistd::{self, dup, Pid};
 use std::env;
 use std::ffi::OsStr;
 use std::fs::{self, File};
-use std::os::unix::ffi::OsStrExt;
 use std::os::unix::fs as unixfs;
 use std::os::unix::io::AsRawFd;
 use std::path::Path;
@@ -111,10 +110,6 @@ enum SubCommand {
 fn announce(logger: &Logger, config: &AgentConfig) {
     info!(logger, "announce";
     "agent-commit" => version::VERSION_COMMIT,
-
-    // Avoid any possibility of confusion with the old agent
-    "agent-type" => "rust",
-
     "agent-version" =>  version::AGENT_VERSION,
     "api-version" => version::API_VERSION,
     "config" => format!("{:?}", config),
@@ -133,7 +128,7 @@ async fn create_logger_task(rfd: RawFd, vsock_port: u32, shutdown: Receiver<bool
             None,
         )?;
 
-        let addr = SockAddr::new_vsock(libc::VMADDR_CID_ANY, vsock_port);
+        let addr = VsockAddr::new(libc::VMADDR_CID_ANY, vsock_port);
         socket::bind(listenfd, &addr)?;
         socket::listen(listenfd, 1)?;
 
@@ -214,7 +209,7 @@ async fn real_main() -> std::result::Result<(), Box<dyn std::error::Error>> {
 
     if config.log_level == slog::Level::Trace {
         // Redirect ttrpc log calls to slog iff full debug requested
-        ttrpc_log_guard = Ok(slog_stdlog::init().map_err(|e| e)?);
+        ttrpc_log_guard = Ok(slog_stdlog::init()?);
     }
 
     if config.tracing {
@@ -382,27 +377,13 @@ fn init_agent_as_init(logger: &Logger, unified_cgroup_hierarchy: bool) -> Result
     let contents_array: Vec<&str> = contents.split(' ').collect();
     let hostname = contents_array[0].trim();
 
-    if sethostname(OsStr::new(hostname)).is_err() {
+    if unistd::sethostname(OsStr::new(hostname)).is_err() {
         warn!(logger, "failed to set hostname");
     }
 
     Ok(())
 }
 
-#[instrument]
-fn sethostname(hostname: &OsStr) -> Result<()> {
-    let size = hostname.len() as usize;
-
-    let result =
-        unsafe { libc::sethostname(hostname.as_bytes().as_ptr() as *const libc::c_char, size) };
-
-    if result != 0 {
-        Err(anyhow!("failed to set hostname"))
-    } else {
-        Ok(())
-    }
-}
-
 // The Rust standard library had suppressed the default SIGPIPE behavior,
 // see https://github.com/rust-lang/rust/pull/13158.
 // Since the parent's signal handler would be inherited by it's child process,

src/agent/src/mount.rs

@@ -840,15 +840,13 @@ pub fn get_mount_fs_type_from_file(mount_file: &str, mount_point: &str) -> Resul
         return Err(anyhow!("Invalid mount point {}", mount_point));
     }
 
-    let file = File::open(mount_file)?;
-    let reader = BufReader::new(file);
+    let content = fs::read_to_string(mount_file)?;
 
     let re = Regex::new(format!("device .+ mounted on {} with fstype (.+)", mount_point).as_str())?;
 
     // Read the file line by line using the lines() iterator from std::io::BufRead.
-    for (_index, line) in reader.lines().enumerate() {
-        let line = line?;
-        let capes = match re.captures(line.as_str()) {
+    for (_index, line) in content.lines().enumerate() {
+        let capes = match re.captures(line) {
             Some(c) => c,
             None => continue,
         };
@@ -859,8 +857,9 @@ pub fn get_mount_fs_type_from_file(mount_file: &str, mount_point: &str) -> Resul
     }
 
     Err(anyhow!(
-        "failed to find FS type for mount point {}",
-        mount_point
+        "failed to find FS type for mount point {}, mount file content: {:?}",
+        mount_point,
+        content
     ))
 }
 
@@ -1018,7 +1017,7 @@ fn parse_options(option_list: Vec<String>) -> HashMap<String, String> {
 mod tests {
     use super::*;
     use crate::test_utils::test_utils::TestUserType;
-    use crate::{skip_if_not_root, skip_loop_if_not_root, skip_loop_if_root};
+    use crate::{skip_if_not_root, skip_loop_by_user, skip_loop_if_not_root, skip_loop_if_root};
     use protobuf::RepeatedField;
     use protocols::agent::FSGroup;
     use std::fs::File;
@@ -1112,11 +1111,7 @@ mod tests {
         for (i, d) in tests.iter().enumerate() {
             let msg = format!("test[{}]: {:?}", i, d);
 
-            if d.test_user == TestUserType::RootOnly {
-                skip_loop_if_not_root!(msg);
-            } else if d.test_user == TestUserType::NonRootOnly {
-                skip_loop_if_root!(msg);
-            }
+            skip_loop_by_user!(msg, d.test_user);
 
             let src: PathBuf;
             let dest: PathBuf;
@@ -1649,11 +1644,8 @@ mod tests {
 
         for (i, d) in tests.iter().enumerate() {
             let msg = format!("test[{}]: {:?}", i, d);
-            if d.test_user == TestUserType::RootOnly {
-                skip_loop_if_not_root!(msg);
-            } else if d.test_user == TestUserType::NonRootOnly {
-                skip_loop_if_root!(msg);
-            }
+
+            skip_loop_by_user!(msg, d.test_user);
 
             let drain = slog::Discard;
             let logger = slog::Logger::root(drain, o!());
@@ -1762,11 +1754,7 @@ mod tests {
 
         for (i, d) in tests.iter().enumerate() {
             let msg = format!("test[{}]: {:?}", i, d);
-            if d.test_user == TestUserType::RootOnly {
-                skip_loop_if_not_root!(msg);
-            } else if d.test_user == TestUserType::NonRootOnly {
-                skip_loop_if_root!(msg);
-            }
+            skip_loop_by_user!(msg, d.test_user);
 
             let drain = slog::Discard;
             let logger = slog::Logger::root(drain, o!());

src/agent/src/netlink.rs

@@ -4,7 +4,7 @@
 //
 
 use anyhow::{anyhow, Context, Result};
-use futures::{future, StreamExt, TryStreamExt};
+use futures::{future, TryStreamExt};
 use ipnetwork::{IpNetwork, Ipv4Network, Ipv6Network};
 use nix::errno::Errno;
 use protobuf::RepeatedField;
@@ -64,7 +64,7 @@ impl Handle {
     pub async fn update_interface(&mut self, iface: &Interface) -> Result<()> {
         // The reliable way to find link is using hardware address
         // as filter. However, hardware filter might not be supported
-        // by netlink, we may have to dump link list and the find the
+        // by netlink, we may have to dump link list and then find the
         // target link. filter using name or family is supported, but
         // we cannot use that to find target link.
         // let's try if hardware address filter works. -_-
@@ -164,7 +164,7 @@ impl Handle {
         let request = self.handle.link().get();
 
         let filtered = match filter {
-            LinkFilter::Name(name) => request.set_name_filter(name.to_owned()),
+            LinkFilter::Name(name) => request.match_name(name.to_owned()),
             LinkFilter::Index(index) => request.match_index(index),
             _ => request, // Post filters
         };
@@ -178,7 +178,7 @@ impl Handle {
                 .with_context(|| format!("Failed to parse MAC address: {}", addr))?;
 
             // Hardware filter might not be supported by netlink,
-            // we may have to dump link list and the find the target link.
+            // we may have to dump link list and then find the target link.
             stream
                 .try_filter(|f| {
                     let result = f.nlas.iter().any(|n| match n {
@@ -516,70 +516,24 @@ impl Handle {
     }
 
     /// Adds an ARP neighbor.
-    /// TODO: `rtnetlink` has no neighbours API, remove this after https://github.com/little-dude/netlink/pull/135
     async fn add_arp_neighbor(&mut self, neigh: &ARPNeighbor) -> Result<()> {
         let ip_address = neigh
             .toIPAddress
             .as_ref()
             .map(|to| to.address.as_str()) // Extract address field
             .and_then(|addr| if addr.is_empty() { None } else { Some(addr) }) // Make sure it's not empty
-            .ok_or(anyhow!(nix::Error::EINVAL))?;
+            .ok_or_else(|| anyhow!(nix::Error::EINVAL))?;
 
         let ip = IpAddr::from_str(ip_address)
             .map_err(|e| anyhow!("Failed to parse IP {}: {:?}", ip_address, e))?;
 
-        // Import rtnetlink objects that make sense only for this function
-        use packet::constants::{NDA_UNSPEC, NLM_F_ACK, NLM_F_CREATE, NLM_F_EXCL, NLM_F_REQUEST};
-        use packet::neighbour::{NeighbourHeader, NeighbourMessage};
-        use packet::nlas::neighbour::Nla;
-        use packet::{NetlinkMessage, NetlinkPayload, RtnlMessage};
-        use rtnetlink::Error;
-
-        const IFA_F_PERMANENT: u16 = 0x80; // See https://github.com/little-dude/netlink/blob/0185b2952505e271805902bf175fee6ea86c42b8/netlink-packet-route/src/rtnl/constants.rs#L770
-
         let link = self.find_link(LinkFilter::Name(&neigh.device)).await?;
 
-        let message = NeighbourMessage {
-            header: NeighbourHeader {
-                family: match ip {
-                    IpAddr::V4(_) => packet::AF_INET,
-                    IpAddr::V6(_) => packet::AF_INET6,
-                } as u8,
-                ifindex: link.index(),
-                state: if neigh.state != 0 {
-                    neigh.state as u16
-                } else {
-                    IFA_F_PERMANENT
-                },
-                flags: neigh.flags as u8,
-                ntype: NDA_UNSPEC as u8,
-            },
-            nlas: {
-                let mut nlas = vec![Nla::Destination(match ip {
-                    IpAddr::V4(v4) => v4.octets().to_vec(),
-                    IpAddr::V6(v6) => v6.octets().to_vec(),
-                })];
-
-                if !neigh.lladdr.is_empty() {
-                    nlas.push(Nla::LinkLocalAddress(
-                        parse_mac_address(&neigh.lladdr)?.to_vec(),
-                    ));
-                }
-
-                nlas
-            },
-        };
-
-        // Send request and ACK
-        let mut req = NetlinkMessage::from(RtnlMessage::NewNeighbour(message));
-        req.header.flags = NLM_F_REQUEST | NLM_F_ACK | NLM_F_EXCL | NLM_F_CREATE;
-
-        let mut response = self.handle.request(req)?;
-        while let Some(message) = response.next().await {
-            if let NetlinkPayload::Error(err) = message.payload {
-                return Err(anyhow!(Error::NetlinkError(err)));
-            }
-        }
+        self.handle
+            .neighbours()
+            .add(link.index(), ip)
+            .execute()
+            .await?;
 
         Ok(())
     }
@@ -612,7 +566,7 @@ fn parse_mac_address(addr: &str) -> Result<[u8; 6]> {
 
     // Parse single Mac address block
     let mut parse_next = || -> Result<u8> {
-        let v = u8::from_str_radix(split.next().ok_or(anyhow!(nix::Error::EINVAL))?, 16)?;
+        let v = u8::from_str_radix(split.next().ok_or_else(|| anyhow!(nix::Error::EINVAL))?, 16)?;
         Ok(v)
     };
 
@@ -950,7 +904,7 @@ mod tests {
             .expect("prepare: failed to delete neigh");
     }
 
-    fn prepare_env_for_test_add_one_arp_neighbor(dummy_name: &str, ip: &str) {
+    fn prepare_env_for_test_add_one_arp_neighbor(dummy_name: &str, ip: &str, mac: &str) {
         clean_env_for_test_add_one_arp_neighbor(dummy_name, ip);
         // modprobe dummy
         Command::new("modprobe")
@@ -964,6 +918,12 @@ mod tests {
             .output()
             .expect("failed to add dummy interface");
 
+        // ip link set dummy address 6a:92:3a:59:70:aa
+        Command::new("ip")
+            .args(&["link", "set", dummy_name, "address", mac])
+            .output()
+            .expect("failed to add dummy interface");
+
         // ip addr add 192.168.0.2/16 dev dummy
         Command::new("ip")
             .args(&["addr", "add", "192.168.0.2/16", "dev", dummy_name])
@@ -985,7 +945,7 @@ mod tests {
         let to_ip = "169.254.1.1";
         let dummy_name = "dummy_for_arp";
 
-        prepare_env_for_test_add_one_arp_neighbor(dummy_name, to_ip);
+        prepare_env_for_test_add_one_arp_neighbor(dummy_name, to_ip, mac);
 
         let mut ip_address = IPAddress::new();
         ip_address.set_address(to_ip.to_string());

src/agent/src/random.rs

@@ -82,7 +82,7 @@ mod tests {
         if nix::unistd::Uid::effective().is_root() {
             assert!(ret.is_ok());
         } else {
-            assert!(!ret.is_ok());
+            assert!(ret.is_err());
         }
     }
 
@@ -90,6 +90,6 @@ mod tests {
     fn test_reseed_rng_zero_data() {
         let seed = [];
         let ret = reseed_rng(&seed);
-        assert!(!ret.is_ok());
+        assert!(ret.is_err());
     }
 }

src/agent/src/rpc.rs

@@ -23,8 +23,9 @@ use cgroups::freezer::FreezerState;
 use oci::{LinuxNamespace, Root, Spec};
 use protobuf::{Message, RepeatedField, SingularPtrField};
 use protocols::agent::{
-    AddSwapRequest, AgentDetails, CopyFileRequest, GuestDetailsResponse, Interfaces, Metrics,
-    OOMEvent, ReadStreamResponse, Routes, StatsContainerResponse, VolumeStatsRequest,
+    AddSwapRequest, AgentDetails, CopyFileRequest, GetIPTablesRequest, GetIPTablesResponse,
+    GuestDetailsResponse, Interfaces, Metrics, OOMEvent, ReadStreamResponse, Routes,
+    SetIPTablesRequest, SetIPTablesResponse, StatsContainerResponse, VolumeStatsRequest,
     WaitProcessResponse, WriteStreamResponse,
 };
 use protocols::csi::{VolumeCondition, VolumeStatsResponse, VolumeUsage, VolumeUsage_Unit};
@@ -33,6 +34,7 @@ use protocols::health::{
     HealthCheckResponse, HealthCheckResponse_ServingStatus, VersionCheckResponse,
 };
 use protocols::types::Interface;
+use protocols::{agent_ttrpc_async as agent_ttrpc, health_ttrpc_async as health_ttrpc};
 use rustjail::cgroups::notifier;
 use rustjail::container::{BaseContainer, Container, LinuxContainer};
 use rustjail::process::Process;
@@ -40,13 +42,11 @@ use rustjail::specconv::CreateOpts;
 
 use nix::errno::Errno;
 use nix::mount::MsFlags;
-use nix::sys::stat;
+use nix::sys::{stat, statfs};
 use nix::unistd::{self, Pid};
 use rustjail::cgroups::Manager;
 use rustjail::process::ProcessOperations;
 
-use sysinfo::{DiskExt, System, SystemExt};
-
 use crate::device::{
     add_devices, get_virtio_blk_pci_device_name, update_device_cgroup, update_env_pci,
 };
@@ -71,25 +71,35 @@ use tracing::instrument;
 
 use libc::{self, c_char, c_ushort, pid_t, winsize, TIOCSWINSZ};
 use std::fs;
-use std::os::unix::fs::MetadataExt;
 use std::os::unix::prelude::PermissionsExt;
 use std::process::{Command, Stdio};
 use std::time::Duration;
 
 use nix::unistd::{Gid, Uid};
 use std::fs::{File, OpenOptions};
-use std::io::{BufRead, BufReader};
+use std::io::{BufRead, BufReader, Write};
 use std::os::unix::fs::FileExt;
 use std::path::PathBuf;
 
 const CONTAINER_BASE: &str = "/run/kata-containers";
 const MODPROBE_PATH: &str = "/sbin/modprobe";
 
+const IPTABLES_SAVE: &str = "/sbin/iptables-save";
+const IPTABLES_RESTORE: &str = "/sbin/iptables-restore";
+const IP6TABLES_SAVE: &str = "/sbin/ip6tables-save";
+const IP6TABLES_RESTORE: &str = "/sbin/ip6tables-restore";
+
 const ERR_CANNOT_GET_WRITER: &str = "Cannot get writer";
 const ERR_INVALID_BLOCK_SIZE: &str = "Invalid block size";
 const ERR_NO_LINUX_FIELD: &str = "Spec does not contain linux field";
 const ERR_NO_SANDBOX_PIDNS: &str = "Sandbox does not have sandbox_pidns";
 
+// IPTABLES_RESTORE_WAIT_SEC is the timeout value provided to iptables-restore --wait. Since we
+// don't expect other writers to iptables, we don't expect contention for grabbing the iptables
+// filesystem lock. Based on this, 5 seconds seems a resonable timeout period in case the lock is
+// not available.
+const IPTABLES_RESTORE_WAIT_SEC: u64 = 5;
+
 // Convenience macro to obtain the scope logger
 macro_rules! sl {
     () => {
@@ -124,30 +134,6 @@ pub struct AgentService {
     sandbox: Arc<Mutex<Sandbox>>,
 }
 
-// A container ID must match this regex:
-//
-//     ^[a-zA-Z0-9][a-zA-Z0-9_.-]+$
-//
-fn verify_cid(id: &str) -> Result<()> {
-    let mut chars = id.chars();
-
-    let valid = match chars.next() {
-        Some(first)
-            if first.is_alphanumeric()
-                && id.len() > 1
-                && chars.all(|c| c.is_alphanumeric() || ['.', '-', '_'].contains(&c)) =>
-        {
-            true
-        }
-        _ => false,
-    };
-
-    match valid {
-        true => Ok(()),
-        false => Err(anyhow!("invalid container ID: {:?}", id)),
-    }
-}
-
 impl AgentService {
     #[instrument]
     async fn do_create_container(
@@ -156,7 +142,7 @@ impl AgentService {
     ) -> Result<()> {
         let cid = req.container_id.clone();
 
-        verify_cid(&cid)?;
+        kata_sys_util::validate::verify_id(&cid)?;
 
         let mut oci_spec = req.OCI.clone();
         let use_sandbox_pidns = req.get_sandbox_pidns();
@@ -260,7 +246,7 @@ impl AgentService {
             .get_container(&cid)
             .ok_or_else(|| anyhow!("Invalid container id"))?;
 
-        ctr.exec()?;
+        ctr.exec().await?;
 
         if sid == cid {
             return Ok(());
@@ -641,7 +627,7 @@ impl AgentService {
 }
 
 #[async_trait]
-impl protocols::agent_ttrpc::AgentService for AgentService {
+impl agent_ttrpc::AgentService for AgentService {
     async fn create_container(
         &self,
         ctx: &TtrpcContext,
@@ -997,6 +983,140 @@ impl protocols::agent_ttrpc::AgentService for AgentService {
         })
     }
 
+    async fn get_ip_tables(
+        &self,
+        ctx: &TtrpcContext,
+        req: GetIPTablesRequest,
+    ) -> ttrpc::Result<GetIPTablesResponse> {
+        trace_rpc_call!(ctx, "get_iptables", req);
+        is_allowed!(req);
+
+        info!(sl!(), "get_ip_tables: request received");
+
+        let cmd = if req.is_ipv6 {
+            IP6TABLES_SAVE
+        } else {
+            IPTABLES_SAVE
+        }
+        .to_string();
+
+        match Command::new(cmd.clone()).output() {
+            Ok(output) => Ok(GetIPTablesResponse {
+                data: output.stdout,
+                ..Default::default()
+            }),
+            Err(e) => {
+                warn!(sl!(), "failed to run {}: {:?}", cmd, e.kind());
+                return Err(ttrpc_error!(ttrpc::Code::INTERNAL, e));
+            }
+        }
+    }
+
+    async fn set_ip_tables(
+        &self,
+        ctx: &TtrpcContext,
+        req: SetIPTablesRequest,
+    ) -> ttrpc::Result<SetIPTablesResponse> {
+        trace_rpc_call!(ctx, "set_iptables", req);
+        is_allowed!(req);
+
+        info!(sl!(), "set_ip_tables request received");
+
+        let cmd = if req.is_ipv6 {
+            IP6TABLES_RESTORE
+        } else {
+            IPTABLES_RESTORE
+        }
+        .to_string();
+
+        let mut child = match Command::new(cmd.clone())
+            .arg("--wait")
+            .arg(IPTABLES_RESTORE_WAIT_SEC.to_string())
+            .stdin(Stdio::piped())
+            .stdout(Stdio::piped())
+            .stderr(Stdio::piped())
+            .spawn()
+        {
+            Ok(child) => child,
+            Err(e) => {
+                warn!(sl!(), "failure to spawn {}: {:?}", cmd, e.kind());
+                return Err(ttrpc_error!(ttrpc::Code::INTERNAL, e));
+            }
+        };
+
+        let mut stdin = match child.stdin.take() {
+            Some(si) => si,
+            None => {
+                println!("failed to get stdin from child");
+                return Err(ttrpc_error!(
+                    ttrpc::Code::INTERNAL,
+                    "failed to take stdin from child".to_string()
+                ));
+            }
+        };
+
+        let (tx, rx) = tokio::sync::oneshot::channel::<i32>();
+        let handle = tokio::spawn(async move {
+            let _ = match stdin.write_all(&req.data) {
+                Ok(o) => o,
+                Err(e) => {
+                    warn!(sl!(), "error writing stdin: {:?}", e.kind());
+                    return;
+                }
+            };
+            if tx.send(1).is_err() {
+                warn!(sl!(), "stdin writer thread receiver dropped");
+            };
+        });
+
+        if tokio::time::timeout(Duration::from_secs(IPTABLES_RESTORE_WAIT_SEC), rx)
+            .await
+            .is_err()
+        {
+            return Err(ttrpc_error!(
+                ttrpc::Code::INTERNAL,
+                "timeout waiting for stdin writer to complete".to_string()
+            ));
+        }
+
+        if handle.await.is_err() {
+            return Err(ttrpc_error!(
+                ttrpc::Code::INTERNAL,
+                "stdin writer thread failure".to_string()
+            ));
+        }
+
+        let output = match child.wait_with_output() {
+            Ok(o) => o,
+            Err(e) => {
+                warn!(
+                    sl!(),
+                    "failure waiting for spawned {} to complete: {:?}",
+                    cmd,
+                    e.kind()
+                );
+                return Err(ttrpc_error!(ttrpc::Code::INTERNAL, e));
+            }
+        };
+
+        if !output.status.success() {
+            warn!(sl!(), "{} failed: {:?}", cmd, output.stderr);
+            return Err(ttrpc_error!(
+                ttrpc::Code::INTERNAL,
+                format!(
+                    "{} failed: {:?}",
+                    cmd,
+                    String::from_utf8_lossy(&output.stderr)
+                )
+            ));
+        }
+
+        Ok(SetIPTablesResponse {
+            data: output.stdout,
+            ..Default::default()
+        })
+    }
+
     async fn list_interfaces(
         &self,
         ctx: &TtrpcContext,
@@ -1393,7 +1513,7 @@ impl protocols::agent_ttrpc::AgentService for AgentService {
 struct HealthService;
 
 #[async_trait]
-impl protocols::health_ttrpc::Health for HealthService {
+impl health_ttrpc::Health for HealthService {
     async fn check(
         &self,
         _ctx: &TtrpcContext,
@@ -1468,20 +1588,12 @@ fn get_memory_info(
 fn get_volume_capacity_stats(path: &str) -> Result<VolumeUsage> {
     let mut usage = VolumeUsage::new();
 
-    let s = System::new();
-    for disk in s.disks() {
-        if let Some(v) = disk.name().to_str() {
-            if v.to_string().eq(path) {
-                usage.available = disk.available_space();
-                usage.total = disk.total_space();
-                usage.used = usage.total - usage.available;
-                usage.unit = VolumeUsage_Unit::BYTES; // bytes
-                break;
-            }
-        } else {
-            return Err(anyhow!(nix::Error::EINVAL));
-        }
-    }
+    let stat = statfs::statfs(path)?;
+    let block_size = stat.block_size() as u64;
+    usage.total = stat.blocks() * block_size;
+    usage.available = stat.blocks_free() * block_size;
+    usage.used = usage.total - usage.available;
+    usage.unit = VolumeUsage_Unit::BYTES;
 
     Ok(usage)
 }
@@ -1489,20 +1601,11 @@ fn get_volume_capacity_stats(path: &str) -> Result<VolumeUsage> {
 fn get_volume_inode_stats(path: &str) -> Result<VolumeUsage> {
     let mut usage = VolumeUsage::new();
 
-    let s = System::new();
-    for disk in s.disks() {
-        if let Some(v) = disk.name().to_str() {
-            if v.to_string().eq(path) {
-                let meta = fs::metadata(disk.mount_point())?;
-                let inode = meta.ino();
-                usage.used = inode;
-                usage.unit = VolumeUsage_Unit::INODES;
-                break;
-            }
-        } else {
-            return Err(anyhow!(nix::Error::EINVAL));
-        }
-    }
+    let stat = statfs::statfs(path)?;
+    usage.total = stat.files();
+    usage.available = stat.files_free();
+    usage.used = usage.total - usage.available;
+    usage.unit = VolumeUsage_Unit::INODES;
 
     Ok(usage)
 }
@@ -1549,18 +1652,17 @@ async fn read_stream(reader: Arc<Mutex<ReadHalf<PipeStream>>>, l: usize) -> Resu
 }
 
 pub fn start(s: Arc<Mutex<Sandbox>>, server_address: &str) -> Result<TtrpcServer> {
-    let agent_service = Box::new(AgentService { sandbox: s })
-        as Box<dyn protocols::agent_ttrpc::AgentService + Send + Sync>;
+    let agent_service =
+        Box::new(AgentService { sandbox: s }) as Box<dyn agent_ttrpc::AgentService + Send + Sync>;
 
     let agent_worker = Arc::new(agent_service);
 
-    let health_service =
-        Box::new(HealthService {}) as Box<dyn protocols::health_ttrpc::Health + Send + Sync>;
+    let health_service = Box::new(HealthService {}) as Box<dyn health_ttrpc::Health + Send + Sync>;
     let health_worker = Arc::new(health_service);
 
-    let aservice = protocols::agent_ttrpc::create_agent_service(agent_worker);
+    let aservice = agent_ttrpc::create_agent_service(agent_worker);
 
-    let hservice = protocols::health_ttrpc::create_health(health_worker);
+    let hservice = health_ttrpc::create_health(health_worker);
 
     let server = TtrpcServer::new()
         .bind(server_address)?
@@ -1667,34 +1769,25 @@ fn is_signal_handled(proc_status_file: &str, signum: u32) -> bool {
     let sig_mask: u64 = 1 << shift_count;
     let reader = BufReader::new(file);
 
-    // Read the file line by line using the lines() iterator from std::io::BufRead.
-    for (_index, line) in reader.lines().enumerate() {
-        let line = match line {
-            Ok(l) => l,
-            Err(_) => {
-                warn!(sl!(), "failed to read file {}", proc_status_file);
-                return false;
-            }
-        };
-        if line.starts_with("SigCgt:") {
+    // read lines start with SigBlk/SigIgn/SigCgt and check any match the signal mask
+    reader
+        .lines()
+        .flatten()
+        .filter(|line| {
+            line.starts_with("SigBlk:")
+                || line.starts_with("SigIgn:")
+                || line.starts_with("SigCgt:")
+        })
+        .any(|line| {
             let mask_vec: Vec<&str> = line.split(':').collect();
-            if mask_vec.len() != 2 {
-                warn!(sl!(), "parse the SigCgt field failed");
-                return false;
-            }
-            let sig_cgt_str = mask_vec[1];
-            let sig_cgt_mask = match u64::from_str_radix(sig_cgt_str, 16) {
-                Ok(h) => h,
-                Err(_) => {
-                    warn!(sl!(), "failed to parse the str {} to hex", sig_cgt_str);
-                    return false;
+            if mask_vec.len() == 2 {
+                let sig_str = mask_vec[1].trim();
+                if let Ok(sig) = u64::from_str_radix(sig_str, 16) {
+                    return sig & sig_mask == sig_mask;
                 }
-            };
-
-            return (sig_cgt_mask & sig_mask) == sig_mask;
-        }
-    }
-    false
+            }
+            false
+        })
 }
 
 fn do_mem_hotplug_by_probe(addrs: &[u64]) -> Result<()> {
@@ -1894,7 +1987,12 @@ fn load_kernel_module(module: &protocols::agent::KernelModule) -> Result<()> {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::{assert_result, namespace::Namespace, protocols::agent_ttrpc::AgentService as _};
+    use crate::{
+        assert_result, namespace::Namespace, protocols::agent_ttrpc_async::AgentService as _,
+        skip_if_not_root,
+    };
+    use nix::mount;
+    use nix::sched::{unshare, CloneFlags};
     use oci::{Hook, Hooks, Linux, LinuxNamespace};
     use tempfile::{tempdir, TempDir};
     use ttrpc::{r#async::TtrpcContext, MessageHeader};
@@ -2140,6 +2238,7 @@ mod tests {
                     if d.has_fd {
                         Some(wfd)
                     } else {
+                        unistd::close(wfd).unwrap();
                         None
                     }
                 };
@@ -2174,13 +2273,14 @@ mod tests {
             if !d.break_pipe {
                 unistd::close(rfd).unwrap();
             }
-            unistd::close(wfd).unwrap();
+            // XXX: Do not close wfd.
+            // the fd will be closed on Process's dropping.
+            // unistd::close(wfd).unwrap();
 
             let msg = format!("{}, result: {:?}", msg, result);
             assert_result!(d.result, result, msg);
         }
     }
-
     #[tokio::test]
     async fn test_update_container_namespaces() {
         #[derive(Debug)]
@@ -2453,6 +2553,26 @@ OtherField:other
                 signum: 4,
                 result: true,
             },
+            TestData {
+                status_file_data: Some("SigCgt:\t000000004b813efb"),
+                signum: 4,
+                result: true,
+            },
+            TestData {
+                status_file_data: Some("SigCgt: 000000004b813efb"),
+                signum: 4,
+                result: true,
+            },
+            TestData {
+                status_file_data: Some("SigCgt:000000004b813efb "),
+                signum: 4,
+                result: true,
+            },
+            TestData {
+                status_file_data: Some("SigCgt:\t000000004b813efb "),
+                signum: 4,
+                result: true,
+            },
             TestData {
                 status_file_data: Some("SigCgt:000000004b813efb"),
                 signum: 3,
@@ -2491,7 +2611,12 @@ OtherField:other
             TestData {
                 status_file_data: Some("SigBlk:0000000000000001"),
                 signum: 1,
-                result: false,
+                result: true,
+            },
+            TestData {
+                status_file_data: Some("SigIgn:0000000000000001"),
+                signum: 1,
+                result: true,
             },
             TestData {
                 status_file_data: None,
@@ -2524,229 +2649,222 @@ OtherField:other
     }
 
     #[tokio::test]
-    async fn test_verify_cid() {
-        #[derive(Debug)]
-        struct TestData<'a> {
-            id: &'a str,
-            expect_error: bool,
-        }
+    async fn test_volume_capacity_stats() {
+        skip_if_not_root!();
 
-        let tests = &[
-            TestData {
-                // Cannot be blank
-                id: "",
-                expect_error: true,
-            },
-            TestData {
-                // Cannot be a space
-                id: " ",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: ".",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "-",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "_",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: " a",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: ".a",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "-a",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "_a",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "..",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "a",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "z",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "A",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "Z",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "0",
-                expect_error: true,
-            },
-            TestData {
-                // Too short
-                id: "9",
-                expect_error: true,
-            },
-            TestData {
-                // Must start with an alphanumeric
-                id: "-1",
-                expect_error: true,
-            },
-            TestData {
-                id: "/",
-                expect_error: true,
-            },
-            TestData {
-                id: "a/",
-                expect_error: true,
-            },
-            TestData {
-                id: "a/../",
-                expect_error: true,
-            },
-            TestData {
-                id: "../a",
-                expect_error: true,
-            },
-            TestData {
-                id: "../../a",
-                expect_error: true,
-            },
-            TestData {
-                id: "../../../a",
-                expect_error: true,
-            },
-            TestData {
-                id: "foo/../bar",
-                expect_error: true,
-            },
-            TestData {
-                id: "foo bar",
-                expect_error: true,
-            },
-            TestData {
-                id: "a.",
-                expect_error: false,
-            },
-            TestData {
-                id: "a..",
-                expect_error: false,
-            },
-            TestData {
-                id: "aa",
-                expect_error: false,
-            },
-            TestData {
-                id: "aa.",
-                expect_error: false,
-            },
-            TestData {
-                id: "hello..world",
-                expect_error: false,
-            },
-            TestData {
-                id: "hello/../world",
-                expect_error: true,
-            },
-            TestData {
-                id: "aa1245124sadfasdfgasdga.",
-                expect_error: false,
-            },
-            TestData {
-                id: "aAzZ0123456789_.-",
-                expect_error: false,
-            },
-            TestData {
-                id: "abcdefghijklmnopqrstuvwxyz0123456789.-_",
-                expect_error: false,
-            },
-            TestData {
-                id: "0123456789abcdefghijklmnopqrstuvwxyz.-_",
-                expect_error: false,
-            },
-            TestData {
-                id: " abcdefghijklmnopqrstuvwxyz0123456789.-_",
-                expect_error: true,
-            },
-            TestData {
-                id: ".abcdefghijklmnopqrstuvwxyz0123456789.-_",
-                expect_error: true,
-            },
-            TestData {
-                id: "ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.-_",
-                expect_error: false,
-            },
-            TestData {
-                id: "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ.-_",
-                expect_error: false,
-            },
-            TestData {
-                id: " ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.-_",
-                expect_error: true,
-            },
-            TestData {
-                id: ".ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.-_",
-                expect_error: true,
-            },
-            TestData {
-                id: "/a/b/c",
-                expect_error: true,
-            },
-            TestData {
-                id: "a/b/c",
-                expect_error: true,
-            },
-            TestData {
-                id: "foo/../../../etc/passwd",
-                expect_error: true,
-            },
-            TestData {
-                id: "../../../../../../etc/motd",
-                expect_error: true,
-            },
-            TestData {
-                id: "/etc/passwd",
-                expect_error: true,
-            },
-        ];
+        // Verify error if path does not exist
+        assert!(get_volume_capacity_stats("/does-not-exist").is_err());
 
-        for (i, d) in tests.iter().enumerate() {
-            let msg = format!("test[{}]: {:?}", i, d);
+        // Create a new tmpfs mount, and verify the initial values
+        let mount_dir = tempfile::tempdir().unwrap();
+        mount::mount(
+            Some("tmpfs"),
+            mount_dir.path().to_str().unwrap(),
+            Some("tmpfs"),
+            mount::MsFlags::empty(),
+            None::<&str>,
+        )
+        .unwrap();
+        let mut stats = get_volume_capacity_stats(mount_dir.path().to_str().unwrap()).unwrap();
+        assert_eq!(stats.used, 0);
+        assert_ne!(stats.available, 0);
+        let available = stats.available;
 
-            let result = verify_cid(d.id);
+        // Verify that writing a file will result in increased utilization
+        fs::write(mount_dir.path().join("file.dat"), "foobar").unwrap();
+        stats = get_volume_capacity_stats(mount_dir.path().to_str().unwrap()).unwrap();
 
-            let msg = format!("{}, result: {:?}", msg, result);
+        assert_eq!(stats.used, 4 * 1024);
+        assert_eq!(stats.available, available - 4 * 1024);
+    }
 
-            if result.is_ok() {
-                assert!(!d.expect_error, "{}", msg);
-            } else {
-                assert!(d.expect_error, "{}", msg);
-            }
-        }
+    #[tokio::test]
+    async fn test_get_volume_inode_stats() {
+        skip_if_not_root!();
+
+        // Verify error if path does not exist
+        assert!(get_volume_inode_stats("/does-not-exist").is_err());
+
+        // Create a new tmpfs mount, and verify the initial values
+        let mount_dir = tempfile::tempdir().unwrap();
+        mount::mount(
+            Some("tmpfs"),
+            mount_dir.path().to_str().unwrap(),
+            Some("tmpfs"),
+            mount::MsFlags::empty(),
+            None::<&str>,
+        )
+        .unwrap();
+        let mut stats = get_volume_inode_stats(mount_dir.path().to_str().unwrap()).unwrap();
+        assert_eq!(stats.used, 1);
+        assert_ne!(stats.available, 0);
+        let available = stats.available;
+
+        // Verify that creating a directory and writing a file will result in increased utilization
+        let dir = mount_dir.path().join("foobar");
+        fs::create_dir_all(&dir).unwrap();
+        fs::write(dir.as_path().join("file.dat"), "foobar").unwrap();
+        stats = get_volume_inode_stats(mount_dir.path().to_str().unwrap()).unwrap();
+
+        assert_eq!(stats.used, 3);
+        assert_eq!(stats.available, available - 2);
+    }
+
+    #[tokio::test]
+    async fn test_ip_tables() {
+        skip_if_not_root!();
+
+        let logger = slog::Logger::root(slog::Discard, o!());
+        let sandbox = Sandbox::new(&logger).unwrap();
+        let agent_service = Box::new(AgentService {
+            sandbox: Arc::new(Mutex::new(sandbox)),
+        });
+
+        let ctx = mk_ttrpc_context();
+
+        // Move to a new netns in order to ensure we don't trash the hosts' iptables
+        unshare(CloneFlags::CLONE_NEWNET).unwrap();
+
+        // Get initial iptables, we expect to be empty:
+        let result = agent_service
+            .get_ip_tables(
+                &ctx,
+                GetIPTablesRequest {
+                    is_ipv6: false,
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_ok(), "get ip tables should succeed");
+        assert_eq!(
+            result.unwrap().data.len(),
+            0,
+            "ip tables should be empty initially"
+        );
+
+        // Initial ip6 ip tables should also be empty:
+        let result = agent_service
+            .get_ip_tables(
+                &ctx,
+                GetIPTablesRequest {
+                    is_ipv6: true,
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_ok(), "get ip6 tables should succeed");
+        assert_eq!(
+            result.unwrap().data.len(),
+            0,
+            "ip tables should be empty initially"
+        );
+
+        // Verify that attempting to write 'empty' iptables results in no error:
+        let empty_rules = "";
+        let result = agent_service
+            .set_ip_tables(
+                &ctx,
+                SetIPTablesRequest {
+                    is_ipv6: false,
+                    data: empty_rules.as_bytes().to_vec(),
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_ok(), "set ip tables with no data should succeed");
+
+        // Verify that attempting to write "garbage" iptables results in an error:
+        let garbage_rules = r#"
+this
+is
+just garbage
+"#;
+        let result = agent_service
+            .set_ip_tables(
+                &ctx,
+                SetIPTablesRequest {
+                    is_ipv6: false,
+                    data: garbage_rules.as_bytes().to_vec(),
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_err(), "set iptables with garbage should fail");
+
+        // Verify setup of valid iptables:Setup  valid set of iptables:
+        let valid_rules = r#"
+*nat
+-A PREROUTING -d 192.168.103.153/32 -j DNAT --to-destination 192.168.188.153
+
+COMMIT
+
+"#;
+        let result = agent_service
+            .set_ip_tables(
+                &ctx,
+                SetIPTablesRequest {
+                    is_ipv6: false,
+                    data: valid_rules.as_bytes().to_vec(),
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_ok(), "set ip tables should succeed");
+
+        let result = agent_service
+            .get_ip_tables(
+                &ctx,
+                GetIPTablesRequest {
+                    is_ipv6: false,
+                    ..Default::default()
+                },
+            )
+            .await
+            .unwrap();
+        assert!(!result.data.is_empty(), "we should have non-zero output:");
+        assert!(
+            std::str::from_utf8(&*result.data).unwrap().contains(
+                "PREROUTING -d 192.168.103.153/32 -j DNAT --to-destination 192.168.188.153"
+            ),
+            "We should see the resulting rule"
+        );
+
+        // Verify setup of valid ip6tables:
+        let valid_ipv6_rules = r#"
+*filter
+-A INPUT -s 2001:db8:100::1/128 -i sit+ -p tcp -m tcp --sport 512:65535
+
+COMMIT
+
+"#;
+        let result = agent_service
+            .set_ip_tables(
+                &ctx,
+                SetIPTablesRequest {
+                    is_ipv6: true,
+                    data: valid_ipv6_rules.as_bytes().to_vec(),
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(result.is_ok(), "set ip6 tables should succeed");
+
+        let result = agent_service
+            .get_ip_tables(
+                &ctx,
+                GetIPTablesRequest {
+                    is_ipv6: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .unwrap();
+        assert!(!result.data.is_empty(), "we should have non-zero output:");
+        assert!(
+            std::str::from_utf8(&*result.data)
+                .unwrap()
+                .contains("INPUT -s 2001:db8:100::1/128 -i sit+ -p tcp -m tcp --sport 512:65535"),
+            "We should see the resulting rule"
+        );
     }
 }

src/agent/src/sandbox.rs

@@ -470,7 +470,7 @@ fn online_memory(logger: &Logger) -> Result<()> {
 
 #[cfg(test)]
 mod tests {
-    use super::Sandbox;
+    use super::*;
     use crate::{mount::baremount, skip_if_not_root};
     use anyhow::{anyhow, Error};
     use nix::mount::MsFlags;
@@ -480,6 +480,7 @@ mod tests {
     use rustjail::specconv::CreateOpts;
     use slog::Logger;
     use std::fs::{self, File};
+    use std::io::prelude::*;
     use std::os::unix::fs::PermissionsExt;
     use std::path::Path;
     use tempfile::{tempdir, Builder, TempDir};
@@ -847,4 +848,259 @@ mod tests {
         let p = s.find_container_process("not-exist-cid", "");
         assert!(p.is_err(), "Expecting Error, Got {:?}", p);
     }
+
+    #[tokio::test]
+    async fn test_find_process() {
+        let logger = slog::Logger::root(slog::Discard, o!());
+
+        let test_pids = [std::i32::MIN, -1, 0, 1, std::i32::MAX];
+
+        for test_pid in test_pids {
+            let mut s = Sandbox::new(&logger).unwrap();
+            let (mut linux_container, _root) = create_linuxcontainer();
+
+            let mut test_process = Process::new(
+                &logger,
+                &oci::Process::default(),
+                "this_is_a_test_process",
+                true,
+                1,
+            )
+            .unwrap();
+            // processes interally only have pids when manually set
+            test_process.pid = test_pid;
+
+            linux_container.processes.insert(test_pid, test_process);
+
+            s.add_container(linux_container);
+
+            let find_result = s.find_process(test_pid);
+
+            // test first if it finds anything
+            assert!(find_result.is_some(), "Should be able to find a process");
+
+            let found_process = find_result.unwrap();
+
+            // then test if it founds the correct process
+            assert_eq!(
+                found_process.pid, test_pid,
+                "Should be able to find correct process"
+            );
+        }
+
+        // to test for nonexistent pids, any pid that isn't the one set
+        // above should work, as linuxcontainer starts with no processes
+        let mut s = Sandbox::new(&logger).unwrap();
+
+        let nonexistent_test_pid = 1234;
+
+        let find_result = s.find_process(nonexistent_test_pid);
+
+        assert!(
+            find_result.is_none(),
+            "Shouldn't find a process for non existent pid"
+        );
+    }
+
+    #[tokio::test]
+    async fn test_online_resources() {
+        #[derive(Debug, Default)]
+        struct TestFile {
+            name: String,
+            content: String,
+        }
+
+        #[derive(Debug, Default)]
+        struct TestDirectory<'a> {
+            name: String,
+            files: &'a [TestFile],
+        }
+
+        #[derive(Debug)]
+        struct TestData<'a> {
+            directory_autogen_name: String,
+            number_autogen_directories: u32,
+
+            extra_directories: &'a [TestDirectory<'a>],
+            pattern: String,
+            to_enable: i32,
+
+            result: Result<i32>,
+        }
+
+        impl Default for TestData<'_> {
+            fn default() -> Self {
+                TestData {
+                    directory_autogen_name: Default::default(),
+                    number_autogen_directories: Default::default(),
+                    extra_directories: Default::default(),
+                    pattern: Default::default(),
+                    to_enable: Default::default(),
+                    result: Ok(Default::default()),
+                }
+            }
+        }
+
+        let tests = &[
+            // 4 well formed directories, request enabled 4,
+            // correct result 4 enabled, should pass
+            TestData {
+                directory_autogen_name: String::from("cpu"),
+                number_autogen_directories: 4,
+                pattern: String::from(r"cpu[0-9]+"),
+                to_enable: 4,
+                result: Ok(4),
+                ..Default::default()
+            },
+            // 0 well formed directories, request enabled 4,
+            // correct result 0 enabled, should pass
+            TestData {
+                number_autogen_directories: 0,
+                to_enable: 4,
+                result: Ok(0),
+                ..Default::default()
+            },
+            // 10 well formed directories, request enabled 4,
+            // correct result 4 enabled, should pass
+            TestData {
+                directory_autogen_name: String::from("cpu"),
+                number_autogen_directories: 10,
+                pattern: String::from(r"cpu[0-9]+"),
+                to_enable: 4,
+                result: Ok(4),
+                ..Default::default()
+            },
+            // 0 well formed directories, request enabled 0,
+            // correct result 0 enabled, should pass
+            TestData {
+                number_autogen_directories: 0,
+                pattern: String::from(r"cpu[0-9]+"),
+                to_enable: 0,
+                result: Ok(0),
+                ..Default::default()
+            },
+            // 4 well formed directories, 1 malformed (no online file),
+            // request enable 5, correct result 4
+            TestData {
+                directory_autogen_name: String::from("cpu"),
+                number_autogen_directories: 4,
+                pattern: String::from(r"cpu[0-9]+"),
+                extra_directories: &[TestDirectory {
+                    name: String::from("cpu4"),
+                    files: &[],
+                }],
+                to_enable: 5,
+                result: Ok(4),
+            },
+            // 3 malformed directories (no online files),
+            // request enable 3, correct result 0
+            TestData {
+                pattern: String::from(r"cpu[0-9]+"),
+                extra_directories: &[
+                    TestDirectory {
+                        name: String::from("cpu0"),
+                        files: &[],
+                    },
+                    TestDirectory {
+                        name: String::from("cpu1"),
+                        files: &[],
+                    },
+                    TestDirectory {
+                        name: String::from("cpu2"),
+                        files: &[],
+                    },
+                ],
+                to_enable: 3,
+                result: Ok(0),
+                ..Default::default()
+            },
+            // 1 malformed directories (online file with content "1"),
+            // request enable 1, correct result 0
+            TestData {
+                pattern: String::from(r"cpu[0-9]+"),
+                extra_directories: &[TestDirectory {
+                    name: String::from("cpu0"),
+                    files: &[TestFile {
+                        name: SYSFS_ONLINE_FILE.to_string(),
+                        content: String::from("1"),
+                    }],
+                }],
+                to_enable: 1,
+                result: Ok(0),
+                ..Default::default()
+            },
+            // 2 well formed directories, 1 malformed (online file with content "1"),
+            // request enable 3, correct result 2
+            TestData {
+                directory_autogen_name: String::from("cpu"),
+                number_autogen_directories: 2,
+                pattern: String::from(r"cpu[0-9]+"),
+                extra_directories: &[TestDirectory {
+                    name: String::from("cpu2"),
+                    files: &[TestFile {
+                        name: SYSFS_ONLINE_FILE.to_string(),
+                        content: String::from("1"),
+                    }],
+                }],
+                to_enable: 3,
+                result: Ok(2),
+            },
+        ];
+
+        let logger = slog::Logger::root(slog::Discard, o!());
+        let tmpdir = Builder::new().tempdir().unwrap();
+        let tmpdir_path = tmpdir.path().to_str().unwrap();
+
+        for (i, d) in tests.iter().enumerate() {
+            let current_test_dir_path = format!("{}/test_{}", tmpdir_path, i);
+            fs::create_dir(&current_test_dir_path).unwrap();
+
+            // create numbered directories and fill using root name
+            for j in 0..d.number_autogen_directories {
+                let subdir_path = format!(
+                    "{}/{}{}",
+                    current_test_dir_path, d.directory_autogen_name, j
+                );
+                let subfile_path = format!("{}/{}", subdir_path, SYSFS_ONLINE_FILE);
+                fs::create_dir(&subdir_path).unwrap();
+                let mut subfile = File::create(subfile_path).unwrap();
+                subfile.write_all(b"0").unwrap();
+            }
+            // create extra directories and fill to specification
+            for j in d.extra_directories {
+                let subdir_path = format!("{}/{}", current_test_dir_path, j.name);
+                fs::create_dir(&subdir_path).unwrap();
+                for file in j.files {
+                    let subfile_path = format!("{}/{}", subdir_path, file.name);
+                    let mut subfile = File::create(&subfile_path).unwrap();
+                    subfile.write_all(file.content.as_bytes()).unwrap();
+                }
+            }
+
+            // run created directory structure against online_resources
+            let result = online_resources(&logger, &current_test_dir_path, &d.pattern, d.to_enable);
+
+            let mut msg = format!(
+                "test[{}]: {:?}, expected {}, actual {}",
+                i,
+                d,
+                d.result.is_ok(),
+                result.is_ok()
+            );
+
+            assert_eq!(result.is_ok(), d.result.is_ok(), "{}", msg);
+
+            if d.result.is_ok() {
+                let test_result_val = *d.result.as_ref().ok().unwrap();
+                let result_val = result.ok().unwrap();
+
+                msg = format!(
+                    "test[{}]: {:?}, expected {}, actual {}",
+                    i, d, test_result_val, result_val
+                );
+
+                assert_eq!(test_result_val, result_val, "{}", msg);
+            }
+        }
+    }
 }

src/agent/src/test_utils.rs

@@ -85,4 +85,15 @@ pub mod test_utils {
             }
         };
     }
+
+    #[macro_export]
+    macro_rules! skip_loop_by_user {
+        ($msg:expr, $user:expr) => {
+            if $user == TestUserType::RootOnly {
+                skip_loop_if_not_root!($msg);
+            } else if $user == TestUserType::NonRootOnly {
+                skip_loop_if_root!($msg);
+            }
+        };
+    }
 }

src/agent/src/tracer.rs

@@ -69,6 +69,8 @@ macro_rules! trace_rpc_call {
             propagator.extract(&extract_carrier_from_ttrpc($ctx))
         });
 
+        info!(sl!(), "rpc call from shim to agent: {:?}", $name);
+
         // generate tracing span
         let rpc_span = span!(tracing::Level::INFO, $name, "mod"="rpc.rs", req=?$req);
 

src/dragonball/.gitignore

@@ -0,0 +1,3 @@
+target
+Cargo.lock
+.idea

src/dragonball/Cargo.toml

@@ -0,0 +1,65 @@
+[package]
+name = "dragonball"
+version = "0.1.0"
+authors = ["The Kata Containers community <kata-dev@lists.katacontainers.io>"]
+description = "A secure sandbox for Kata Containers"
+keywords = ["kata-containers", "sandbox", "vmm", "dragonball"]
+homepage = "https://katacontainers.io/"
+repository = "https://github.com/kata-containers/kata-containers.git"
+license = "Apache-2.0"
+edition = "2018"
+
+[dependencies]
+arc-swap = "1.5.0"
+bytes = "1.1.0"
+dbs-address-space = "0.1.0"
+dbs-allocator = "0.1.0"
+dbs-arch = "0.1.0"
+dbs-boot = "0.2.0"
+dbs-device = "0.1.0"
+dbs-interrupt = { version = "0.1.0", features = ["kvm-irq"] }
+dbs-legacy-devices = "0.1.0"
+dbs-upcall = { version = "0.1.0", optional = true }
+dbs-utils = "0.1.0"
+dbs-virtio-devices = { version = "0.1.0", optional = true, features = ["virtio-mmio"] }
+kvm-bindings = "0.5.0"
+kvm-ioctls = "0.11.0"
+lazy_static = "1.2"
+libc = "0.2.39"
+linux-loader = "0.4.0"
+log = "0.4.14"
+nix = "0.23.1"
+seccompiler = "0.2.0"
+serde = "1.0.27"
+serde_derive = "1.0.27"
+serde_json = "1.0.9"
+slog = "2.5.2"
+slog-scope = "4.4.0"
+thiserror = "1"
+vmm-sys-util = "0.9.0"
+virtio-queue = { version = "0.1.0", optional = true }
+vm-memory = { version = "0.7.0", features = ["backend-mmap"] }
+
+[dev-dependencies]
+slog-term = "2.9.0"
+slog-async = "2.7.0"
+
+[features]
+acpi = []
+atomic-guest-memory = []
+hotplug = ["virtio-vsock"]
+virtio-vsock = ["dbs-virtio-devices/virtio-vsock", "virtio-queue"]
+virtio-blk = ["dbs-virtio-devices/virtio-blk", "virtio-queue"]
+virtio-net = ["dbs-virtio-devices/virtio-net", "virtio-queue"]
+# virtio-fs only work on atomic-guest-memory
+virtio-fs = ["dbs-virtio-devices/virtio-fs", "virtio-queue", "atomic-guest-memory"]
+
+[patch.'crates-io']
+dbs-device = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-interrupt = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-legacy-devices = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-upcall = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-utils = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-virtio-devices = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-boot = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }
+dbs-arch = { git = "https://github.com/openanolis/dragonball-sandbox.git", rev = "7a8e832b53d66994d6a16f0513d69f540583dcd0" }

src/dragonball/LICENSE

@@ -0,0 +1 @@
+../../LICENSE

src/dragonball/Makefile

@@ -0,0 +1,29 @@
+# Copyright (c) 2019-2022 Alibaba Cloud. All rights reserved.
+# Copyright (c) 2019-2022 Ant Group. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+default: build
+
+build:
+	# FIXME: This line will be removed when we solve the vm-memory dependency problem in Dragonball Sandbox
+	cargo update -p vm-memory:0.8.0 --precise 0.7.0
+	cargo build --all-features
+
+check: clippy format
+
+clippy:
+	@echo "INFO: cargo clippy..."
+	cargo clippy --all-targets --all-features \
+		-- \
+		-D warnings
+
+format:
+	@echo "INFO: cargo fmt..."
+	cargo fmt -- --check
+
+clean:
+	cargo clean
+
+test:
+	@echo "INFO: testing dragonball for development build"
+	cargo test --all-features -- --nocapture

src/dragonball/README.md

@@ -0,0 +1,40 @@
+# Introduction
+`Dragonball Sandbox` is a light-weight virtual machine manager (VMM) based on Linux Kernel-based Virtual Machine (KVM),
+which is optimized for container workloads with:
+- container image management and acceleration service
+- flexible and high-performance virtual device drivers
+- low CPU and memory overhead
+- minimal startup time
+- optimized concurrent startup speed
+
+`Dragonball Sandbox` aims to provide a simple solution for the Kata Containers community. It is integrated into Kata 3.0
+runtime as a built-in VMM and gives users an out-of-the-box Kata Containers experience without complex environment setup
+and configuration process.
+
+# Getting Started
+[TODO](https://github.com/kata-containers/kata-containers/issues/4302)
+
+# Documentation
+
+Device: [Device Document](docs/device.md)
+vCPU: [vCPU Document](docs/vcpu.md)
+API: [API Document](docs/api.md)
+
+Currently, the documents are still actively adding.
+You could see the [official documentation](docs/) page for more details.
+
+# Supported Architectures
+- x86-64
+- aarch64
+ 
+# Supported Kernel
+[TODO](https://github.com/kata-containers/kata-containers/issues/4303)
+
+# Acknowledgement
+Part of the code is based on the [Cloud Hypervisor](https://github.com/cloud-hypervisor/cloud-hypervisor) project, [`crosvm`](https://github.com/google/crosvm) project and [Firecracker](https://github.com/firecracker-microvm/firecracker) project. They are all rust written virtual machine managers with advantages on safety and security.
+
+`Dragonball sandbox` is designed to be a VMM that is customized for Kata Containers and we will focus on optimizing container workloads for Kata ecosystem. The focus on the Kata community is what differentiates us from other rust written virtual machines.
+
+# License
+
+`Dragonball` is licensed under [Apache License](http://www.apache.org/licenses/LICENSE-2.0), Version 2.0.

src/dragonball/THIRD-PARTY

@@ -0,0 +1,27 @@
+// Copyright 2017 The Chromium OS Authors. All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//    * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//    * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//    * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

src/dragonball/docs/api.md

@@ -0,0 +1,27 @@
+# API
+
+We provide plenty API for Kata runtime to interact with `Dragonball` virtual machine manager.
+This document provides the introduction for each of them.
+
+## `ConfigureBootSource`
+Configure the boot source of the VM using `BootSourceConfig`. This action can only be called before the VM has booted.
+
+### Boot Source Config
+1. `kernel_path`: Path of the kernel image. `Dragonball` only supports compressed kernel image for now.
+2. `initrd_path`: Path of the initrd (could be None)
+3. `boot_args`: Boot arguments passed to the kernel (could be None)
+
+## `SetVmConfiguration`
+Set virtual machine configuration using `VmConfigInfo` to initialize VM.
+
+### VM Config Info
+1. `vcpu_count`: Number of vCPU to start. Currently we only support up to 255 vCPUs.
+2. `max_vcpu_count`: Max number of vCPU can be added through CPU hotplug.
+3. `cpu_pm`: CPU power management.
+4. `cpu_topology`: CPU topology information (including `threads_per_core`, `cores_per_die`, `dies_per_socket` and `sockets`).
+5. `vpmu_feature`: `vPMU` feature level.
+6. `mem_type`: Memory type that can be either `hugetlbfs` or `shmem`, default is `shmem`.
+7. `mem_file_path` : Memory file path.
+8. `mem_size_mib`: The memory size in MiB. The maximum memory size is 1TB.
+9. `serial_path`: Optional sock path.
+

src/dragonball/docs/device.md

@@ -0,0 +1,20 @@
+# Device
+
+## Device Manager
+
+Currently we have following device manager:
+| Name | Description | 
+| --- | --- |
+| [address space manager](../src/address_space_manager.rs) | abstracts virtual machine's physical management and provide mapping for guest virtual memory and MMIO ranges of emulated virtual devices, pass-through devices and vCPU |
+| [config manager](../src/config_manager.rs) | provides abstractions for configuration information | 
+| [console manager](../src/device_manager/console_manager.rs) | provides management for all console devices | 
+| [resource manager](../src/resource_manager.rs) |provides resource management for `legacy_irq_pool`, `msi_irq_pool`, `pio_pool`, `mmio_pool`, `mem_pool`, `kvm_mem_slot_pool` with builder `ResourceManagerBuilder` | 
+| [VSOCK device manager](../src/device_manager/vsock_dev_mgr.rs) | provides configuration info for `VIRTIO-VSOCK` and management for all VSOCK devices | 
+   
+
+## Device supported
+`VIRTIO-VSOCK`
+`i8042`
+`COM1`
+`COM2`
+

src/dragonball/docs/vcpu.md

@@ -0,0 +1,42 @@
+# vCPU
+
+## vCPU Manager
+The vCPU manager is to manage all vCPU related actions, we will dive into some of the important structure members in this doc.
+
+For now, aarch64 vCPU support is still under development, we'll introduce it when we merge `runtime-rs` to the master branch. (issue: #4445)
+
+### vCPU config
+`VcpuConfig` is used to configure guest overall CPU info.
+
+`boot_vcpu_count` is used to define the initial vCPU number.
+
+`max_vcpu_count` is used to define the maximum vCPU number and it's used for the upper boundary for CPU hotplug feature
+
+`thread_per_core`, `cores_per_die`, `dies_per_socket` and `socket` are used to define CPU topology.
+
+`vpmu_feature` is used to define `vPMU` feature level.
+If `vPMU` feature is `Disabled`, it means `vPMU` feature is off (by default).
+If `vPMU` feature is `LimitedlyEnabled`, it means minimal `vPMU` counters are supported (cycles and instructions).
+If `vPMU` feature is `FullyEnabled`, it means all `vPMU` counters are supported
+
+## vCPU State
+
+There are four states for vCPU state machine: `running`, `paused`, `waiting_exit`, `exited`. There is a state machine to maintain the task flow.
+
+When the vCPU is created, it'll turn to `paused` state. After vCPU resource is ready at VMM, it'll send a `Resume` event to the vCPU thread, and then vCPU state will change to `running`.
+
+During the `running` state, VMM will catch vCPU exit and execute different logic according to the exit reason.
+
+If the VMM catch some exit reasons that it cannot handle, the state will change to `waiting_exit` and VMM will stop the virtual machine. 
+When the state switches to `waiting_exit`, an exit event will be sent to vCPU `exit_evt`, event manager will detect the change in `exit_evt` and set VMM `exit_evt_flag` as 1. A thread serving for VMM event loop will check `exit_evt_flag` and if the flag is 1, it'll stop the VMM.
+
+When the VMM is stopped / destroyed, the state will change to `exited`.
+   
+## vCPU Hot plug
+Since `Dragonball Sandbox` doesn't support virtualization of ACPI system, we use [`upcall`](https://github.com/openanolis/dragonball-sandbox/tree/main/crates/dbs-upcall) to establish a direct communication channel between `Dragonball` and Guest in order to trigger vCPU hotplug.
+
+To use `upcall`, kernel patches are needed, you can get the patches from [`upcall`](https://github.com/openanolis/dragonball-sandbox/tree/main/crates/dbs-upcall) page, and we'll provide a ready-to-use guest kernel binary for you to try.
+
+vCPU hot plug / hot unplug range is [1, `max_vcpu_count`]. Operations not in this range will be invalid.
+
+

src/dragonball/src/address_space_manager.rs

@@ -0,0 +1,892 @@
+// Copyright (C) 2019-2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+//! Address space abstraction to manage virtual machine's physical address space.
+//!
+//! The AddressSpace abstraction is introduced to manage virtual machine's physical address space.
+//! The regions in virtual machine's physical address space may be used to:
+//! 1) map guest virtual memory
+//! 2) map MMIO ranges for emulated virtual devices, such as virtio-fs DAX window.
+//! 3) map MMIO ranges for pass-through devices, such as PCI device BARs.
+//! 4) map MMIO ranges for to vCPU, such as local APIC.
+//! 5) not used/available
+//!
+//! A related abstraction, vm_memory::GuestMemory, is used to access guest virtual memory only.
+//! In other words, AddressSpace is the resource owner, and GuestMemory is an accessor for guest
+//! virtual memory.
+
+use std::collections::{BTreeMap, HashMap};
+use std::fs::File;
+use std::os::unix::io::{AsRawFd, FromRawFd};
+use std::sync::atomic::{AtomicBool, AtomicU8, Ordering};
+use std::sync::{Arc, Mutex};
+use std::thread;
+
+use dbs_address_space::{
+    AddressSpace, AddressSpaceError, AddressSpaceLayout, AddressSpaceRegion,
+    AddressSpaceRegionType, NumaNode, NumaNodeInfo, MPOL_MF_MOVE, MPOL_PREFERRED,
+};
+use dbs_allocator::Constraint;
+use kvm_bindings::kvm_userspace_memory_region;
+use kvm_ioctls::VmFd;
+use log::{debug, error, info, warn};
+use nix::sys::mman;
+use nix::unistd::dup;
+#[cfg(feature = "atomic-guest-memory")]
+use vm_memory::atomic::GuestMemoryAtomic;
+use vm_memory::{
+    Address, FileOffset, GuestAddress, GuestAddressSpace, GuestMemoryMmap, GuestMemoryRegion,
+    GuestRegionMmap, GuestUsize, MemoryRegionAddress, MmapRegion,
+};
+
+use crate::resource_manager::ResourceManager;
+use crate::vm::NumaRegionInfo;
+
+#[cfg(not(feature = "atomic-guest-memory"))]
+/// Concrete GuestAddressSpace type used by the VMM.
+pub type GuestAddressSpaceImpl = Arc<GuestMemoryMmap>;
+
+#[cfg(feature = "atomic-guest-memory")]
+/// Concrete GuestAddressSpace type used by the VMM.
+pub type GuestAddressSpaceImpl = GuestMemoryAtomic<GuestMemoryMmap>;
+
+/// Concrete GuestMemory type used by the VMM.
+pub type GuestMemoryImpl = <Arc<vm_memory::GuestMemoryMmap> as GuestAddressSpace>::M;
+/// Concrete GuestRegion type used by the VMM.
+pub type GuestRegionImpl = GuestRegionMmap;
+
+// Maximum number of working threads for memory pre-allocation.
+const MAX_PRE_ALLOC_THREAD: u64 = 16;
+
+// Control the actual number of pre-allocating threads. After several performance tests, we decide to use one thread to do pre-allocating for every 4G memory.
+const PRE_ALLOC_GRANULARITY: u64 = 32;
+
+// We don't have plan to support mainframe computer and only focus on PC servers.
+// 64 as max nodes should be enough for now.
+const MAX_NODE: u32 = 64;
+
+// We will split the memory region if it conflicts with the MMIO hole.
+// But if the space below the MMIO hole is smaller than the MINIMAL_SPLIT_SPACE, we won't split the memory region in order to enhance performance.
+const MINIMAL_SPLIT_SPACE: u64 = 128 << 20;
+
+/// Errors associated with virtual machine address space management.
+#[derive(Debug, thiserror::Error)]
+pub enum AddressManagerError {
+    /// Invalid address space operation.
+    #[error("invalid address space operation")]
+    InvalidOperation,
+
+    /// Invalid address range.
+    #[error("invalid address space region (0x{0:x}, 0x{1:x})")]
+    InvalidAddressRange(u64, GuestUsize),
+
+    /// No available mem address.
+    #[error("no available mem address")]
+    NoAvailableMemAddress,
+
+    /// No available kvm slotse.
+    #[error("no available kvm slots")]
+    NoAvailableKvmSlot,
+
+    /// Address manager failed to create memfd to map anonymous memory.
+    #[error("address manager failed to create memfd to map anonymous memory")]
+    CreateMemFd(#[source] nix::Error),
+
+    /// Address manager failed to open memory file.
+    #[error("address manager failed to open memory file")]
+    OpenFile(#[source] std::io::Error),
+
+    /// Memory file provided is invalid due to empty file path, non-existent file path and other possible mistakes.
+    #[error("memory file provided to address manager {0} is invalid")]
+    FileInvalid(String),
+
+    /// Memory file provided is invalid due to empty memory type
+    #[error("memory type provided to address manager {0} is invalid")]
+    TypeInvalid(String),
+
+    /// Failed to set size for memory file.
+    #[error("address manager failed to set size for memory file")]
+    SetFileSize(#[source] std::io::Error),
+
+    /// Failed to unlink memory file.
+    #[error("address manager failed to unlink memory file")]
+    UnlinkFile(#[source] nix::Error),
+
+    /// Failed to duplicate fd of memory file.
+    #[error("address manager failed to duplicate memory file descriptor")]
+    DupFd(#[source] nix::Error),
+
+    /// Failure in accessing the memory located at some address.
+    #[error("address manager failed to access guest memory located at 0x{0:x}")]
+    AccessGuestMemory(u64, #[source] vm_memory::mmap::Error),
+
+    /// Failed to create GuestMemory
+    #[error("address manager failed to create guest memory object")]
+    CreateGuestMemory(#[source] vm_memory::Error),
+
+    /// Failure in initializing guest memory.
+    #[error("address manager failed to initialize guest memory")]
+    GuestMemoryNotInitialized,
+
+    /// Failed to mmap() guest memory
+    #[error("address manager failed to mmap() guest memory into current process")]
+    MmapGuestMemory(#[source] vm_memory::mmap::MmapRegionError),
+
+    /// Failed to set KVM memory slot.
+    #[error("address manager failed to configure KVM memory slot")]
+    KvmSetMemorySlot(#[source] kvm_ioctls::Error),
+
+    /// Failed to set madvise on AddressSpaceRegion
+    #[error("address manager failed to set madvice() on guest memory region")]
+    Madvise(#[source] nix::Error),
+
+    /// join threads fail
+    #[error("address manager failed to join threads")]
+    JoinFail,
+
+    /// Failed to create Address Space Region
+    #[error("address manager failed to create Address Space Region {0}")]
+    CreateAddressSpaceRegion(#[source] AddressSpaceError),
+}
+
+type Result<T> = std::result::Result<T, AddressManagerError>;
+
+/// Parameters to configure address space creation operations.
+pub struct AddressSpaceMgrBuilder<'a> {
+    mem_type: &'a str,
+    mem_file: &'a str,
+    mem_index: u32,
+    mem_suffix: bool,
+    mem_prealloc: bool,
+    dirty_page_logging: bool,
+    vmfd: Option<Arc<VmFd>>,
+}
+
+impl<'a> AddressSpaceMgrBuilder<'a> {
+    /// Create a new [`AddressSpaceMgrBuilder`] object.
+    pub fn new(mem_type: &'a str, mem_file: &'a str) -> Result<Self> {
+        if mem_type.is_empty() {
+            return Err(AddressManagerError::TypeInvalid(mem_type.to_string()));
+        }
+        Ok(AddressSpaceMgrBuilder {
+            mem_type,
+            mem_file,
+            mem_index: 0,
+            mem_suffix: true,
+            mem_prealloc: false,
+            dirty_page_logging: false,
+            vmfd: None,
+        })
+    }
+
+    /// Enable/disable adding numbered suffix to memory file path.
+    /// This feature could be useful to generate hugetlbfs files with number suffix. (e.g. shmem0, shmem1)
+    pub fn toggle_file_suffix(&mut self, enabled: bool) {
+        self.mem_suffix = enabled;
+    }
+
+    /// Enable/disable memory pre-allocation.
+    /// Enable this feature could improve performance stability at the start of workload by avoiding page fault.
+    /// Disable this feature may influence performance stability but the cpu resource consumption and start-up time will decrease.
+    pub fn toggle_prealloc(&mut self, prealloc: bool) {
+        self.mem_prealloc = prealloc;
+    }
+
+    /// Enable/disable KVM dirty page logging.
+    pub fn toggle_dirty_page_logging(&mut self, logging: bool) {
+        self.dirty_page_logging = logging;
+    }
+
+    /// Set KVM [`VmFd`] handle to configure memory slots.
+    pub fn set_kvm_vm_fd(&mut self, vmfd: Arc<VmFd>) -> Option<Arc<VmFd>> {
+        let mut existing_vmfd = None;
+        if self.vmfd.is_some() {
+            existing_vmfd = self.vmfd.clone();
+        }
+        self.vmfd = Some(vmfd);
+        existing_vmfd
+    }
+
+    /// Build a ['AddressSpaceMgr'] using the configured parameters.
+    pub fn build(
+        self,
+        res_mgr: &ResourceManager,
+        numa_region_infos: &[NumaRegionInfo],
+    ) -> Result<AddressSpaceMgr> {
+        let mut mgr = AddressSpaceMgr::default();
+        mgr.create_address_space(res_mgr, numa_region_infos, self)?;
+        Ok(mgr)
+    }
+
+    fn get_next_mem_file(&mut self) -> String {
+        if self.mem_suffix {
+            let path = format!("{}{}", self.mem_file, self.mem_index);
+            self.mem_index += 1;
+            path
+        } else {
+            self.mem_file.to_string()
+        }
+    }
+}
+
+/// Struct to manage virtual machine's physical address space.
+pub struct AddressSpaceMgr {
+    address_space: Option<AddressSpace>,
+    vm_as: Option<GuestAddressSpaceImpl>,
+    base_to_slot: Arc<Mutex<HashMap<u64, u32>>>,
+    prealloc_handlers: Vec<thread::JoinHandle<()>>,
+    prealloc_exit: Arc<AtomicBool>,
+    numa_nodes: BTreeMap<u32, NumaNode>,
+}
+
+impl AddressSpaceMgr {
+    /// Query address space manager is initialized or not
+    pub fn is_initialized(&self) -> bool {
+        self.address_space.is_some()
+    }
+
+    /// Gets address space.
+    pub fn address_space(&self) -> Option<&AddressSpace> {
+        self.address_space.as_ref()
+    }
+
+    /// Create the address space for a virtual machine.
+    ///
+    /// This method is designed to be called when starting up a virtual machine instead of at
+    /// runtime, so it's expected the virtual machine will be tore down and no strict error recover.
+    pub fn create_address_space(
+        &mut self,
+        res_mgr: &ResourceManager,
+        numa_region_infos: &[NumaRegionInfo],
+        mut param: AddressSpaceMgrBuilder,
+    ) -> Result<()> {
+        let mut regions = Vec::new();
+        let mut start_addr = dbs_boot::layout::GUEST_MEM_START;
+
+        // Create address space regions.
+        for info in numa_region_infos.iter() {
+            info!("numa_region_info {:?}", info);
+            // convert size_in_mib to bytes
+            let size = info
+                .size
+                .checked_shl(20)
+                .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+
+            // Guest memory does not intersect with the MMIO hole.
+            // TODO: make it work for ARM (issue #4307)
+            if start_addr > dbs_boot::layout::MMIO_LOW_END
+                || start_addr + size <= dbs_boot::layout::MMIO_LOW_START
+            {
+                let region = self.create_region(start_addr, size, info, &mut param)?;
+                regions.push(region);
+                start_addr = start_addr
+                    .checked_add(size)
+                    .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+            } else {
+                // Add guest memory below the MMIO hole, avoid splitting the memory region
+                // if the available address region is small than MINIMAL_SPLIT_SPACE MiB.
+                let mut below_size = dbs_boot::layout::MMIO_LOW_START
+                    .checked_sub(start_addr)
+                    .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+                if below_size < (MINIMAL_SPLIT_SPACE) {
+                    below_size = 0;
+                } else {
+                    let region = self.create_region(start_addr, below_size, info, &mut param)?;
+                    regions.push(region);
+                }
+
+                // Add guest memory above the MMIO hole
+                let above_start = dbs_boot::layout::MMIO_LOW_END + 1;
+                let above_size = size
+                    .checked_sub(below_size)
+                    .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+                let region = self.create_region(above_start, above_size, info, &mut param)?;
+                regions.push(region);
+                start_addr = above_start
+                    .checked_add(above_size)
+                    .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+            }
+        }
+
+        // Create GuestMemory object
+        let mut vm_memory = GuestMemoryMmap::new();
+        for reg in regions.iter() {
+            // Allocate used guest memory addresses.
+            // These addresses are statically allocated, resource allocation/update should not fail.
+            let constraint = Constraint::new(reg.len())
+                .min(reg.start_addr().raw_value())
+                .max(reg.last_addr().raw_value());
+            let _key = res_mgr
+                .allocate_mem_address(&constraint)
+                .ok_or(AddressManagerError::NoAvailableMemAddress)?;
+            let mmap_reg = self.create_mmap_region(reg.clone())?;
+
+            vm_memory = vm_memory
+                .insert_region(mmap_reg.clone())
+                .map_err(AddressManagerError::CreateGuestMemory)?;
+            self.map_to_kvm(res_mgr, &param, reg, mmap_reg)?;
+        }
+
+        #[cfg(feature = "atomic-guest-memory")]
+        {
+            self.vm_as = Some(AddressSpace::convert_into_vm_as(vm_memory));
+        }
+        #[cfg(not(feature = "atomic-guest-memory"))]
+        {
+            self.vm_as = Some(Arc::new(vm_memory));
+        }
+
+        let layout = AddressSpaceLayout::new(
+            *dbs_boot::layout::GUEST_PHYS_END,
+            dbs_boot::layout::GUEST_MEM_START,
+            *dbs_boot::layout::GUEST_MEM_END,
+        );
+        self.address_space = Some(AddressSpace::from_regions(regions, layout));
+
+        Ok(())
+    }
+
+    // size unit: Byte
+    fn create_region(
+        &mut self,
+        start_addr: u64,
+        size_bytes: u64,
+        info: &NumaRegionInfo,
+        param: &mut AddressSpaceMgrBuilder,
+    ) -> Result<Arc<AddressSpaceRegion>> {
+        let mem_file_path = param.get_next_mem_file();
+        let region = AddressSpaceRegion::create_default_memory_region(
+            GuestAddress(start_addr),
+            size_bytes,
+            info.host_numa_node_id,
+            param.mem_type,
+            &mem_file_path,
+            param.mem_prealloc,
+            false,
+        )
+        .map_err(AddressManagerError::CreateAddressSpaceRegion)?;
+        let region = Arc::new(region);
+
+        self.insert_into_numa_nodes(
+            &region,
+            info.guest_numa_node_id.unwrap_or(0),
+            &info.vcpu_ids,
+        );
+        info!(
+            "create new region: guest addr 0x{:x}-0x{:x} size {}",
+            start_addr,
+            start_addr + size_bytes,
+            size_bytes
+        );
+
+        Ok(region)
+    }
+
+    fn map_to_kvm(
+        &mut self,
+        res_mgr: &ResourceManager,
+        param: &AddressSpaceMgrBuilder,
+        reg: &Arc<AddressSpaceRegion>,
+        mmap_reg: Arc<GuestRegionImpl>,
+    ) -> Result<()> {
+        // Build mapping between GPA <-> HVA, by adding kvm memory slot.
+        let slot = res_mgr
+            .allocate_kvm_mem_slot(1, None)
+            .ok_or(AddressManagerError::NoAvailableKvmSlot)?;
+
+        if let Some(vmfd) = param.vmfd.as_ref() {
+            let host_addr = mmap_reg
+                .get_host_address(MemoryRegionAddress(0))
+                .map_err(|_e| AddressManagerError::InvalidOperation)?;
+            let flags = 0u32;
+
+            let mem_region = kvm_userspace_memory_region {
+                slot: slot as u32,
+                guest_phys_addr: reg.start_addr().raw_value(),
+                memory_size: reg.len() as u64,
+                userspace_addr: host_addr as u64,
+                flags,
+            };
+
+            info!(
+                "VM: guest memory region {:x} starts at {:x?}",
+                reg.start_addr().raw_value(),
+                host_addr
+            );
+            // Safe because the guest regions are guaranteed not to overlap.
+            unsafe { vmfd.set_user_memory_region(mem_region) }
+                .map_err(AddressManagerError::KvmSetMemorySlot)?;
+        }
+
+        self.base_to_slot
+            .lock()
+            .unwrap()
+            .insert(reg.start_addr().raw_value(), slot as u32);
+
+        Ok(())
+    }
+
+    /// Mmap the address space region into current process.
+    pub fn create_mmap_region(
+        &mut self,
+        region: Arc<AddressSpaceRegion>,
+    ) -> Result<Arc<GuestRegionImpl>> {
+        // Special check for 32bit host with 64bit virtual machines.
+        if region.len() > usize::MAX as u64 {
+            return Err(AddressManagerError::InvalidAddressRange(
+                region.start_addr().raw_value(),
+                region.len(),
+            ));
+        }
+        // The device MMIO regions may not be backed by memory files, so refuse to mmap them.
+        if region.region_type() == AddressSpaceRegionType::DeviceMemory {
+            return Err(AddressManagerError::InvalidOperation);
+        }
+
+        // The GuestRegionMmap/MmapRegion will take ownership of the FileOffset object,
+        // so we have to duplicate the fd here. It's really a dirty design.
+        let file_offset = match region.file_offset().as_ref() {
+            Some(fo) => {
+                let fd = dup(fo.file().as_raw_fd()).map_err(AddressManagerError::DupFd)?;
+                // Safe because we have just duplicated the raw fd.
+                let file = unsafe { File::from_raw_fd(fd) };
+                let file_offset = FileOffset::new(file, fo.start());
+                Some(file_offset)
+            }
+            None => None,
+        };
+        let perm_flags = if (region.perm_flags() & libc::MAP_POPULATE) != 0 && region.is_hugepage()
+        {
+            // mmap(MAP_POPULATE) conflicts with madive(MADV_HUGEPAGE) because mmap(MAP_POPULATE)
+            // will pre-fault in all memory with normal pages before madive(MADV_HUGEPAGE) gets
+            // called. So remove the MAP_POPULATE flag and memory will be faulted in by working
+            // threads.
+            region.perm_flags() & (!libc::MAP_POPULATE)
+        } else {
+            region.perm_flags()
+        };
+        let mmap_reg = MmapRegion::build(
+            file_offset,
+            region.len() as usize,
+            libc::PROT_READ | libc::PROT_WRITE,
+            perm_flags,
+        )
+        .map_err(AddressManagerError::MmapGuestMemory)?;
+
+        if region.is_anonpage() {
+            self.configure_anon_mem(&mmap_reg)?;
+        }
+        if let Some(node_id) = region.host_numa_node_id() {
+            self.configure_numa(&mmap_reg, node_id)?;
+        }
+        if region.is_hugepage() {
+            self.configure_thp_and_prealloc(&region, &mmap_reg)?;
+        }
+
+        let reg = GuestRegionImpl::new(mmap_reg, region.start_addr())
+            .map_err(AddressManagerError::CreateGuestMemory)?;
+        Ok(Arc::new(reg))
+    }
+
+    fn configure_anon_mem(&self, mmap_reg: &MmapRegion) -> Result<()> {
+        unsafe {
+            mman::madvise(
+                mmap_reg.as_ptr() as *mut libc::c_void,
+                mmap_reg.size(),
+                mman::MmapAdvise::MADV_DONTFORK,
+            )
+        }
+        .map_err(AddressManagerError::Madvise)
+    }
+
+    fn configure_numa(&self, mmap_reg: &MmapRegion, node_id: u32) -> Result<()> {
+        let nodemask = 1_u64
+            .checked_shl(node_id)
+            .ok_or_else(|| AddressManagerError::InvalidOperation)?;
+        let res = unsafe {
+            libc::syscall(
+                libc::SYS_mbind,
+                mmap_reg.as_ptr() as *mut libc::c_void,
+                mmap_reg.size(),
+                MPOL_PREFERRED,
+                &nodemask as *const u64,
+                MAX_NODE,
+                MPOL_MF_MOVE,
+            )
+        };
+        if res < 0 {
+            warn!(
+                "failed to mbind memory to host_numa_node_id {}: this may affect performance",
+                node_id
+            );
+        }
+        Ok(())
+    }
+
+    // We set Transparent Huge Page (THP) through mmap to increase performance.
+    // In order to reduce the impact of page fault on performance, we start several threads (up to MAX_PRE_ALLOC_THREAD) to touch every 4k page of the memory region to manually do memory pre-allocation.
+    // The reason why we don't use mmap to enable THP and pre-alloction is that THP setting won't take effect in this operation (tested in kernel 4.9)
+    fn configure_thp_and_prealloc(
+        &mut self,
+        region: &Arc<AddressSpaceRegion>,
+        mmap_reg: &MmapRegion,
+    ) -> Result<()> {
+        debug!(
+            "Setting MADV_HUGEPAGE on AddressSpaceRegion addr {:x?} len {:x?}",
+            mmap_reg.as_ptr(),
+            mmap_reg.size()
+        );
+
+        // Safe because we just create the MmapRegion
+        unsafe {
+            mman::madvise(
+                mmap_reg.as_ptr() as *mut libc::c_void,
+                mmap_reg.size(),
+                mman::MmapAdvise::MADV_HUGEPAGE,
+            )
+        }
+        .map_err(AddressManagerError::Madvise)?;
+
+        if region.perm_flags() & libc::MAP_POPULATE > 0 {
+            // Touch every 4k page to trigger allocation. The step is 4K instead of 2M to ensure
+            // pre-allocation when running out of huge pages.
+            const PAGE_SIZE: u64 = 4096;
+            const PAGE_SHIFT: u32 = 12;
+            let addr = mmap_reg.as_ptr() as u64;
+            // Here we use >> PAGE_SHIFT to calculate how many 4K pages in the memory region.
+            let npage = (mmap_reg.size() as u64) >> PAGE_SHIFT;
+
+            let mut touch_thread = ((mmap_reg.size() as u64) >> PRE_ALLOC_GRANULARITY) + 1;
+            if touch_thread > MAX_PRE_ALLOC_THREAD {
+                touch_thread = MAX_PRE_ALLOC_THREAD;
+            }
+
+            let per_npage = npage / touch_thread;
+            for n in 0..touch_thread {
+                let start_npage = per_npage * n;
+                let end_npage = if n == (touch_thread - 1) {
+                    npage
+                } else {
+                    per_npage * (n + 1)
+                };
+                let mut per_addr = addr + (start_npage * PAGE_SIZE);
+                let should_stop = self.prealloc_exit.clone();
+
+                let handler = thread::Builder::new()
+                    .name("PreallocThread".to_string())
+                    .spawn(move || {
+                        info!("PreallocThread start start_npage: {:?}, end_npage: {:?}, per_addr: {:?}, thread_number: {:?}",
+                              start_npage, end_npage, per_addr, touch_thread );
+                        for _ in start_npage..end_npage {
+                            if should_stop.load(Ordering::Acquire) {
+                                info!("PreallocThread stop start_npage: {:?}, end_npage: {:?}, per_addr: {:?}, thread_number: {:?}",
+                                      start_npage, end_npage, per_addr, touch_thread);
+                                break;
+                            }
+
+                            // Reading from a THP page may be served by the zero page, so only
+                            // write operation could ensure THP memory allocation. So use
+                            // the compare_exchange(old_val, old_val) trick to trigger allocation.
+                            let addr_ptr = per_addr as *mut u8;
+                            let read_byte = unsafe { std::ptr::read_volatile(addr_ptr) };
+                            let atomic_u8 : &AtomicU8 = unsafe {&*(addr_ptr as *mut AtomicU8)};
+                            let _ = atomic_u8.compare_exchange(read_byte, read_byte, Ordering::SeqCst, Ordering::SeqCst);
+                            per_addr += PAGE_SIZE;
+                        }
+
+                        info!("PreallocThread done start_npage: {:?}, end_npage: {:?}, per_addr: {:?}, thread_number: {:?}",
+                              start_npage, end_npage, per_addr, touch_thread );
+                    });
+
+                match handler {
+                    Err(e) => error!(
+                        "Failed to create working thread for async pre-allocation, {:?}. This may affect performance stability at the start of the workload.",
+                        e
+                    ),
+                    Ok(hdl) => self.prealloc_handlers.push(hdl),
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Get the address space object
+    pub fn get_address_space(&self) -> Option<&AddressSpace> {
+        self.address_space.as_ref()
+    }
+
+    /// Get the default guest memory object, which will be used to access virtual machine's default
+    /// guest memory.
+    pub fn get_vm_as(&self) -> Option<&GuestAddressSpaceImpl> {
+        self.vm_as.as_ref()
+    }
+
+    /// Get the base to slot map
+    pub fn get_base_to_slot_map(&self) -> Arc<Mutex<HashMap<u64, u32>>> {
+        self.base_to_slot.clone()
+    }
+
+    /// get numa nodes infos from address space manager.
+    pub fn get_numa_nodes(&self) -> &BTreeMap<u32, NumaNode> {
+        &self.numa_nodes
+    }
+
+    /// add cpu and memory numa informations to BtreeMap
+    fn insert_into_numa_nodes(
+        &mut self,
+        region: &Arc<AddressSpaceRegion>,
+        guest_numa_node_id: u32,
+        vcpu_ids: &[u32],
+    ) {
+        let node = self
+            .numa_nodes
+            .entry(guest_numa_node_id)
+            .or_insert_with(NumaNode::new);
+        node.add_info(&NumaNodeInfo {
+            base: region.start_addr(),
+            size: region.len(),
+        });
+        node.add_vcpu_ids(vcpu_ids);
+    }
+
+    /// get address space layout from address space manager.
+    pub fn get_layout(&self) -> Result<AddressSpaceLayout> {
+        self.address_space
+            .as_ref()
+            .map(|v| v.layout())
+            .ok_or(AddressManagerError::GuestMemoryNotInitialized)
+    }
+
+    /// Wait for the pre-allocation working threads to finish work.
+    ///
+    /// Force all working threads to exit if `stop` is true.
+    pub fn wait_prealloc(&mut self, stop: bool) -> Result<()> {
+        if stop {
+            self.prealloc_exit.store(true, Ordering::Release);
+        }
+        while let Some(handlers) = self.prealloc_handlers.pop() {
+            if let Err(e) = handlers.join() {
+                error!("wait_prealloc join fail {:?}", e);
+                return Err(AddressManagerError::JoinFail);
+            }
+        }
+        Ok(())
+    }
+}
+
+impl Default for AddressSpaceMgr {
+    /// Create a new empty AddressSpaceMgr
+    fn default() -> Self {
+        AddressSpaceMgr {
+            address_space: None,
+            vm_as: None,
+            base_to_slot: Arc::new(Mutex::new(HashMap::new())),
+            prealloc_handlers: Vec::new(),
+            prealloc_exit: Arc::new(AtomicBool::new(false)),
+            numa_nodes: BTreeMap::new(),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use dbs_boot::layout::GUEST_MEM_START;
+    use std::ops::Deref;
+
+    use vm_memory::{Bytes, GuestAddressSpace, GuestMemory, GuestMemoryRegion};
+    use vmm_sys_util::tempfile::TempFile;
+
+    use super::*;
+
+    #[test]
+    fn test_create_address_space() {
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = 128 << 20;
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: vec![1, 2],
+        }];
+        let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+        let as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        let vm_as = as_mgr.get_vm_as().unwrap();
+        let guard = vm_as.memory();
+        let gmem = guard.deref();
+        assert_eq!(gmem.num_regions(), 1);
+
+        let reg = gmem
+            .find_region(GuestAddress(GUEST_MEM_START + mem_size - 1))
+            .unwrap();
+        assert_eq!(reg.start_addr(), GuestAddress(GUEST_MEM_START));
+        assert_eq!(reg.len(), mem_size);
+        assert!(gmem
+            .find_region(GuestAddress(GUEST_MEM_START + mem_size))
+            .is_none());
+        assert!(reg.file_offset().is_some());
+
+        let buf = [0x1u8, 0x2u8, 0x3u8, 0x4u8, 0x5u8];
+        gmem.write_slice(&buf, GuestAddress(GUEST_MEM_START))
+            .unwrap();
+
+        // Update middle of mapped memory region
+        let mut val = 0xa5u8;
+        gmem.write_obj(val, GuestAddress(GUEST_MEM_START + 0x1))
+            .unwrap();
+        val = gmem.read_obj(GuestAddress(GUEST_MEM_START + 0x1)).unwrap();
+        assert_eq!(val, 0xa5);
+        val = gmem.read_obj(GuestAddress(GUEST_MEM_START)).unwrap();
+        assert_eq!(val, 1);
+        val = gmem.read_obj(GuestAddress(GUEST_MEM_START + 0x2)).unwrap();
+        assert_eq!(val, 3);
+        val = gmem.read_obj(GuestAddress(GUEST_MEM_START + 0x5)).unwrap();
+        assert_eq!(val, 0);
+
+        // Read ahead of mapped memory region
+        assert!(gmem
+            .read_obj::<u8>(GuestAddress(GUEST_MEM_START + mem_size))
+            .is_err());
+
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = dbs_boot::layout::MMIO_LOW_START + (1 << 30);
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: vec![1, 2],
+        }];
+        let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+        let as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        let vm_as = as_mgr.get_vm_as().unwrap();
+        let guard = vm_as.memory();
+        let gmem = guard.deref();
+        #[cfg(target_arch = "x86_64")]
+        assert_eq!(gmem.num_regions(), 2);
+        #[cfg(target_arch = "aarch64")]
+        assert_eq!(gmem.num_regions(), 1);
+
+        // Test dropping GuestMemoryMmap object releases all resources.
+        for _ in 0..10000 {
+            let res_mgr = ResourceManager::new(None);
+            let mem_size = 1 << 20;
+            let numa_region_infos = vec![NumaRegionInfo {
+                size: mem_size >> 20,
+                host_numa_node_id: None,
+                guest_numa_node_id: Some(0),
+                vcpu_ids: vec![1, 2],
+            }];
+            let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+            let _as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        }
+        let file = TempFile::new().unwrap().into_file();
+        let fd = file.as_raw_fd();
+        // fd should be small enough if there's no leaking of fds.
+        assert!(fd < 1000);
+    }
+
+    #[test]
+    fn test_address_space_mgr_get_boundary() {
+        let layout = AddressSpaceLayout::new(
+            *dbs_boot::layout::GUEST_PHYS_END,
+            dbs_boot::layout::GUEST_MEM_START,
+            *dbs_boot::layout::GUEST_MEM_END,
+        );
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = 128 << 20;
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: vec![1, 2],
+        }];
+        let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+        let as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        assert_eq!(as_mgr.get_layout().unwrap(), layout);
+    }
+
+    #[test]
+    fn test_address_space_mgr_get_numa_nodes() {
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = 128 << 20;
+        let cpu_vec = vec![1, 2];
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: cpu_vec.clone(),
+        }];
+        let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+        let as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        let mut numa_node = NumaNode::new();
+        numa_node.add_info(&NumaNodeInfo {
+            base: GuestAddress(GUEST_MEM_START),
+            size: mem_size,
+        });
+        numa_node.add_vcpu_ids(&cpu_vec);
+
+        assert_eq!(*as_mgr.get_numa_nodes().get(&0).unwrap(), numa_node);
+    }
+
+    #[test]
+    fn test_address_space_mgr_async_prealloc() {
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = 2 << 20;
+        let cpu_vec = vec![1, 2];
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: cpu_vec,
+        }];
+        let mut builder = AddressSpaceMgrBuilder::new("hugeshmem", "").unwrap();
+        builder.toggle_prealloc(true);
+        let mut as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        as_mgr.wait_prealloc(false).unwrap();
+    }
+
+    #[test]
+    fn test_address_space_mgr_builder() {
+        let mut builder = AddressSpaceMgrBuilder::new("shmem", "/tmp/shmem").unwrap();
+
+        assert_eq!(builder.mem_type, "shmem");
+        assert_eq!(builder.mem_file, "/tmp/shmem");
+        assert_eq!(builder.mem_index, 0);
+        assert!(builder.mem_suffix);
+        assert!(!builder.mem_prealloc);
+        assert!(!builder.dirty_page_logging);
+        assert!(builder.vmfd.is_none());
+
+        assert_eq!(&builder.get_next_mem_file(), "/tmp/shmem0");
+        assert_eq!(&builder.get_next_mem_file(), "/tmp/shmem1");
+        assert_eq!(&builder.get_next_mem_file(), "/tmp/shmem2");
+        assert_eq!(builder.mem_index, 3);
+
+        builder.toggle_file_suffix(false);
+        assert_eq!(&builder.get_next_mem_file(), "/tmp/shmem");
+        assert_eq!(&builder.get_next_mem_file(), "/tmp/shmem");
+        assert_eq!(builder.mem_index, 3);
+
+        builder.toggle_prealloc(true);
+        builder.toggle_dirty_page_logging(true);
+        assert!(builder.mem_prealloc);
+        assert!(builder.dirty_page_logging);
+    }
+
+    #[test]
+    fn test_configure_invalid_numa() {
+        let res_mgr = ResourceManager::new(None);
+        let mem_size = 128 << 20;
+        let numa_region_infos = vec![NumaRegionInfo {
+            size: mem_size >> 20,
+            host_numa_node_id: None,
+            guest_numa_node_id: Some(0),
+            vcpu_ids: vec![1, 2],
+        }];
+        let builder = AddressSpaceMgrBuilder::new("shmem", "").unwrap();
+        let as_mgr = builder.build(&res_mgr, &numa_region_infos).unwrap();
+        let mmap_reg = MmapRegion::new(8).unwrap();
+
+        assert!(as_mgr.configure_numa(&mmap_reg, u32::MAX).is_err());
+    }
+}

src/dragonball/src/api/mod.rs

@@ -0,0 +1,6 @@
+// Copyright (C) 2019-2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+//! API related data structures to configure the vmm.
+
+pub mod v1;

src/dragonball/src/api/v1/boot_source.rs

@@ -0,0 +1,55 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+use serde_derive::{Deserialize, Serialize};
+
+/// Default guest kernel command line:
+/// - `reboot=k` shutdown the guest on reboot, instead of well... rebooting;
+/// - `panic=1` on panic, reboot after 1 second;
+/// - `pci=off` do not scan for PCI devices (ser boot time);
+/// - `nomodules` disable loadable kernel module support;
+/// - `8250.nr_uarts=0` disable 8250 serial interface;
+/// - `i8042.noaux` do not probe the i8042 controller for an attached mouse (ser boot time);
+/// - `i8042.nomux` do not probe i8042 for a multiplexing controller (ser boot time);
+/// - `i8042.nopnp` do not use ACPIPnP to discover KBD/AUX controllers (ser boot time);
+/// - `i8042.dumbkbd` do not attempt to control kbd state via the i8042 (ser boot time).
+pub const DEFAULT_KERNEL_CMDLINE: &str = "reboot=k panic=1 pci=off nomodules 8250.nr_uarts=0 \
+                                          i8042.noaux i8042.nomux i8042.nopnp i8042.dumbkbd";
+
+/// Strongly typed data structure used to configure the boot source of the microvm.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize, Default)]
+#[serde(deny_unknown_fields)]
+pub struct BootSourceConfig {
+    /// Path of the kernel image.
+    /// We only support uncompressed kernel for Dragonball.
+    pub kernel_path: String,
+    /// Path of the initrd, if there is one.
+    /// ps. rootfs is set in BlockDeviceConfigInfo
+    pub initrd_path: Option<String>,
+    /// The boot arguments to pass to the kernel.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub boot_args: Option<String>,
+}
+
+/// Errors associated with actions on `BootSourceConfig`.
+#[derive(Debug, thiserror::Error)]
+pub enum BootSourceConfigError {
+    /// The kernel file cannot be opened.
+    #[error(
+        "the kernel file cannot be opened due to invalid kernel path or invalid permissions: {0}"
+    )]
+    InvalidKernelPath(#[source] std::io::Error),
+
+    /// The initrd file cannot be opened.
+    #[error("the initrd file cannot be opened due to invalid path or invalid permissions: {0}")]
+    InvalidInitrdPath(#[source] std::io::Error),
+
+    /// The kernel command line is invalid.
+    #[error("the kernel command line is invalid: {0}")]
+    InvalidKernelCommandLine(#[source] linux_loader::cmdline::Error),
+
+    /// The boot source cannot be update post boot.
+    #[error("the update operation is not allowed after boot")]
+    UpdateNotAllowedPostBoot,
+}

src/dragonball/src/api/v1/instance_info.rs

@@ -0,0 +1,88 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+//
+// SPDX-License-Identifier: Apache-2.0
+
+use serde_derive::{Deserialize, Serialize};
+
+/// The microvm state.
+///
+/// When Dragonball starts, the instance state is Uninitialized. Once start_microvm method is
+/// called, the state goes from Uninitialized to Starting. The state is changed to Running until
+/// the start_microvm method ends. Halting and Halted are currently unsupported.
+#[derive(Copy, Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub enum InstanceState {
+    /// Microvm is not initialized.
+    Uninitialized,
+    /// Microvm is starting.
+    Starting,
+    /// Microvm is running.
+    Running,
+    /// Microvm is Paused.
+    Paused,
+    /// Microvm received a halt instruction.
+    Halting,
+    /// Microvm is halted.
+    Halted,
+    /// Microvm exit instead of process exit.
+    Exited(i32),
+}
+
+/// The state of async actions
+#[derive(Debug, Deserialize, Serialize, Clone, PartialEq)]
+pub enum AsyncState {
+    /// Uninitialized
+    Uninitialized,
+    /// Success
+    Success,
+    /// Failure
+    Failure,
+}
+
+/// The strongly typed that contains general information about the microVM.
+#[derive(Debug, Deserialize, Serialize)]
+pub struct InstanceInfo {
+    /// The ID of the microVM.
+    pub id: String,
+    /// The state of the microVM.
+    pub state: InstanceState,
+    /// The version of the VMM that runs the microVM.
+    pub vmm_version: String,
+    /// The pid of the current VMM process.
+    pub pid: u32,
+    /// The state of async actions.
+    pub async_state: AsyncState,
+    /// List of tids of vcpu threads (vcpu index, tid)
+    pub tids: Vec<(u8, u32)>,
+    /// Last instance downtime
+    pub last_instance_downtime: u64,
+}
+
+impl InstanceInfo {
+    /// create instance info object with given id, version, and platform type
+    pub fn new(id: String, vmm_version: String) -> Self {
+        InstanceInfo {
+            id,
+            state: InstanceState::Uninitialized,
+            vmm_version,
+            pid: std::process::id(),
+            async_state: AsyncState::Uninitialized,
+            tids: Vec::new(),
+            last_instance_downtime: 0,
+        }
+    }
+}
+
+impl Default for InstanceInfo {
+    fn default() -> Self {
+        InstanceInfo {
+            id: String::from(""),
+            state: InstanceState::Uninitialized,
+            vmm_version: env!("CARGO_PKG_VERSION").to_string(),
+            pid: std::process::id(),
+            async_state: AsyncState::Uninitialized,
+            tids: Vec::new(),
+            last_instance_downtime: 0,
+        }
+    }
+}

src/dragonball/src/api/v1/machine_config.rs

@@ -0,0 +1,86 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+/// We only support this number of vcpus for now. Mostly because we have set all vcpu related metrics as u8
+/// and breaking u8 will take extra efforts.
+pub const MAX_SUPPORTED_VCPUS: u8 = 254;
+
+/// Memory hotplug value should have alignment in this size (unit: MiB)
+pub const MEMORY_HOTPLUG_ALIGHMENT: u8 = 64;
+
+/// Errors associated with configuring the microVM.
+#[derive(Debug, PartialEq, thiserror::Error)]
+pub enum VmConfigError {
+    /// Cannot update the configuration of the microvm post boot.
+    #[error("update operation is not allowed after boot")]
+    UpdateNotAllowedPostBoot,
+
+    /// The max vcpu count is invalid.
+    #[error("the vCPU number shouldn't large than {}", MAX_SUPPORTED_VCPUS)]
+    VcpuCountExceedsMaximum,
+
+    /// The vcpu count is invalid. When hyperthreading is enabled, the `cpu_count` must be either
+    /// 1 or an even number.
+    #[error(
+        "the vCPU number '{0}' can only be 1 or an even number when hyperthreading is enabled"
+    )]
+    InvalidVcpuCount(u8),
+
+    /// The threads_per_core is invalid. It should be either 1 or 2.
+    #[error("the threads_per_core number '{0}' can only be 1 or 2")]
+    InvalidThreadsPerCore(u8),
+
+    /// The cores_per_die is invalid. It should be larger than 0.
+    #[error("the cores_per_die number '{0}' can only be larger than 0")]
+    InvalidCoresPerDie(u8),
+
+    /// The dies_per_socket is invalid. It should be larger than 0.
+    #[error("the dies_per_socket number '{0}' can only be larger than 0")]
+    InvalidDiesPerSocket(u8),
+
+    /// The socket number is invalid. It should be either 1 or 2.
+    #[error("the socket number '{0}' can only be 1 or 2")]
+    InvalidSocket(u8),
+
+    /// max vcpu count inferred from cpu topology(threads_per_core * cores_per_die * dies_per_socket * sockets) should be larger or equal to vcpu_count
+    #[error("the max vcpu count inferred from cpu topology '{0}' (threads_per_core * cores_per_die * dies_per_socket * sockets) should be larger or equal to vcpu_count")]
+    InvalidCpuTopology(u8),
+
+    /// The max vcpu count is invalid.
+    #[error(
+        "the max vCPU number '{0}' shouldn't less than vCPU count and can only be 1 or an even number when hyperthreading is enabled"
+    )]
+    InvalidMaxVcpuCount(u8),
+
+    /// The memory size is invalid. The memory can only be an unsigned integer.
+    #[error("the memory size 0x{0:x}MiB is invalid")]
+    InvalidMemorySize(usize),
+
+    /// The hotplug memory size is invalid. The memory can only be an unsigned integer.
+    #[error(
+        "the hotplug memory size '{0}' (MiB) is invalid, must be multiple of {}",
+        MEMORY_HOTPLUG_ALIGHMENT
+    )]
+    InvalidHotplugMemorySize(usize),
+
+    /// The memory type is invalid.
+    #[error("the memory type '{0}' is invalid")]
+    InvalidMemType(String),
+
+    /// The memory file path is invalid.
+    #[error("the memory file path is invalid")]
+    InvalidMemFilePath(String),
+
+    /// NUMA region memory size is invalid
+    #[error("Total size of memory in NUMA regions: {0}, should matches memory size in config")]
+    InvalidNumaRegionMemorySize(usize),
+
+    /// NUMA region vCPU count is invalid
+    #[error("Total counts of vCPUs in NUMA regions: {0}, should matches max vcpu count in config")]
+    InvalidNumaRegionCpuCount(u16),
+
+    /// NUMA region vCPU count is invalid
+    #[error("Max id of vCPUs in NUMA regions: {0}, should matches max vcpu count in config")]
+    InvalidNumaRegionCpuMaxId(u16),
+}

src/dragonball/src/api/v1/mod.rs

@@ -0,0 +1,19 @@
+// Copyright (C) 2019-2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+//! API Version 1 related data structures to configure the vmm.
+
+mod vmm_action;
+pub use self::vmm_action::*;
+
+/// Wrapper for configuring the microVM boot source.
+mod boot_source;
+pub use self::boot_source::{BootSourceConfig, BootSourceConfigError, DEFAULT_KERNEL_CMDLINE};
+
+/// Wrapper over the microVM general information.
+mod instance_info;
+pub use self::instance_info::{InstanceInfo, InstanceState};
+
+/// Wrapper for configuring the memory and CPU of the microVM.
+mod machine_config;
+pub use self::machine_config::{VmConfigError, MAX_SUPPORTED_VCPUS};

src/dragonball/src/api/v1/vmm_action.rs

@@ -0,0 +1,636 @@
+// Copyright (C) 2020-2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+use std::fs::File;
+use std::sync::mpsc::{Receiver, Sender, TryRecvError};
+
+use log::{debug, error, info, warn};
+
+use crate::error::{Result, StartMicroVmError, StopMicrovmError};
+use crate::event_manager::EventManager;
+use crate::vm::{CpuTopology, KernelConfigInfo, VmConfigInfo};
+use crate::vmm::Vmm;
+
+use self::VmConfigError::*;
+use self::VmmActionError::MachineConfig;
+
+#[cfg(feature = "virtio-blk")]
+pub use crate::device_manager::blk_dev_mgr::{
+    BlockDeviceConfigInfo, BlockDeviceConfigUpdateInfo, BlockDeviceError, BlockDeviceMgr,
+};
+#[cfg(feature = "virtio-fs")]
+pub use crate::device_manager::fs_dev_mgr::{
+    FsDeviceConfigInfo, FsDeviceConfigUpdateInfo, FsDeviceError, FsDeviceMgr, FsMountConfigInfo,
+};
+#[cfg(feature = "virtio-net")]
+pub use crate::device_manager::virtio_net_dev_mgr::{
+    VirtioNetDeviceConfigInfo, VirtioNetDeviceConfigUpdateInfo, VirtioNetDeviceError,
+    VirtioNetDeviceMgr,
+};
+#[cfg(feature = "virtio-vsock")]
+pub use crate::device_manager::vsock_dev_mgr::{VsockDeviceConfigInfo, VsockDeviceError};
+
+use super::*;
+
+/// Wrapper for all errors associated with VMM actions.
+#[derive(Debug, thiserror::Error)]
+pub enum VmmActionError {
+    /// Invalid virtual machine instance ID.
+    #[error("the virtual machine instance ID is invalid")]
+    InvalidVMID,
+
+    /// Failed to hotplug, due to Upcall not ready.
+    #[error("Upcall not ready, can't hotplug device.")]
+    UpcallNotReady,
+
+    /// The action `ConfigureBootSource` failed either because of bad user input or an internal
+    /// error.
+    #[error("failed to configure boot source for VM: {0}")]
+    BootSource(#[source] BootSourceConfigError),
+
+    /// The action `StartMicroVm` failed either because of bad user input or an internal error.
+    #[error("failed to boot the VM: {0}")]
+    StartMicroVm(#[source] StartMicroVmError),
+
+    /// The action `StopMicroVm` failed either because of bad user input or an internal error.
+    #[error("failed to shutdown the VM: {0}")]
+    StopMicrovm(#[source] StopMicrovmError),
+
+    /// One of the actions `GetVmConfiguration` or `SetVmConfiguration` failed either because of bad
+    /// input or an internal error.
+    #[error("failed to set configuration for the VM: {0}")]
+    MachineConfig(#[source] VmConfigError),
+
+    #[cfg(feature = "virtio-vsock")]
+    /// The action `InsertVsockDevice` failed either because of bad user input or an internal error.
+    #[error("failed to add virtio-vsock device: {0}")]
+    Vsock(#[source] VsockDeviceError),
+
+    #[cfg(feature = "virtio-blk")]
+    /// Block device related errors.
+    #[error("virtio-blk device error: {0}")]
+    Block(#[source] BlockDeviceError),
+
+    #[cfg(feature = "virtio-net")]
+    /// Net device related errors.
+    #[error("virtio-net device error: {0}")]
+    VirtioNet(#[source] VirtioNetDeviceError),
+
+    #[cfg(feature = "virtio-fs")]
+    /// The action `InsertFsDevice` failed either because of bad user input or an internal error.
+    #[error("virtio-fs device: {0}")]
+    FsDevice(#[source] FsDeviceError),
+}
+
+/// This enum represents the public interface of the VMM. Each action contains various
+/// bits of information (ids, paths, etc.).
+#[derive(Clone, Debug, PartialEq)]
+pub enum VmmAction {
+    /// Configure the boot source of the microVM using `BootSourceConfig`.
+    /// This action can only be called before the microVM has booted.
+    ConfigureBootSource(BootSourceConfig),
+
+    /// Launch the microVM. This action can only be called before the microVM has booted.
+    StartMicroVm,
+
+    /// Shutdown the vmicroVM. This action can only be called after the microVM has booted.
+    /// When vmm is used as the crate by the other process, which is need to
+    /// shutdown the vcpu threads and destory all of the object.
+    ShutdownMicroVm,
+
+    /// Get the configuration of the microVM.
+    GetVmConfiguration,
+
+    /// Set the microVM configuration (memory & vcpu) using `VmConfig` as input. This
+    /// action can only be called before the microVM has booted.
+    SetVmConfiguration(VmConfigInfo),
+
+    #[cfg(feature = "virtio-vsock")]
+    /// Add a new vsock device or update one that already exists using the
+    /// `VsockDeviceConfig` as input. This action can only be called before the microVM has
+    /// booted. The response is sent using the `OutcomeSender`.
+    InsertVsockDevice(VsockDeviceConfigInfo),
+
+    #[cfg(feature = "virtio-blk")]
+    /// Add a new block device or update one that already exists using the `BlockDeviceConfig` as
+    /// input. This action can only be called before the microVM has booted.
+    InsertBlockDevice(BlockDeviceConfigInfo),
+
+    #[cfg(feature = "virtio-blk")]
+    /// Remove a new block device for according to given drive_id
+    RemoveBlockDevice(String),
+
+    #[cfg(feature = "virtio-blk")]
+    /// Update a block device, after microVM start. Currently, the only updatable properties
+    /// are the RX and TX rate limiters.
+    UpdateBlockDevice(BlockDeviceConfigUpdateInfo),
+
+    #[cfg(feature = "virtio-net")]
+    /// Add a new network interface config or update one that already exists using the
+    /// `NetworkInterfaceConfig` as input. This action can only be called before the microVM has
+    /// booted. The response is sent using the `OutcomeSender`.
+    InsertNetworkDevice(VirtioNetDeviceConfigInfo),
+
+    #[cfg(feature = "virtio-net")]
+    /// Update a network interface, after microVM start. Currently, the only updatable properties
+    /// are the RX and TX rate limiters.
+    UpdateNetworkInterface(VirtioNetDeviceConfigUpdateInfo),
+
+    #[cfg(feature = "virtio-fs")]
+    /// Add a new shared fs device or update one that already exists using the
+    /// `FsDeviceConfig` as input. This action can only be called before the microVM has
+    /// booted.
+    InsertFsDevice(FsDeviceConfigInfo),
+
+    #[cfg(feature = "virtio-fs")]
+    /// Attach a new virtiofs Backend fs or detach an existing virtiofs Backend fs using the
+    /// `FsMountConfig` as input. This action can only be called _after_ the microVM has
+    /// booted.
+    ManipulateFsBackendFs(FsMountConfigInfo),
+
+    #[cfg(feature = "virtio-fs")]
+    /// Update fs rate limiter, after microVM start.
+    UpdateFsDevice(FsDeviceConfigUpdateInfo),
+}
+
+/// The enum represents the response sent by the VMM in case of success. The response is either
+/// empty, when no data needs to be sent, or an internal VMM structure.
+#[derive(Debug)]
+pub enum VmmData {
+    /// No data is sent on the channel.
+    Empty,
+    /// The microVM configuration represented by `VmConfigInfo`.
+    MachineConfiguration(Box<VmConfigInfo>),
+}
+
+/// Request data type used to communicate between the API and the VMM.
+pub type VmmRequest = Box<VmmAction>;
+
+/// Data type used to communicate between the API and the VMM.
+pub type VmmRequestResult = std::result::Result<VmmData, VmmActionError>;
+
+/// Response data type used to communicate between the API and the VMM.
+pub type VmmResponse = Box<VmmRequestResult>;
+
+/// VMM Service to handle requests from the API server.
+///
+/// There are two levels of API servers as below:
+/// API client <--> VMM API Server <--> VMM Core
+pub struct VmmService {
+    from_api: Receiver<VmmRequest>,
+    to_api: Sender<VmmResponse>,
+    machine_config: VmConfigInfo,
+}
+
+impl VmmService {
+    /// Create a new VMM API server instance.
+    pub fn new(from_api: Receiver<VmmRequest>, to_api: Sender<VmmResponse>) -> Self {
+        VmmService {
+            from_api,
+            to_api,
+            machine_config: VmConfigInfo::default(),
+        }
+    }
+
+    /// Handle requests from the HTTP API Server and send back replies.
+    pub fn run_vmm_action(&mut self, vmm: &mut Vmm, event_mgr: &mut EventManager) -> Result<()> {
+        let request = match self.from_api.try_recv() {
+            Ok(t) => *t,
+            Err(TryRecvError::Empty) => {
+                warn!("Got a spurious notification from api thread");
+                return Ok(());
+            }
+            Err(TryRecvError::Disconnected) => {
+                panic!("The channel's sending half was disconnected. Cannot receive data.");
+            }
+        };
+        debug!("receive vmm action: {:?}", request);
+
+        let response = match request {
+            VmmAction::ConfigureBootSource(boot_source_body) => {
+                self.configure_boot_source(vmm, boot_source_body)
+            }
+            VmmAction::StartMicroVm => self.start_microvm(vmm, event_mgr),
+            VmmAction::ShutdownMicroVm => self.shutdown_microvm(vmm),
+            VmmAction::GetVmConfiguration => Ok(VmmData::MachineConfiguration(Box::new(
+                self.machine_config.clone(),
+            ))),
+            VmmAction::SetVmConfiguration(machine_config) => {
+                self.set_vm_configuration(vmm, machine_config)
+            }
+            #[cfg(feature = "virtio-vsock")]
+            VmmAction::InsertVsockDevice(vsock_cfg) => self.add_vsock_device(vmm, vsock_cfg),
+            #[cfg(feature = "virtio-blk")]
+            VmmAction::InsertBlockDevice(block_device_config) => {
+                self.add_block_device(vmm, event_mgr, block_device_config)
+            }
+            #[cfg(feature = "virtio-blk")]
+            VmmAction::UpdateBlockDevice(blk_update) => {
+                self.update_blk_rate_limiters(vmm, blk_update)
+            }
+            #[cfg(feature = "virtio-blk")]
+            VmmAction::RemoveBlockDevice(drive_id) => {
+                self.remove_block_device(vmm, event_mgr, &drive_id)
+            }
+            #[cfg(feature = "virtio-net")]
+            VmmAction::InsertNetworkDevice(virtio_net_cfg) => {
+                self.add_virtio_net_device(vmm, event_mgr, virtio_net_cfg)
+            }
+            #[cfg(feature = "virtio-net")]
+            VmmAction::UpdateNetworkInterface(netif_update) => {
+                self.update_net_rate_limiters(vmm, netif_update)
+            }
+            #[cfg(feature = "virtio-fs")]
+            VmmAction::InsertFsDevice(fs_cfg) => self.add_fs_device(vmm, fs_cfg),
+
+            #[cfg(feature = "virtio-fs")]
+            VmmAction::ManipulateFsBackendFs(fs_mount_cfg) => {
+                self.manipulate_fs_backend_fs(vmm, fs_mount_cfg)
+            }
+            #[cfg(feature = "virtio-fs")]
+            VmmAction::UpdateFsDevice(fs_update_cfg) => {
+                self.update_fs_rate_limiters(vmm, fs_update_cfg)
+            }
+        };
+
+        debug!("send vmm response: {:?}", response);
+        self.send_response(response)
+    }
+
+    fn send_response(&self, result: VmmRequestResult) -> Result<()> {
+        self.to_api
+            .send(Box::new(result))
+            .map_err(|_| ())
+            .expect("vmm: one-shot API result channel has been closed");
+
+        Ok(())
+    }
+
+    fn configure_boot_source(
+        &self,
+        vmm: &mut Vmm,
+        boot_source_config: BootSourceConfig,
+    ) -> VmmRequestResult {
+        use super::BootSourceConfigError::{
+            InvalidInitrdPath, InvalidKernelCommandLine, InvalidKernelPath,
+            UpdateNotAllowedPostBoot,
+        };
+        use super::VmmActionError::BootSource;
+
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        if vm.is_vm_initialized() {
+            return Err(BootSource(UpdateNotAllowedPostBoot));
+        }
+
+        let kernel_file = File::open(&boot_source_config.kernel_path)
+            .map_err(|e| BootSource(InvalidKernelPath(e)))?;
+
+        let initrd_file = match boot_source_config.initrd_path {
+            None => None,
+            Some(ref path) => Some(File::open(path).map_err(|e| BootSource(InvalidInitrdPath(e)))?),
+        };
+
+        let mut cmdline = linux_loader::cmdline::Cmdline::new(dbs_boot::layout::CMDLINE_MAX_SIZE);
+        let boot_args = boot_source_config
+            .boot_args
+            .clone()
+            .unwrap_or_else(|| String::from(DEFAULT_KERNEL_CMDLINE));
+        cmdline
+            .insert_str(boot_args)
+            .map_err(|e| BootSource(InvalidKernelCommandLine(e)))?;
+
+        let kernel_config = KernelConfigInfo::new(kernel_file, initrd_file, cmdline);
+        vm.set_kernel_config(kernel_config);
+
+        Ok(VmmData::Empty)
+    }
+
+    fn start_microvm(&mut self, vmm: &mut Vmm, event_mgr: &mut EventManager) -> VmmRequestResult {
+        use self::StartMicroVmError::MicroVMAlreadyRunning;
+        use self::VmmActionError::StartMicroVm;
+
+        let vmm_seccomp_filter = vmm.vmm_seccomp_filter();
+        let vcpu_seccomp_filter = vmm.vcpu_seccomp_filter();
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        if vm.is_vm_initialized() {
+            return Err(StartMicroVm(MicroVMAlreadyRunning));
+        }
+
+        vm.start_microvm(event_mgr, vmm_seccomp_filter, vcpu_seccomp_filter)
+            .map(|_| VmmData::Empty)
+            .map_err(StartMicroVm)
+    }
+
+    fn shutdown_microvm(&mut self, vmm: &mut Vmm) -> VmmRequestResult {
+        vmm.event_ctx.exit_evt_triggered = true;
+
+        Ok(VmmData::Empty)
+    }
+
+    /// Set virtual machine configuration.
+    pub fn set_vm_configuration(
+        &mut self,
+        vmm: &mut Vmm,
+        machine_config: VmConfigInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        if vm.is_vm_initialized() {
+            return Err(MachineConfig(UpdateNotAllowedPostBoot));
+        }
+
+        // If the check is successful, set it up together.
+        let mut config = vm.vm_config().clone();
+        if config.vcpu_count != machine_config.vcpu_count {
+            let vcpu_count = machine_config.vcpu_count;
+            // Check that the vcpu_count value is >=1.
+            if vcpu_count == 0 {
+                return Err(MachineConfig(InvalidVcpuCount(vcpu_count)));
+            }
+            config.vcpu_count = vcpu_count;
+        }
+
+        if config.cpu_topology != machine_config.cpu_topology {
+            let cpu_topology = &machine_config.cpu_topology;
+            config.cpu_topology = handle_cpu_topology(cpu_topology, config.vcpu_count)?.clone();
+        } else {
+            // the same default
+            let mut default_cpu_topology = CpuTopology {
+                threads_per_core: 1,
+                cores_per_die: config.vcpu_count,
+                dies_per_socket: 1,
+                sockets: 1,
+            };
+            if machine_config.max_vcpu_count > config.vcpu_count {
+                default_cpu_topology.cores_per_die = machine_config.max_vcpu_count;
+            }
+            config.cpu_topology = default_cpu_topology;
+        }
+        let cpu_topology = &config.cpu_topology;
+        let max_vcpu_from_topo = cpu_topology.threads_per_core
+            * cpu_topology.cores_per_die
+            * cpu_topology.dies_per_socket
+            * cpu_topology.sockets;
+        // If the max_vcpu_count inferred by cpu_topology is not equal to
+        // max_vcpu_count, max_vcpu_count will be changed. currently, max vcpu size
+        // is used when cpu_topology is not defined and help define the cores_per_die
+        // for the default cpu topology.
+        let mut max_vcpu_count = machine_config.max_vcpu_count;
+        if max_vcpu_count < config.vcpu_count {
+            return Err(MachineConfig(InvalidMaxVcpuCount(max_vcpu_count)));
+        }
+        if max_vcpu_from_topo != max_vcpu_count {
+            max_vcpu_count = max_vcpu_from_topo;
+            info!("Since max_vcpu_count is not equal to cpu topo information, we have changed the max vcpu count to {}", max_vcpu_from_topo);
+        }
+        config.max_vcpu_count = max_vcpu_count;
+
+        config.cpu_pm = machine_config.cpu_pm;
+        config.mem_type = machine_config.mem_type;
+
+        let mem_size_mib_value = machine_config.mem_size_mib;
+        // Support 1TB memory at most, 2MB aligned for huge page.
+        if mem_size_mib_value == 0 || mem_size_mib_value > 0x10_0000 || mem_size_mib_value % 2 != 0
+        {
+            return Err(MachineConfig(InvalidMemorySize(mem_size_mib_value)));
+        }
+        config.mem_size_mib = mem_size_mib_value;
+
+        config.mem_file_path = machine_config.mem_file_path.clone();
+
+        if config.mem_type == "hugetlbfs" && config.mem_file_path.is_empty() {
+            return Err(MachineConfig(InvalidMemFilePath("".to_owned())));
+        }
+        config.vpmu_feature = machine_config.vpmu_feature;
+
+        let vm_id = vm.shared_info().read().unwrap().id.clone();
+        let serial_path = match machine_config.serial_path {
+            Some(value) => value,
+            None => {
+                if config.serial_path.is_none() {
+                    String::from("/run/dragonball/") + &vm_id + "_com1"
+                } else {
+                    // Safe to unwrap() because we have checked it has a value.
+                    config.serial_path.as_ref().unwrap().clone()
+                }
+            }
+        };
+        config.serial_path = Some(serial_path);
+
+        vm.set_vm_config(config.clone());
+        self.machine_config = config;
+
+        Ok(VmmData::Empty)
+    }
+
+    #[cfg(feature = "virtio-vsock")]
+    fn add_vsock_device(&self, vmm: &mut Vmm, config: VsockDeviceConfigInfo) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        if vm.is_vm_initialized() {
+            return Err(VmmActionError::Vsock(
+                VsockDeviceError::UpdateNotAllowedPostBoot,
+            ));
+        }
+
+        // VMADDR_CID_ANY (-1U) means any address for binding;
+        // VMADDR_CID_HYPERVISOR (0) is reserved for services built into the hypervisor;
+        // VMADDR_CID_RESERVED (1) must not be used;
+        // VMADDR_CID_HOST (2) is the well-known address of the host.
+        if config.guest_cid <= 2 {
+            return Err(VmmActionError::Vsock(VsockDeviceError::GuestCIDInvalid(
+                config.guest_cid,
+            )));
+        }
+
+        info!("add_vsock_device: {:?}", config);
+        let ctx = vm.create_device_op_context(None).map_err(|e| {
+            info!("create device op context error: {:?}", e);
+            VmmActionError::Vsock(VsockDeviceError::UpdateNotAllowedPostBoot)
+        })?;
+
+        vm.device_manager_mut()
+            .vsock_manager
+            .insert_device(ctx, config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::Vsock)
+    }
+
+    #[cfg(feature = "virtio-blk")]
+    // Only call this function as part of the API.
+    // If the drive_id does not exist, a new Block Device Config is added to the list.
+    fn add_block_device(
+        &mut self,
+        vmm: &mut Vmm,
+        event_mgr: &mut EventManager,
+        config: BlockDeviceConfigInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        let ctx = vm
+            .create_device_op_context(Some(event_mgr.epoll_manager()))
+            .map_err(|e| {
+                if let StartMicroVmError::UpcallNotReady = e {
+                    return VmmActionError::UpcallNotReady;
+                }
+                VmmActionError::Block(BlockDeviceError::UpdateNotAllowedPostBoot)
+            })?;
+
+        BlockDeviceMgr::insert_device(vm.device_manager_mut(), ctx, config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::Block)
+    }
+
+    #[cfg(feature = "virtio-blk")]
+    /// Updates configuration for an emulated net device as described in `config`.
+    fn update_blk_rate_limiters(
+        &mut self,
+        vmm: &mut Vmm,
+        config: BlockDeviceConfigUpdateInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+
+        BlockDeviceMgr::update_device_ratelimiters(vm.device_manager_mut(), config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::Block)
+    }
+
+    #[cfg(feature = "virtio-blk")]
+    // Remove the device
+    fn remove_block_device(
+        &mut self,
+        vmm: &mut Vmm,
+        event_mgr: &mut EventManager,
+        drive_id: &str,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        let ctx = vm
+            .create_device_op_context(Some(event_mgr.epoll_manager()))
+            .map_err(|_| VmmActionError::Block(BlockDeviceError::UpdateNotAllowedPostBoot))?;
+
+        BlockDeviceMgr::remove_device(vm.device_manager_mut(), ctx, drive_id)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::Block)
+    }
+
+    #[cfg(feature = "virtio-net")]
+    fn add_virtio_net_device(
+        &mut self,
+        vmm: &mut Vmm,
+        event_mgr: &mut EventManager,
+        config: VirtioNetDeviceConfigInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        let ctx = vm
+            .create_device_op_context(Some(event_mgr.epoll_manager()))
+            .map_err(|e| {
+                if let StartMicroVmError::MicroVMAlreadyRunning = e {
+                    VmmActionError::VirtioNet(VirtioNetDeviceError::UpdateNotAllowedPostBoot)
+                } else if let StartMicroVmError::UpcallNotReady = e {
+                    VmmActionError::UpcallNotReady
+                } else {
+                    VmmActionError::StartMicroVm(e)
+                }
+            })?;
+
+        VirtioNetDeviceMgr::insert_device(vm.device_manager_mut(), ctx, config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::VirtioNet)
+    }
+
+    #[cfg(feature = "virtio-net")]
+    fn update_net_rate_limiters(
+        &mut self,
+        vmm: &mut Vmm,
+        config: VirtioNetDeviceConfigUpdateInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+
+        VirtioNetDeviceMgr::update_device_ratelimiters(vm.device_manager_mut(), config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::VirtioNet)
+    }
+
+    #[cfg(feature = "virtio-fs")]
+    fn add_fs_device(&mut self, vmm: &mut Vmm, config: FsDeviceConfigInfo) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+        let hotplug = vm.is_vm_initialized();
+        if !cfg!(feature = "hotplug") && hotplug {
+            return Err(VmmActionError::FsDevice(
+                FsDeviceError::UpdateNotAllowedPostBoot,
+            ));
+        }
+
+        let ctx = vm.create_device_op_context(None).map_err(|e| {
+            info!("create device op context error: {:?}", e);
+            VmmActionError::FsDevice(FsDeviceError::UpdateNotAllowedPostBoot)
+        })?;
+        FsDeviceMgr::insert_device(vm.device_manager_mut(), ctx, config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::FsDevice)
+    }
+
+    #[cfg(feature = "virtio-fs")]
+    fn manipulate_fs_backend_fs(
+        &self,
+        vmm: &mut Vmm,
+        config: FsMountConfigInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+
+        if !vm.is_vm_initialized() {
+            return Err(VmmActionError::FsDevice(FsDeviceError::MicroVMNotRunning));
+        }
+
+        FsDeviceMgr::manipulate_backend_fs(vm.device_manager_mut(), config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::FsDevice)
+    }
+
+    #[cfg(feature = "virtio-fs")]
+    fn update_fs_rate_limiters(
+        &self,
+        vmm: &mut Vmm,
+        config: FsDeviceConfigUpdateInfo,
+    ) -> VmmRequestResult {
+        let vm = vmm.get_vm_mut().ok_or(VmmActionError::InvalidVMID)?;
+
+        if !vm.is_vm_initialized() {
+            return Err(VmmActionError::FsDevice(FsDeviceError::MicroVMNotRunning));
+        }
+
+        FsDeviceMgr::update_device_ratelimiters(vm.device_manager_mut(), config)
+            .map(|_| VmmData::Empty)
+            .map_err(VmmActionError::FsDevice)
+    }
+}
+
+fn handle_cpu_topology(
+    cpu_topology: &CpuTopology,
+    vcpu_count: u8,
+) -> std::result::Result<&CpuTopology, VmmActionError> {
+    // Check if dies_per_socket, cores_per_die, threads_per_core and socket number is valid
+    if cpu_topology.threads_per_core < 1 || cpu_topology.threads_per_core > 2 {
+        return Err(MachineConfig(InvalidThreadsPerCore(
+            cpu_topology.threads_per_core,
+        )));
+    }
+    let vcpu_count_from_topo = cpu_topology
+        .sockets
+        .checked_mul(cpu_topology.dies_per_socket)
+        .ok_or(MachineConfig(VcpuCountExceedsMaximum))?
+        .checked_mul(cpu_topology.cores_per_die)
+        .ok_or(MachineConfig(VcpuCountExceedsMaximum))?
+        .checked_mul(cpu_topology.threads_per_core)
+        .ok_or(MachineConfig(VcpuCountExceedsMaximum))?;
+    if vcpu_count_from_topo > MAX_SUPPORTED_VCPUS {
+        return Err(MachineConfig(VcpuCountExceedsMaximum));
+    }
+    if vcpu_count_from_topo < vcpu_count {
+        return Err(MachineConfig(InvalidCpuTopology(vcpu_count_from_topo)));
+    }
+
+    Ok(cpu_topology)
+}

src/dragonball/src/config_manager.rs

@@ -0,0 +1,760 @@
+// Copyright (C) 2020-2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+use std::convert::TryInto;
+use std::io;
+use std::ops::{Index, IndexMut};
+use std::sync::Arc;
+
+use dbs_device::DeviceIo;
+use dbs_utils::rate_limiter::{RateLimiter, TokenBucket};
+use serde_derive::{Deserialize, Serialize};
+
+/// Get bucket update for rate limiter.
+#[macro_export]
+macro_rules! get_bucket_update {
+    ($self:ident, $rate_limiter: ident, $metric: ident) => {{
+        match &$self.$rate_limiter {
+            Some(rl_cfg) => {
+                let tb_cfg = &rl_cfg.$metric;
+                dbs_utils::rate_limiter::RateLimiter::make_bucket(
+                    tb_cfg.size,
+                    tb_cfg.one_time_burst,
+                    tb_cfg.refill_time,
+                )
+                // Updated active rate-limiter.
+                .map(dbs_utils::rate_limiter::BucketUpdate::Update)
+                // Updated/deactivated rate-limiter
+                .unwrap_or(dbs_utils::rate_limiter::BucketUpdate::Disabled)
+            }
+            // No update to the rate-limiter.
+            None => dbs_utils::rate_limiter::BucketUpdate::None,
+        }
+    }};
+}
+
+/// Trait for generic configuration information.
+pub trait ConfigItem {
+    /// Related errors.
+    type Err;
+
+    /// Get the unique identifier of the configuration item.
+    fn id(&self) -> &str;
+
+    /// Check whether current configuration item conflicts with another one.
+    fn check_conflicts(&self, other: &Self) -> std::result::Result<(), Self::Err>;
+}
+
+/// Struct to manage a group of configuration items.
+#[derive(Debug, Default, Deserialize, PartialEq, Serialize)]
+pub struct ConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    configs: Vec<T>,
+}
+
+impl<T> ConfigInfos<T>
+where
+    T: ConfigItem + Clone + Default,
+{
+    /// Constructor
+    pub fn new() -> Self {
+        ConfigInfos::default()
+    }
+
+    /// Insert a configuration item in the group.
+    pub fn insert(&mut self, config: T) -> std::result::Result<(), T::Err> {
+        for item in self.configs.iter() {
+            config.check_conflicts(item)?;
+        }
+        self.configs.push(config);
+
+        Ok(())
+    }
+
+    /// Update a configuration item in the group.
+    pub fn update(&mut self, config: T, err: T::Err) -> std::result::Result<(), T::Err> {
+        match self.get_index_by_id(&config) {
+            None => Err(err),
+            Some(index) => {
+                for (idx, item) in self.configs.iter().enumerate() {
+                    if idx != index {
+                        config.check_conflicts(item)?;
+                    }
+                }
+                self.configs[index] = config;
+                Ok(())
+            }
+        }
+    }
+
+    /// Insert or update a configuration item in the group.
+    pub fn insert_or_update(&mut self, config: T) -> std::result::Result<(), T::Err> {
+        match self.get_index_by_id(&config) {
+            None => {
+                for item in self.configs.iter() {
+                    config.check_conflicts(item)?;
+                }
+
+                self.configs.push(config)
+            }
+            Some(index) => {
+                for (idx, item) in self.configs.iter().enumerate() {
+                    if idx != index {
+                        config.check_conflicts(item)?;
+                    }
+                }
+                self.configs[index] = config;
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Remove the matching configuration entry.
+    pub fn remove(&mut self, config: &T) -> Option<T> {
+        if let Some(index) = self.get_index_by_id(config) {
+            Some(self.configs.remove(index))
+        } else {
+            None
+        }
+    }
+
+    /// Returns an immutable iterator over the config items
+    pub fn iter(&self) -> ::std::slice::Iter<T> {
+        self.configs.iter()
+    }
+
+    /// Get the configuration entry with matching ID.
+    pub fn get_by_id(&self, item: &T) -> Option<&T> {
+        let id = item.id();
+
+        self.configs.iter().rfind(|cfg| cfg.id() == id)
+    }
+
+    fn get_index_by_id(&self, item: &T) -> Option<usize> {
+        let id = item.id();
+        self.configs.iter().position(|cfg| cfg.id() == id)
+    }
+}
+
+impl<T> Clone for ConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    fn clone(&self) -> Self {
+        ConfigInfos {
+            configs: self.configs.clone(),
+        }
+    }
+}
+
+/// Struct to maintain configuration information for a device.
+pub struct DeviceConfigInfo<T>
+where
+    T: ConfigItem + Clone,
+{
+    /// Configuration information for the device object.
+    pub config: T,
+    /// The associated device object.
+    pub device: Option<Arc<dyn DeviceIo>>,
+}
+
+impl<T> DeviceConfigInfo<T>
+where
+    T: ConfigItem + Clone,
+{
+    /// Create a new instance of ['DeviceInfoGroup'].
+    pub fn new(config: T) -> Self {
+        DeviceConfigInfo {
+            config,
+            device: None,
+        }
+    }
+
+    /// Create a new instance of ['DeviceInfoGroup'] with optional device.
+    pub fn new_with_device(config: T, device: Option<Arc<dyn DeviceIo>>) -> Self {
+        DeviceConfigInfo { config, device }
+    }
+
+    /// Set the device object associated with the configuration.
+    pub fn set_device(&mut self, device: Arc<dyn DeviceIo>) {
+        self.device = Some(device);
+    }
+}
+
+impl<T> Clone for DeviceConfigInfo<T>
+where
+    T: ConfigItem + Clone,
+{
+    fn clone(&self) -> Self {
+        DeviceConfigInfo::new_with_device(self.config.clone(), self.device.clone())
+    }
+}
+
+/// Struct to maintain configuration information for a group of devices.
+pub struct DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    info_list: Vec<DeviceConfigInfo<T>>,
+}
+
+impl<T> Default for DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl<T> DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    /// Create a new instance of ['DeviceConfigInfos'].
+    pub fn new() -> Self {
+        DeviceConfigInfos {
+            info_list: Vec::new(),
+        }
+    }
+
+    /// Insert or update configuration information for a device.
+    pub fn insert_or_update(&mut self, config: &T) -> std::result::Result<usize, T::Err> {
+        let device_info = DeviceConfigInfo::new(config.clone());
+        Ok(match self.get_index_by_id(config) {
+            Some(index) => {
+                for (idx, info) in self.info_list.iter().enumerate() {
+                    if idx != index {
+                        info.config.check_conflicts(config)?;
+                    }
+                }
+                self.info_list[index] = device_info;
+                index
+            }
+            None => {
+                for info in self.info_list.iter() {
+                    info.config.check_conflicts(config)?;
+                }
+                self.info_list.push(device_info);
+                self.info_list.len() - 1
+            }
+        })
+    }
+
+    /// Remove a device configuration information object.
+    pub fn remove(&mut self, index: usize) -> Option<DeviceConfigInfo<T>> {
+        if self.info_list.len() > index {
+            Some(self.info_list.remove(index))
+        } else {
+            None
+        }
+    }
+
+    /// Get number of device configuration information objects.
+    pub fn len(&self) -> usize {
+        self.info_list.len()
+    }
+
+    /// Returns true if the device configuration information objects is empty.
+    pub fn is_empty(&self) -> bool {
+        self.info_list.len() == 0
+    }
+
+    /// Add a device configuration information object at the tail.
+    pub fn push(&mut self, info: DeviceConfigInfo<T>) {
+        self.info_list.push(info);
+    }
+
+    /// Iterator for configuration information objects.
+    pub fn iter(&self) -> std::slice::Iter<DeviceConfigInfo<T>> {
+        self.info_list.iter()
+    }
+
+    /// Mutable iterator for configuration information objects.
+    pub fn iter_mut(&mut self) -> std::slice::IterMut<DeviceConfigInfo<T>> {
+        self.info_list.iter_mut()
+    }
+
+    fn get_index_by_id(&self, config: &T) -> Option<usize> {
+        self.info_list
+            .iter()
+            .position(|info| info.config.id().eq(config.id()))
+    }
+}
+
+impl<T> Index<usize> for DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    type Output = DeviceConfigInfo<T>;
+    fn index(&self, idx: usize) -> &Self::Output {
+        &self.info_list[idx]
+    }
+}
+
+impl<T> IndexMut<usize> for DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    fn index_mut(&mut self, idx: usize) -> &mut Self::Output {
+        &mut self.info_list[idx]
+    }
+}
+
+impl<T> Clone for DeviceConfigInfos<T>
+where
+    T: ConfigItem + Clone,
+{
+    fn clone(&self) -> Self {
+        DeviceConfigInfos {
+            info_list: self.info_list.clone(),
+        }
+    }
+}
+
+/// Configuration information for RateLimiter token bucket.
+#[derive(Clone, Debug, Default, Deserialize, PartialEq, Serialize)]
+pub struct TokenBucketConfigInfo {
+    /// The size for the token bucket. A TokenBucket of `size` total capacity will take `refill_time`
+    /// milliseconds to go from zero tokens to total capacity.
+    pub size: u64,
+    /// Number of free initial tokens, that can be consumed at no cost.
+    pub one_time_burst: u64,
+    /// Complete refill time in milliseconds.
+    pub refill_time: u64,
+}
+
+impl TokenBucketConfigInfo {
+    fn resize(&mut self, n: u64) {
+        if n != 0 {
+            self.size /= n;
+            self.one_time_burst /= n;
+        }
+    }
+}
+
+impl From<TokenBucketConfigInfo> for TokenBucket {
+    fn from(t: TokenBucketConfigInfo) -> TokenBucket {
+        (&t).into()
+    }
+}
+
+impl From<&TokenBucketConfigInfo> for TokenBucket {
+    fn from(t: &TokenBucketConfigInfo) -> TokenBucket {
+        TokenBucket::new(t.size, t.one_time_burst, t.refill_time)
+    }
+}
+
+/// Configuration information for RateLimiter objects.
+#[derive(Clone, Debug, Default, Deserialize, PartialEq, Serialize)]
+pub struct RateLimiterConfigInfo {
+    /// Data used to initialize the RateLimiter::bandwidth bucket.
+    pub bandwidth: TokenBucketConfigInfo,
+    /// Data used to initialize the RateLimiter::ops bucket.
+    pub ops: TokenBucketConfigInfo,
+}
+
+impl RateLimiterConfigInfo {
+    /// Update the bandwidth budget configuration.
+    pub fn update_bandwidth(&mut self, new_config: TokenBucketConfigInfo) {
+        self.bandwidth = new_config;
+    }
+
+    /// Update the ops budget configuration.
+    pub fn update_ops(&mut self, new_config: TokenBucketConfigInfo) {
+        self.ops = new_config;
+    }
+
+    /// resize the limiter to its 1/n.
+    pub fn resize(&mut self, n: u64) {
+        self.bandwidth.resize(n);
+        self.ops.resize(n);
+    }
+}
+
+impl TryInto<RateLimiter> for &RateLimiterConfigInfo {
+    type Error = io::Error;
+
+    fn try_into(self) -> Result<RateLimiter, Self::Error> {
+        RateLimiter::new(
+            self.bandwidth.size,
+            self.bandwidth.one_time_burst,
+            self.bandwidth.refill_time,
+            self.ops.size,
+            self.ops.one_time_burst,
+            self.ops.refill_time,
+        )
+    }
+}
+
+impl TryInto<RateLimiter> for RateLimiterConfigInfo {
+    type Error = io::Error;
+
+    fn try_into(self) -> Result<RateLimiter, Self::Error> {
+        RateLimiter::new(
+            self.bandwidth.size,
+            self.bandwidth.one_time_burst,
+            self.bandwidth.refill_time,
+            self.ops.size,
+            self.ops.one_time_burst,
+            self.ops.refill_time,
+        )
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[derive(Debug, thiserror::Error)]
+    pub enum DummyError {
+        #[error("configuration entry exists")]
+        Exist,
+    }
+
+    #[derive(Clone, Debug, Default)]
+    pub struct DummyConfigInfo {
+        id: String,
+        content: String,
+    }
+
+    impl ConfigItem for DummyConfigInfo {
+        type Err = DummyError;
+
+        fn id(&self) -> &str {
+            &self.id
+        }
+
+        fn check_conflicts(&self, other: &Self) -> Result<(), DummyError> {
+            if self.id == other.id || self.content == other.content {
+                Err(DummyError::Exist)
+            } else {
+                Ok(())
+            }
+        }
+    }
+
+    type DummyConfigInfos = ConfigInfos<DummyConfigInfo>;
+
+    #[test]
+    fn test_insert_config_info() {
+        let mut configs = DummyConfigInfos::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert(config1).unwrap();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+
+        // Test case: cannot insert new item with the same id.
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.insert(config2).unwrap_err();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert(config3).unwrap();
+        assert_eq!(configs.configs.len(), 2);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+        assert_eq!(configs.configs[1].id, "2");
+        assert_eq!(configs.configs[1].content, "c");
+
+        // Test case: cannot insert new item with the same content.
+        let config4 = DummyConfigInfo {
+            id: "3".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert(config4).unwrap_err();
+        assert_eq!(configs.configs.len(), 2);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+        assert_eq!(configs.configs[1].id, "2");
+        assert_eq!(configs.configs[1].content, "c");
+    }
+
+    #[test]
+    fn test_update_config_info() {
+        let mut configs = DummyConfigInfos::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert(config1).unwrap();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+
+        // Test case: succeed to update an existing entry
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.update(config2, DummyError::Exist).unwrap();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+
+        // Test case: cannot update a non-existing entry
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.update(config3, DummyError::Exist).unwrap_err();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+
+        // Test case: cannot update an entry with conflicting content
+        let config4 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert(config4).unwrap();
+        let config5 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.update(config5, DummyError::Exist).unwrap_err();
+    }
+
+    #[test]
+    fn test_insert_or_update_config_info() {
+        let mut configs = DummyConfigInfos::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert_or_update(config1).unwrap();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "a");
+
+        // Test case: succeed to update an existing entry
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.insert_or_update(config2.clone()).unwrap();
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+
+        // Add a second entry
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(config3.clone()).unwrap();
+        assert_eq!(configs.configs.len(), 2);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+        assert_eq!(configs.configs[1].id, "2");
+        assert_eq!(configs.configs[1].content, "c");
+
+        // Lookup the first entry
+        let config4 = configs
+            .get_by_id(&DummyConfigInfo {
+                id: "1".to_owned(),
+                content: "b".to_owned(),
+            })
+            .unwrap();
+        assert_eq!(config4.id, config2.id);
+        assert_eq!(config4.content, config2.content);
+
+        // Lookup the second entry
+        let config5 = configs
+            .get_by_id(&DummyConfigInfo {
+                id: "2".to_owned(),
+                content: "c".to_owned(),
+            })
+            .unwrap();
+        assert_eq!(config5.id, config3.id);
+        assert_eq!(config5.content, config3.content);
+
+        // Test case: can't insert an entry with conflicting content
+        let config6 = DummyConfigInfo {
+            id: "3".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(config6).unwrap_err();
+        assert_eq!(configs.configs.len(), 2);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+        assert_eq!(configs.configs[1].id, "2");
+        assert_eq!(configs.configs[1].content, "c");
+    }
+
+    #[test]
+    fn test_remove_config_info() {
+        let mut configs = DummyConfigInfos::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert_or_update(config1).unwrap();
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.insert_or_update(config2.clone()).unwrap();
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(config3.clone()).unwrap();
+        assert_eq!(configs.configs.len(), 2);
+        assert_eq!(configs.configs[0].id, "1");
+        assert_eq!(configs.configs[0].content, "b");
+        assert_eq!(configs.configs[1].id, "2");
+        assert_eq!(configs.configs[1].content, "c");
+
+        let config4 = configs
+            .remove(&DummyConfigInfo {
+                id: "1".to_owned(),
+                content: "no value".to_owned(),
+            })
+            .unwrap();
+        assert_eq!(config4.id, config2.id);
+        assert_eq!(config4.content, config2.content);
+        assert_eq!(configs.configs.len(), 1);
+        assert_eq!(configs.configs[0].id, "2");
+        assert_eq!(configs.configs[0].content, "c");
+
+        let config5 = configs
+            .remove(&DummyConfigInfo {
+                id: "2".to_owned(),
+                content: "no value".to_owned(),
+            })
+            .unwrap();
+        assert_eq!(config5.id, config3.id);
+        assert_eq!(config5.content, config3.content);
+        assert_eq!(configs.configs.len(), 0);
+    }
+
+    type DummyDeviceInfoList = DeviceConfigInfos<DummyConfigInfo>;
+
+    #[test]
+    fn test_insert_or_update_device_info() {
+        let mut configs = DummyDeviceInfoList::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert_or_update(&config1).unwrap();
+        assert_eq!(configs.len(), 1);
+        assert_eq!(configs[0].config.id, "1");
+        assert_eq!(configs[0].config.content, "a");
+
+        // Test case: succeed to update an existing entry
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.insert_or_update(&config2 /*  */).unwrap();
+        assert_eq!(configs.len(), 1);
+        assert_eq!(configs[0].config.id, "1");
+        assert_eq!(configs[0].config.content, "b");
+
+        // Add a second entry
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(&config3).unwrap();
+        assert_eq!(configs.len(), 2);
+        assert_eq!(configs[0].config.id, "1");
+        assert_eq!(configs[0].config.content, "b");
+        assert_eq!(configs[1].config.id, "2");
+        assert_eq!(configs[1].config.content, "c");
+
+        // Lookup the first entry
+        let config4_id = configs
+            .get_index_by_id(&DummyConfigInfo {
+                id: "1".to_owned(),
+                content: "b".to_owned(),
+            })
+            .unwrap();
+        let config4 = &configs[config4_id].config;
+        assert_eq!(config4.id, config2.id);
+        assert_eq!(config4.content, config2.content);
+
+        // Lookup the second entry
+        let config5_id = configs
+            .get_index_by_id(&DummyConfigInfo {
+                id: "2".to_owned(),
+                content: "c".to_owned(),
+            })
+            .unwrap();
+        let config5 = &configs[config5_id].config;
+        assert_eq!(config5.id, config3.id);
+        assert_eq!(config5.content, config3.content);
+
+        // Test case: can't insert an entry with conflicting content
+        let config6 = DummyConfigInfo {
+            id: "3".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(&config6).unwrap_err();
+        assert_eq!(configs.len(), 2);
+        assert_eq!(configs[0].config.id, "1");
+        assert_eq!(configs[0].config.content, "b");
+        assert_eq!(configs[1].config.id, "2");
+        assert_eq!(configs[1].config.content, "c");
+    }
+
+    #[test]
+    fn test_remove_device_info() {
+        let mut configs = DummyDeviceInfoList::new();
+
+        let config1 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "a".to_owned(),
+        };
+        configs.insert_or_update(&config1).unwrap();
+        let config2 = DummyConfigInfo {
+            id: "1".to_owned(),
+            content: "b".to_owned(),
+        };
+        configs.insert_or_update(&config2).unwrap();
+        let config3 = DummyConfigInfo {
+            id: "2".to_owned(),
+            content: "c".to_owned(),
+        };
+        configs.insert_or_update(&config3).unwrap();
+        assert_eq!(configs.len(), 2);
+        assert_eq!(configs[0].config.id, "1");
+        assert_eq!(configs[0].config.content, "b");
+        assert_eq!(configs[1].config.id, "2");
+        assert_eq!(configs[1].config.content, "c");
+
+        let config4 = configs.remove(0).unwrap().config;
+        assert_eq!(config4.id, config2.id);
+        assert_eq!(config4.content, config2.content);
+        assert_eq!(configs.len(), 1);
+        assert_eq!(configs[0].config.id, "2");
+        assert_eq!(configs[0].config.content, "c");
+
+        let config5 = configs.remove(0).unwrap().config;
+        assert_eq!(config5.id, config3.id);
+        assert_eq!(config5.content, config3.content);
+        assert_eq!(configs.len(), 0);
+    }
+}

src/dragonball/src/device_manager/blk_dev_mgr.rs

@@ -0,0 +1,773 @@
+// Copyright 2020-2022 Alibaba, Inc. or its affiliates. All Rights Reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+//! Device manager for virtio-blk and vhost-user-blk devices.
+use std::collections::{vec_deque, VecDeque};
+use std::convert::TryInto;
+use std::fs::OpenOptions;
+use std::os::unix::fs::OpenOptionsExt;
+use std::os::unix::io::AsRawFd;
+use std::path::{Path, PathBuf};
+use std::sync::Arc;
+
+use dbs_virtio_devices as virtio;
+use dbs_virtio_devices::block::{aio::Aio, io_uring::IoUring, Block, LocalFile, Ufile};
+use serde_derive::{Deserialize, Serialize};
+
+use crate::address_space_manager::GuestAddressSpaceImpl;
+use crate::config_manager::{ConfigItem, DeviceConfigInfo, RateLimiterConfigInfo};
+use crate::device_manager::blk_dev_mgr::BlockDeviceError::InvalidDeviceId;
+use crate::device_manager::{DeviceManager, DeviceMgrError, DeviceOpContext};
+use crate::get_bucket_update;
+use crate::vm::KernelConfigInfo;
+
+use super::DbsMmioV2Device;
+
+// The flag of whether to use the shared irq.
+const USE_SHARED_IRQ: bool = true;
+// The flag of whether to use the generic irq.
+const USE_GENERIC_IRQ: bool = true;
+
+macro_rules! info(
+    ($l:expr, $($args:tt)+) => {
+        slog::info!($l, $($args)+; slog::o!("subsystem" => "block_manager"))
+    };
+);
+
+macro_rules! error(
+    ($l:expr, $($args:tt)+) => {
+        slog::error!($l, $($args)+; slog::o!("subsystem" => "block_manager"))
+    };
+);
+
+/// Default queue size for VirtIo block devices.
+pub const QUEUE_SIZE: u16 = 128;
+
+/// Errors associated with the operations allowed on a drive.
+#[derive(Debug, thiserror::Error)]
+pub enum BlockDeviceError {
+    /// Invalid VM instance ID.
+    #[error("invalid VM instance id")]
+    InvalidVMID,
+
+    /// The block device path is invalid.
+    #[error("invalid block device path '{0}'")]
+    InvalidBlockDevicePath(PathBuf),
+
+    /// The block device type is invalid.
+    #[error("invalid block device type")]
+    InvalidBlockDeviceType,
+
+    /// The block device path was already used for a different drive.
+    #[error("block device path '{0}' already exists")]
+    BlockDevicePathAlreadyExists(PathBuf),
+
+    /// The device id doesn't exist.
+    #[error("invalid block device id '{0}'")]
+    InvalidDeviceId(String),
+
+    /// Cannot perform the requested operation after booting the microVM.
+    #[error("block device does not support runtime update")]
+    UpdateNotAllowedPostBoot,
+
+    /// A root block device was already added.
+    #[error("could not add multiple virtual machine root devices")]
+    RootBlockDeviceAlreadyAdded,
+
+    /// Failed to send patch message to block epoll handler.
+    #[error("could not send patch message to the block epoll handler")]
+    BlockEpollHanderSendFail,
+
+    /// Failure from device manager,
+    #[error("device manager errors: {0}")]
+    DeviceManager(#[from] DeviceMgrError),
+
+    /// Failure from virtio subsystem.
+    #[error(transparent)]
+    Virtio(virtio::Error),
+
+    /// Unable to seek the block device backing file due to invalid permissions or
+    /// the file was deleted/corrupted.
+    #[error("cannot create block device: {0}")]
+    CreateBlockDevice(#[source] virtio::Error),
+
+    /// Cannot open the block device backing file.
+    #[error("cannot open the block device backing file: {0}")]
+    OpenBlockDevice(#[source] std::io::Error),
+
+    /// Cannot initialize a MMIO Block Device or add a device to the MMIO Bus.
+    #[error("failure while registering block device: {0}")]
+    RegisterBlockDevice(#[source] DeviceMgrError),
+}
+
+/// Type of low level storage device/protocol for virtio-blk devices.
+#[derive(Clone, Copy, Debug, PartialEq, Serialize, Deserialize)]
+pub enum BlockDeviceType {
+    /// Unknown low level device type.
+    Unknown,
+    /// Vhost-user-blk based low level device.
+    /// SPOOL is a reliable NVMe virtualization system for the cloud environment.
+    /// You could learn more SPOOL here: https://www.usenix.org/conference/atc20/presentation/xue
+    Spool,
+    /// Local disk/file based low level device.
+    RawBlock,
+}
+
+impl BlockDeviceType {
+    /// Get type of low level storage device/protocol by parsing `path`.
+    pub fn get_type(path: &str) -> BlockDeviceType {
+        // SPOOL path should be started with "spool", e.g. "spool:/device1"
+        if path.starts_with("spool:/") {
+            BlockDeviceType::Spool
+        } else {
+            BlockDeviceType::RawBlock
+        }
+    }
+}
+
+/// Configuration information for a block device.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct BlockDeviceConfigUpdateInfo {
+    /// Unique identifier of the drive.
+    pub drive_id: String,
+    /// Rate Limiter for I/O operations.
+    pub rate_limiter: Option<RateLimiterConfigInfo>,
+}
+
+impl BlockDeviceConfigUpdateInfo {
+    /// Provides a `BucketUpdate` description for the bandwidth rate limiter.
+    pub fn bytes(&self) -> dbs_utils::rate_limiter::BucketUpdate {
+        get_bucket_update!(self, rate_limiter, bandwidth)
+    }
+    /// Provides a `BucketUpdate` description for the ops rate limiter.
+    pub fn ops(&self) -> dbs_utils::rate_limiter::BucketUpdate {
+        get_bucket_update!(self, rate_limiter, ops)
+    }
+}
+
+/// Configuration information for a block device.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct BlockDeviceConfigInfo {
+    /// Unique identifier of the drive.
+    pub drive_id: String,
+    /// Type of low level storage/protocol.
+    pub device_type: BlockDeviceType,
+    /// Path of the drive.
+    pub path_on_host: PathBuf,
+    /// If set to true, it makes the current device the root block device.
+    /// Setting this flag to true will mount the block device in the
+    /// guest under /dev/vda unless the part_uuid is present.
+    pub is_root_device: bool,
+    /// Part-UUID. Represents the unique id of the boot partition of this device.
+    /// It is optional and it will be used only if the `is_root_device` field is true.
+    pub part_uuid: Option<String>,
+    /// If set to true, the drive is opened in read-only mode. Otherwise, the
+    /// drive is opened as read-write.
+    pub is_read_only: bool,
+    /// If set to false, the drive is opened with buffered I/O mode. Otherwise, the
+    /// drive is opened with direct I/O mode.
+    pub is_direct: bool,
+    /// Don't close `path_on_host` file when dropping the device.
+    pub no_drop: bool,
+    /// Block device multi-queue
+    pub num_queues: usize,
+    /// Virtio queue size. Size: byte
+    pub queue_size: u16,
+    /// Rate Limiter for I/O operations.
+    pub rate_limiter: Option<RateLimiterConfigInfo>,
+    /// Use shared irq
+    pub use_shared_irq: Option<bool>,
+    /// Use generic irq
+    pub use_generic_irq: Option<bool>,
+}
+
+impl std::default::Default for BlockDeviceConfigInfo {
+    fn default() -> Self {
+        Self {
+            drive_id: String::default(),
+            device_type: BlockDeviceType::RawBlock,
+            path_on_host: PathBuf::default(),
+            is_root_device: false,
+            part_uuid: None,
+            is_read_only: false,
+            is_direct: Self::default_direct(),
+            no_drop: Self::default_no_drop(),
+            num_queues: Self::default_num_queues(),
+            queue_size: 256,
+            rate_limiter: None,
+            use_shared_irq: None,
+            use_generic_irq: None,
+        }
+    }
+}
+
+impl BlockDeviceConfigInfo {
+    /// Get default queue numbers
+    pub fn default_num_queues() -> usize {
+        1
+    }
+
+    /// Get default value of is_direct switch
+    pub fn default_direct() -> bool {
+        true
+    }
+
+    /// Get default value of no_drop switch
+    pub fn default_no_drop() -> bool {
+        false
+    }
+
+    /// Get type of low level storage/protocol.
+    pub fn device_type(&self) -> BlockDeviceType {
+        self.device_type
+    }
+
+    /// Returns a reference to `path_on_host`.
+    pub fn path_on_host(&self) -> &PathBuf {
+        &self.path_on_host
+    }
+
+    /// Returns a reference to the part_uuid.
+    pub fn get_part_uuid(&self) -> Option<&String> {
+        self.part_uuid.as_ref()
+    }
+
+    /// Checks whether the drive had read only permissions.
+    pub fn is_read_only(&self) -> bool {
+        self.is_read_only
+    }
+
+    /// Checks whether the drive uses direct I/O
+    pub fn is_direct(&self) -> bool {
+        self.is_direct
+    }
+
+    /// Get number and size of queues supported.
+    pub fn queue_sizes(&self) -> Vec<u16> {
+        (0..self.num_queues)
+            .map(|_| self.queue_size)
+            .collect::<Vec<u16>>()
+    }
+}
+
+impl ConfigItem for BlockDeviceConfigInfo {
+    type Err = BlockDeviceError;
+
+    fn id(&self) -> &str {
+        &self.drive_id
+    }
+
+    fn check_conflicts(&self, other: &Self) -> Result<(), BlockDeviceError> {
+        if self.drive_id == other.drive_id {
+            Ok(())
+        } else if self.path_on_host == other.path_on_host {
+            Err(BlockDeviceError::BlockDevicePathAlreadyExists(
+                self.path_on_host.clone(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+}
+
+impl std::fmt::Debug for BlockDeviceInfo {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{:?}", self.config)
+    }
+}
+
+/// Block Device Info
+pub type BlockDeviceInfo = DeviceConfigInfo<BlockDeviceConfigInfo>;
+
+/// Wrapper for the collection that holds all the Block Devices Configs
+//#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+#[derive(Clone)]
+pub struct BlockDeviceMgr {
+    /// A list of `BlockDeviceInfo` objects.
+    info_list: VecDeque<BlockDeviceInfo>,
+    has_root_block: bool,
+    has_part_uuid_root: bool,
+    read_only_root: bool,
+    part_uuid: Option<String>,
+    use_shared_irq: bool,
+}
+
+impl BlockDeviceMgr {
+    /// returns a front-to-back iterator.
+    pub fn iter(&self) -> vec_deque::Iter<BlockDeviceInfo> {
+        self.info_list.iter()
+    }
+
+    /// Checks whether any of the added BlockDevice is the root.
+    pub fn has_root_block_device(&self) -> bool {
+        self.has_root_block
+    }
+
+    /// Checks whether the root device is configured using a part UUID.
+    pub fn has_part_uuid_root(&self) -> bool {
+        self.has_part_uuid_root
+    }
+
+    /// Checks whether the root device has read-only permisssions.
+    pub fn is_read_only_root(&self) -> bool {
+        self.read_only_root
+    }
+
+    /// Gets the index of the device with the specified `drive_id` if it exists in the list.
+    pub fn get_index_of_drive_id(&self, id: &str) -> Option<usize> {
+        self.info_list
+            .iter()
+            .position(|info| info.config.id().eq(id))
+    }
+
+    /// Gets the 'BlockDeviceConfigInfo' of the device with the specified `drive_id` if it exists in the list.
+    pub fn get_config_of_drive_id(&self, drive_id: &str) -> Option<BlockDeviceConfigInfo> {
+        match self.get_index_of_drive_id(drive_id) {
+            Some(index) => {
+                let config = self.info_list.get(index).unwrap().config.clone();
+                Some(config)
+            }
+            None => None,
+        }
+    }
+
+    /// Inserts `block_device_config` in the block device configuration list.
+    /// If an entry with the same id already exists, it will attempt to update
+    /// the existing entry.
+    /// Inserting a secondary root block device will fail.
+    pub fn insert_device(
+        device_mgr: &mut DeviceManager,
+        mut ctx: DeviceOpContext,
+        config: BlockDeviceConfigInfo,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        if !cfg!(feature = "hotplug") && ctx.is_hotplug {
+            return Err(BlockDeviceError::UpdateNotAllowedPostBoot);
+        }
+
+        let mgr = &mut device_mgr.block_manager;
+
+        // If the id of the drive already exists in the list, the operation is update.
+        match mgr.get_index_of_drive_id(config.id()) {
+            Some(index) => {
+                // No support for runtime update yet.
+                if ctx.is_hotplug {
+                    Err(BlockDeviceError::BlockDevicePathAlreadyExists(
+                        config.path_on_host.clone(),
+                    ))
+                } else {
+                    for (idx, info) in mgr.info_list.iter().enumerate() {
+                        if idx != index {
+                            info.config.check_conflicts(&config)?;
+                        }
+                    }
+                    mgr.update(index, config)
+                }
+            }
+            None => {
+                for info in mgr.info_list.iter() {
+                    info.config.check_conflicts(&config)?;
+                }
+                let index = mgr.create(config.clone())?;
+                if !ctx.is_hotplug {
+                    return Ok(());
+                }
+
+                match config.device_type {
+                    BlockDeviceType::RawBlock => {
+                        let device = Self::create_blk_device(&config, &mut ctx)
+                            .map_err(BlockDeviceError::Virtio)?;
+                        let dev = DeviceManager::create_mmio_virtio_device(
+                            device,
+                            &mut ctx,
+                            config.use_shared_irq.unwrap_or(mgr.use_shared_irq),
+                            config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+                        )
+                        .map_err(BlockDeviceError::DeviceManager)?;
+                        mgr.update_device_by_index(index, Arc::clone(&dev))?;
+                        // live-upgrade need save/restore device from info.device.
+                        mgr.info_list[index].set_device(dev.clone());
+                        ctx.insert_hotplug_mmio_device(&dev, None).map_err(|e| {
+                            let logger = ctx.logger().new(slog::o!());
+                            BlockDeviceMgr::remove_device(device_mgr, ctx, &config.drive_id)
+                                .unwrap();
+                            error!(
+                                logger,
+                                "failed to hot-add virtio block device {}, {:?}",
+                                &config.drive_id,
+                                e
+                            );
+                            BlockDeviceError::DeviceManager(e)
+                        })
+                    }
+                    _ => Err(BlockDeviceError::InvalidBlockDeviceType),
+                }
+            }
+        }
+    }
+
+    /// Attaches all block devices from the BlockDevicesConfig.
+    pub fn attach_devices(
+        &mut self,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        for info in self.info_list.iter_mut() {
+            match info.config.device_type {
+                BlockDeviceType::RawBlock => {
+                    info!(
+                        ctx.logger(),
+                        "attach virtio-blk device, drive_id {}, path {}",
+                        info.config.drive_id,
+                        info.config.path_on_host.to_str().unwrap_or("<unknown>")
+                    );
+                    let device = Self::create_blk_device(&info.config, ctx)
+                        .map_err(BlockDeviceError::Virtio)?;
+                    let device = DeviceManager::create_mmio_virtio_device(
+                        device,
+                        ctx,
+                        info.config.use_shared_irq.unwrap_or(self.use_shared_irq),
+                        info.config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+                    )
+                    .map_err(BlockDeviceError::RegisterBlockDevice)?;
+                    info.device = Some(device);
+                }
+                _ => {
+                    return Err(BlockDeviceError::OpenBlockDevice(
+                        std::io::Error::from_raw_os_error(libc::EINVAL),
+                    ));
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Removes all virtio-blk devices
+    pub fn remove_devices(&mut self, ctx: &mut DeviceOpContext) -> Result<(), DeviceMgrError> {
+        while let Some(mut info) = self.info_list.pop_back() {
+            info!(ctx.logger(), "remove drive {}", info.config.drive_id);
+            if let Some(device) = info.device.take() {
+                DeviceManager::destroy_mmio_virtio_device(device, ctx)?;
+            }
+        }
+
+        Ok(())
+    }
+
+    fn remove(&mut self, drive_id: &str) -> Option<BlockDeviceInfo> {
+        match self.get_index_of_drive_id(drive_id) {
+            Some(index) => self.info_list.remove(index),
+            None => None,
+        }
+    }
+
+    /// remove a block device, it basically is the inverse operation of `insert_device``
+    pub fn remove_device(
+        dev_mgr: &mut DeviceManager,
+        mut ctx: DeviceOpContext,
+        drive_id: &str,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        if !cfg!(feature = "hotplug") {
+            return Err(BlockDeviceError::UpdateNotAllowedPostBoot);
+        }
+
+        let mgr = &mut dev_mgr.block_manager;
+        match mgr.remove(drive_id) {
+            Some(mut info) => {
+                info!(ctx.logger(), "remove drive {}", info.config.drive_id);
+                if let Some(device) = info.device.take() {
+                    DeviceManager::destroy_mmio_virtio_device(device, &mut ctx)
+                        .map_err(BlockDeviceError::DeviceManager)?;
+                }
+            }
+            None => return Err(BlockDeviceError::InvalidDeviceId(drive_id.to_owned())),
+        }
+
+        Ok(())
+    }
+
+    fn create_blk_device(
+        cfg: &BlockDeviceConfigInfo,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<Box<Block<GuestAddressSpaceImpl>>, virtio::Error> {
+        let epoll_mgr = ctx.epoll_mgr.clone().ok_or(virtio::Error::InvalidInput)?;
+
+        let mut block_files: Vec<Box<dyn Ufile>> = vec![];
+
+        match cfg.device_type {
+            BlockDeviceType::RawBlock => {
+                let custom_flags = if cfg.is_direct() {
+                    info!(
+                        ctx.logger(),
+                        "Open block device \"{}\" in direct mode.",
+                        cfg.path_on_host().display()
+                    );
+                    libc::O_DIRECT
+                } else {
+                    info!(
+                        ctx.logger(),
+                        "Open block device \"{}\" in buffer mode.",
+                        cfg.path_on_host().display(),
+                    );
+                    0
+                };
+                let io_uring_supported = IoUring::is_supported();
+                for i in 0..cfg.num_queues {
+                    let queue_size = cfg.queue_sizes()[i] as u32;
+                    let file = OpenOptions::new()
+                        .read(true)
+                        .custom_flags(custom_flags)
+                        .write(!cfg.is_read_only())
+                        .open(cfg.path_on_host())?;
+                    info!(ctx.logger(), "Queue {}: block file opened", i);
+
+                    if io_uring_supported {
+                        info!(
+                            ctx.logger(),
+                            "Queue {}: Using io_uring Raw disk file, queue size {}.", i, queue_size
+                        );
+                        let io_engine = IoUring::new(file.as_raw_fd(), queue_size)?;
+                        block_files.push(Box::new(LocalFile::new(file, cfg.no_drop, io_engine)?));
+                    } else {
+                        info!(
+                            ctx.logger(),
+                            "Queue {}: Since io_uring_supported is not enabled, change to default support of Aio Raw disk file, queue size {}", i, queue_size
+                        );
+                        let io_engine = Aio::new(file.as_raw_fd(), queue_size)?;
+                        block_files.push(Box::new(LocalFile::new(file, cfg.no_drop, io_engine)?));
+                    }
+                }
+            }
+            _ => {
+                error!(
+                    ctx.logger(),
+                    "invalid block device type: {:?}", cfg.device_type
+                );
+                return Err(virtio::Error::InvalidInput);
+            }
+        };
+
+        let mut limiters = vec![];
+        for _i in 0..cfg.num_queues {
+            if let Some(limiter) = cfg.rate_limiter.clone().map(|mut v| {
+                v.resize(cfg.num_queues as u64);
+                v.try_into().unwrap()
+            }) {
+                limiters.push(limiter);
+            }
+        }
+
+        Ok(Box::new(Block::new(
+            block_files,
+            cfg.is_read_only,
+            Arc::new(cfg.queue_sizes()),
+            epoll_mgr,
+            limiters,
+        )?))
+    }
+
+    /// Generated guest kernel commandline related to root block device.
+    pub fn generate_kernel_boot_args(
+        &self,
+        kernel_config: &mut KernelConfigInfo,
+    ) -> std::result::Result<(), DeviceMgrError> {
+        // Respect user configuration if kernel_cmdline contains "root=",
+        // special attention for the case when kernel command line starting with "root=xxx"
+        let old_kernel_cmdline = format!(" {}", kernel_config.kernel_cmdline().as_str());
+        if !old_kernel_cmdline.contains(" root=") && self.has_root_block {
+            let cmdline = kernel_config.kernel_cmdline_mut();
+            if let Some(ref uuid) = self.part_uuid {
+                cmdline
+                    .insert("root", &format!("PART_UUID={}", uuid))
+                    .map_err(DeviceMgrError::Cmdline)?;
+            } else {
+                cmdline
+                    .insert("root", "/dev/vda")
+                    .map_err(DeviceMgrError::Cmdline)?;
+            }
+            if self.read_only_root {
+                if old_kernel_cmdline.contains(" rw") {
+                    return Err(DeviceMgrError::InvalidOperation);
+                }
+                cmdline.insert_str("ro").map_err(DeviceMgrError::Cmdline)?;
+            }
+        }
+
+        Ok(())
+    }
+
+    /// insert a block device's config. return index on success.
+    fn create(
+        &mut self,
+        block_device_config: BlockDeviceConfigInfo,
+    ) -> std::result::Result<usize, BlockDeviceError> {
+        self.check_data_file_present(&block_device_config)?;
+        if self
+            .get_index_of_drive_path(&block_device_config.path_on_host)
+            .is_some()
+        {
+            return Err(BlockDeviceError::BlockDevicePathAlreadyExists(
+                block_device_config.path_on_host,
+            ));
+        }
+
+        // check whether the Device Config belongs to a root device
+        // we need to satisfy the condition by which a VMM can only have on root device
+        if block_device_config.is_root_device {
+            if self.has_root_block {
+                return Err(BlockDeviceError::RootBlockDeviceAlreadyAdded);
+            } else {
+                self.has_root_block = true;
+                self.read_only_root = block_device_config.is_read_only;
+                self.has_part_uuid_root = block_device_config.part_uuid.is_some();
+                self.part_uuid = block_device_config.part_uuid.clone();
+                // Root Device should be the first in the list whether or not PART_UUID is specified
+                // in order to avoid bugs in case of switching from part_uuid boot scenarios to
+                // /dev/vda boot type.
+                self.info_list
+                    .push_front(BlockDeviceInfo::new(block_device_config));
+                Ok(0)
+            }
+        } else {
+            self.info_list
+                .push_back(BlockDeviceInfo::new(block_device_config));
+            Ok(self.info_list.len() - 1)
+        }
+    }
+
+    /// Updates a Block Device Config. The update fails if it would result in two
+    /// root block devices.
+    fn update(
+        &mut self,
+        mut index: usize,
+        new_config: BlockDeviceConfigInfo,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        // Check if the path exists
+        self.check_data_file_present(&new_config)?;
+        if let Some(idx) = self.get_index_of_drive_path(&new_config.path_on_host) {
+            if idx != index {
+                return Err(BlockDeviceError::BlockDevicePathAlreadyExists(
+                    new_config.path_on_host.clone(),
+                ));
+            }
+        }
+
+        if self.info_list.get(index).is_none() {
+            return Err(InvalidDeviceId(index.to_string()));
+        }
+        // Check if the root block device is being updated.
+        if self.info_list[index].config.is_root_device {
+            self.has_root_block = new_config.is_root_device;
+            self.read_only_root = new_config.is_root_device && new_config.is_read_only;
+            self.has_part_uuid_root = new_config.part_uuid.is_some();
+            self.part_uuid = new_config.part_uuid.clone();
+        } else if new_config.is_root_device {
+            // Check if a second root block device is being added.
+            if self.has_root_block {
+                return Err(BlockDeviceError::RootBlockDeviceAlreadyAdded);
+            } else {
+                // One of the non-root blocks is becoming root.
+                self.has_root_block = true;
+                self.read_only_root = new_config.is_read_only;
+                self.has_part_uuid_root = new_config.part_uuid.is_some();
+                self.part_uuid = new_config.part_uuid.clone();
+
+                // Make sure the root device is on the first position.
+                self.info_list.swap(0, index);
+                // Block config to be updated has moved to first position.
+                index = 0;
+            }
+        }
+        // Update the config.
+        self.info_list[index].config = new_config;
+
+        Ok(())
+    }
+
+    fn check_data_file_present(
+        &self,
+        block_device_config: &BlockDeviceConfigInfo,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        if block_device_config.device_type == BlockDeviceType::RawBlock
+            && !block_device_config.path_on_host.exists()
+        {
+            Err(BlockDeviceError::InvalidBlockDevicePath(
+                block_device_config.path_on_host.clone(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+
+    fn get_index_of_drive_path(&self, drive_path: &Path) -> Option<usize> {
+        self.info_list
+            .iter()
+            .position(|info| info.config.path_on_host.eq(drive_path))
+    }
+
+    /// update devce information in `info_list`. The caller of this method is
+    /// `insert_device` when hotplug is true.
+    pub fn update_device_by_index(
+        &mut self,
+        index: usize,
+        device: Arc<DbsMmioV2Device>,
+    ) -> Result<(), BlockDeviceError> {
+        if let Some(info) = self.info_list.get_mut(index) {
+            info.device = Some(device);
+            return Ok(());
+        }
+
+        Err(BlockDeviceError::InvalidDeviceId("".to_owned()))
+    }
+
+    /// Update the ratelimiter settings of a virtio blk device.
+    pub fn update_device_ratelimiters(
+        device_mgr: &mut DeviceManager,
+        new_cfg: BlockDeviceConfigUpdateInfo,
+    ) -> std::result::Result<(), BlockDeviceError> {
+        let mgr = &mut device_mgr.block_manager;
+        match mgr.get_index_of_drive_id(&new_cfg.drive_id) {
+            Some(index) => {
+                let config = &mut mgr.info_list[index].config;
+                config.rate_limiter = new_cfg.rate_limiter.clone();
+                let device = mgr.info_list[index]
+                    .device
+                    .as_mut()
+                    .ok_or_else(|| BlockDeviceError::InvalidDeviceId("".to_owned()))?;
+                if let Some(mmio_dev) = device.as_any().downcast_ref::<DbsMmioV2Device>() {
+                    let guard = mmio_dev.state();
+                    let inner_dev = guard.get_inner_device();
+                    if let Some(blk_dev) = inner_dev
+                        .as_any()
+                        .downcast_ref::<virtio::block::Block<GuestAddressSpaceImpl>>()
+                    {
+                        return blk_dev
+                            .set_patch_rate_limiters(new_cfg.bytes(), new_cfg.ops())
+                            .map(|_p| ())
+                            .map_err(|_e| BlockDeviceError::BlockEpollHanderSendFail);
+                    }
+                }
+                Ok(())
+            }
+            None => Err(BlockDeviceError::InvalidDeviceId(new_cfg.drive_id)),
+        }
+    }
+}
+
+impl Default for BlockDeviceMgr {
+    /// Constructor for the BlockDeviceMgr. It initializes an empty LinkedList.
+    fn default() -> BlockDeviceMgr {
+        BlockDeviceMgr {
+            info_list: VecDeque::<BlockDeviceInfo>::new(),
+            has_root_block: false,
+            has_part_uuid_root: false,
+            read_only_root: false,
+            part_uuid: None,
+            use_shared_irq: USE_SHARED_IRQ,
+        }
+    }
+}

src/dragonball/src/device_manager/console_manager.rs

@@ -0,0 +1,440 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+//! Virtual machine console device manager.
+//!
+//! A virtual console are composed up of two parts: frontend in virtual machine and backend in
+//! host OS. A frontend may be serial port, virtio-console etc, a backend may be stdio or Unix
+//! domain socket. The manager connects the frontend with the backend.
+use std::io::{self, Read};
+use std::os::unix::net::{UnixListener, UnixStream};
+use std::path::Path;
+use std::sync::{Arc, Mutex};
+
+use bytes::{BufMut, BytesMut};
+use dbs_legacy_devices::{ConsoleHandler, SerialDevice};
+use dbs_utils::epoll_manager::{
+    EpollManager, EventOps, EventSet, Events, MutEventSubscriber, SubscriberId,
+};
+use vmm_sys_util::terminal::Terminal;
+
+use super::{DeviceMgrError, Result};
+
+const EPOLL_EVENT_SERIAL: u32 = 0;
+const EPOLL_EVENT_SERIAL_DATA: u32 = 1;
+const EPOLL_EVENT_STDIN: u32 = 2;
+// Maximal backend throughput for every data transaction.
+const MAX_BACKEND_THROUGHPUT: usize = 64;
+
+/// Errors related to Console manager operations.
+#[derive(Debug, thiserror::Error)]
+pub enum ConsoleManagerError {
+    /// Cannot create unix domain socket for serial port
+    #[error("cannot create socket for serial console")]
+    CreateSerialSock(#[source] std::io::Error),
+
+    /// An operation on the epoll instance failed due to resource exhaustion or bad configuration.
+    #[error("failure while managing epoll event for console fd")]
+    EpollMgr(#[source] dbs_utils::epoll_manager::Error),
+
+    /// Cannot set mode for terminal.
+    #[error("failure while setting attribute for terminal")]
+    StdinHandle(#[source] vmm_sys_util::errno::Error),
+}
+
+enum Backend {
+    StdinHandle(std::io::Stdin),
+    SockPath(String),
+}
+
+/// Console manager to manage frontend and backend console devices.
+pub struct ConsoleManager {
+    epoll_mgr: EpollManager,
+    logger: slog::Logger,
+    subscriber_id: Option<SubscriberId>,
+    backend: Option<Backend>,
+}
+
+impl ConsoleManager {
+    /// Create a console manager instance.
+    pub fn new(epoll_mgr: EpollManager, logger: &slog::Logger) -> Self {
+        let logger = logger.new(slog::o!("subsystem" => "console_manager"));
+        ConsoleManager {
+            epoll_mgr,
+            logger,
+            subscriber_id: Default::default(),
+            backend: None,
+        }
+    }
+
+    /// Create a console backend device by using stdio streams.
+    pub fn create_stdio_console(&mut self, device: Arc<Mutex<SerialDevice>>) -> Result<()> {
+        let stdin_handle = std::io::stdin();
+        stdin_handle
+            .lock()
+            .set_raw_mode()
+            .map_err(|e| DeviceMgrError::ConsoleManager(ConsoleManagerError::StdinHandle(e)))?;
+
+        let handler = ConsoleEpollHandler::new(device, Some(stdin_handle), None, &self.logger);
+        self.subscriber_id = Some(self.epoll_mgr.add_subscriber(Box::new(handler)));
+        self.backend = Some(Backend::StdinHandle(std::io::stdin()));
+
+        Ok(())
+    }
+
+    /// Create s console backend device by using Unix Domain socket.
+    pub fn create_socket_console(
+        &mut self,
+        device: Arc<Mutex<SerialDevice>>,
+        sock_path: String,
+    ) -> Result<()> {
+        let sock_listener = Self::bind_domain_socket(&sock_path).map_err(|e| {
+            DeviceMgrError::ConsoleManager(ConsoleManagerError::CreateSerialSock(e))
+        })?;
+        let handler = ConsoleEpollHandler::new(device, None, Some(sock_listener), &self.logger);
+
+        self.subscriber_id = Some(self.epoll_mgr.add_subscriber(Box::new(handler)));
+        self.backend = Some(Backend::SockPath(sock_path));
+
+        Ok(())
+    }
+
+    /// Reset the host side terminal to canonical mode.
+    pub fn reset_console(&self) -> Result<()> {
+        if let Some(Backend::StdinHandle(stdin_handle)) = self.backend.as_ref() {
+            stdin_handle
+                .lock()
+                .set_canon_mode()
+                .map_err(|e| DeviceMgrError::ConsoleManager(ConsoleManagerError::StdinHandle(e)))?;
+        }
+
+        Ok(())
+    }
+
+    fn bind_domain_socket(serial_path: &str) -> std::result::Result<UnixListener, std::io::Error> {
+        let path = Path::new(serial_path);
+        if path.is_file() {
+            let _ = std::fs::remove_file(serial_path);
+        }
+
+        UnixListener::bind(path)
+    }
+}
+
+struct ConsoleEpollHandler {
+    device: Arc<Mutex<SerialDevice>>,
+    stdin_handle: Option<std::io::Stdin>,
+    sock_listener: Option<UnixListener>,
+    sock_conn: Option<UnixStream>,
+    logger: slog::Logger,
+}
+
+impl ConsoleEpollHandler {
+    fn new(
+        device: Arc<Mutex<SerialDevice>>,
+        stdin_handle: Option<std::io::Stdin>,
+        sock_listener: Option<UnixListener>,
+        logger: &slog::Logger,
+    ) -> Self {
+        ConsoleEpollHandler {
+            device,
+            stdin_handle,
+            sock_listener,
+            sock_conn: None,
+            logger: logger.new(slog::o!("subsystem" => "console_manager")),
+        }
+    }
+
+    fn uds_listener_accept(&mut self, ops: &mut EventOps) -> std::io::Result<()> {
+        if self.sock_conn.is_some() {
+            slog::warn!(self.logger,
+                "UDS for serial port 1 already exists, reject the new connection";
+                "subsystem" => "console_mgr",
+            );
+            // Do not expected poisoned lock.
+            let _ = self.sock_listener.as_mut().unwrap().accept();
+        } else {
+            // Safe to unwrap() because self.sock_conn is Some().
+            let (conn_sock, _) = self.sock_listener.as_ref().unwrap().accept()?;
+            let events = Events::with_data(&conn_sock, EPOLL_EVENT_SERIAL_DATA, EventSet::IN);
+            if let Err(e) = ops.add(events) {
+                slog::error!(self.logger,
+                    "failed to register epoll event for serial, {:?}", e;
+                    "subsystem" => "console_mgr",
+                );
+                return Err(std::io::Error::last_os_error());
+            }
+
+            let conn_sock_copy = conn_sock.try_clone()?;
+            // Do not expected poisoned lock.
+            self.device
+                .lock()
+                .unwrap()
+                .set_output_stream(Some(Box::new(conn_sock_copy)));
+
+            self.sock_conn = Some(conn_sock);
+        }
+
+        Ok(())
+    }
+
+    fn uds_read_in(&mut self, ops: &mut EventOps) -> std::io::Result<()> {
+        let mut should_drop = true;
+
+        if let Some(conn_sock) = self.sock_conn.as_mut() {
+            let mut out = [0u8; MAX_BACKEND_THROUGHPUT];
+            match conn_sock.read(&mut out[..]) {
+                Ok(0) => {
+                    // Zero-length read means EOF. Remove this conn sock.
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .set_output_stream(None);
+                }
+                Ok(count) => {
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .raw_input(&out[..count])?;
+                    should_drop = false;
+                }
+                Err(e) => {
+                    slog::warn!(self.logger,
+                        "error while reading serial conn sock: {:?}", e;
+                        "subsystem" => "console_mgr"
+                    );
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .set_output_stream(None);
+                }
+            }
+        }
+
+        if should_drop {
+            assert!(self.sock_conn.is_some());
+            // Safe to unwrap() because self.sock_conn is Some().
+            let sock_conn = self.sock_conn.take().unwrap();
+            let events = Events::with_data(&sock_conn, EPOLL_EVENT_SERIAL_DATA, EventSet::IN);
+            if let Err(e) = ops.remove(events) {
+                slog::error!(self.logger,
+                    "failed deregister epoll event for UDS, {:?}", e;
+                    "subsystem" => "console_mgr"
+                );
+            }
+        }
+
+        Ok(())
+    }
+
+    fn stdio_read_in(&mut self, ops: &mut EventOps) -> std::io::Result<()> {
+        let mut should_drop = true;
+
+        if let Some(handle) = self.stdin_handle.as_ref() {
+            let mut out = [0u8; MAX_BACKEND_THROUGHPUT];
+            // Safe to unwrap() because self.stdin_handle is Some().
+            let stdin_lock = handle.lock();
+            match stdin_lock.read_raw(&mut out[..]) {
+                Ok(0) => {
+                    // Zero-length read indicates EOF. Remove from pollables.
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .set_output_stream(None);
+                }
+                Ok(count) => {
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .raw_input(&out[..count])?;
+                    should_drop = false;
+                }
+                Err(e) => {
+                    slog::warn!(self.logger,
+                        "error while reading stdin: {:?}", e;
+                        "subsystem" => "console_mgr"
+                    );
+                    self.device
+                        .lock()
+                        .expect("console: poisoned console lock")
+                        .set_output_stream(None);
+                }
+            }
+        }
+
+        if should_drop {
+            let events = Events::with_data_raw(libc::STDIN_FILENO, EPOLL_EVENT_STDIN, EventSet::IN);
+            if let Err(e) = ops.remove(events) {
+                slog::error!(self.logger,
+                    "failed to deregister epoll event for stdin, {:?}", e;
+                    "subsystem" => "console_mgr"
+                );
+            }
+        }
+
+        Ok(())
+    }
+}
+
+impl MutEventSubscriber for ConsoleEpollHandler {
+    fn process(&mut self, events: Events, ops: &mut EventOps) {
+        slog::trace!(self.logger, "ConsoleEpollHandler::process()");
+        let slot = events.data();
+        match slot {
+            EPOLL_EVENT_SERIAL => {
+                if let Err(e) = self.uds_listener_accept(ops) {
+                    slog::warn!(self.logger, "failed to accept incoming connection, {:?}", e);
+                }
+            }
+            EPOLL_EVENT_SERIAL_DATA => {
+                if let Err(e) = self.uds_read_in(ops) {
+                    slog::warn!(self.logger, "failed to read data from UDS, {:?}", e);
+                }
+            }
+            EPOLL_EVENT_STDIN => {
+                if let Err(e) = self.stdio_read_in(ops) {
+                    slog::warn!(self.logger, "failed to read data from stdin, {:?}", e);
+                }
+            }
+            _ => slog::error!(self.logger, "unknown epoll slot number {}", slot),
+        }
+    }
+
+    fn init(&mut self, ops: &mut EventOps) {
+        slog::trace!(self.logger, "ConsoleEpollHandler::init()");
+
+        if self.stdin_handle.is_some() {
+            slog::info!(self.logger, "ConsoleEpollHandler: stdin handler");
+            let events = Events::with_data_raw(libc::STDIN_FILENO, EPOLL_EVENT_STDIN, EventSet::IN);
+            if let Err(e) = ops.add(events) {
+                slog::error!(
+                    self.logger,
+                    "failed to register epoll event for stdin, {:?}",
+                    e
+                );
+            }
+        }
+        if let Some(sock) = self.sock_listener.as_ref() {
+            slog::info!(self.logger, "ConsoleEpollHandler: sock listener");
+            let events = Events::with_data(sock, EPOLL_EVENT_SERIAL, EventSet::IN);
+            if let Err(e) = ops.add(events) {
+                slog::error!(
+                    self.logger,
+                    "failed to register epoll event for UDS listener, {:?}",
+                    e
+                );
+            }
+        }
+
+        if let Some(conn) = self.sock_conn.as_ref() {
+            slog::info!(self.logger, "ConsoleEpollHandler: sock connection");
+            let events = Events::with_data(conn, EPOLL_EVENT_SERIAL_DATA, EventSet::IN);
+            if let Err(e) = ops.add(events) {
+                slog::error!(
+                    self.logger,
+                    "failed to register epoll event for UDS connection, {:?}",
+                    e
+                );
+            }
+        }
+    }
+}
+
+/// Writer to process guest kernel dmesg.
+pub struct DmesgWriter {
+    buf: BytesMut,
+    logger: slog::Logger,
+}
+
+impl DmesgWriter {
+    /// Creates a new instance.
+    pub fn new(logger: &slog::Logger) -> Self {
+        Self {
+            buf: BytesMut::with_capacity(1024),
+            logger: logger.new(slog::o!("subsystem" => "dmesg")),
+        }
+    }
+}
+
+impl io::Write for DmesgWriter {
+    /// 0000000   [                   0   .   0   3   4   9   1   6   ]       R
+    ///          5b  20  20  20  20  30  2e  30  33  34  39  31  36  5d  20  52
+    /// 0000020   u   n       /   s   b   i   n   /   i   n   i   t       a   s
+    ///          75  6e  20  2f  73  62  69  6e  2f  69  6e  69  74  20  61  73
+    /// 0000040       i   n   i   t       p   r   o   c   e   s   s  \r  \n   [
+    ///
+    /// dmesg message end a line with /r/n . When redirect message to logger, we should
+    /// remove the /r/n .
+    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+        let arr: Vec<&[u8]> = buf.split(|c| *c == b'\n').collect();
+        let count = arr.len();
+
+        for (i, sub) in arr.iter().enumerate() {
+            if sub.is_empty() {
+                if !self.buf.is_empty() {
+                    slog::info!(
+                        self.logger,
+                        "{}",
+                        String::from_utf8_lossy(self.buf.as_ref()).trim_end()
+                    );
+                    self.buf.clear();
+                }
+            } else if sub.len() < buf.len() && i < count - 1 {
+                slog::info!(
+                    self.logger,
+                    "{}{}",
+                    String::from_utf8_lossy(self.buf.as_ref()).trim_end(),
+                    String::from_utf8_lossy(sub).trim_end(),
+                );
+                self.buf.clear();
+            } else {
+                self.buf.put_slice(sub);
+            }
+        }
+
+        Ok(buf.len())
+    }
+
+    fn flush(&mut self) -> io::Result<()> {
+        Ok(())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use slog::Drain;
+    use std::io::Write;
+
+    fn create_logger() -> slog::Logger {
+        let decorator = slog_term::TermDecorator::new().build();
+        let drain = slog_term::FullFormat::new(decorator).build().fuse();
+        let drain = slog_async::Async::new(drain).build().fuse();
+        slog::Logger::root(drain, slog::o!())
+    }
+
+    #[test]
+    fn test_dmesg_writer() {
+        let mut writer = DmesgWriter {
+            buf: Default::default(),
+            logger: create_logger(),
+        };
+
+        writer.flush().unwrap();
+        writer.write_all("".as_bytes()).unwrap();
+        writer.write_all("\n".as_bytes()).unwrap();
+        writer.write_all("\n\n".as_bytes()).unwrap();
+        writer.write_all("\n\n\n".as_bytes()).unwrap();
+        writer.write_all("12\n23\n34\n56".as_bytes()).unwrap();
+        writer.write_all("78".as_bytes()).unwrap();
+        writer.write_all("90\n".as_bytes()).unwrap();
+        writer.flush().unwrap();
+    }
+
+    // TODO: add unit tests for console manager
+}

src/dragonball/src/device_manager/fs_dev_mgr.rs

@@ -0,0 +1,528 @@
+// Copyright 2020-2022 Alibaba Cloud. All Rights Reserved.
+// Copyright 2019 Intel Corporation. All Rights Reserved.
+//
+// SPDX-License-Identifier: Apache-2.0
+
+use std::convert::TryInto;
+
+use dbs_utils::epoll_manager::EpollManager;
+use dbs_virtio_devices::{self as virtio, Error as VirtIoError};
+use serde_derive::{Deserialize, Serialize};
+use slog::{error, info};
+
+use crate::address_space_manager::GuestAddressSpaceImpl;
+use crate::config_manager::{
+    ConfigItem, DeviceConfigInfo, DeviceConfigInfos, RateLimiterConfigInfo,
+};
+use crate::device_manager::{
+    DbsMmioV2Device, DeviceManager, DeviceMgrError, DeviceOpContext, DeviceVirtioRegionHandler,
+};
+use crate::get_bucket_update;
+
+use super::DbsVirtioDevice;
+
+// The flag of whether to use the shared irq.
+const USE_SHARED_IRQ: bool = true;
+// The flag of whether to use the generic irq.
+const USE_GENERIC_IRQ: bool = true;
+// Default cache size is 2 Gi since this is a typical VM memory size.
+const DEFAULT_CACHE_SIZE: u64 = 2 * 1024 * 1024 * 1024;
+// We have 2 supported fs device mode, vhostuser and virtio
+const VHOSTUSER_FS_MODE: &str = "vhostuser";
+// We have 2 supported fs device mode, vhostuser and virtio
+const VIRTIO_FS_MODE: &str = "virtio";
+
+/// Errors associated with `FsDeviceConfig`.
+#[derive(Debug, thiserror::Error)]
+pub enum FsDeviceError {
+    /// Invalid fs, "virtio" or "vhostuser" is allowed.
+    #[error("the fs type is invalid, virtio or vhostuser is allowed")]
+    InvalidFs,
+
+    /// Cannot access address space.
+    #[error("Cannot access address space.")]
+    AddressSpaceNotInitialized,
+
+    /// Cannot convert RateLimterConfigInfo into RateLimiter.
+    #[error("failure while converting RateLimterConfigInfo into RateLimiter: {0}")]
+    RateLimterConfigInfoTryInto(#[source] std::io::Error),
+
+    /// The fs device tag was already used for a different fs.
+    #[error("VirtioFs device tag {0} already exists")]
+    FsDeviceTagAlreadyExists(String),
+
+    /// The fs device path was already used for a different fs.
+    #[error("VirtioFs device tag {0} already exists")]
+    FsDevicePathAlreadyExists(String),
+
+    /// The update is not allowed after booting the microvm.
+    #[error("update operation is not allowed after boot")]
+    UpdateNotAllowedPostBoot,
+
+    /// The attachbackendfs operation fails.
+    #[error("Fs device attach a backend fs failed")]
+    AttachBackendFailed(String),
+
+    /// attach backend fs must be done when vm is running.
+    #[error("vm is not running when attaching a backend fs")]
+    MicroVMNotRunning,
+
+    /// The mount tag doesn't exist.
+    #[error("fs tag'{0}' doesn't exist")]
+    TagNotExists(String),
+
+    /// Failed to send patch message to VirtioFs epoll handler.
+    #[error("could not send patch message to the VirtioFs epoll handler")]
+    VirtioFsEpollHanderSendFail,
+
+    /// Creating a shared-fs device fails (if the vhost-user socket cannot be open.)
+    #[error("cannot create shared-fs device: {0}")]
+    CreateFsDevice(#[source] VirtIoError),
+
+    /// Cannot initialize a shared-fs device or add a device to the MMIO Bus.
+    #[error("failure while registering shared-fs device: {0}")]
+    RegisterFsDevice(#[source] DeviceMgrError),
+
+    /// The device manager errors.
+    #[error("DeviceManager error: {0}")]
+    DeviceManager(#[source] DeviceMgrError),
+}
+
+/// Configuration information for a vhost-user-fs device.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct FsDeviceConfigInfo {
+    /// vhost-user socket path.
+    pub sock_path: String,
+    /// virtiofs mount tag name used inside the guest.
+    /// used as the device name during mount.
+    pub tag: String,
+    /// Number of virtqueues to use.
+    pub num_queues: usize,
+    /// Size of each virtqueue. Unit: byte.
+    pub queue_size: u16,
+    /// DAX cache window size
+    pub cache_size: u64,
+    /// Number of thread pool workers.
+    pub thread_pool_size: u16,
+    /// The caching policy the file system should use (auto, always or never).
+    /// This cache policy is set for virtio-fs, visit https://gitlab.com/virtio-fs/virtiofsd to get further information.
+    pub cache_policy: String,
+    /// Writeback cache
+    pub writeback_cache: bool,
+    /// Enable no_open or not
+    pub no_open: bool,
+    /// Enable xattr or not
+    pub xattr: bool,
+    /// Drop CAP_SYS_RESOURCE or not
+    pub drop_sys_resource: bool,
+    /// virtio fs or vhostuser fs.
+    pub mode: String,
+    /// Enable kill_priv_v2 or not
+    pub fuse_killpriv_v2: bool,
+    /// Enable no_readdir or not
+    pub no_readdir: bool,
+    /// Rate Limiter for I/O operations.
+    pub rate_limiter: Option<RateLimiterConfigInfo>,
+    /// Use shared irq
+    pub use_shared_irq: Option<bool>,
+    /// Use generic irq
+    pub use_generic_irq: Option<bool>,
+}
+
+impl std::default::Default for FsDeviceConfigInfo {
+    fn default() -> Self {
+        Self {
+            sock_path: String::default(),
+            tag: String::default(),
+            num_queues: 1,
+            queue_size: 1024,
+            cache_size: DEFAULT_CACHE_SIZE,
+            thread_pool_size: 0,
+            cache_policy: Self::default_cache_policy(),
+            writeback_cache: Self::default_writeback_cache(),
+            no_open: Self::default_no_open(),
+            fuse_killpriv_v2: Self::default_fuse_killpriv_v2(),
+            no_readdir: Self::default_no_readdir(),
+            xattr: Self::default_xattr(),
+            drop_sys_resource: Self::default_drop_sys_resource(),
+            mode: Self::default_fs_mode(),
+            rate_limiter: Some(RateLimiterConfigInfo::default()),
+            use_shared_irq: None,
+            use_generic_irq: None,
+        }
+    }
+}
+
+impl FsDeviceConfigInfo {
+    /// The default mode is set to 'virtio' for 'virtio-fs' device.
+    pub fn default_fs_mode() -> String {
+        String::from(VIRTIO_FS_MODE)
+    }
+
+    /// The default cache policy
+    pub fn default_cache_policy() -> String {
+        "always".to_string()
+    }
+
+    /// The default setting of writeback cache
+    pub fn default_writeback_cache() -> bool {
+        true
+    }
+
+    /// The default setting of no_open
+    pub fn default_no_open() -> bool {
+        true
+    }
+
+    /// The default setting of killpriv_v2
+    pub fn default_fuse_killpriv_v2() -> bool {
+        false
+    }
+
+    /// The default setting of xattr
+    pub fn default_xattr() -> bool {
+        false
+    }
+
+    /// The default setting of drop_sys_resource
+    pub fn default_drop_sys_resource() -> bool {
+        false
+    }
+
+    /// The default setting of no_readdir
+    pub fn default_no_readdir() -> bool {
+        false
+    }
+
+    /// The default setting of rate limiter
+    pub fn default_fs_rate_limiter() -> Option<RateLimiterConfigInfo> {
+        None
+    }
+}
+
+/// Configuration information for virtio-fs.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct FsDeviceConfigUpdateInfo {
+    /// virtiofs mount tag name used inside the guest.
+    /// used as the device name during mount.
+    pub tag: String,
+    /// Rate Limiter for I/O operations.
+    pub rate_limiter: Option<RateLimiterConfigInfo>,
+}
+
+impl FsDeviceConfigUpdateInfo {
+    /// Provides a `BucketUpdate` description for the bandwidth rate limiter.
+    pub fn bytes(&self) -> dbs_utils::rate_limiter::BucketUpdate {
+        get_bucket_update!(self, rate_limiter, bandwidth)
+    }
+    /// Provides a `BucketUpdate` description for the ops rate limiter.
+    pub fn ops(&self) -> dbs_utils::rate_limiter::BucketUpdate {
+        get_bucket_update!(self, rate_limiter, ops)
+    }
+}
+
+impl ConfigItem for FsDeviceConfigInfo {
+    type Err = FsDeviceError;
+
+    fn id(&self) -> &str {
+        &self.tag
+    }
+
+    fn check_conflicts(&self, other: &Self) -> Result<(), FsDeviceError> {
+        if self.tag == other.tag {
+            Err(FsDeviceError::FsDeviceTagAlreadyExists(self.tag.clone()))
+        } else if self.mode.as_str() == VHOSTUSER_FS_MODE && self.sock_path == other.sock_path {
+            Err(FsDeviceError::FsDevicePathAlreadyExists(
+                self.sock_path.clone(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+}
+
+/// Configuration information of manipulating backend fs for a virtiofs device.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct FsMountConfigInfo {
+    /// Mount operations, mount, update, umount
+    pub ops: String,
+    /// The backend fs type to mount.
+    pub fstype: Option<String>,
+    /// the source file/directory the backend fs points to
+    pub source: Option<String>,
+    /// where the backend fs gets mounted
+    pub mountpoint: String,
+    /// backend fs config content in json format
+    pub config: Option<String>,
+    /// virtiofs mount tag name used inside the guest.
+    /// used as the device name during mount.
+    pub tag: String,
+    /// Path to file that contains file lists that should be prefetched by rafs
+    pub prefetch_list_path: Option<String>,
+    /// What size file supports dax
+    pub dax_threshold_size_kb: Option<u64>,
+}
+
+pub(crate) type FsDeviceInfo = DeviceConfigInfo<FsDeviceConfigInfo>;
+
+impl ConfigItem for FsDeviceInfo {
+    type Err = FsDeviceError;
+    fn id(&self) -> &str {
+        &self.config.tag
+    }
+
+    fn check_conflicts(&self, other: &Self) -> Result<(), FsDeviceError> {
+        if self.config.tag == other.config.tag {
+            Err(FsDeviceError::FsDeviceTagAlreadyExists(
+                self.config.tag.clone(),
+            ))
+        } else if self.config.sock_path == other.config.sock_path {
+            Err(FsDeviceError::FsDevicePathAlreadyExists(
+                self.config.sock_path.clone(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+}
+
+/// Wrapper for the collection that holds all the Fs Devices Configs
+pub struct FsDeviceMgr {
+    /// A list of `FsDeviceConfig` objects.
+    pub(crate) info_list: DeviceConfigInfos<FsDeviceConfigInfo>,
+    pub(crate) use_shared_irq: bool,
+}
+
+impl FsDeviceMgr {
+    /// Inserts `fs_cfg` in the shared-fs device configuration list.
+    pub fn insert_device(
+        device_mgr: &mut DeviceManager,
+        ctx: DeviceOpContext,
+        fs_cfg: FsDeviceConfigInfo,
+    ) -> std::result::Result<(), FsDeviceError> {
+        // It's too complicated to manage life cycle of shared-fs service process for hotplug.
+        if ctx.is_hotplug {
+            error!(
+                ctx.logger(),
+                "no support of shared-fs device hotplug";
+                "subsystem" => "shared-fs",
+                "tag" => &fs_cfg.tag,
+            );
+            return Err(FsDeviceError::UpdateNotAllowedPostBoot);
+        }
+
+        info!(
+            ctx.logger(),
+            "add shared-fs device configuration";
+            "subsystem" => "shared-fs",
+            "tag" => &fs_cfg.tag,
+        );
+        device_mgr
+            .fs_manager
+            .lock()
+            .unwrap()
+            .info_list
+            .insert_or_update(&fs_cfg)?;
+
+        Ok(())
+    }
+
+    /// Attaches all vhost-user-fs devices from the FsDevicesConfig.
+    pub fn attach_devices(
+        &mut self,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<(), FsDeviceError> {
+        let epoll_mgr = ctx
+            .epoll_mgr
+            .clone()
+            .ok_or(FsDeviceError::CreateFsDevice(virtio::Error::InvalidInput))?;
+
+        for info in self.info_list.iter_mut() {
+            let device = Self::create_fs_device(&info.config, ctx, epoll_mgr.clone())?;
+            let mmio_device = DeviceManager::create_mmio_virtio_device(
+                device,
+                ctx,
+                info.config.use_shared_irq.unwrap_or(self.use_shared_irq),
+                info.config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+            )
+            .map_err(FsDeviceError::RegisterFsDevice)?;
+
+            info.set_device(mmio_device);
+        }
+
+        Ok(())
+    }
+
+    fn create_fs_device(
+        config: &FsDeviceConfigInfo,
+        ctx: &mut DeviceOpContext,
+        epoll_mgr: EpollManager,
+    ) -> std::result::Result<DbsVirtioDevice, FsDeviceError> {
+        match &config.mode as &str {
+            VIRTIO_FS_MODE => Self::attach_virtio_fs_devices(config, ctx, epoll_mgr),
+            _ => Err(FsDeviceError::CreateFsDevice(virtio::Error::InvalidInput)),
+        }
+    }
+
+    fn attach_virtio_fs_devices(
+        config: &FsDeviceConfigInfo,
+        ctx: &mut DeviceOpContext,
+        epoll_mgr: EpollManager,
+    ) -> std::result::Result<DbsVirtioDevice, FsDeviceError> {
+        info!(
+            ctx.logger(),
+            "add virtio-fs device configuration";
+            "subsystem" => "virito-fs",
+            "tag" => &config.tag,
+            "dax_window_size" => &config.cache_size,
+        );
+
+        let limiter = if let Some(rlc) = config.rate_limiter.clone() {
+            Some(
+                rlc.try_into()
+                    .map_err(FsDeviceError::RateLimterConfigInfoTryInto)?,
+            )
+        } else {
+            None
+        };
+
+        let vm_as = ctx.get_vm_as().map_err(|e| {
+            error!(ctx.logger(), "virtio-fs get vm_as error: {:?}", e; 
+                "subsystem" => "virito-fs");
+            FsDeviceError::DeviceManager(e)
+        })?;
+        let address_space = match ctx.address_space.as_ref() {
+            Some(address_space) => address_space.clone(),
+            None => {
+                error!(ctx.logger(), "virtio-fs get address_space error"; "subsystem" => "virito-fs");
+                return Err(FsDeviceError::AddressSpaceNotInitialized);
+            }
+        };
+        let handler = DeviceVirtioRegionHandler {
+            vm_as,
+            address_space,
+        };
+
+        let device = Box::new(
+            virtio::fs::VirtioFs::new(
+                &config.tag,
+                config.num_queues,
+                config.queue_size,
+                config.cache_size,
+                &config.cache_policy,
+                config.thread_pool_size,
+                config.writeback_cache,
+                config.no_open,
+                config.fuse_killpriv_v2,
+                config.xattr,
+                config.drop_sys_resource,
+                config.no_readdir,
+                Box::new(handler),
+                epoll_mgr,
+                limiter,
+            )
+            .map_err(FsDeviceError::CreateFsDevice)?,
+        );
+
+        Ok(device)
+    }
+
+    /// Attach a backend fs to a VirtioFs device or detach a backend
+    /// fs from a Virtiofs device
+    pub fn manipulate_backend_fs(
+        device_mgr: &mut DeviceManager,
+        config: FsMountConfigInfo,
+    ) -> std::result::Result<(), FsDeviceError> {
+        let mut found = false;
+
+        let mgr = &mut device_mgr.fs_manager.lock().unwrap();
+        for info in mgr
+            .info_list
+            .iter()
+            .filter(|info| info.config.tag.as_str() == config.tag.as_str())
+        {
+            found = true;
+            if let Some(device) = info.device.as_ref() {
+                if let Some(mmio_dev) = device.as_any().downcast_ref::<DbsMmioV2Device>() {
+                    let mut guard = mmio_dev.state();
+                    let inner_dev = guard.get_inner_device_mut();
+                    if let Some(virtio_fs_dev) = inner_dev
+                        .as_any_mut()
+                        .downcast_mut::<virtio::fs::VirtioFs<GuestAddressSpaceImpl>>()
+                    {
+                        return virtio_fs_dev
+                            .manipulate_backend_fs(
+                                config.source,
+                                config.fstype,
+                                &config.mountpoint,
+                                config.config,
+                                &config.ops,
+                                config.prefetch_list_path,
+                                config.dax_threshold_size_kb,
+                            )
+                            .map(|_p| ())
+                            .map_err(|e| FsDeviceError::AttachBackendFailed(e.to_string()));
+                    }
+                }
+            }
+        }
+        if !found {
+            Err(FsDeviceError::AttachBackendFailed(
+                "fs tag not found".to_string(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+
+    /// Gets the index of the device with the specified `tag` if it exists in the list.
+    pub fn get_index_of_tag(&self, tag: &str) -> Option<usize> {
+        self.info_list
+            .iter()
+            .position(|info| info.config.id().eq(tag))
+    }
+
+    /// Update the ratelimiter settings of a virtio fs device.
+    pub fn update_device_ratelimiters(
+        device_mgr: &mut DeviceManager,
+        new_cfg: FsDeviceConfigUpdateInfo,
+    ) -> std::result::Result<(), FsDeviceError> {
+        let mgr = &mut device_mgr.fs_manager.lock().unwrap();
+        match mgr.get_index_of_tag(&new_cfg.tag) {
+            Some(index) => {
+                let config = &mut mgr.info_list[index].config;
+                config.rate_limiter = new_cfg.rate_limiter.clone();
+                let device = mgr.info_list[index]
+                    .device
+                    .as_mut()
+                    .ok_or_else(|| FsDeviceError::TagNotExists("".to_owned()))?;
+
+                if let Some(mmio_dev) = device.as_any().downcast_ref::<DbsMmioV2Device>() {
+                    let guard = mmio_dev.state();
+                    let inner_dev = guard.get_inner_device();
+                    if let Some(fs_dev) = inner_dev
+                        .as_any()
+                        .downcast_ref::<virtio::fs::VirtioFs<GuestAddressSpaceImpl>>()
+                    {
+                        return fs_dev
+                            .set_patch_rate_limiters(new_cfg.bytes(), new_cfg.ops())
+                            .map(|_p| ())
+                            .map_err(|_e| FsDeviceError::VirtioFsEpollHanderSendFail);
+                    }
+                }
+                Ok(())
+            }
+            None => Err(FsDeviceError::TagNotExists(new_cfg.tag)),
+        }
+    }
+}
+
+impl Default for FsDeviceMgr {
+    /// Create a new `FsDeviceMgr` object..
+    fn default() -> Self {
+        FsDeviceMgr {
+            info_list: DeviceConfigInfos::new(),
+            use_shared_irq: USE_SHARED_IRQ,
+        }
+    }
+}

src/dragonball/src/device_manager/legacy.rs

@@ -0,0 +1,246 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+//! Device Manager for Legacy Devices.
+
+use std::io;
+use std::sync::{Arc, Mutex};
+
+use dbs_device::device_manager::Error as IoManagerError;
+#[cfg(target_arch = "aarch64")]
+use dbs_legacy_devices::RTCDevice;
+use dbs_legacy_devices::SerialDevice;
+use vmm_sys_util::eventfd::EventFd;
+
+// The I8042 Data Port (IO Port 0x60) is used for reading data that was received from a I8042 device or from the I8042 controller itself and writing data to a I8042 device or to the I8042 controller itself.
+const I8042_DATA_PORT: u16 = 0x60;
+
+/// Errors generated by legacy device manager.
+#[derive(Debug, thiserror::Error)]
+pub enum Error {
+    /// Cannot add legacy device to Bus.
+    #[error("bus failure while managing legacy device")]
+    BusError(#[source] IoManagerError),
+
+    /// Cannot create EventFd.
+    #[error("failure while reading EventFd file descriptor")]
+    EventFd(#[source] io::Error),
+
+    /// Failed to register/deregister interrupt.
+    #[error("failure while managing interrupt for legacy device")]
+    IrqManager(#[source] vmm_sys_util::errno::Error),
+}
+
+/// The `LegacyDeviceManager` is a wrapper that is used for registering legacy devices
+/// on an I/O Bus.
+///
+/// It currently manages the uart and i8042 devices. The `LegacyDeviceManger` should be initialized
+/// only by using the constructor.
+pub struct LegacyDeviceManager {
+    #[cfg(target_arch = "x86_64")]
+    i8042_reset_eventfd: EventFd,
+    #[cfg(target_arch = "aarch64")]
+    pub(crate) _rtc_device: Arc<Mutex<RTCDevice>>,
+    #[cfg(target_arch = "aarch64")]
+    _rtc_eventfd: EventFd,
+    pub(crate) com1_device: Arc<Mutex<SerialDevice>>,
+    _com1_eventfd: EventFd,
+    pub(crate) com2_device: Arc<Mutex<SerialDevice>>,
+    _com2_eventfd: EventFd,
+}
+
+impl LegacyDeviceManager {
+    /// Get the serial device for com1.
+    pub fn get_com1_serial(&self) -> Arc<Mutex<SerialDevice>> {
+        self.com1_device.clone()
+    }
+
+    /// Get the serial device for com2
+    pub fn get_com2_serial(&self) -> Arc<Mutex<SerialDevice>> {
+        self.com2_device.clone()
+    }
+}
+
+#[cfg(target_arch = "x86_64")]
+pub(crate) mod x86_64 {
+    use super::*;
+    use dbs_device::device_manager::IoManager;
+    use dbs_device::resources::Resource;
+    use dbs_legacy_devices::{EventFdTrigger, I8042Device, I8042DeviceMetrics};
+    use kvm_ioctls::VmFd;
+
+    pub(crate) const COM1_IRQ: u32 = 4;
+    pub(crate) const COM1_PORT1: u16 = 0x3f8;
+    pub(crate) const COM2_IRQ: u32 = 3;
+    pub(crate) const COM2_PORT1: u16 = 0x2f8;
+
+    type Result<T> = ::std::result::Result<T, Error>;
+
+    impl LegacyDeviceManager {
+        /// Create a LegacyDeviceManager instance handling legacy devices (uart, i8042).
+        pub fn create_manager(bus: &mut IoManager, vm_fd: Option<Arc<VmFd>>) -> Result<Self> {
+            let (com1_device, com1_eventfd) =
+                Self::create_com_device(bus, vm_fd.as_ref(), COM1_IRQ, COM1_PORT1)?;
+            let (com2_device, com2_eventfd) =
+                Self::create_com_device(bus, vm_fd.as_ref(), COM2_IRQ, COM2_PORT1)?;
+
+            let exit_evt = EventFd::new(libc::EFD_NONBLOCK).map_err(Error::EventFd)?;
+            let i8042_device = Arc::new(Mutex::new(I8042Device::new(
+                EventFdTrigger::new(exit_evt.try_clone().map_err(Error::EventFd)?),
+                Arc::new(I8042DeviceMetrics::default()),
+            )));
+            let resources = [Resource::PioAddressRange {
+                // 0x60 and 0x64 are the io ports that i8042 devices used.
+                // We register pio address range from 0x60 - 0x64 with base I8042_DATA_PORT for i8042 to use.
+                base: I8042_DATA_PORT,
+                size: 0x5,
+            }];
+            bus.register_device_io(i8042_device, &resources)
+                .map_err(Error::BusError)?;
+
+            Ok(LegacyDeviceManager {
+                i8042_reset_eventfd: exit_evt,
+                com1_device,
+                _com1_eventfd: com1_eventfd,
+                com2_device,
+                _com2_eventfd: com2_eventfd,
+            })
+        }
+
+        /// Get the eventfd for exit notification.
+        pub fn get_reset_eventfd(&self) -> Result<EventFd> {
+            self.i8042_reset_eventfd.try_clone().map_err(Error::EventFd)
+        }
+
+        fn create_com_device(
+            bus: &mut IoManager,
+            vm_fd: Option<&Arc<VmFd>>,
+            irq: u32,
+            port_base: u16,
+        ) -> Result<(Arc<Mutex<SerialDevice>>, EventFd)> {
+            let eventfd = EventFd::new(libc::EFD_NONBLOCK).map_err(Error::EventFd)?;
+            let device = Arc::new(Mutex::new(SerialDevice::new(
+                eventfd.try_clone().map_err(Error::EventFd)?,
+            )));
+            // port_base defines the base port address for the COM devices.
+            // Since every COM device has 8 data registers so we register the pio address range as size 0x8.
+            let resources = [Resource::PioAddressRange {
+                base: port_base,
+                size: 0x8,
+            }];
+            bus.register_device_io(device.clone(), &resources)
+                .map_err(Error::BusError)?;
+
+            if let Some(fd) = vm_fd {
+                fd.register_irqfd(&eventfd, irq)
+                    .map_err(Error::IrqManager)?;
+            }
+
+            Ok((device, eventfd))
+        }
+    }
+}
+
+#[cfg(target_arch = "aarch64")]
+pub(crate) mod aarch64 {
+    use super::*;
+    use dbs_device::device_manager::IoManager;
+    use dbs_device::resources::DeviceResources;
+    use kvm_ioctls::VmFd;
+    use std::collections::HashMap;
+
+    type Result<T> = ::std::result::Result<T, Error>;
+
+    /// LegacyDeviceType: com1
+    pub const COM1: &str = "com1";
+    /// LegacyDeviceType: com2
+    pub const COM2: &str = "com2";
+    /// LegacyDeviceType: rtc
+    pub const RTC: &str = "rtc";
+
+    impl LegacyDeviceManager {
+        /// Create a LegacyDeviceManager instance handling legacy devices.
+        pub fn create_manager(
+            bus: &mut IoManager,
+            vm_fd: Option<Arc<VmFd>>,
+            resources: &HashMap<String, DeviceResources>,
+        ) -> Result<Self> {
+            let (com1_device, com1_eventfd) =
+                Self::create_com_device(bus, vm_fd.as_ref(), resources.get(COM1).unwrap())?;
+            let (com2_device, com2_eventfd) =
+                Self::create_com_device(bus, vm_fd.as_ref(), resources.get(COM2).unwrap())?;
+            let (rtc_device, rtc_eventfd) =
+                Self::create_rtc_device(bus, vm_fd.as_ref(), resources.get(RTC).unwrap())?;
+
+            Ok(LegacyDeviceManager {
+                _rtc_device: rtc_device,
+                _rtc_eventfd: rtc_eventfd,
+                com1_device,
+                _com1_eventfd: com1_eventfd,
+                com2_device,
+                _com2_eventfd: com2_eventfd,
+            })
+        }
+
+        fn create_com_device(
+            bus: &mut IoManager,
+            vm_fd: Option<&Arc<VmFd>>,
+            resources: &DeviceResources,
+        ) -> Result<(Arc<Mutex<SerialDevice>>, EventFd)> {
+            let eventfd = EventFd::new(libc::EFD_NONBLOCK).map_err(Error::EventFd)?;
+            let device = Arc::new(Mutex::new(SerialDevice::new(
+                eventfd.try_clone().map_err(Error::EventFd)?,
+            )));
+
+            bus.register_device_io(device.clone(), resources.get_all_resources())
+                .map_err(Error::BusError)?;
+
+            if let Some(fd) = vm_fd {
+                let irq = resources.get_legacy_irq().unwrap();
+                fd.register_irqfd(&eventfd, irq)
+                    .map_err(Error::IrqManager)?;
+            }
+
+            Ok((device, eventfd))
+        }
+
+        fn create_rtc_device(
+            bus: &mut IoManager,
+            vm_fd: Option<&Arc<VmFd>>,
+            resources: &DeviceResources,
+        ) -> Result<(Arc<Mutex<RTCDevice>>, EventFd)> {
+            let eventfd = EventFd::new(libc::EFD_NONBLOCK).map_err(Error::EventFd)?;
+            let device = Arc::new(Mutex::new(RTCDevice::new()));
+
+            bus.register_device_io(device.clone(), resources.get_all_resources())
+                .map_err(Error::BusError)?;
+
+            if let Some(fd) = vm_fd {
+                let irq = resources.get_legacy_irq().unwrap();
+                fd.register_irqfd(&eventfd, irq)
+                    .map_err(Error::IrqManager)?;
+            }
+
+            Ok((device, eventfd))
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    #[cfg(target_arch = "x86_64")]
+    use super::*;
+
+    #[test]
+    #[cfg(target_arch = "x86_64")]
+    fn test_create_legacy_device_manager() {
+        let mut bus = dbs_device::device_manager::IoManager::new();
+        let mgr = LegacyDeviceManager::create_manager(&mut bus, None).unwrap();
+        let _exit_fd = mgr.get_reset_eventfd().unwrap();
+    }
+}

src/dragonball/src/device_manager/memory_region_handler.rs

@@ -0,0 +1,110 @@
+// Copyright 2022 Alibaba, Inc. or its affiliates. All Rights Reserved.
+//
+// SPDX-License-Identifier: Apache-2.0
+
+use std::io;
+use std::sync::Arc;
+
+use dbs_address_space::{AddressSpace, AddressSpaceRegion, AddressSpaceRegionType};
+use dbs_virtio_devices::{Error as VirtIoError, VirtioRegionHandler};
+use log::{debug, error};
+use vm_memory::{FileOffset, GuestAddressSpace, GuestMemoryRegion, GuestRegionMmap};
+
+use crate::address_space_manager::GuestAddressSpaceImpl;
+
+/// This struct implements the VirtioRegionHandler trait, which inserts the memory
+/// region of the virtio device into vm_as and address_space.
+///
+/// * After region is inserted into the vm_as, the virtio device can read guest memory
+///   data using vm_as.get_slice with GuestAddress.
+///
+/// * Insert virtio memory into address_space so that the correct guest last address can
+///   be found when initializing the e820 table. The e820 table is a table that describes
+///   guest memory prepared before the guest startup. we need to config the correct guest
+///   memory address and length in the table. The virtio device memory belongs to the MMIO
+///   space and does not belong to the Guest Memory space. Therefore, it cannot be configured
+///   into the e820 table. When creating AddressSpaceRegion we use
+///   AddressSpaceRegionType::ReservedMemory type, in this way, address_space will know that
+///   this region a special memory, it will don't put the this memory in e820 table.
+///
+/// This function relies on the atomic-guest-memory feature. Without this feature enabled, memory
+/// regions cannot be inserted into vm_as. Because the insert_region interface of vm_as does
+/// not insert regions in place, but returns an array of inserted regions. We need to manually
+/// replace this array of regions with vm_as, and that's what atomic-guest-memory feature does.
+/// So we rely on the atomic-guest-memory feature here
+pub struct DeviceVirtioRegionHandler {
+    pub(crate) vm_as: GuestAddressSpaceImpl,
+    pub(crate) address_space: AddressSpace,
+}
+
+impl DeviceVirtioRegionHandler {
+    fn insert_address_space(
+        &mut self,
+        region: Arc<GuestRegionMmap>,
+    ) -> std::result::Result<(), VirtIoError> {
+        let file_offset = match region.file_offset() {
+            // TODO: use from_arc
+            Some(f) => Some(FileOffset::new(f.file().try_clone()?, 0)),
+            None => None,
+        };
+
+        let as_region = Arc::new(AddressSpaceRegion::build(
+            AddressSpaceRegionType::DAXMemory,
+            region.start_addr(),
+            region.size() as u64,
+            None,
+            file_offset,
+            region.flags(),
+            false,
+        ));
+
+        self.address_space.insert_region(as_region).map_err(|e| {
+            error!("inserting address apace error: {}", e);
+            // dbs-virtio-devices should not depend on dbs-address-space.
+            // So here io::Error is used instead of AddressSpaceError directly.
+            VirtIoError::IOError(io::Error::new(
+                io::ErrorKind::Other,
+                format!(
+                    "invalid address space region ({0:#x}, {1:#x})",
+                    region.start_addr().0,
+                    region.len()
+                ),
+            ))
+        })?;
+        Ok(())
+    }
+
+    fn insert_vm_as(
+        &mut self,
+        region: Arc<GuestRegionMmap>,
+    ) -> std::result::Result<(), VirtIoError> {
+        let vm_as_new = self.vm_as.memory().insert_region(region).map_err(|e| {
+            error!(
+                "DeviceVirtioRegionHandler failed to insert guest memory region: {:?}.",
+                e
+            );
+            VirtIoError::InsertMmap(e)
+        })?;
+        // Do not expect poisoned lock here, so safe to unwrap().
+        self.vm_as.lock().unwrap().replace(vm_as_new);
+
+        Ok(())
+    }
+}
+
+impl VirtioRegionHandler for DeviceVirtioRegionHandler {
+    fn insert_region(
+        &mut self,
+        region: Arc<GuestRegionMmap>,
+    ) -> std::result::Result<(), VirtIoError> {
+        debug!(
+            "add geust memory region to address_space/vm_as, new region: {:?}",
+            region
+        );
+
+        self.insert_address_space(region.clone())?;
+        self.insert_vm_as(region)?;
+
+        Ok(())
+    }
+}

src/dragonball/src/device_manager/virtio_net_dev_mgr.rs

@@ -0,0 +1,387 @@
+// Copyright 2020-2022 Alibaba, Inc. or its affiliates. All Rights Reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+use std::convert::TryInto;
+use std::sync::Arc;
+
+use dbs_utils::net::{MacAddr, Tap, TapError};
+use dbs_utils::rate_limiter::BucketUpdate;
+use dbs_virtio_devices as virtio;
+use dbs_virtio_devices::net::Net;
+use dbs_virtio_devices::Error as VirtioError;
+use serde_derive::{Deserialize, Serialize};
+
+use crate::address_space_manager::GuestAddressSpaceImpl;
+use crate::config_manager::{
+    ConfigItem, DeviceConfigInfo, DeviceConfigInfos, RateLimiterConfigInfo,
+};
+use crate::device_manager::{DeviceManager, DeviceMgrError, DeviceOpContext};
+use crate::get_bucket_update;
+
+use super::DbsMmioV2Device;
+
+/// Default number of virtio queues, one rx/tx pair.
+pub const NUM_QUEUES: usize = 2;
+/// Default size of virtio queues.
+pub const QUEUE_SIZE: u16 = 256;
+// The flag of whether to use the shared irq.
+const USE_SHARED_IRQ: bool = true;
+// The flag of whether to use the generic irq.
+const USE_GENERIC_IRQ: bool = true;
+
+/// Errors associated with virtio net device operations.
+#[derive(Debug, thiserror::Error)]
+pub enum VirtioNetDeviceError {
+    /// The virtual machine instance ID is invalid.
+    #[error("the virtual machine instance ID is invalid")]
+    InvalidVMID,
+
+    /// The iface ID is invalid.
+    #[error("invalid virtio-net iface id '{0}'")]
+    InvalidIfaceId(String),
+
+    /// Invalid queue number configuration for virtio_net device.
+    #[error("invalid queue number {0} for virtio-net device")]
+    InvalidQueueNum(usize),
+
+    /// Failure from device manager,
+    #[error("failure in device manager operations, {0}")]
+    DeviceManager(#[source] DeviceMgrError),
+
+    /// The Context Identifier is already in use.
+    #[error("the device ID {0} already exists")]
+    DeviceIDAlreadyExist(String),
+
+    /// The MAC address is already in use.
+    #[error("the guest MAC address {0} is already in use")]
+    GuestMacAddressInUse(String),
+
+    /// The host device name is already in use.
+    #[error("the host device name {0} is already in use")]
+    HostDeviceNameInUse(String),
+
+    /// Cannot open/create tap device.
+    #[error("cannot open TAP device")]
+    OpenTap(#[source] TapError),
+
+    /// Failure from virtio subsystem.
+    #[error(transparent)]
+    Virtio(VirtioError),
+
+    /// Failed to send patch message to net epoll handler.
+    #[error("could not send patch message to the net epoll handler")]
+    NetEpollHanderSendFail,
+
+    /// The update is not allowed after booting the microvm.
+    #[error("update operation is not allowed after boot")]
+    UpdateNotAllowedPostBoot,
+
+    /// Split this at some point.
+    /// Internal errors are due to resource exhaustion.
+    /// Users errors are due to invalid permissions.
+    #[error("cannot create network device: {0}")]
+    CreateNetDevice(#[source] VirtioError),
+
+    /// Cannot initialize a MMIO Network Device or add a device to the MMIO Bus.
+    #[error("failure while registering network device: {0}")]
+    RegisterNetDevice(#[source] DeviceMgrError),
+}
+
+/// Configuration information for virtio net devices.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct VirtioNetDeviceConfigUpdateInfo {
+    /// ID of the guest network interface.
+    pub iface_id: String,
+    /// Rate Limiter for received packages.
+    pub rx_rate_limiter: Option<RateLimiterConfigInfo>,
+    /// Rate Limiter for transmitted packages.
+    pub tx_rate_limiter: Option<RateLimiterConfigInfo>,
+}
+
+impl VirtioNetDeviceConfigUpdateInfo {
+    /// Provides a `BucketUpdate` description for the RX bandwidth rate limiter.
+    pub fn rx_bytes(&self) -> BucketUpdate {
+        get_bucket_update!(self, rx_rate_limiter, bandwidth)
+    }
+    /// Provides a `BucketUpdate` description for the RX ops rate limiter.
+    pub fn rx_ops(&self) -> BucketUpdate {
+        get_bucket_update!(self, rx_rate_limiter, ops)
+    }
+    /// Provides a `BucketUpdate` description for the TX bandwidth rate limiter.
+    pub fn tx_bytes(&self) -> BucketUpdate {
+        get_bucket_update!(self, tx_rate_limiter, bandwidth)
+    }
+    /// Provides a `BucketUpdate` description for the TX ops rate limiter.
+    pub fn tx_ops(&self) -> BucketUpdate {
+        get_bucket_update!(self, tx_rate_limiter, ops)
+    }
+}
+
+/// Configuration information for virtio net devices.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize, Default)]
+pub struct VirtioNetDeviceConfigInfo {
+    /// ID of the guest network interface.
+    pub iface_id: String,
+    /// Host level path for the guest network interface.
+    pub host_dev_name: String,
+    /// Number of virtqueues to use.
+    pub num_queues: usize,
+    /// Size of each virtqueue. Unit: byte.
+    pub queue_size: u16,
+    /// Guest MAC address.
+    pub guest_mac: Option<MacAddr>,
+    /// Rate Limiter for received packages.
+    pub rx_rate_limiter: Option<RateLimiterConfigInfo>,
+    /// Rate Limiter for transmitted packages.
+    pub tx_rate_limiter: Option<RateLimiterConfigInfo>,
+    /// allow duplicate mac
+    pub allow_duplicate_mac: bool,
+    /// Use shared irq
+    pub use_shared_irq: Option<bool>,
+    /// Use generic irq
+    pub use_generic_irq: Option<bool>,
+}
+
+impl VirtioNetDeviceConfigInfo {
+    /// Returns the tap device that `host_dev_name` refers to.
+    pub fn open_tap(&self) -> std::result::Result<Tap, VirtioNetDeviceError> {
+        Tap::open_named(self.host_dev_name.as_str(), false).map_err(VirtioNetDeviceError::OpenTap)
+    }
+
+    /// Returns a reference to the mac address. It the mac address is not configured, it
+    /// return None.
+    pub fn guest_mac(&self) -> Option<&MacAddr> {
+        self.guest_mac.as_ref()
+    }
+
+    ///Rx and Tx queue and max queue sizes
+    pub fn queue_sizes(&self) -> Vec<u16> {
+        let mut queue_size = self.queue_size;
+        if queue_size == 0 {
+            queue_size = QUEUE_SIZE;
+        }
+        let num_queues = if self.num_queues > 0 {
+            self.num_queues
+        } else {
+            NUM_QUEUES
+        };
+
+        (0..num_queues).map(|_| queue_size).collect::<Vec<u16>>()
+    }
+}
+
+impl ConfigItem for VirtioNetDeviceConfigInfo {
+    type Err = VirtioNetDeviceError;
+
+    fn id(&self) -> &str {
+        &self.iface_id
+    }
+
+    fn check_conflicts(&self, other: &Self) -> Result<(), VirtioNetDeviceError> {
+        if self.iface_id == other.iface_id {
+            Err(VirtioNetDeviceError::DeviceIDAlreadyExist(
+                self.iface_id.clone(),
+            ))
+        } else if !other.allow_duplicate_mac
+            && self.guest_mac.is_some()
+            && self.guest_mac == other.guest_mac
+        {
+            Err(VirtioNetDeviceError::GuestMacAddressInUse(
+                self.guest_mac.as_ref().unwrap().to_string(),
+            ))
+        } else if self.host_dev_name == other.host_dev_name {
+            Err(VirtioNetDeviceError::HostDeviceNameInUse(
+                self.host_dev_name.clone(),
+            ))
+        } else {
+            Ok(())
+        }
+    }
+}
+
+/// Virtio Net Device Info
+pub type VirtioNetDeviceInfo = DeviceConfigInfo<VirtioNetDeviceConfigInfo>;
+
+/// Device manager to manage all virtio net devices.
+pub struct VirtioNetDeviceMgr {
+    pub(crate) info_list: DeviceConfigInfos<VirtioNetDeviceConfigInfo>,
+    pub(crate) use_shared_irq: bool,
+}
+
+impl VirtioNetDeviceMgr {
+    /// Gets the index of the device with the specified `drive_id` if it exists in the list.
+    pub fn get_index_of_iface_id(&self, if_id: &str) -> Option<usize> {
+        self.info_list
+            .iter()
+            .position(|info| info.config.iface_id.eq(if_id))
+    }
+
+    /// Insert or update a virtio net device into the manager.
+    pub fn insert_device(
+        device_mgr: &mut DeviceManager,
+        mut ctx: DeviceOpContext,
+        config: VirtioNetDeviceConfigInfo,
+    ) -> std::result::Result<(), VirtioNetDeviceError> {
+        if config.num_queues % 2 != 0 {
+            return Err(VirtioNetDeviceError::InvalidQueueNum(config.num_queues));
+        }
+        if !cfg!(feature = "hotplug") && ctx.is_hotplug {
+            return Err(VirtioNetDeviceError::UpdateNotAllowedPostBoot);
+        }
+
+        let mgr = &mut device_mgr.virtio_net_manager;
+
+        slog::info!(
+            ctx.logger(),
+            "add virtio-net device configuration";
+            "subsystem" => "net_dev_mgr",
+            "id" => &config.iface_id,
+            "host_dev_name" => &config.host_dev_name,
+        );
+
+        let device_index = mgr.info_list.insert_or_update(&config)?;
+
+        if ctx.is_hotplug {
+            slog::info!(
+                ctx.logger(),
+                "attach virtio-net device";
+                "subsystem" => "net_dev_mgr",
+                "id" => &config.iface_id,
+                "host_dev_name" => &config.host_dev_name,
+            );
+
+            match Self::create_device(&config, &mut ctx) {
+                Ok(device) => {
+                    let dev = DeviceManager::create_mmio_virtio_device(
+                        device,
+                        &mut ctx,
+                        config.use_shared_irq.unwrap_or(mgr.use_shared_irq),
+                        config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+                    )
+                    .map_err(VirtioNetDeviceError::DeviceManager)?;
+                    ctx.insert_hotplug_mmio_device(&dev.clone(), None)
+                        .map_err(VirtioNetDeviceError::DeviceManager)?;
+                    // live-upgrade need save/restore device from info.device.
+                    mgr.info_list[device_index].set_device(dev);
+                }
+                Err(e) => {
+                    mgr.info_list.remove(device_index);
+                    return Err(VirtioNetDeviceError::Virtio(e));
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Update the ratelimiter settings of a virtio net device.
+    pub fn update_device_ratelimiters(
+        device_mgr: &mut DeviceManager,
+        new_cfg: VirtioNetDeviceConfigUpdateInfo,
+    ) -> std::result::Result<(), VirtioNetDeviceError> {
+        let mgr = &mut device_mgr.virtio_net_manager;
+        match mgr.get_index_of_iface_id(&new_cfg.iface_id) {
+            Some(index) => {
+                let config = &mut mgr.info_list[index].config;
+                config.rx_rate_limiter = new_cfg.rx_rate_limiter.clone();
+                config.tx_rate_limiter = new_cfg.tx_rate_limiter.clone();
+                let device = mgr.info_list[index].device.as_mut().ok_or_else(|| {
+                    VirtioNetDeviceError::InvalidIfaceId(new_cfg.iface_id.clone())
+                })?;
+
+                if let Some(mmio_dev) = device.as_any().downcast_ref::<DbsMmioV2Device>() {
+                    let guard = mmio_dev.state();
+                    let inner_dev = guard.get_inner_device();
+                    if let Some(net_dev) = inner_dev
+                        .as_any()
+                        .downcast_ref::<virtio::net::Net<GuestAddressSpaceImpl>>()
+                    {
+                        return net_dev
+                            .set_patch_rate_limiters(
+                                new_cfg.rx_bytes(),
+                                new_cfg.rx_ops(),
+                                new_cfg.tx_bytes(),
+                                new_cfg.tx_ops(),
+                            )
+                            .map(|_p| ())
+                            .map_err(|_e| VirtioNetDeviceError::NetEpollHanderSendFail);
+                    }
+                }
+                Ok(())
+            }
+            None => Err(VirtioNetDeviceError::InvalidIfaceId(
+                new_cfg.iface_id.clone(),
+            )),
+        }
+    }
+
+    /// Attach all configured vsock device to the virtual machine instance.
+    pub fn attach_devices(
+        &mut self,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<(), VirtioNetDeviceError> {
+        for info in self.info_list.iter_mut() {
+            slog::info!(
+                ctx.logger(),
+                "attach virtio-net device";
+                "subsystem" => "net_dev_mgr",
+                "id" => &info.config.iface_id,
+                "host_dev_name" => &info.config.host_dev_name,
+            );
+
+            let device = Self::create_device(&info.config, ctx)
+                .map_err(VirtioNetDeviceError::CreateNetDevice)?;
+            let device = DeviceManager::create_mmio_virtio_device(
+                device,
+                ctx,
+                info.config.use_shared_irq.unwrap_or(self.use_shared_irq),
+                info.config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+            )
+            .map_err(VirtioNetDeviceError::RegisterNetDevice)?;
+            info.set_device(device);
+        }
+
+        Ok(())
+    }
+
+    fn create_device(
+        cfg: &VirtioNetDeviceConfigInfo,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<Box<Net<GuestAddressSpaceImpl>>, virtio::Error> {
+        let epoll_mgr = ctx.epoll_mgr.clone().ok_or(virtio::Error::InvalidInput)?;
+        let rx_rate_limiter = match cfg.rx_rate_limiter.as_ref() {
+            Some(rl) => Some(rl.try_into().map_err(virtio::Error::IOError)?),
+            None => None,
+        };
+        let tx_rate_limiter = match cfg.tx_rate_limiter.as_ref() {
+            Some(rl) => Some(rl.try_into().map_err(virtio::Error::IOError)?),
+            None => None,
+        };
+
+        let net_device = Net::new(
+            cfg.host_dev_name.clone(),
+            cfg.guest_mac(),
+            Arc::new(cfg.queue_sizes()),
+            epoll_mgr,
+            rx_rate_limiter,
+            tx_rate_limiter,
+        )?;
+
+        Ok(Box::new(net_device))
+    }
+}
+
+impl Default for VirtioNetDeviceMgr {
+    /// Create a new virtio net device manager.
+    fn default() -> Self {
+        VirtioNetDeviceMgr {
+            info_list: DeviceConfigInfos::new(),
+            use_shared_irq: USE_SHARED_IRQ,
+        }
+    }
+}

src/dragonball/src/device_manager/vsock_dev_mgr.rs

@@ -0,0 +1,299 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+use std::sync::Arc;
+
+use dbs_virtio_devices as virtio;
+use dbs_virtio_devices::mmio::DRAGONBALL_FEATURE_INTR_USED;
+use dbs_virtio_devices::vsock::backend::{
+    VsockInnerBackend, VsockInnerConnector, VsockTcpBackend, VsockUnixStreamBackend,
+};
+use dbs_virtio_devices::vsock::Vsock;
+use dbs_virtio_devices::Error as VirtioError;
+use serde_derive::{Deserialize, Serialize};
+
+use super::StartMicroVmError;
+use crate::config_manager::{ConfigItem, DeviceConfigInfo, DeviceConfigInfos};
+use crate::device_manager::{DeviceManager, DeviceOpContext};
+
+pub use dbs_virtio_devices::vsock::QUEUE_SIZES;
+
+const SUBSYSTEM: &str = "vsock_dev_mgr";
+// The flag of whether to use the shared irq.
+const USE_SHARED_IRQ: bool = true;
+// The flag of whether to use the generic irq.
+const USE_GENERIC_IRQ: bool = true;
+
+/// Errors associated with `VsockDeviceConfigInfo`.
+#[derive(Debug, thiserror::Error)]
+pub enum VsockDeviceError {
+    /// The virtual machine instance ID is invalid.
+    #[error("the virtual machine instance ID is invalid")]
+    InvalidVMID,
+
+    /// The Context Identifier is already in use.
+    #[error("the device ID {0} already exists")]
+    DeviceIDAlreadyExist(String),
+
+    /// The Context Identifier is invalid.
+    #[error("the guest CID {0} is invalid")]
+    GuestCIDInvalid(u32),
+
+    /// The Context Identifier is already in use.
+    #[error("the guest CID {0} is already in use")]
+    GuestCIDAlreadyInUse(u32),
+
+    /// The Unix Domain Socket path is already in use.
+    #[error("the Unix Domain Socket path {0} is already in use")]
+    UDSPathAlreadyInUse(String),
+
+    /// The net address is already in use.
+    #[error("the net address {0} is already in use")]
+    NetAddrAlreadyInUse(String),
+
+    /// The update is not allowed after booting the microvm.
+    #[error("update operation is not allowed after boot")]
+    UpdateNotAllowedPostBoot,
+
+    /// The VsockId Already Exists
+    #[error("vsock id {0} already exists")]
+    VsockIdAlreadyExists(String),
+
+    /// Inner backend create error
+    #[error("vsock inner backend create error: {0}")]
+    CreateInnerBackend(#[source] std::io::Error),
+}
+
+/// Configuration information for a vsock device.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct VsockDeviceConfigInfo {
+    /// ID of the vsock device.
+    pub id: String,
+    /// A 32-bit Context Identifier (CID) used to identify the guest.
+    pub guest_cid: u32,
+    /// unix domain socket path.
+    pub uds_path: Option<String>,
+    /// tcp socket address.
+    pub tcp_addr: Option<String>,
+    /// Virtio queue size.
+    pub queue_size: Vec<u16>,
+    /// Use shared irq
+    pub use_shared_irq: Option<bool>,
+    /// Use generic irq
+    pub use_generic_irq: Option<bool>,
+}
+
+impl Default for VsockDeviceConfigInfo {
+    fn default() -> Self {
+        Self {
+            id: String::default(),
+            guest_cid: 0,
+            uds_path: None,
+            tcp_addr: None,
+            queue_size: Vec::from(QUEUE_SIZES),
+            use_shared_irq: None,
+            use_generic_irq: None,
+        }
+    }
+}
+
+impl VsockDeviceConfigInfo {
+    /// Get number and size of queues supported.
+    pub fn queue_sizes(&self) -> Vec<u16> {
+        self.queue_size.clone()
+    }
+}
+
+impl ConfigItem for VsockDeviceConfigInfo {
+    type Err = VsockDeviceError;
+
+    fn id(&self) -> &str {
+        &self.id
+    }
+
+    fn check_conflicts(&self, other: &Self) -> Result<(), VsockDeviceError> {
+        if self.id == other.id {
+            return Err(VsockDeviceError::DeviceIDAlreadyExist(self.id.clone()));
+        }
+        if self.guest_cid == other.guest_cid {
+            return Err(VsockDeviceError::GuestCIDAlreadyInUse(self.guest_cid));
+        }
+        if let (Some(self_uds_path), Some(other_uds_path)) =
+            (self.uds_path.as_ref(), other.uds_path.as_ref())
+        {
+            if self_uds_path == other_uds_path {
+                return Err(VsockDeviceError::UDSPathAlreadyInUse(self_uds_path.clone()));
+            }
+        }
+        if let (Some(self_net_addr), Some(other_net_addr)) =
+            (self.tcp_addr.as_ref(), other.tcp_addr.as_ref())
+        {
+            if self_net_addr == other_net_addr {
+                return Err(VsockDeviceError::NetAddrAlreadyInUse(self_net_addr.clone()));
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// Vsock Device Info
+pub type VsockDeviceInfo = DeviceConfigInfo<VsockDeviceConfigInfo>;
+
+/// Device manager to manage all vsock devices.
+pub struct VsockDeviceMgr {
+    pub(crate) info_list: DeviceConfigInfos<VsockDeviceConfigInfo>,
+    pub(crate) default_inner_backend: Option<VsockInnerBackend>,
+    pub(crate) default_inner_connector: Option<VsockInnerConnector>,
+    pub(crate) use_shared_irq: bool,
+}
+
+impl VsockDeviceMgr {
+    /// Insert or update a vsock device into the manager.
+    pub fn insert_device(
+        &mut self,
+        ctx: DeviceOpContext,
+        config: VsockDeviceConfigInfo,
+    ) -> std::result::Result<(), VsockDeviceError> {
+        if ctx.is_hotplug {
+            slog::error!(
+                ctx.logger(),
+                "no support of virtio-vsock device hotplug";
+                "subsystem" => SUBSYSTEM,
+                "id" => &config.id,
+                "uds_path" => &config.uds_path,
+            );
+
+            return Err(VsockDeviceError::UpdateNotAllowedPostBoot);
+        }
+
+        // VMADDR_CID_ANY (-1U) means any address for binding;
+        // VMADDR_CID_HYPERVISOR (0) is reserved for services built into the hypervisor;
+        // VMADDR_CID_RESERVED (1) must not be used;
+        // VMADDR_CID_HOST (2) is the well-known address of the host.
+        if config.guest_cid <= 2 {
+            return Err(VsockDeviceError::GuestCIDInvalid(config.guest_cid));
+        }
+
+        slog::info!(
+            ctx.logger(),
+            "add virtio-vsock device configuration";
+            "subsystem" => SUBSYSTEM,
+            "id" => &config.id,
+            "uds_path" => &config.uds_path,
+        );
+
+        self.lazy_make_default_connector()?;
+
+        self.info_list.insert_or_update(&config)?;
+
+        Ok(())
+    }
+
+    /// Attach all configured vsock device to the virtual machine instance.
+    pub fn attach_devices(
+        &mut self,
+        ctx: &mut DeviceOpContext,
+    ) -> std::result::Result<(), StartMicroVmError> {
+        let epoll_mgr = ctx
+            .epoll_mgr
+            .clone()
+            .ok_or(StartMicroVmError::CreateVsockDevice(
+                virtio::Error::InvalidInput,
+            ))?;
+
+        for info in self.info_list.iter_mut() {
+            slog::info!(
+                ctx.logger(),
+                "attach virtio-vsock device";
+                "subsystem" => SUBSYSTEM,
+                "id" => &info.config.id,
+                "uds_path" => &info.config.uds_path,
+            );
+
+            let mut device = Box::new(
+                Vsock::new(
+                    info.config.guest_cid as u64,
+                    Arc::new(info.config.queue_sizes()),
+                    epoll_mgr.clone(),
+                )
+                .map_err(VirtioError::VirtioVsockError)
+                .map_err(StartMicroVmError::CreateVsockDevice)?,
+            );
+            if let Some(uds_path) = info.config.uds_path.as_ref() {
+                let unix_backend = VsockUnixStreamBackend::new(uds_path.clone())
+                    .map_err(VirtioError::VirtioVsockError)
+                    .map_err(StartMicroVmError::CreateVsockDevice)?;
+                device
+                    .add_backend(Box::new(unix_backend), true)
+                    .map_err(VirtioError::VirtioVsockError)
+                    .map_err(StartMicroVmError::CreateVsockDevice)?;
+            }
+            if let Some(tcp_addr) = info.config.tcp_addr.as_ref() {
+                let tcp_backend = VsockTcpBackend::new(tcp_addr.clone())
+                    .map_err(VirtioError::VirtioVsockError)
+                    .map_err(StartMicroVmError::CreateVsockDevice)?;
+                device
+                    .add_backend(Box::new(tcp_backend), false)
+                    .map_err(VirtioError::VirtioVsockError)
+                    .map_err(StartMicroVmError::CreateVsockDevice)?;
+            }
+            // add inner backend to the the first added vsock device
+            if let Some(inner_backend) = self.default_inner_backend.take() {
+                device
+                    .add_backend(Box::new(inner_backend), false)
+                    .map_err(VirtioError::VirtioVsockError)
+                    .map_err(StartMicroVmError::CreateVsockDevice)?;
+            }
+            let device = DeviceManager::create_mmio_virtio_device_with_features(
+                device,
+                ctx,
+                Some(DRAGONBALL_FEATURE_INTR_USED),
+                info.config.use_shared_irq.unwrap_or(self.use_shared_irq),
+                info.config.use_generic_irq.unwrap_or(USE_GENERIC_IRQ),
+            )
+            .map_err(StartMicroVmError::RegisterVsockDevice)?;
+            info.device = Some(device);
+        }
+
+        Ok(())
+    }
+
+    // check the default connector is present, or build it.
+    fn lazy_make_default_connector(&mut self) -> std::result::Result<(), VsockDeviceError> {
+        if self.default_inner_connector.is_none() {
+            let inner_backend =
+                VsockInnerBackend::new().map_err(VsockDeviceError::CreateInnerBackend)?;
+            self.default_inner_connector = Some(inner_backend.get_connector());
+            self.default_inner_backend = Some(inner_backend);
+        }
+        Ok(())
+    }
+
+    /// Get the default vsock inner connector.
+    pub fn get_default_connector(
+        &mut self,
+    ) -> std::result::Result<VsockInnerConnector, VsockDeviceError> {
+        self.lazy_make_default_connector()?;
+
+        // safe to unwrap, because we created the inner connector before
+        Ok(self.default_inner_connector.clone().unwrap())
+    }
+}
+
+impl Default for VsockDeviceMgr {
+    /// Create a new Vsock device manager.
+    fn default() -> Self {
+        VsockDeviceMgr {
+            info_list: DeviceConfigInfos::new(),
+            default_inner_backend: None,
+            default_inner_connector: None,
+            use_shared_irq: USE_SHARED_IRQ,
+        }
+    }
+}

src/dragonball/src/error.rs

@@ -0,0 +1,224 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file
+
+//! Error codes for the virtual machine monitor subsystem.
+
+#[cfg(feature = "dbs-virtio-devices")]
+use dbs_virtio_devices::Error as VirtIoError;
+
+use crate::{address_space_manager, device_manager, resource_manager, vcpu, vm};
+
+/// Shorthand result type for internal VMM commands.
+pub type Result<T> = std::result::Result<T, Error>;
+
+/// Errors associated with the VMM internal logic.
+///
+/// These errors cannot be generated by direct user input, but can result from bad configuration
+/// of the host (for example if Dragonball doesn't have permissions to open the KVM fd).
+#[derive(Debug, thiserror::Error)]
+pub enum Error {
+    /// Empty AddressSpace from parameters.
+    #[error("Empty AddressSpace from parameters")]
+    AddressSpace,
+
+    /// The zero page extends past the end of guest_mem.
+    #[error("the guest zero page extends past the end of guest memory")]
+    ZeroPagePastRamEnd,
+
+    /// Error writing the zero page of guest memory.
+    #[error("failed to write to guest zero page")]
+    ZeroPageSetup,
+
+    /// Failure occurs in issuing KVM ioctls and errors will be returned from kvm_ioctls lib.
+    #[error("failure in issuing KVM ioctl command: {0}")]
+    Kvm(#[source] kvm_ioctls::Error),
+
+    /// The host kernel reports an unsupported KVM API version.
+    #[error("unsupported KVM version {0}")]
+    KvmApiVersion(i32),
+
+    /// Cannot initialize the KVM context due to missing capabilities.
+    #[error("missing KVM capability: {0:?}")]
+    KvmCap(kvm_ioctls::Cap),
+
+    #[cfg(target_arch = "x86_64")]
+    #[error("failed to configure MSRs: {0:?}")]
+    /// Cannot configure MSRs
+    GuestMSRs(dbs_arch::msr::Error),
+
+    /// MSR inner error
+    #[error("MSR inner error")]
+    Msr(vmm_sys_util::fam::Error),
+
+    /// Error writing MP table to memory.
+    #[cfg(target_arch = "x86_64")]
+    #[error("failed to write MP table to guest memory: {0}")]
+    MpTableSetup(#[source] dbs_boot::mptable::Error),
+
+    /// Fail to boot system
+    #[error("failed to boot system: {0}")]
+    BootSystem(#[source] dbs_boot::Error),
+
+    /// Cannot open the VM file descriptor.
+    #[error(transparent)]
+    Vm(vm::VmError),
+}
+
+/// Errors associated with starting the instance.
+#[derive(Debug, thiserror::Error)]
+pub enum StartMicroVmError {
+    /// Failed to allocate resources.
+    #[error("cannot allocate resources")]
+    AllocateResource(#[source] resource_manager::ResourceError),
+
+    /// Cannot read from an Event file descriptor.
+    #[error("failure while reading from EventFd file descriptor")]
+    EventFd,
+
+    /// Cannot add event to Epoll.
+    #[error("failure while registering epoll event for file descriptor")]
+    RegisterEvent,
+
+    /// The start command was issued more than once.
+    #[error("the virtual machine is already running")]
+    MicroVMAlreadyRunning,
+
+    /// Cannot start the VM because the kernel was not configured.
+    #[error("cannot start the virtual machine without kernel configuration")]
+    MissingKernelConfig,
+
+    #[cfg(feature = "hotplug")]
+    /// Upcall initialize miss vsock device.
+    #[error("the upcall client needs a virtio-vsock device for communication")]
+    UpcallMissVsock,
+
+    /// Upcall is not ready
+    #[error("the upcall client is not ready")]
+    UpcallNotReady,
+
+    /// Configuration passed in is invalidate.
+    #[error("invalid virtual machine configuration: {0} ")]
+    ConfigureInvalid(String),
+
+    /// This error is thrown by the minimal boot loader implementation.
+    /// It is related to a faulty memory configuration.
+    #[error("failure while configuring boot information for the virtual machine: {0}")]
+    ConfigureSystem(#[source] Error),
+
+    /// Cannot configure the VM.
+    #[error("failure while configuring the virtual machine: {0}")]
+    ConfigureVm(#[source] vm::VmError),
+
+    /// Cannot load initrd.
+    #[error("cannot load Initrd into guest memory: {0}")]
+    InitrdLoader(#[from] LoadInitrdError),
+
+    /// Cannot load kernel due to invalid memory configuration or invalid kernel image.
+    #[error("cannot load guest kernel into guest memory: {0}")]
+    KernelLoader(#[source] linux_loader::loader::Error),
+
+    /// Cannot load command line string.
+    #[error("failure while configuring guest kernel commandline: {0}")]
+    LoadCommandline(#[source] linux_loader::loader::Error),
+
+    /// The device manager was not configured.
+    #[error("the device manager failed to manage devices: {0}")]
+    DeviceManager(#[source] device_manager::DeviceMgrError),
+
+    /// Cannot add devices to the Legacy I/O Bus.
+    #[error("failure in managing legacy device: {0}")]
+    LegacyDevice(#[source] device_manager::LegacyDeviceError),
+
+    #[cfg(feature = "virtio-vsock")]
+    /// Failed to create the vsock device.
+    #[error("cannot create virtio-vsock device: {0}")]
+    CreateVsockDevice(#[source] VirtIoError),
+
+    #[cfg(feature = "virtio-vsock")]
+    /// Cannot initialize a MMIO Vsock Device or add a device to the MMIO Bus.
+    #[error("failure while registering virtio-vsock device: {0}")]
+    RegisterVsockDevice(#[source] device_manager::DeviceMgrError),
+
+    /// Address space manager related error, e.g.cannot access guest address space manager.
+    #[error("address space manager related error: {0}")]
+    AddressManagerError(#[source] address_space_manager::AddressManagerError),
+
+    /// Cannot create a new vCPU file descriptor.
+    #[error("vCPU related error: {0}")]
+    Vcpu(#[source] vcpu::VcpuManagerError),
+
+    #[cfg(all(feature = "hotplug", feature = "dbs-upcall"))]
+    /// Upcall initialize Error.
+    #[error("failure while initializing the upcall client: {0}")]
+    UpcallInitError(#[source] dbs_upcall::UpcallClientError),
+
+    #[cfg(all(feature = "hotplug", feature = "dbs-upcall"))]
+    /// Upcall connect Error.
+    #[error("failure while connecting the upcall client: {0}")]
+    UpcallConnectError(#[source] dbs_upcall::UpcallClientError),
+
+    #[cfg(feature = "virtio-blk")]
+    /// Virtio-blk errors.
+    #[error("virtio-blk errors: {0}")]
+    BlockDeviceError(#[source] device_manager::blk_dev_mgr::BlockDeviceError),
+
+    #[cfg(feature = "virtio-net")]
+    /// Virtio-net errors.
+    #[error("virtio-net errors: {0}")]
+    VirtioNetDeviceError(#[source] device_manager::virtio_net_dev_mgr::VirtioNetDeviceError),
+
+    #[cfg(feature = "virtio-fs")]
+    /// Virtio-fs errors.
+    #[error("virtio-fs errors: {0}")]
+    FsDeviceError(#[source] device_manager::fs_dev_mgr::FsDeviceError),
+}
+
+/// Errors associated with starting the instance.
+#[derive(Debug, thiserror::Error)]
+pub enum StopMicrovmError {
+    /// Guest memory has not been initialized.
+    #[error("Guest memory has not been initialized")]
+    GuestMemoryNotInitialized,
+
+    /// Cannnot remove devices
+    #[error("Failed to remove devices in device_manager {0}")]
+    DeviceManager(#[source] device_manager::DeviceMgrError),
+}
+
+/// Errors associated with loading initrd
+#[derive(Debug, thiserror::Error)]
+pub enum LoadInitrdError {
+    /// Cannot load initrd due to an invalid memory configuration.
+    #[error("failed to load the initrd image to guest memory")]
+    LoadInitrd,
+    /// Cannot load initrd due to an invalid image.
+    #[error("failed to read the initrd image: {0}")]
+    ReadInitrd(#[source] std::io::Error),
+}
+
+/// A dedicated error type to glue with the vmm_epoll crate.
+#[derive(Debug, thiserror::Error)]
+pub enum EpollError {
+    /// Generic internal error.
+    #[error("unclassfied internal error")]
+    InternalError,
+
+    /// Errors from the epoll subsystem.
+    #[error("failed to issue epoll syscall: {0}")]
+    EpollMgr(#[from] dbs_utils::epoll_manager::Error),
+
+    /// Generic IO errors.
+    #[error(transparent)]
+    IOError(std::io::Error),
+
+    #[cfg(feature = "dbs-virtio-devices")]
+    /// Errors from virtio devices.
+    #[error("failed to manager Virtio device: {0}")]
+    VirtIoDevice(#[source] VirtIoError),
+}

src/dragonball/src/event_manager.rs

@@ -0,0 +1,169 @@
+// Copyright (C) 2020-2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+
+//! Event manager to manage and handle IO events and requests from API server .
+
+use std::sync::atomic::{AtomicUsize, Ordering};
+use std::sync::{Arc, Mutex};
+
+use dbs_utils::epoll_manager::{
+    EpollManager, EventOps, EventSet, Events, MutEventSubscriber, SubscriberId,
+};
+use log::{error, warn};
+use vmm_sys_util::eventfd::EventFd;
+
+use crate::error::{EpollError, Result};
+use crate::vmm::Vmm;
+
+// Statically assigned epoll slot for VMM events.
+pub(crate) const EPOLL_EVENT_EXIT: u32 = 0;
+pub(crate) const EPOLL_EVENT_API_REQUEST: u32 = 1;
+
+/// Shared information between vmm::vmm_thread_event_loop() and VmmEpollHandler.
+pub(crate) struct EventContext {
+    pub api_event_fd: EventFd,
+    pub api_event_triggered: bool,
+    pub exit_evt_triggered: bool,
+}
+
+impl EventContext {
+    /// Create a new instance of [`EventContext`].
+    pub fn new(api_event_fd: EventFd) -> Result<Self> {
+        Ok(EventContext {
+            api_event_fd,
+            api_event_triggered: false,
+            exit_evt_triggered: false,
+        })
+    }
+}
+
+/// Event manager for VMM to handle API requests and IO events.
+pub struct EventManager {
+    epoll_mgr: EpollManager,
+    subscriber_id: SubscriberId,
+    vmm_event_count: Arc<AtomicUsize>,
+}
+
+impl Drop for EventManager {
+    fn drop(&mut self) {
+        // Vmm -> Vm -> EpollManager ->  VmmEpollHandler -> Vmm
+        // We need to remove VmmEpollHandler to break the circular reference
+        // so that Vmm can drop.
+        self.epoll_mgr
+            .remove_subscriber(self.subscriber_id)
+            .map_err(|e| {
+                error!("event_manager: remove_subscriber err. {:?}", e);
+                e
+            })
+            .ok();
+    }
+}
+
+impl EventManager {
+    /// Create a new event manager associated with the VMM object.
+    pub fn new(vmm: &Arc<Mutex<Vmm>>, epoll_mgr: EpollManager) -> Result<Self> {
+        let vmm_event_count = Arc::new(AtomicUsize::new(0));
+        let handler: Box<dyn MutEventSubscriber + Send> = Box::new(VmmEpollHandler {
+            vmm: vmm.clone(),
+            vmm_event_count: vmm_event_count.clone(),
+        });
+        let subscriber_id = epoll_mgr.add_subscriber(handler);
+
+        Ok(EventManager {
+            epoll_mgr,
+            subscriber_id,
+            vmm_event_count,
+        })
+    }
+
+    /// Get the underlying epoll event manager.
+    pub fn epoll_manager(&self) -> EpollManager {
+        self.epoll_mgr.clone()
+    }
+
+    /// Registry the eventfd for exit notification.
+    pub fn register_exit_eventfd(
+        &mut self,
+        exit_evt: &EventFd,
+    ) -> std::result::Result<(), EpollError> {
+        let events = Events::with_data(exit_evt, EPOLL_EVENT_EXIT, EventSet::IN);
+
+        self.epoll_mgr
+            .add_event(self.subscriber_id, events)
+            .map_err(EpollError::EpollMgr)
+    }
+
+    /// Poll pending events and invoke registered event handler.
+    ///
+    /// # Arguments:
+    /// * max_events: maximum number of pending events to handle
+    /// * timeout: maximum time in milliseconds to wait
+    pub fn handle_events(&self, timeout: i32) -> std::result::Result<usize, EpollError> {
+        self.epoll_mgr
+            .handle_events(timeout)
+            .map_err(EpollError::EpollMgr)
+    }
+
+    /// Fetch the VMM event count and reset it to zero.
+    pub fn fetch_vmm_event_count(&self) -> usize {
+        self.vmm_event_count.swap(0, Ordering::AcqRel)
+    }
+}
+
+struct VmmEpollHandler {
+    vmm: Arc<Mutex<Vmm>>,
+    vmm_event_count: Arc<AtomicUsize>,
+}
+
+impl MutEventSubscriber for VmmEpollHandler {
+    fn process(&mut self, events: Events, _ops: &mut EventOps) {
+        // Do not try to recover when the lock has already been poisoned.
+        // And be careful to avoid deadlock between process() and vmm::vmm_thread_event_loop().
+        let mut vmm = self.vmm.lock().unwrap();
+
+        match events.data() {
+            EPOLL_EVENT_API_REQUEST => {
+                if let Err(e) = vmm.event_ctx.api_event_fd.read() {
+                    error!("event_manager: failed to read API eventfd, {:?}", e);
+                }
+                vmm.event_ctx.api_event_triggered = true;
+                self.vmm_event_count.fetch_add(1, Ordering::AcqRel);
+            }
+            EPOLL_EVENT_EXIT => {
+                let vm = vmm.get_vm().unwrap();
+                match vm.get_reset_eventfd() {
+                    Some(ev) => {
+                        if let Err(e) = ev.read() {
+                            error!("event_manager: failed to read exit eventfd, {:?}", e);
+                        }
+                    }
+                    None => warn!("event_manager: leftover exit event in epoll context!"),
+                }
+                vmm.event_ctx.exit_evt_triggered = true;
+                self.vmm_event_count.fetch_add(1, Ordering::AcqRel);
+            }
+            _ => error!("event_manager: unknown epoll slot number {}", events.data()),
+        }
+    }
+
+    fn init(&mut self, ops: &mut EventOps) {
+        // Do not expect poisoned lock.
+        let vmm = self.vmm.lock().unwrap();
+        let events = Events::with_data(
+            &vmm.event_ctx.api_event_fd,
+            EPOLL_EVENT_API_REQUEST,
+            EventSet::IN,
+        );
+        if let Err(e) = ops.add(events) {
+            error!(
+                "event_manager: failed to register epoll event for API server, {:?}",
+                e
+            );
+        }
+    }
+}

src/dragonball/src/io_manager.rs

@@ -0,0 +1,60 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+use std::sync::Arc;
+
+use arc_swap::{ArcSwap, Cache};
+use dbs_device::device_manager::Error;
+use dbs_device::device_manager::IoManager;
+
+/// A specialized version of [`std::result::Result`] for IO manager related operations.
+pub type Result<T> = std::result::Result<T, Error>;
+
+/// Wrapper over IoManager to support device hotplug with [`ArcSwap`] and [`Cache`].
+#[derive(Clone)]
+pub struct IoManagerCached(pub(crate) Cache<Arc<ArcSwap<IoManager>>, Arc<IoManager>>);
+
+impl IoManagerCached {
+    /// Create a new instance of [`IoManagerCached`].
+    pub fn new(io_manager: Arc<ArcSwap<IoManager>>) -> Self {
+        IoManagerCached(Cache::new(io_manager))
+    }
+
+    #[cfg(target_arch = "x86_64")]
+    #[inline]
+    /// Read data from IO ports.
+    pub fn pio_read(&mut self, addr: u16, data: &mut [u8]) -> Result<()> {
+        self.0.load().pio_read(addr, data)
+    }
+
+    #[cfg(target_arch = "x86_64")]
+    #[inline]
+    /// Write data to IO ports.
+    pub fn pio_write(&mut self, addr: u16, data: &[u8]) -> Result<()> {
+        self.0.load().pio_write(addr, data)
+    }
+
+    #[inline]
+    /// Read data to MMIO address.
+    pub fn mmio_read(&mut self, addr: u64, data: &mut [u8]) -> Result<()> {
+        self.0.load().mmio_read(addr, data)
+    }
+
+    #[inline]
+    /// Write data to MMIO address.
+    pub fn mmio_write(&mut self, addr: u64, data: &[u8]) -> Result<()> {
+        self.0.load().mmio_write(addr, data)
+    }
+
+    #[inline]
+    /// Revalidate the inner cache
+    pub fn revalidate_cache(&mut self) {
+        let _ = self.0.load();
+    }
+
+    #[inline]
+    /// Get immutable reference to underlying [`IoManager`].
+    pub fn load(&mut self) -> &IoManager {
+        self.0.load()
+    }
+}

src/dragonball/src/kvm_context.rs

@@ -0,0 +1,251 @@
+// Copyright (C) 2022 Alibaba Cloud. All rights reserved.
+// Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// Portions Copyright 2017 The Chromium OS Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the THIRD-PARTY file.
+#![allow(dead_code)]
+use kvm_bindings::KVM_API_VERSION;
+use kvm_ioctls::{Cap, Kvm, VmFd};
+use std::os::unix::io::{FromRawFd, RawFd};
+
+use crate::error::{Error, Result};
+
+/// Describes a KVM context that gets attached to the micro VM instance.
+/// It gives access to the functionality of the KVM wrapper as long as every required
+/// KVM capability is present on the host.
+pub struct KvmContext {
+    kvm: Kvm,
+    max_memslots: usize,
+    #[cfg(target_arch = "x86_64")]
+    supported_msrs: kvm_bindings::MsrList,
+}
+
+impl KvmContext {
+    /// Create a new KVM context object, using the provided `kvm_fd` if one is presented.
+    pub fn new(kvm_fd: Option<RawFd>) -> Result<Self> {
+        let kvm = if let Some(fd) = kvm_fd {
+            // Safe because we expect kvm_fd to contain a valid fd number when is_some() == true.
+            unsafe { Kvm::from_raw_fd(fd) }
+        } else {
+            Kvm::new().map_err(Error::Kvm)?
+        };
+
+        if kvm.get_api_version() != KVM_API_VERSION as i32 {
+            return Err(Error::KvmApiVersion(kvm.get_api_version()));
+        }
+
+        Self::check_cap(&kvm, Cap::Irqchip)?;
+        Self::check_cap(&kvm, Cap::Irqfd)?;
+        Self::check_cap(&kvm, Cap::Ioeventfd)?;
+        Self::check_cap(&kvm, Cap::UserMemory)?;
+        #[cfg(target_arch = "x86_64")]
+        Self::check_cap(&kvm, Cap::SetTssAddr)?;
+
+        #[cfg(target_arch = "x86_64")]
+        let supported_msrs = dbs_arch::msr::supported_guest_msrs(&kvm).map_err(Error::GuestMSRs)?;
+        let max_memslots = kvm.get_nr_memslots();
+
+        Ok(KvmContext {
+            kvm,
+            max_memslots,
+            #[cfg(target_arch = "x86_64")]
+            supported_msrs,
+        })
+    }
+
+    /// Get underlying KVM object to access kvm-ioctls interfaces.
+    pub fn kvm(&self) -> &Kvm {
+        &self.kvm
+    }
+
+    /// Get the maximum number of memory slots reported by this KVM context.
+    pub fn max_memslots(&self) -> usize {
+        self.max_memslots
+    }
+
+    /// Create a virtual machine object.
+    pub fn create_vm(&self) -> Result<VmFd> {
+        self.kvm.create_vm().map_err(Error::Kvm)
+    }
+
+    /// Get the max vcpu count supported by kvm
+    pub fn get_max_vcpus(&self) -> usize {
+        self.kvm.get_max_vcpus()
+    }
+
+    fn check_cap(kvm: &Kvm, cap: Cap) -> std::result::Result<(), Error> {
+        if !kvm.check_extension(cap) {
+            return Err(Error::KvmCap(cap));
+        }
+        Ok(())
+    }
+}
+
+#[cfg(target_arch = "x86_64")]
+mod x86_64 {
+    use super::*;
+    use dbs_arch::msr::*;
+    use kvm_bindings::{kvm_msr_entry, CpuId, MsrList, Msrs};
+    use std::collections::HashSet;
+
+    impl KvmContext {
+        /// Get information about supported CPUID of x86 processor.
+        pub fn supported_cpuid(
+            &self,
+            max_entries_count: usize,
+        ) -> std::result::Result<CpuId, kvm_ioctls::Error> {
+            self.kvm.get_supported_cpuid(max_entries_count)
+        }
+
+        /// Get information about supported MSRs of x86 processor.
+        pub fn supported_msrs(
+            &self,
+            _max_entries_count: usize,
+        ) -> std::result::Result<MsrList, kvm_ioctls::Error> {
+            Ok(self.supported_msrs.clone())
+        }
+
+        // It's very sensible to manipulate MSRs, so please be careful to change code below.
+        fn build_msrs_list(kvm: &Kvm) -> Result<Msrs> {
+            let mut mset: HashSet<u32> = HashSet::new();
+            let supported_msr_list = kvm.get_msr_index_list().map_err(super::Error::Kvm)?;
+            for msr in supported_msr_list.as_slice() {
+                mset.insert(*msr);
+            }
+
+            let mut msrs = vec![
+                MSR_IA32_APICBASE,
+                MSR_IA32_SYSENTER_CS,
+                MSR_IA32_SYSENTER_ESP,
+                MSR_IA32_SYSENTER_EIP,
+                MSR_IA32_CR_PAT,
+            ];
+
+            let filters_list = vec![
+                MSR_STAR,
+                MSR_VM_HSAVE_PA,
+                MSR_TSC_AUX,
+                MSR_IA32_TSC_ADJUST,
+                MSR_IA32_TSCDEADLINE,
+                MSR_IA32_MISC_ENABLE,
+                MSR_IA32_BNDCFGS,
+                MSR_IA32_SPEC_CTRL,
+            ];
+            for msr in filters_list {
+                if mset.contains(&msr) {
+                    msrs.push(msr);
+                }
+            }
+
+            // TODO: several msrs are optional.
+
+            // TODO: Since our guests don't support nested-vmx, LMCE nor SGX for now.
+            // msrs.push(MSR_IA32_FEATURE_CONTROL);
+
+            msrs.push(MSR_CSTAR);
+            msrs.push(MSR_KERNEL_GS_BASE);
+            msrs.push(MSR_SYSCALL_MASK);
+            msrs.push(MSR_LSTAR);
+            msrs.push(MSR_IA32_TSC);
+
+            msrs.push(MSR_KVM_SYSTEM_TIME_NEW);
+            msrs.push(MSR_KVM_WALL_CLOCK_NEW);
+
+            // FIXME: check if it's supported.
+            msrs.push(MSR_KVM_ASYNC_PF_EN);
+            msrs.push(MSR_KVM_PV_EOI_EN);
+            msrs.push(MSR_KVM_STEAL_TIME);
+
+            msrs.push(MSR_CORE_PERF_FIXED_CTR_CTRL);
+            msrs.push(MSR_CORE_PERF_GLOBAL_CTRL);
+            msrs.push(MSR_CORE_PERF_GLOBAL_STATUS);
+            msrs.push(MSR_CORE_PERF_GLOBAL_OVF_CTRL);
+
+            const MAX_FIXED_COUNTERS: u32 = 3;
+            for i in 0..MAX_FIXED_COUNTERS {
+                msrs.push(MSR_CORE_PERF_FIXED_CTR0 + i);
+            }
+
+            // FIXME: skip MCE for now.
+
+            let mtrr_msrs = vec![
+                MSR_MTRRdefType,
+                MSR_MTRRfix64K_00000,
+                MSR_MTRRfix16K_80000,
+                MSR_MTRRfix16K_A0000,
+                MSR_MTRRfix4K_C0000,
+                MSR_MTRRfix4K_C8000,
+                MSR_MTRRfix4K_D0000,
+                MSR_MTRRfix4K_D8000,
+                MSR_MTRRfix4K_E0000,
+                MSR_MTRRfix4K_E8000,
+                MSR_MTRRfix4K_F0000,
+                MSR_MTRRfix4K_F8000,
+            ];
+            for mtrr in mtrr_msrs {
+                msrs.push(mtrr);
+            }
+
+            const MSR_MTRRCAP_VCNT: u32 = 8;
+            for i in 0..MSR_MTRRCAP_VCNT {
+                msrs.push(0x200 + 2 * i);
+                msrs.push(0x200 + 2 * i + 1);
+            }
+
+            let msrs: Vec<kvm_msr_entry> = msrs
+                .iter()
+                .map(|reg| kvm_msr_entry {
+                    index: *reg,
+                    reserved: 0,
+                    data: 0,
+                })
+                .collect();
+
+            Msrs::from_entries(&msrs).map_err(super::Error::Msr)
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use kvm_ioctls::Kvm;
+    use std::fs::File;
+    use std::os::unix::fs::MetadataExt;
+    use std::os::unix::io::{AsRawFd, FromRawFd};
+
+    #[test]
+    fn test_create_kvm_context() {
+        let c = KvmContext::new(None).unwrap();
+
+        assert!(c.max_memslots >= 32);
+
+        let kvm = Kvm::new().unwrap();
+        let f = unsafe { File::from_raw_fd(kvm.as_raw_fd()) };
+        let m1 = f.metadata().unwrap();
+        let m2 = File::open("/dev/kvm").unwrap().metadata().unwrap();
+
+        assert_eq!(m1.dev(), m2.dev());
+        assert_eq!(m1.ino(), m2.ino());
+    }
+
+    #[cfg(target_arch = "x86_64")]
+    #[test]
+    fn test_get_supported_cpu_id() {
+        let c = KvmContext::new(None).unwrap();
+
+        let _ = c
+            .supported_cpuid(kvm_bindings::KVM_MAX_CPUID_ENTRIES)
+            .expect("failed to get supported CPUID");
+        assert!(c.supported_cpuid(0).is_err());
+    }
+
+    #[test]
+    fn test_create_vm() {
+        let c = KvmContext::new(None).unwrap();
+
+        let _ = c.create_vm().unwrap();
+    }
+}