build: Release v31.2 (bug fix release)

Signed-off-by: Bo Chen <chen.bo@intel.com>

.config/nextest.toml

@@ -1,3 +0,0 @@
-[profile.default]
-# Don't let one individual test run for more than 10 minutes
-slow-timeout = { period = "60s", terminate-after = 10 }

.editorconfig

@@ -1,25 +0,0 @@
-# https://editorconfig.org/
-#
-# Hints for editors to assist with correct formatting as you type.
-root = true
-
-# Unix-style newlines with a newline ending every file
-[*]
-charset = utf-8
-indent_size = 4
-end_of_line = lf
-indent_style = space
-insert_final_newline = true
-trim_trailing_whitespace = true
-# Recommendation, not enforced.
-max_line_length = 80
-
-[Makefile]
-indent_style = tab
-
-# Inherited as default
-# [*.sh]
-# indent_size = 4
-
-[{Cargo.lock,*.md,*.toml,*.yml,*.yaml}]
-indent_size = 2

.github/dependabot.yml

@@ -1,75 +1,18 @@
 version: 2
 updates:
 - package-ecosystem: cargo
-  directories:
-    - "/"
-    - "/fuzz"
-  schedule:
-    interval: weekly
-  allow:
-    - dependency-name: "acpi_tables"
-    - dependency-name: "kvm-bindings"
-    - dependency-name: "kvm-ioctls"
-    - dependency-name: "linux-loader"
-    - dependency-name: "micro_http"
-    - dependency-name: "mshv-bindings"
-    - dependency-name: "mshv-ioctls"
-    - dependency-name: "seccompiler"
-    - dependency-name: "vfio-bindings"
-    - dependency-name: "vfio-ioctls"
-    - dependency-name: "vfio_user"
-    - dependency-name: "vhost"
-    - dependency-name: "vhost-user-backend"
-    - dependency-name: "virtio-bindings"
-    - dependency-name: "virtio-queue"
-    - dependency-name: "vm-fdt"
-    - dependency-name: "vm-memory"
-    - dependency-name: "vmm-sys-util"
-  groups:
-      rust-vmm:
-        patterns:
-          - "*"
-- package-ecosystem: cargo
-  directories:
-    - "/"
-    - "/fuzz"
-  schedule:
-    interval: weekly
-  allow:
-    - dependency-type: all
-  cooldown:
-    default-days: 7
-    semver-major-days: 14
-    semver-minor-days: 7
-    semver-patch-days: 3
-  ignore:
-    - dependency-name: "acpi_tables"
-    - dependency-name: "kvm-bindings"
-    - dependency-name: "kvm-ioctls"
-    - dependency-name: "linux-loader"
-    - dependency-name: "micro_http"
-    - dependency-name: "mshv-bindings"
-    - dependency-name: "mshv-ioctls"
-    - dependency-name: "seccompiler"
-    - dependency-name: "vfio-bindings"
-    - dependency-name: "vfio-ioctls"
-    - dependency-name: "vfio_user"
-    - dependency-name: "vhost"
-    - dependency-name: "vhost-user-backend"
-    - dependency-name: "virtio-bindings"
-    - dependency-name: "virtio-queue"
-    - dependency-name: "vm-fdt"
-    - dependency-name: "vm-memory"
-    - dependency-name: "vmm-sys-util"
-  groups:
-      non-rust-vmm:
-        patterns:
-          - "*"
-  # Makes it possible to have another config for the same directory.
-  # https://github.com/dependabot/dependabot-core/issues/1778#issuecomment-1988140219
-  target-branch: main
-- package-ecosystem: github-actions
   directory: "/"
   schedule:
     interval: daily
-  open-pull-requests-limit: 1
+  open-pull-requests-limit: 1
+  allow:
+  - dependency-type: direct
+  - dependency-type: indirect
+- package-ecosystem: cargo
+  directory: "/fuzz"
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 1
+  allow:
+  - dependency-type: direct
+  - dependency-type: indirect

.github/workflows/audit.yaml

@@ -0,0 +1,15 @@
+name: Cloud Hypervisor Dependency Audit
+on:
+  pull_request:
+    paths: 
+      - '**/Cargo.toml'
+      - '**/Cargo.lock'
+jobs:
+  security_audit:
+    name: Audit
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v1
+      - uses: actions-rs/audit-check@v1
+        with:
+          token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/build.yaml

@@ -0,0 +1,58 @@
+name: Cloud Hypervisor Build
+on: [pull_request, create]
+
+jobs:
+  build:
+    if: github.event_name == 'pull_request'
+    name: Build
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        rust:
+          - stable
+          - beta
+          - nightly
+          - "1.62"
+        target:
+          - x86_64-unknown-linux-gnu
+          - x86_64-unknown-linux-musl
+    steps:
+      - name: Code checkout
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 0
+
+      - name: Install musl-gcc
+        run: sudo apt install -y musl-tools
+
+      - name: Install Rust toolchain (${{ matrix.rust }})
+        uses: actions-rs/toolchain@v1
+        with:
+          toolchain: ${{ matrix.rust }}
+          target: ${{ matrix.target }}
+          override: true
+
+      - name: Build (default features)
+        run: cargo rustc --locked --bin cloud-hypervisor -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Build (kvm)
+        run: cargo rustc --locked --bin cloud-hypervisor --no-default-features --features "kvm"  -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Build (default features + tdx)
+        run: cargo rustc --locked --bin cloud-hypervisor --features "tdx" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Build (default features + guest_debug)
+        run: cargo rustc --locked --bin cloud-hypervisor --features "guest_debug" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Build (mshv)
+        run: cargo rustc --locked --bin cloud-hypervisor --no-default-features --features "mshv"  -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Build (mshv + kvm)
+        run: cargo rustc --locked --bin cloud-hypervisor --no-default-features --features "mshv,kvm"  -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Release Build (default features)
+        run: cargo build --locked --all --release --target=${{ matrix.target }}
+
+      - name: Check build did not modify any files
+        run: test -z "$(git status --porcelain)"

.github/workflows/ci.yaml

@@ -1,770 +0,0 @@
-name: CI
-on: [pull_request, merge_group]
-permissions:
-  contents: read
-  pull-requests: read
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-${{ github.event_name }}
-  cancel-in-progress: true
-jobs:
-  preflight:
-    name: preflight
-    runs-on: ubuntu-latest
-    outputs:
-      full: ${{ steps.classify.outputs.full }}
-      rust: ${{ steps.changes.outputs.rust }}
-      cargo: ${{ steps.changes.outputs.cargo }}
-      openapi: ${{ steps.changes.outputs.openapi }}
-      dockerfile: ${{ steps.changes.outputs.dockerfile }}
-      shell: ${{ steps.changes.outputs.shell }}
-      ci: ${{ steps.changes.outputs.ci }}
-      docs: ${{ steps.changes.outputs.docs }}
-    steps:
-      - uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - id: changes
-        uses: dorny/paths-filter@fbd0ab8f3e69293af611ebaee6363fc25e6d187d # v4.0.1
-        with:
-          filters: |
-            rust:
-              - '**/*.rs'
-              - 'build.rs'
-              - '**/Cargo.toml'
-              - '**/Cargo.lock'
-              - 'rust-toolchain.toml'
-            cargo:
-              - '**/Cargo.toml'
-              - '**/Cargo.lock'
-            openapi:
-              - 'vmm/src/api/openapi/**'
-            dockerfile:
-              - 'resources/Dockerfile'
-            shell:
-              - '**/*.sh'
-              - 'scripts/**'
-            ci:
-              - '.github/workflows/**'
-            docs:
-              - 'docs/**'
-              - '**/*.md'
-              - '.github/ISSUE_TEMPLATE/**'
-              - 'LICENSES/**'
-              - 'CODEOWNERS'
-      - id: classify
-        name: Classify changes
-        run: |
-          set -eufo pipefail
-          full=false
-          if [[ "${{ steps.changes.outputs.rust }}" == "true" \
-             || "${{ steps.changes.outputs.dockerfile }}" == "true" \
-             || "${{ steps.changes.outputs.shell }}" == "true" \
-             || "${{ steps.changes.outputs.ci }}" == "true" ]]; then
-            full=true
-          fi
-          echo "full=$full" >> "$GITHUB_OUTPUT"
-          echo "full=$full"
-  dco:
-    name: dco
-    needs: [preflight]
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - name: Set up Python 3.x
-        uses: actions/setup-python@v6
-        with:
-          python-version: '3.x'
-      - name: Check DCO
-        if: github.event_name == 'pull_request'
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          set -eufo pipefail
-          pip3 install -U dco-check
-          dco-check -e "49699333+dependabot[bot]@users.noreply.github.com"
-  gitlint:
-    name: gitlint
-    needs: [preflight]
-    # PR-only: gitlint needs GITHUB_BASE_REF, unset on merge_group.
-    if: github.event_name == 'pull_request'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-        with:
-          # PR head, not the merge ref, so gitlint sees the PR's commits.
-          ref: ${{ github.event.pull_request.head.sha }}
-          fetch-depth: 0
-      - name: Set up Python 3.10
-        uses: actions/setup-python@v6
-        with:
-          python-version: "3.10"
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install --upgrade gitlint
-      - name: Lint git commit messages
-        run: |
-          gitlint --commits "origin/$GITHUB_BASE_REF.."
-  lychee:
-    name: lychee
-    needs: [preflight]
-    if: needs.preflight.outputs.docs == 'true' || needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Get changed files in PR
-        id: changed-files
-        uses: tj-actions/changed-files@9426d40962ed5378910ee2e21d5f8c6fcbf2dd96 # v47.0.6
-        with:
-          base_sha: ${{ github.event.pull_request.base.sha }}
-      - name: Verify Changed Files
-        run: |
-          set -eufo pipefail
-          echo "--- tj-actions/changed-files Outputs ---"
-          echo "any_changed: ${{ steps.changed-files.outputs.any_changed }}"
-          echo "all_changed_files: ${{ steps.changed-files.outputs.all_changed_files }}"
-          echo "added_files: ${{ steps.changed-files.outputs.added_files }}"
-          echo "modified_files: ${{ steps.changed-files.outputs.modified_files }}"
-          echo "deleted_files: ${{ steps.changed-files.outputs.deleted_files }}"
-          echo "renamed_files: ${{ steps.changed-files.outputs.renamed_files }}"
-          echo "----------------------------------------"
-          if [ -n "${{ steps.changed-files.outputs.all_changed_files }}" ]; then
-            echo "Detected changes: all_changed_files output is NOT empty."
-          else
-            echo "No changes detected: all_changed_files output IS empty."
-          fi
-      - name: Link Availability Check (Diff Only)
-        if: ${{ steps.changed-files.outputs.all_changed_files != '' }}
-        uses: lycheeverse/lychee-action@8646ba30535128ac92d33dfc9133794bfdd9b411 # v2.8.0
-        with:
-          args: --verbose --config .lychee.toml ${{ steps.changed-files.outputs.all_changed_files }}
-          failIfEmpty: false
-          fail: true
-  taplo:
-    name: taplo
-    needs: [preflight]
-    if: needs.preflight.outputs.cargo == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-      - name: Install Rust toolchain
-        uses: dtolnay/rust-toolchain@stable
-      - name: Install build dependencies
-        run: sudo apt-get update && sudo apt-get -yqq install build-essential libssl-dev
-      - name: Install taplo
-        run: cargo install taplo-cli --locked
-      - name: Check formatting
-        run: taplo fmt --check
-  audit:
-    name: audit
-    needs: [preflight]
-    if: needs.preflight.outputs.cargo == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - uses: actions-rust-lang/audit@v1
-        with:
-          token: ${{ secrets.GITHUB_TOKEN }}
-  shlint:
-    name: shlint
-    needs: [preflight]
-    if: needs.preflight.outputs.shell == 'true' || needs.preflight.outputs.ci == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-      - name: Run the shell script checkers
-        uses: luizm/action-sh-checker@883217215b11c1fabbf00eb1a9a041f62d74c744 # v0.10.0
-        env:
-          SHFMT_OPTS: -i 4 -d
-          SHELLCHECK_OPTS: -x --source-path scripts
-  hadolint:
-    name: hadolint
-    needs: [preflight]
-    if: needs.preflight.outputs.dockerfile == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v6
-      - name: Lint Dockerfile
-        uses: hadolint/hadolint-action@2332a7b74a6de0dda2e2221d575162eba76ba5e5 # v3.3.0
-        with:
-          dockerfile: ./resources/Dockerfile
-          format: tty
-          no-fail: false
-          verbose: true
-          failure-threshold: info
-  reuse:
-    name: reuse
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true' || needs.preflight.outputs.cargo == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - name: REUSE Compliance Check
-        uses: fsfe/reuse-action@v6
-  formatting:
-    name: formatting
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        rust: [nightly]
-        target:
-          - x86_64-unknown-linux-gnu
-          - aarch64-unknown-linux-musl
-    env:
-      RUSTFLAGS: -D warnings
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-      - name: Install Rust toolchain (${{ matrix.rust }})
-        uses: dtolnay/rust-toolchain@stable
-        with:
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          components: rustfmt
-      - name: Formatting (rustfmt)
-        run: cargo fmt --all -- --check
-      - name: Formatting (fuzz) (rustfmt)
-        run: cargo fmt --all --manifest-path fuzz/Cargo.toml -- --check
-  package-consistency:
-    name: package-consistency
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install dependencies
-        run: sudo apt install -y python3
-      - name: Install Rust toolchain stable
-        uses: dtolnay/rust-toolchain@stable
-        with:
-          toolchain: stable
-      - name: Check Rust VMM Package Consistency of root Workspace
-        run: python3 scripts/package-consistency-check.py github.com/rust-vmm
-      - name: Check Rust VMM Package Consistency of fuzz Workspace
-        run: |
-          set -eufo pipefail
-          pushd fuzz
-          python3 ../scripts/package-consistency-check.py github.com/rust-vmm
-          popd
-  fuzz-build:
-    name: fuzz-build
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        rust: [nightly]
-        target: [x86_64-unknown-linux-gnu]
-    env:
-      RUSTFLAGS: -D warnings
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-      - name: Install Rust toolchain (${{ matrix.rust }})
-        uses: dtolnay/rust-toolchain@stable
-        with:
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-      - name: Install Cargo fuzz
-        run: cargo install cargo-fuzz
-      - name: Fuzz Build
-        run: cargo fuzz build
-      - name: Fuzz Check
-        run: cargo fuzz check
-  openapi:
-    name: openapi
-    needs: [preflight]
-    if: needs.preflight.outputs.openapi == 'true'
-    runs-on: ubuntu-latest
-    container: openapitools/openapi-generator-cli
-    steps:
-      - uses: actions/checkout@v6
-      - name: Validate OpenAPI
-        run: |
-          /usr/local/bin/docker-entrypoint.sh validate -i vmm/src/api/openapi/cloud-hypervisor.yaml
-  typos:
-    name: typos
-    needs: [preflight]
-    if: github.event_name == 'pull_request'
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - uses: crate-ci/typos@5374cbf686e897b15713110e233094e2874de7ef # v1.46.1
-  quality:
-    name: quality
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    # Beta clippy is non-blocking; continue-on-error below keeps the
-    # aggregated needs.quality.result green when only beta fails.
-    continue-on-error: ${{ matrix.experimental }}
-    strategy:
-      fail-fast: false
-      matrix:
-        rust:
-          - beta
-          - stable
-        target:
-          - aarch64-unknown-linux-gnu
-          - aarch64-unknown-linux-musl
-          - x86_64-unknown-linux-gnu
-          - x86_64-unknown-linux-musl
-        include:
-          - rust: beta
-            experimental: true
-          - rust: stable
-            experimental: false
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install Rust toolchain (${{ matrix.rust }})
-        uses: dtolnay/rust-toolchain@stable
-        with:
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          override: true
-          components: clippy
-      - name: Bisectability Check (default features)
-        if: ${{ github.event_name == 'pull_request' && matrix.target == 'x86_64-unknown-linux-gnu' }}
-        run: |
-          set -eufo pipefail
-          commits=$(git rev-list origin/${{ github.base_ref }}..${{ github.sha }})
-          for commit in $commits; do git checkout $commit; cargo check --tests --examples --all --target=${{ matrix.target }}; done
-          git checkout ${{ github.sha }}
-      - name: Clippy (kvm)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "kvm" -- -D warnings
-      - name: Clippy (mshv)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "mshv" -- -D warnings
-      - name: Clippy (mshv + kvm)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "mshv,kvm" -- -D warnings
-      - name: Clippy (default features)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples -- -D warnings
-      - name: Clippy (default features + guest_debug)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples --features "guest_debug" -- -D warnings
-      - name: Clippy (default features + pvmemcontrol)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples --features "pvmemcontrol" -- -D warnings
-      - name: Clippy (default features + tracing)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples --features "tracing" -- -D warnings
-      - name: Clippy (default features + fw_cfg)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --target=${{ matrix.target }} --locked --all --all-targets --tests --examples --features "fw_cfg" -- -D warnings
-      - name: Clippy (default features + ivshmem)
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples --features "ivshmem" -- -D warnings
-      - name: Clippy (sev_snp)
-        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "sev_snp" -- -D warnings
-      - name: Clippy (igvm)
-        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "igvm" -- -D warnings
-      - name: Clippy (kvm + tdx)
-        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "tdx,kvm" -- -D warnings
-      - name: Clippy (kvm + igvm + sev_snp + fw_cfg)
-        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          cross-version: 3e0957637b49b1bbced23ad909170650c5b70635
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --no-default-features --tests --examples --features "kvm,igvm,sev_snp,fw_cfg" -- -D warnings
-      - name: Clippy (default features + sev_snp + igvm + fw_cfg)
-        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
-        uses: houseabsolute/actions-rust-cross@v1
-        with:
-          command: clippy
-          cross-version: 3e0957637b49b1bbced23ad909170650c5b70635
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-          args: --locked --all --all-targets --tests --examples --features "sev_snp,igvm,fw_cfg" -- -D warnings
-      - name: Check build did not modify any files
-        run: test -z "$(git status --porcelain)"
-  build:
-    name: build
-    needs: [preflight]
-    if: needs.preflight.outputs.full == 'true'
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        rust:
-          - stable
-          - beta
-          - nightly
-          - "1.89.0" # MSRV — keep quoted.
-        target:
-          - x86_64-unknown-linux-gnu
-          - x86_64-unknown-linux-musl
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install musl-gcc
-        run: sudo apt install -y musl-tools
-      - name: Install Rust toolchain (${{ matrix.rust }})
-        uses: dtolnay/rust-toolchain@stable
-        with:
-          toolchain: ${{ matrix.rust }}
-          target: ${{ matrix.target }}
-      - name: Build (default features)
-        run: cargo build --locked --bin cloud-hypervisor
-      - name: Build (kvm)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "kvm"
-      - name: Build (default features + tdx)
-        run: cargo build --locked --bin cloud-hypervisor --features "tdx"
-      - name: Build (default features + dbus_api)
-        run: cargo build --locked --bin cloud-hypervisor --features "dbus_api"
-      - name: Build (default features + guest_debug)
-        run: cargo build --locked --bin cloud-hypervisor --features "guest_debug"
-      - name: Build (default features + pvmemcontrol)
-        run: cargo build --locked --bin cloud-hypervisor --features "pvmemcontrol"
-      - name: Build (default features + fw_cfg)
-        run: cargo build --locked --bin cloud-hypervisor --features "fw_cfg"
-      - name: Build (default features + ivshmem)
-        run: cargo build --locked --bin cloud-hypervisor --features "ivshmem"
-      - name: Build (mshv)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "mshv"
-      - name: Build (sev_snp)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "sev_snp"
-      - name: Build (kvm + igvm + sev_snp + fw_cfg)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "kvm,igvm,sev_snp,fw_cfg"
-      - name: Build (igvm)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "igvm"
-      - name: Build (mshv + kvm)
-        run: cargo build --locked --bin cloud-hypervisor --no-default-features --features "mshv,kvm"
-      - name: Release Build (default features)
-        run: cargo build --locked --all --release --target=${{ matrix.target }}
-      - name: Check build did not modify any files
-        run: test -z "$(git status --porcelain)"
-  # garm-jammy + gnu: runs on PR and MQ. Other 3 matrix entries are in
-  # integration-x86-64-mq (sibling, MQ-only, runs in parallel).
-  integration-x86-64-pr:
-    name: integration-x86-64-pr
-    needs: [preflight, dco, quality, build]
-    if: >-
-      needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    timeout-minutes: 80
-    env:
-      # Our runner has 16 cores (nproc).
-      # We limit parallelism only to avoid exhausting disk space and memory
-      # resources, not to save CPU resources.
-      PARALLEL_INTEGRATION_TESTS_NUM: 12
-    runs-on: garm-jammy-16
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install Docker
-        run: |
-          set -eufo pipefail
-          sudo apt-get update
-          sudo apt-get -y install ca-certificates curl gnupg
-          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
-          sudo chmod a+r /usr/share/keyrings/docker-archive-keyring.gpg
-          echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
-          sudo apt-get update
-          sudo apt install -y docker-ce docker-ce-cli
-      - name: Prepare for VDPA
-        run: scripts/prepare_vdpa.sh
-      - name: Run unit tests
-        run: scripts/dev_cli.sh tests --unit --libc gnu
-      - name: Load openvswitch module
-        run: sudo modprobe openvswitch
-      - name: Run integration tests
-        timeout-minutes: 60
-        run: scripts/dev_cli.sh tests --integration --libc gnu
-  # MQ-only: the 3 matrix entries that integration-x86-64-pr does not cover.
-  integration-x86-64-mq:
-    name: integration-x86-64-mq
-    needs: [preflight, dco, quality, build]
-    if: >-
-      github.event_name == 'merge_group' && needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    timeout-minutes: 80
-    env:
-      # Our runner has 16 cores (nproc).
-      # We limit parallelism only to avoid exhausting disk space and memory
-      # resources, not to save CPU resources.
-      PARALLEL_INTEGRATION_TESTS_NUM: 12
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - {runner: garm-jammy, libc: musl}
-          - {runner: garm-jammy-amd, libc: gnu}
-          - {runner: garm-jammy-amd, libc: musl}
-    # format() because `${{ matrix.runner }}-16` is not valid in runs-on.
-    runs-on: ${{ format('{0}-16', matrix.runner) }}
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install Docker
-        run: |
-          set -eufo pipefail
-          sudo apt-get update
-          sudo apt-get -y install ca-certificates curl gnupg
-          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
-          sudo chmod a+r /usr/share/keyrings/docker-archive-keyring.gpg
-          echo "deb [arch=$(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
-          sudo apt-get update
-          sudo apt install -y docker-ce docker-ce-cli
-      - name: Prepare for VDPA
-        run: scripts/prepare_vdpa.sh
-      - name: Run unit tests
-        run: scripts/dev_cli.sh tests --unit --libc ${{ matrix.libc }}
-      - name: Load openvswitch module
-        run: sudo modprobe openvswitch
-      - name: Run integration tests
-        timeout-minutes: 60
-        run: scripts/dev_cli.sh tests --integration --libc ${{ matrix.libc }}
-  integration-arm64:
-    name: integration-arm64
-    needs: [preflight, dco, quality, build]
-    if: >-
-      github.event_name == 'merge_group' && needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    timeout-minutes: 120
-    env:
-      # Our runner has 80 cores (nproc).
-      # We limit parallelism only to avoid exhausting disk space and memory
-      # resources, not to save CPU resources.
-      PARALLEL_INTEGRATION_TESTS_NUM: 25
-    runs-on: bookworm-arm64
-    steps:
-      # arm64 runner user is "runner" (vfio's is "github-runner").
-      - name: Fix workspace permissions
-        run: sudo chown -R runner:runner ${GITHUB_WORKSPACE}
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Run unit tests (musl)
-        run: scripts/dev_cli.sh tests --unit --libc musl
-      - name: Load openvswitch module
-        run: sudo modprobe openvswitch
-      - name: Run integration tests (musl)
-        timeout-minutes: 60
-        run: scripts/dev_cli.sh tests --integration --libc musl
-      - name: Install Azure CLI
-        run: |
-          set -eufo pipefail
-          sudo apt install -y ca-certificates curl apt-transport-https lsb-release gnupg
-          curl -sL https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor | sudo tee /etc/apt/trusted.gpg.d/microsoft.gpg > /dev/null
-          echo "deb [arch=arm64] https://packages.microsoft.com/repos/azure-cli/ bookworm main" | sudo tee /etc/apt/sources.list.d/azure-cli.list
-          sudo apt update
-          sudo apt install -y azure-cli
-      - name: Download Windows image
-        shell: bash
-        run: |
-          set -eufo pipefail
-          IMG_BASENAME=windows-11-iot-enterprise-aarch64.raw
-          IMG_PATH=$HOME/workloads/$IMG_BASENAME
-          IMG_GZ_PATH=$HOME/workloads/$IMG_BASENAME.gz
-          IMG_GZ_BLOB_NAME=windows-11-iot-enterprise-aarch64-9-min.raw.gz
-          cp "scripts/$IMG_BASENAME.sha1" "$HOME/workloads/"
-          pushd "$HOME/workloads"
-          if sha1sum "$IMG_BASENAME.sha1" --check; then
-              exit
-          fi
-          popd
-          mkdir -p "$HOME/workloads"
-          az storage blob download --container-name private-images --file "$IMG_GZ_PATH" --name "$IMG_GZ_BLOB_NAME" --connection-string "${{ secrets.CH_PRIVATE_IMAGES }}"
-          gzip -d "$IMG_GZ_PATH"
-      - name: Run Windows guest integration tests
-        timeout-minutes: 30
-        run: scripts/dev_cli.sh tests --integration-windows --libc musl
-  integration-vfio:
-    name: integration-vfio
-    needs: [preflight, dco, quality, build]
-    if: >-
-      github.event_name == 'merge_group' && needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    runs-on: vfio-nvidia
-    env:
-      AUTH_DOWNLOAD_TOKEN: ${{ secrets.AUTH_DOWNLOAD_TOKEN }}
-    steps:
-      # vfio-nvidia runner user is "github-runner" (not "runner" like arm64).
-      - name: Fix workspace permissions
-        run: sudo chown -R github-runner:github-runner "${GITHUB_WORKSPACE}"
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Run VFIO integration tests
-        timeout-minutes: 25
-        run: scripts/dev_cli.sh tests --integration-vfio
-    # Most tests are failing with musl, see #6790
-    # - name: Run VFIO integration tests for musl
-    #   timeout-minutes: 25
-    #   run: scripts/dev_cli.sh tests --integration-vfio --libc musl
-  integration-windows:
-    name: integration-windows
-    needs: [preflight, dco, quality, build]
-    if: >-
-      github.event_name == 'merge_group' && needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    runs-on: garm-jammy-16
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Install Docker
-        run: |
-          set -eufo pipefail
-          sudo apt-get update
-          sudo apt-get -y install ca-certificates curl gnupg
-          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
-          sudo chmod a+r /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
-          sudo apt-get update
-          sudo apt install -y docker-ce docker-ce-cli
-      - name: Install Azure CLI
-        run: |
-          set -eufo pipefail
-          sudo apt install -y ca-certificates curl apt-transport-https lsb-release gnupg
-          curl -sL https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor | sudo tee /etc/apt/trusted.gpg.d/microsoft.gpg > /dev/null
-          echo "deb [arch=amd64] https://packages.microsoft.com/repos/azure-cli/ jammy main" | sudo tee /etc/apt/sources.list.d/azure-cli.list
-          sudo apt update
-          sudo apt install -y azure-cli
-      - name: Download Windows image
-        run: |
-          set -eufo pipefail
-          mkdir $HOME/workloads
-          az storage blob download --container-name private-images --file "$HOME/workloads/windows-server-2025-amd64-1.raw" --name windows-server-2025-amd64-1.raw --connection-string "${{ secrets.CH_PRIVATE_IMAGES }}"
-      - name: Run Windows guest integration tests
-        timeout-minutes: 15
-        run: scripts/dev_cli.sh tests --integration-windows
-      - name: Run Windows guest integration tests for musl
-        timeout-minutes: 15
-        run: scripts/dev_cli.sh tests --integration-windows --libc musl
-  integration-rate-limiter:
-    name: integration-rate-limiter
-    needs: [preflight, dco, quality, build]
-    if: >-
-      github.event_name == 'merge_group' && needs.preflight.outputs.full == 'true' && needs.dco.result == 'success' && needs.quality.result == 'success' && needs.build.result == 'success'
-    runs-on: bare-metal-9950x
-    env:
-      AUTH_DOWNLOAD_TOKEN: ${{ secrets.AUTH_DOWNLOAD_TOKEN }}
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Run rate-limiter integration tests
-        timeout-minutes: 20
-        run: scripts/dev_cli.sh tests --integration-rate-limiter
-  # The single required-status check. Branch protection requires this one job.
-  all-green:
-    name: all-green
-    needs:
-      - audit
-      - build
-      - dco
-      - formatting
-      - fuzz-build
-      - gitlint
-      - hadolint
-      - integration-arm64
-      # VFIO worker is failing #8160
-      # - integration-vfio
-      # See: #8211
-      # - integration-windows
-      - integration-x86-64-mq
-      - integration-x86-64-pr
-      - openapi
-      - package-consistency
-      - preflight
-      - quality
-      - reuse
-      - shlint
-      - taplo
-      - typos
-    if: always()
-    runs-on: ubuntu-latest
-    steps:
-      - name: Verify all dependencies succeeded or were skipped
-        env:
-          NEEDS_JSON: ${{ toJson(needs) }}
-        run: |
-          set -eufo pipefail
-          echo "$NEEDS_JSON" | jq .
-          # success or skipped = pass; failure or cancelled = red.
-          echo "$NEEDS_JSON" | jq -e '
-            to_entries
-            | map(select(.value.result != "success" and .value.result != "skipped"))
-            | length == 0
-          ' >/dev/null

.github/workflows/dco.yaml

@@ -0,0 +1,18 @@
+name: DCO
+on:
+  pull_request:
+jobs:
+  check:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - name: Set up Python 3.x
+      uses: actions/setup-python@v1
+      with:
+        python-version: '3.x'
+    - name: Check DCO
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      run: |
+        pip3 install -U dco-check
+        dco-check -e "49699333+dependabot[bot]@users.noreply.github.com"

.github/workflows/docker-image.yaml

@@ -1,13 +1,11 @@
 name: Cloud Hypervisor's Docker image update
+
 on:
   push:
     branches: main
     paths: resources/Dockerfile
   pull_request:
     paths: resources/Dockerfile
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-${{ github.event_name }}
-  cancel-in-progress: true
 
 env:
   REGISTRY: ghcr.io
@@ -18,16 +16,16 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@v2
 
       - name: Set up QEMU
-        uses: docker/setup-qemu-action@v4
+        uses: docker/setup-qemu-action@v1
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v4
+        uses: docker/setup-buildx-action@v1
 
       - name: Login to ghcr
-        uses: docker/login-action@v4
+        uses: docker/login-action@v2
         with:
           registry: ${{ env.REGISTRY }}
           username: ${{ github.actor }}
@@ -36,17 +34,17 @@ jobs:
 
       - name: Docker meta
         id: meta
-        uses: docker/metadata-action@v6
+        uses: docker/metadata-action@v4
         with:
           images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
           # generate Docker tags based on the following events/attributes
           tags: |
-            type=raw,value=20251114-0
+            type=raw,value={{date 'YYYYMMDD'}}-0
             type=sha
 
       - name: Build and push
         if: ${{ github.event_name == 'push' }}
-        uses: docker/build-push-action@v7
+        uses: docker/build-push-action@v2
         with:
           file: ./resources/Dockerfile
           platforms: linux/amd64,linux/arm64
@@ -55,7 +53,7 @@ jobs:
 
       - name: Build only
         if: ${{ github.event_name == 'pull_request' }}
-        uses: docker/build-push-action@v7
+        uses: docker/build-push-action@v2
         with:
           file: ./resources/Dockerfile
           platforms: linux/amd64,linux/arm64

.github/workflows/fuzz-build.yaml

@@ -0,0 +1,29 @@
+name: Cloud Hypervisor Cargo Fuzz Build
+on: [pull_request, create]
+
+jobs:
+  build:
+    if: github.event_name == 'pull_request'
+    name: Cargo Fuzz Build
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        rust:
+          - nightly
+        target:
+          - x86_64-unknown-linux-gnu
+    steps:
+      - name: Code checkout
+        uses: actions/checkout@v2
+      - name: Install Rust toolchain (${{ matrix.rust }})
+        uses: actions-rs/toolchain@v1
+        with:
+            toolchain: ${{ matrix.rust }}
+            target: ${{ matrix.target }}
+            override: true
+      - name: Install Cargo fuzz
+        # Temporary fix for cargo-fuzz on latest nightly: https://github.com/rust-fuzz/cargo-fuzz/issues/276
+        #run: cargo install cargo-fuzz
+        run: cargo install --git https://github.com/rust-fuzz/cargo-fuzz --rev b4df3e58f767b5cad8d1aa6753961003f56f3609
+      - name: Cargo Fuzz Build
+        run: cargo fuzz build

.github/workflows/integration-metrics.yaml

@@ -1,22 +0,0 @@
-name: Cloud Hypervisor Tests (Metrics)
-on:
-  push:
-    branches:
-      - main
-
-jobs:
-  build:
-    name: Tests (Metrics)
-    runs-on: bare-metal-9950x
-    env:
-      METRICS_PUBLISH_KEY: ${{ secrets.METRICS_PUBLISH_KEY }}
-    steps:
-      - name: Code checkout
-        uses: actions/checkout@v6
-        with:
-          fetch-depth: 0
-      - name: Run metrics tests
-        timeout-minutes: 60
-        run: scripts/dev_cli.sh tests --metrics -- --test-exclude micro_ -- --report-file /root/workloads/metrics.json
-      - name: Upload metrics report
-        run: 'curl -X PUT https://ch-metrics.azurewebsites.net/api/publishmetrics -H "x-functions-key: $METRICS_PUBLISH_KEY" -T ~/workloads/metrics.json'

.github/workflows/mshv-infra.yaml

@@ -1,246 +0,0 @@
-name: MSHV Infra Setup
-on:
-  workflow_call:
-    inputs:
-      ARCH:
-        description: 'Architecture for the VM'
-        required: true
-        type: string
-      KEY:
-        description: 'SSH Key Name'
-        required: true
-        type: string
-      OS_DISK_SIZE:
-        description: 'OS Disk Size in GB'
-        required: true
-        type: number
-      RG:
-        description: 'Resource Group Name'
-        required: true
-        type: string
-      VM_SKU:
-        description: 'VM SKU'
-        required: true
-        type: string
-    secrets:
-      MI_CLIENT_ID:
-        required: true
-      RUNNER_RG:
-        required: true
-      STORAGE_ACCOUNT_PATHS:
-        required: true
-      ARCH_SOURCE_PATH:
-        required: true
-      USERNAME:
-        required: true
-    outputs:
-      RG_NAME:
-        description: 'Resource group of the VM'
-        value: ${{ jobs.infra-setup.outputs.RG_NAME }}
-      VM_NAME:
-        description: 'Name of the VM'
-        value: ${{ jobs.infra-setup.outputs.VM_NAME }}
-      PRIVATE_IP:
-        description: 'Private IP of the VM'
-        value: ${{ jobs.infra-setup.outputs.PRIVATE_IP }}
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-${{ github.event_name }}
-  cancel-in-progress: true
-jobs:
-  infra-setup:
-    name: ${{ inputs.ARCH }} VM Provision
-    runs-on: mshv
-    outputs:
-      RG_NAME: ${{ steps.rg-setup.outputs.RG_NAME }}
-      VM_NAME: ${{ steps.vm-setup.outputs.VM_NAME }}
-      PRIVATE_IP: ${{ steps.get-vm-ip.outputs.PRIVATE_IP }}
-    steps:
-      - name: Install & login to AZ CLI
-        env:
-          MI_CLIENT_ID: ${{ secrets.MI_CLIENT_ID }}
-        run: |
-          set -eufo pipefail
-          echo "Installing Azure CLI if not already installed"
-          if ! command -v az &>/dev/null; then
-            curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash
-          else
-            echo "Azure CLI already installed"
-          fi
-          az --version
-          echo "Logging into Azure CLI using Managed Identity"
-          az login --identity --client-id "${MI_CLIENT_ID}"
-
-      - name: Get Location
-        id: get-location
-        env:
-          SKU: ${{ inputs.VM_SKU }}
-          STORAGE_ACCOUNT_PATHS: ${{ secrets.STORAGE_ACCOUNT_PATHS }}
-        run: |
-          set -eufo pipefail
-          # Extract vCPU count from SKU (e.g., "Standard_D2s_v3" => 2)
-          if ! [[ "$SKU" =~ ^Standard_[A-Za-z]+([1-9][0-9]*) ]]; then
-            printf 'Cannot extract vCPU count from SKU: %q\n' "$SKU"
-            exit 1
-          fi
-          vcpu=${BASH_REMATCH[1]}
-
-          SUPPORTED_LOCATIONS=$(echo "$STORAGE_ACCOUNT_PATHS" | jq -r 'to_entries[] | .key')
-
-          for location in $SUPPORTED_LOCATIONS; do
-            family=$(az vm list-skus --size "$SKU" --location "$location" --resource-type "virtualMachines" --query '[0].family' -o tsv)
-            if [[ -z "$family" ]]; then
-              echo "Cannot determine VM family for SKU: $SKU in $location"
-              continue
-            fi
-
-            remaining=$(az vm list-usage --location "$location" --query "[?name.value=='$family'] | [0]" -o json |
-                        jq '(.limit | tonumber) - (.currentValue | tonumber) >= ($ARGS.positional[0] | tonumber)' --jsonargs "$vcpu")
-            if [[ "$remaining" = true ]]; then
-              echo "Sufficient quota found in $location"
-              echo "location=$location" >> "$GITHUB_OUTPUT"
-              exit 0
-            fi
-          done
-
-          echo "No location found with sufficient vCPU quota for SKU: $SKU"
-          exit 1
-
-      - name: Create Resource Group
-        id: rg-setup
-        env:
-          LOCATION: ${{ steps.get-location.outputs.location }}
-          RG: ${{ inputs.RG }}
-          STORAGE_ACCOUNT_PATHS: ${{ secrets.STORAGE_ACCOUNT_PATHS }}
-        run: |
-          set -eufo pipefail
-          echo "Creating Resource Group: $RG"
-          # Create the resource group
-          echo "Creating resource group in location: ${LOCATION}"
-          az group create --name "${RG}" --location "${LOCATION}"
-          echo "RG_NAME=${RG}" >> $GITHUB_OUTPUT
-          echo "Resource group created successfully."
-
-      - name: Generate SSH Key
-        id: generate-ssh-key
-        env:
-          KEY: ${{ inputs.KEY }}
-        run: |
-          set -eufo pipefail
-          echo "Generating SSH key: $KEY"
-          mkdir -p ~/.ssh
-          ssh-keygen -t rsa -b 4096 -f ~/.ssh/"${KEY}" -N ""
-
-      - name: Create VM
-        id: vm-setup
-        env:
-          KEY: ${{ inputs.KEY }}
-          LOCATION: ${{ steps.get-location.outputs.location }}
-          OS_DISK_SIZE: ${{ inputs.OS_DISK_SIZE }}
-          RG: ${{ inputs.RG }}
-          RUNNER_RG: ${{ secrets.RUNNER_RG }}
-          USERNAME: ${{ secrets.USERNAME }}
-          VM_SKU: ${{ inputs.VM_SKU }}
-          VM_IMAGE_NAME: ${{ inputs.ARCH }}_${{ steps.get-location.outputs.location }}_image
-          VM_NAME: ${{ inputs.ARCH }}_${{ steps.get-location.outputs.location }}_${{ github.run_id }}
-        run: |
-          set -eufo pipefail
-          echo "Creating $VM_SKU VM: $VM_NAME"
-
-          # Extract subnet ID from the runner VM
-          echo "Retrieving subnet ID..."
-          SUBNET_ID=$(az network vnet list --resource-group "$RUNNER_RG" --query "[?contains(location, '${LOCATION}')].{SUBNETS:subnets}" | jq -r ".[0].SUBNETS[0].id")
-          if [[ -z "${SUBNET_ID}" ]]; then
-            echo "ERROR: Failed to retrieve Subnet ID."
-            exit 1
-          fi
-
-          # Extract image ID from the runner VM
-          echo "Retrieving image ID..."
-          IMAGE_ID=$(az image show --resource-group "$RUNNER_RG" --name "$VM_IMAGE_NAME" --query "id" -o tsv)
-          if [[ -z "${IMAGE_ID}" ]]; then
-            echo "ERROR: Failed to retrieve Image ID."
-            exit 1
-          fi
-
-          # Create VM
-          az vm create \
-            --resource-group "${RG}" \
-            --name "${VM_NAME}" \
-            --subnet "${SUBNET_ID}" \
-            --size "${VM_SKU}" \
-            --location "${LOCATION}" \
-            --image "${IMAGE_ID}" \
-            --os-disk-size-gb "${OS_DISK_SIZE}" \
-            --public-ip-sku Standard \
-            --storage-sku Premium_LRS \
-            --public-ip-address "" \
-            --admin-username "${USERNAME}" \
-            --ssh-key-value  ~/.ssh/"${KEY}".pub \
-            --security-type Standard \
-            --output json
-
-          az vm boot-diagnostics enable --name "${VM_NAME}" --resource-group "${RG}"
-
-          echo "VM_NAME=${VM_NAME}" >> "$GITHUB_OUTPUT"
-          echo "VM creation process completed successfully."
-
-      - name: Get VM Private IP
-        id: get-vm-ip
-        env:
-          RG: ${{ inputs.RG }}
-          VM_NAME: ${{ inputs.ARCH }}_${{ steps.get-location.outputs.location }}_${{ github.run_id }}
-        run: |
-          set -eufo pipefail
-          echo "Retrieving VM Private IP address..."
-          # Retrieve VM Private IP address
-          PRIVATE_IP=$(az vm show -g "${RG}" -n "${VM_NAME}" -d --query privateIps -o tsv)
-          if [[ -z "$PRIVATE_IP" ]]; then
-            echo "ERROR: Failed to retrieve private IP address."
-            exit 1
-          fi
-          echo "PRIVATE_IP=$PRIVATE_IP" >> "$GITHUB_OUTPUT"
-
-      - name: Wait for SSH availability
-        env:
-          KEY: ${{ inputs.KEY }}
-          PRIVATE_IP: ${{ steps.get-vm-ip.outputs.PRIVATE_IP }}
-          USERNAME: ${{ secrets.USERNAME }}
-        run: |
-          echo "Waiting for SSH to be accessible..."
-          timeout 120 bash -c 'until ssh -o StrictHostKeyChecking=no -i ~/.ssh/"${KEY}" -- "${USERNAME}@${PRIVATE_IP}" "exit" 2>/dev/null; do sleep 5; done'
-          echo "VM is accessible!"
-
-      - name: Remove Old Host Key
-        env:
-          PRIVATE_IP: ${{ steps.get-vm-ip.outputs.PRIVATE_IP }}
-        run: |
-          set -eufo pipefail
-          echo "Removing the old host key"
-          ssh-keygen -R "$PRIVATE_IP"
-
-      - name: SSH into VM and Install Dependencies
-        env:
-          KEY: ${{ inputs.KEY }}
-          PRIVATE_IP: ${{ steps.get-vm-ip.outputs.PRIVATE_IP }}
-          USERNAME: ${{ secrets.USERNAME }}
-        run: |
-          set -eufo pipefail
-          ssh -i ~/.ssh/"${KEY}" -o StrictHostKeyChecking=no -- "${USERNAME}@${PRIVATE_IP}" << EOF
-            set -eufo pipefail
-            echo "Logged in successfully."
-            echo "Installing dependencies..."
-            sudo tdnf install -y git moby-engine moby-cli clang llvm pkg-config make gcc glibc-devel
-            echo "Installing Rust..."
-            curl -sSf https://sh.rustup.rs | sh -s -- --default-toolchain stable --profile default -y
-            export PATH="\$HOME/.cargo/bin:\$PATH"
-            cargo --version
-            sudo mkdir -p /etc/docker/
-            echo '{"default-ulimits":{"nofile":{"Hard":65535,"Name":"nofile","Soft":65535}}}' | sudo tee /etc/docker/daemon.json
-            sudo systemctl stop docker
-            sudo systemctl enable docker.service
-            sudo systemctl enable containerd.service
-            sudo systemctl start docker
-            sudo groupadd -f docker
-            sudo usermod -a -G docker "${USERNAME}"
-            sudo systemctl restart docker
-          EOF

.github/workflows/mshv-integration.yaml

@@ -1,129 +0,0 @@
-name: Cloud Hypervisor Tests (MSHV) (x86_64)
-on: [pull_request_target, merge_group]
-permissions: {}
-
-jobs:
-  infra-setup:
-    name: MSHV Infra Setup (x86_64)
-    uses: ./.github/workflows/mshv-infra.yaml
-    with:
-      ARCH: x86_64
-      KEY: azure_key_${{ github.run_id }}
-      OS_DISK_SIZE: 512
-      RG: MSHV-INTEGRATION-${{ github.run_id }}
-      VM_SKU: Standard_D16s_v5
-    secrets:
-      MI_CLIENT_ID: ${{ secrets.MSHV_MI_CLIENT_ID }}
-      RUNNER_RG: ${{ secrets.MSHV_RUNNER_RG }}
-      STORAGE_ACCOUNT_PATHS: ${{ secrets.MSHV_STORAGE_ACCOUNT_PATHS }}
-      ARCH_SOURCE_PATH: ${{ secrets.MSHV_X86_SOURCE_PATH }}
-      USERNAME: ${{ secrets.MSHV_USERNAME }}
-
-  run-tests:
-    name: Integration Tests (x86_64)
-    needs: infra-setup
-    if: ${{ always() && needs.infra-setup.result == 'success' }}
-    runs-on: mshv
-    steps:
-      - name: Run integration tests
-        timeout-minutes: 60
-        env:
-          KEY: azure_key_${{ github.run_id }}
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-          REPO_URL: https://github.com/cloud-hypervisor/cloud-hypervisor.git
-          REPO_DIR: cloud-hypervisor
-          PRIVATE_IP: ${{ needs.infra-setup.outputs.PRIVATE_IP }}
-          RG: MSHV-${{ github.run_id }}
-          USERNAME: ${{ secrets.MSHV_USERNAME }}
-        run: |
-          set -eufo pipefail
-          echo "Connecting to the VM via SSH..."
-          ssh -i ~/.ssh/"${KEY}" -o StrictHostKeyChecking=no -- "${USERNAME}@${PRIVATE_IP}" << EOF
-            set -e
-            echo "Logged in successfully."
-            export PATH="\$HOME/.cargo/bin:\$PATH"
-
-            if [[ "${{ github.event_name }}" == "pull_request_target" ]]; then
-              git clone --depth 1 "$REPO_URL" "$REPO_DIR"
-              cd "$REPO_DIR"
-              git fetch origin pull/${{ github.event.pull_request.number }}/merge
-              git checkout FETCH_HEAD
-            else
-              git clone --depth 1 --single-branch --branch "${{ github.ref_name }}" "$REPO_URL" "$REPO_DIR"
-              cd "$REPO_DIR"
-            fi
-
-            echo "Loading VDPA kernel modules..."
-            sudo modprobe vdpa
-            sudo modprobe vhost_vdpa
-            sudo modprobe vdpa_sim
-            sudo modprobe vdpa_sim_blk
-            sudo modprobe vdpa_sim_net
-
-            echo "Creating VDPA devices..."
-            sudo vdpa dev add name vdpa-blk0 mgmtdev vdpasim_blk
-            sudo vdpa dev add name vdpa-blk1 mgmtdev vdpasim_blk
-            sudo vdpa dev add name vdpa-blk2 mgmtdev vdpasim_net
-
-            echo "Setting permissions..."
-            for i in 0 1 2; do
-              dev="/dev/vhost-vdpa-\$i"
-              if [ -e "\$dev" ]; then
-                sudo chown \$USER:\$USER "\$dev"
-                sudo chmod 660 "\$dev"
-              else
-                echo "Warning: Device \$dev not found"
-              fi
-            done
-
-            sudo ./scripts/dev_cli.sh tests --hypervisor mshv --integration
-          EOF
-
-      - name: Dump dmesg
-        if: always()
-        continue-on-error: true
-        env:
-          KEY: azure_key_${{ github.run_id }}
-          PRIVATE_IP: ${{ needs.infra-setup.outputs.PRIVATE_IP }}
-          USERNAME: ${{ secrets.MSHV_USERNAME }}
-        run: |
-          ssh -i ~/.ssh/"${KEY}" -o StrictHostKeyChecking=no -- "${USERNAME}@${PRIVATE_IP}" sudo dmesg
-
-      - name: Dump serial console logs
-        if: always()
-        continue-on-error: true
-        env:
-          RG_NAME: ${{ needs.infra-setup.outputs.RG_NAME }}
-          VM_NAME: ${{ needs.infra-setup.outputs.VM_NAME }}
-        run: |
-          set -eufo pipefail
-          az vm boot-diagnostics get-boot-log --name "${VM_NAME}" --resource-group "${RG_NAME}" | jq -r
-
-  cleanup:
-    name: Cleanup
-    needs: run-tests
-    if: always()
-    runs-on: mshv
-    steps:
-      - name: Delete RG
-        env:
-          RG: MSHV-INTEGRATION-${{ github.run_id }}
-        run: |
-          if az group exists --name "${RG}"; then
-            az group delete --name "${RG}" --yes --no-wait
-          else
-            echo "Resource Group ${RG} does not exist. Skipping deletion."
-          fi
-          echo "Cleanup process completed."
-
-      - name: Delete SSH Key
-        env:
-          KEY: azure_key_${{ github.run_id }}
-        run: |
-          if [ -f ~/.ssh/"${KEY}" ]; then
-            rm -f ~/.ssh/"${KEY}" ~/.ssh/"${KEY}.pub"
-            echo "SSH key deleted successfully."
-          else
-            echo "SSH key does not exist. Skipping deletion."
-          fi
-          echo "Cleanup process completed."

.github/workflows/quality.yaml

@@ -0,0 +1,113 @@
+name: Cloud Hypervisor Quality Checks
+on: [pull_request, create]
+
+jobs:
+  build:
+    if: github.event_name == 'pull_request'
+    name: Quality (clippy, rustfmt)
+    runs-on: ubuntu-latest
+    continue-on-error: ${{ matrix.experimental }}
+    strategy:
+      fail-fast: false
+      matrix:
+        rust:
+          - stable
+        target:
+          - aarch64-unknown-linux-gnu
+          - aarch64-unknown-linux-musl
+          - x86_64-unknown-linux-gnu
+          - x86_64-unknown-linux-musl
+
+        experimental: [false]
+        include:
+          - rust: beta
+            target: aarch64-unknown-linux-gnu
+            experimental: true
+          - rust: beta
+            target: aarch64-unknown-linux-musl
+            experimental: true
+          - rust: beta
+            target: x86_64-unknown-linux-gnu
+            experimental: true
+          - rust: beta
+            target: x86_64-unknown-linux-musl
+            experimental: true
+    steps:
+      - name: Code checkout
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      - name: Install Rust toolchain (${{ matrix.rust }})
+        uses: actions-rs/toolchain@v1
+        with:
+          toolchain: ${{ matrix.rust }}
+          target: ${{ matrix.target }}
+          override: true
+          components: rustfmt, clippy
+
+      - name: Debug Check (default features)
+        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
+        run: |
+          set -e
+          commits=$(git rev-list origin/${{ github.base_ref }}..${{ github.sha }})
+          for commit in $commits; do git checkout $commit; cargo check --tests --examples --all --target=${{ matrix.target }}; done
+          git checkout ${{ github.sha }}
+
+      - name: Formatting (rustfmt)
+        run: cargo fmt -- --check
+
+      - name: Clippy (kvm)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --no-default-features --tests --examples --features "kvm" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (default features)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --tests --examples -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (default features + guest_debug)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --tests --examples --features "guest_debug" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (default features + tracing)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --tests --examples --features "tracing" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (mshv)
+        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --no-default-features --tests --examples --features "mshv" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (mshv + kvm)
+        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --no-default-features --tests --examples --features "mshv,kvm" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Clippy (kvm + tdx)
+        if: ${{ matrix.target == 'x86_64-unknown-linux-gnu' }}
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: ${{ matrix.target != 'x86_64-unknown-linux-gnu' }}
+          command: clippy
+          args: --target=${{ matrix.target }} --locked --all --all-targets --no-default-features --tests --examples --features "tdx,kvm" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
+      - name: Check build did not modify any files
+        run: test -z "$(git status --porcelain)"

.github/workflows/release.yaml

@@ -1,69 +1,134 @@
 name: Cloud Hypervisor Release
-on: [create, merge_group]
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-${{ github.event_name }}
-  cancel-in-progress: true
-env:
-  GITHUB_TOKEN: ${{ github.token }}
+on: [pull_request, create]
 
 jobs:
   release:
-    if: (github.event_name == 'create' && github.event.ref_type == 'tag') || github.event_name == 'merge_group'
-    name: Release ${{ matrix.platform.target }}
-    strategy:
-      fail-fast: false
-      matrix:
-        platform:
-        - target: x86_64-unknown-linux-gnu
-          args: --all --release --features mshv
-          name_ch: cloud-hypervisor
-          name_ch_remote: ch-remote
-        - target: x86_64-unknown-linux-musl
-          args: --all --release --features mshv
-          name_ch: cloud-hypervisor-static
-          name_ch_remote: ch-remote-static
-        - target: aarch64-unknown-linux-musl
-          args:  --all --release
-          name_ch: cloud-hypervisor-static-aarch64
-          name_ch_remote: ch-remote-static-aarch64
+    if: (github.event_name == 'create' && github.event.ref_type == 'tag') || github.event_name == 'pull_request'
+    name: Release
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@v2
       - name: Install musl-gcc
-        if: contains(matrix.platform.target, 'musl')
         run: sudo apt install -y musl-tools
       - name: Create release directory
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
         run: rsync -rv --exclude=.git . ../cloud-hypervisor-${{ github.event.ref }}
-      - name: Build ${{ matrix.platform.target }}
-        uses: houseabsolute/actions-rust-cross@v1
+      - name: Install Rust toolchain (x86_64-unknown-linux-gnu)
+        uses: actions-rs/toolchain@v1
         with:
+          toolchain: "1.67.1"
+          target: x86_64-unknown-linux-gnu
+      - name: Install Rust toolchain (x86_64-unknown-linux-musl)
+        uses: actions-rs/toolchain@v1
+        with:
+          toolchain: "1.67.1"
+          target: x86_64-unknown-linux-musl
+      - name: Build
+        uses: actions-rs/cargo@v1
+        with:
+          toolchain: "1.67.1"
           command: build
-          target: ${{ matrix.platform.target }}
-          args: ${{ matrix.platform.args  }}
-          strip: true
-          toolchain: "1.89.0"
-      - name: Copy Release Binaries
-        if: github.event_name == 'create' && github.event.ref_type == 'tag'
-        shell: bash
-        run: |
-          cp target/${{ matrix.platform.target }}/release/cloud-hypervisor ./${{ matrix.platform.name_ch }}
-          cp target/${{ matrix.platform.target }}/release/ch-remote ./${{ matrix.platform.name_ch_remote }}
-      - name: Upload Release Artifacts
-        if: github.event_name == 'create' && github.event.ref_type == 'tag'
-        uses: actions/upload-artifact@v7
+          args: --all --release --no-default-features --features "kvm,mshv" --target=x86_64-unknown-linux-gnu
+      - name: Static Build
+        uses: actions-rs/cargo@v1
         with:
-          name: Artifacts for ${{ matrix.platform.target }}
-          path: |
-            ./${{ matrix.platform.name_ch }}
-            ./${{ matrix.platform.name_ch_remote }}
+          toolchain: "1.67.1"
+          command: build
+          args: --all --release --no-default-features --features "kvm,mshv" --target=x86_64-unknown-linux-musl
+      - name: Install Rust toolchain (aarch64-unknown-linux-musl)
+        uses: actions-rs/toolchain@v1
+        with:
+          toolchain: "1.67.1"
+          target: aarch64-unknown-linux-musl
+          override: true
+      - name: Create Release
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: create_release
+        uses: actions/create-release@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          tag_name: ${{ github.ref }}
+          release_name: ${{ github.ref }}
+          draft: true
+          prerelease: true
+      - name: Upload cloud-hypervisor
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-cloud-hypervisor
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/x86_64-unknown-linux-gnu/release/cloud-hypervisor
+          asset_name: cloud-hypervisor
+          asset_content_type: application/octet-stream
+      - name: Upload static cloud-hypervisor
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-static-cloud-hypervisor
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/x86_64-unknown-linux-musl/release/cloud-hypervisor
+          asset_name: cloud-hypervisor-static
+          asset_content_type: application/octet-stream
+      - name: Upload ch-remote
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-ch-remote
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/x86_64-unknown-linux-gnu/release/ch-remote
+          asset_name: ch-remote
+          asset_content_type: application/octet-stream
+      - name: Upload static-ch-remote
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-static-ch-remote
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/x86_64-unknown-linux-musl/release/ch-remote
+          asset_name: ch-remote-static
+          asset_content_type: application/octet-stream
+      - name: Clean build tree ahead of cross build
+        uses: actions-rs/cargo@v1
+        with:
+          command: clean
+      - name: Static Build (AArch64)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: true
+          command: build
+          args: --all --release --target=aarch64-unknown-linux-musl
+      - name: Upload static AArch64 cloud-hypervisor
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-static-aarch64-cloud-hypervisor
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/aarch64-unknown-linux-musl/release/cloud-hypervisor
+          asset_name: cloud-hypervisor-static-aarch64
+          asset_content_type: application/octet-stream
+      - name: Upload static AArch64 ch-remote
+        if: github.event_name == 'create' && github.event.ref_type == 'tag'
+        id: upload-release-static-aarch64-ch-remote
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: target/aarch64-unknown-linux-musl/release/ch-remote
+          asset_name: ch-remote-static-aarch64
+          asset_content_type: application/octet-stream
       - name: Vendor
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
         working-directory: ../cloud-hypervisor-${{ github.event.ref }}
         run: |
           mkdir ../vendor-cargo-home
@@ -71,25 +136,16 @@ jobs:
           mkdir .cargo
           cargo vendor > .cargo/config.toml
       - name: Create vendored source archive
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
-        run: tar cJf cloud-hypervisor-${{ github.event.ref }}.tar.xz ../cloud-hypervisor-${{ github.event.ref }}
+        working-directory: ../
+        run: tar cJf cloud-hypervisor-${{ github.event.ref }}.tar.xz cloud-hypervisor-${{ github.event.ref }}
       - name: Upload cloud-hypervisor vendored source archive
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
-        id: upload-release-cloud-hypervisor-vendored-sources
-        uses: actions/upload-artifact@v7
-        with:
-          path: cloud-hypervisor-${{ github.event.ref }}.tar.xz
-          name: cloud-hypervisor-${{ github.event.ref }}.tar.xz
-      - name: Create GitHub Release
         if: github.event_name == 'create' && github.event.ref_type == 'tag'
-        uses: softprops/action-gh-release@v3
+        id: upload-release-cloud-hypervisor-vendored-sources
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
-          draft: true
-          files: |
-            ./${{ matrix.platform.name_ch }}
-            ./${{ matrix.platform.name_ch_remote }}
-            ./cloud-hypervisor-${{ github.event.ref }}.tar.xz
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: ../cloud-hypervisor-${{ github.event.ref }}.tar.xz
+          asset_name: cloud-hypervisor-${{ github.event.ref }}.tar.xz
+          asset_content_type: application/x-xz

.gitignore

@@ -1,13 +1,8 @@
+/build
+/.cargo
+/target
 **/*.rs.bk
 **/Cargo.lock
 **/rusty-tags.vi
-/.agents
-/.cargo
-/.claude
-/.codex
-/.vscode
-/build
 /rpm/SOURCES
-/target
-/vendor
-__pycache__
+/.vscode

.gitlint

@@ -1,13 +0,0 @@
-[general]
-extra-path=scripts/gitlint/rules
-regex-style-search=true
-ignore=body-max-line-length,body-hard-tab
-
-[ignore-by-author-name]
-regex=dependabot
-ignore=all
-
-# default 72
-[title-max-length]
-line-length=72
-

.lychee.toml

@@ -1,33 +0,0 @@
-verbose = "info"
-
-exclude_path = [".lychee.toml"]
-
-exclude = [
-  # Availability of links below should be manually verified.
-  # Page for intel TDX support, returns 403 while querying.
-  '^https://www.intel.com/content/www/us/en/developer/tools/trust-domain-extensions/overview.html',
-  # Page for TPM, returns 403 while querying.
-  '^https://trustedcomputinggroup.org/wp-content/uploads/PC-Client-Specific-Platform-TPM-Profile-for-TPM-2p0-v1p05p_r14_pub.pdf',
-  # GitHub user smibarber referenced in `CREDITS.md` no longer exist
-  '^https://github.com/smibarber',
-  # OSDev has added bot protection and accesses my result in 403 Forbidden.
-  '^https://wiki.osdev.org',
-  # Exclude all pages with $ in the URL since $XXX is a variable
-  "\\$.*",
-  # Exclude local files
-  "file://.*",
-  # ARM documentation returns 403 Forbidden for automated CI checks.
-  '^http://infocenter\.arm\.com',
-  '^https://developer\.arm\.com',
-  # Ignore internal/unsupported protocols seen in logs
-  '^tcp://192\.168\.1\.10',
-  # Slack invite endpoints reject automated GETs and return 403.
-  '^https://join\.slack\.com/t/',
-]
-
-# Exclude loopback addresses
-exclude_loopback = true
-
-max_retries = 3
-
-retry_wait_time = 5

.reuse/dep5

@@ -1,12 +0,0 @@
-Format: https://www.debian.org/doc/packaging-manuals/copyright-format/1.0/
-Upstream-Name: cloud-hypervisor
-Upstream-Contact:  <>
-Source: https://www.cloudhypervisor.org
-
-Files: docs/*.md *.md
-Copyright: 2024
-License: CC-BY-4.0
-
-Files: scripts/* test_data/* *.toml .git* .editorconfig fuzz/Cargo.lock fuzz/.gitignore resources/linux-config-* vmm/src/api/openapi/cloud-hypervisor.yaml CODEOWNERS Cargo.lock
-Copyright: 2024
-License: Apache-2.0

.rustfmt.toml

@@ -1,4 +1 @@
-edition = "2024"
-group_imports="StdExternalCrate"
-imports_granularity="Module"
-
+edition = "2021"

.taplo.toml

@@ -1,6 +0,0 @@
-include = ["**/Cargo.toml"]
-
-[formatting]
-indent_string = "  " # 2 spaces: keep in sync with .editorconfig
-reorder_arrays = true
-reorder_keys = true

.typos.toml

@@ -1,28 +0,0 @@
-# Configuration for https://github.com/crate-ci/typos
-
-[files]
-extend-exclude = [
-  "hypervisor/src/kvm/x86_64/mod.rs",
-  "resources/linux-config-*",
-]
-
-[default.extend-words]
-CLASSE = "CLASSE"
-Dake = "Dake"
-EXTINT = "EXTINT"
-INOUT = "INOUT"
-MSIS = "MSIS" # MSIs (Message Signaled Interrupt)
-SME = "SME" # Secure Memory Encryption
-THR = "THR" # Transmitter Holding Register
-TRANSLATER = "TRANSLATER"
-ba = "ba"
-conectix = "conectix"
-liness = "liness"
-outout = "outout"
-
-[default.extend-identifiers]
-consts = "consts"
-fo = "fo"
-fpr = "fpr"
-# Public Linux API
-msg_controllen = "msg_controllen"

AGENTS.md

@@ -1,81 +0,0 @@
-## For Humans
-
-This is a compact [AGENTS.md](https://agents.md/) file for Cloud Hypervisor.
-It is meant to help automated coding agents make useful changes that stay safe,
-reviewable, and compatible with the project's normal engineering constraints.
-
-## For LLMs
-
-### Project Context
-
-- Start with `README.md` for the project shape and `CONTRIBUTING.md` for the
-  contribution rules, coding style, commit message guidance, and LLM assistance
-  disclosure policy. Following `CONTRIBUTING.md` is crucial!
-- The main supported architectures are `x86_64` and `aarch64`; the main
-  hypervisor backends are KVM and MSHV. `x86_64` with KVM gets the most regular
-  exercise, but changes must not make the other first-class targets worse.
-
-### Change Guidelines
-
-- Prefer correctness, safety, and readability over micro-optimizations. Keep
-  changes small, reviewable, and aligned with the existing crate/module
-  boundaries. Avoid speculative changes and unrelated refactoring.
-- For API, config, migration, device model, or hypervisor boundary changes,
-  consider the effect on all architectures and all backends. Changes to one
-  backend can be okay if the other backend still functions properly and could
-  be extended or modified later.
-- Follow Rust best practices and the style already present in the touched code.
-- Avoid new dependencies unless the benefit is clear and local alternatives are
-  not enough.
-- Preserve existing behavior unless the requested change explicitly needs a
-  behavior change; refactors must preserve behavior. Call out compatibility or
-  migration implications.
-- Do not invent APIs, behavior, or requirements. If something is uncertain,
-  state the uncertainty and proceed only with minimal, explicit assumptions.
-
-### Safety and Domain Notes
-
-- Prefer safe Rust. If `unsafe` is necessary, keep it narrow, add a `SAFETY:`
-  comment with the invariants, and make sure the surrounding code upholds them.
-- Assume concurrency matters. Avoid races, unsynchronized shared state, and
-  implicit ordering assumptions; prefer clear ownership and synchronization.
-- Keep docs and comments short and useful. Document non-trivial invariants at
-  struct definitions and critical state transitions.
-- Logging should be minimal and high signal. Use `info!` for important normal
-  state changes that matter in production; use `warn!` or `error!` only for
-  abnormal conditions. Keep `debug!` for focused diagnostics.
-
-### Build and Test Notes
-
-- Some workspace members require the `kvm` feature to build or test correctly.
-  When a default build failure looks feature-related, retry the narrow command
-  with `--features kvm` before widening the diagnosis.
-- Prefer narrow crate/test commands while iterating, then broaden verification
-  when the touched surface justifies it.
-- Formatting currently needs nightly-only rustfmt features; use
-  `cargo +nightly fmt --all`.
-- Add targeted unit tests for bug fixes and non-trivial logic where practical.
-  Keep test scaffolding minimal and focused.
-- Integration tests live in `./cloud-hypervisor/tests/` and are normally driven
-  by `./scripts/dev_cli.sh` / `./scripts/run_integration_tests_*.sh`. They need
-  host privileges, workloads, and container setup. To build the integration-test
-  code directly without the infrastructure from `./scripts`, set the Rust cfg
-  `devcli_testenv` or simply build through `clippy` which automatically includes
-  these code paths; otherwise the integration-test code is not included. Do not
-  assume the tests can be run directly in a restricted agent environment; ask
-  the developer to run them when real integration coverage is needed.
-
-### Commit and Patch Formatting
-
-- Follow the rules in `CONTRIBUTING.md`, including reviewable commit structure,
-  valid component prefixes, 72-column commit messages, and a `Signed-off-by`
-  trailer.
-- Lines in a commit message that are allowed to exceed the 72-column limit are
-  specified in `./scripts/gitlint/rules`.
-- For LLM-assisted changes, follow the disclosure guidance in `CONTRIBUTING.md`:
-  use the project's `Assisted-by:` trailer when disclosure is needed, and do not
-  add `Co-authored-by` or similar trailers unless that policy changes.
-- Temporary allowances such as `#[allow(unused)]` or ignored tests are only
-  acceptable if resolved within the same commit series or paired with a clear
-  TODO referencing a ticket. Ask the developer if in doubt.
-

CONTRIBUTING.md

@@ -5,61 +5,20 @@ License](https://opensource.org/licenses/Apache-2.0) and the [BSD 3
 Clause](https://opensource.org/licenses/BSD-3-Clause) license. Individual files
 contain details of their licensing and changes to that file are under the same
 license unless the contribution changes the license of the file. When importing
-code from a third party project (e.g. Firecracker or crosvm) please respect the
+code from a third party project (e.g. Firecracker or CrosVM) please respect the
 license of those projects.
 
 New code should be under the [Apache v2
 License](https://opensource.org/licenses/Apache-2.0).
 
-## Coding Style & Code Comments
+## Coding Style
 
-We use the [Rust Style] guide and enforce formatting and linting in CI,
-including `rustfmt`, `clippy`, and other common Rust quality checks, for every
-pull request. We adapt to best practices, new lints and new tooling as the
-ecosystem evolves.
-
-Code should **speak for itself** (for example, by using descriptive identifiers)
-and be **easy to read and maintain**. Beyond the conventions and tooling
-described above, contributors have _some_ room to apply their own style and
-preferred structure. Maintainers may still suggest refactorings where they
-believe readability, consistency, or maintainability can be improved.
-
-For new code, add documentation and comments where they **provide additional value**:
-
-* **Rustdoc** explains the API to its users.
-* **Inline comments** explain the code the reader, especially *why* it is
-  written that way.
-* **Commit messages** explain the broader context of a change (for more
-  information on commit messages, see below).
-
-Comments should be concise and add additional context or information to the code.
-
-[Rust Style]: https://github.com/rust-lang/rust/tree/HEAD/src/doc/style-guide/src
+We follow the [Rust Style](https://github.com/rust-dev-tools/fmt-rfcs/blob/master/guide/guide.md)
+convention and enforce it through the Continuous Integration (CI) process calling into `rustfmt`
+for each submitted Pull Request (PR).
 
 ## Basic Checks
 
-```sh
-# We currently rely on nightly-only formatting features
-cargo +nightly fmt --all
-cargo check --all-targets --tests
-cargo clippy --all-targets --tests
-# Please note that this will not execute integration tests.
-cargo test --all-targets --tests
-
-# To lint your last three commits
-gitlint --commits "HEAD~3..HEAD"
-```
-
-### \[Optional\] Run Integration Tests
-
-_Caution: These tests are taking a long time to complete (40+ mins) and need special setup._
-
-```sh
- bash ./scripts/dev_cli.sh tests --integration -- --test-filter '<optionally filter test by name pattern>'
-```
-
-### Setup Commit Hook
-
 Please consider creating the following hook as `.git/hooks/pre-commit` in order
 to ensure basic correctness of your code. You can extend this further if you
 have specific features that you regularly develop against.
@@ -67,9 +26,9 @@ have specific features that you regularly develop against.
 ```sh
 #!/bin/sh
 
-cargo +nightly fmt --all -- --check || exit 1
-cargo check --locked --all-targets --tests || exit 1
-cargo clippy --locked --all-targets --tests -- -D warnings || exit 1
+cargo fmt -- --check || exit 1
+cargo check --locked --all --all-targets --tests || exit 1
+cargo clippy --locked --all --all-targets --tests -- -D warnings || exit 1
 ```
 
 You will need to `chmod +x .git/hooks/pre-commit` to have it run on every
@@ -77,65 +36,42 @@ commit you make.
 
 ## Certificate of Origin
 
-In order to get a clear contribution chain of trust we use the [signed-off-by language](https://www.kernel.org/doc/Documentation/process/submitting-patches.rst)
+In order to get a clear contribution chain of trust we use the [signed-off-by language](https://01.org/community/signed-process)
 used by the Linux kernel project.
 
-## Patch format & Git Commit Hygiene
+## Patch format
 
-_We use **Patch** as synonym for **Commit**._
+Beside the signed-off-by footer, we expect each patch to comply with the following format:
 
-We require patches to:
+```
+<component>: Change summary
 
-- Have a `Signed-off-by: Name <email>` footer
-- Follow the pattern: \
-  ```
-   <component>: Change summary
+More detailed explanation of your changes: Why and how.
+Wrap it to 72 characters.
+See http://chris.beams.io/posts/git-commit/
+for some more good pieces of advice.
 
-   More detailed explanation of your changes: Why and how.
-   Wrap it to 72 characters.
-   See http://chris.beams.io/posts/git-commit/
-   for some more good pieces of advice.
+Signed-off-by: <contributor@foo.com>
+```
 
-   Signed-off-by: <contributor@foo.com>
-   ```
-
-
-Valid components are listed in `TitleStartsWithComponent.py`. In short, each
-cargo workspace member is a valid component as well as `build`, `ci`, `docs` and
-`misc`.
-
-Example patch:
+For example:
 
 ```
 vm-virtio: Reset underlying device on driver request
-
+    
 If the driver triggers a reset by writing zero into the status register
 then reset the underlying device if supported. A device reset also
 requires resetting various aspects of the queue.
-
+    
 In order to be able to do a subsequent reactivate it is required to
 reclaim certain resources (interrupt and queue EventFDs.) If a device
 reset is requested by the driver but the underlying device does not
 support it then generate an error as the driver would not be able to
 configure it anyway.
-
+    
 Signed-off-by: Rob Bradford <robert.bradford@intel.com>
 ```
 
-### Git Commit History
-
-We value a clean, **reviewable** commit history. Each commit should represent
-a self-contained, logical step that guides reviewers clearly from A to B.
-
-Avoid patterns like `init A -> init B -> fix A` or \
-`init design A -> revert A -> use design B`. Commits must be independently
-reviewable - don't leave "fix previous commit" or earlier design attempts in
-the history.
-
-Intermediate work-in-progress changes are acceptable only if a subsequent
-commit in the same series cleans them up (e.g. a temporary `#[allow(unused)]`
-removed in the next commit).
-
 ## Pull requests
 
 Cloud Hypervisor uses the “fork-and-pull” development model. Follow these steps if
@@ -143,16 +79,12 @@ you want to merge your changes to `cloud-hypervisor`:
 
 1. Fork the [cloud-hypervisor](https://github.com/cloud-hypervisor/cloud-hypervisor) project
    into your github organization.
-1. Within your fork, create a branch for your contribution.
-1. [Create a pull request](https://help.github.com/articles/creating-a-pull-request-from-a-fork/)
+2. Within your fork, create a branch for your contribution.
+3. [Create a pull request](https://help.github.com/articles/creating-a-pull-request-from-a-fork/)
    against the main branch of the Cloud Hypervisor repository.
-1. Each commit must comply with the Commit Hygiene guidelines above.
-1. A pull request should address a single component or concern to keep review
-   focused and approvals straightforward.
-1. Once the pull request is approved it can be integrated.
-
-Please squash any changes done during review already into the corresponding
-commits instead of pushing `<component>: addressing review for A`-style commits.
+4. To update your pull request amend existing commits whenever applicable and
+   then push the new changes to your pull request branch.
+5. Once the pull request is approved it can be integrated.
 
 ## Issue tracking
 
@@ -169,83 +101,16 @@ comments or by adding the `Fixes` keyword to your commit message:
 
 ```
 serial: Set terminal in raw mode
-
+    
 In order to have proper output from the serial, we need to setup the
 terminal in raw mode. When the VM is shutting down, it is also the
 VMM responsibility to set the terminal back into canonical mode if we
 don't want to get any weird behavior from the shell.
-
+    
 Fixes #88
-
+	
 Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
 ```
 
-Then, after the corresponding PR is merged, GitHub will automatically close that issue when parsing the
+Then, after the corresponding PR is merged, Github will automatically close that issue when parsing the
 [commit message](https://help.github.com/articles/closing-issues-via-commit-messages/).
-
-## AI/LLM Assistance & Generated Code
-
-We recommend **a careful and conservative approach** to LLM usage, guided by
-sound engineering judgment. Please use AI/LLM-assisted tooling thoughtfully and
-responsibly to ensure efficient use of limited project resources, particularly
-in code review and long-term maintenance. Our primary goals are to avoid
-ambiguity in license compliance and to keep contributions clear and easy to
-review.
-
-Or in other words: please apply common sense and don't blindly accept LLM
-suggestions.
-
-This policy can be revisited as LLMs evolve and mature.
-
-### Code Review
-
-We generally recommend doing early coarse-grained reviews using state-of-the-art
-LLMs. This can help identify rough edges, copy & paste errors, and typos early
-on. This reduces review cycles for human reviewers.
-
-Please **do not** use GitHub Copilot directly in PRs to keep discussions clean.
-Instead, ask an LLM of your choice for a review. A convenient way to do this is
-
-- appending `.patch` to the GitHub PR URL
-  (e.g., `https://github.com/cloud-hypervisor/cloud-hypervisor/pull/1234.patch`)
-  and pasting it into the LLM of your choice, or
-- using a local agent in your terminal, such as `codex` or `claude`.
-
-### Contributions assisted by LLMs
-
-All contributions **must** be submitted by a human contributor. Automated or
-bot-driven PRs are not accepted.
-
-You are responsible for every piece of code you submit, and you must understand
-both the design and the implementation details. LLMs are useful for prototyping
-and generating boilerplate code. However, large or complex logic must be
-authored and fully understood by the contributor - LLM output should not be
-submitted without careful review and comprehension.
-
-Please disclose LLM use in your commit message and PR description if it
-meaningfully contributed to the submitted code. Again, we recommend careful and
-conservative use of LLMs, guided by common sense.
-
-Use the following tag to disclose LLM assistance in your commit message:
-
-```
-Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]
-```
-
-Where:
-
-- ``AGENT_NAME`` is the name of the AI tool or framework
-- ``MODEL_VERSION`` is the specific model version used
-- ``[TOOL1] [TOOL2]`` are optional specialized analysis tools used
-
-Basic development tools (git, make, editors) should not be listed.
-
-Example:
-
-```
-Assisted-by: Claude:Opus-4.6 CodeQL
-```
-
-Maintainers reserve the right to request additional clarification or decline
-contributions where LLM usage raises concerns. Ultimately, acceptance of any
-contribution is at the maintainers' discretion.

Cargo.toml

@@ -1,131 +1,102 @@
-# Cloud Hypervisor Workspace
-#
-# The main crate producing the binaries is in `./cloud-hypervisor`.
+[package]
+name = "cloud-hypervisor"
+version = "31.2.0"
+authors = ["The Cloud Hypervisor Authors"]
+edition = "2021"
+default-run = "cloud-hypervisor"
+build = "build.rs"
+license = "LICENSE-APACHE & LICENSE-BSD-3-Clause"
+description = "Open source Virtual Machine Monitor (VMM) that runs on top of KVM"
+homepage = "https://github.com/cloud-hypervisor/cloud-hypervisor"
+# Minimum buildable version:
+# Keep in sync with version in .github/workflows/build.yaml
+# Policy on MSRV (see #4318):
+# Can only be bumped by:
+# a.) A dependency requires it,
+# b.) If we want to use a new feature and that MSRV is at least 6 months old,
+# c.) There is a security issue that is addressed by the toolchain update.
+rust-version = "1.62"
 
 [profile.release]
-codegen-units = 1
 lto = true
+codegen-units = 1
 opt-level = "s"
 strip = true
 
 [profile.profiling]
-debug = true
 inherits = "release"
 strip = false
+debug = true
+
+[dependencies]
+anyhow = "1.0.69"
+api_client = { path = "api_client" }
+argh = "0.1.9"
+dhat = { version = "0.3.2", optional = true }
+epoll = "4.3.1"
+event_monitor = { path = "event_monitor" }
+hypervisor = { path = "hypervisor" }
+libc = "0.2.139"
+log = { version = "0.4.17", features = ["std"] }
+option_parser = { path = "option_parser" }
+seccompiler = "0.3.0"
+serde_json = "1.0.95"
+signal-hook = "0.3.15"
+thiserror = "1.0.39"
+tpm = { path = "tpm"}
+tracer = { path = "tracer" }
+vmm = { path = "vmm" }
+vmm-sys-util = "0.11.0"
+vm-memory = "0.10.0"
+
+# List of patched crates
+[patch.crates-io]
+kvm-bindings = { git = "https://github.com/cloud-hypervisor/kvm-bindings", branch = "ch-v0.6.0-tdx" }
+kvm-ioctls = { git = "https://github.com/rust-vmm/kvm-ioctls", branch = "main" }
+versionize_derive = { git = "https://github.com/cloud-hypervisor/versionize_derive", branch = "ch" }
+
+[dev-dependencies]
+dirs = "4.0.0"
+net_util = { path = "net_util" }
+once_cell = "1.17.1"
+serde_json = "1.0.95"
+test_infra = { path = "test_infra" }
+wait-timeout = "0.2.0"
+
+[features]
+default = ["kvm"]
+dhat-heap = ["dhat"] # For heap profiling
+guest_debug = ["vmm/guest_debug"]
+kvm = ["vmm/kvm"]
+mshv = ["vmm/mshv"]
+tdx = ["vmm/tdx"]
+tracing = ["vmm/tracing", "tracer/tracing"]
 
 [workspace]
 members = [
-  "api_client",
-  "arch",
-  "block",
-  "cloud-hypervisor",
-  "devices",
-  "event_monitor",
-  "hypervisor",
-  "net_util",
-  "option_parser",
-  "pci",
-  "performance-metrics",
-  "rate_limiter",
-  "serial_buffer",
-  "test_infra",
-  "tracer",
-  "vhost_user_block",
-  "vhost_user_net",
-  "virtio-devices",
-  "vm-allocator",
-  "vm-device",
-  "vm-migration",
-  "vm-virtio",
-  "vmm",
+    "api_client",
+    "arch",
+    "block_util",
+    "devices",
+    "event_monitor",
+    "hypervisor",
+    "net_gen",
+    "net_util",
+    "option_parser",
+    "pci",
+    "performance-metrics",
+    "qcow",
+    "rate_limiter",
+    "serial_buffer",
+    "test_infra",
+    "tracer",
+    "vhdx",
+    "vhost_user_block",
+    "vhost_user_net",
+    "virtio-devices",
+    "vmm",
+    "vm-allocator",
+    "vm-device",
+    "vm-migration",
+    "vm-virtio"
 ]
-package.edition = "2024"
-# Minimum buildable version:
-# Keep in sync with version in .github/workflows/build.yaml
-# Policy on MSRV (see #4318):
-# Can only be bumped if satisfying any of the following:
-# a.) A dependency requires it,
-# b.) If we want to use a new feature and that MSRV is at least 6 months old,
-# c.) There is a security issue that is addressed by the toolchain update.
-package.rust-version = "1.89.0"
-resolver = "3"
-
-[workspace.dependencies]
-# rust-vmm crates
-acpi_tables = "0.2.0"
-iommufd-ioctls = "0.1.0"
-kvm-bindings = "0.14.0"
-kvm-ioctls = "0.24.0"
-linux-loader = "0.13.2"
-mshv-bindings = "0.6.9"
-mshv-ioctls = "0.6.9"
-seccompiler = "0.5.0"
-vfio-bindings = { version = "0.6.2", default-features = false }
-vfio-ioctls = { version = "0.6.0", default-features = false }
-vfio_user = { version = "0.1.3", default-features = false }
-vhost = { version = "0.16.0", default-features = false }
-vhost-user-backend = { version = "0.22.0", default-features = false }
-virtio-bindings = "0.2.6"
-virtio-queue = "0.17.0"
-vm-fdt = "0.3.0"
-vm-memory = "0.17.1"
-vmm-sys-util = "0.15.0"
-
-# igvm crates
-igvm = "0.4.0"
-igvm_defs = "0.4.0"
-
-# serde crates
-serde = "1.0.228"
-serde_json = "1.0.149"
-serde_with = { version = "3.18.0", default-features = false }
-
-# other crates
-anyhow = "1.0.102"
-bitflags = "2.11.1"
-byteorder = "1.5.0"
-cfg-if = "1.0.4"
-clap = "4.6.1"
-dhat = "0.3.3"
-dirs = "6.0.0"
-env_logger = "0.11.10"
-epoll = "4.4.0"
-flume = "0.12.0"
-itertools = "0.14.0"
-jiff = { version = "0.2", default-features = false, features = ["std"] }
-libc = "0.2.186"
-log = "0.4.29"
-sha2 = "0.11.0"
-signal-hook = "0.4.4"
-thiserror = "2.0.18"
-uuid = { version = "1.23.1" }
-wait-timeout = "0.2.1"
-zerocopy = { version = "0.8.48", default-features = false }
-
-[workspace.lints.clippy]
-# Any clippy lint (group) in alphabetical order:
-# https://rust-lang.github.io/rust-clippy/master/index.html
-
-# Groups
-all = "deny"         # shorthand for the other groups but here for compleness
-complexity = "deny"
-correctness = "deny"
-perf = "deny"
-style = "deny"
-suspicious = "deny"
-
-# Individual Lints
-assertions_on_result_states = "deny"
-if_not_else = "deny"
-manual_string_new = "deny"
-map_unwrap_or = "deny"
-needless_pass_by_value = "deny"
-redundant_else = "deny"
-semicolon_if_nothing_returned = "deny"
-undocumented_unsafe_blocks = "deny"
-uninlined_format_args = "deny"
-unnecessary_semicolon = "deny"
-
-[workspace.lints.rust]
-# `level = warn` is irrelevant here but mandatory for rustc/cargo
-unexpected_cfgs = { level = "warn", check-cfg = ['cfg(devcli_testenv)'] }

Jenkinsfile

@@ -0,0 +1,452 @@
+def runWorkers = true
+pipeline {
+    agent none
+    options {
+        timeout(time: 4, unit: 'HOURS')
+    }
+    stages {
+        stage('Early checks') {
+            agent { node { label 'built-in' } }
+            stages {
+                stage('Checkout') {
+                    steps {
+                        checkout scm
+                    }
+                }
+                stage('Check if worker build can be skipped') {
+                    when {
+                        expression {
+                            return skipWorkerBuild()
+                        }
+                    }
+                    steps {
+                        script {
+                            runWorkers = false
+                            echo 'No changes requring a build'
+                        }
+                    }
+                }
+                stage('Check for RFC/WIP builds') {
+                    when {
+                        changeRequest comparator: 'REGEXP', title: '.*(rfc|RFC|wip|WIP).*'
+                        beforeAgent true
+                    }
+                    steps {
+                        error('Failing as this is marked as a WIP or RFC PR.')
+                    }
+                }
+                stage('Cancel older builds') {
+                    when { not { branch 'main' } }
+                    steps {
+                        cancelPreviousBuilds()
+                    }
+                }
+            }
+        }
+        stage('Build') {
+            parallel {
+                stage('Worker build') {
+                    agent { node { label 'jammy' } }
+                    when {
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Prepare environment') {
+                            steps {
+                                sh 'scripts/prepare_vdpa.sh'
+                            }
+                        }
+                        stage('Run OpenAPI tests') {
+                            steps {
+                                sh 'scripts/run_openapi_tests.sh'
+                            }
+                        }
+                        stage('Run unit tests') {
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --unit'
+                            }
+                        }
+                        stage('Run integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration'
+                            }
+                        }
+                        stage('Run live-migration integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration-live-migration'
+                            }
+                        }
+                        stage('Run unit tests for musl') {
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --unit --libc musl'
+                            }
+                        }
+                        stage('Run integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration --libc musl'
+                            }
+                        }
+                        stage('Run live-migration integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration-live-migration --libc musl'
+                            }
+                        }
+                    }
+                }
+                stage('AArch64 worker build') {
+                    agent { node { label 'bionic-arm64' } }
+                    when {
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    environment {
+                            AZURE_CONNECTION_STRING = credentials('46b4e7d6-315f-4cc1-8333-b58780863b9b')
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Run unit tests') {
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --unit --libc musl'
+                            }
+                        }
+                        stage('Run integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration --libc musl'
+                            }
+                        }
+                        stage('Install azure-cli') {
+                            steps {
+                                installAzureCli('bionic', 'arm64')
+                            }
+                        }
+                        stage('Download Windows image') {
+                            steps {
+                                sh '''#!/bin/bash -x
+                                    IMG_BASENAME=windows-11-iot-enterprise-aarch64.raw
+                                    IMG_PATH=$HOME/workloads/$IMG_BASENAME
+                                    IMG_GZ_PATH=$HOME/workloads/$IMG_BASENAME.gz
+                                    IMG_GZ_BLOB_NAME=windows-11-iot-enterprise-aarch64-9-min.raw.gz
+                                    cp "scripts/$IMG_BASENAME.sha1" "$HOME/workloads/"
+                                    pushd "$HOME/workloads"
+                                    if sha1sum "$IMG_BASENAME.sha1" --check; then
+                                        exit
+                                    fi
+                                    popd
+                                    mkdir -p "$HOME/workloads"
+                                    az storage blob download \
+                                        --container-name private-images \
+                                        --file "$IMG_GZ_PATH" \
+                                        --name "$IMG_GZ_BLOB_NAME" \
+                                        --connection-string "$AZURE_CONNECTION_STRING"
+                                    gzip -d $IMG_GZ_PATH
+                                '''
+                            }
+                        }
+                        stage('Run Windows guest integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-windows --libc musl'
+                            }
+                        }
+                    }
+                    post {
+                        always {
+                            sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+                            deleteDir()
+                        }
+                    }
+                }
+                stage('Worker build - Windows guest') {
+                    agent { node { label 'jammy' } }
+                    when {
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    environment {
+                            AZURE_CONNECTION_STRING = credentials('46b4e7d6-315f-4cc1-8333-b58780863b9b')
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Install azure-cli') {
+                            steps {
+                                installAzureCli('jammy', 'amd64')
+                            }
+                        }
+                        stage('Download assets') {
+                            steps {
+                                sh "mkdir ${env.HOME}/workloads"
+                                sh 'az storage blob download --container-name private-images --file "$HOME/workloads/windows-server-2022-amd64-2.raw" --name windows-server-2022-amd64-2.raw --connection-string "$AZURE_CONNECTION_STRING"'
+                            }
+                        }
+                        stage('Run Windows guest integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-windows'
+                            }
+                        }
+                        stage('Run Windows guest integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-windows --libc musl'
+                            }
+                        }
+                    }
+                }
+                stage('Worker build - Metrics') {
+                    agent { node { label 'jammy-metrics' } }
+                    when {
+                        branch 'main'
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    environment {
+                        METRICS_PUBLISH_KEY = credentials('52e0945f-ce7a-43d1-87af-67d1d87cc40f')
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Run metrics tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --metrics -- -- --report-file /root/workloads/metrics.json'
+                            }
+                        }
+                        stage('Upload metrics report') {
+                            steps {
+                                sh 'curl -X PUT https://cloud-hypervisor-metrics.azurewebsites.net/api/publishmetrics -H "x-functions-key: $METRICS_PUBLISH_KEY" -T ~/workloads/metrics.json'
+                            }
+                        }
+                    }
+                }
+                stage('Worker build - Rate Limiter') {
+                    agent { node { label 'focal-metrics' } }
+                    when {
+                        branch 'main'
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Run rate-limiter integration tests') {
+                            options {
+                                timeout(time: 10, unit: 'MINUTES')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-rate-limiter'
+                            }
+                        }
+                    }
+                }
+                stage('Worker build - SGX') {
+                    agent { node { label 'jammy-sgx' } }
+                    when {
+                        beforeAgent true
+                        allOf {
+                            branch 'main'
+                            expression {
+                                return runWorkers
+                            }
+                        }
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Run SGX integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-sgx'
+                            }
+                        }
+                        stage('Run SGX integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-sgx --libc musl'
+                            }
+                        }
+                    }
+                    post {
+                        always {
+                            sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+                            deleteDir()
+                        }
+                    }
+                }
+                stage('Worker build - VFIO') {
+                    agent { node { label 'jammy-vfio' } }
+                    when {
+                        beforeAgent true
+                        allOf {
+                            branch 'main'
+                            expression {
+                                return runWorkers
+                            }
+                        }
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Run VFIO integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-vfio'
+                            }
+                        }
+                        stage('Run VFIO integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'scripts/dev_cli.sh tests --integration-vfio --libc musl'
+                            }
+                        }
+                    }
+                    post {
+                        always {
+                            sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+                            deleteDir()
+                        }
+                    }
+                }
+            }
+        }
+    }
+    post {
+        regression {
+            script {
+                if (env.BRANCH_NAME == 'main') {
+                    slackSend(color: '#ff0000', message: '"main" branch build is now failing', channel: '#jenkins-ci')
+                }
+            }
+        }
+        fixed {
+            script {
+                if (env.BRANCH_NAME == 'main') {
+                    slackSend(color: '#00ff00', message: '"main" branch build is now fixed', channel: '#jenkins-ci')
+                }
+            }
+        }
+    }
+}
+
+def cancelPreviousBuilds() {
+    // Check for other instances of this particular build, cancel any that are older than the current one
+    def jobName = env.JOB_NAME
+    def currentBuildNumber = env.BUILD_NUMBER.toInteger()
+    def currentJob = Jenkins.instance.getItemByFullName(jobName)
+
+    // Loop through all instances of this particular job/branch
+    for (def build : currentJob.builds) {
+        if (build.isBuilding() && (build.number.toInteger() < currentBuildNumber)) {
+            echo "Older build still queued. Sending kill signal to build number: ${build.number}"
+            build.doStop()
+        }
+    }
+}
+
+def installAzureCli(distro, arch) {
+    sh 'sudo apt install -y ca-certificates curl apt-transport-https lsb-release gnupg'
+    sh 'curl -sL https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor | sudo tee /etc/apt/trusted.gpg.d/microsoft.gpg > /dev/null'
+    sh "echo \"deb [arch=${arch}] https://packages.microsoft.com/repos/azure-cli/ ${distro} main\" | sudo tee /etc/apt/sources.list.d/azure-cli.list"
+    sh 'sudo apt update'
+    sh 'sudo apt install -y azure-cli'
+}
+
+def boolean skipWorkerBuild() {
+    if (env.CHANGE_TARGET == null) {
+        return false
+    }
+
+    if (sh(
+        returnStatus: true,
+        script: "git diff --name-only origin/${env.CHANGE_TARGET}... | grep -v '\\.md'"
+    ) != 0) {
+        return true
+    }
+
+    if (sh(
+        returnStatus: true,
+        script: "git diff --name-only origin/${env.CHANGE_TARGET}... | grep -v -E 'fuzz/'"
+    ) != 0) {
+        return true
+    }
+
+    if (sh(
+        returnStatus: true,
+        script: "git diff --name-only origin/${env.CHANGE_TARGET}... | grep -v -E '.github/'"
+    ) != 0) {
+        return true
+    }
+
+    return false
+}

README.md

@@ -59,13 +59,9 @@ based on the [Rust VMM](https://github.com/rust-vmm) crates.
 
 ### Architectures
 
-Cloud Hypervisor supports the `x86-64`, `AArch64` and `riscv64`
-architectures, with functionality varying across these platforms. The
-functionality differences between `x86-64` and `AArch64` are documented
-in [#1125](https://github.com/cloud-hypervisor/cloud-hypervisor/issues/1125).
-The `riscv64` architecture support is experimental and offers limited
-functionality. For more details and instructions, please refer to [riscv
-documentation](docs/riscv.md).
+Cloud Hypervisor supports the `x86-64` and `AArch64` architectures. There are
+minor differences in functionality between the two architectures
+(see [#1125](https://github.com/cloud-hypervisor/cloud-hypervisor/issues/1125)).
 
 ### Guest OS
 
@@ -82,9 +78,9 @@ The following sections describe how to build and run Cloud Hypervisor.
 
 ## Host OS
 
-For required KVM functionality and adequate performance the recommended host
-kernel version is 5.13. The majority of the CI currently tests with kernel
-version 5.15.
+For required KVM functionality the minimum host kernel version is 4.11.  For
+adequate performance the minimum recommended host kernel version is 5.6. The
+majority of the CI currently tests with kernel version 5.15.
 
 ## Use Pre-built Binaries
 
@@ -111,25 +107,19 @@ do not wish to use the pre-built binaries.
 
 ## Booting Linux
 
-Cloud Hypervisor boots guests in one of two ways. The first is direct
-kernel boot, where a kernel image is passed to `--kernel`. The x86-64
-kernel must be built with PVH support or be a bzImage. The second is
-firmware boot, where a firmware image is passed to `--firmware` and
-brings up the guest's normal boot loader.
+Cloud Hypervisor supports direct kernel boot (the x86-64 kernel requires the kernel
+built with PVH support) or booting via a firmware (either [Rust Hypervisor
+Firmware](https://github.com/cloud-hypervisor/rust-hypervisor-firmware) or an
+edk2 UEFI firmware called `CLOUDHV` / `CLOUDHV_EFI`.)
 
-Two firmware options are supported, and which one works best depends
-on the guest OS. [Rust Hypervisor
-Firmware](https://github.com/cloud-hypervisor/rust-hypervisor-firmware)
-is a lightweight Rust-based PVH firmware. The edk2 UEFI firmware is
-called `CLOUDHV.fd` for x86-64 and `CLOUDHV_EFI.fd` for AArch64.
-Prebuilt binaries for both are available at their respective releases
-pages, [Rust Hypervisor
+Binary builds of the firmware files are available for the latest release of
+[Rust Hypervisor
 Firmware](https://github.com/cloud-hypervisor/rust-hypervisor-firmware/releases/latest)
 and [our edk2
-fork](https://github.com/cloud-hypervisor/edk2/releases/latest).
-The edk2 fork carries customizations required to boot AArch64 guests
-on cloud-hypervisor. See [docs/uefi.md](docs/uefi.md) for differences
-with upstream tianocore/edk2.
+repository](https://github.com/cloud-hypervisor/edk2/releases/latest)
+
+The choice of firmware depends on your guest OS choice; some experimentation
+may be required.
 
 ### Firmware Booting
 
@@ -150,17 +140,14 @@ The Ubuntu cloud images do not ship with a default password so it necessary to
 use a `cloud-init` disk image to customise the image on the first boot. A basic
 `cloud-init` image is generated by this [script](scripts/create-cloud-init.sh).
 This seeds the image with a default username/password of `cloud/cloud123`. It
-is only necessary to add this disk image on the first boot. Script also assigns
-default IP address using `test_data/cloud-init/ubuntu/local/network-config` details
-with `--net "mac=12:34:56:78:90:ab,tap="` option. Then the matching mac address
-interface will be enabled as per `network-config` details.
+is only necessary to add this disk image on the first boot.
 
 ```shell
 $ sudo setcap cap_net_admin+ep ./cloud-hypervisor
 $ ./create-cloud-init.sh
 $ ./cloud-hypervisor \
-	--firmware ./hypervisor-fw \
-	--disk path=focal-server-cloudimg-amd64.raw path=/tmp/ubuntu-cloudinit.img \
+	--kernel ./hypervisor-fw \
+	--disk path=focal-server-cloudimg-amd64.raw --disk path=/tmp/ubuntu-cloudinit.img \
 	--cpus boot=4 \
 	--memory size=1024M \
 	--net "tap=,mac=,ip=,mask="
@@ -173,7 +160,7 @@ GRUB) is required then it necessary to switch to the serial console instead of
 ```shell
 $ ./cloud-hypervisor \
 	--kernel ./hypervisor-fw \
-	--disk path=focal-server-cloudimg-amd64.raw path=/tmp/ubuntu-cloudinit.img \
+	--disk path=focal-server-cloudimg-amd64.raw --disk path=/tmp/ubuntu-cloudinit.img \
 	--cpus boot=4 \
 	--memory size=1024M \
 	--net "tap=,mac=,ip=,mask=" \
@@ -181,31 +168,24 @@ $ ./cloud-hypervisor \
 	--console off
 ```
 
-## Booting: `--firmware` vs `--kernel`
-
-The following scenarios are supported by Cloud Hypervisor to bootstrap a VM, i.e.,
-to load a payload/bootitem(s):
-
-- Provide firmware
-- Provide kernel \[+ cmdline\]\ [+ initrd\]
-
-Please note that our Cloud Hypervisor firmware (`hypervisor-fw`) has a Xen PVH
-boot entry, therefore it can also be booted via the `--kernel` parameter, as 
-seen in some examples.
-
 ### Custom Kernel and Disk Image
 
 #### Building your Kernel
 
-Cloud Hypervisor also supports direct kernel boot. For x86-64, a `vmlinux` ELF kernel (compiled with PVH support) or a regular bzImage are supported. In order to support development there is a custom branch; however provided the required options are enabled any recent kernel will suffice.
+Cloud Hypervisor also supports direct kernel boot. For x86-64, a `vmlinux` ELF kernel (compiled with PVH support) is needed. In order to support development there is a custom branch; however provided the required options are enabled any recent kernel will suffice.
 
 To build the kernel:
 
 ```shell
 # Clone the Cloud Hypervisor Linux branch
-$ git clone --depth 1 https://github.com/cloud-hypervisor/linux.git -b ch-6.12.8 linux-cloud-hypervisor
+$ git clone --depth 1 https://github.com/cloud-hypervisor/linux.git -b ch-6.2 linux-cloud-hypervisor
 $ pushd linux-cloud-hypervisor
-$ make ch_defconfig
+# Use the x86-64 cloud-hypervisor kernel config to build your kernel for x86-64
+$ wget https://raw.githubusercontent.com/cloud-hypervisor/cloud-hypervisor/main/resources/linux-config-x86_64
+# Use the AArch64 cloud-hypervisor kernel config to build your kernel for AArch64
+$ wget https://raw.githubusercontent.com/cloud-hypervisor/cloud-hypervisor/main/resources/linux-config-aarch64
+$ cp linux-config-x86_64 .config  # x86-64
+$ cp linux-config-aarch64 .config # AArch64
 # Do native build of the x86-64 kernel
 $ KCFLAGS="-Wa,-mx86-used-note=no" make bzImage -j `nproc`
 # Do native build of the AArch64 kernel
@@ -242,7 +222,7 @@ $ sudo setcap cap_net_admin+ep ./cloud-hypervisor
 $ ./create-cloud-init.sh
 $ ./cloud-hypervisor \
 	--kernel ./linux-cloud-hypervisor/arch/x86/boot/compressed/vmlinux.bin \
-	--disk path=focal-server-cloudimg-amd64.raw path=/tmp/ubuntu-cloudinit.img \
+	--disk path=focal-server-cloudimg-amd64.raw --disk path=/tmp/ubuntu-cloudinit.img \
 	--cmdline "console=hvc0 root=/dev/vda1 rw" \
 	--cpus boot=4 \
 	--memory size=1024M \
@@ -256,7 +236,7 @@ $ sudo setcap cap_net_admin+ep ./cloud-hypervisor
 $ ./create-cloud-init.sh
 $ ./cloud-hypervisor \
 	--kernel ./linux-cloud-hypervisor/arch/arm64/boot/Image \
-	--disk path=focal-server-cloudimg-arm64.raw path=/tmp/ubuntu-cloudinit.img \
+	--disk path=focal-server-cloudimg-arm64.raw --disk path=/tmp/ubuntu-cloudinit.img \
 	--cmdline "console=hvc0 root=/dev/vda1 rw" \
 	--cpus boot=4 \
 	--memory size=1024M \
@@ -319,9 +299,8 @@ Further details can be found in the [release documentation](docs/releases.md).
 As of 2023-01-03, the following cloud images are supported:
 
 - [Ubuntu Focal](https://cloud-images.ubuntu.com/focal/current/) (focal-server-cloudimg-{amd64,arm64}.img)
-- [Ubuntu Jammy](https://cloud-images.ubuntu.com/jammy/current/) (jammy-server-cloudimg-{amd64,arm64}.img)
-- [Ubuntu Noble](https://cloud-images.ubuntu.com/noble/current/) (noble-server-cloudimg-{amd64,arm64}.img)
-- [Fedora 36](https://archives.fedoraproject.org/pub/archive/fedora/linux/releases/36/Cloud/) ([Fedora-Cloud-Base-36-1.5.x86_64.raw.xz](https://archives.fedoraproject.org/pub/archive/fedora/linux/releases/36/Cloud/x86_64/images/) / [Fedora-Cloud-Base-36-1.5.aarch64.raw.xz](https://archives.fedoraproject.org/pub/archive/fedora/linux/releases/36/Cloud/aarch64/images/))
+- [Ubuntu Jammy](https://cloud-images.ubuntu.com/jammy/current/) (jammy-server-cloudimg-{amd64,arm64}.img )
+- [Fedora 36](https://fedora.mirrorservice.org/fedora/linux/releases/36/Cloud/) ([Fedora-Cloud-Base-36-1.5.x86_64.raw.xz](https://fedora.mirrorservice.org/fedora/linux/releases/36/Cloud/x86_64/images/) / [Fedora-Cloud-Base-36-1.5.aarch64.raw.xz](https://fedora.mirrorservice.org/fedora/linux/releases/36/Cloud/aarch64/images/))
 
 Direct kernel boot to userspace should work with a rootfs from most
 distributions although you may need to enable exotic filesystem types in the

api_client/Cargo.toml

@@ -1,14 +1,8 @@
 [package]
-authors = ["The Cloud Hypervisor Authors"]
-edition.workspace = true
-license = "Apache-2.0"
 name = "api_client"
-rust-version.workspace = true
 version = "0.1.0"
+authors = ["The Cloud Hypervisor Authors"]
+edition = "2021"
 
 [dependencies]
-thiserror = { workspace = true }
-vmm-sys-util = { workspace = true }
-
-[lints]
-workspace = true
+vmm-sys-util = "0.11.0"

api_client/src/lib.rs

@@ -3,32 +3,39 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+use std::fmt;
 use std::io::{Read, Write};
 use std::os::unix::io::RawFd;
-
-use thiserror::Error;
 use vmm_sys_util::sock_ctrl_msg::ScmSocket;
 
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
-    #[error("Error writing to or reading from HTTP socket")]
-    Socket(#[source] std::io::Error),
-    #[error("Error sending file descriptors")]
-    SocketSendFds(#[source] vmm_sys_util::errno::Error),
-    #[error("Error parsing HTTP status code")]
-    StatusCodeParsing(#[source] std::num::ParseIntError),
-    #[error("HTTP output is missing protocol statement")]
+    Socket(std::io::Error),
+    SocketSendFds(vmm_sys_util::errno::Error),
+    StatusCodeParsing(std::num::ParseIntError),
     MissingProtocol,
-    #[error("Error parsing HTTP Content-Length field")]
-    ContentLengthParsing(#[source] std::num::ParseIntError),
-    #[error("Server responded with error {0:?}: {1:?}")]
-    ServerResponse(
-        StatusCode,
-        // TODO: Move `api` module from `vmm` to dedicated crate and use a common type definition
-        Option<
-            String, /* Untyped: Currently Vec<String> of error messages from top to root cause */
-        >,
-    ),
+    ContentLengthParsing(std::num::ParseIntError),
+    ServerResponse(StatusCode, Option<String>),
+}
+
+impl fmt::Display for Error {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        use Error::*;
+        match self {
+            Socket(e) => write!(f, "Error writing to or reading from HTTP socket: {e}"),
+            SocketSendFds(e) => write!(f, "Error writing to or reading from HTTP socket: {e}"),
+            StatusCodeParsing(e) => write!(f, "Error parsing HTTP status code: {e}"),
+            MissingProtocol => write!(f, "HTTP output is missing protocol statement"),
+            ContentLengthParsing(e) => write!(f, "Error parsing HTTP Content-Length field: {e}"),
+            ServerResponse(s, o) => {
+                if let Some(o) = o {
+                    write!(f, "Server responded with an error: {s:?}: {o}")
+                } else {
+                    write!(f, "Server responded with an error: {s:?}")
+                }
+            }
+        }
+    }
 }
 
 #[derive(Clone, Copy, Debug)]
@@ -38,7 +45,6 @@ pub enum StatusCode {
     NoContent,
     BadRequest,
     NotFound,
-    TooManyRequests,
     InternalServerError,
     NotImplemented,
     Unknown,
@@ -52,7 +58,6 @@ impl StatusCode {
             204 => StatusCode::NoContent,
             400 => StatusCode::BadRequest,
             404 => StatusCode::NotFound,
-            429 => StatusCode::TooManyRequests,
             500 => StatusCode::InternalServerError,
             501 => StatusCode::NotImplemented,
             _ => StatusCode::Unknown,
@@ -118,11 +123,12 @@ fn parse_http_response(socket: &mut dyn Read) -> Result<Option<String>, Error> {
             }
         }
 
-        if let Some(body_offset) = body_offset
-            && let Some(content_length) = content_length
-            && res.len() >= content_length + body_offset
-        {
-            break;
+        if let Some(body_offset) = body_offset {
+            if let Some(content_length) = content_length {
+                if res.len() >= content_length + body_offset {
+                    break;
+                }
+            }
         }
     }
     let body_string = content_length.and(body_offset.map(|o| String::from(&res[o..])));
@@ -142,7 +148,7 @@ pub fn simple_api_full_command_with_fds_and_response<T: Read + Write + ScmSocket
     method: &str,
     full_command: &str,
     request_body: Option<&str>,
-    request_fds: &[RawFd],
+    request_fds: Vec<RawFd>,
 ) -> Result<Option<String>, Error> {
     socket
         .send_with_fds(
@@ -150,7 +156,7 @@ pub fn simple_api_full_command_with_fds_and_response<T: Read + Write + ScmSocket
                 "{method} /api/v1/{full_command} HTTP/1.1\r\nHost: localhost\r\nAccept: */*\r\n"
             )
             .as_bytes()],
-            request_fds,
+            &request_fds,
         )
         .map_err(Error::SocketSendFds)?;
 
@@ -178,7 +184,7 @@ pub fn simple_api_full_command_with_fds<T: Read + Write + ScmSocket>(
     method: &str,
     full_command: &str,
     request_body: Option<&str>,
-    request_fds: &[RawFd],
+    request_fds: Vec<RawFd>,
 ) -> Result<(), Error> {
     let response = simple_api_full_command_with_fds_and_response(
         socket,
@@ -188,8 +194,8 @@ pub fn simple_api_full_command_with_fds<T: Read + Write + ScmSocket>(
         request_fds,
     )?;
 
-    if let Some(response) = response {
-        println!("{response}");
+    if response.is_some() {
+        println!("{}", response.unwrap());
     }
 
     Ok(())
@@ -201,7 +207,7 @@ pub fn simple_api_full_command<T: Read + Write + ScmSocket>(
     full_command: &str,
     request_body: Option<&str>,
 ) -> Result<(), Error> {
-    simple_api_full_command_with_fds(socket, method, full_command, request_body, &[])
+    simple_api_full_command_with_fds(socket, method, full_command, request_body, Vec::new())
 }
 
 pub fn simple_api_full_command_and_response<T: Read + Write + ScmSocket>(
@@ -210,7 +216,13 @@ pub fn simple_api_full_command_and_response<T: Read + Write + ScmSocket>(
     full_command: &str,
     request_body: Option<&str>,
 ) -> Result<Option<String>, Error> {
-    simple_api_full_command_with_fds_and_response(socket, method, full_command, request_body, &[])
+    simple_api_full_command_with_fds_and_response(
+        socket,
+        method,
+        full_command,
+        request_body,
+        Vec::new(),
+    )
 }
 
 pub fn simple_api_command_with_fds<T: Read + Write + ScmSocket>(
@@ -218,7 +230,7 @@ pub fn simple_api_command_with_fds<T: Read + Write + ScmSocket>(
     method: &str,
     c: &str,
     request_body: Option<&str>,
-    request_fds: &[RawFd],
+    request_fds: Vec<RawFd>,
 ) -> Result<(), Error> {
     // Create the full VM command. For VMM commands, use
     // simple_api_full_command().
@@ -233,5 +245,5 @@ pub fn simple_api_command<T: Read + Write + ScmSocket>(
     c: &str,
     request_body: Option<&str>,
 ) -> Result<(), Error> {
-    simple_api_command_with_fds(socket, method, c, request_body, &[])
+    simple_api_command_with_fds(socket, method, c, request_body, Vec::new())
 }

arch/Cargo.toml

@@ -1,33 +1,29 @@
 [package]
-authors = ["The Chromium OS Authors"]
-edition.workspace = true
 name = "arch"
-rust-version.workspace = true
 version = "0.1.0"
+authors = ["The Chromium OS Authors"]
+edition = "2021"
 
 [features]
 default = []
-fw_cfg = []
-kvm = ["hypervisor/kvm"]
-sev_snp = []
 tdx = []
 
 [dependencies]
-anyhow = { workspace = true }
-byteorder = { workspace = true }
+anyhow = "1.0.69"
+byteorder = "1.4.3"
 hypervisor = { path = "../hypervisor" }
-libc = { workspace = true }
-linux-loader = { workspace = true, features = ["bzimage", "elf", "pe"] }
-log = { workspace = true }
-serde = { workspace = true, features = ["derive", "rc"] }
-thiserror = { workspace = true }
-uuid = { workspace = true }
-vm-memory = { workspace = true, features = ["backend-bitmap", "backend-mmap"] }
-vmm-sys-util = { workspace = true, features = ["with-serde"] }
+libc = "0.2.139"
+linux-loader = { version = "0.8.1", features = ["elf", "bzimage", "pe"] }
+log = "0.4.17"
+serde = { version = "1.0.151", features = ["rc", "derive"] }
+thiserror = "1.0.39"
+uuid = "1.3.0"
+versionize = "0.1.10"
+versionize_derive = "0.1.4"
+vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-bitmap"] }
+vm-migration = { path = "../vm-migration" }
+vmm-sys-util = { version = "0.11.0", features = ["with-serde"] }
 
-[target.'cfg(any(target_arch = "aarch64", target_arch = "riscv64"))'.dependencies]
-fdt_parser = { version = "0.1.5", package = "fdt" }
-vm-fdt = { workspace = true }
-
-[lints]
-workspace = true
+[target.'cfg(target_arch = "aarch64")'.dependencies]
+fdt_parser = { version = "0.1.4", package = "fdt" }
+vm-fdt = { git = "https://github.com/rust-vmm/vm-fdt", branch = "main" }

arch/src/aarch64/fdt.rs

@@ -6,32 +6,27 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the THIRD-PARTY file.
 
+use crate::{NumaNodes, PciSpaceInfo};
+use byteorder::{BigEndian, ByteOrder};
+use hypervisor::arch::aarch64::gic::Vgic;
+use std::cmp;
 use std::collections::HashMap;
 use std::ffi::CStr;
 use std::fmt::Debug;
-use std::path::Path;
+use std::result;
+use std::str;
 use std::sync::{Arc, Mutex};
-use std::{cmp, fs, result, str};
 
-use byteorder::{BigEndian, ByteOrder};
-use hypervisor::arch::aarch64::gic::Vgic;
-use hypervisor::arch::aarch64::regs::{
-    AARCH64_ARCH_TIMER_HYP_IRQ, AARCH64_ARCH_TIMER_PHYS_NONSECURE_IRQ,
-    AARCH64_ARCH_TIMER_PHYS_SECURE_IRQ, AARCH64_ARCH_TIMER_VIRT_IRQ, AARCH64_PMU_IRQ,
+use super::super::DeviceType;
+use super::super::GuestMemoryMmap;
+use super::super::InitramfsConfig;
+use super::layout::{
+    IRQ_BASE, MEM_32BIT_DEVICES_SIZE, MEM_32BIT_DEVICES_START, MEM_PCI_IO_SIZE, MEM_PCI_IO_START,
+    PCI_HIGH_BASE, PCI_MMIO_CONFIG_SIZE_PER_SEGMENT,
 };
-use log::{debug, info, warn};
-use thiserror::Error;
 use vm_fdt::{FdtWriter, FdtWriterResult};
 use vm_memory::{Address, Bytes, GuestMemory, GuestMemoryError, GuestMemoryRegion};
 
-use super::super::{DeviceType, GuestMemoryMmap, InitramfsConfig};
-use super::layout::{
-    GIC_V2M_COMPATIBLE, GICV2M_SPI_BASE, GICV2M_SPI_NUM, IRQ_BASE, MEM_32BIT_DEVICES_SIZE,
-    MEM_32BIT_DEVICES_START, MEM_PCI_IO_SIZE, MEM_PCI_IO_START, PCI_HIGH_BASE,
-    PCI_MMIO_CONFIG_SIZE_PER_SEGMENT,
-};
-use crate::{NumaNodes, PciSpaceInfo};
-
 // This is a value for uniquely identifying the FDT node declaring the interrupt controller.
 const GIC_PHANDLE: u32 = 1;
 // This is a value for uniquely identifying the FDT node declaring the MSI controller.
@@ -45,12 +40,8 @@ const VIRTIO_IOMMU_PHANDLE: u32 = 5;
 // NOTE: Keep FIRST_VCPU_PHANDLE the last PHANDLE defined.
 // This is a value for uniquely identifying the FDT node containing the first vCPU.
 // The last number of vCPU phandle depends on the number of vCPUs.
-const FIRST_VCPU_PHANDLE: u32 = 8;
+const FIRST_VCPU_PHANDLE: u32 = 6;
 
-// This is a value for uniquely identifying the FDT node containing the L2 cache info
-const L2_CACHE_PHANDLE: u32 = 6;
-// This is a value for uniquely identifying the FDT node containing the L3 cache info
-const L3_CACHE_PHANDLE: u32 = 7;
 // Read the documentation specified when appending the root node to the FDT.
 const ADDRESS_CELLS: u32 = 0x2;
 const SIZE_CELLS: u32 = 0x2;
@@ -65,6 +56,9 @@ const GIC_FDT_IRQ_TYPE_PPI: u32 = 1;
 const IRQ_TYPE_EDGE_RISING: u32 = 1;
 const IRQ_TYPE_LEVEL_HI: u32 = 4;
 
+// PMU PPI interrupt number
+pub const AARCH64_PMU_IRQ: u32 = 7;
+
 // Keys and Buttons
 // System Power Down
 const KEY_POWER: u32 = 116;
@@ -80,136 +74,20 @@ pub trait DeviceInfoForFdt {
 }
 
 /// Errors thrown while configuring the Flattened Device Tree for aarch64.
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// Failure in writing FDT in memory.
-    #[error("Failure in writing FDT in memory")]
-    WriteFdtToMemory(#[source] GuestMemoryError),
+    WriteFdtToMemory(GuestMemoryError),
 }
 type Result<T> = result::Result<T, Error>;
 
-#[derive(Copy, Clone)]
-pub enum CacheLevel {
-    /// L1 data cache
-    L1D = 0,
-    /// L1 instruction cache
-    L1I = 1,
-    /// L2 cache
-    L2 = 2,
-    /// L3 cache
-    L3 = 3,
-}
-
-/// NOTE: cache size file directory example,
-/// "/sys/devices/system/cpu/cpu0/cache/index0/size".
-pub fn get_cache_size(cache_level: CacheLevel) -> u32 {
-    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
-    match cache_level {
-        CacheLevel::L1D => file_directory += "/index0/size",
-        CacheLevel::L1I => file_directory += "/index1/size",
-        CacheLevel::L2 => file_directory += "/index2/size",
-        CacheLevel::L3 => file_directory += "/index3/size",
-    }
-
-    let file_path = Path::new(&file_directory);
-    if file_path.exists() {
-        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
-        // The content of the file is as simple as a size, like: "32K"
-        let src = src.trim();
-        let src_digits: u32 = src[0..src.len() - 1].parse().unwrap();
-        let src_unit = &src[src.len() - 1..];
-
-        src_digits
-            * match src_unit {
-                "K" => 1024,
-                "M" => 1024u32.pow(2),
-                "G" => 1024u32.pow(3),
-                _ => 1,
-            }
-    } else {
-        0
-    }
-}
-
-/// NOTE: coherency_line_size file directory example,
-/// "/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size".
-pub fn get_cache_coherency_line_size(cache_level: CacheLevel) -> u32 {
-    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
-    match cache_level {
-        CacheLevel::L1D => file_directory += "/index0/coherency_line_size",
-        CacheLevel::L1I => file_directory += "/index1/coherency_line_size",
-        CacheLevel::L2 => file_directory += "/index2/coherency_line_size",
-        CacheLevel::L3 => file_directory += "/index3/coherency_line_size",
-    }
-
-    let file_path = Path::new(&file_directory);
-    if file_path.exists() {
-        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
-        src.trim().parse::<u32>().unwrap()
-    } else {
-        0
-    }
-}
-
-/// NOTE: number_of_sets file directory example,
-/// "/sys/devices/system/cpu/cpu0/cache/index0/number_of_sets".
-pub fn get_cache_number_of_sets(cache_level: CacheLevel) -> u32 {
-    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
-    match cache_level {
-        CacheLevel::L1D => file_directory += "/index0/number_of_sets",
-        CacheLevel::L1I => file_directory += "/index1/number_of_sets",
-        CacheLevel::L2 => file_directory += "/index2/number_of_sets",
-        CacheLevel::L3 => file_directory += "/index3/number_of_sets",
-    }
-
-    let file_path = Path::new(&file_directory);
-    if file_path.exists() {
-        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
-        src.trim().parse::<u32>().unwrap()
-    } else {
-        0
-    }
-}
-
-/// NOTE: shared_cpu_list file directory example,
-/// "/sys/devices/system/cpu/cpu0/cache/index0/shared_cpu_list".
-pub fn get_cache_shared(cache_level: CacheLevel) -> bool {
-    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
-    let mut result = true;
-
-    match cache_level {
-        CacheLevel::L1D | CacheLevel::L1I => result = false,
-        CacheLevel::L2 => file_directory += "/index2/shared_cpu_list",
-        CacheLevel::L3 => file_directory += "/index3/shared_cpu_list",
-    }
-
-    if !result {
-        return false;
-    }
-
-    let file_path = Path::new(&file_directory);
-    if file_path.exists() {
-        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
-        let src = src.trim();
-        if src.is_empty() {
-            result = false;
-        } else {
-            result = src.contains('-') || src.contains(',');
-        }
-    } else {
-        result = false;
-    }
-
-    result
-}
-
 /// Creates the flattened device tree for this aarch64 VM.
 #[allow(clippy::too_many_arguments)]
 pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
     guest_mem: &GuestMemoryMmap,
     cmdline: &str,
-    vcpu_mpidr: &[u64],
-    vcpu_topology: Option<(u16, u16, u16, u16)>,
+    vcpu_mpidr: Vec<u64>,
+    vcpu_topology: Option<(u8, u8, u8)>,
     device_info: &HashMap<(DeviceType, String), T, S>,
     gic_device: &Arc<Mutex<dyn Vgic>>,
     initrd: &Option<InitramfsConfig>,
@@ -222,8 +100,8 @@ pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHash
     let mut fdt = FdtWriter::new().unwrap();
 
     // For an explanation why these nodes were introduced in the blob take a look at
-    // the "Device Node Requirements" chapter of the Devicetree Specification.
-    // https://www.devicetree.org/specifications/
+    // https://github.com/torvalds/linux/blob/master/Documentation/devicetree/booting-without-of.txt#L845
+    // Look for "Required nodes and properties".
 
     // Header or the root node as per above mentioned documentation.
     let root_node = fdt.begin_node("")?;
@@ -235,7 +113,7 @@ pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHash
     // This is not mandatory but we use it to point the root node to the node
     // containing description of the interrupt controller for this VM.
     fdt.property_u32("interrupt-parent", GIC_PHANDLE)?;
-    create_cpu_nodes(&mut fdt, vcpu_mpidr, vcpu_topology, numa_nodes)?;
+    create_cpu_nodes(&mut fdt, &vcpu_mpidr, vcpu_topology, numa_nodes)?;
     create_memory_node(&mut fdt, guest_mem, numa_nodes)?;
     create_chosen_node(&mut fdt, cmdline, initrd)?;
     create_gic_node(&mut fdt, gic_device)?;
@@ -259,10 +137,10 @@ pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHash
     Ok(fdt_final)
 }
 
-pub fn write_fdt_to_memory(fdt_final: &[u8], guest_mem: &GuestMemoryMmap) -> Result<()> {
+pub fn write_fdt_to_memory(fdt_final: Vec<u8>, guest_mem: &GuestMemoryMmap) -> Result<()> {
     // Write FDT to memory.
     guest_mem
-        .write_slice(fdt_final, super::layout::FDT_START)
+        .write_slice(fdt_final.as_slice(), super::layout::FDT_START)
         .map_err(Error::WriteFdtToMemory)?;
     Ok(())
 }
@@ -271,7 +149,7 @@ pub fn write_fdt_to_memory(fdt_final: &[u8], guest_mem: &GuestMemoryMmap) -> Res
 fn create_cpu_nodes(
     fdt: &mut FdtWriter,
     vcpu_mpidr: &[u64],
-    vcpu_topology: Option<(u16, u16, u16, u16)>,
+    vcpu_topology: Option<(u8, u8, u8)>,
     numa_nodes: &NumaNodes,
 ) -> FdtWriterResult<()> {
     // See https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/arm/cpus.yaml.
@@ -280,81 +158,6 @@ fn create_cpu_nodes(
     fdt.property_u32("#size-cells", 0x0)?;
 
     let num_cpus = vcpu_mpidr.len();
-    let (threads_per_core, cores_per_die, dies_per_package, packages) =
-        vcpu_topology.unwrap_or((1, 1, 1, 1));
-    let cores_per_package = cores_per_die * dies_per_package;
-    let max_cpus: u32 =
-        threads_per_core as u32 * cores_per_die as u32 * dies_per_package as u32 * packages as u32;
-
-    // Add cache info.
-    // L1 Data Cache Info.
-    let mut l1_d_cache_size: u32 = 0;
-    let mut l1_d_cache_line_size: u32 = 0;
-    let mut l1_d_cache_sets: u32 = 0;
-
-    // L1 Instruction Cache Info.
-    let mut l1_i_cache_size: u32 = 0;
-    let mut l1_i_cache_line_size: u32 = 0;
-    let mut l1_i_cache_sets: u32 = 0;
-
-    // L2 Cache Info.
-    let mut l2_cache_size: u32 = 0;
-    let mut l2_cache_line_size: u32 = 0;
-    let mut l2_cache_sets: u32 = 0;
-
-    // L3 Cache Info.
-    let mut l3_cache_size: u32 = 0;
-    let mut l3_cache_line_size: u32 = 0;
-    let mut l3_cache_sets: u32 = 0;
-
-    // Cache Shared Info.
-    let mut l2_cache_shared: bool = false;
-    let mut l3_cache_shared: bool = false;
-
-    let cache_path = Path::new("/sys/devices/system/cpu/cpu0/cache");
-    let cache_exist: bool = cache_path.exists();
-    if cache_exist {
-        // L1 Data Cache Info.
-        l1_d_cache_size = get_cache_size(CacheLevel::L1D);
-        l1_d_cache_line_size = get_cache_coherency_line_size(CacheLevel::L1D);
-        l1_d_cache_sets = get_cache_number_of_sets(CacheLevel::L1D);
-
-        // L1 Instruction Cache Info.
-        l1_i_cache_size = get_cache_size(CacheLevel::L1I);
-        l1_i_cache_line_size = get_cache_coherency_line_size(CacheLevel::L1I);
-        l1_i_cache_sets = get_cache_number_of_sets(CacheLevel::L1I);
-
-        // L2 Cache Info.
-        l2_cache_size = get_cache_size(CacheLevel::L2);
-        l2_cache_line_size = get_cache_coherency_line_size(CacheLevel::L2);
-        l2_cache_sets = get_cache_number_of_sets(CacheLevel::L2);
-
-        // L3 Cache Info.
-        l3_cache_size = get_cache_size(CacheLevel::L3);
-        l3_cache_line_size = get_cache_coherency_line_size(CacheLevel::L3);
-        l3_cache_sets = get_cache_number_of_sets(CacheLevel::L3);
-
-        // Cache Shared Info.
-        if l2_cache_size != 0 {
-            l2_cache_shared = get_cache_shared(CacheLevel::L2);
-        }
-        if l3_cache_size != 0 {
-            l3_cache_shared = get_cache_shared(CacheLevel::L3);
-        }
-    } else {
-        warn!("cache sysfs system does not exist.");
-    }
-
-    // Arm boot protocol requires a minimal Device Tree
-    // https://docs.kernel.org/arch/arm64/booting.html
-    // As Generic initiators are supported only in ACPI
-    // When a guest kernel does not boot under "acpi=force" mode it can
-    // hang due to conflicting numa information present in FDT which
-    // does not support Generic Initiators
-    let has_generic_initiator = numa_nodes.values().any(|node| node.device_id.is_some());
-    if has_generic_initiator {
-        info!("Skipping NUMA CPU node encoding in FDT with Generic Initiator devices");
-    }
 
     for (cpu_id, mpidr) in vcpu_mpidr.iter().enumerate().take(num_cpus) {
         let cpu_name = format!("cpu@{cpu_id:x}");
@@ -370,103 +173,21 @@ fn create_cpu_nodes(
         fdt.property_u32("reg", (mpidr & 0x7FFFFF) as u32)?;
         fdt.property_u32("phandle", cpu_id as u32 + FIRST_VCPU_PHANDLE)?;
 
-        // Skipping NUMA encoding in FDT when Generic Initiator devices
-        // are present allowed such guest kernels to boot properly and
-        // rely solely on ACPI tables to setup NUMA
-        if numa_nodes.len() > 1 && !has_generic_initiator {
+        // Add `numa-node-id` property if there is any numa config.
+        if numa_nodes.len() > 1 {
             for numa_node_idx in 0..numa_nodes.len() {
                 let numa_node = numa_nodes.get(&(numa_node_idx as u32));
-                if numa_node.unwrap().cpus.contains(&(cpu_id as u32)) {
+                if numa_node.unwrap().cpus.contains(&(cpu_id as u8)) {
                     fdt.property_u32("numa-node-id", numa_node_idx as u32)?;
                 }
             }
         }
 
-        if cache_exist && l1_d_cache_size != 0 && l1_i_cache_size != 0 {
-            // Add cache info.
-            fdt.property_u32("d-cache-size", l1_d_cache_size)?;
-            fdt.property_u32("d-cache-line-size", l1_d_cache_line_size)?;
-            fdt.property_u32("d-cache-sets", l1_d_cache_sets)?;
-
-            fdt.property_u32("i-cache-size", l1_i_cache_size)?;
-            fdt.property_u32("i-cache-line-size", l1_i_cache_line_size)?;
-            fdt.property_u32("i-cache-sets", l1_i_cache_sets)?;
-
-            if l2_cache_size != 0 && !l2_cache_shared {
-                fdt.property_u32(
-                    "next-level-cache",
-                    cpu_id as u32 + max_cpus + FIRST_VCPU_PHANDLE + L2_CACHE_PHANDLE,
-                )?;
-
-                let l2_cache_name = "l2-cache0";
-                let l2_cache_node = fdt.begin_node(l2_cache_name)?;
-                // PHANDLE is used to mark device node, and PHANDLE is unique. To avoid phandle
-                // conflicts with other device nodes, consider the previous CPU PHANDLE, so the
-                // CPU L2 cache PHANDLE must start from the largest CPU PHANDLE plus 1.
-                fdt.property_u32(
-                    "phandle",
-                    cpu_id as u32 + max_cpus + FIRST_VCPU_PHANDLE + L2_CACHE_PHANDLE,
-                )?;
-
-                fdt.property_string("compatible", "cache")?;
-                fdt.property_u32("cache-size", l2_cache_size)?;
-                fdt.property_u32("cache-line-size", l2_cache_line_size)?;
-                fdt.property_u32("cache-sets", l2_cache_sets)?;
-                fdt.property_u32("cache-level", 2)?;
-
-                if l3_cache_size != 0 && l3_cache_shared {
-                    let package_id: u32 = cpu_id as u32 / cores_per_package as u32;
-                    fdt.property_u32(
-                        "next-level-cache",
-                        package_id
-                            + num_cpus as u32
-                            + max_cpus
-                            + FIRST_VCPU_PHANDLE
-                            + L2_CACHE_PHANDLE
-                            + L3_CACHE_PHANDLE,
-                    )?;
-                }
-
-                fdt.end_node(l2_cache_node)?;
-            }
-        }
-
         fdt.end_node(cpu_node)?;
     }
 
-    if cache_exist && l3_cache_size != 0 && !l2_cache_shared && l3_cache_shared {
-        let mut i: u32 = 0;
-        while i < packages.into() {
-            let l3_cache_name = "l3-cache0";
-            let l3_cache_node = fdt.begin_node(l3_cache_name)?;
-            // ARM L3 cache is generally shared within the package (socket), so the
-            // L3 cache node pointed to by the CPU in the package has the same L3
-            // cache PHANDLE. The L3 cache phandle must start from the largest L2
-            // cache PHANDLE plus 1 to avoid duplication.
-            fdt.property_u32(
-                "phandle",
-                i + num_cpus as u32
-                    + max_cpus
-                    + FIRST_VCPU_PHANDLE
-                    + L2_CACHE_PHANDLE
-                    + L3_CACHE_PHANDLE,
-            )?;
-
-            fdt.property_string("compatible", "cache")?;
-            fdt.property_null("cache-unified")?;
-            fdt.property_u32("cache-size", l3_cache_size)?;
-            fdt.property_u32("cache-line-size", l3_cache_line_size)?;
-            fdt.property_u32("cache-sets", l3_cache_sets)?;
-            fdt.property_u32("cache-level", 3)?;
-            fdt.end_node(l3_cache_node)?;
-
-            i += 1;
-        }
-    }
-
     if let Some(topology) = vcpu_topology {
-        let (threads_per_core, cores_per_die, dies_per_package, packages) = topology;
-        let cores_per_package = cores_per_die * dies_per_package;
+        let (threads_per_core, cores_per_package, packages) = topology;
         let cpu_map_node = fdt.begin_node("cpu-map")?;
 
         // Create device tree nodes with regard of above mapping.
@@ -514,14 +235,7 @@ fn create_memory_node(
 ) -> FdtWriterResult<()> {
     // See https://github.com/torvalds/linux/blob/58ae0b51506802713aa0e9956d1853ba4c722c98/Documentation/devicetree/bindings/numa.txt
     // for NUMA setting in memory node.
-    let has_generic_initiator = numa_nodes.values().any(|node| node.device_id.is_some());
-    if has_generic_initiator {
-        info!("Skipping NUMA memory node encoding in FDT with Generic Initiator devices");
-    }
-    // Skipping NUMA encoding in FDT when Generic Initiator devices
-    // are present allowed guest kernels to boot and
-    // rely solely on ACPI tables to setup NUMA
-    if numa_nodes.len() > 1 && !has_generic_initiator {
+    if numa_nodes.len() > 1 {
         for numa_node_idx in 0..numa_nodes.len() {
             let numa_node = numa_nodes.get(&(numa_node_idx as u32));
             let mut mem_reg_prop: Vec<u64> = Vec::new();
@@ -533,20 +247,17 @@ fn create_memory_node(
                 let memory_region_size: u64 = memory_region.size() as u64;
                 mem_reg_prop.push(memory_region_start_addr);
                 mem_reg_prop.push(memory_region_size);
-                // Set the node address the first non-zero region address
+                // Set the node address the first non-zero regison address
                 if node_memory_addr == 0 {
                     node_memory_addr = memory_region_start_addr;
                 }
             }
-            // Only create a memory node if this NUMA node has memory regions
-            if !mem_reg_prop.is_empty() {
-                let memory_node_name = format!("memory@{node_memory_addr:x}");
-                let memory_node = fdt.begin_node(&memory_node_name)?;
-                fdt.property_string("device_type", "memory")?;
-                fdt.property_array_u64("reg", &mem_reg_prop)?;
-                fdt.property_u32("numa-node-id", numa_node_idx as u32)?;
-                fdt.end_node(memory_node)?;
-            }
+            let memory_node_name = format!("memory@{node_memory_addr:x}");
+            let memory_node = fdt.begin_node(&memory_node_name)?;
+            fdt.property_string("device_type", "memory")?;
+            fdt.property_array_u64("reg", &mem_reg_prop)?;
+            fdt.property_u32("numa-node-id", numa_node_idx as u32)?;
+            fdt.end_node(memory_node)?;
         }
     } else {
         // Note: memory regions from "GuestMemory" are sorted and non-zero sized.
@@ -582,7 +293,7 @@ fn create_memory_node(
 
         if ram_regions.len() > 2 {
             panic!(
-                "There should be up to two non-continuous regions, divided by the
+                "There should be up to two non-continuous regions, devidided by the
                     gap at the end of 32bit address space."
             );
         }
@@ -600,14 +311,15 @@ fn create_memory_node(
                 && (first_region_end <= &mem_32bit_reserved_start))
             {
                 panic!(
-                    "Unexpected first memory region layout: (start: 0x{first_region_start:08x}, end: 0x{first_region_end:08x}).
-                    ram_start: 0x{ram_start:08x}, mem_32bit_reserved_start: 0x{mem_32bit_reserved_start:08x}"
+                    "Unexpected first memory region layout: (start: 0x{:08x}, end: 0x{:08x}).
+                    ram_start: 0x{:08x}, mem_32bit_reserved_start: 0x{:08x}",
+                    first_region_start, first_region_end, ram_start, mem_32bit_reserved_start
                 );
             }
 
             let mem_size = first_region_end - ram_start;
             let mem_reg_prop = [ram_start, mem_size];
-            let memory_node_name = format!("memory@{ram_start:x}");
+            let memory_node_name = format!("memory@{:x}", ram_start);
             let memory_node = fdt.begin_node(&memory_node_name)?;
             fdt.property_string("device_type", "memory")?;
             fdt.property_array_u64("reg", &mem_reg_prop)?;
@@ -620,13 +332,14 @@ fn create_memory_node(
 
             if second_region_start != &ram_64bit_start {
                 panic!(
-                    "Unexpected second memory region layout: start: 0x{second_region_start:08x}, ram_64bit_start: 0x{ram_64bit_start:08x}"
+                    "Unexpected second memory region layout: start: 0x{:08x}, ram_64bit_start: 0x{:08x}",
+                    second_region_start, ram_64bit_start
                 );
             }
 
             let mem_size = second_region_end - ram_64bit_start;
             let mem_reg_prop = [ram_64bit_start, mem_size];
-            let memory_node_name = format!("memory@{ram_64bit_start:x}");
+            let memory_node_name = format!("memory@{:x}", ram_64bit_start);
             let memory_node = fdt.begin_node(&memory_node_name)?;
             fdt.property_string("device_type", "memory")?;
             fdt.property_array_u64("reg", &mem_reg_prop)?;
@@ -683,19 +396,11 @@ fn create_gic_node(fdt: &mut FdtWriter, gic_device: &Arc<Mutex<dyn Vgic>>) -> Fd
 
     if gic_device.lock().unwrap().msi_compatible() {
         let msic_node = fdt.begin_node("msic")?;
-        let msi_compatibility = gic_device.lock().unwrap().msi_compatibility().to_string();
-
-        fdt.property_string("compatible", msi_compatibility.as_str())?;
+        fdt.property_string("compatible", gic_device.lock().unwrap().msi_compatibility())?;
         fdt.property_null("msi-controller")?;
         fdt.property_u32("phandle", MSI_PHANDLE)?;
         let msi_reg_prop = gic_device.lock().unwrap().msi_properties();
         fdt.property_array_u64("reg", &msi_reg_prop)?;
-
-        if msi_compatibility == GIC_V2M_COMPATIBLE {
-            fdt.property_u32("arm,msi-base-spi", GICV2M_SPI_BASE)?;
-            fdt.property_u32("arm,msi-num-spis", GICV2M_SPI_NUM)?;
-        }
-
         fdt.end_node(msic_node)?;
     }
 
@@ -722,14 +427,9 @@ fn create_clock_node(fdt: &mut FdtWriter) -> FdtWriterResult<()> {
 
 fn create_timer_node(fdt: &mut FdtWriter) -> FdtWriterResult<()> {
     // See
-    // https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/timer/arm%2Carch_timer.yaml
+    // https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/interrupt-controller/arch_timer.txt
     // These are fixed interrupt numbers for the timer device.
-    let irqs = [
-        AARCH64_ARCH_TIMER_PHYS_SECURE_IRQ,
-        AARCH64_ARCH_TIMER_PHYS_NONSECURE_IRQ,
-        AARCH64_ARCH_TIMER_VIRT_IRQ,
-        AARCH64_ARCH_TIMER_HYP_IRQ,
-    ];
+    let irqs = [13, 14, 11, 10];
     let compatible = "arm,armv8-timer";
 
     let mut timer_reg_cells: Vec<u32> = Vec::new();
@@ -864,21 +564,6 @@ fn create_gpio_node<T: DeviceInfoForFdt + Clone + Debug>(
     Ok(())
 }
 
-// https://www.kernel.org/doc/Documentation/devicetree/bindings/arm/fw-cfg.txt
-#[cfg(feature = "fw_cfg")]
-fn create_fw_cfg_node<T: DeviceInfoForFdt + Clone + Debug>(
-    fdt: &mut FdtWriter,
-    dev_info: &T,
-) -> FdtWriterResult<()> {
-    // FwCfg node
-    let fw_cfg_node = fdt.begin_node(&format!("fw-cfg@{:x}", dev_info.addr()))?;
-    fdt.property("compatible", b"qemu,fw-cfg-mmio\0")?;
-    fdt.property_array_u64("reg", &[dev_info.addr(), dev_info.length()])?;
-    fdt.end_node(fw_cfg_node)?;
-
-    Ok(())
-}
-
 fn create_devices_node<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
     fdt: &mut FdtWriter,
     dev_info: &HashMap<(DeviceType, String), T, S>,
@@ -894,8 +579,6 @@ fn create_devices_node<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::Buil
             DeviceType::Virtio(_) => {
                 ordered_virtio_device.push(info);
             }
-            #[cfg(feature = "fw_cfg")]
-            DeviceType::FwCfg => create_fw_cfg_node(fdt, info)?,
         }
     }
 
@@ -935,7 +618,7 @@ fn create_pci_nodes(
     for pci_device_info_elem in pci_device_info.iter() {
         // EDK2 requires the PCIe high space above 4G address.
         // The actual space in CLH follows the RAM. If the RAM space is small, the PCIe high space
-        // could fall below 4G.
+        // could fall bellow 4G.
         // Here we cut off PCI device space below 8G in FDT to workaround the EDK2 check.
         // But the address written in ACPI is not impacted.
         let (pci_device_base_64bit, pci_device_size_64bit) =
@@ -1025,39 +708,39 @@ fn create_pci_nodes(
         fdt.property_array_u32("msi-map", &msi_map)?;
         fdt.property_u32("msi-parent", MSI_PHANDLE)?;
 
-        if pci_device_info_elem.pci_segment_id == 0
-            && let Some(virtio_iommu_bdf) = virtio_iommu_bdf
-        {
-            // See kernel document Documentation/devicetree/bindings/pci/pci-iommu.txt
-            // for 'iommu-map' attribute setting.
-            let iommu_map = [
-                0_u32,
-                VIRTIO_IOMMU_PHANDLE,
-                0_u32,
-                virtio_iommu_bdf,
-                virtio_iommu_bdf + 1,
-                VIRTIO_IOMMU_PHANDLE,
-                virtio_iommu_bdf + 1,
-                0xffff - virtio_iommu_bdf,
-            ];
-            fdt.property_array_u32("iommu-map", &iommu_map)?;
+        if pci_device_info_elem.pci_segment_id == 0 {
+            if let Some(virtio_iommu_bdf) = virtio_iommu_bdf {
+                // See kernel document Documentation/devicetree/bindings/pci/pci-iommu.txt
+                // for 'iommu-map' attribute setting.
+                let iommu_map = [
+                    0_u32,
+                    VIRTIO_IOMMU_PHANDLE,
+                    0_u32,
+                    virtio_iommu_bdf,
+                    virtio_iommu_bdf + 1,
+                    VIRTIO_IOMMU_PHANDLE,
+                    virtio_iommu_bdf + 1,
+                    0xffff - virtio_iommu_bdf,
+                ];
+                fdt.property_array_u32("iommu-map", &iommu_map)?;
 
-            // See kernel document Documentation/devicetree/bindings/virtio/iommu.txt
-            // for virtio-iommu node settings.
-            let virtio_iommu_node_name = format!("virtio_iommu@{virtio_iommu_bdf:x}");
-            let virtio_iommu_node = fdt.begin_node(&virtio_iommu_node_name)?;
-            fdt.property_u32("#iommu-cells", 1)?;
-            fdt.property_string("compatible", "virtio,pci-iommu")?;
+                // See kernel document Documentation/devicetree/bindings/virtio/iommu.txt
+                // for virtio-iommu node settings.
+                let virtio_iommu_node_name = format!("virtio_iommu@{virtio_iommu_bdf:x}");
+                let virtio_iommu_node = fdt.begin_node(&virtio_iommu_node_name)?;
+                fdt.property_u32("#iommu-cells", 1)?;
+                fdt.property_string("compatible", "virtio,pci-iommu")?;
 
-            // 'reg' is a five-cell address encoded as
-            // (phys.hi phys.mid phys.lo size.hi size.lo). phys.hi should contain the
-            // device's BDF as 0b00000000 bbbbbbbb dddddfff 00000000. The other cells
-            // should be zero.
-            let reg = [virtio_iommu_bdf << 8, 0_u32, 0_u32, 0_u32, 0_u32];
-            fdt.property_array_u32("reg", &reg)?;
-            fdt.property_u32("phandle", VIRTIO_IOMMU_PHANDLE)?;
+                // 'reg' is a five-cell address encoded as
+                // (phys.hi phys.mid phys.lo size.hi size.lo). phys.hi should contain the
+                // device's BDF as 0b00000000 bbbbbbbb dddddfff 00000000. The other cells
+                // should be zero.
+                let reg = [virtio_iommu_bdf << 8, 0_u32, 0_u32, 0_u32, 0_u32];
+                fdt.property_array_u32("reg", &reg)?;
+                fdt.property_u32("phandle", VIRTIO_IOMMU_PHANDLE)?;
 
-            fdt.end_node(virtio_iommu_node)?;
+                fdt.end_node(virtio_iommu_node)?;
+            }
         }
 
         fdt.end_node(pci_node)?;
@@ -1067,22 +750,6 @@ fn create_pci_nodes(
 }
 
 fn create_distance_map_node(fdt: &mut FdtWriter, numa_nodes: &NumaNodes) -> FdtWriterResult<()> {
-    // When Generic Initiator nodes are present, skip ALL FDT NUMA information.
-    // Let ACPI (which supports Generic Initiator via SRAT Type 5) handle the entire NUMA topology.
-    // FDT cannot represent Generic Initiator nodes, and mixing FDT + ACPI NUMA info causes conflicts.
-    let has_generic_initiator = numa_nodes.values().any(|node| node.device_id.is_some());
-    if has_generic_initiator {
-        info!("Skipping NUMA distance map encoding in FDT with Generic Initiator devices");
-        return Ok(());
-    }
-    // At this point, we know there are no Generic Initiator nodes
-    let mut numa_ids: Vec<u32> = numa_nodes.keys().cloned().collect();
-
-    // If we only have one node, no distance map is needed
-    if numa_ids.len() <= 1 {
-        return Ok(());
-    }
-
     let distance_map_node = fdt.begin_node("distance-map")?;
     fdt.property_string("compatible", "numa-distance-map-v1")?;
     // Construct the distance matrix.
@@ -1095,33 +762,26 @@ fn create_distance_map_node(fdt: &mut FdtWriter, numa_nodes: &NumaNodes) -> FdtW
     // a value greater than 10.
     // 4. distance-matrix should have entries in lexicographical ascending
     // order of nodes.
-    numa_ids.sort_unstable(); // lexicographical order
     let mut distance_matrix = Vec::new();
-    // Iterate over actual numa IDs instead of 0..len()
-    for numa_id in numa_ids.iter() {
-        let numa_node = &numa_nodes[numa_id];
-        for dest_numa_id in numa_ids.iter() {
-            if *numa_id == *dest_numa_id {
-                distance_matrix.push(*numa_id);
-                distance_matrix.push(*dest_numa_id);
+    for numa_node_idx in 0..numa_nodes.len() {
+        let numa_node = numa_nodes.get(&(numa_node_idx as u32));
+        for dest_numa_node in 0..numa_node.unwrap().distances.len() + 1 {
+            if numa_node_idx == dest_numa_node {
+                distance_matrix.push(numa_node_idx as u32);
+                distance_matrix.push(dest_numa_node as u32);
                 distance_matrix.push(10_u32);
                 continue;
             }
 
-            distance_matrix.push(*numa_id);
-            distance_matrix.push(*dest_numa_id);
-            // Use user-specified distance, checking both directions for symmetry
-            let distance = if let Some(&dist) = numa_node.distances.get(dest_numa_id) {
-                // Forward direction: current node -> dest node
-                dist
-            } else if let Some(dest_node) = numa_nodes.get(dest_numa_id) {
-                // Reverse direction for symmetry: dest node -> current node
-                dest_node.distances.get(numa_id).copied().unwrap_or(20)
-            } else {
-                // Default distance when neither direction is specified
-                20
-            };
-            distance_matrix.push(distance as u32);
+            distance_matrix.push(numa_node_idx as u32);
+            distance_matrix.push(dest_numa_node as u32);
+            distance_matrix.push(
+                *numa_node
+                    .unwrap()
+                    .distances
+                    .get(&(dest_numa_node as u32))
+                    .unwrap() as u32,
+            );
         }
     }
     fdt.property_array_u32("distance-matrix", distance_matrix.as_ref())?;
@@ -1175,7 +835,10 @@ fn print_node(node: fdt_parser::node::FdtNode<'_, '_>, n_spaces: usize) {
         //   - At first, try to convert it to CStr and print,
         //   - If failed, print it as u32 array.
         let value_result = match CStr::from_bytes_with_nul(value) {
-            Ok(value_cstr) => value_cstr.to_str().ok(),
+            Ok(value_cstr) => match value_cstr.to_str() {
+                Ok(value_str) => Some(value_str),
+                Err(_e) => None,
+            },
             Err(_e) => None,
         };
 
@@ -1198,7 +861,7 @@ fn print_node(node: fdt_parser::node::FdtNode<'_, '_>, n_spaces: usize) {
                 array,
                 indent = (n_spaces + 2)
             );
-        }
+        };
     }
 
     // Print children nodes if there is any
@@ -1206,118 +869,3 @@ fn print_node(node: fdt_parser::node::FdtNode<'_, '_>, n_spaces: usize) {
         print_node(child, n_spaces + 2);
     }
 }
-
-#[cfg(test)]
-mod tests {
-    use std::collections::BTreeMap;
-
-    use super::*;
-    use crate::NumaNode;
-
-    // Helper function to create a simple NumaNode for testing
-    fn create_test_numa_node(cpus: Vec<u32>, device_id: Option<String>) -> NumaNode {
-        NumaNode {
-            memory_regions: Vec::new(),
-            hotplug_regions: Vec::new(),
-            cpus,
-            pci_segments: Vec::new(),
-            distances: BTreeMap::new(),
-            memory_zones: Vec::new(),
-            device_id,
-        }
-    }
-
-    #[test]
-    fn test_fdt_generic_initiator_detection_and_skip() {
-        // No Generic Initiator - should not skip FDT NUMA
-        let mut numa_nodes = BTreeMap::new();
-        numa_nodes.insert(0, create_test_numa_node(vec![0, 1], None));
-        numa_nodes.insert(1, create_test_numa_node(vec![2, 3], None));
-
-        let has_gi = numa_nodes.values().any(|node| node.device_id.is_some());
-        assert!(
-            !has_gi,
-            "Should not detect Generic Initiator when none present"
-        );
-
-        // One Generic Initiator - should skip FDT NUMA
-        let mut numa_nodes = BTreeMap::new();
-        numa_nodes.insert(0, create_test_numa_node(vec![0, 1], None));
-        numa_nodes.insert(1, create_test_numa_node(vec![], Some("vfio0".to_string())));
-
-        let has_gi = numa_nodes.values().any(|node| node.device_id.is_some());
-        assert!(has_gi, "Should detect Generic Initiator when present");
-
-        let mut fdt = FdtWriter::new().unwrap();
-        let result = create_distance_map_node(&mut fdt, &numa_nodes);
-        assert!(result.is_ok(), "Should skip distance map when GI present");
-
-        // Multiple Generic Initiators - should skip FDT NUMA
-        let mut numa_nodes = BTreeMap::new();
-        numa_nodes.insert(0, create_test_numa_node(vec![0, 1], None));
-        numa_nodes.insert(1, create_test_numa_node(vec![], Some("vfio0".to_string())));
-        numa_nodes.insert(2, create_test_numa_node(vec![], Some("vfio1".to_string())));
-
-        let has_gi = numa_nodes.values().any(|node| node.device_id.is_some());
-        assert!(has_gi, "Should detect multiple Generic Initiators");
-    }
-
-    #[test]
-    fn test_fdt_distance_map() {
-        // Single NUMA node - should skip distance map
-        let mut numa_nodes = BTreeMap::new();
-        numa_nodes.insert(0, create_test_numa_node(vec![0, 1], None));
-
-        let mut fdt = FdtWriter::new().unwrap();
-        let result = create_distance_map_node(&mut fdt, &numa_nodes);
-        assert!(result.is_ok(), "Should skip distance map for single node");
-
-        // Empty NUMA nodes - should handle gracefully
-        let numa_nodes = BTreeMap::new();
-        let mut fdt = FdtWriter::new().unwrap();
-        let result = create_distance_map_node(&mut fdt, &numa_nodes);
-        assert!(result.is_ok(), "Should handle empty NUMA nodes");
-
-        // Non-contiguous NUMA IDs (0, 2, 5) with distance symmetry
-        let mut numa_nodes = BTreeMap::new();
-
-        let mut node0 = create_test_numa_node(vec![0], None);
-        node0.distances.insert(2, 20);
-        // node0 has no explicit distance to node5
-
-        let mut node2 = create_test_numa_node(vec![1], None);
-        node2.distances.insert(0, 20);
-        node2.distances.insert(5, 25);
-
-        let mut node5 = create_test_numa_node(vec![2], None);
-        node5.distances.insert(0, 30);
-        node5.distances.insert(2, 25);
-        // node5->node0 (should be used for node0->node5)
-
-        numa_nodes.insert(0, node0);
-        numa_nodes.insert(2, node2);
-        numa_nodes.insert(5, node5);
-
-        // Verify IDs are sorted lexicographically
-        let mut numa_ids: Vec<u32> = numa_nodes.keys().cloned().collect();
-        numa_ids.sort_unstable();
-        assert_eq!(numa_ids, vec![0, 2, 5]);
-
-        let mut fdt = FdtWriter::new().unwrap();
-        let result = create_distance_map_node(&mut fdt, &numa_nodes);
-        assert!(
-            result.is_ok(),
-            "Should handle non-contiguous IDs and symmetry"
-        );
-
-        // Default distance (20) when no distance specified in either direction
-        let mut numa_nodes = BTreeMap::new();
-        numa_nodes.insert(0, create_test_numa_node(vec![0], None));
-        numa_nodes.insert(1, create_test_numa_node(vec![1], None));
-        // Neither node has distance to the other
-
-        let mut fdt = FdtWriter::new().unwrap();
-        let result = create_distance_map_node(&mut fdt, &numa_nodes);
-        assert!(result.is_ok(), "Should default to 20 for missing distances");
-    }
-}

arch/src/aarch64/layout.rs

@@ -111,9 +111,8 @@ pub const RAM_64BIT_START: GuestAddress = GuestAddress(0x1_0000_0000);
 pub const CMDLINE_MAX_SIZE: usize = 2048;
 
 /// FDT is at the beginning of RAM.
+/// Maximum size of the device tree blob as specified in https://www.kernel.org/doc/Documentation/arm64/booting.txt.
 pub const FDT_START: GuestAddress = RAM_START;
-/// Maximum size of the device tree blob as specified in [the kernel
-/// documentation](https://www.kernel.org/doc/Documentation/arm64/booting.txt).
 pub const FDT_MAX_SIZE: u64 = 0x20_0000;
 
 /// Put ACPI table above dtb
@@ -138,12 +137,3 @@ pub const IRQ_BASE: u32 = 32;
 
 /// Number of supported interrupts
 pub const IRQ_NUM: u32 = 256;
-
-/// Base SPI interrupt number for the GICv2M MSI frame
-pub const GICV2M_SPI_BASE: u32 = 128;
-
-/// Total number of SPIs for the GICv2M MSI frame
-pub const GICV2M_SPI_NUM: u32 = 64;
-
-/// GICv2M compatible string
-pub const GIC_V2M_COMPATIBLE: &str = "arm,gic-v2m-frame";

arch/src/aarch64/mod.rs

@@ -6,56 +6,54 @@
 pub mod fdt;
 /// Layout for this aarch64 system.
 pub mod layout;
+/// Module for system registers definition
+pub mod regs;
 /// Module for loading UEFI binary.
 pub mod uefi;
 
-use std::collections::HashMap;
-use std::fmt::Debug;
-use std::sync::{Arc, Mutex};
-
-use hypervisor::arch::aarch64::gic::Vgic;
-use hypervisor::arch::aarch64::regs::MPIDR_EL1;
-use log::{Level, log_enabled};
-use thiserror::Error;
-use vm_memory::{Address, GuestAddress, GuestMemory, GuestMemoryAtomic};
-
 pub use self::fdt::DeviceInfoForFdt;
 use crate::{DeviceType, GuestMemoryMmap, NumaNodes, PciSpaceInfo, RegionType};
+use hypervisor::arch::aarch64::gic::Vgic;
+use log::{log_enabled, Level};
+use std::collections::HashMap;
+use std::convert::TryInto;
+use std::fmt::Debug;
+use std::sync::{Arc, Mutex};
+use vm_memory::{Address, GuestAddress, GuestMemory, GuestMemoryAtomic};
 
 pub const _NSIG: i32 = 65;
 
 /// Errors thrown while configuring aarch64 system.
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// Failed to create a FDT.
-    #[error("Failed to create a FDT")]
     SetupFdt,
 
     /// Failed to write FDT to memory.
-    #[error("Failed to write FDT to memory")]
-    WriteFdtToMemory(#[source] fdt::Error),
+    WriteFdtToMemory(fdt::Error),
 
     /// Failed to create a GIC.
-    #[error("Failed to create a GIC")]
     SetupGic,
 
     /// Failed to compute the initramfs address.
-    #[error("Failed to compute the initramfs address")]
     InitramfsAddress,
 
     /// Error configuring the general purpose registers
-    #[error("Error configuring the general purpose registers")]
-    RegsConfiguration(#[source] hypervisor::HypervisorCpuError),
+    RegsConfiguration(hypervisor::HypervisorCpuError),
 
     /// Error configuring the MPIDR register
-    #[error("Error configuring the MPIDR register")]
-    VcpuRegMpidr(#[source] hypervisor::HypervisorCpuError),
+    VcpuRegMpidr(hypervisor::HypervisorCpuError),
 
     /// Error initializing PMU for vcpu
-    #[error("Error initializing PMU for vcpu")]
     VcpuInitPmu,
 }
 
+impl From<Error> for super::Error {
+    fn from(e: Error) -> super::Error {
+        super::Error::PlatformSpecific(e)
+    }
+}
+
 #[derive(Debug, Copy, Clone)]
 /// Specifies the entry point address where the guest must start
 /// executing code.
@@ -66,8 +64,8 @@ pub struct EntryPoint {
 
 /// Configure the specified VCPU, and return its MPIDR.
 pub fn configure_vcpu(
-    vcpu: &dyn hypervisor::Vcpu,
-    id: u32,
+    vcpu: &Arc<dyn hypervisor::Vcpu>,
+    id: u8,
     boot_setup: Option<(EntryPoint, &GuestMemoryAtomic<GuestMemoryMmap>)>,
 ) -> super::Result<u64> {
     if let Some((kernel_entry_point, _guest_memory)) = boot_setup {
@@ -79,7 +77,9 @@ pub fn configure_vcpu(
         .map_err(Error::RegsConfiguration)?;
     }
 
-    let mpidr = vcpu.get_sys_reg(MPIDR_EL1).map_err(Error::VcpuRegMpidr)?;
+    let mpidr = vcpu
+        .get_sys_reg(regs::MPIDR_EL1)
+        .map_err(Error::VcpuRegMpidr)?;
     Ok(mpidr)
 }
 
@@ -125,8 +125,8 @@ pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
 pub fn configure_system<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
     guest_mem: &GuestMemoryMmap,
     cmdline: &str,
-    vcpu_mpidr: &[u64],
-    vcpu_topology: Option<(u16, u16, u16, u16)>,
+    vcpu_mpidr: Vec<u64>,
+    vcpu_topology: Option<(u8, u8, u8)>,
     device_info: &HashMap<(DeviceType, String), T, S>,
     initrd: &Option<super::InitramfsConfig>,
     pci_space_info: &[PciSpaceInfo],
@@ -154,7 +154,7 @@ pub fn configure_system<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::Bui
         fdt::print_fdt(&fdt_final);
     }
 
-    fdt::write_fdt_to_memory(&fdt_final, guest_mem).map_err(Error::WriteFdtToMemory)?;
+    fdt::write_fdt_to_memory(fdt_final, guest_mem).map_err(Error::WriteFdtToMemory)?;
 
     Ok(())
 }
@@ -180,8 +180,11 @@ pub fn initramfs_load_addr(
     }
 }
 
-pub fn get_host_cpu_phys_bits(hypervisor: &dyn hypervisor::Hypervisor) -> u8 {
-    let host_cpu_phys_bits = hypervisor.get_host_ipa_limit().try_into().unwrap();
+pub fn get_host_cpu_phys_bits() -> u8 {
+    // A dummy hypervisor created only for querying the host IPA size and will
+    // be freed after the query.
+    let hv = hypervisor::new().unwrap();
+    let host_cpu_phys_bits = hv.get_host_ipa_limit().try_into().unwrap();
     if host_cpu_phys_bits == 0 {
         // Host kernel does not support `get_host_ipa_limit`,
         // we return the default value 40 here.
@@ -192,7 +195,7 @@ pub fn get_host_cpu_phys_bits(hypervisor: &dyn hypervisor::Hypervisor) -> u8 {
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]

arch/src/aarch64/regs.rs

@@ -0,0 +1,43 @@
+// Copyright 2022 Arm Limited (or its affiliates). All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+// AArch64 system register encoding:
+// See https://developer.arm.com/documentation/ddi0487 (chapter D12)
+//
+//   31      22  21 20 19 18 16 15 12 11  8 7   5 4  0
+//  +----------+---+-----+-----+-----+-----+-----+----+
+//  |1101010100| L | op0 | op1 | CRn | CRm | op2 | Rt |
+//  +----------+---+-----+-----+-----+-----+-----+----+
+//
+// Notes:
+// - L and Rt are reserved as implementation defined fields, ignored.
+
+const SYSREG_HEAD: u32 = 0b1101010100u32 << 22;
+const SYSREG_OP0_SHIFT: u32 = 19;
+const SYSREG_OP0_MASK: u32 = 0b11u32 << 19;
+const SYSREG_OP1_SHIFT: u32 = 16;
+const SYSREG_OP1_MASK: u32 = 0b111u32 << 16;
+const SYSREG_CRN_SHIFT: u32 = 12;
+const SYSREG_CRN_MASK: u32 = 0b1111u32 << 12;
+const SYSREG_CRM_SHIFT: u32 = 8;
+const SYSREG_CRM_MASK: u32 = 0b1111u32 << 8;
+const SYSREG_OP2_SHIFT: u32 = 5;
+const SYSREG_OP2_MASK: u32 = 0b111u32 << 5;
+
+/// Define the ID of system registers
+#[macro_export]
+macro_rules! arm64_sys_reg {
+    ($name: tt, $op0: tt, $op1: tt, $crn: tt, $crm: tt, $op2: tt) => {
+        pub const $name: u32 = SYSREG_HEAD
+            | ((($op0 as u32) << SYSREG_OP0_SHIFT) & SYSREG_OP0_MASK as u32)
+            | ((($op1 as u32) << SYSREG_OP1_SHIFT) & SYSREG_OP1_MASK as u32)
+            | ((($crn as u32) << SYSREG_CRN_SHIFT) & SYSREG_CRN_MASK as u32)
+            | ((($crm as u32) << SYSREG_CRM_SHIFT) & SYSREG_CRM_MASK as u32)
+            | ((($op2 as u32) << SYSREG_OP2_SHIFT) & SYSREG_OP2_MASK as u32);
+    };
+}
+
+arm64_sys_reg!(MPIDR_EL1, 3, 0, 0, 0, 5);
+arm64_sys_reg!(ID_AA64MMFR0_EL1, 3, 0, 0, 7, 0);
+arm64_sys_reg!(TTBR1_EL1, 3, 0, 2, 0, 1);
+arm64_sys_reg!(TCR_EL1, 3, 0, 2, 0, 2);

arch/src/aarch64/uefi.rs

@@ -1,28 +1,19 @@
 // Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
 
 use std::io::{Read, Seek, SeekFrom};
-use std::os::fd::AsFd;
 use std::result;
-
-use thiserror::Error;
 use vm_memory::{Bytes, GuestAddress, GuestMemory};
 
 /// Errors thrown while loading UEFI binary
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// Unable to seek to UEFI image start.
-    #[error("Unable to seek to UEFI image start")]
     SeekUefiStart,
     /// Unable to seek to UEFI image end.
-    #[error("Unable to seek to UEFI image end")]
     SeekUefiEnd,
     /// UEFI image too big.
-    #[error("UEFI image too big")]
     UefiTooBig,
     /// Unable to read UEFI image
-    #[error("Unable to read UEFI image")]
     ReadUefiImage,
 }
 type Result<T> = result::Result<T, Error>;
@@ -33,7 +24,7 @@ pub fn load_uefi<F, M: GuestMemory>(
     uefi_image: &mut F,
 ) -> Result<()>
 where
-    F: Read + Seek + AsFd,
+    F: Read + Seek,
 {
     let uefi_size = uefi_image
         .seek(SeekFrom::End(0))
@@ -45,6 +36,6 @@ where
     }
     uefi_image.rewind().map_err(|_| Error::SeekUefiStart)?;
     guest_mem
-        .read_exact_volatile_from(guest_addr, &mut uefi_image.as_fd(), uefi_size)
+        .read_exact_from(guest_addr, uefi_image, uefi_size)
         .map_err(|_| Error::ReadUefiImage)
 }

arch/src/lib.rs

@@ -1,4 +1,3 @@
-// Copyright © 2024 Institute of Software, CAS. All rights reserved.
 // Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
 // Copyright © 2020, Oracle and/or its affiliates.
 //
@@ -6,14 +5,22 @@
 // SPDX-License-Identifier: Apache-2.0
 
 //! Implements platform specific functionality.
-//! Supported platforms: x86_64, aarch64, riscv64.
+//! Supported platforms: x86_64, aarch64.
 
-use std::collections::BTreeMap;
-use std::sync::Arc;
-use std::{fmt, result};
+#[macro_use]
+extern crate log;
 
+#[cfg(target_arch = "x86_64")]
+use crate::x86_64::SgxEpcSection;
 use serde::{Deserialize, Serialize};
+use std::collections::BTreeMap;
+use std::fmt;
+use std::result;
+use std::sync::Arc;
 use thiserror::Error;
+use versionize::{VersionMap, Versionize, VersionizeError, VersionizeResult};
+use versionize_derive::Versionize;
+use vm_migration::VersionMapped;
 
 type GuestMemoryMmap = vm_memory::GuestMemoryMmap<vm_memory::bitmap::AtomicBitmap>;
 type GuestRegionMmap = vm_memory::GuestRegionMmap<vm_memory::bitmap::AtomicBitmap>;
@@ -22,14 +29,11 @@ type GuestRegionMmap = vm_memory::GuestRegionMmap<vm_memory::bitmap::AtomicBitma
 #[derive(Debug, Error)]
 pub enum Error {
     #[cfg(target_arch = "x86_64")]
-    #[error("Platform specific error (x86_64)")]
-    PlatformSpecific(#[from] x86_64::Error),
+    #[error("Platform specific error (x86_64): {0:?}")]
+    PlatformSpecific(x86_64::Error),
     #[cfg(target_arch = "aarch64")]
-    #[error("Platform specific error (aarch64)")]
-    PlatformSpecific(#[from] aarch64::Error),
-    #[cfg(target_arch = "riscv64")]
-    #[error("Platform specific error (riscv64)")]
-    PlatformSpecific(#[from] riscv64::Error),
+    #[error("Platform specific error (aarch64): {0:?}")]
+    PlatformSpecific(aarch64::Error),
     #[error("The memory map table extends past the end of guest memory")]
     MemmapTablePastRamEnd,
     #[error("Error writing memory map table to guest memory")]
@@ -40,21 +44,17 @@ pub enum Error {
     StartInfoSetup,
     #[error("Failed to compute initramfs address")]
     InitramfsAddress,
-    #[error("Error writing module entry to guest memory")]
+    #[error("Error writing module entry to guest memory: {0}")]
     ModlistSetup(#[source] vm_memory::GuestMemoryError),
     #[error("RSDP extends past the end of guest memory")]
     RsdpPastRamEnd,
-    #[error("Failed to setup Zero Page for bzImage")]
-    ZeroPageSetup(#[source] vm_memory::GuestMemoryError),
-    #[error("Zero Page for bzImage past RAM end")]
-    ZeroPagePastRamEnd,
 }
 
 /// Type for returning public functions outcome.
 pub type Result<T> = result::Result<T, Error>;
 
 /// Type for memory region types.
-#[derive(Clone, Copy, PartialEq, Eq, Debug, Serialize, Deserialize)]
+#[derive(Clone, Copy, PartialEq, Eq, Debug, Serialize, Deserialize, Versionize)]
 pub enum RegionType {
     /// RAM type
     Ram,
@@ -72,26 +72,17 @@ pub enum RegionType {
     Reserved,
 }
 
+impl VersionMapped for RegionType {}
+
 /// Module for aarch64 related functionality.
 #[cfg(target_arch = "aarch64")]
 pub mod aarch64;
 
 #[cfg(target_arch = "aarch64")]
 pub use aarch64::{
-    _NSIG, EntryPoint, arch_memory_regions, configure_system, configure_vcpu,
-    fdt::DeviceInfoForFdt, get_host_cpu_phys_bits, initramfs_load_addr, layout,
-    layout::CMDLINE_MAX_SIZE, layout::IRQ_BASE, uefi,
-};
-
-/// Module for riscv64 related functionality.
-#[cfg(target_arch = "riscv64")]
-pub mod riscv64;
-
-#[cfg(target_arch = "riscv64")]
-pub use riscv64::{
-    _NSIG, EntryPoint, arch_memory_regions, configure_system, configure_vcpu,
-    fdt::DeviceInfoForFdt, get_host_cpu_phys_bits, initramfs_load_addr, layout,
-    layout::CMDLINE_MAX_SIZE, layout::IRQ_BASE, uefi,
+    arch_memory_regions, configure_system, configure_vcpu, fdt::DeviceInfoForFdt,
+    get_host_cpu_phys_bits, initramfs_load_addr, layout, layout::CMDLINE_MAX_SIZE,
+    layout::IRQ_BASE, uefi, EntryPoint, _NSIG,
 };
 
 #[cfg(target_arch = "x86_64")]
@@ -99,9 +90,9 @@ pub mod x86_64;
 
 #[cfg(target_arch = "x86_64")]
 pub use x86_64::{
-    _NSIG, CpuidConfig, CpuidFeatureEntry, EntryPoint, arch_memory_regions, configure_system,
-    configure_vcpu, generate_common_cpuid, generate_ram_ranges, get_host_cpu_phys_bits,
-    initramfs_load_addr, layout, layout::CMDLINE_MAX_SIZE, layout::CMDLINE_START, regs,
+    arch_memory_regions, configure_system, configure_vcpu, generate_common_cpuid,
+    get_host_cpu_phys_bits, initramfs_load_addr, layout, layout::CMDLINE_MAX_SIZE,
+    layout::CMDLINE_START, regs, CpuidFeatureEntry, EntryPoint, _NSIG,
 };
 
 /// Safe wrapper for `sysconf(_SC_PAGESIZE)`.
@@ -116,11 +107,11 @@ fn pagesize() -> usize {
 pub struct NumaNode {
     pub memory_regions: Vec<Arc<GuestRegionMmap>>,
     pub hotplug_regions: Vec<Arc<GuestRegionMmap>>,
-    pub cpus: Vec<u32>,
-    pub pci_segments: Vec<u16>,
+    pub cpus: Vec<u8>,
     pub distances: BTreeMap<u32, u8>,
     pub memory_zones: Vec<String>,
-    pub device_id: Option<String>,
+    #[cfg(target_arch = "x86_64")]
+    pub sgx_epc_sections: Vec<SgxEpcSection>,
 }
 
 pub type NumaNodes = BTreeMap<u32, NumaNode>;
@@ -139,7 +130,7 @@ pub enum DeviceType {
     /// Device Type: Virtio.
     Virtio(u32),
     /// Device Type: Serial.
-    #[cfg(any(target_arch = "aarch64", target_arch = "riscv64"))]
+    #[cfg(target_arch = "aarch64")]
     Serial,
     /// Device Type: RTC.
     #[cfg(target_arch = "aarch64")]
@@ -147,9 +138,6 @@ pub enum DeviceType {
     /// Device Type: GPIO.
     #[cfg(target_arch = "aarch64")]
     Gpio,
-    /// Device Type: fw_cfg.
-    #[cfg(feature = "fw_cfg")]
-    FwCfg,
 }
 
 /// Default (smallest) memory page size for the supported architectures.
@@ -163,7 +151,7 @@ impl fmt::Display for DeviceType {
 
 /// Structure to describe MMIO device information
 #[derive(Clone, Debug)]
-#[cfg(any(target_arch = "aarch64", target_arch = "riscv64"))]
+#[cfg(target_arch = "aarch64")]
 pub struct MmioDeviceInfo {
     pub addr: u64,
     pub len: u64,
@@ -172,7 +160,7 @@ pub struct MmioDeviceInfo {
 
 /// Structure to describe PCI space information
 #[derive(Clone, Debug)]
-#[cfg(any(target_arch = "aarch64", target_arch = "riscv64"))]
+#[cfg(target_arch = "aarch64")]
 pub struct PciSpaceInfo {
     pub pci_segment_id: u16,
     pub mmio_config_address: u64,
@@ -180,7 +168,7 @@ pub struct PciSpaceInfo {
     pub pci_device_space_size: u64,
 }
 
-#[cfg(any(target_arch = "aarch64", target_arch = "riscv64"))]
+#[cfg(target_arch = "aarch64")]
 impl DeviceInfoForFdt for MmioDeviceInfo {
     fn addr(&self) -> u64 {
         self.addr

arch/src/riscv64/fdt.rs

@@ -1,483 +0,0 @@
-// Copyright © 2024 Institute of Software, CAS. All rights reserved.
-// Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
-// Copyright 2019 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::collections::HashMap;
-use std::ffi::CStr;
-use std::fmt::Debug;
-use std::sync::{Arc, Mutex};
-use std::{cmp, result, str};
-
-use byteorder::{BigEndian, ByteOrder};
-use hypervisor::arch::riscv64::aia::Vaia;
-use log::debug;
-use thiserror::Error;
-use vm_fdt::{FdtWriter, FdtWriterResult};
-use vm_memory::{Address, Bytes, GuestMemory, GuestMemoryError, GuestMemoryRegion};
-
-use super::super::{DeviceType, GuestMemoryMmap, InitramfsConfig};
-use super::layout::{
-    IRQ_BASE, MEM_32BIT_DEVICES_SIZE, MEM_32BIT_DEVICES_START, MEM_PCI_IO_SIZE, MEM_PCI_IO_START,
-    PCI_HIGH_BASE, PCI_MMIO_CONFIG_SIZE_PER_SEGMENT,
-};
-use crate::PciSpaceInfo;
-
-const AIA_APLIC_PHANDLE: u32 = 1;
-const AIA_IMSIC_PHANDLE: u32 = 2;
-const CPU_INTC_BASE_PHANDLE: u32 = 3;
-const CPU_BASE_PHANDLE: u32 = 256 + CPU_INTC_BASE_PHANDLE;
-// Read the documentation specified when appending the root node to the FDT.
-const ADDRESS_CELLS: u32 = 0x2;
-const SIZE_CELLS: u32 = 0x2;
-
-// From https://elixir.bootlin.com/linux/v6.10/source/include/dt-bindings/interrupt-controller/irq.h#L14
-const _IRQ_TYPE_EDGE_RISING: u32 = 1;
-const IRQ_TYPE_LEVEL_HI: u32 = 4;
-
-const S_MODE_EXT_IRQ: u32 = 9;
-
-/// Trait for devices to be added to the Flattened Device Tree.
-pub trait DeviceInfoForFdt {
-    /// Returns the address where this device will be loaded.
-    fn addr(&self) -> u64;
-    /// Returns the associated interrupt for this device.
-    fn irq(&self) -> u32;
-    /// Returns the amount of memory that needs to be reserved for this device.
-    fn length(&self) -> u64;
-}
-
-/// Errors thrown while configuring the Flattened Device Tree for riscv64.
-#[derive(Debug, Error)]
-pub enum Error {
-    /// Failure in writing FDT in memory.
-    #[error("Failure in writing FDT in memory")]
-    WriteFdtToMemory(#[source] GuestMemoryError),
-}
-type Result<T> = result::Result<T, Error>;
-
-/// Creates the flattened device tree for this riscv64 VM.
-#[allow(clippy::too_many_arguments)]
-pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
-    guest_mem: &GuestMemoryMmap,
-    cmdline: &str,
-    num_vcpu: u32,
-    isa_string: &str,
-    device_info: &HashMap<(DeviceType, String), T, S>,
-    aia_device: &Arc<Mutex<dyn Vaia>>,
-    initrd: &Option<InitramfsConfig>,
-    pci_space_info: &[PciSpaceInfo],
-) -> FdtWriterResult<Vec<u8>> {
-    // Allocate stuff necessary for the holding the blob.
-    let mut fdt = FdtWriter::new()?;
-
-    // For an explanation why these nodes were introduced in the blob take a look at
-    // https://github.com/devicetree-org/devicetree-specification/releases/tag/v0.4
-    // In chapter 3.
-
-    // Header or the root node as per above mentioned documentation.
-    let root_node = fdt.begin_node("")?;
-    fdt.property_string("compatible", "linux,dummy-virt")?;
-    // For info on #address-cells and size-cells resort to Table 3.1 Root Node
-    // Properties
-    fdt.property_u32("#address-cells", ADDRESS_CELLS)?;
-    fdt.property_u32("#size-cells", SIZE_CELLS)?;
-    create_cpu_nodes(&mut fdt, num_vcpu, isa_string)?;
-    create_memory_node(&mut fdt, guest_mem)?;
-    create_chosen_node(&mut fdt, cmdline, initrd)?;
-    create_aia_node(&mut fdt, aia_device)?;
-    create_devices_node(&mut fdt, device_info)?;
-    create_pci_nodes(&mut fdt, pci_space_info)?;
-
-    // End Header node.
-    fdt.end_node(root_node)?;
-
-    let fdt_final = fdt.finish()?;
-
-    Ok(fdt_final)
-}
-
-pub fn write_fdt_to_memory(fdt_final: &[u8], guest_mem: &GuestMemoryMmap) -> Result<()> {
-    // Write FDT to memory.
-    guest_mem
-        .write_slice(fdt_final, super::layout::FDT_START)
-        .map_err(Error::WriteFdtToMemory)?;
-    Ok(())
-}
-
-// Following are the auxiliary function for creating the different nodes that we append to our FDT.
-fn create_cpu_nodes(fdt: &mut FdtWriter, num_cpus: u32, isa_string: &str) -> FdtWriterResult<()> {
-    // See https://elixir.bootlin.com/linux/v6.10/source/Documentation/devicetree/bindings/riscv/cpus.yaml
-    let cpus = fdt.begin_node("cpus")?;
-    // As per documentation, on RISC-V 64-bit systems value should be set to 1.
-    fdt.property_u32("#address-cells", 0x01)?;
-    fdt.property_u32("#size-cells", 0x0)?;
-    // TODO: Retrieve CPU frequency from cpu timer regs
-    let timebase_frequency: u32 = 0x989680;
-    fdt.property_u32("timebase-frequency", timebase_frequency)?;
-
-    for cpu_index in 0..num_cpus {
-        let cpu = fdt.begin_node(&format!("cpu@{cpu_index:x}"))?;
-        fdt.property_string("device_type", "cpu")?;
-        fdt.property_string("compatible", "riscv")?;
-        fdt.property_string("mmu-type", "sv48")?;
-        fdt.property_string("riscv,isa", isa_string)?;
-        fdt.property_string("status", "okay")?;
-        fdt.property_u32("reg", cpu_index)?;
-        fdt.property_u32("phandle", CPU_BASE_PHANDLE + cpu_index)?;
-
-        // interrupt controller node
-        let intc_node = fdt.begin_node("interrupt-controller")?;
-        fdt.property_string("compatible", "riscv,cpu-intc")?;
-        fdt.property_u32("#interrupt-cells", 1u32)?;
-        fdt.property_null("interrupt-controller")?;
-        fdt.property_u32("phandle", CPU_INTC_BASE_PHANDLE + cpu_index)?;
-        fdt.end_node(intc_node)?;
-
-        fdt.end_node(cpu)?;
-    }
-
-    fdt.end_node(cpus)?;
-
-    Ok(())
-}
-
-fn create_memory_node(fdt: &mut FdtWriter, guest_mem: &GuestMemoryMmap) -> FdtWriterResult<()> {
-    // Note: memory regions from "GuestMemory" are sorted and non-zero sized.
-    let ram_regions = {
-        let mut ram_regions = Vec::new();
-        let mut current_start = guest_mem
-            .iter()
-            .next()
-            .map(GuestMemoryRegion::start_addr)
-            .expect("GuestMemory must have one memory region at least")
-            .raw_value();
-        let mut current_end = current_start;
-
-        for (start, size) in guest_mem
-            .iter()
-            .map(|m| (m.start_addr().raw_value(), m.len()))
-        {
-            if current_end == start {
-                // This zone is continuous with the previous one.
-                current_end += size;
-            } else {
-                ram_regions.push((current_start, current_end));
-
-                current_start = start;
-                current_end = start + size;
-            }
-        }
-
-        ram_regions.push((current_start, current_end));
-
-        ram_regions
-    };
-
-    let mut mem_reg_property = Vec::new();
-    for region in ram_regions {
-        let mem_size = region.1 - region.0;
-        mem_reg_property.push(region.0);
-        mem_reg_property.push(mem_size);
-    }
-
-    let ram_start = super::layout::RAM_START.raw_value();
-    let memory_node_name = format!("memory@{ram_start:x}");
-    let memory_node = fdt.begin_node(&memory_node_name)?;
-    fdt.property_string("device_type", "memory")?;
-    fdt.property_array_u64("reg", &mem_reg_property)?;
-    fdt.end_node(memory_node)?;
-
-    Ok(())
-}
-
-fn create_chosen_node(
-    fdt: &mut FdtWriter,
-    cmdline: &str,
-    initrd: &Option<InitramfsConfig>,
-) -> FdtWriterResult<()> {
-    let chosen_node = fdt.begin_node("chosen")?;
-    fdt.property_string("bootargs", cmdline)?;
-
-    if let Some(initrd_config) = initrd {
-        let initrd_start = initrd_config.address.raw_value();
-        let initrd_end = initrd_config.address.raw_value() + initrd_config.size as u64;
-        fdt.property_u64("linux,initrd-start", initrd_start)?;
-        fdt.property_u64("linux,initrd-end", initrd_end)?;
-    }
-
-    fdt.end_node(chosen_node)?;
-
-    Ok(())
-}
-
-fn create_aia_node(fdt: &mut FdtWriter, aia_device: &Arc<Mutex<dyn Vaia>>) -> FdtWriterResult<()> {
-    // IMSIC
-    if aia_device.lock().unwrap().msi_compatible() {
-        use super::layout::IMSIC_START;
-        let imsic_name = format!("imsics@{:x}", IMSIC_START.0);
-        let imsic_node = fdt.begin_node(&imsic_name)?;
-
-        fdt.property_string(
-            "compatible",
-            aia_device.lock().unwrap().imsic_compatibility(),
-        )?;
-        let imsic_reg_prop = aia_device.lock().unwrap().imsic_properties();
-        fdt.property_array_u32("reg", &imsic_reg_prop)?;
-        fdt.property_u32("#interrupt-cells", 0u32)?;
-        fdt.property_null("interrupt-controller")?;
-        fdt.property_null("msi-controller")?;
-        // TODO complete num-ids
-        fdt.property_u32("riscv,num-ids", 2047u32)?;
-        fdt.property_u32("phandle", AIA_IMSIC_PHANDLE)?;
-
-        let mut irq_cells = Vec::new();
-        let num_cpus = aia_device.lock().unwrap().vcpu_count();
-        for i in 0..num_cpus {
-            irq_cells.push(CPU_INTC_BASE_PHANDLE + i);
-            irq_cells.push(S_MODE_EXT_IRQ);
-        }
-        fdt.property_array_u32("interrupts-extended", &irq_cells)?;
-
-        fdt.end_node(imsic_node)?;
-    }
-
-    // APLIC
-    use super::layout::APLIC_START;
-    let aplic_name = format!("aplic@{:x}", APLIC_START.0);
-    let aplic_node = fdt.begin_node(&aplic_name)?;
-
-    fdt.property_string(
-        "compatible",
-        aia_device.lock().unwrap().aplic_compatibility(),
-    )?;
-    let reg_cells = aia_device.lock().unwrap().aplic_properties();
-    fdt.property_array_u32("reg", &reg_cells)?;
-    fdt.property_u32("#interrupt-cells", 2u32)?;
-    fdt.property_null("interrupt-controller")?;
-    // TODO complete num-srcs
-    fdt.property_u32("riscv,num-sources", 96u32)?;
-    fdt.property_u32("phandle", AIA_APLIC_PHANDLE)?;
-    fdt.property_u32("msi-parent", AIA_IMSIC_PHANDLE)?;
-
-    fdt.end_node(aplic_node)?;
-
-    Ok(())
-}
-
-fn create_serial_node<T: DeviceInfoForFdt + Clone + Debug>(
-    fdt: &mut FdtWriter,
-    dev_info: &T,
-) -> FdtWriterResult<()> {
-    let serial_reg_prop = [dev_info.addr(), dev_info.length()];
-    let irq = [dev_info.irq() - IRQ_BASE, IRQ_TYPE_LEVEL_HI];
-
-    let serial_node = fdt.begin_node(&format!("serial@{:x}", dev_info.addr()))?;
-    fdt.property_string("compatible", "ns16550a")?;
-    fdt.property_array_u64("reg", &serial_reg_prop)?;
-    fdt.property_u32("clock-frequency", 3686400)?;
-    fdt.property_u32("interrupt-parent", AIA_APLIC_PHANDLE)?;
-    fdt.property_array_u32("interrupts", &irq)?;
-    fdt.end_node(serial_node)?;
-
-    Ok(())
-}
-
-fn create_devices_node<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
-    fdt: &mut FdtWriter,
-    dev_info: &HashMap<(DeviceType, String), T, S>,
-) -> FdtWriterResult<()> {
-    for ((device_type, _device_id), info) in dev_info {
-        match device_type {
-            DeviceType::Serial => create_serial_node(fdt, info)?,
-            DeviceType::Virtio(_) => unreachable!(),
-        }
-    }
-
-    Ok(())
-}
-
-fn create_pci_nodes(fdt: &mut FdtWriter, pci_device_info: &[PciSpaceInfo]) -> FdtWriterResult<()> {
-    // Add node for PCIe controller.
-    // See Documentation/devicetree/bindings/pci/host-generic-pci.txt in the kernel
-    // and https://elinux.org/Device_Tree_Usage.
-    // In multiple PCI segments setup, each PCI segment needs a PCI node.
-    for pci_device_info_elem in pci_device_info.iter() {
-        // EDK2 requires the PCIe high space above 4G address.
-        // The actual space in CLH follows the RAM. If the RAM space is small, the PCIe high space
-        // could fall below 4G.
-        // Here we cut off PCI device space below 8G in FDT to workaround the EDK2 check.
-        // But the address written in ACPI is not impacted.
-        let (pci_device_base_64bit, pci_device_size_64bit) =
-            if pci_device_info_elem.pci_device_space_start < PCI_HIGH_BASE.raw_value() {
-                (
-                    PCI_HIGH_BASE.raw_value(),
-                    pci_device_info_elem.pci_device_space_size
-                        - (PCI_HIGH_BASE.raw_value() - pci_device_info_elem.pci_device_space_start),
-                )
-            } else {
-                (
-                    pci_device_info_elem.pci_device_space_start,
-                    pci_device_info_elem.pci_device_space_size,
-                )
-            };
-        // There is no specific requirement of the 32bit MMIO range, and
-        // therefore at least we can make these ranges 4K aligned.
-        let pci_device_size_32bit: u64 =
-            MEM_32BIT_DEVICES_SIZE / ((1 << 12) * pci_device_info.len() as u64) * (1 << 12);
-        let pci_device_base_32bit: u64 = MEM_32BIT_DEVICES_START.0
-            + pci_device_size_32bit * pci_device_info_elem.pci_segment_id as u64;
-
-        let ranges = [
-            // io addresses. Since AArch64 will not use IO address,
-            // we can set the same IO address range for every segment.
-            0x1000000,
-            0_u32,
-            0_u32,
-            (MEM_PCI_IO_START.0 >> 32) as u32,
-            MEM_PCI_IO_START.0 as u32,
-            (MEM_PCI_IO_SIZE >> 32) as u32,
-            MEM_PCI_IO_SIZE as u32,
-            // mmio addresses
-            0x2000000,                            // (ss = 10: 32-bit memory space)
-            (pci_device_base_32bit >> 32) as u32, // PCI address
-            pci_device_base_32bit as u32,
-            (pci_device_base_32bit >> 32) as u32, // CPU address
-            pci_device_base_32bit as u32,
-            (pci_device_size_32bit >> 32) as u32, // size
-            pci_device_size_32bit as u32,
-            // device addresses
-            0x3000000,                            // (ss = 11: 64-bit memory space)
-            (pci_device_base_64bit >> 32) as u32, // PCI address
-            pci_device_base_64bit as u32,
-            (pci_device_base_64bit >> 32) as u32, // CPU address
-            pci_device_base_64bit as u32,
-            (pci_device_size_64bit >> 32) as u32, // size
-            pci_device_size_64bit as u32,
-        ];
-        let bus_range = [0, 0]; // Only bus 0
-        let reg = [
-            pci_device_info_elem.mmio_config_address,
-            PCI_MMIO_CONFIG_SIZE_PER_SEGMENT,
-        ];
-        // See kernel document Documentation/devicetree/bindings/pci/pci-msi.txt
-        let msi_map = [
-            // rid-base: A single cell describing the first RID matched by the entry.
-            0x0,
-            // msi-controller: A single phandle to an MSI controller.
-            AIA_IMSIC_PHANDLE,
-            // msi-base: An msi-specifier describing the msi-specifier produced for the
-            // first RID matched by the entry.
-            (pci_device_info_elem.pci_segment_id as u32) << 8,
-            // length: A single cell describing how many consecutive RIDs are matched
-            // following the rid-base.
-            0x100,
-        ];
-
-        let pci_node_name = format!("pci@{:x}", pci_device_info_elem.mmio_config_address);
-        let pci_node = fdt.begin_node(&pci_node_name)?;
-
-        fdt.property_string("compatible", "pci-host-ecam-generic")?;
-        fdt.property_string("device_type", "pci")?;
-        fdt.property_array_u32("ranges", &ranges)?;
-        fdt.property_array_u32("bus-range", &bus_range)?;
-        fdt.property_u32(
-            "linux,pci-domain",
-            pci_device_info_elem.pci_segment_id as u32,
-        )?;
-        fdt.property_u32("#address-cells", 3)?;
-        fdt.property_u32("#size-cells", 2)?;
-        fdt.property_array_u64("reg", &reg)?;
-        fdt.property_u32("#interrupt-cells", 1)?;
-        fdt.property_null("interrupt-map")?;
-        fdt.property_null("interrupt-map-mask")?;
-        fdt.property_null("dma-coherent")?;
-        fdt.property_array_u32("msi-map", &msi_map)?;
-        fdt.property_u32("msi-parent", AIA_IMSIC_PHANDLE)?;
-
-        fdt.end_node(pci_node)?;
-    }
-
-    Ok(())
-}
-
-// Parse the DTB binary and print for debugging
-pub fn print_fdt(dtb: &[u8]) {
-    match fdt_parser::Fdt::new(dtb) {
-        Ok(fdt) => {
-            if let Some(root) = fdt.find_node("/") {
-                debug!("Printing the FDT:");
-                print_node(root, 0);
-            } else {
-                debug!("Failed to find root node in FDT for debugging.");
-            }
-        }
-        Err(_) => debug!("Failed to parse FDT for debugging."),
-    }
-}
-
-fn print_node(node: fdt_parser::node::FdtNode<'_, '_>, n_spaces: usize) {
-    debug!("{:indent$}{}/", "", node.name, indent = n_spaces);
-    for property in node.properties() {
-        let name = property.name;
-
-        // If the property is 'compatible', its value requires special handling.
-        // The u8 array could contain multiple null-terminated strings.
-        // We copy the original array and simply replace all 'null' characters with spaces.
-        let value = if name == "compatible" {
-            let mut compatible = vec![0u8; 256];
-            let handled_value = property
-                .value
-                .iter()
-                .map(|&c| if c == 0 { b' ' } else { c })
-                .collect::<Vec<_>>();
-            let len = cmp::min(255, handled_value.len());
-            compatible[..len].copy_from_slice(&handled_value[..len]);
-            compatible[..(len + 1)].to_vec()
-        } else {
-            property.value.to_vec()
-        };
-        let value = &value;
-
-        // Now the value can be either:
-        //   - A null-terminated C string, or
-        //   - Binary data
-        // We follow a very simple logic to present the value:
-        //   - At first, try to convert it to CStr and print,
-        //   - If failed, print it as u32 array.
-        let value_result = match CStr::from_bytes_with_nul(value) {
-            Ok(value_cstr) => value_cstr.to_str().ok(),
-            Err(_e) => None,
-        };
-
-        if let Some(value_str) = value_result {
-            debug!(
-                "{:indent$}{} : {:#?}",
-                "",
-                name,
-                value_str,
-                indent = (n_spaces + 2)
-            );
-        } else {
-            let mut array = Vec::with_capacity(256);
-            array.resize(value.len() / 4, 0u32);
-            BigEndian::read_u32_into(value, &mut array);
-            debug!(
-                "{:indent$}{} : {:X?}",
-                "",
-                name,
-                array,
-                indent = (n_spaces + 2)
-            );
-        }
-    }
-
-    // Print children nodes if there is any
-    for child in node.children() {
-        print_node(child, n_spaces + 2);
-    }
-}

arch/src/riscv64/layout.rs

@@ -1,117 +0,0 @@
-// Copyright © 2024 Institute of Software, CAS. All rights reserved.
-// Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
-// Copyright 2019 Amazon.com, Inc. or its affiliates. All Rights Reserved.
-// SPDX-License-Identifier: Apache-2.0
-
-//
-// Memory layout of RISC-V 64-bit guest:
-//
-// Physical  +---------------------------------------------------------------+
-// address   |                                                               |
-// end       |                                                               |
-//           ~                   ~                       ~                   ~
-//           |                                                               |
-//           |                    Highmem PCI MMIO space                     |
-//           |                                                               |
-// RAM end   +---------------------------------------------------------------+
-// (dynamic, |                                                               |
-// including |                                                               |
-// hotplug   ~                   ~                       ~                   ~
-// memory)   |                                                               |
-//           |                             DRAM                              |
-//           |                                                               |
-//           |                                                               |
-//           |                                                               |
-//           |                                                               |
-//    1 GB   +---------------------------------------------------------------+
-//           |                                                               |
-//           |                      PCI MMCONFIG space                       |
-//           |                                                               |
-//  768 MB   +---------------------------------------------------------------+
-//           |                                                               |
-//           |                                                               |
-//           |                        PCI MMIO space                         |
-//           |                                                               |
-//  256 MB   +---------------------------------------------------------------|
-//           |                                                               |
-//           |                     Legacy devices space                      |
-//           |                                                               |
-//  128 MB   +---------------------------------------------------------------|
-//           |                                                               |
-//           |                            IMSICs                             |
-//           |                                                               |
-//   64 MB   +---------------------------------------------------------------+
-//           |                                                               |
-//           |                            APLICs                             |
-//           |                                                               |
-//    4 MB   +---------------------------------------------------------------+
-//           |                          UEFI flash                           |
-//    0 GB   +---------------------------------------------------------------+
-//
-//
-
-use vm_memory::GuestAddress;
-
-/// 0x0 ~ 0x40_0000 (4 MiB) is reserved to UEFI
-/// UEFI binary size is required less than 3 MiB, reserving 4 MiB is enough.
-pub const UEFI_START: GuestAddress = GuestAddress(0);
-pub const UEFI_SIZE: u64 = 0x040_0000;
-
-/// AIA related devices
-/// See https://elixir.bootlin.com/linux/v6.10/source/arch/riscv/include/uapi/asm/kvm.h
-/// 0x40_0000 ~ 0x0400_0000 (64 MiB) resides APLICs
-pub const APLIC_START: GuestAddress = GuestAddress(0x40_0000);
-pub const APLIC_SIZE: u64 = 0x4000;
-
-/// 0x0400_0000 ~ 0x0800_0000 (64 MiB) resides IMSICs
-pub const IMSIC_START: GuestAddress = GuestAddress(0x0400_0000);
-pub const IMSIC_SIZE: u64 = 0x1000;
-
-/// Below this address will reside the AIA, above this address will reside the MMIO devices.
-const MAPPED_IO_START: GuestAddress = GuestAddress(0x0800_0000);
-
-/// Space 0x0800_0000 ~ 0x1000_0000 is reserved for legacy devices.
-pub const LEGACY_SERIAL_MAPPED_IO_START: GuestAddress = MAPPED_IO_START;
-
-/// Space 0x0905_0000 ~ 0x0906_0000 is reserved for pcie io address
-pub const MEM_PCI_IO_START: GuestAddress = GuestAddress(0x0905_0000);
-pub const MEM_PCI_IO_SIZE: u64 = 0x1_0000;
-
-/// Starting from 0x1000_0000 (256MiB) to 0x3000_0000 (768MiB) is used for PCIE MMIO
-pub const MEM_32BIT_DEVICES_START: GuestAddress = GuestAddress(0x1000_0000);
-pub const MEM_32BIT_DEVICES_SIZE: u64 = 0x2000_0000;
-
-/// PCI MMCONFIG space (start: after the device space at 768MiB, length: 256MiB)
-pub const PCI_MMCONFIG_START: GuestAddress = GuestAddress(0x3000_0000);
-pub const PCI_MMCONFIG_SIZE: u64 = 256 << 20;
-// One bus with potentially 256 devices (32 slots x 8 functions).
-pub const PCI_MMIO_CONFIG_SIZE_PER_SEGMENT: u64 = 4096 * 256;
-
-/// Start of RAM.
-pub const RAM_START: GuestAddress = GuestAddress(0x4000_0000);
-
-/// Kernel command line maximum size on RISC-V.
-/// See https://elixir.bootlin.com/linux/v6.10/source/arch/riscv/include/uapi/asm/setup.h
-pub const CMDLINE_MAX_SIZE: usize = 1024;
-
-/// FDT is at the beginning of RAM.
-pub const FDT_START: GuestAddress = RAM_START;
-pub const FDT_MAX_SIZE: u64 = 0x1_0000;
-
-/// Put ACPI table above dtb
-pub const ACPI_START: GuestAddress = GuestAddress(RAM_START.0 + FDT_MAX_SIZE);
-pub const ACPI_MAX_SIZE: u64 = 0x20_0000;
-pub const RSDP_POINTER: GuestAddress = ACPI_START;
-
-/// Kernel start after FDT and ACPI
-pub const KERNEL_START: GuestAddress = GuestAddress(RAM_START.0 + FDT_MAX_SIZE);
-
-/// Pci high memory base
-pub const PCI_HIGH_BASE: GuestAddress = GuestAddress(0x2_0000_0000);
-
-/// First usable interrupt on riscv64
-pub const IRQ_BASE: u32 = 0;
-
-// As per https://elixir.bootlin.com/linux/v6.10/source/arch/riscv/include/asm/kvm_host.h#L31
-/// Number of supported interrupts
-pub const IRQ_NUM: u32 = 1023;

arch/src/riscv64/mod.rs

@@ -1,231 +0,0 @@
-// Copyright © 2024 Institute of Software, CAS. All rights reserved.
-// Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
-// Copyright 2019 Amazon.com, Inc. or its affiliates. All Rights Reserved.
-// SPDX-License-Identifier: Apache-2.0
-
-/// Module for the flattened device tree.
-pub mod fdt;
-/// Layout for this riscv64 system.
-pub mod layout;
-/// Module for loading UEFI binary.
-pub mod uefi;
-
-use std::collections::HashMap;
-use std::fmt::Debug;
-use std::fs::File;
-use std::io::{BufRead, BufReader};
-use std::sync::{Arc, Mutex};
-
-use hypervisor::arch::riscv64::aia::Vaia;
-use log::{Level, log_enabled};
-use thiserror::Error;
-use vm_memory::{Address, GuestAddress, GuestMemory, GuestMemoryAtomic};
-
-pub use self::fdt::DeviceInfoForFdt;
-use crate::{DeviceType, GuestMemoryMmap, PciSpaceInfo, RegionType};
-
-pub const CLOUDHV_IRQCHIP_NUM_MSIS: u16 = 255;
-pub const CLOUDHV_IRQCHIP_NUM_SOURCES: u8 = 96;
-pub const CLOUDHV_IRQCHIP_NUM_PRIO_BITS: u8 = 3;
-pub const CLOUDHV_IRQCHIP_MAX_GUESTS_BITS: u8 = 3;
-pub const CLOUDHV_IRQCHIP_MAX_GUESTS: u8 = (1 << CLOUDHV_IRQCHIP_MAX_GUESTS_BITS) - 1;
-pub const _NSIG: i32 = 65;
-
-/// Errors thrown while configuring riscv64 system.
-#[derive(Debug, Error)]
-pub enum Error {
-    /// Failed to create a FDT.
-    #[error("Failed to create a FDT")]
-    SetupFdt,
-
-    /// Failed to write FDT to memory.
-    #[error("Failed to write FDT to memory")]
-    WriteFdtToMemory(#[source] fdt::Error),
-
-    /// Failed to create a AIA.
-    #[error("Failed to create a AIA")]
-    SetupAia,
-
-    /// Failed to compute the initramfs address.
-    #[error("Failed to compute the initramfs address")]
-    InitramfsAddress,
-
-    /// Error configuring the general purpose registers
-    #[error("Error configuring the general purpose registers")]
-    RegsConfiguration(#[source] hypervisor::HypervisorCpuError),
-
-    /// Error opening /proc/cpuinfo
-    #[error("Error opening /proc/cpuinfo")]
-    OpenCpuInfo(#[source] std::io::Error),
-
-    /// Error reading /proc/cpuinfo
-    #[error("Error reading /proc/cpuinfo")]
-    ReadCpuInfo(#[source] std::io::Error),
-
-    /// Invalid ISA string
-    #[error("Invalid ISA string: {0}")]
-    InvalidIsaString(String),
-
-    /// Error parsing /proc/cpuinfo
-    #[error("Error parsing /proc/cpuinfo")]
-    CpuInfoParsing,
-}
-
-#[derive(Debug, Copy, Clone)]
-/// Specifies the entry point address where the guest must start
-/// executing code.
-pub struct EntryPoint {
-    /// Address in guest memory where the guest must start execution
-    pub entry_addr: GuestAddress,
-}
-
-/// Configure the specified VCPU, and return its MPIDR.
-pub fn configure_vcpu(
-    vcpu: &dyn hypervisor::Vcpu,
-    id: u32,
-    boot_setup: Option<(EntryPoint, &GuestMemoryAtomic<GuestMemoryMmap>)>,
-) -> super::Result<()> {
-    if let Some((kernel_entry_point, _guest_memory)) = boot_setup {
-        vcpu.setup_regs(
-            id,
-            kernel_entry_point.entry_addr.raw_value(),
-            layout::FDT_START.raw_value(),
-        )
-        .map_err(Error::RegsConfiguration)?;
-    }
-
-    Ok(())
-}
-
-pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
-    vec![
-        // 0 MiB ~ 256 MiB: AIA and legacy devices
-        (
-            GuestAddress(0),
-            layout::MEM_32BIT_DEVICES_START.0 as usize,
-            RegionType::Reserved,
-        ),
-        // 256 MiB ~ 768 MiB: MMIO space
-        (
-            layout::MEM_32BIT_DEVICES_START,
-            layout::MEM_32BIT_DEVICES_SIZE as usize,
-            RegionType::SubRegion,
-        ),
-        // 768 MiB ~ 1 GiB: reserved. The leading 256M for PCIe MMCONFIG space
-        (
-            layout::PCI_MMCONFIG_START,
-            layout::PCI_MMCONFIG_SIZE as usize,
-            RegionType::Reserved,
-        ),
-        // 1GiB ~ inf: RAM
-        (layout::RAM_START, usize::MAX, RegionType::Ram),
-    ]
-}
-
-// Read the first "isa" string from /proc/cpuinfo and filter out the H extension,
-// while correctly preserving multi-letter extensions.
-fn isa_string_from_host() -> Result<String, Error> {
-    let file = File::open("/proc/cpuinfo").map_err(Error::OpenCpuInfo)?;
-    let reader = BufReader::new(file);
-
-    for line in reader.lines() {
-        let line = line.map_err(Error::ReadCpuInfo)?;
-        let trimmed_line = line.trim();
-
-        if trimmed_line.starts_with("isa") {
-            let parts: Vec<&str> = trimmed_line.split(':').collect();
-            if parts.len() == 2 {
-                let isa_string = parts[1].trim();
-
-                // Split the string by underscores to separate single letter vs long-form
-                // extensions
-                let mut components: Vec<String> =
-                    isa_string.split('_').map(|s| s.to_string()).collect();
-
-                if components.is_empty() {
-                    return Err(Error::InvalidIsaString(isa_string.to_string()));
-                }
-
-                // Remove H extension if present in single letter extensions
-                let first_component = components[0].chars().filter(|&c| c != 'h').collect();
-
-                components[0] = first_component;
-
-                return Ok(components.join("_"));
-            }
-        }
-    }
-
-    Err(Error::CpuInfoParsing)
-}
-
-/// Configures the system and should be called once per vm before starting vcpu threads.
-#[allow(clippy::too_many_arguments)]
-pub fn configure_system<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
-    guest_mem: &GuestMemoryMmap,
-    cmdline: &str,
-    num_vcpu: u32,
-    device_info: &HashMap<(DeviceType, String), T, S>,
-    initrd: &Option<super::InitramfsConfig>,
-    pci_space_info: &[PciSpaceInfo],
-    aia_device: &Arc<Mutex<dyn Vaia>>,
-) -> super::Result<()> {
-    let isa_string = isa_string_from_host()?;
-    let fdt_final = fdt::create_fdt(
-        guest_mem,
-        cmdline,
-        num_vcpu,
-        &isa_string,
-        device_info,
-        aia_device,
-        initrd,
-        pci_space_info,
-    )
-    .map_err(|_| Error::SetupFdt)?;
-
-    if log_enabled!(Level::Debug) {
-        fdt::print_fdt(&fdt_final);
-    }
-
-    fdt::write_fdt_to_memory(&fdt_final, guest_mem).map_err(Error::WriteFdtToMemory)?;
-
-    Ok(())
-}
-
-/// Returns the memory address where the initramfs could be loaded.
-pub fn initramfs_load_addr(
-    guest_mem: &GuestMemoryMmap,
-    initramfs_size: usize,
-) -> super::Result<u64> {
-    let round_to_pagesize = |size| (size + (super::PAGE_SIZE - 1)) & !(super::PAGE_SIZE - 1);
-    match guest_mem
-        .last_addr()
-        .checked_sub(round_to_pagesize(initramfs_size) as u64 - 1)
-    {
-        Some(offset) => {
-            if guest_mem.address_in_range(offset) {
-                Ok(offset.raw_value())
-            } else {
-                Err(super::Error::PlatformSpecific(Error::InitramfsAddress))
-            }
-        }
-        None => Err(super::Error::PlatformSpecific(Error::InitramfsAddress)),
-    }
-}
-
-pub fn get_host_cpu_phys_bits(_hypervisor: &dyn hypervisor::Hypervisor) -> u8 {
-    40
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use super::*;
-
-    #[test]
-    fn test_arch_memory_regions_dram() {
-        let regions = arch_memory_regions();
-        assert_eq!(4, regions.len());
-        assert_eq!(layout::RAM_START, regions[3].0);
-        assert_eq!(RegionType::Ram, regions[3].2);
-    }
-}

arch/src/riscv64/uefi.rs

@@ -1,50 +0,0 @@
-// Copyright 2020 Arm Limited (or its affiliates). All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::io::{Read, Seek, SeekFrom};
-use std::os::fd::AsFd;
-use std::result;
-
-use thiserror::Error;
-use vm_memory::{GuestAddress, GuestMemory};
-
-/// Errors thrown while loading UEFI binary
-#[derive(Debug, Error)]
-pub enum Error {
-    /// Unable to seek to UEFI image start.
-    #[error("Unable to seek to UEFI image start")]
-    SeekUefiStart,
-    /// Unable to seek to UEFI image end.
-    #[error("Unable to seek to UEFI image end")]
-    SeekUefiEnd,
-    /// UEFI image too big.
-    #[error("UEFI image too big")]
-    UefiTooBig,
-    /// Unable to read UEFI image
-    #[error("Unable to read UEFI image")]
-    ReadUefiImage,
-}
-type Result<T> = result::Result<T, Error>;
-
-pub fn load_uefi<F, M: GuestMemory>(
-    guest_mem: &M,
-    guest_addr: GuestAddress,
-    uefi_image: &mut F,
-) -> Result<()>
-where
-    F: Read + Seek + AsFd,
-{
-    let uefi_size = uefi_image
-        .seek(SeekFrom::End(0))
-        .map_err(|_| Error::SeekUefiEnd)? as usize;
-
-    // edk2 image on virtual platform is smaller than 3M
-    if uefi_size > 0x300000 {
-        return Err(Error::UefiTooBig);
-    }
-    uefi_image.rewind().map_err(|_| Error::SeekUefiStart)?;
-    guest_mem
-        .read_exact_volatile_from(guest_addr, &mut uefi_image.as_fd(), uefi_size)
-        .map_err(|_| Error::ReadUefiImage)
-}

arch/src/x86_64/interrupts.rs

@@ -6,6 +6,7 @@
 // found in the LICENSE-BSD-3-Clause file.
 
 use std::result;
+use std::sync::Arc;
 
 pub type Result<T> = result::Result<T, hypervisor::HypervisorCpuError>;
 
@@ -23,7 +24,7 @@ pub fn set_apic_delivery_mode(reg: u32, mode: u32) -> u32 {
 ///
 /// # Arguments
 /// * `vcpu` - The VCPU object to configure.
-pub fn set_lint(vcpu: &dyn hypervisor::Vcpu) -> Result<()> {
+pub fn set_lint(vcpu: &Arc<dyn hypervisor::Vcpu>) -> Result<()> {
     let mut klapic = vcpu.get_lapic()?;
 
     let lvt_lint0 = klapic.get_klapic_reg(APIC_LVT0);

arch/src/x86_64/mpspec.rs

@@ -1,10 +1,6 @@
 // Copyright 2017 The Chromium OS Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE-BSD-3-Clause file.
-#![allow(non_camel_case_types)]
-use vm_memory::ByteValued;
 
 pub const MP_PROCESSOR: ::std::os::raw::c_uint = 0;
 pub const MP_BUS: ::std::os::raw::c_uint = 1;
@@ -19,7 +15,7 @@ pub const MP_IRQDIR_DEFAULT: ::std::os::raw::c_uint = 0;
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpf_intel {
-    pub signature: [::std::os::raw::c_uchar; 4usize],
+    pub signature: [::std::os::raw::c_char; 4usize],
     pub physptr: ::std::os::raw::c_uint,
     pub length: ::std::os::raw::c_uchar,
     pub specification: ::std::os::raw::c_uchar,
@@ -31,22 +27,15 @@ pub struct mpf_intel {
     pub feature5: ::std::os::raw::c_uchar,
 }
 
-const _: () = assert!(::core::mem::size_of::<mpf_intel>() == 16);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpf_intel {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_table {
-    pub signature: [::std::os::raw::c_uchar; 4usize],
+    pub signature: [::std::os::raw::c_char; 4usize],
     pub length: ::std::os::raw::c_ushort,
-    pub spec: ::std::os::raw::c_uchar,
-    pub checksum: ::std::os::raw::c_uchar,
-    pub oem: [::std::os::raw::c_uchar; 8usize],
-    pub productid: [::std::os::raw::c_uchar; 12usize],
+    pub spec: ::std::os::raw::c_char,
+    pub checksum: ::std::os::raw::c_char,
+    pub oem: [::std::os::raw::c_char; 8usize],
+    pub productid: [::std::os::raw::c_char; 12usize],
     pub oemptr: ::std::os::raw::c_uint,
     pub oemsize: ::std::os::raw::c_ushort,
     pub oemcount: ::std::os::raw::c_ushort,
@@ -54,19 +43,6 @@ pub struct mpc_table {
     pub reserved: ::std::os::raw::c_uint,
 }
 
-const _: () = {
-    assert!(::core::mem::size_of::<mpc_table>() == 4 + 2 + 1 + 1 + 8 + 12 + 4 + 2 + 2 + 4 + 4);
-    assert!(::core::mem::size_of::<::std::os::raw::c_uint>() == 4);
-    assert!(::core::mem::size_of::<::std::os::raw::c_ushort>() == 2);
-    assert!(::core::mem::size_of::<::std::os::raw::c_uchar>() == 1);
-};
-
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_table {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_cpu {
@@ -79,13 +55,6 @@ pub struct mpc_cpu {
     pub reserved: [::std::os::raw::c_uint; 2usize],
 }
 
-const _: () = assert!(::core::mem::size_of::<mpc_cpu>() == 20);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_cpu {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_bus {
@@ -94,13 +63,6 @@ pub struct mpc_bus {
     pub bustype: [::std::os::raw::c_uchar; 6usize],
 }
 
-const _: () = assert!(::core::mem::size_of::<mpc_bus>() == 8);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_bus {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_ioapic {
@@ -111,13 +73,6 @@ pub struct mpc_ioapic {
     pub apicaddr: ::std::os::raw::c_uint,
 }
 
-const _: () = assert!(::core::mem::size_of::<mpc_ioapic>() == 8);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_ioapic {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_intsrc {
@@ -130,13 +85,6 @@ pub struct mpc_intsrc {
     pub dstirq: ::std::os::raw::c_uchar,
 }
 
-const _: () = assert!(::core::mem::size_of::<mpc_intsrc>() == 8);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_intsrc {}
-
 pub const MP_IRQ_SOURCE_TYPES_MP_INT: ::std::os::raw::c_uint = 0;
 pub const MP_IRQ_SOURCE_TYPES_MP_NMI: ::std::os::raw::c_uint = 1;
 pub const MP_IRQ_SOURCE_TYPES_MP_EXT_INT: ::std::os::raw::c_uint = 3;
@@ -153,26 +101,12 @@ pub struct mpc_lintsrc {
     pub destapiclint: ::std::os::raw::c_uchar,
 }
 
-const _: () = assert!(::core::mem::size_of::<mpc_lintsrc>() == 8);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_lintsrc {}
-
 #[repr(C)]
 #[derive(Debug, Default, Copy, Clone)]
 pub struct mpc_oemtable {
-    pub signature: [::std::os::raw::c_uchar; 4usize],
+    pub signature: [::std::os::raw::c_char; 4usize],
     pub length: ::std::os::raw::c_ushort,
-    pub rev: ::std::os::raw::c_uchar,
-    pub checksum: ::std::os::raw::c_uchar,
-    pub mpc: [::std::os::raw::c_uchar; 8usize],
+    pub rev: ::std::os::raw::c_char,
+    pub checksum: ::std::os::raw::c_char,
+    pub mpc: [::std::os::raw::c_char; 8usize],
 }
-
-const _: () = assert!(::core::mem::size_of::<mpc_oemtable>() == 16);
-// SAFETY: all members of this struct are plain integers
-// and the sum of their sizes is the size of the struct, so
-// padding and reserved values are not possible as there
-// would be nowhere for them to exist.
-unsafe impl ByteValued for mpc_oemtable {}

arch/src/x86_64/mptable.rs

@@ -5,17 +5,15 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE-BSD-3-Clause file.
 
-use std::{mem, result, slice};
-
-use libc::c_uchar;
-use log::{info, warn};
-use thiserror::Error;
-use vm_memory::{Address, ByteValued, Bytes, GuestAddress, GuestMemory, GuestMemoryError};
-
-use super::MAX_SUPPORTED_CPUS_LEGACY;
-use crate::GuestMemoryMmap;
 use crate::layout::{APIC_START, HIGH_RAM_START, IOAPIC_START};
-use crate::x86_64::{get_x2apic_id, mpspec};
+use crate::x86_64::mpspec;
+use crate::GuestMemoryMmap;
+use libc::c_char;
+use std::io;
+use std::mem;
+use std::result;
+use std::slice;
+use vm_memory::{Address, ByteValued, Bytes, GuestAddress, GuestMemory, GuestMemoryError};
 
 // This is a workaround to the Rust enforcement specifying that any implementation of a foreign
 // trait (in this case `ByteValued`) where:
@@ -52,58 +50,59 @@ unsafe impl ByteValued for MpcLintsrcWrapper {}
 // SAFETY: see above
 unsafe impl ByteValued for MpfIntelWrapper {}
 
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// There was too little guest memory to store the entire MP table.
-    #[error("There was too little guest memory to store the entire MP table")]
     NotEnoughMemory,
     /// The MP table has too little address space to be stored.
-    #[error("The MP table has too little address space to be stored")]
     AddressOverflow,
     /// Failure while zeroing out the memory for the MP table.
-    #[error("Failure while zeroing out the memory for the MP table")]
-    Clear(#[source] GuestMemoryError),
+    Clear(GuestMemoryError),
+    /// Number of CPUs exceeds the maximum supported CPUs
+    TooManyCpus,
     /// Failure to write the MP floating pointer.
-    #[error("Failure to write the MP floating pointer")]
-    WriteMpfIntel(#[source] GuestMemoryError),
+    WriteMpfIntel(GuestMemoryError),
     /// Failure to write MP CPU entry.
-    #[error("Failure to write MP CPU entry")]
-    WriteMpcCpu(#[source] GuestMemoryError),
+    WriteMpcCpu(GuestMemoryError),
     /// Failure to write MP ioapic entry.
-    #[error("Failure to write MP ioapic entry")]
-    WriteMpcIoapic(#[source] GuestMemoryError),
+    WriteMpcIoapic(GuestMemoryError),
     /// Failure to write MP bus entry.
-    #[error("Failure to write MP bus entry")]
-    WriteMpcBus(#[source] GuestMemoryError),
+    WriteMpcBus(GuestMemoryError),
     /// Failure to write MP interrupt source entry.
-    #[error("Failure to write MP interrupt source entry")]
-    WriteMpcIntsrc(#[source] GuestMemoryError),
+    WriteMpcIntsrc(GuestMemoryError),
     /// Failure to write MP local interrupt source entry.
-    #[error("Failure to write MP local interrupt source entry")]
-    WriteMpcLintsrc(#[source] GuestMemoryError),
+    WriteMpcLintsrc(GuestMemoryError),
     /// Failure to write MP table header.
-    #[error("Failure to write MP table header")]
-    WriteMpcTable(#[source] GuestMemoryError),
+    WriteMpcTable(GuestMemoryError),
 }
 
 pub type Result<T> = result::Result<T, Error>;
 
+// With APIC/xAPIC, there are only 255 APIC IDs available. And IOAPIC occupies
+// one APIC ID, so only 254 CPUs at maximum may be supported. Actually it's
+// a large number for FC usecases.
+pub const MAX_SUPPORTED_CPUS: u32 = 254;
+
+// Convenience macro for making arrays of diverse character types.
+macro_rules! char_array {
+    ($t:ty; $( $c:expr ),*) => ( [ $( $c as $t ),* ] )
+}
+
 // Most of these variables are sourced from the Intel MP Spec 1.4.
-const SMP_MAGIC_IDENT: &[c_uchar; 4] = b"_MP_";
-const MPC_SIGNATURE: &[c_uchar; 4] = b"PCMP";
-const MPC_SPEC: u8 = 4;
-const MPC_OEM: &[c_uchar; 8] = b"FC      ";
-const MPC_PRODUCT_ID: &[c_uchar; 12] = &[b'0'; 12];
-const BUS_TYPE_ISA: &[c_uchar; 6] = b"ISA   ";
+const SMP_MAGIC_IDENT: [c_char; 4] = char_array!(c_char; '_', 'M', 'P', '_');
+const MPC_SIGNATURE: [c_char; 4] = char_array!(c_char; 'P', 'C', 'M', 'P');
+const MPC_SPEC: i8 = 4;
+const MPC_OEM: [c_char; 8] = char_array!(c_char; 'F', 'C', ' ', ' ', ' ', ' ', ' ', ' ');
+const MPC_PRODUCT_ID: [c_char; 12] = ['0' as c_char; 12];
+const BUS_TYPE_ISA: [u8; 6] = char_array!(u8; 'I', 'S', 'A', ' ', ' ', ' ');
 const APIC_VERSION: u8 = 0x14;
 const CPU_STEPPING: u32 = 0x600;
 const CPU_FEATURE_APIC: u32 = 0x200;
 const CPU_FEATURE_FPU: u32 = 0x001;
 
-fn compute_checksum<T: Copy + ByteValued>(v: &T) -> u8 {
-    let v: *const T = v;
+fn compute_checksum<T: Copy>(v: &T) -> u8 {
     // SAFETY: we are only reading the bytes within the size of the `T` reference `v`.
-    let v_slice = unsafe { slice::from_raw_parts(v.cast(), mem::size_of::<T>()) };
+    let v_slice = unsafe { slice::from_raw_parts(v as *const T as *const u8, mem::size_of::<T>()) };
     let mut checksum: u8 = 0;
     for i in v_slice.iter() {
         checksum = checksum.wrapping_add(*i);
@@ -116,7 +115,7 @@ fn mpf_intel_compute_checksum(v: &mpspec::mpf_intel) -> u8 {
     (!checksum).wrapping_add(1)
 }
 
-fn compute_mp_size(num_cpus: u32) -> usize {
+fn compute_mp_size(num_cpus: u8) -> usize {
     mem::size_of::<MpfIntelWrapper>()
         + mem::size_of::<MpcTableWrapper>()
         + mem::size_of::<MpcCpuWrapper>() * (num_cpus as usize)
@@ -127,19 +126,9 @@ fn compute_mp_size(num_cpus: u32) -> usize {
 }
 
 /// Performs setup of the MP table for the given `num_cpus`.
-pub fn setup_mptable(
-    offset: GuestAddress,
-    mem: &GuestMemoryMmap,
-    num_cpus: u32,
-    topology: Option<(u16, u16, u16, u16)>,
-) -> Result<()> {
-    if num_cpus > 0 {
-        let cpu_id_max = num_cpus - 1;
-        let x2apic_id_max = get_x2apic_id(cpu_id_max, topology);
-        if x2apic_id_max >= MAX_SUPPORTED_CPUS_LEGACY {
-            info!("Skipping mptable creation due to too many CPUs");
-            return Ok(());
-        }
+pub fn setup_mptable(offset: GuestAddress, mem: &GuestMemoryMmap, num_cpus: u8) -> Result<()> {
+    if num_cpus as u32 > MAX_SUPPORTED_CPUS {
+        return Err(Error::TooManyCpus);
     }
 
     // Used to keep track of the next base pointer into the MP table.
@@ -153,7 +142,7 @@ pub fn setup_mptable(
     }
 
     let mut checksum: u8 = 0;
-    let ioapicid: u8 = MAX_SUPPORTED_CPUS_LEGACY as u8 + 1;
+    let ioapicid: u8 = num_cpus + 1;
 
     // The checked_add here ensures the all of the following base_mp.unchecked_add's will be without
     // overflow.
@@ -165,13 +154,13 @@ pub fn setup_mptable(
         return Err(Error::AddressOverflow);
     }
 
-    mem.read_exact_volatile_from(base_mp, &mut vec![0; mp_size].as_slice(), mp_size)
+    mem.read_exact_from(base_mp, &mut io::repeat(0), mp_size)
         .map_err(Error::Clear)?;
 
     {
         let mut mpf_intel = MpfIntelWrapper(mpspec::mpf_intel::default());
         let size = mem::size_of::<MpfIntelWrapper>() as u64;
-        mpf_intel.0.signature = *SMP_MAGIC_IDENT;
+        mpf_intel.0.signature = SMP_MAGIC_IDENT;
         mpf_intel.0.length = 1;
         mpf_intel.0.specification = 4;
         mpf_intel.0.physptr = (base_mp.raw_value() + size) as u32;
@@ -191,7 +180,7 @@ pub fn setup_mptable(
         for cpu_id in 0..num_cpus {
             let mut mpc_cpu = MpcCpuWrapper(mpspec::mpc_cpu::default());
             mpc_cpu.0.type_ = mpspec::MP_PROCESSOR as u8;
-            mpc_cpu.0.apicid = get_x2apic_id(cpu_id, topology) as u8;
+            mpc_cpu.0.apicid = cpu_id;
             mpc_cpu.0.apicver = APIC_VERSION;
             mpc_cpu.0.cpuflag = mpspec::CPU_ENABLED as u8
                 | if cpu_id == 0 {
@@ -212,7 +201,7 @@ pub fn setup_mptable(
         let mut mpc_bus = MpcBusWrapper(mpspec::mpc_bus::default());
         mpc_bus.0.type_ = mpspec::MP_BUS as u8;
         mpc_bus.0.busid = 0;
-        mpc_bus.0.bustype = *BUS_TYPE_ISA;
+        mpc_bus.0.bustype = BUS_TYPE_ISA;
         mem.write_obj(mpc_bus, base_mp)
             .map_err(Error::WriteMpcBus)?;
         base_mp = base_mp.unchecked_add(size as u64);
@@ -283,14 +272,14 @@ pub fn setup_mptable(
 
     {
         let mut mpc_table = MpcTableWrapper(mpspec::mpc_table::default());
-        mpc_table.0.signature = *MPC_SIGNATURE;
+        mpc_table.0.signature = MPC_SIGNATURE;
         mpc_table.0.length = table_end.unchecked_offset_from(table_base) as u16;
         mpc_table.0.spec = MPC_SPEC;
-        mpc_table.0.oem = *MPC_OEM;
-        mpc_table.0.productid = *MPC_PRODUCT_ID;
+        mpc_table.0.oem = MPC_OEM;
+        mpc_table.0.productid = MPC_PRODUCT_ID;
         mpc_table.0.lapic = APIC_START.0 as u32;
         checksum = checksum.wrapping_add(compute_checksum(&mpc_table.0));
-        mpc_table.0.checksum = (!checksum).wrapping_add(1);
+        mpc_table.0.checksum = (!checksum).wrapping_add(1) as i8;
         mem.write_obj(mpc_table, table_base)
             .map_err(Error::WriteMpcTable)?;
     }
@@ -299,12 +288,10 @@ pub fn setup_mptable(
 }
 
 #[cfg(test)]
-mod unit_tests {
-    use vm_memory::bitmap::BitmapSlice;
-    use vm_memory::{GuestUsize, VolatileMemoryError, VolatileSlice, WriteVolatile};
-
+mod tests {
     use super::*;
     use crate::layout::MPTABLE_START;
+    use vm_memory::{GuestAddress, GuestUsize};
 
     fn table_entry_size(type_: u8) -> usize {
         match type_ as u32 {
@@ -323,7 +310,7 @@ mod unit_tests {
         let mem =
             GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(num_cpus))]).unwrap();
 
-        setup_mptable(MPTABLE_START, &mem, num_cpus, None).unwrap();
+        setup_mptable(MPTABLE_START, &mem, num_cpus).unwrap();
     }
 
     #[test]
@@ -332,7 +319,7 @@ mod unit_tests {
         let mem = GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(num_cpus) - 1)])
             .unwrap();
 
-        setup_mptable(MPTABLE_START, &mem, num_cpus, None).unwrap_err();
+        assert!(setup_mptable(MPTABLE_START, &mem, num_cpus).is_err());
     }
 
     #[test]
@@ -341,7 +328,7 @@ mod unit_tests {
         let mem =
             GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(num_cpus))]).unwrap();
 
-        setup_mptable(MPTABLE_START, &mem, num_cpus, None).unwrap();
+        setup_mptable(MPTABLE_START, &mem, num_cpus).unwrap();
 
         let mpf_intel: MpfIntelWrapper = mem.read_obj(MPTABLE_START).unwrap();
 
@@ -357,31 +344,27 @@ mod unit_tests {
         let mem =
             GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(num_cpus))]).unwrap();
 
-        setup_mptable(MPTABLE_START, &mem, num_cpus, None).unwrap();
+        setup_mptable(MPTABLE_START, &mem, num_cpus).unwrap();
 
         let mpf_intel: MpfIntelWrapper = mem.read_obj(MPTABLE_START).unwrap();
         let mpc_offset = GuestAddress(mpf_intel.0.physptr as GuestUsize);
         let mpc_table: MpcTableWrapper = mem.read_obj(mpc_offset).unwrap();
 
         struct Sum(u8);
-        impl WriteVolatile for Sum {
-            fn write_volatile<B: BitmapSlice>(
-                &mut self,
-                buf: &VolatileSlice<B>,
-            ) -> result::Result<usize, VolatileMemoryError> {
-                let mut tmp = vec![0u8; buf.len()];
-                tmp.write_all_volatile(buf)?;
-
-                for v in tmp.iter() {
+        impl io::Write for Sum {
+            fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+                for v in buf.iter() {
                     self.0 = self.0.wrapping_add(*v);
                 }
-
                 Ok(buf.len())
             }
+            fn flush(&mut self) -> io::Result<()> {
+                Ok(())
+            }
         }
 
         let mut sum = Sum(0);
-        mem.write_volatile_to(mpc_offset, &mut sum, mpc_table.0.length as usize)
+        mem.write_to(mpc_offset, &mut sum, mpc_table.0.length as usize)
             .unwrap();
         assert_eq!(sum.0, 0);
     }
@@ -390,12 +373,12 @@ mod unit_tests {
     fn cpu_entry_count() {
         let mem = GuestMemoryMmap::from_ranges(&[(
             MPTABLE_START,
-            compute_mp_size(MAX_SUPPORTED_CPUS_LEGACY),
+            compute_mp_size(MAX_SUPPORTED_CPUS as u8),
         )])
         .unwrap();
 
-        for i in 0..MAX_SUPPORTED_CPUS_LEGACY {
-            setup_mptable(MPTABLE_START, &mem, i, None).unwrap();
+        for i in 0..MAX_SUPPORTED_CPUS as u8 {
+            setup_mptable(MPTABLE_START, &mem, i).unwrap();
 
             let mpf_intel: MpfIntelWrapper = mem.read_obj(MPTABLE_START).unwrap();
             let mpc_offset = GuestAddress(mpf_intel.0.physptr as GuestUsize);
@@ -424,9 +407,11 @@ mod unit_tests {
 
     #[test]
     fn cpu_entry_count_max() {
-        let cpus = MAX_SUPPORTED_CPUS_LEGACY + 1;
-        let mem = GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(cpus))]).unwrap();
+        let cpus = MAX_SUPPORTED_CPUS + 1;
+        let mem =
+            GuestMemoryMmap::from_ranges(&[(MPTABLE_START, compute_mp_size(cpus as u8))]).unwrap();
 
-        setup_mptable(MPTABLE_START, &mem, cpus, None).unwrap();
+        let result = setup_mptable(MPTABLE_START, &mem, cpus as u8);
+        assert!(result.is_err());
     }
 }

arch/src/x86_64/regs.rs

@@ -6,57 +6,41 @@
 // 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 LICENSE-BSD-3-Clause file.
-use std::{mem, result};
-
+use crate::layout::{BOOT_GDT_START, BOOT_IDT_START, PVH_INFO_START};
+use crate::GuestMemoryMmap;
 use hypervisor::arch::x86::gdt::{gdt_entry, segment_from_gdt};
 use hypervisor::arch::x86::regs::CR0_PE;
-use hypervisor::arch::x86::{FpuState, SpecialRegisters};
-use thiserror::Error;
+use hypervisor::arch::x86::{FpuState, SpecialRegisters, StandardRegisters};
+use std::sync::Arc;
+use std::{mem, result};
 use vm_memory::{Address, Bytes, GuestMemory, GuestMemoryError};
 
-use crate::layout::{
-    BOOT_GDT_START, BOOT_IDT_START, BOOT_STACK_POINTER, PVH_INFO_START, ZERO_PAGE_START,
-};
-use crate::{EntryPoint, GuestMemoryMmap};
-
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// Failed to get SREGs for this CPU.
-    #[error("Failed to get SREGs for this CPU")]
-    GetStatusRegisters(#[source] hypervisor::HypervisorCpuError),
+    GetStatusRegisters(hypervisor::HypervisorCpuError),
     /// Failed to set base registers for this CPU.
-    #[error("Failed to set base registers for this CPU")]
-    SetBaseRegisters(#[source] hypervisor::HypervisorCpuError),
+    SetBaseRegisters(hypervisor::HypervisorCpuError),
     /// Failed to configure the FPU.
-    #[error("Failed to configure the FPU")]
-    SetFpuRegisters(#[source] hypervisor::HypervisorCpuError),
+    SetFpuRegisters(hypervisor::HypervisorCpuError),
     /// Setting up MSRs failed.
-    #[error("Setting up MSRs failed")]
-    SetModelSpecificRegisters(#[source] hypervisor::HypervisorCpuError),
+    SetModelSpecificRegisters(hypervisor::HypervisorCpuError),
     /// Failed to set SREGs for this CPU.
-    #[error("Failed to set SREGs for this CPU")]
-    SetStatusRegisters(#[source] hypervisor::HypervisorCpuError),
+    SetStatusRegisters(hypervisor::HypervisorCpuError),
     /// Checking the GDT address failed.
-    #[error("Checking the GDT address failed")]
     CheckGdtAddr,
     /// Writing the GDT to RAM failed.
-    #[error("Writing the GDT to RAM failed")]
-    WriteGdt(#[source] GuestMemoryError),
+    WriteGdt(GuestMemoryError),
     /// Writing the IDT to RAM failed.
-    #[error("Writing the IDT to RAM failed")]
-    WriteIdt(#[source] GuestMemoryError),
+    WriteIdt(GuestMemoryError),
     /// Writing PDPTE to RAM failed.
-    #[error("Writing PDPTE to RAM failed")]
-    WritePdpteAddress(#[source] GuestMemoryError),
+    WritePdpteAddress(GuestMemoryError),
     /// Writing PDE to RAM failed.
-    #[error("Writing PDE to RAM failed")]
-    WritePdeAddress(#[source] GuestMemoryError),
+    WritePdeAddress(GuestMemoryError),
     /// Writing PML4 to RAM failed.
-    #[error("Writing PML4 to RAM failed")]
-    WritePml4Address(#[source] GuestMemoryError),
+    WritePml4Address(GuestMemoryError),
     /// Writing PML5 to RAM failed.
-    #[error("Writing PML5 to RAM failed")]
-    WritePml5Address(#[source] GuestMemoryError),
+    WritePml5Address(GuestMemoryError),
 }
 
 pub type Result<T> = result::Result<T, Error>;
@@ -66,7 +50,7 @@ pub type Result<T> = result::Result<T, Error>;
 /// # Arguments
 ///
 /// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
-pub fn setup_fpu(vcpu: &dyn hypervisor::Vcpu) -> Result<()> {
+pub fn setup_fpu(vcpu: &Arc<dyn hypervisor::Vcpu>) -> Result<()> {
     let fpu: FpuState = FpuState {
         fcw: 0x37f,
         mxcsr: 0x1f80,
@@ -81,8 +65,8 @@ pub fn setup_fpu(vcpu: &dyn hypervisor::Vcpu) -> Result<()> {
 /// # Arguments
 ///
 /// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
-pub fn setup_msrs(vcpu: &dyn hypervisor::Vcpu) -> Result<()> {
-    vcpu.set_msrs(vcpu.boot_msr_entries())
+pub fn setup_msrs(vcpu: &Arc<dyn hypervisor::Vcpu>) -> Result<()> {
+    vcpu.set_msrs(&vcpu.boot_msr_entries())
         .map_err(Error::SetModelSpecificRegisters)?;
 
     Ok(())
@@ -93,22 +77,14 @@ pub fn setup_msrs(vcpu: &dyn hypervisor::Vcpu) -> Result<()> {
 /// # Arguments
 ///
 /// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
-/// * `entry_point` - Description of the boot entry to set up.
-pub fn setup_regs(vcpu: &dyn hypervisor::Vcpu, entry_point: EntryPoint) -> Result<()> {
-    let mut regs = vcpu.create_standard_regs();
-    match entry_point.setup_header {
-        None => {
-            regs.set_rflags(0x0000000000000002u64);
-            regs.set_rip(entry_point.entry_addr.raw_value());
-            regs.set_rbx(PVH_INFO_START.raw_value());
-        }
-        Some(_) => {
-            regs.set_rflags(0x0000000000000002u64);
-            regs.set_rip(entry_point.entry_addr.raw_value());
-            regs.set_rsp(BOOT_STACK_POINTER.raw_value());
-            regs.set_rsi(ZERO_PAGE_START.raw_value());
-        }
-    }
+/// * `boot_ip` - Starting instruction pointer.
+pub fn setup_regs(vcpu: &Arc<dyn hypervisor::Vcpu>, boot_ip: u64) -> Result<()> {
+    let regs = StandardRegisters {
+        rflags: 0x0000000000000002u64,
+        rbx: PVH_INFO_START.raw_value(),
+        rip: boot_ip,
+        ..Default::default()
+    };
     vcpu.set_regs(&regs).map_err(Error::SetBaseRegisters)
 }
 
@@ -118,13 +94,9 @@ pub fn setup_regs(vcpu: &dyn hypervisor::Vcpu, entry_point: EntryPoint) -> Resul
 ///
 /// * `mem` - The memory that will be passed to the guest.
 /// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
-pub fn setup_sregs(
-    mem: &GuestMemoryMmap,
-    vcpu: &dyn hypervisor::Vcpu,
-    enable_x2_apic_mode: bool,
-) -> Result<()> {
+pub fn setup_sregs(mem: &GuestMemoryMmap, vcpu: &Arc<dyn hypervisor::Vcpu>) -> Result<()> {
     let mut sregs: SpecialRegisters = vcpu.get_sregs().map_err(Error::GetStatusRegisters)?;
-    configure_segments_and_sregs(mem, &mut sregs, enable_x2_apic_mode)?;
+    configure_segments_and_sregs(mem, &mut sregs)?;
     vcpu.set_sregs(&sregs).map_err(Error::SetStatusRegisters)
 }
 
@@ -151,7 +123,6 @@ fn write_idt_value(val: u64, guest_mem: &GuestMemoryMmap) -> Result<()> {
 pub fn configure_segments_and_sregs(
     mem: &GuestMemoryMmap,
     sregs: &mut SpecialRegisters,
-    enable_x2_apic_mode: bool,
 ) -> Result<()> {
     let gdt_table: [u64; BOOT_GDT_MAX] = {
         // Configure GDT entries as specified by PVH boot protocol
@@ -187,19 +158,14 @@ pub fn configure_segments_and_sregs(
     sregs.cr0 = CR0_PE;
     sregs.cr4 = 0;
 
-    if enable_x2_apic_mode {
-        const X2APIC_ENABLE_BIT: u64 = 1 << 10;
-        sregs.apic_base |= X2APIC_ENABLE_BIT;
-    }
-
     Ok(())
 }
 
 #[cfg(test)]
-mod unit_tests {
-    use vm_memory::GuestAddress;
-
+mod tests {
     use super::*;
+    use crate::GuestMemoryMmap;
+    use vm_memory::GuestAddress;
 
     fn create_guest_mem() -> GuestMemoryMmap {
         GuestMemoryMmap::from_ranges(&[(GuestAddress(0), 0x10000)]).unwrap()
@@ -213,7 +179,7 @@ mod unit_tests {
     fn segments_and_sregs() {
         let mut sregs: SpecialRegisters = Default::default();
         let gm = create_guest_mem();
-        configure_segments_and_sregs(&gm, &mut sregs, false).unwrap();
+        configure_segments_and_sregs(&gm, &mut sregs).unwrap();
         assert_eq!(0x0, read_u64(&gm, BOOT_GDT_START));
         assert_eq!(
             0xcf9b000000ffff,

arch/src/x86_64/smbios.rs

@@ -6,35 +6,53 @@
 //
 // SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
 
-use std::{mem, result, slice};
-
-use thiserror::Error;
-use uuid::Uuid;
-use vm_memory::{Address, ByteValued, Bytes, GuestAddress};
-
-use crate::GuestMemoryMmap;
 use crate::layout::SMBIOS_START;
+use crate::GuestMemoryMmap;
+use std::fmt::{self, Display};
+use std::mem;
+use std::result;
+use std::slice;
+use uuid::Uuid;
+use vm_memory::ByteValued;
+use vm_memory::{Address, Bytes, GuestAddress};
 
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     /// There was too little guest memory to store the entire SMBIOS table.
-    #[error("There was too little guest memory to store the SMBIOS table")]
     NotEnoughMemory,
     /// The SMBIOS table has too little address space to be stored.
-    #[error("The SMBIOS table has too little address space to be stored")]
     AddressOverflow,
     /// Failure while zeroing out the memory for the SMBIOS table.
-    #[error("Failure while zeroing out the memory for the SMBIOS table")]
     Clear,
     /// Failure to write SMBIOS entrypoint structure
-    #[error("Failure to write SMBIOS entrypoint structure")]
     WriteSmbiosEp,
     /// Failure to write additional data to memory
-    #[error("Failure to write additional data to memory")]
     WriteData,
     /// Failure to parse uuid, uuid format may be error
-    #[error("Failure to parse uuid: {1}")]
-    ParseUuid(#[source] uuid::Error, String),
+    ParseUuid(uuid::Error),
+}
+
+impl std::error::Error for Error {}
+
+impl Display for Error {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        use self::Error::*;
+
+        let description = match self {
+            NotEnoughMemory => {
+                "There was too little guest memory to store the SMBIOS table".to_string()
+            }
+            AddressOverflow => {
+                "The SMBIOS table has too little address space to be stored".to_string()
+            }
+            Clear => "Failure while zeroing out the memory for the SMBIOS table".to_string(),
+            WriteSmbiosEp => "Failure to write SMBIOS entrypoint structure".to_string(),
+            WriteData => "Failure to write additional data to memory".to_string(),
+            ParseUuid(e) => format!("Failure to parse uuid: {e}"),
+        };
+
+        write!(f, "SMBIOS error: {description}")
+    }
 }
 
 pub type Result<T> = result::Result<T, Error>;
@@ -49,9 +67,8 @@ const PCI_SUPPORTED: u64 = 1 << 7;
 const IS_VIRTUAL_MACHINE: u8 = 1 << 4;
 
 fn compute_checksum<T: Copy>(v: &T) -> u8 {
-    let v: *const T = v;
     // SAFETY: we are only reading the bytes within the size of the `T` reference `v`.
-    let v_slice = unsafe { slice::from_raw_parts(v.cast(), mem::size_of::<T>()) };
+    let v_slice = unsafe { slice::from_raw_parts(v as *const T as *const u8, mem::size_of::<T>()) };
     let mut checksum: u8 = 0;
     for i in v_slice.iter() {
         checksum = checksum.wrapping_add(*i);
@@ -199,7 +216,7 @@ pub fn setup_smbios(
         let uuid_number = uuid
             .map(Uuid::parse_str)
             .transpose()
-            .map_err(|e| Error::ParseUuid(e, uuid.unwrap().to_string()))?
+            .map_err(Error::ParseUuid)?
             .unwrap_or(Uuid::nil());
         let smbios_sysinfo = SmbiosSysInfo {
             r#type: SYSTEM_INFORMATION,
@@ -273,7 +290,7 @@ pub fn setup_smbios(
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]

arch/src/x86_64/tdx/mod.rs

@@ -1,24 +1,21 @@
 // Copyright © 2021 Intel Corporation
 //
 // SPDX-License-Identifier: Apache-2.0
+use crate::GuestMemoryMmap;
 use std::fs::File;
 use std::io::{Read, Seek, SeekFrom};
 use std::str::FromStr;
-
-use log::{debug, info};
 use thiserror::Error;
 use uuid::Uuid;
 use vm_memory::{ByteValued, Bytes, GuestAddress, GuestMemoryError};
 
-use crate::GuestMemoryMmap;
-
 #[derive(Error, Debug)]
 pub enum TdvfError {
-    #[error("Failed read TDVF descriptor")]
+    #[error("Failed read TDVF descriptor: {0}")]
     ReadDescriptor(#[source] std::io::Error),
-    #[error("Failed read TDVF descriptor offset")]
+    #[error("Failed read TDVF descriptor offset: {0}")]
     ReadDescriptorOffset(#[source] std::io::Error),
-    #[error("Failed read GUID table")]
+    #[error("Failed read GUID table: {0}")]
     ReadGuidTable(#[source] std::io::Error),
     #[error("Invalid descriptor signature")]
     InvalidDescriptorSignature,
@@ -26,9 +23,9 @@ pub enum TdvfError {
     InvalidDescriptorSize,
     #[error("Invalid descriptor version")]
     InvalidDescriptorVersion,
-    #[error("Failed to write HOB details to guest memory")]
+    #[error("Failed to write HOB details to guest memory: {0}")]
     GuestMemoryWriteHob(#[source] GuestMemoryError),
-    #[error("Failed to create Uuid")]
+    #[error("Failed to create Uuid: {0}")]
     UuidCreation(#[source] uuid::Error),
 }
 
@@ -36,7 +33,7 @@ const TABLE_FOOTER_GUID: &str = "96b582de-1fb2-45f7-baea-a366c55a082d";
 const TDVF_METADATA_OFFSET_GUID: &str = "e47a6535-984a-4798-865e-4685a7bf8ec2";
 
 // TDVF_DESCRIPTOR
-#[repr(C, packed)]
+#[repr(packed)]
 #[derive(Default)]
 pub struct TdvfDescriptor {
     signature: [u8; 4],
@@ -46,7 +43,7 @@ pub struct TdvfDescriptor {
 }
 
 // TDVF_SECTION
-#[repr(C, packed)]
+#[repr(packed)]
 #[derive(Clone, Copy, Default, Debug)]
 pub struct TdvfSection {
     pub data_offset: u32,
@@ -101,7 +98,7 @@ fn tdvf_descriptor_offset(file: &mut File) -> Result<(SeekFrom, bool), TdvfError
         // We start after the footer GUID and the table length.
         let mut offset = table_size - 18;
 
-        debug!("Parsing GUID structure");
+        debug!("Parsing GUIDed structure");
         while offset >= 18 {
             let entry_uuid = Uuid::from_slice_le(&table[offset - 16..offset])
                 .map_err(TdvfError::UuidCreation)?;
@@ -109,7 +106,7 @@ fn tdvf_descriptor_offset(file: &mut File) -> Result<(SeekFrom, bool), TdvfError
                 u16::from_le_bytes(table[offset - 18..offset - 16].try_into().unwrap()) as usize;
             debug!(
                 "Entry GUID = {}, size = {}",
-                entry_uuid.hyphenated(),
+                entry_uuid.hyphenated().to_string(),
                 entry_size
             );
 
@@ -163,7 +160,7 @@ pub fn parse_tdvf_sections(file: &mut File) -> Result<(Vec<TdvfSection>, bool),
     // SAFETY: we read exactly the size of the descriptor header
     file.read_exact(unsafe {
         std::slice::from_raw_parts_mut(
-            (&raw mut descriptor).cast(),
+            &mut descriptor as *mut _ as *mut u8,
             std::mem::size_of::<TdvfDescriptor>(),
         )
     })
@@ -190,7 +187,7 @@ pub fn parse_tdvf_sections(file: &mut File) -> Result<(Vec<TdvfSection>, bool),
     // SAFETY: we read exactly the advertised sections
     file.read_exact(unsafe {
         std::slice::from_raw_parts_mut(
-            sections.as_mut_ptr().cast(),
+            sections.as_mut_ptr() as *mut u8,
             descriptor.num_sections as usize * std::mem::size_of::<TdvfSection>(),
         )
     })
@@ -210,7 +207,7 @@ enum HobType {
     EndOfHobList = 0xffff,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobHeader {
     r#type: HobType,
@@ -218,7 +215,7 @@ struct HobHeader {
     reserved: u32,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobHandoffInfoTable {
     header: HobHeader,
@@ -231,7 +228,7 @@ struct HobHandoffInfoTable {
     efi_end_of_hob_list: u64,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct EfiGuid {
     data1: u32,
@@ -240,7 +237,7 @@ struct EfiGuid {
     data4: [u8; 8],
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobResourceDescriptor {
     header: HobHeader,
@@ -251,7 +248,7 @@ struct HobResourceDescriptor {
     resource_length: u64,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobGuidType {
     header: HobHeader,
@@ -267,14 +264,14 @@ pub enum PayloadImageType {
     RawVmLinux,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 pub struct PayloadInfo {
     pub image_type: PayloadImageType,
     pub entry_point: u64,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct TdPayload {
     guid_type: HobGuidType,
@@ -300,12 +297,12 @@ pub struct TdHob {
 }
 
 fn align_hob(v: u64) -> u64 {
-    v.div_ceil(8) * 8
+    (v + 7) / 8 * 8
 }
 
 impl TdHob {
     fn update_offset<T>(&mut self) {
-        self.current_offset = align_hob(self.current_offset + std::mem::size_of::<T>() as u64);
+        self.current_offset = align_hob(self.current_offset + std::mem::size_of::<T>() as u64)
     }
 
     pub fn start(offset: u64) -> TdHob {
@@ -520,7 +517,7 @@ impl TdHob {
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]
@@ -529,7 +526,7 @@ mod unit_tests {
         let mut f = std::fs::File::open("tdvf.fd").unwrap();
         let (sections, _) = parse_tdvf_sections(&mut f).unwrap();
         for section in sections {
-            eprintln!("{section:x?}");
+            eprintln!("{section:x?}")
         }
     }
 }

block/Cargo.toml

@@ -1,40 +0,0 @@
-[package]
-authors = ["The Chromium OS Authors", "The Cloud Hypervisor Authors"]
-edition.workspace = true
-name = "block"
-rust-version.workspace = true
-version = "0.1.0"
-
-[features]
-default = []
-io_uring = ["dep:io-uring"]
-
-[dependencies]
-bitflags = { workspace = true }
-byteorder = { workspace = true }
-crc-any = "2.5.0"
-flate2 = "1.1"
-io-uring = { version = "0.7.12", optional = true }
-libc = { workspace = true }
-log = { workspace = true }
-remain = "0.2.15"
-serde = { workspace = true, features = ["derive"] }
-smallvec = "1.15.1"
-thiserror = { workspace = true }
-uuid = { workspace = true, features = ["v4"] }
-virtio-bindings = { workspace = true }
-virtio-queue = { workspace = true }
-vm-memory = { workspace = true, features = [
-  "backend-atomic",
-  "backend-bitmap",
-  "backend-mmap",
-] }
-vm-virtio = { path = "../vm-virtio" }
-vmm-sys-util = { workspace = true }
-zstd = "0.13"
-
-[dev-dependencies]
-cfg-if = { workspace = true }
-
-[lints]
-workspace = true

block/src/aligned_operation.rs

@@ -1,90 +0,0 @@
-// Copyright (c) 2026 Meta Platforms, Inc. and affiliates.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::alloc::{Layout, alloc_zeroed, dealloc};
-use std::io;
-
-use vm_memory::GuestAddress;
-
-/// Owns an aligned bounce buffer used when a guest descriptor's host VA
-/// does not meet the disk backend's alignment requirement.
-#[derive(Debug)]
-pub struct AlignedOperation {
-    data_addr: GuestAddress,
-    aligned_ptr: *mut u8,
-    size: usize,
-    layout: Layout,
-}
-
-impl AlignedOperation {
-    /// Allocate a zero-initialized buffer of `size` bytes aligned to
-    /// `alignment`. Returns `InvalidInput` if `size` is zero;
-    /// `alignment` must be a power of two and not exceed `isize::MAX`
-    /// after rounding up.
-    pub fn new(data_addr: GuestAddress, size: usize, alignment: usize) -> io::Result<Self> {
-        if size == 0 {
-            return Err(io::Error::new(
-                io::ErrorKind::InvalidInput,
-                "AlignedOperation requires a non-zero size",
-            ));
-        }
-        let layout = Layout::from_size_align(size, alignment)
-            .map_err(|e| io::Error::new(io::ErrorKind::InvalidInput, e))?;
-        // SAFETY: size is non-zero (checked above) and Layout::from_size_align
-        // rejects alignments that are not a power of two or that overflow.
-        let aligned_ptr = unsafe { alloc_zeroed(layout) };
-        if aligned_ptr.is_null() {
-            return Err(io::Error::last_os_error());
-        }
-        Ok(Self {
-            data_addr,
-            aligned_ptr,
-            size,
-            layout,
-        })
-    }
-
-    /// Gets the raw pointer to the aligned buffer.
-    pub fn as_mut_ptr(&mut self) -> *mut u8 {
-        self.aligned_ptr
-    }
-
-    /// Returns the aligned buffer as a slice.
-    pub fn as_bytes(&self) -> &[u8] {
-        // SAFETY: `new` allocates `size` bytes via alloc_zeroed (so they
-        // are initialized) and AlignedOperation owns the buffer
-        // exclusively.
-        unsafe { std::slice::from_raw_parts(self.aligned_ptr, self.size) }
-    }
-
-    /// Returns the aligned buffer as a mutable slice.
-    pub fn as_bytes_mut(&mut self) -> &mut [u8] {
-        // SAFETY: same invariant as as_bytes; &mut self rules out other
-        // simultaneous borrows.
-        unsafe { std::slice::from_raw_parts_mut(self.aligned_ptr, self.size) }
-    }
-
-    /// Returns the guest address for this op.
-    pub fn data_addr(&self) -> GuestAddress {
-        self.data_addr
-    }
-}
-
-impl Drop for AlignedOperation {
-    fn drop(&mut self) {
-        // SAFETY: `new` is the only constructor, and it stores a pointer
-        // returned by `alloc_zeroed` paired with the exact `layout` used
-        // for that allocation. Ownership has not escaped (the type is
-        // neither `Clone` nor `Copy`).
-        unsafe {
-            dealloc(self.aligned_ptr, self.layout);
-        }
-    }
-}
-
-// SAFETY: AlignedOperation owns its heap allocation exclusively (no Clone/
-// Copy, no shared aliases) and the allocation's lifetime is tied to the
-// value's. Moving an AlignedOperation between threads transfers that
-// ownership — the same rationale Box<T> uses for its Send impl.
-unsafe impl Send for AlignedOperation {}

block/src/async_io.rs

@@ -1,112 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::marker::PhantomData;
-use std::os::fd::{AsRawFd, OwnedFd, RawFd};
-
-use thiserror::Error;
-use vmm_sys_util::eventfd::EventFd;
-
-use crate::{BatchRequest, SECTOR_SIZE};
-
-#[derive(Error, Debug)]
-pub enum DiskFileError {
-    /// Failed getting disk file size.
-    #[error("Failed getting disk file size")]
-    Size(#[source] std::io::Error),
-    /// Failed creating a new AsyncIo.
-    #[error("Failed creating a new AsyncIo")]
-    NewAsyncIo(#[source] std::io::Error),
-    /// Unsupported operation.
-    #[error("Unsupported operation")]
-    Unsupported,
-    /// Resize failed
-    #[error("Resize failed")]
-    ResizeError(#[source] std::io::Error),
-    #[error("Failed cloning disk file")]
-    Clone(#[source] std::io::Error),
-}
-
-pub type DiskFileResult<T> = std::result::Result<T, DiskFileError>;
-
-/// A wrapper for [`RawFd`] capturing the lifetime of a corresponding disk file.
-///
-/// This fulfills the same role as [`BorrowedFd`] but is tailored to the limitations
-/// by some disk implementations, which wrap the effective [`File`]
-/// in an `Arc<Mutex<T>>`, making the use of [`BorrowedFd`] impossible.
-///
-/// [`BorrowedFd`]: std::os::fd::BorrowedFd
-#[derive(Copy, Clone, Debug)]
-pub struct BorrowedDiskFd<'fd> {
-    raw_fd: RawFd,
-    _lifetime: PhantomData<&'fd OwnedFd>,
-}
-
-impl BorrowedDiskFd<'_> {
-    pub(super) fn new(raw_fd: RawFd) -> Self {
-        Self {
-            raw_fd,
-            _lifetime: PhantomData,
-        }
-    }
-}
-
-impl AsRawFd for BorrowedDiskFd<'_> {
-    fn as_raw_fd(&self) -> RawFd {
-        self.raw_fd
-    }
-}
-
-#[derive(Error, Debug)]
-pub enum AsyncIoError {
-    /// Failed vectored reading from file.
-    #[error("Failed vectored reading from file")]
-    ReadVectored(#[source] std::io::Error),
-    /// Failed vectored writing to file.
-    #[error("Failed vectored writing to file")]
-    WriteVectored(#[source] std::io::Error),
-    /// Failed synchronizing file.
-    #[error("Failed synchronizing file")]
-    Fsync(#[source] std::io::Error),
-    /// Failed punching hole.
-    #[error("Failed punching hole")]
-    PunchHole(#[source] std::io::Error),
-    /// Failed writing zeroes.
-    #[error("Failed writing zeroes")]
-    WriteZeroes(#[source] std::io::Error),
-    /// Failed submitting batch requests.
-    #[error("Failed submitting batch requests")]
-    SubmitBatchRequests(#[source] std::io::Error),
-}
-
-pub type AsyncIoResult<T> = std::result::Result<T, AsyncIoError>;
-
-pub trait AsyncIo: Send {
-    fn notifier(&self) -> &EventFd;
-    fn read_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()>;
-    fn write_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()>;
-    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()>;
-    fn punch_hole(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()>;
-    fn write_zeroes(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()>;
-    fn next_completed_request(&mut self) -> Option<(u64, i32)>;
-    fn batch_requests_enabled(&self) -> bool {
-        false
-    }
-    fn submit_batch_requests(&mut self, _batch_request: &[BatchRequest]) -> AsyncIoResult<()> {
-        Ok(())
-    }
-    fn alignment(&self) -> u64 {
-        SECTOR_SIZE
-    }
-}

block/src/disk_file.rs

@@ -1,158 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-//! Composable disk capability traits for the block crate.
-//!
-//! Small traits define individual capabilities:
-//!
-//! - [`DiskSize`] - reported capacity (logical size)
-//! - [`PhysicalSize`] - host allocation size
-//! - [`DiskFd`] - backing file descriptor access
-//! - [`Geometry`] - sector/cluster geometry (default 512B)
-//! - [`SparseCapable`] - sparse and zero flag support
-//! - [`Resizable`] - online resize
-//!
-//! [`DiskFile`] is a supertrait that bundles the universal capabilities
-//! (`DiskSize` + `Geometry`). [`FullDiskFile`] adds all optional
-//! capabilities. [`AsyncDiskFile`] extends `DiskFile` with async I/O
-//! construction for virtio queue workers. [`AsyncFullDiskFile`]
-//! combines both axes.
-//!
-//! ```text
-//!         DiskFile: DiskSize + Geometry + Sync
-//!         /                                     \
-//! FullDiskFile:                           AsyncDiskFile:
-//!   DiskFile + PhysicalSize +               DiskFile + Unpin
-//!   DiskFd + SparseCapable +               try_clone, create_async_io
-//!   Resizable
-//!         \                                     /
-//!          AsyncFullDiskFile: FullDiskFile + AsyncDiskFile
-//! ```
-//!
-//! Readonly accessors take `&self`. Only [`Resizable::resize`] requires
-//! `&mut self`. Errors are returned as [`BlockResult`].
-
-use std::fmt::Debug;
-
-use crate::async_io::{AsyncIo, BorrowedDiskFd};
-use crate::{BlockResult, DiskTopology};
-
-/// Reported capacity of a disk image.
-pub trait DiskSize: Send + Debug {
-    /// Virtual size of the disk image in bytes (reported capacity).
-    fn logical_size(&self) -> BlockResult<u64>;
-}
-
-/// Host allocation size of a file-backed disk image.
-pub trait PhysicalSize: Send + Debug {
-    /// Actual bytes occupied on the host filesystem.
-    fn physical_size(&self) -> BlockResult<u64>;
-}
-
-/// Backing file descriptor access for disk images backed by a file.
-pub trait DiskFd: Send + Debug {
-    /// Borrows the underlying file descriptor.
-    fn fd(&self) -> BorrowedDiskFd<'_>;
-}
-
-/// Sector and cluster geometry of a disk image.
-///
-/// Default returns `DiskTopology::default()` (512B logical/physical).
-pub trait Geometry: Send + Debug {
-    /// Returns the disk topology.
-    fn topology(&self) -> DiskTopology {
-        DiskTopology::default()
-    }
-}
-
-/// Sparse and zero flag support for thin provisioned disk images.
-pub trait SparseCapable: Send + Debug {
-    /// Indicates support for sparse operations (punch hole, write zeroes, discard).
-    fn supports_sparse_operations(&self) -> bool {
-        false
-    }
-
-    /// Indicates support for a metadata level zero flag optimization in
-    /// virtio `VIRTIO_BLK_T_WRITE_ZEROES` requests. When true, the format
-    /// can mark regions as reading zeros via a metadata bit rather than
-    /// writing actual zero bytes to disk.
-    fn supports_zero_flag(&self) -> bool {
-        false
-    }
-}
-
-/// Live disk resize support.
-///
-/// Implementations may return an error if the backend does not
-/// support resizing (e.g. fixed size formats).
-pub trait Resizable: Send + Debug {
-    /// Resizes the disk image to the given size in bytes, if the backend supports it.
-    fn resize(&mut self, size: u64) -> BlockResult<()>;
-}
-
-/// Supertrait bundling universal disk capabilities.
-///
-/// Every disk format implements `DiskSize` and `Geometry`.
-/// `Sync` is required so that `Arc<dyn DiskFile>` can be shared
-/// across threads for concurrent readonly access.
-pub trait DiskFile: DiskSize + Geometry + Sync {}
-
-/// Full capability disk file trait.
-///
-/// Bundles all optional capabilities on top of [`DiskFile`]:
-/// file descriptor access, physical size, sparse operations, and resize.
-/// Used by consumers that need feature negotiation without async I/O
-/// (e.g. vhost user block).
-pub trait FullDiskFile: DiskFile + PhysicalSize + DiskFd + SparseCapable + Resizable {}
-
-/// Blanket implementation: any type implementing all constituent traits
-/// automatically satisfies [`FullDiskFile`].
-impl<T: DiskFile + PhysicalSize + DiskFd + SparseCapable + Resizable> FullDiskFile for T {}
-
-/// Extended disk file trait for virtio queue workers.
-///
-/// Adds cloning and async I/O construction on top of [`DiskFile`].
-/// `Unpin` is required so trait objects can be moved freely.
-pub trait AsyncDiskFile: DiskFile + Unpin {
-    /// Creates an independent handle for a queue worker.
-    ///
-    /// The clone shares internally reference counted state (e.g.
-    /// `Arc<Metadata>`) with the original, but owns its own file
-    /// descriptor and I/O completion resources. Each virtio queue
-    /// gets one clone so that workers can operate in parallel
-    /// without contending on I/O state.
-    ///
-    /// Returns `Box<dyn AsyncDiskFile>` (not `AsyncFullDiskFile`)
-    /// because clones only serve as data plane handles for queue
-    /// workers. The original remains the control plane for feature
-    /// negotiation and configuration.
-    fn try_clone(&self) -> BlockResult<Box<dyn AsyncDiskFile>>;
-
-    /// Constructs a per queue async I/O engine.
-    ///
-    /// # Arguments
-    ///
-    /// * `ring_depth` - maximum number of in flight I/O operations.
-    ///   Callers typically pass the virtio queue size. Must be greater
-    ///   than zero. Backends that do not use an async ring (e.g. sync
-    ///   fallback implementations) may ignore this value.
-    fn create_async_io(&self, ring_depth: u32) -> BlockResult<Box<dyn AsyncIo>>;
-}
-
-/// Full capability async disk file trait.
-///
-/// Combines [`FullDiskFile`] (all optional capabilities) with
-/// [`AsyncDiskFile`] (async I/O construction). This is the top level
-/// trait for virtio block devices that need both feature negotiation
-/// and async queue workers.
-///
-/// The type narrowing on [`AsyncDiskFile::try_clone`] is intentional:
-/// clones only serve as data plane handles for queue workers, while
-/// the original `AsyncFullDiskFile` handle remains the control plane
-/// for feature negotiation and configuration.
-pub trait AsyncFullDiskFile: FullDiskFile + AsyncDiskFile {}
-
-/// Blanket implementation: any type implementing both [`FullDiskFile`]
-/// and [`AsyncDiskFile`] automatically satisfies [`AsyncFullDiskFile`].
-impl<T: FullDiskFile + AsyncDiskFile> AsyncFullDiskFile for T {}

block/src/error.rs

@@ -1,245 +0,0 @@
-// Copyright 2025 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-//! Unified error handling for the block crate.
-//!
-//! # Architecture
-//!
-//! ```text
-//! BlockError                 -- single public error type
-//!  |-- BlockErrorKind        -- small, stable, matchable classification
-//!  |-- ErrorContext          -- optional diagnostic metadata (path, offset, op)
-//!  +-- source                -- format-specific error (boxed)
-//!         |-- QcowError
-//!         |-- VhdError / RawError / ...
-//!         +-- io::Error / etc.
-//! ```
-
-use std::error::Error as StdError;
-use std::fmt::{self, Display, Formatter};
-use std::io;
-use std::path::PathBuf;
-
-/// Small, stable classification of block errors.
-///
-/// Callers match on this for control flow. Adding new format specific
-/// errors does not require new variants here.
-#[derive(Debug, Copy, Clone, Eq, PartialEq)]
-#[non_exhaustive]
-pub enum BlockErrorKind {
-    /// An underlying I/O operation failed.
-    Io,
-    /// The disk image format is structurally invalid.
-    InvalidFormat,
-    /// The disk image requires a feature that is not implemented.
-    UnsupportedFeature,
-    /// The image is marked or detected as corrupt.
-    CorruptImage,
-    /// An address, offset, or index is outside the valid range.
-    OutOfBounds,
-    /// A file or required internal structure could not be found.
-    NotFound,
-    /// An internal counter or limit was exceeded.
-    Overflow,
-}
-
-impl Display for BlockErrorKind {
-    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
-        match self {
-            Self::Io => write!(f, "I/O error"),
-            Self::InvalidFormat => write!(f, "Invalid format"),
-            Self::UnsupportedFeature => write!(f, "Unsupported feature"),
-            Self::CorruptImage => write!(f, "Corrupt image"),
-            Self::OutOfBounds => write!(f, "Out of bounds"),
-            Self::NotFound => write!(f, "Not found"),
-            Self::Overflow => write!(f, "Overflow"),
-        }
-    }
-}
-
-/// Classification of the operation that was in progress when an error occurred.
-#[derive(Debug, Copy, Clone, Eq, PartialEq)]
-#[non_exhaustive]
-pub enum ErrorOp {
-    /// Opening a disk image file.
-    Open,
-    /// Detecting the image format.
-    DetectImageType,
-    /// Duplicating a backing-file descriptor.
-    DupBackingFd,
-    /// Resizing a disk image.
-    Resize,
-}
-
-impl Display for ErrorOp {
-    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
-        match self {
-            Self::Open => write!(f, "open"),
-            Self::DetectImageType => write!(f, "detect_image_type"),
-            Self::DupBackingFd => write!(f, "dup_backing_fd"),
-            Self::Resize => write!(f, "resize"),
-        }
-    }
-}
-
-/// Optional diagnostic context attached to a [`BlockError`].
-#[derive(Debug, Default, Clone)]
-pub struct ErrorContext {
-    pub path: Option<PathBuf>,
-    pub offset: Option<u64>,
-    pub op: Option<ErrorOp>,
-}
-
-impl Display for ErrorContext {
-    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
-        let mut first = true;
-        if let Some(path) = &self.path {
-            write!(f, "path={}", path.display())?;
-            first = false;
-        }
-        if let Some(offset) = self.offset {
-            if !first {
-                write!(f, " ")?;
-            }
-            write!(f, "offset={offset:#x}")?;
-            first = false;
-        }
-        if let Some(op) = self.op {
-            if !first {
-                write!(f, " ")?;
-            }
-            write!(f, "op={op}")?;
-        }
-        Ok(())
-    }
-}
-
-/// Unified error type for the block crate.
-///
-/// Pairs a stable [`BlockErrorKind`] classification with an optional
-/// boxed source error (format-specific) and optional [`ErrorContext`].
-///
-/// Display renders kind + context only; the underlying cause is
-/// exposed via [`std::error::Error::source()`] for reporters that
-/// walk the chain.
-#[derive(Debug)]
-pub struct BlockError {
-    kind: BlockErrorKind,
-    source: Option<Box<dyn StdError + Send + Sync + 'static>>,
-    ctx: Option<ErrorContext>,
-}
-
-impl BlockError {
-    /// Create a new `BlockError` from a kind and a source error.
-    pub fn new<E>(kind: BlockErrorKind, source: E) -> Self
-    where
-        E: StdError + Send + Sync + 'static,
-    {
-        Self {
-            kind,
-            source: Some(Box::new(source)),
-            ctx: None,
-        }
-    }
-
-    /// Create a `BlockError` from just a kind, with no underlying cause.
-    pub fn from_kind(kind: BlockErrorKind) -> Self {
-        Self {
-            kind,
-            source: None,
-            ctx: None,
-        }
-    }
-
-    /// Attach or replace the source error (builder-style).
-    pub fn with_source<E>(mut self, source: E) -> Self
-    where
-        E: StdError + Send + Sync + 'static,
-    {
-        self.source = Some(Box::new(source));
-        self
-    }
-
-    /// Attach diagnostic context.
-    pub fn with_ctx(mut self, ctx: ErrorContext) -> Self {
-        self.ctx = Some(ctx);
-        self
-    }
-
-    /// Replace the error classification (builder-style).
-    pub fn with_kind(mut self, kind: BlockErrorKind) -> Self {
-        self.kind = kind;
-        self
-    }
-
-    /// Shorthand: attach an operation name.
-    pub fn with_op(mut self, op: ErrorOp) -> Self {
-        self.ctx.get_or_insert_with(ErrorContext::default).op = Some(op);
-        self
-    }
-
-    /// Shorthand: attach a file path.
-    pub fn with_path(mut self, path: impl Into<PathBuf>) -> Self {
-        self.ctx.get_or_insert_with(ErrorContext::default).path = Some(path.into());
-        self
-    }
-
-    /// Shorthand: attach a byte offset.
-    pub fn with_offset(mut self, offset: u64) -> Self {
-        self.ctx.get_or_insert_with(ErrorContext::default).offset = Some(offset);
-        self
-    }
-
-    /// The error classification.
-    pub fn kind(&self) -> BlockErrorKind {
-        self.kind
-    }
-
-    /// The diagnostic context, if any.
-    pub fn context(&self) -> Option<&ErrorContext> {
-        self.ctx.as_ref()
-    }
-
-    /// Access the underlying source error, if any.
-    pub fn source_ref(&self) -> Option<&(dyn StdError + Send + Sync + 'static)> {
-        self.source.as_deref()
-    }
-
-    /// Try to downcast the source to a concrete type.
-    pub fn downcast_ref<T: StdError + 'static>(&self) -> Option<&T> {
-        self.source.as_ref()?.downcast_ref::<T>()
-    }
-
-    /// Consume the error and return the boxed source, if any.
-    pub fn into_source(self) -> Option<Box<dyn StdError + Send + Sync + 'static>> {
-        self.source
-    }
-}
-
-impl Display for BlockError {
-    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
-        write!(f, "{}", self.kind)?;
-        if let Some(ctx) = &self.ctx {
-            write!(f, " ({ctx})")?;
-        }
-        Ok(())
-    }
-}
-
-impl StdError for BlockError {
-    fn source(&self) -> Option<&(dyn StdError + 'static)> {
-        self.source
-            .as_ref()
-            .map(|e| e.as_ref() as &(dyn StdError + 'static))
-    }
-}
-
-/// Convenience: wrap an `io::Error` as `BlockErrorKind::Io`.
-impl From<io::Error> for BlockError {
-    fn from(e: io::Error) -> Self {
-        Self::new(BlockErrorKind::Io, e)
-    }
-}
-
-pub type BlockResult<T> = Result<T, BlockError>;

block/src/factory.rs

@@ -1,292 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-//! Disk image factory.
-//!
-//! [`open_disk`] is the single entry point for opening a disk image.
-//! It opens the file, detects the image format, probes async I/O
-//! support, and constructs the appropriate backend. Callers receive
-//! a trait object that is ready for use by virtio queue workers.
-
-use std::os::unix::fs::OpenOptionsExt;
-use std::path::Path;
-use std::sync::OnceLock;
-use std::{fmt, fs};
-
-use log::info;
-
-#[cfg(feature = "io_uring")]
-use crate::block_io_uring_is_supported;
-use crate::disk_file::AsyncFullDiskFile;
-use crate::error::{BlockError, BlockErrorKind, BlockResult};
-use crate::fixed_vhd_disk::FixedVhdDisk;
-use crate::qcow_disk::QcowDisk;
-use crate::raw_disk::{RawBackend, RawDisk};
-use crate::vhdx_sync::VhdxDiskSync;
-use crate::{
-    ImageType, block_aio_is_supported, detect_image_type, open_disk_image, preallocate_disk,
-};
-
-/// Options for opening a disk image via [`open_disk`].
-pub struct DiskOpenOptions<'a> {
-    pub path: &'a Path,
-    pub readonly: bool,
-    pub direct: bool,
-    pub sparse: bool,
-    pub backing_files: bool,
-    pub disable_io_uring: bool,
-    pub disable_aio: bool,
-}
-
-/// Result of [`open_disk`], carrying the detected image type alongside
-/// the constructed backend.
-pub struct OpenedDisk {
-    pub image_type: ImageType,
-    pub disk: Box<dyn AsyncFullDiskFile>,
-}
-
-impl fmt::Debug for OpenedDisk {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        f.debug_struct("OpenedDisk")
-            .field("image_type", &self.image_type)
-            .finish_non_exhaustive()
-    }
-}
-
-/// Returns true when io_uring is supported on the running kernel.
-///
-/// The result is cached so the probe runs at most once per process.
-#[cfg(feature = "io_uring")]
-fn io_uring_supported() -> bool {
-    static SUPPORTED: OnceLock<bool> = OnceLock::new();
-    *SUPPORTED.get_or_init(block_io_uring_is_supported)
-}
-
-/// Returns true when Linux AIO is supported on the running kernel.
-///
-/// The result is cached so the probe runs at most once per process.
-fn aio_supported() -> bool {
-    static SUPPORTED: OnceLock<bool> = OnceLock::new();
-    *SUPPORTED.get_or_init(block_aio_is_supported)
-}
-
-/// Open a disk image and construct the appropriate async backend.
-///
-/// - Opens the file with the requested access mode and flags.
-/// - Detects the image format from the file header.
-/// - Probes io_uring and Linux AIO support on the running kernel.
-/// - Constructs the most capable backend available for the detected
-///   format, preferring io_uring over AIO over synchronous fallback.
-///
-/// The returned [`OpenedDisk`] exposes the detected [`ImageType`] so
-/// callers can perform post construction validation (e.g. type mismatch
-/// checks, configuration warnings).
-pub fn open_disk(options: &DiskOpenOptions<'_>) -> BlockResult<OpenedDisk> {
-    let mut fs_options = fs::OpenOptions::new();
-    fs_options.read(true);
-    fs_options.write(!options.readonly);
-    if options.direct {
-        fs_options.custom_flags(libc::O_DIRECT);
-    }
-
-    let mut file = open_disk_image(options.path, &fs_options)?;
-    let image_type = detect_image_type(&mut file)?;
-
-    let disk: Box<dyn AsyncFullDiskFile> = match image_type {
-        ImageType::FixedVhd => open_fixed_vhd(file, options)?,
-        ImageType::Raw => open_raw(file, options)?,
-        ImageType::Qcow2 => open_qcow2(file, options)?,
-        ImageType::Vhdx => open_vhdx(file, options)?,
-        ImageType::Unknown => {
-            return Err(
-                BlockError::from_kind(BlockErrorKind::UnsupportedFeature).with_path(options.path)
-            );
-        }
-    };
-
-    Ok(OpenedDisk { image_type, disk })
-}
-
-fn open_vhdx(
-    file: fs::File,
-    options: &DiskOpenOptions<'_>,
-) -> BlockResult<Box<dyn AsyncFullDiskFile>> {
-    info!("Opening VHDX disk file with synchronous backend");
-    Ok(Box::new(
-        VhdxDiskSync::new(file).map_err(|e| e.with_path(options.path))?,
-    ))
-}
-
-fn open_fixed_vhd(
-    file: fs::File,
-    options: &DiskOpenOptions<'_>,
-) -> BlockResult<Box<dyn AsyncFullDiskFile>> {
-    #[cfg(feature = "io_uring")]
-    if !options.disable_io_uring {
-        if io_uring_supported() {
-            info!("Opening fixed VHD disk file with io_uring backend");
-            return Ok(Box::new(
-                FixedVhdDisk::new(file, true).map_err(|e| e.with_path(options.path))?,
-            ));
-        }
-        info!("io_uring runtime probe failed for fixed VHD, using synchronous backend");
-    }
-
-    info!("Opening fixed VHD disk file with synchronous backend");
-    Ok(Box::new(
-        FixedVhdDisk::new(file, false).map_err(|e| e.with_path(options.path))?,
-    ))
-}
-
-fn open_raw(
-    file: fs::File,
-    options: &DiskOpenOptions<'_>,
-) -> BlockResult<Box<dyn AsyncFullDiskFile>> {
-    if !options.readonly && !options.sparse {
-        preallocate_disk(&file, options.path);
-    }
-
-    #[cfg(feature = "io_uring")]
-    if !options.disable_io_uring {
-        if io_uring_supported() {
-            info!("Opening RAW disk file with io_uring backend");
-            return Ok(Box::new(RawDisk::new(file, RawBackend::IoUring)));
-        }
-        info!("io_uring runtime probe failed for RAW, trying next backend");
-    }
-
-    if !options.disable_aio {
-        if aio_supported() {
-            info!("Opening RAW disk file with AIO backend");
-            return Ok(Box::new(RawDisk::new(file, RawBackend::Aio)));
-        }
-        info!("AIO runtime probe failed for RAW, using synchronous backend");
-    }
-
-    info!("Opening RAW disk file with synchronous backend");
-    Ok(Box::new(RawDisk::new(file, RawBackend::Sync)))
-}
-
-fn open_qcow2(
-    file: fs::File,
-    options: &DiskOpenOptions<'_>,
-) -> BlockResult<Box<dyn AsyncFullDiskFile>> {
-    #[cfg(feature = "io_uring")]
-    if !options.disable_io_uring {
-        if io_uring_supported() {
-            info!("Opening QCOW2 disk file with io_uring backend");
-            return Ok(Box::new(
-                QcowDisk::new(
-                    file,
-                    options.direct,
-                    options.backing_files,
-                    options.sparse,
-                    true,
-                )
-                .map_err(|e| e.with_path(options.path))?,
-            ));
-        }
-        info!("io_uring runtime probe failed for QCOW2, using synchronous backend");
-    }
-
-    info!("Opening QCOW2 disk file with synchronous backend");
-    Ok(Box::new(
-        QcowDisk::new(
-            file,
-            options.direct,
-            options.backing_files,
-            options.sparse,
-            false,
-        )
-        .map_err(|e| e.with_path(options.path))?,
-    ))
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use std::io::Write;
-    use std::path::Path;
-
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::qcow::{QcowFile, RawFile};
-
-    fn default_options(path: &Path) -> DiskOpenOptions<'_> {
-        DiskOpenOptions {
-            path,
-            readonly: false,
-            direct: false,
-            sparse: false,
-            backing_files: false,
-            disable_io_uring: true,
-            disable_aio: true,
-        }
-    }
-
-    #[test]
-    fn nonexistent_path_returns_error() {
-        let path = Path::new("/tmp/no_such_disk_image.raw");
-        let options = default_options(path);
-        match open_disk(&options) {
-            Err(e) => assert_eq!(e.kind(), BlockErrorKind::Io),
-            Ok(_) => panic!("expected error for nonexistent path"),
-        }
-    }
-
-    #[test]
-    fn detect_raw_image() {
-        let tmp = TempFile::new().unwrap();
-        tmp.as_file().set_len(1 << 20).unwrap();
-        let path = tmp.as_path().to_owned();
-        let options = default_options(&path);
-        let opened = open_disk(&options).unwrap();
-        assert_eq!(opened.image_type, ImageType::Raw);
-    }
-
-    #[test]
-    fn detect_qcow2_image() {
-        let tmp = TempFile::new().unwrap();
-        {
-            let raw = RawFile::new(tmp.as_file().try_clone().unwrap(), false);
-            let mut qcow = QcowFile::new(raw, 3, 100 * 1024 * 1024, true).unwrap();
-            qcow.flush().unwrap();
-        }
-        let path = tmp.as_path().to_owned();
-        let options = default_options(&path);
-        let opened = open_disk(&options).unwrap();
-        assert_eq!(opened.image_type, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn open_readonly() {
-        let tmp = TempFile::new().unwrap();
-        tmp.as_file().set_len(1 << 20).unwrap();
-        let path = tmp.as_path().to_owned();
-        let mut options = default_options(&path);
-        options.readonly = true;
-        let opened = open_disk(&options).unwrap();
-        assert_eq!(opened.image_type, ImageType::Raw);
-    }
-
-    #[test]
-    fn sync_fallback_when_async_disabled() {
-        let tmp = TempFile::new().unwrap();
-        let size = 1u64 << 20;
-        tmp.as_file().set_len(size).unwrap();
-        let path = tmp.as_path().to_owned();
-        let options = DiskOpenOptions {
-            path: &path,
-            readonly: false,
-            direct: false,
-            sparse: false,
-            backing_files: false,
-            disable_io_uring: true,
-            disable_aio: true,
-        };
-        let opened = open_disk(&options).unwrap();
-        assert_eq!(opened.image_type, ImageType::Raw);
-        assert_eq!(opened.disk.logical_size().unwrap(), size);
-    }
-}

block/src/fcntl.rs

@@ -1,254 +0,0 @@
-// Copyright © 2025 Cyberus Technology GmbH
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-
-//! Helpers for advisory file locking.
-//!
-//! Under the hood, the implementation uses OFD locks for the entire file,
-//! as described in [[0]]. The advantage over `F_SETLKW` (currently used by
-//! Rust std: `File::try_lock()`) is that only the very last `close()` on a
-//! file descriptor releases the lock. This prevents mistakes and unexpected
-//! behavior.
-//!
-//! [0]: <https://apenwarr.ca/log/20101213>.
-
-use std::fmt::Debug;
-use std::io;
-use std::os::fd::{AsRawFd, RawFd};
-use std::str::FromStr;
-
-use thiserror::Error;
-
-/// Errors that can happen when working with file locks.
-#[derive(Error, Debug)]
-pub enum LockError {
-    /// The file is already locked.
-    ///
-    /// A call to [`get_lock_state`] can help to identify the reason.
-    #[error("The file is already locked")]
-    AlreadyLocked,
-    /// IO error.
-    #[error("The lock state could not be checked or set")]
-    Io(#[source] io::Error),
-}
-
-/// Commands for use with [`fcntl`].
-#[allow(non_camel_case_types)]
-enum FcntlArg<'a> {
-    /// Set an OFD lock from the given lock description.
-    F_OFD_SETLK(&'a libc::flock),
-    /// Get the first OFD lock for the given lock description.
-    F_OFD_GETLK(&'a mut libc::flock),
-}
-
-/// Wrapper for [`libc::fcntl`] that properly sets the function arguments.
-fn fcntl(fd: RawFd, arg: FcntlArg) -> libc::c_int {
-    // SAFETY: We use a valid FD.
-    unsafe {
-        match arg {
-            FcntlArg::F_OFD_SETLK(flock) => libc::fcntl(fd, libc::F_OFD_SETLK, flock),
-            FcntlArg::F_OFD_GETLK(flock) => libc::fcntl(fd, libc::F_OFD_GETLK, flock),
-        }
-    }
-}
-
-/// Describes the type of lock you want to set.
-#[derive(Clone, Copy, Debug)]
-pub enum LockType {
-    /// Clear a lock.
-    Unlock,
-    /// Set a write lock (exclusive).
-    Write,
-    /// Set a read lock (shared).
-    Read,
-}
-
-impl LockType {
-    pub const fn to_libc_val(self) -> libc::c_int {
-        match self {
-            Self::Unlock => libc::F_UNLCK as libc::c_int,
-            Self::Write => libc::F_WRLCK as libc::c_int,
-            Self::Read => libc::F_RDLCK as libc::c_int,
-        }
-    }
-}
-
-/// Describes the current state of a lock.
-#[derive(Debug)]
-pub enum LockState {
-    /// No lock set.
-    Unlocked,
-    /// Locked for reading (non-exclusive).
-    SharedRead,
-    /// Locked for writing (exclusive mode).
-    ExclusiveWrite,
-}
-
-impl LockState {
-    fn new(value: libc::c_int) -> Self {
-        const F_UNLCK: libc::c_int = libc::F_UNLCK as libc::c_int;
-        const F_WRLCK: libc::c_int = libc::F_WRLCK as libc::c_int;
-        const F_RDLCK: libc::c_int = libc::F_RDLCK as libc::c_int;
-        match value {
-            F_UNLCK => Self::Unlocked,
-            F_WRLCK => Self::ExclusiveWrite,
-            F_RDLCK => Self::SharedRead,
-            // This is so unlikely that we want to avoid the complexity of
-            // coping with this error case. Can only fail if either Linux
-            // is broken or memory is messed up.
-            other => panic!("Unexpected lock state: {other}"),
-        }
-    }
-}
-
-/// The granularity of the advisory lock.
-///
-/// The granularity has significant implications in typical cloud deployments
-/// with network storage. The Linux kernel will sync advisory locks to network
-/// file systems, but these backends may have different policies and handle
-/// locks differently. For example, Netapp speaks a NFS API but will treat
-/// advisory OFD locks for the whole file as mandatory locks, whereas byte-range
-/// locks for the whole file will remain advisory [0].
-///
-/// As it is a valid use case to prevent multiple CHV instances from accessing
-/// the same disk but disk management software (e.g., Cinder in OpenStack)
-/// should be able to snapshot disks while VMs are running, we need special
-/// control over the lock granularity. Therefore, it is a valid use case to lock
-/// the whole byte range of a disk image without technically locking the whole
-/// file - to get the best of both worlds.
-///
-/// [0] https://kb.netapp.com/on-prem/ontap/da/NAS/NAS-KBs/How_is_Mandatory_Locking_supported_for_NFSv4_on_ONTAP_9
-#[derive(Clone, Copy, Debug)]
-pub enum LockGranularity {
-    WholeFile,
-    ByteRange(u64 /* from, inclusive */, u64 /* len */),
-}
-
-impl LockGranularity {
-    const fn l_start(self) -> u64 {
-        match self {
-            LockGranularity::WholeFile => 0,
-            LockGranularity::ByteRange(start, _) => start,
-        }
-    }
-
-    const fn l_len(self) -> u64 {
-        match self {
-            LockGranularity::WholeFile => 0, /* EOF */
-            LockGranularity::ByteRange(_, len) => len,
-        }
-    }
-}
-
-/// User-facing choice for the lock granularity.
-///
-/// This allows external management software to create snapshots of the disk
-/// image. Without a byte-range lock, some NFS implementations may treat the
-/// entire file as exclusively locked and prevent such operations (e.g. NetApp).
-#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, serde::Deserialize, serde::Serialize)]
-pub enum LockGranularityChoice {
-    /// Byte-range lock covering [0, size).
-    #[default]
-    ByteRange,
-    /// Whole-file lock (l_start=0, l_len=0) - original OFD whole-file lock behavior.
-    Full,
-}
-
-/// Error returned when parsing a [`LockGranularityChoice`] from a string.
-#[derive(Error, Debug)]
-#[error("Invalid lock granularity value: {0}, expected 'byte-range' or 'full'")]
-pub struct LockGranularityParseError(String);
-
-impl FromStr for LockGranularityChoice {
-    type Err = LockGranularityParseError;
-
-    fn from_str(s: &str) -> Result<Self, Self::Err> {
-        match s {
-            "byte-range" => Ok(LockGranularityChoice::ByteRange),
-            "full" => Ok(LockGranularityChoice::Full),
-            _ => Err(LockGranularityParseError(s.to_owned())),
-        }
-    }
-}
-
-/// Returns a [`struct@libc::flock`] structure for the whole file.
-const fn get_flock(lock_type: LockType, granularity: LockGranularity) -> libc::flock {
-    libc::flock {
-        l_type: lock_type.to_libc_val() as libc::c_short,
-        l_whence: libc::SEEK_SET as libc::c_short,
-        l_start: granularity.l_start() as libc::c_long,
-        l_len: granularity.l_len() as libc::c_long,
-        l_pid: 0, /* filled by callee */
-    }
-}
-
-/// Tries to acquire a lock using [`fcntl`] with respect to the given
-/// parameters.
-///
-/// Please note that `fcntl()` OFD locks are **advisory locks**, which do not
-/// prevent to `open()` a file if a lock is already placed.
-///
-/// # Parameters
-/// - `file`: The file to acquire a lock for [`LockType`]. The file's state will
-///   be logically mutated, but not technically.
-/// - `lock_type`: The [`LockType`]
-/// - `granularity`: The [`LockGranularity`].
-pub fn try_acquire_lock<Fd: AsRawFd>(
-    file: &Fd,
-    lock_type: LockType,
-    granularity: LockGranularity,
-) -> Result<(), LockError> {
-    let flock = get_flock(lock_type, granularity);
-
-    let res = fcntl(file.as_raw_fd(), FcntlArg::F_OFD_SETLK(&flock));
-    match res {
-        0 => Ok(()),
-        -1 => {
-            let io_error = io::Error::last_os_error();
-            let errno = io_error.raw_os_error().unwrap();
-            match errno {
-                // See man page for error code:
-                // <https://man7.org/linux/man-pages/man2/fcntl.2.html>
-                libc::EAGAIN | libc::EACCES => Err(LockError::AlreadyLocked),
-                _ => Err(LockError::Io(io_error)),
-            }
-        }
-        val => panic!("Unexpected return value from fcntl(): {val}"),
-    }
-}
-
-/// Clears a lock.
-///
-/// # Parameters
-/// - `file`: The file to clear all locks for [`LockType`].
-/// - `granularity`: The [`LockGranularity`].
-pub fn clear_lock<Fd: AsRawFd>(file: &Fd, granularity: LockGranularity) -> Result<(), LockError> {
-    try_acquire_lock(file, LockType::Unlock, granularity)
-}
-
-/// Returns the current lock state using [`fcntl`] with respect to the given
-/// parameters.
-///
-/// # Parameters
-/// - `file`: The file for which to get the lock state.
-/// - `granularity`: The [`LockGranularity`].
-pub fn get_lock_state<Fd: AsRawFd>(
-    file: &Fd,
-    granularity: LockGranularity,
-) -> Result<LockState, LockError> {
-    let mut flock = get_flock(LockType::Write, granularity);
-    let res = fcntl(file.as_raw_fd(), FcntlArg::F_OFD_GETLK(&mut flock));
-    match res {
-        0 => {
-            let state = flock.l_type as libc::c_int;
-            let state = LockState::new(state);
-            Ok(state)
-        }
-        -1 => {
-            let io_error = io::Error::last_os_error();
-            Err(LockError::Io(io_error))
-        }
-        val => panic!("Unexpected return value from fcntl(): {val}"),
-    }
-}

block/src/fixed_vhd.rs

@@ -1,99 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::fs::File;
-use std::io::{Read, Seek, SeekFrom, Write};
-use std::os::unix::io::{AsRawFd, RawFd};
-
-use crate::BlockBackend;
-use crate::vhd::VhdFooter;
-
-#[derive(Debug)]
-pub struct FixedVhd {
-    file: File,
-    size: u64,
-    position: u64,
-}
-
-impl FixedVhd {
-    pub fn new(mut file: File) -> std::io::Result<Self> {
-        let footer = VhdFooter::new(&mut file)?;
-
-        Ok(Self {
-            file,
-            size: footer.current_size(),
-            position: 0,
-        })
-    }
-}
-
-impl AsRawFd for FixedVhd {
-    fn as_raw_fd(&self) -> RawFd {
-        self.file.as_raw_fd()
-    }
-}
-
-impl Read for FixedVhd {
-    fn read(&mut self, buf: &mut [u8]) -> std::io::Result<usize> {
-        match self.file.read(buf) {
-            Ok(r) => {
-                self.position = self.position.checked_add(r.try_into().unwrap()).unwrap();
-                Ok(r)
-            }
-            Err(e) => Err(e),
-        }
-    }
-}
-
-impl Write for FixedVhd {
-    fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
-        match self.file.write(buf) {
-            Ok(r) => {
-                self.position = self.position.checked_add(r.try_into().unwrap()).unwrap();
-                Ok(r)
-            }
-            Err(e) => Err(e),
-        }
-    }
-
-    fn flush(&mut self) -> std::io::Result<()> {
-        self.file.sync_all()
-    }
-}
-
-impl Seek for FixedVhd {
-    fn seek(&mut self, newpos: SeekFrom) -> std::io::Result<u64> {
-        match self.file.seek(newpos) {
-            Ok(pos) => {
-                self.position = pos;
-                Ok(pos)
-            }
-            Err(e) => Err(e),
-        }
-    }
-}
-
-impl BlockBackend for FixedVhd {
-    fn logical_size(&self) -> Result<u64, crate::Error> {
-        Ok(self.size)
-    }
-
-    /// Returns the physical size of the underlying file.
-    fn physical_size(&self) -> Result<u64, crate::Error> {
-        self.file
-            .metadata()
-            .map(|m| m.len())
-            .map_err(crate::Error::GetFileMetadata)
-    }
-}
-
-impl Clone for FixedVhd {
-    fn clone(&self) -> Self {
-        Self {
-            file: self.file.try_clone().expect("FixedVhd cloning failed"),
-            size: self.size,
-            position: self.position,
-        }
-    }
-}

block/src/fixed_vhd_disk.rs

@@ -1,222 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::fs::File;
-use std::io;
-use std::os::unix::io::AsRawFd;
-
-use crate::async_io::{AsyncIo, BorrowedDiskFd, DiskFileError};
-use crate::disk_file::DiskSize;
-use crate::error::{BlockError, BlockErrorKind, BlockResult, ErrorOp};
-use crate::fixed_vhd::FixedVhd;
-#[cfg(feature = "io_uring")]
-use crate::fixed_vhd_async::FixedVhdAsync;
-use crate::fixed_vhd_sync::FixedVhdSync;
-use crate::{BlockBackend, Error, disk_file};
-
-#[derive(Debug)]
-pub struct FixedVhdDisk {
-    inner: FixedVhd,
-    use_io_uring: bool,
-}
-
-impl FixedVhdDisk {
-    pub fn new(file: File, use_io_uring: bool) -> BlockResult<Self> {
-        #[cfg(not(feature = "io_uring"))]
-        if use_io_uring {
-            return Err(BlockError::new(
-                BlockErrorKind::UnsupportedFeature,
-                DiskFileError::NewAsyncIo(io::Error::other(
-                    "io_uring requested but feature is not enabled",
-                )),
-            ));
-        }
-
-        Ok(Self {
-            inner: FixedVhd::new(file).map_err(|e| BlockError::from(e).with_op(ErrorOp::Open))?,
-            use_io_uring,
-        })
-    }
-}
-
-impl disk_file::DiskSize for FixedVhdDisk {
-    fn logical_size(&self) -> BlockResult<u64> {
-        self.inner
-            .logical_size()
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, e))
-    }
-}
-
-impl disk_file::PhysicalSize for FixedVhdDisk {
-    fn physical_size(&self) -> BlockResult<u64> {
-        self.inner.physical_size().map_err(|e| match e {
-            Error::GetFileMetadata(io) => {
-                BlockError::new(BlockErrorKind::Io, Error::GetFileMetadata(io))
-            }
-            _ => unreachable!("unexpected error from FixedVhd::physical_size(): {e}"),
-        })
-    }
-}
-
-impl disk_file::DiskFd for FixedVhdDisk {
-    fn fd(&self) -> BorrowedDiskFd<'_> {
-        BorrowedDiskFd::new(self.inner.as_raw_fd())
-    }
-}
-
-impl disk_file::Geometry for FixedVhdDisk {}
-
-impl disk_file::SparseCapable for FixedVhdDisk {}
-
-impl disk_file::Resizable for FixedVhdDisk {
-    fn resize(&mut self, _size: u64) -> BlockResult<()> {
-        Err(BlockError::new(
-            BlockErrorKind::UnsupportedFeature,
-            DiskFileError::ResizeError(io::Error::other("resize not supported for fixed VHD")),
-        )
-        .with_op(ErrorOp::Resize))
-    }
-}
-
-impl disk_file::DiskFile for FixedVhdDisk {}
-
-impl disk_file::AsyncDiskFile for FixedVhdDisk {
-    fn try_clone(&self) -> BlockResult<Box<dyn disk_file::AsyncDiskFile>> {
-        Ok(Box::new(FixedVhdDisk {
-            inner: self.inner.clone(),
-            use_io_uring: self.use_io_uring,
-        }))
-    }
-
-    fn create_async_io(&self, ring_depth: u32) -> BlockResult<Box<dyn AsyncIo>> {
-        let size = self.logical_size()?;
-
-        if self.use_io_uring {
-            #[cfg(feature = "io_uring")]
-            {
-                return Ok(Box::new(FixedVhdAsync::new(
-                    self.inner.as_raw_fd(),
-                    ring_depth,
-                    size,
-                )?));
-            }
-
-            #[cfg(not(feature = "io_uring"))]
-            unreachable!("use_io_uring is set but io_uring feature is not enabled");
-        }
-
-        let _ = ring_depth;
-        Ok(Box::new(
-            FixedVhdSync::new(self.inner.as_raw_fd(), size).map_err(|e| {
-                BlockError::new(BlockErrorKind::Io, DiskFileError::NewAsyncIo(e))
-                    .with_op(ErrorOp::Open)
-            })?,
-        ))
-    }
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use std::fs::File;
-    use std::io::{Seek, SeekFrom, Write};
-
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::async_io::AsyncIo;
-    use crate::disk_file::{AsyncDiskFile, DiskSize, PhysicalSize, Resizable};
-
-    /// Minimal fixed VHD footer (disk type = 2, current_size = 0x11223344).
-    fn fixed_vhd_footer() -> &'static [u8] {
-        &[
-            0x63, 0x6f, 0x6e, 0x65, 0x63, 0x74, 0x69, 0x78, // cookie
-            0x00, 0x00, 0x00, 0x02, // features
-            0x00, 0x01, 0x00, 0x00, // file format version
-            0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, // data offset
-            0x27, 0xa6, 0xa6, 0x5d, // time stamp
-            0x71, 0x65, 0x6d, 0x75, // creator application
-            0x00, 0x05, 0x00, 0x03, // creator version
-            0x57, 0x69, 0x32, 0x6b, // creator host os
-            0x00, 0x00, 0x00, 0x00, 0x11, 0x22, 0x33, 0x44, // original size
-            0x00, 0x00, 0x00, 0x00, 0x11, 0x22, 0x33, 0x44, // current size
-            0x11, 0xe0, 0x10, 0x3f, // disk geometry
-            0x00, 0x00, 0x00, 0x02, // disk type
-            0x00, 0x00, 0x00, 0x00, // checksum
-            0x98, 0x7b, 0xb1, 0xcd, 0x84, 0x14, 0x41, 0xfc, // unique id
-            0xa4, 0xab, 0xd0, 0x69, 0x45, 0x2b, 0xf2, 0x23, 0x00, // saved state
-        ]
-    }
-
-    fn make_vhd_file() -> File {
-        let mut file: File = TempFile::new().unwrap().into_file();
-        let data_size: u64 = 0x1122_3344;
-        file.set_len(data_size + 0x200).unwrap();
-        file.seek(SeekFrom::Start(data_size)).unwrap();
-        file.write_all(fixed_vhd_footer()).unwrap();
-        file
-    }
-
-    #[test]
-    fn new_sync_returns_correct_size() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, false).unwrap();
-        assert_eq!(disk.logical_size().unwrap(), 0x1122_3344);
-    }
-
-    fn assert_async_io_from_dyn(disk: &dyn AsyncDiskFile, expect_batch: bool) {
-        let io: Box<dyn AsyncIo> = disk.create_async_io(128).unwrap();
-        assert_eq!(io.batch_requests_enabled(), expect_batch);
-    }
-
-    fn assert_async_io(disk: &FixedVhdDisk, expect_batch: bool) {
-        assert_async_io_from_dyn(disk, expect_batch);
-    }
-
-    #[test]
-    fn sync_backend_disables_batch_requests() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, false).unwrap();
-        assert_async_io(&disk, false);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn io_uring_backend_enables_batch_requests() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, true).unwrap();
-        assert_async_io(&disk, true);
-    }
-
-    #[test]
-    fn try_clone_preserves_sync_dispatch() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, false).unwrap();
-        let cloned = disk.try_clone().unwrap();
-        assert_async_io_from_dyn(cloned.as_ref(), false);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn try_clone_preserves_io_uring_dispatch() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, true).unwrap();
-        let cloned = disk.try_clone().unwrap();
-        assert_async_io_from_dyn(cloned.as_ref(), true);
-    }
-
-    #[test]
-    fn resize_returns_error() {
-        let file = make_vhd_file();
-        let mut disk = FixedVhdDisk::new(file, false).unwrap();
-        assert!(disk.resize(0x2000_0000).is_err());
-    }
-
-    #[test]
-    fn physical_size_includes_footer() {
-        let file = make_vhd_file();
-        let disk = FixedVhdDisk::new(file, false).unwrap();
-        // Data region (0x1122_3344) + VHD footer (0x200).
-        assert_eq!(disk.physical_size().unwrap(), 0x1122_3344 + 0x200);
-    }
-}

block/src/qcow/backing.rs

@@ -1,184 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-//! Thread safe backing file readers for QCOW2 images.
-
-use std::io;
-use std::os::fd::{AsFd, AsRawFd, BorrowedFd, OwnedFd};
-use std::sync::Arc;
-
-use crate::error::{BlockError, BlockErrorKind, BlockResult, ErrorOp};
-use crate::qcow::decoder::Decoder;
-use crate::qcow::metadata::{BackingRead, ClusterReadMapping, QcowMetadata};
-use crate::qcow::{BackingFile, BackingKind, Error as QcowError};
-use crate::qcow_common::{decompress_cluster, pread_alloc, pread_exact};
-
-/// Raw backing file using pread64 on a duplicated fd.
-pub(crate) struct RawBacking {
-    pub(crate) fd: OwnedFd,
-    pub(crate) virtual_size: u64,
-}
-
-// SAFETY: The only I/O operation is pread64 which is position independent
-// and safe for concurrent use from multiple threads.
-unsafe impl Sync for RawBacking {}
-
-impl BackingRead for RawBacking {
-    fn read_at(&self, address: u64, buf: &mut [u8]) -> io::Result<()> {
-        if address >= self.virtual_size {
-            buf.fill(0);
-            return Ok(());
-        }
-        let available = (self.virtual_size - address) as usize;
-        if available >= buf.len() {
-            pread_exact(self.fd.as_raw_fd(), buf, address)
-        } else {
-            pread_exact(self.fd.as_raw_fd(), &mut buf[..available], address)?;
-            buf[available..].fill(0);
-            Ok(())
-        }
-    }
-}
-
-/// QCOW2 image used as a backing file for another QCOW2 image.
-///
-/// Resolves guest offsets through the QCOW2 cluster mapping (L1/L2
-/// tables, refcounts) before reading the underlying data. Read only
-/// because backing files never receive writes. Nested backing chains
-/// are handled recursively via the optional `backing_file` field.
-pub(crate) struct Qcow2Backing {
-    pub(crate) metadata: Arc<QcowMetadata>,
-    pub(crate) data_fd: OwnedFd,
-    pub(crate) backing_file: Option<Arc<dyn BackingRead>>,
-    pub(crate) cluster_size: u64,
-    pub(crate) decoder: Arc<dyn Decoder>,
-}
-
-// SAFETY: All reads go through QcowMetadata which uses RwLock
-// and pread64 which is position independent and thread safe.
-unsafe impl Sync for Qcow2Backing {}
-
-impl BackingRead for Qcow2Backing {
-    fn read_at(&self, address: u64, buf: &mut [u8]) -> io::Result<()> {
-        let virtual_size = self.metadata.virtual_size();
-        if address >= virtual_size {
-            buf.fill(0);
-            return Ok(());
-        }
-        let available = (virtual_size - address) as usize;
-        if available < buf.len() {
-            self.read_clusters(address, &mut buf[..available])?;
-            buf[available..].fill(0);
-            return Ok(());
-        }
-        self.read_clusters(address, buf)
-    }
-}
-
-impl Qcow2Backing {
-    /// Resolve cluster mappings via metadata then read allocated clusters
-    /// with pread64.
-    fn read_clusters(&self, address: u64, buf: &mut [u8]) -> io::Result<()> {
-        let total_len = buf.len();
-        let has_backing = self.backing_file.is_some();
-
-        let mappings = self
-            .metadata
-            .map_clusters_for_read(address, total_len, has_backing)?;
-
-        let mut buf_offset = 0usize;
-        for mapping in mappings {
-            match mapping {
-                ClusterReadMapping::Zero { length } => {
-                    buf[buf_offset..buf_offset + length as usize].fill(0);
-                    buf_offset += length as usize;
-                }
-                ClusterReadMapping::Allocated {
-                    offset: host_offset,
-                    length,
-                } => {
-                    pread_exact(
-                        self.data_fd.as_raw_fd(),
-                        &mut buf[buf_offset..buf_offset + length as usize],
-                        host_offset,
-                    )?;
-                    buf_offset += length as usize;
-                }
-                ClusterReadMapping::Compressed {
-                    host_offset,
-                    compressed_size,
-                    cluster_offset,
-                    length,
-                } => {
-                    let compressed =
-                        pread_alloc(self.data_fd.as_raw_fd(), host_offset, compressed_size)?;
-                    let decompressed = decompress_cluster(
-                        &compressed,
-                        self.cluster_size as usize,
-                        &*self.decoder,
-                    )?;
-                    buf[buf_offset..buf_offset + length]
-                        .copy_from_slice(&decompressed[cluster_offset..cluster_offset + length]);
-                    buf_offset += length;
-                }
-                ClusterReadMapping::Backing {
-                    offset: backing_offset,
-                    length,
-                } => {
-                    self.backing_file.as_ref().unwrap().read_at(
-                        backing_offset,
-                        &mut buf[buf_offset..buf_offset + length as usize],
-                    )?;
-                    buf_offset += length as usize;
-                }
-            }
-        }
-        Ok(())
-    }
-}
-
-impl Drop for Qcow2Backing {
-    fn drop(&mut self) {
-        self.metadata.shutdown();
-    }
-}
-
-/// Construct a thread safe backing file reader.
-pub fn shared_backing_from(bf: BackingFile) -> BlockResult<Arc<dyn BackingRead>> {
-    let (kind, virtual_size) = bf.into_kind();
-
-    let dup_fd = |fd: BorrowedFd<'_>| -> BlockResult<OwnedFd> {
-        fd.try_clone_to_owned().map_err(|e| {
-            BlockError::new(
-                BlockErrorKind::Io,
-                QcowError::BackingFileIo(String::new(), e),
-            )
-            .with_op(ErrorOp::DupBackingFd)
-        })
-    };
-
-    match kind {
-        BackingKind::Raw(raw_file) => {
-            let fd = dup_fd(raw_file.as_fd())?;
-            Ok(Arc::new(RawBacking { fd, virtual_size }))
-        }
-        BackingKind::Qcow { inner, backing } => {
-            let data_fd = dup_fd(inner.raw_file.as_fd())?;
-            let metadata = Arc::new(QcowMetadata::new(*inner));
-            Ok(Arc::new(Qcow2Backing {
-                cluster_size: metadata.cluster_size(),
-                decoder: metadata.decoder(),
-                metadata,
-                data_fd,
-                backing_file: backing.map(|bf| shared_backing_from(*bf)).transpose()?,
-            }))
-        }
-        #[cfg(test)]
-        BackingKind::QcowFile(_) => {
-            unreachable!("QcowFile variant is only used by set_backing_file() in tests")
-        }
-    }
-}

block/src/qcow/decoder.rs

@@ -1,87 +0,0 @@
-// Copyright 2025 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use thiserror::Error;
-
-#[derive(Debug, Error)]
-pub enum Error {
-    #[error("Zlib decompress error")]
-    ZlibDecompress(#[source] flate2::DecompressError),
-    #[error("Zlib unexpected status: {0:?}")]
-    ZlibUnexpectedStatus(flate2::Status),
-    #[error("Zstd decompress error")]
-    ZstdDecompress(#[source] std::io::Error),
-    #[error("Zstd: failed to fill buffer")]
-    ZstdFillBuffer(#[source] std::io::Error),
-}
-
-pub type Result<T> = std::result::Result<T, Error>;
-
-/// Generic trait for decoding zlib/zstd formats
-pub trait Decoder: Send + Sync {
-    fn decode(&self, input: &[u8], output: &mut [u8]) -> Result<usize>;
-}
-
-#[derive(Default)]
-pub struct ZlibDecoder {}
-
-impl Decoder for ZlibDecoder {
-    fn decode(&self, input: &[u8], output: &mut [u8]) -> Result<usize> {
-        use flate2::{Decompress, FlushDecompress, Status};
-
-        let mut decompressor = Decompress::new(false);
-        let status = decompressor
-            .decompress(input, output, FlushDecompress::Finish)
-            .map_err(Error::ZlibDecompress)?;
-        if status == Status::StreamEnd {
-            Ok(decompressor.total_out() as usize)
-        } else {
-            Err(Error::ZlibUnexpectedStatus(status))
-        }
-    }
-}
-
-#[derive(Default)]
-pub struct ZstdDecoder {}
-
-impl Decoder for ZstdDecoder {
-    fn decode(&self, input: &[u8], output: &mut [u8]) -> Result<usize> {
-        use std::io::Read;
-
-        let mut decoder = zstd::stream::read::Decoder::new(input).map_err(Error::ZstdDecompress)?;
-        let decoded_size = decoder.read(output).map_err(Error::ZstdFillBuffer)?;
-        Ok(decoded_size)
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn test_zlib_decode() {
-        let d = ZlibDecoder::default();
-        let valid_input = vec![99, 96, 100, 98, 6, 0];
-        let mut output1 = vec![0; 4];
-        d.decode(&valid_input, &mut output1).unwrap();
-        assert_eq!(&output1, b"\x00\x01\x02\x03");
-
-        let invalid_input = vec![1, 2, 3, 4];
-        let mut output2 = vec![0; 1024];
-        d.decode(&invalid_input, &mut output2).unwrap_err();
-    }
-
-    #[test]
-    fn test_zstd_decode() {
-        let d = ZstdDecoder::default();
-        let valid_input = vec![40, 181, 47, 253, 32, 2, 17, 0, 0, 1, 254];
-        let mut output1 = vec![0; 2];
-        d.decode(&valid_input, &mut output1).unwrap();
-        assert_eq!(&output1, b"\x01\xfe");
-
-        let invalid_input = vec![1, 2, 3, 4];
-        let mut output2 = vec![0; 1024];
-        d.decode(&invalid_input, &mut output2).unwrap_err();
-    }
-}

block/src/qcow/header.rs

@@ -1,605 +0,0 @@
-// Copyright 2018 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 LICENSE-BSD-3-Clause file.
-//
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-//! QCOW2 header parsing, validation, and creation.
-
-use std::fmt::{Display, Formatter, Result as FmtResult};
-use std::io::{Read, Seek, SeekFrom, Write};
-use std::mem::size_of;
-use std::str::FromStr;
-
-use bitflags::bitflags;
-use vmm_sys_util::file_traits::FileSync;
-
-use super::decoder::{Decoder, ZlibDecoder, ZstdDecoder};
-use super::qcow_raw_file::BeUint;
-use super::raw_file::RawFile;
-use super::{Error, Result, div_round_up_u32, div_round_up_u64};
-use crate::error::{BlockError, BlockErrorKind, BlockResult};
-
-#[derive(Copy, Clone, Debug, PartialEq, Eq)]
-pub enum ImageType {
-    Raw,
-    Qcow2,
-}
-
-impl Display for ImageType {
-    fn fmt(&self, f: &mut Formatter<'_>) -> FmtResult {
-        match self {
-            ImageType::Raw => write!(f, "raw"),
-            ImageType::Qcow2 => write!(f, "qcow2"),
-        }
-    }
-}
-
-impl FromStr for ImageType {
-    type Err = Error;
-
-    fn from_str(s: &str) -> Result<Self> {
-        match s {
-            "raw" => Ok(ImageType::Raw),
-            "qcow2" => Ok(ImageType::Qcow2),
-            _ => Err(Error::UnsupportedBackingFileFormat(s.to_string())),
-        }
-    }
-}
-
-#[derive(Clone, Debug)]
-pub enum CompressionType {
-    Zlib,
-    Zstd,
-}
-
-#[derive(Debug, Clone)]
-pub struct BackingFileConfig {
-    pub path: String,
-    // If this is None, we will autodetect it.
-    pub format: Option<ImageType>,
-}
-
-// Maximum data size supported.
-pub(super) const MAX_QCOW_FILE_SIZE: u64 = 0x01 << 44; // 16 TB.
-
-// QCOW magic constant that starts the header.
-pub(super) const QCOW_MAGIC: u32 = 0x5146_49fb;
-// Default to a cluster size of 2^DEFAULT_CLUSTER_BITS
-pub(super) const DEFAULT_CLUSTER_BITS: u32 = 16;
-// Limit clusters to reasonable sizes. Choose the same limits as qemu. Making the clusters smaller
-// increases the amount of overhead for book keeping.
-pub(super) const MIN_CLUSTER_BITS: u32 = 9;
-pub(super) const MAX_CLUSTER_BITS: u32 = 21;
-// The L1 and RefCount table are kept in RAM, only handle files that require less than 35M entries.
-// This easily covers 1 TB files. When support for bigger files is needed the assumptions made to
-// keep these tables in RAM needs to be thrown out.
-pub(super) const MAX_RAM_POINTER_TABLE_SIZE: u64 = 35_000_000;
-// 16-bit refcounts.
-pub(super) const DEFAULT_REFCOUNT_ORDER: u32 = 4;
-
-pub(super) const V2_BARE_HEADER_SIZE: u32 = 72;
-pub(super) const V3_BARE_HEADER_SIZE: u32 = 104;
-pub(super) const AUTOCLEAR_FEATURES_OFFSET: u64 = 88;
-
-pub(super) const COMPATIBLE_FEATURES_LAZY_REFCOUNTS: u64 = 1;
-
-// Compression types as defined in https://www.qemu.org/docs/master/interop/qcow2.html
-const COMPRESSION_TYPE_ZLIB: u64 = 0; // zlib/deflate <https://www.ietf.org/rfc/rfc1951.txt>
-const COMPRESSION_TYPE_ZSTD: u64 = 1; // zstd <http://github.com/facebook/zstd>
-
-// Header extension types
-pub(super) const HEADER_EXT_END: u32 = 0x00000000;
-// Backing file format name (raw, qcow2)
-pub(super) const HEADER_EXT_BACKING_FORMAT: u32 = 0xe2792aca;
-// Feature name table
-const HEADER_EXT_FEATURE_NAME_TABLE: u32 = 0x6803f857;
-
-// Feature name table entry type incompatible
-const FEAT_TYPE_INCOMPATIBLE: u8 = 0;
-
-bitflags! {
-    #[derive(Debug, Clone, Copy, PartialEq, Eq)]
-    pub struct IncompatFeatures: u64 {
-        const DIRTY = 1 << 0;
-        const CORRUPT = 1 << 1;
-        const DATA_FILE = 1 << 2;
-        const COMPRESSION = 1 << 3;
-        const EXTENDED_L2 = 1 << 4;
-    }
-}
-
-impl IncompatFeatures {
-    /// Features supported by this implementation.
-    pub(super) const SUPPORTED: IncompatFeatures = IncompatFeatures::DIRTY
-        .union(IncompatFeatures::CORRUPT)
-        .union(IncompatFeatures::COMPRESSION);
-
-    /// Get the fallback name for a known feature bit.
-    fn flag_name(bit: u8) -> Option<&'static str> {
-        Some(match Self::from_bits_truncate(1u64 << bit) {
-            Self::DIRTY => "dirty bit",
-            Self::CORRUPT => "corrupt bit",
-            Self::DATA_FILE => "external data file",
-            Self::EXTENDED_L2 => "extended L2 entries",
-            _ => return None,
-        })
-    }
-}
-
-/// Error type for unsupported incompatible features.
-#[derive(Debug, Clone, thiserror::Error)]
-pub struct MissingFeatureError {
-    /// Unsupported feature bits.
-    features: IncompatFeatures,
-    /// Feature name table from the qcow2 image.
-    feature_names: Vec<(u8, String)>,
-}
-
-impl MissingFeatureError {
-    pub(super) fn new(features: IncompatFeatures, feature_names: Vec<(u8, String)>) -> Self {
-        Self {
-            features,
-            feature_names,
-        }
-    }
-}
-
-impl Display for MissingFeatureError {
-    fn fmt(&self, f: &mut Formatter<'_>) -> FmtResult {
-        let names: Vec<String> = (0u8..64)
-            .filter(|&bit| self.features.bits() & (1u64 << bit) != 0)
-            .map(|bit| {
-                // First try the image's feature name table
-                self.feature_names
-                    .iter()
-                    .find(|(b, _)| *b == bit)
-                    .map(|(_, name)| name.clone())
-                    // Then try hardcoded fallback names
-                    .or_else(|| IncompatFeatures::flag_name(bit).map(|s| s.to_string()))
-                    // Finally, use generic description
-                    .unwrap_or_else(|| format!("unknown feature bit {bit}"))
-            })
-            .collect();
-        write!(f, "Missing features: {}", names.join(", "))
-    }
-}
-
-// The format supports a "header extension area", that crosvm does not use.
-const QCOW_EMPTY_HEADER_EXTENSION_SIZE: u32 = 8;
-
-// Defined by the specification
-const MAX_BACKING_FILE_SIZE: u32 = 1023;
-
-/// Contains the information from the header of a qcow file.
-#[derive(Clone, Debug)]
-pub struct QcowHeader {
-    pub magic: u32,
-    pub version: u32,
-
-    pub backing_file_offset: u64,
-    pub backing_file_size: u32,
-
-    pub cluster_bits: u32,
-    pub size: u64,
-    pub crypt_method: u32,
-
-    pub l1_size: u32,
-    pub l1_table_offset: u64,
-
-    pub refcount_table_offset: u64,
-    pub refcount_table_clusters: u32,
-
-    pub nb_snapshots: u32,
-    pub snapshots_offset: u64,
-
-    // v3 entries
-    pub incompatible_features: u64,
-    pub compatible_features: u64,
-    pub autoclear_features: u64,
-    pub refcount_order: u32,
-    pub header_size: u32,
-    pub compression_type: CompressionType,
-
-    // Post-header entries
-    pub backing_file: Option<BackingFileConfig>,
-}
-
-impl QcowHeader {
-    /// Read header extensions, optionally collecting feature names for error reporting.
-    pub(super) fn read_header_extensions(
-        f: &mut RawFile,
-        header: &mut QcowHeader,
-        mut feature_table: Option<&mut Vec<(u8, String)>>,
-    ) -> Result<()> {
-        // Extensions start directly after the header
-        f.seek(SeekFrom::Start(header.header_size as u64))
-            .map_err(Error::ReadingHeader)?;
-
-        loop {
-            let ext_type = u32::read_be(f).map_err(Error::ReadingHeader)?;
-            if ext_type == HEADER_EXT_END {
-                break;
-            }
-
-            let ext_length = u32::read_be(f).map_err(Error::ReadingHeader)?;
-
-            match ext_type {
-                HEADER_EXT_BACKING_FORMAT => {
-                    let mut format_bytes = vec![0u8; ext_length as usize];
-                    f.read_exact(&mut format_bytes)
-                        .map_err(Error::ReadingHeader)?;
-                    let format_str = String::from_utf8(format_bytes)
-                        .map_err(|err| Error::InvalidBackingFileName(err.utf8_error()))?;
-                    if let Some(backing_file) = &mut header.backing_file {
-                        backing_file.format = Some(format_str.parse()?);
-                    }
-                }
-                HEADER_EXT_FEATURE_NAME_TABLE if feature_table.is_some() => {
-                    const FEATURE_NAME_ENTRY_SIZE: usize = 1 + 1 + 46; // type + bit + name
-                    let mut data = vec![0u8; ext_length as usize];
-                    f.read_exact(&mut data).map_err(Error::ReadingHeader)?;
-                    let table = feature_table.as_mut().unwrap();
-                    for entry in data.chunks_exact(FEATURE_NAME_ENTRY_SIZE) {
-                        if entry[0] == FEAT_TYPE_INCOMPATIBLE {
-                            let bit_number = entry[1];
-                            let name_bytes = &entry[2..];
-                            let name_len = name_bytes.iter().position(|&b| b == 0).unwrap_or(46);
-                            let name = String::from_utf8_lossy(&name_bytes[..name_len]).to_string();
-                            table.push((bit_number, name));
-                        }
-                    }
-                }
-                _ => {
-                    // Skip unknown extension
-                    f.seek(SeekFrom::Current(ext_length as i64))
-                        .map_err(Error::ReadingHeader)?;
-                }
-            }
-
-            // Skip to the next 8 byte boundary
-            let padding = (8 - (ext_length % 8)) % 8;
-            f.seek(SeekFrom::Current(padding as i64))
-                .map_err(Error::ReadingHeader)?;
-        }
-
-        Ok(())
-    }
-
-    /// Creates a QcowHeader from a reference to a file.
-    pub fn new(f: &mut RawFile) -> Result<QcowHeader> {
-        f.rewind().map_err(Error::ReadingHeader)?;
-        let magic = u32::read_be(f).map_err(Error::ReadingHeader)?;
-        if magic != QCOW_MAGIC {
-            return Err(Error::InvalidMagic);
-        }
-
-        // Reads the next u32 from the file.
-        fn read_u32_be(f: &mut RawFile) -> Result<u32> {
-            u32::read_be(f).map_err(Error::ReadingHeader)
-        }
-
-        // Reads the next u64 from the file.
-        fn read_u64_be(f: &mut RawFile) -> Result<u64> {
-            u64::read_be(f).map_err(Error::ReadingHeader)
-        }
-
-        let version = read_u32_be(f)?;
-
-        let mut header = QcowHeader {
-            magic,
-            version,
-            backing_file_offset: read_u64_be(f)?,
-            backing_file_size: read_u32_be(f)?,
-            cluster_bits: read_u32_be(f)?,
-            size: read_u64_be(f)?,
-            crypt_method: read_u32_be(f)?,
-            l1_size: read_u32_be(f)?,
-            l1_table_offset: read_u64_be(f)?,
-            refcount_table_offset: read_u64_be(f)?,
-            refcount_table_clusters: read_u32_be(f)?,
-            nb_snapshots: read_u32_be(f)?,
-            snapshots_offset: read_u64_be(f)?,
-            incompatible_features: if version == 2 { 0 } else { read_u64_be(f)? },
-            compatible_features: if version == 2 { 0 } else { read_u64_be(f)? },
-            autoclear_features: if version == 2 { 0 } else { read_u64_be(f)? },
-            refcount_order: if version == 2 {
-                DEFAULT_REFCOUNT_ORDER
-            } else {
-                read_u32_be(f)?
-            },
-            header_size: if version == 2 {
-                V2_BARE_HEADER_SIZE
-            } else {
-                read_u32_be(f)?
-            },
-            compression_type: CompressionType::Zlib,
-            backing_file: None,
-        };
-        if version == 3 && header.header_size > V3_BARE_HEADER_SIZE {
-            let raw_compression_type = read_u64_be(f)? >> (64 - 8);
-            header.compression_type = if raw_compression_type == COMPRESSION_TYPE_ZLIB {
-                Ok(CompressionType::Zlib)
-            } else if raw_compression_type == COMPRESSION_TYPE_ZSTD {
-                Ok(CompressionType::Zstd)
-            } else {
-                Err(Error::UnsupportedCompressionType)
-            }?;
-        }
-        if header.backing_file_size > MAX_BACKING_FILE_SIZE {
-            return Err(Error::BackingFileTooLong(header.backing_file_size as usize));
-        }
-        if header.backing_file_offset != 0 {
-            f.seek(SeekFrom::Start(header.backing_file_offset))
-                .map_err(Error::ReadingHeader)?;
-            let mut backing_file_name_bytes = vec![0u8; header.backing_file_size as usize];
-            f.read_exact(&mut backing_file_name_bytes)
-                .map_err(Error::ReadingHeader)?;
-            let path = String::from_utf8(backing_file_name_bytes)
-                .map_err(|err| Error::InvalidBackingFileName(err.utf8_error()))?;
-            header.backing_file = Some(BackingFileConfig { path, format: None });
-        }
-
-        if version == 3 {
-            // Check for unsupported incompatible features first
-            let features = IncompatFeatures::from_bits_retain(header.incompatible_features);
-            let unsupported = features - IncompatFeatures::SUPPORTED;
-            if !unsupported.is_empty() {
-                // Read extensions only to get feature names for error reporting
-                let mut feature_table = Vec::new();
-                if header.header_size > V3_BARE_HEADER_SIZE {
-                    let _ = Self::read_header_extensions(f, &mut header, Some(&mut feature_table));
-                }
-                return Err(Error::UnsupportedFeature(MissingFeatureError::new(
-                    unsupported,
-                    feature_table,
-                )));
-            }
-
-            // Features OK, now read extensions normally
-            if header.header_size > V3_BARE_HEADER_SIZE {
-                Self::read_header_extensions(f, &mut header, None)?;
-            }
-        }
-
-        Ok(header)
-    }
-
-    pub fn get_decoder(&self) -> Box<dyn Decoder> {
-        match self.compression_type {
-            CompressionType::Zlib => Box::new(ZlibDecoder {}),
-            CompressionType::Zstd => Box::new(ZstdDecoder {}),
-        }
-    }
-
-    pub fn create_for_size_and_path(
-        version: u32,
-        size: u64,
-        backing_file: Option<&str>,
-    ) -> Result<QcowHeader> {
-        let header_size = if version == 2 {
-            V2_BARE_HEADER_SIZE
-        } else {
-            V3_BARE_HEADER_SIZE + QCOW_EMPTY_HEADER_EXTENSION_SIZE
-        };
-        let cluster_bits: u32 = DEFAULT_CLUSTER_BITS;
-        let cluster_size: u32 = 0x01 << cluster_bits;
-        let max_length: usize = (cluster_size - header_size) as usize;
-        if let Some(path) = backing_file
-            && path.len() > max_length
-        {
-            return Err(Error::BackingFileTooLong(path.len() - max_length));
-        }
-
-        // L2 blocks are always one cluster long. They contain cluster_size/sizeof(u64) addresses.
-        let entries_per_cluster: u32 = cluster_size / size_of::<u64>() as u32;
-        let num_clusters: u32 = div_round_up_u64(size, u64::from(cluster_size)) as u32;
-        let num_l2_clusters: u32 = div_round_up_u32(num_clusters, entries_per_cluster);
-        let l1_clusters: u32 = div_round_up_u32(num_l2_clusters, entries_per_cluster);
-        let header_clusters = div_round_up_u32(size_of::<QcowHeader>() as u32, cluster_size);
-        Ok(QcowHeader {
-            magic: QCOW_MAGIC,
-            version,
-            backing_file_offset: backing_file.map_or(0, |_| {
-                header_size
-                    + if version == 3 {
-                        QCOW_EMPTY_HEADER_EXTENSION_SIZE
-                    } else {
-                        0
-                    }
-            }) as u64,
-            backing_file_size: backing_file.map_or(0, |x| x.len()) as u32,
-            cluster_bits: DEFAULT_CLUSTER_BITS,
-            size,
-            crypt_method: 0,
-            l1_size: num_l2_clusters,
-            l1_table_offset: u64::from(cluster_size),
-            // The refcount table is after l1 + header.
-            refcount_table_offset: u64::from(cluster_size * (l1_clusters + 1)),
-            refcount_table_clusters: {
-                // Pre-allocate enough clusters for the entire refcount table as it must be
-                // continuous in the file. Allocate enough space to refcount all clusters, including
-                // the refcount clusters.
-                let max_refcount_clusters = max_refcount_clusters(
-                    DEFAULT_REFCOUNT_ORDER,
-                    cluster_size,
-                    num_clusters + l1_clusters + num_l2_clusters + header_clusters,
-                ) as u32;
-                // The refcount table needs to store the offset of each refcount cluster.
-                div_round_up_u32(
-                    max_refcount_clusters * size_of::<u64>() as u32,
-                    cluster_size,
-                )
-            },
-            nb_snapshots: 0,
-            snapshots_offset: 0,
-            incompatible_features: 0,
-            compatible_features: 0,
-            autoclear_features: 0,
-            refcount_order: DEFAULT_REFCOUNT_ORDER,
-            header_size,
-            compression_type: CompressionType::Zlib,
-            backing_file: backing_file.map(|path| BackingFileConfig {
-                path: String::from(path),
-                format: None,
-            }),
-        })
-    }
-
-    /// Write the header to `file`.
-    pub fn write_to<F: Write + Seek>(&self, file: &mut F) -> Result<()> {
-        // Writes the next u32 to the file.
-        fn write_u32_be<F: Write>(f: &mut F, value: u32) -> Result<()> {
-            u32::write_be(f, value).map_err(Error::WritingHeader)
-        }
-
-        // Writes the next u64 to the file.
-        fn write_u64_be<F: Write>(f: &mut F, value: u64) -> Result<()> {
-            u64::write_be(f, value).map_err(Error::WritingHeader)
-        }
-
-        write_u32_be(file, self.magic)?;
-        write_u32_be(file, self.version)?;
-        write_u64_be(file, self.backing_file_offset)?;
-        write_u32_be(file, self.backing_file_size)?;
-        write_u32_be(file, self.cluster_bits)?;
-        write_u64_be(file, self.size)?;
-        write_u32_be(file, self.crypt_method)?;
-        write_u32_be(file, self.l1_size)?;
-        write_u64_be(file, self.l1_table_offset)?;
-        write_u64_be(file, self.refcount_table_offset)?;
-        write_u32_be(file, self.refcount_table_clusters)?;
-        write_u32_be(file, self.nb_snapshots)?;
-        write_u64_be(file, self.snapshots_offset)?;
-
-        if self.version == 3 {
-            write_u64_be(file, self.incompatible_features)?;
-            write_u64_be(file, self.compatible_features)?;
-            write_u64_be(file, self.autoclear_features)?;
-            write_u32_be(file, self.refcount_order)?;
-            write_u32_be(file, self.header_size)?;
-
-            if self.header_size > V3_BARE_HEADER_SIZE {
-                write_u64_be(file, 0)?; // no compression
-            }
-
-            write_u32_be(file, 0)?; // header extension type: end of header extension area
-            write_u32_be(file, 0)?; // length of header extension data: 0
-        }
-
-        if let Some(backing_file_path) = self.backing_file.as_ref().map(|bf| &bf.path) {
-            if self.backing_file_offset > 0 {
-                file.seek(SeekFrom::Start(self.backing_file_offset))
-                    .map_err(Error::WritingHeader)?;
-            }
-            write!(file, "{backing_file_path}").map_err(Error::WritingHeader)?;
-        }
-
-        // Set the file length by seeking and writing a zero to the last byte. This avoids needing
-        // a `File` instead of anything that implements seek as the `file` argument.
-        // Zeros out the l1 and refcount table clusters.
-        let cluster_size = 0x01u64 << self.cluster_bits;
-        let refcount_blocks_size = u64::from(self.refcount_table_clusters) * cluster_size;
-        file.seek(SeekFrom::Start(
-            self.refcount_table_offset + refcount_blocks_size - 2,
-        ))
-        .map_err(Error::WritingHeader)?;
-        file.write(&[0u8]).map_err(Error::WritingHeader)?;
-
-        Ok(())
-    }
-
-    /// Write only the incompatible_features field to the file at its fixed offset.
-    fn write_incompatible_features<F: Seek + Write>(&self, file: &mut F) -> BlockResult<()> {
-        if self.version != 3 {
-            return Ok(());
-        }
-        file.seek(SeekFrom::Start(V2_BARE_HEADER_SIZE as u64))
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, Error::WritingHeader(e)))?;
-        u64::write_be(file, self.incompatible_features)
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, Error::WritingHeader(e)))?;
-        Ok(())
-    }
-
-    /// Set or clear the dirty bit for QCOW2 v3 images.
-    ///
-    /// When `dirty` is true, sets the bit to indicate the image is in use.
-    /// When `dirty` is false, clears the bit to indicate a clean shutdown.
-    pub fn set_dirty_bit<F: Seek + Write + FileSync>(
-        &mut self,
-        file: &mut F,
-        dirty: bool,
-    ) -> BlockResult<()> {
-        if self.version == 3 {
-            if dirty {
-                self.incompatible_features |= IncompatFeatures::DIRTY.bits();
-            } else {
-                self.incompatible_features &= !IncompatFeatures::DIRTY.bits();
-            }
-            self.write_incompatible_features(file)?;
-            file.fsync()
-                .map_err(|e| BlockError::new(BlockErrorKind::Io, Error::SyncingHeader(e)))?;
-        }
-        Ok(())
-    }
-
-    /// Set the corrupt bit for QCOW2 v3 images.
-    ///
-    /// This marks the image as corrupted. Once set, the image can only be
-    /// opened read-only until repaired.
-    pub fn set_corrupt_bit<F: Seek + Write + FileSync>(&mut self, file: &mut F) -> BlockResult<()> {
-        if self.version == 3 {
-            self.incompatible_features |= IncompatFeatures::CORRUPT.bits();
-            self.write_incompatible_features(file)?;
-            file.fsync()
-                .map_err(|e| BlockError::new(BlockErrorKind::Io, Error::SyncingHeader(e)))?;
-        }
-        Ok(())
-    }
-
-    pub fn is_corrupt(&self) -> bool {
-        IncompatFeatures::from_bits_truncate(self.incompatible_features)
-            .contains(IncompatFeatures::CORRUPT)
-    }
-
-    /// Clear all autoclear feature bits for QCOW2 v3 images.
-    ///
-    /// These bits indicate features that can be safely disabled when modified
-    /// by software that doesn't understand them.
-    pub fn clear_autoclear_features<F: Seek + Write + FileSync>(
-        &mut self,
-        file: &mut F,
-    ) -> Result<()> {
-        if self.version == 3 && self.autoclear_features != 0 {
-            self.autoclear_features = 0;
-            file.seek(SeekFrom::Start(AUTOCLEAR_FEATURES_OFFSET))
-                .map_err(Error::WritingHeader)?;
-            u64::write_be(file, 0).map_err(Error::WritingHeader)?;
-            file.fsync().map_err(Error::SyncingHeader)?;
-        }
-        Ok(())
-    }
-}
-
-pub(super) fn max_refcount_clusters(
-    refcount_order: u32,
-    cluster_size: u32,
-    num_clusters: u32,
-) -> u64 {
-    // Use u64 as the product of the u32 inputs can overflow.
-    let refcount_bits = 0x01u64 << u64::from(refcount_order);
-    let cluster_bits = u64::from(cluster_size) * 8;
-    let for_data = div_round_up_u64(u64::from(num_clusters) * refcount_bits, cluster_bits);
-    let for_refcounts = div_round_up_u64(for_data * refcount_bits, cluster_bits);
-    for_data + for_refcounts
-}
-
-/// Returns an Error if the given offset doesn't align to a cluster boundary.
-pub(super) fn offset_is_cluster_boundary(offset: u64, cluster_bits: u32) -> Result<()> {
-    if offset & ((0x01 << cluster_bits) - 1) != 0 {
-        return Err(Error::InvalidOffset(offset));
-    }
-    Ok(())
-}

block/src/qcow/qcow_raw_file.rs

@@ -1,369 +0,0 @@
-// Copyright 2018 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 LICENSE-BSD-3-Clause file.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::fmt::Debug;
-use std::io::{self, BufWriter, Read, Seek, SeekFrom, Write};
-use std::mem::size_of;
-use std::os::fd::{AsFd, AsRawFd, BorrowedFd, RawFd};
-
-use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
-use vmm_sys_util::write_zeroes::WriteZeroes;
-
-use super::RawFile;
-
-// Type aliases for the refcount read/write function pointers
-type RefcountReader = fn(&mut RawFile, usize) -> io::Result<Vec<u64>>;
-type RefcountWriter = fn(&mut RawFile, &[u64]) -> io::Result<()>;
-
-/// Big-endian file access trait.
-pub(super) trait BeUint: Sized + Copy {
-    fn from_be_slice(bytes: &[u8]) -> u64;
-    fn read_be<R: Read>(r: &mut R) -> io::Result<Self>;
-    fn write_be<W: Write>(w: &mut W, val: Self) -> io::Result<()>;
-}
-
-impl BeUint for u8 {
-    #[inline(always)]
-    fn from_be_slice(bytes: &[u8]) -> u64 {
-        bytes[0] as u64
-    }
-    #[inline(always)]
-    fn read_be<R: Read>(r: &mut R) -> io::Result<Self> {
-        r.read_u8()
-    }
-    #[inline(always)]
-    fn write_be<W: Write>(w: &mut W, val: Self) -> io::Result<()> {
-        w.write_u8(val)
-    }
-}
-
-impl BeUint for u16 {
-    #[inline(always)]
-    fn from_be_slice(bytes: &[u8]) -> u64 {
-        u16::from_be_bytes([bytes[0], bytes[1]]) as u64
-    }
-    #[inline(always)]
-    fn read_be<R: Read>(r: &mut R) -> io::Result<Self> {
-        r.read_u16::<BigEndian>()
-    }
-    #[inline(always)]
-    fn write_be<W: Write>(w: &mut W, val: Self) -> io::Result<()> {
-        w.write_u16::<BigEndian>(val)
-    }
-}
-
-impl BeUint for u32 {
-    #[inline(always)]
-    fn from_be_slice(bytes: &[u8]) -> u64 {
-        u32::from_be_bytes([bytes[0], bytes[1], bytes[2], bytes[3]]) as u64
-    }
-    #[inline(always)]
-    fn read_be<R: Read>(r: &mut R) -> io::Result<Self> {
-        r.read_u32::<BigEndian>()
-    }
-    #[inline(always)]
-    fn write_be<W: Write>(w: &mut W, val: Self) -> io::Result<()> {
-        w.write_u32::<BigEndian>(val)
-    }
-}
-
-impl BeUint for u64 {
-    #[inline(always)]
-    fn from_be_slice(bytes: &[u8]) -> u64 {
-        u64::from_be_bytes([
-            bytes[0], bytes[1], bytes[2], bytes[3], bytes[4], bytes[5], bytes[6], bytes[7],
-        ])
-    }
-    #[inline(always)]
-    fn read_be<R: Read>(r: &mut R) -> io::Result<Self> {
-        r.read_u64::<BigEndian>()
-    }
-    #[inline(always)]
-    fn write_be<W: Write>(w: &mut W, val: Self) -> io::Result<()> {
-        w.write_u64::<BigEndian>(val)
-    }
-}
-
-/// Read byte-aligned refcounts.
-fn read_refcount<T: BeUint>(file: &mut RawFile, count: usize) -> io::Result<Vec<u64>> {
-    let bytes_per_entry = size_of::<T>();
-    let mut data = vec![0u8; count * bytes_per_entry];
-    file.read_exact(&mut data)?;
-    Ok(data
-        .chunks_exact(bytes_per_entry)
-        .map(T::from_be_slice)
-        .collect())
-}
-
-/// Write byte-aligned refcounts.
-fn write_refcount<T: BeUint + TryFrom<u64>>(file: &mut RawFile, table: &[u64]) -> io::Result<()>
-where
-    <T as TryFrom<u64>>::Error: Debug,
-{
-    let bytes_per_entry = size_of::<T>();
-    let mut buffer = BufWriter::with_capacity(table.len() * bytes_per_entry, file);
-    for &val in table {
-        let converted = T::try_from(val).expect("refcount values are validated on increment");
-        T::write_be(&mut buffer, converted)?;
-    }
-    buffer.flush()
-}
-
-/// Read sub-byte refcounts. Bit 0 is the least significant bit.
-fn read_refcount_subbyte<const BITS: usize>(
-    file: &mut RawFile,
-    count: usize,
-) -> io::Result<Vec<u64>> {
-    const { assert!(BITS == 1 || BITS == 2 || BITS == 4) };
-    let entries_per_byte = 8 / BITS;
-    let mask = (1u64 << BITS) - 1;
-    let bytes_needed = count.div_ceil(entries_per_byte);
-    let mut bytes = vec![0u8; bytes_needed];
-    file.read_exact(&mut bytes)?;
-
-    let mut table = vec![0u64; count];
-    for (i, val) in table.iter_mut().enumerate() {
-        let byte_idx = i / entries_per_byte;
-        let bit_offset = (i % entries_per_byte) * BITS;
-        *val = (bytes[byte_idx] as u64 >> bit_offset) & mask;
-    }
-    Ok(table)
-}
-
-/// Write sub-byte refcounts. Bit 0 is the least significant bit.
-fn write_refcount_subbyte<const BITS: usize>(file: &mut RawFile, table: &[u64]) -> io::Result<()> {
-    const { assert!(BITS == 1 || BITS == 2 || BITS == 4) };
-    let entries_per_byte = 8 / BITS;
-    let mask = (1u64 << BITS) - 1;
-    let mut buffer = BufWriter::with_capacity(table.len().div_ceil(entries_per_byte), file);
-
-    for chunk in table.chunks(entries_per_byte) {
-        let mut byte = 0u8;
-        for (i, &val) in chunk.iter().enumerate() {
-            let bit_offset = i * BITS;
-            byte |= ((val & mask) << bit_offset) as u8;
-        }
-        buffer.write_u8(byte)?;
-    }
-    buffer.flush()
-}
-
-/// A qcow file. Allows reading/writing clusters and appending clusters.
-#[derive(Debug)]
-pub struct QcowRawFile {
-    file: RawFile,
-    cluster_size: u64,
-    cluster_mask: u64,
-    refcount_block_entries: u64,
-    read_refcount_fn: RefcountReader,
-    write_refcount_fn: RefcountWriter,
-}
-
-impl QcowRawFile {
-    /// Creates a `QcowRawFile` from the given `File`, `None` is returned if `cluster_size` is not
-    /// a power of two or refcount_bits is invalid.
-    pub fn from(file: RawFile, cluster_size: u64, refcount_bits: u64) -> Option<Self> {
-        if !cluster_size.is_power_of_two() {
-            return None;
-        }
-
-        let (read_refcount_fn, write_refcount_fn): (RefcountReader, RefcountWriter) =
-            match refcount_bits {
-                1 => (read_refcount_subbyte::<1>, write_refcount_subbyte::<1>),
-                2 => (read_refcount_subbyte::<2>, write_refcount_subbyte::<2>),
-                4 => (read_refcount_subbyte::<4>, write_refcount_subbyte::<4>),
-                8 => (read_refcount::<u8>, write_refcount::<u8>),
-                16 => (read_refcount::<u16>, write_refcount::<u16>),
-                32 => (read_refcount::<u32>, write_refcount::<u32>),
-                64 => (read_refcount::<u64>, write_refcount::<u64>),
-                _ => return None,
-            };
-
-        // For sub-byte refcounts (1,2,4 bits), entries pack multiple per byte
-        let refcount_block_entries = cluster_size * 8 / refcount_bits;
-
-        Some(QcowRawFile {
-            file,
-            cluster_size,
-            cluster_mask: cluster_size - 1,
-            refcount_block_entries,
-            read_refcount_fn,
-            write_refcount_fn,
-        })
-    }
-
-    /// Reads `count` 64 bit offsets and returns them as a vector.
-    /// `mask` optionally `&`s out some of the bits on the file.
-    pub fn read_pointer_table(
-        &mut self,
-        offset: u64,
-        count: u64,
-        mask: Option<u64>,
-    ) -> io::Result<Vec<u64>> {
-        let mut table = vec![0; count as usize];
-        self.file.seek(SeekFrom::Start(offset))?;
-        self.file.read_u64_into::<BigEndian>(&mut table)?;
-        if let Some(m) = mask {
-            for ptr in &mut table {
-                *ptr &= m;
-            }
-        }
-        Ok(table)
-    }
-
-    /// Reads a cluster's worth of 64 bit offsets and returns them as a vector.
-    /// `mask` optionally `&`s out some of the bits on the file.
-    pub fn read_pointer_cluster(&mut self, offset: u64, mask: Option<u64>) -> io::Result<Vec<u64>> {
-        let count = self.cluster_size / size_of::<u64>() as u64;
-        self.read_pointer_table(offset, count, mask)
-    }
-
-    /// Internal helper for creating a buffered writer for pointer tables.
-    #[inline]
-    fn setup_pointer_table_writer<T>(
-        &mut self,
-        offset: u64,
-        entries: &impl Iterator<Item = T>,
-    ) -> io::Result<BufWriter<RawFile>> {
-        self.file.seek(SeekFrom::Start(offset))?;
-        let my_file = self.file.try_clone()?;
-        let capacity = entries.size_hint().0 * size_of::<u64>();
-        Ok(BufWriter::with_capacity(capacity, my_file))
-    }
-
-    /// Writes a pointer table to `offset` in the file.
-    /// Entries are computed on-the-fly by the callback.
-    pub fn write_pointer_table<'a, T: Copy + 'a>(
-        &mut self,
-        offset: u64,
-        entries: impl Iterator<Item = &'a T>,
-        mut f: impl FnMut(&mut QcowRawFile, T) -> io::Result<u64>,
-    ) -> io::Result<()> {
-        let mut buffer = self.setup_pointer_table_writer(offset, &entries)?;
-
-        for addr in entries {
-            let entry = f(self, *addr)?;
-            u64::write_be(&mut buffer, entry)?;
-        }
-        buffer.flush()?;
-        Ok(())
-    }
-
-    /// Writes a pointer table directly without transforming values.
-    pub fn write_pointer_table_direct<'a>(
-        &mut self,
-        offset: u64,
-        entries: impl Iterator<Item = &'a u64>,
-    ) -> io::Result<()> {
-        let mut buffer = self.setup_pointer_table_writer(offset, &entries)?;
-
-        for &entry in entries {
-            u64::write_be(&mut buffer, entry)?;
-        }
-        buffer.flush()?;
-        Ok(())
-    }
-
-    /// Read a refcount block from the file and returns a Vec containing the block.
-    /// Always returns a cluster's worth of data.
-    #[inline]
-    pub fn read_refcount_block(&mut self, offset: u64) -> io::Result<Vec<u64>> {
-        self.file.seek(SeekFrom::Start(offset))?;
-        (self.read_refcount_fn)(&mut self.file, self.refcount_block_entries as usize)
-    }
-
-    /// Writes a refcount block to the file.
-    #[inline]
-    pub fn write_refcount_block(&mut self, offset: u64, table: &[u64]) -> io::Result<()> {
-        self.file.seek(SeekFrom::Start(offset))?;
-        (self.write_refcount_fn)(&mut self.file, table)
-    }
-
-    /// Allocates a new cluster at the end of the current file, return the address.
-    pub fn add_cluster_end(&mut self, max_valid_cluster_offset: u64) -> io::Result<Option<u64>> {
-        // Determine where the new end of the file should be and set_len, which
-        // translates to truncate(2).
-        let file_end: u64 = self.file.seek(SeekFrom::End(0))?;
-        let new_cluster_address: u64 = (file_end + self.cluster_size - 1) & !self.cluster_mask;
-
-        if new_cluster_address > max_valid_cluster_offset {
-            return Ok(None);
-        }
-
-        self.file.set_len(new_cluster_address + self.cluster_size)?;
-
-        Ok(Some(new_cluster_address))
-    }
-
-    /// Returns a reference to the underlying file.
-    pub fn file(&self) -> &RawFile {
-        &self.file
-    }
-
-    /// Returns a mutable reference to the underlying file.
-    pub fn file_mut(&mut self) -> &mut RawFile {
-        &mut self.file
-    }
-
-    /// Returns the size of the file's clusters.
-    pub fn cluster_size(&self) -> u64 {
-        self.cluster_size
-    }
-
-    /// Returns the offset of `address` within a cluster.
-    pub fn cluster_offset(&self, address: u64) -> u64 {
-        address & self.cluster_mask
-    }
-
-    /// Returns the base address of the cluster containing `address`.
-    pub fn cluster_address(&self, address: u64) -> u64 {
-        address & !self.cluster_mask
-    }
-
-    /// Zeros out a cluster in the file.
-    pub fn zero_cluster(&mut self, address: u64) -> io::Result<()> {
-        let cluster_size = self.cluster_size as usize;
-        self.file.seek(SeekFrom::Start(address))?;
-        self.file.write_zeroes(cluster_size)?;
-        Ok(())
-    }
-
-    /// Writes
-    pub fn write_cluster(&mut self, address: u64, data: &[u8]) -> io::Result<()> {
-        let cluster_size = self.cluster_size as usize;
-        self.file.seek(SeekFrom::Start(address))?;
-        self.file.write_all(&data[0..cluster_size])
-    }
-
-    pub fn physical_size(&self) -> Result<u64, std::io::Error> {
-        self.file.metadata().map(|m| m.len())
-    }
-}
-
-impl Clone for QcowRawFile {
-    fn clone(&self) -> Self {
-        QcowRawFile {
-            file: self.file.try_clone().expect("QcowRawFile cloning failed"),
-            cluster_size: self.cluster_size,
-            cluster_mask: self.cluster_mask,
-            refcount_block_entries: self.refcount_block_entries,
-            read_refcount_fn: self.read_refcount_fn,
-            write_refcount_fn: self.write_refcount_fn,
-        }
-    }
-}
-
-impl AsRawFd for QcowRawFile {
-    fn as_raw_fd(&self) -> RawFd {
-        self.file.as_raw_fd()
-    }
-}
-
-impl AsFd for QcowRawFile {
-    fn as_fd(&self) -> BorrowedFd<'_> {
-        self.file.as_fd()
-    }
-}

block/src/qcow/util.rs

@@ -1,79 +0,0 @@
-// Copyright 2018 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 LICENSE-BSD-3-Clause file.
-//
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-//! Pure helper functions and constants for QCOW2 L1/L2 table entry
-//! manipulation and integer arithmetic. Shared across the `qcow` submodules.
-
-/// Nesting depth limit for disk formats that can open other disk files.
-pub(crate) const MAX_NESTING_DEPTH: u32 = 10;
-
-// bits 0-8 and 56-63 are reserved.
-pub(super) const L1_TABLE_OFFSET_MASK: u64 = 0x00ff_ffff_ffff_fe00;
-pub(super) const L2_TABLE_OFFSET_MASK: u64 = 0x00ff_ffff_ffff_fe00;
-// Flags
-pub(super) const ZERO_FLAG: u64 = 1 << 0;
-pub(super) const COMPRESSED_FLAG: u64 = 1 << 62;
-pub(super) const COMPRESSED_SECTOR_SIZE: u64 = 512;
-pub(super) const CLUSTER_USED_FLAG: u64 = 1 << 63;
-
-/// Check if L2 entry is empty (unallocated).
-pub(super) fn l2_entry_is_empty(l2_entry: u64) -> bool {
-    l2_entry == 0
-}
-
-/// Check bit 0 - only valid for standard clusters.
-pub(super) fn l2_entry_is_zero(l2_entry: u64) -> bool {
-    l2_entry & ZERO_FLAG != 0
-}
-
-/// Check if L2 entry refers to a compressed cluster.
-pub(super) fn l2_entry_is_compressed(l2_entry: u64) -> bool {
-    l2_entry & COMPRESSED_FLAG != 0
-}
-
-/// Get file offset and size of compressed cluster data.
-pub(super) fn l2_entry_compressed_cluster_layout(l2_entry: u64, cluster_bits: u32) -> (u64, usize) {
-    let compressed_size_shift = 62 - (cluster_bits - 8);
-    let compressed_size_mask = (1 << (cluster_bits - 8)) - 1;
-    let compressed_cluster_addr = l2_entry & ((1 << compressed_size_shift) - 1);
-    let nsectors = (l2_entry >> compressed_size_shift & compressed_size_mask) + 1;
-    let compressed_cluster_size = ((nsectors * COMPRESSED_SECTOR_SIZE)
-        - (compressed_cluster_addr & (COMPRESSED_SECTOR_SIZE - 1)))
-        as usize;
-    (compressed_cluster_addr, compressed_cluster_size)
-}
-
-/// Get file offset of standard (non-compressed) cluster.
-pub(super) fn l2_entry_std_cluster_addr(l2_entry: u64) -> u64 {
-    l2_entry & L2_TABLE_OFFSET_MASK
-}
-
-/// Make L2 entry for standard (non-compressed) cluster.
-pub(super) fn l2_entry_make_std(cluster_addr: u64) -> u64 {
-    (cluster_addr & L2_TABLE_OFFSET_MASK) | CLUSTER_USED_FLAG
-}
-
-/// Make L2 entry for preallocated zero cluster.
-pub(super) fn l2_entry_make_zero(cluster_addr: u64) -> u64 {
-    (cluster_addr & L2_TABLE_OFFSET_MASK) | CLUSTER_USED_FLAG | ZERO_FLAG
-}
-
-/// Make L1 entry with optional flags.
-pub(super) fn l1_entry_make(cluster_addr: u64, refcount_is_one: bool) -> u64 {
-    (cluster_addr & L1_TABLE_OFFSET_MASK) | (refcount_is_one as u64 * CLUSTER_USED_FLAG)
-}
-
-/// Ceiling of the division of `dividend`/`divisor`.
-pub(super) fn div_round_up_u32(dividend: u32, divisor: u32) -> u32 {
-    dividend / divisor + u32::from(!dividend.is_multiple_of(divisor))
-}
-
-/// Ceiling of the division of `dividend`/`divisor`.
-pub(super) fn div_round_up_u64(dividend: u64, divisor: u64) -> u64 {
-    dividend / divisor + u64::from(!dividend.is_multiple_of(divisor))
-}

block/src/qcow_common.rs

@@ -1,446 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-//! Shared helpers for QCOW2 sync and async backends.
-//!
-//! Position-independent I/O (`pread_exact`, `pwrite_all`) and iovec
-//! scatter/gather helpers used by both `qcow_sync` and `qcow_async`.
-
-use std::alloc::{Layout, alloc_zeroed, dealloc};
-use std::cmp::min;
-use std::os::fd::RawFd;
-use std::{io, ptr, slice};
-
-use crate::qcow::decoder::Decoder;
-
-// -- Position independent I/O helpers --
-//
-// Duplicated file descriptors share the kernel file description and thus the
-// file position. Using seek then read from multiple queues races on that
-// shared position. pread64 and pwrite64 are atomic and never touch the position.
-
-/// Read exactly the requested bytes at offset, looping on short reads.
-pub fn pread_exact(fd: RawFd, buf: &mut [u8], offset: u64) -> io::Result<()> {
-    let mut total = 0usize;
-    while total < buf.len() {
-        // SAFETY: buf and fd are valid for the lifetime of the call.
-        let ret = unsafe {
-            libc::pread64(
-                fd,
-                buf[total..].as_mut_ptr().cast(),
-                buf.len() - total,
-                (offset + total as u64) as libc::off_t,
-            )
-        };
-        if ret < 0 {
-            return Err(io::Error::last_os_error());
-        }
-        if ret == 0 {
-            return Err(io::Error::from(io::ErrorKind::UnexpectedEof));
-        }
-        total += ret as usize;
-    }
-    Ok(())
-}
-
-/// Allocate a buffer and pread exactly `len` bytes at `offset`.
-pub fn pread_alloc(fd: RawFd, offset: u64, len: usize) -> io::Result<Vec<u8>> {
-    let mut buf = vec![0u8; len];
-    pread_exact(fd, &mut buf, offset)?;
-    Ok(buf)
-}
-
-/// Decompress a full QCOW2 cluster from compressed data.
-///
-/// Returns a `cluster_size` byte buffer with the decompressed cluster
-/// content. Fails if the decoder does not produce exactly `cluster_size`
-/// bytes.
-pub fn decompress_cluster(
-    compressed: &[u8],
-    cluster_size: usize,
-    decoder: &dyn Decoder,
-) -> io::Result<Vec<u8>> {
-    let mut decompressed = vec![0u8; cluster_size];
-    let n = decoder
-        .decode(compressed, &mut decompressed)
-        .map_err(|_| io::Error::from_raw_os_error(libc::EIO))?;
-    if n != cluster_size {
-        return Err(io::Error::from_raw_os_error(libc::EIO));
-    }
-    Ok(decompressed)
-}
-
-/// Write all bytes to fd at offset, looping on short writes.
-pub fn pwrite_all(fd: RawFd, buf: &[u8], offset: u64) -> io::Result<()> {
-    let mut total = 0usize;
-    while total < buf.len() {
-        // SAFETY: buf and fd are valid for the lifetime of the call.
-        let ret = unsafe {
-            libc::pwrite64(
-                fd,
-                buf[total..].as_ptr().cast(),
-                buf.len() - total,
-                (offset + total as u64) as libc::off_t,
-            )
-        };
-        if ret < 0 {
-            return Err(io::Error::last_os_error());
-        }
-        if ret == 0 {
-            return Err(io::Error::other("pwrite64 wrote 0 bytes"));
-        }
-        total += ret as usize;
-    }
-    Ok(())
-}
-
-/// RAII wrapper for an aligned heap buffer required by O_DIRECT.
-pub struct AlignedBuf {
-    ptr: *mut u8,
-    layout: Layout,
-}
-
-impl AlignedBuf {
-    pub fn new(size: usize, alignment: usize) -> io::Result<Self> {
-        let size = size.max(1).next_multiple_of(alignment);
-        let layout = Layout::from_size_align(size, alignment)
-            .map_err(|e| io::Error::other(format!("invalid aligned layout: {e}")))?;
-        // SAFETY: layout has non-zero size.
-        let ptr = unsafe { alloc_zeroed(layout) };
-        if ptr.is_null() {
-            return Err(io::Error::new(
-                io::ErrorKind::OutOfMemory,
-                "aligned allocation failed",
-            ));
-        }
-        Ok(AlignedBuf { ptr, layout })
-    }
-
-    pub fn as_mut_slice(&mut self, len: usize) -> &mut [u8] {
-        let len = len.min(self.layout.size());
-        // SAFETY: ptr is valid for layout.size() bytes; len <= layout.size().
-        unsafe { slice::from_raw_parts_mut(self.ptr, len) }
-    }
-
-    pub fn as_slice(&self, len: usize) -> &[u8] {
-        let len = len.min(self.layout.size());
-        // SAFETY: ptr is valid for layout.size() bytes; len <= layout.size().
-        unsafe { slice::from_raw_parts(self.ptr, len) }
-    }
-
-    #[cfg(test)]
-    pub fn layout(&self) -> &Layout {
-        &self.layout
-    }
-
-    #[cfg(test)]
-    pub fn ptr(&self) -> *const u8 {
-        self.ptr
-    }
-}
-
-impl Drop for AlignedBuf {
-    fn drop(&mut self) {
-        // SAFETY: ptr was allocated by alloc_zeroed with self.layout.
-        unsafe { dealloc(self.ptr, self.layout) };
-    }
-}
-
-/// Read into `buf` via an aligned bounce buffer when O_DIRECT requires it.
-pub fn aligned_pread(fd: RawFd, buf: &mut [u8], offset: u64, alignment: usize) -> io::Result<()> {
-    if alignment == 0
-        || ((buf.as_ptr() as usize).is_multiple_of(alignment)
-            && buf.len().is_multiple_of(alignment)
-            && (offset as usize).is_multiple_of(alignment))
-    {
-        return pread_exact(fd, buf, offset);
-    }
-
-    let aligned_offset = offset & !(alignment as u64 - 1);
-    let head = (offset - aligned_offset) as usize;
-    let aligned_len = (head + buf.len()).next_multiple_of(alignment);
-    let mut bounce = AlignedBuf::new(aligned_len, alignment)?;
-    pread_exact(fd, bounce.as_mut_slice(aligned_len), aligned_offset)?;
-    buf.copy_from_slice(&bounce.as_slice(aligned_len)[head..head + buf.len()]);
-    Ok(())
-}
-
-/// Write `buf` via an aligned bounce buffer when O_DIRECT requires it.
-pub fn aligned_pwrite(fd: RawFd, buf: &[u8], offset: u64, alignment: usize) -> io::Result<()> {
-    if alignment == 0
-        || ((buf.as_ptr() as usize).is_multiple_of(alignment)
-            && buf.len().is_multiple_of(alignment)
-            && (offset as usize).is_multiple_of(alignment))
-    {
-        return pwrite_all(fd, buf, offset);
-    }
-
-    let aligned_offset = offset & !(alignment as u64 - 1);
-    let head = (offset - aligned_offset) as usize;
-    let aligned_len = (head + buf.len()).next_multiple_of(alignment);
-    let mut bounce = AlignedBuf::new(aligned_len, alignment)?;
-
-    // Read-modify-write: read the existing aligned region, overlay our data.
-    pread_exact(fd, bounce.as_mut_slice(aligned_len), aligned_offset)?;
-    bounce.as_mut_slice(aligned_len)[head..head + buf.len()].copy_from_slice(buf);
-    pwrite_all(fd, bounce.as_slice(aligned_len), aligned_offset)
-}
-
-// -- iovec helper functions --
-//
-// Operate on the iovec array as a flat byte stream.
-
-/// Copy data into iovecs starting at the given byte offset.
-///
-/// # Safety
-/// Caller must ensure iovecs point to valid, writable memory of sufficient size.
-pub unsafe fn scatter_to_iovecs(iovecs: &[libc::iovec], start: usize, data: &[u8]) {
-    let mut remaining = data;
-    let mut pos = 0usize;
-    for iov in iovecs {
-        let iov_end = pos + iov.iov_len;
-        if iov_end <= start || remaining.is_empty() {
-            pos = iov_end;
-            continue;
-        }
-        let iov_start = start.saturating_sub(pos);
-        let available = iov.iov_len - iov_start;
-        let count = min(available, remaining.len());
-        // SAFETY: iov_base is valid for iov_len bytes per caller contract.
-        unsafe {
-            let dst = iov.iov_base.cast::<u8>().add(iov_start);
-            ptr::copy_nonoverlapping(remaining.as_ptr(), dst, count);
-        }
-        remaining = &remaining[count..];
-        if remaining.is_empty() {
-            break;
-        }
-        pos = iov_end;
-    }
-}
-
-/// Zero fill iovecs starting at the given byte offset for the given length.
-///
-/// # Safety
-/// Caller must ensure iovecs point to valid, writable memory of sufficient size.
-pub unsafe fn zero_fill_iovecs(iovecs: &[libc::iovec], start: usize, len: usize) {
-    let mut remaining = len;
-    let mut pos = 0usize;
-    for iov in iovecs {
-        let iov_end = pos + iov.iov_len;
-        if iov_end <= start || remaining == 0 {
-            pos = iov_end;
-            continue;
-        }
-        let iov_start = start.saturating_sub(pos);
-        let available = iov.iov_len - iov_start;
-        let count = min(available, remaining);
-        // SAFETY: iov_base is valid for iov_len bytes per caller contract.
-        unsafe {
-            let dst = iov.iov_base.cast::<u8>().add(iov_start);
-            ptr::write_bytes(dst, 0, count);
-        }
-        remaining -= count;
-        if remaining == 0 {
-            break;
-        }
-        pos = iov_end;
-    }
-}
-
-/// Gather bytes from iovecs starting at the given byte offset into `dst`.
-///
-/// # Safety
-/// Caller must ensure iovecs point to valid, readable memory of sufficient size.
-pub unsafe fn gather_from_iovecs_into(iovecs: &[libc::iovec], start: usize, dst: &mut [u8]) {
-    let len = dst.len();
-    let mut written = 0usize;
-    let mut pos = 0usize;
-    for iov in iovecs {
-        let iov_end = pos + iov.iov_len;
-        if iov_end <= start || written == len {
-            pos = iov_end;
-            continue;
-        }
-        let iov_start = start.saturating_sub(pos);
-        let available = iov.iov_len - iov_start;
-        let count = min(available, len - written);
-        // SAFETY: iov_base is valid for iov_len bytes per caller contract.
-        unsafe {
-            let src = iov.iov_base.cast::<u8>().add(iov_start);
-            ptr::copy_nonoverlapping(src, dst.as_mut_ptr().add(written), count);
-        }
-        written += count;
-        if written == len {
-            break;
-        }
-        pos = iov_end;
-    }
-}
-
-/// Gather bytes from iovecs starting at the given byte offset into a Vec.
-///
-/// # Safety
-/// Caller must ensure iovecs point to valid, readable memory of sufficient size.
-pub unsafe fn gather_from_iovecs(iovecs: &[libc::iovec], start: usize, len: usize) -> Vec<u8> {
-    let mut result = vec![0u8; len];
-    // SAFETY: caller guarantees iovecs are valid; result has len bytes.
-    unsafe { gather_from_iovecs_into(iovecs, start, &mut result) };
-    result
-}
-
-#[cfg(test)]
-pub(crate) mod unit_tests {
-    use std::fs::File;
-    use std::io::{Read, Seek, SeekFrom, Write};
-    use std::os::unix::fs::FileExt;
-    use std::os::unix::io::AsRawFd;
-
-    use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
-    use flate2::Compression;
-    use flate2::write::DeflateEncoder;
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::{decompress_cluster, pread_alloc};
-    use crate::qcow::decoder::ZlibDecoder;
-
-    const COMPRESSED_FLAG: u64 = 1 << 62;
-    const CLUSTER_USED_FLAG: u64 = 1 << 63;
-    const COMPRESSED_SECTOR_SIZE: u64 = 512;
-
-    const HEADER_CLUSTER_BITS_OFFSET: u64 = 20;
-    const HEADER_L1_SIZE_OFFSET: u64 = 36;
-    const HEADER_L1_TABLE_OFFSET: u64 = 40;
-
-    const L1_L2_ADDR_MASK: u64 = 0x00ff_ffff_ffff_fe00;
-
-    fn make_compressed_l2_entry(host_offset: u64, compressed_len: usize, cluster_bits: u32) -> u64 {
-        let compressed_size_shift = 62 - (cluster_bits - 8);
-        let intra_sector_offset = host_offset & (COMPRESSED_SECTOR_SIZE - 1);
-        let total_bytes = compressed_len as u64 + intra_sector_offset;
-        let nsectors = total_bytes.div_ceil(COMPRESSED_SECTOR_SIZE);
-        let addr_part = host_offset & ((1 << compressed_size_shift) - 1);
-        let size_part = (nsectors - 1) << compressed_size_shift;
-        COMPRESSED_FLAG | size_part | addr_part
-    }
-
-    /// Compress every allocated cluster in a QCOW2 image file in place.
-    ///
-    /// Walks L1 -> L2 tables, compresses each standard cluster with raw
-    /// deflate, appends the compressed payload at the end of the file,
-    /// and rewrites the L2 entry with the compressed layout.
-    pub fn compress_allocated_clusters(file: &mut File) {
-        file.seek(SeekFrom::Start(HEADER_CLUSTER_BITS_OFFSET))
-            .unwrap();
-        let cluster_bits = file.read_u32::<BigEndian>().unwrap();
-        let cluster_size = 1u64 << cluster_bits;
-
-        file.seek(SeekFrom::Start(HEADER_L1_SIZE_OFFSET)).unwrap();
-        let l1_size = file.read_u32::<BigEndian>().unwrap();
-
-        file.seek(SeekFrom::Start(HEADER_L1_TABLE_OFFSET)).unwrap();
-        let l1_table_offset = file.read_u64::<BigEndian>().unwrap();
-
-        let entries_per_l2 = cluster_size / 8;
-
-        let mut append_offset = file.seek(SeekFrom::End(0)).unwrap();
-        append_offset = (append_offset + 511) & !511;
-
-        for l1_idx in 0..l1_size as u64 {
-            let l1_entry_offset = l1_table_offset + l1_idx * 8;
-            file.seek(SeekFrom::Start(l1_entry_offset)).unwrap();
-            let l1_entry = file.read_u64::<BigEndian>().unwrap();
-
-            let l2_table_addr = l1_entry & L1_L2_ADDR_MASK;
-            if l2_table_addr == 0 {
-                continue;
-            }
-
-            for l2_idx in 0..entries_per_l2 {
-                let l2_entry_offset = l2_table_addr + l2_idx * 8;
-                file.seek(SeekFrom::Start(l2_entry_offset)).unwrap();
-                let l2_entry = file.read_u64::<BigEndian>().unwrap();
-
-                if l2_entry & CLUSTER_USED_FLAG == 0 || l2_entry & COMPRESSED_FLAG != 0 {
-                    continue;
-                }
-
-                let host_cluster_addr = l2_entry & L1_L2_ADDR_MASK;
-                if host_cluster_addr == 0 {
-                    continue;
-                }
-
-                let mut cluster_data = vec![0u8; cluster_size as usize];
-                file.seek(SeekFrom::Start(host_cluster_addr)).unwrap();
-                file.read_exact(&mut cluster_data).unwrap();
-
-                let mut encoder = DeflateEncoder::new(Vec::new(), Compression::default());
-                encoder.write_all(&cluster_data).unwrap();
-                let compressed = encoder.finish().unwrap();
-
-                file.seek(SeekFrom::Start(append_offset)).unwrap();
-                file.write_all(&compressed).unwrap();
-
-                // The L2 entry encodes the compressed size in units of
-                // 512 byte sectors. The reader decodes the sector count
-                // back and computes: nsectors * 512 - (addr & 511).
-                // Because addr is 512 aligned, this yields nsectors * 512
-                // which rounds up to the next sector boundary. The file
-                // must contain enough bytes for that rounded up pread.
-                let padded_len = (compressed.len() + 511) & !511;
-                if padded_len > compressed.len() {
-                    let padding = vec![0u8; padded_len - compressed.len()];
-                    file.write_all(&padding).unwrap();
-                }
-
-                let new_entry =
-                    make_compressed_l2_entry(append_offset, compressed.len(), cluster_bits);
-                file.seek(SeekFrom::Start(l2_entry_offset)).unwrap();
-                file.write_u64::<BigEndian>(new_entry).unwrap();
-
-                append_offset += padded_len as u64;
-            }
-        }
-
-        file.flush().unwrap();
-    }
-
-    #[test]
-    fn test_pread_alloc() {
-        let temp = TempFile::new().unwrap();
-        let file = temp.as_file();
-        let data: Vec<u8> = (0..=255).cycle().take(4096).collect();
-        file.write_all_at(&data, 0).unwrap();
-
-        let buf = pread_alloc(file.as_raw_fd(), 0, 4096).unwrap();
-        assert_eq!(buf, data);
-
-        let buf = pread_alloc(file.as_raw_fd(), 100, 200).unwrap();
-        assert_eq!(buf, &data[100..300]);
-
-        pread_alloc(file.as_raw_fd(), 4000, 200).unwrap_err();
-    }
-
-    #[test]
-    fn test_decompress_cluster() {
-        let cluster_size = 65536;
-        let original: Vec<u8> = (0..=255).cycle().take(cluster_size).collect();
-
-        let mut encoder = DeflateEncoder::new(Vec::new(), Compression::default());
-        encoder.write_all(&original).unwrap();
-        let compressed = encoder.finish().unwrap();
-
-        let result = decompress_cluster(&compressed, cluster_size, &ZlibDecoder {}).unwrap();
-        assert_eq!(result, original);
-    }
-
-    #[test]
-    fn test_decompress_cluster_corrupt_input() {
-        let corrupt = vec![0xffu8; 64];
-        let err = decompress_cluster(&corrupt, 65536, &ZlibDecoder {}).unwrap_err();
-        assert_eq!(err.raw_os_error(), Some(libc::EIO));
-    }
-}

block/src/qcow_disk.rs

@@ -1,266 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::fs::File;
-use std::os::unix::io::AsRawFd;
-use std::sync::Arc;
-use std::{fmt, io};
-
-use crate::async_io::{AsyncIo, BorrowedDiskFd, DiskFileError};
-use crate::disk_file;
-use crate::error::{BlockError, BlockErrorKind, BlockResult, ErrorOp};
-use crate::qcow::backing::shared_backing_from;
-use crate::qcow::metadata::{BackingRead, QcowMetadata};
-use crate::qcow::qcow_raw_file::QcowRawFile;
-use crate::qcow::{MAX_NESTING_DEPTH, RawFile, parse_qcow};
-#[cfg(feature = "io_uring")]
-use crate::qcow_async::QcowAsync;
-use crate::qcow_sync::QcowSync;
-
-/// Unified DiskFile wrapper for QCOW2 disk images.
-///
-/// Holds the in memory QCOW2 metadata, the data file, and an optional
-/// backing file. The metadata is wrapped in an `Arc` because
-/// [`QcowSync`] and [`QcowAsync`] I/O workers receive a clone when
-/// they are created via [`create_async_io`](DiskFile::create_async_io).
-/// The backing file is likewise shared with workers through an `Arc`.
-///
-/// The `sparse` flag controls whether the image advertises discard
-/// support to the guest. The `use_io_uring` flag selects between the
-/// [`QcowSync`] and [`QcowAsync`] I/O backends. Both are recorded at
-/// construction time and propagated through [`try_clone`](DiskFile::try_clone).
-pub struct QcowDisk {
-    metadata: Arc<QcowMetadata>,
-    backing_file: Option<Arc<dyn BackingRead>>,
-    sparse: bool,
-    data_raw_file: QcowRawFile,
-    use_io_uring: bool,
-}
-
-impl fmt::Debug for QcowDisk {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        f.debug_struct("QcowDisk")
-            .field("sparse", &self.sparse)
-            .field("has_backing", &self.backing_file.is_some())
-            .field("use_io_uring", &self.use_io_uring)
-            .finish_non_exhaustive()
-    }
-}
-
-impl QcowDisk {
-    pub fn new(
-        file: File,
-        direct_io: bool,
-        backing_files: bool,
-        sparse: bool,
-        use_io_uring: bool,
-    ) -> BlockResult<Self> {
-        #[cfg(not(feature = "io_uring"))]
-        if use_io_uring {
-            return Err(BlockError::new(
-                BlockErrorKind::UnsupportedFeature,
-                DiskFileError::NewAsyncIo(io::Error::other(
-                    "io_uring requested but feature is not enabled",
-                )),
-            ));
-        }
-
-        let max_nesting_depth = if backing_files { MAX_NESTING_DEPTH } else { 0 };
-        let raw_file = RawFile::new(file, direct_io);
-        let (inner, backing_file, sparse) = parse_qcow(raw_file, max_nesting_depth, sparse)
-            .map_err(|e| {
-                let e = if !backing_files && matches!(e.kind(), BlockErrorKind::Overflow) {
-                    e.with_kind(BlockErrorKind::UnsupportedFeature)
-                } else {
-                    e
-                };
-                e.with_op(ErrorOp::Open)
-            })?;
-        let data_raw_file = inner.raw_file.clone();
-        Ok(QcowDisk {
-            metadata: Arc::new(QcowMetadata::new(inner)),
-            backing_file: backing_file.map(shared_backing_from).transpose()?,
-            sparse,
-            data_raw_file,
-            use_io_uring,
-        })
-    }
-}
-
-impl Drop for QcowDisk {
-    fn drop(&mut self) {
-        self.metadata.shutdown();
-    }
-}
-
-impl disk_file::DiskSize for QcowDisk {
-    fn logical_size(&self) -> BlockResult<u64> {
-        Ok(self.metadata.virtual_size())
-    }
-}
-
-impl disk_file::PhysicalSize for QcowDisk {
-    fn physical_size(&self) -> BlockResult<u64> {
-        Ok(self.data_raw_file.physical_size()?)
-    }
-}
-
-impl disk_file::DiskFd for QcowDisk {
-    fn fd(&self) -> BorrowedDiskFd<'_> {
-        BorrowedDiskFd::new(self.data_raw_file.as_raw_fd())
-    }
-}
-
-impl disk_file::Geometry for QcowDisk {}
-
-impl disk_file::SparseCapable for QcowDisk {
-    fn supports_sparse_operations(&self) -> bool {
-        true
-    }
-
-    fn supports_zero_flag(&self) -> bool {
-        true
-    }
-}
-
-impl disk_file::Resizable for QcowDisk {
-    fn resize(&mut self, size: u64) -> BlockResult<()> {
-        if self.backing_file.is_some() {
-            return Err(BlockError::new(
-                BlockErrorKind::UnsupportedFeature,
-                DiskFileError::ResizeError(io::Error::other(
-                    "resize not supported with backing files",
-                )),
-            )
-            .with_op(ErrorOp::Resize));
-        }
-        self.metadata.resize(size).map_err(|e| {
-            BlockError::new(BlockErrorKind::Io, DiskFileError::ResizeError(e))
-                .with_op(ErrorOp::Resize)
-        })
-    }
-}
-
-impl disk_file::DiskFile for QcowDisk {}
-
-impl disk_file::AsyncDiskFile for QcowDisk {
-    fn try_clone(&self) -> BlockResult<Box<dyn disk_file::AsyncDiskFile>> {
-        Ok(Box::new(QcowDisk {
-            metadata: Arc::clone(&self.metadata),
-            backing_file: self.backing_file.as_ref().map(Arc::clone),
-            sparse: self.sparse,
-            data_raw_file: self.data_raw_file.clone(),
-            use_io_uring: self.use_io_uring,
-        }))
-    }
-
-    fn create_async_io(&self, ring_depth: u32) -> BlockResult<Box<dyn AsyncIo>> {
-        if self.use_io_uring {
-            #[cfg(feature = "io_uring")]
-            {
-                return Ok(Box::new(
-                    QcowAsync::new(
-                        Arc::clone(&self.metadata),
-                        self.data_raw_file.clone(),
-                        self.backing_file.as_ref().map(Arc::clone),
-                        self.sparse,
-                        ring_depth,
-                    )
-                    .map_err(|e| {
-                        BlockError::new(BlockErrorKind::Io, DiskFileError::NewAsyncIo(e))
-                    })?,
-                ));
-            }
-
-            #[cfg(not(feature = "io_uring"))]
-            unreachable!("use_io_uring is set but io_uring feature is not enabled");
-        }
-
-        let _ = ring_depth;
-        Ok(Box::new(QcowSync::new(
-            Arc::clone(&self.metadata),
-            self.data_raw_file.clone(),
-            self.backing_file.as_ref().map(Arc::clone),
-            self.sparse,
-        )))
-    }
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::async_io::AsyncIo;
-    use crate::disk_file::{AsyncDiskFile, DiskSize, PhysicalSize};
-    use crate::qcow::{QcowFile, RawFile};
-
-    const TEST_SIZE: u64 = 0x5566_7788;
-
-    fn make_qcow_file() -> File {
-        let temp_file = TempFile::new().unwrap();
-        {
-            let raw = RawFile::new(temp_file.as_file().try_clone().unwrap(), false);
-            QcowFile::new(raw, 3, TEST_SIZE, true).unwrap();
-        }
-        temp_file.into_file()
-    }
-
-    #[test]
-    fn new_sync_returns_correct_size() {
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, false).unwrap();
-        assert_eq!(disk.logical_size().unwrap(), TEST_SIZE);
-    }
-
-    fn assert_async_io_from_dyn(disk: &dyn AsyncDiskFile, expect_batch: bool) {
-        let io: Box<dyn AsyncIo> = disk.create_async_io(128).unwrap();
-        assert_eq!(io.batch_requests_enabled(), expect_batch);
-    }
-
-    fn assert_async_io(disk: &QcowDisk, expect_batch: bool) {
-        assert_async_io_from_dyn(disk, expect_batch);
-    }
-
-    #[test]
-    fn sync_backend_disables_batch_requests() {
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, false).unwrap();
-        assert_async_io(&disk, false);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn io_uring_backend_enables_batch_requests() {
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, true).unwrap();
-        assert_async_io(&disk, true);
-    }
-
-    #[test]
-    fn try_clone_preserves_sync_dispatch() {
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, false).unwrap();
-        let cloned = disk.try_clone().unwrap();
-        assert_async_io_from_dyn(cloned.as_ref(), false);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn try_clone_preserves_io_uring_dispatch() {
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, true).unwrap();
-        let cloned = disk.try_clone().unwrap();
-        assert_async_io_from_dyn(cloned.as_ref(), true);
-    }
-
-    #[test]
-    fn physical_size_less_than_logical() {
-        // make_qcow_file() writes no guest data, so the file on disk
-        // only contains QCOW2 headers and metadata tables.
-        let file = make_qcow_file();
-        let disk = QcowDisk::new(file, false, false, true, false).unwrap();
-        assert!(disk.physical_size().unwrap() < disk.logical_size().unwrap());
-    }
-}

block/src/raw_async.rs

@@ -1,286 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::io::Error;
-use std::os::unix::io::{AsRawFd, RawFd};
-
-use io_uring::{IoUring, opcode, types};
-use libc::{FALLOC_FL_KEEP_SIZE, FALLOC_FL_PUNCH_HOLE, FALLOC_FL_ZERO_RANGE};
-use vmm_sys_util::eventfd::EventFd;
-
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
-use crate::error::{BlockError, BlockErrorKind, BlockResult};
-use crate::{BatchRequest, RequestType, SECTOR_SIZE};
-
-pub struct RawFileAsync {
-    fd: RawFd,
-    io_uring: IoUring,
-    eventfd: EventFd,
-    alignment: u64,
-}
-
-impl RawFileAsync {
-    pub fn new(fd: RawFd, ring_depth: u32) -> BlockResult<Self> {
-        let io_uring =
-            IoUring::new(ring_depth).map_err(|e| BlockError::new(BlockErrorKind::Io, e))?;
-        let eventfd =
-            EventFd::new(libc::EFD_NONBLOCK).map_err(|e| BlockError::new(BlockErrorKind::Io, e))?;
-
-        // Register the io_uring eventfd that will notify when something in
-        // the completion queue is ready.
-        io_uring
-            .submitter()
-            .register_eventfd(eventfd.as_raw_fd())
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, e))?;
-
-        Ok(RawFileAsync {
-            fd,
-            io_uring,
-            eventfd,
-            alignment: SECTOR_SIZE,
-        })
-    }
-}
-
-impl AsyncIo for RawFileAsync {
-    fn notifier(&self) -> &EventFd {
-        &self.eventfd
-    }
-
-    fn alignment(&self) -> u64 {
-        self.alignment
-    }
-
-    fn read_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        let (submitter, mut sq, _) = self.io_uring.split();
-
-        // SAFETY: we know the file descriptor is valid and we
-        // relied on vm-memory to provide the buffer address.
-        unsafe {
-            sq.push(
-                &opcode::Readv::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
-                    .offset(offset.try_into().unwrap())
-                    .build()
-                    .user_data(user_data),
-            )
-            .map_err(|e| {
-                AsyncIoError::ReadVectored(Error::other(format!("Submission queue is full: {e:?}")))
-            })?;
-        };
-
-        // Update the submission queue and submit new operations to the
-        // io_uring instance.
-        sq.sync();
-        submitter.submit().map_err(AsyncIoError::ReadVectored)?;
-
-        Ok(())
-    }
-
-    fn write_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        let (submitter, mut sq, _) = self.io_uring.split();
-
-        // SAFETY: we know the file descriptor is valid and we
-        // relied on vm-memory to provide the buffer address.
-        unsafe {
-            sq.push(
-                &opcode::Writev::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
-                    .offset(offset.try_into().unwrap())
-                    .build()
-                    .user_data(user_data),
-            )
-            .map_err(|e| {
-                AsyncIoError::WriteVectored(Error::other(format!(
-                    "Submission queue is full: {e:?}"
-                )))
-            })?;
-        };
-
-        // Update the submission queue and submit new operations to the
-        // io_uring instance.
-        sq.sync();
-        submitter.submit().map_err(AsyncIoError::WriteVectored)?;
-
-        Ok(())
-    }
-
-    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
-        if let Some(user_data) = user_data {
-            let (submitter, mut sq, _) = self.io_uring.split();
-
-            // SAFETY: we know the file descriptor is valid.
-            unsafe {
-                sq.push(
-                    &opcode::Fsync::new(types::Fd(self.fd))
-                        .build()
-                        .user_data(user_data),
-                )
-                .map_err(|e| {
-                    AsyncIoError::Fsync(Error::other(format!("Submission queue is full: {e:?}")))
-                })?;
-            };
-
-            // Update the submission queue and submit new operations to the
-            // io_uring instance.
-            sq.sync();
-            submitter.submit().map_err(AsyncIoError::Fsync)?;
-        } else {
-            // SAFETY: FFI call with a valid fd
-            unsafe { libc::fsync(self.fd) };
-        }
-
-        Ok(())
-    }
-
-    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
-        self.io_uring
-            .completion()
-            .next()
-            .map(|entry| (entry.user_data(), entry.result()))
-    }
-
-    fn batch_requests_enabled(&self) -> bool {
-        true
-    }
-
-    fn submit_batch_requests(&mut self, batch_request: &[BatchRequest]) -> AsyncIoResult<()> {
-        if !self.batch_requests_enabled() {
-            return Ok(());
-        }
-
-        let (submitter, mut sq, _) = self.io_uring.split();
-        let mut submitted = false;
-
-        // Refuse the whole batch if it can't fit in the SQ to avoid having to unroll a partially
-        // successful push.
-        if batch_request.len() > sq.capacity() - sq.len() {
-            return Err(AsyncIoError::SubmitBatchRequests(Error::other(
-                "io_uring submission queue is full",
-            )));
-        }
-
-        for req in batch_request {
-            match req.request_type {
-                RequestType::In => {
-                    // SAFETY: we know the file descriptor is valid and we
-                    // relied on vm-memory to provide the buffer address.
-                    unsafe {
-                        sq.push(
-                            &opcode::Readv::new(
-                                types::Fd(self.fd),
-                                req.iovecs.as_ptr(),
-                                req.iovecs.len() as u32,
-                            )
-                            .offset(req.offset as u64)
-                            .build()
-                            .user_data(req.user_data),
-                        )
-                        .map_err(|e| {
-                            AsyncIoError::ReadVectored(Error::other(format!(
-                                "Submission queue is full: {e:?}"
-                            )))
-                        })?;
-                    };
-                    submitted = true;
-                }
-                RequestType::Out => {
-                    // SAFETY: we know the file descriptor is valid and we
-                    // relied on vm-memory to provide the buffer address.
-                    unsafe {
-                        sq.push(
-                            &opcode::Writev::new(
-                                types::Fd(self.fd),
-                                req.iovecs.as_ptr(),
-                                req.iovecs.len() as u32,
-                            )
-                            .offset(req.offset as u64)
-                            .build()
-                            .user_data(req.user_data),
-                        )
-                        .map_err(|e| {
-                            AsyncIoError::WriteVectored(Error::other(format!(
-                                "Submission queue is full: {e:?}"
-                            )))
-                        })?;
-                    };
-                    submitted = true;
-                }
-                _ => {
-                    unreachable!("Unexpected batch request type: {:?}", req.request_type)
-                }
-            }
-        }
-
-        // Only submit if we actually queued something
-        if submitted {
-            // Update the submission queue and submit new operations to the
-            // io_uring instance.
-            sq.sync();
-            submitter
-                .submit()
-                .map_err(AsyncIoError::SubmitBatchRequests)?;
-        }
-
-        Ok(())
-    }
-
-    fn punch_hole(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        let (submitter, mut sq, _) = self.io_uring.split();
-
-        let mode = FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: The file descriptor is known to be valid.
-        unsafe {
-            sq.push(
-                &opcode::Fallocate::new(types::Fd(self.fd), length)
-                    .offset(offset)
-                    .mode(mode)
-                    .build()
-                    .user_data(user_data),
-            )
-            .map_err(|e| {
-                AsyncIoError::PunchHole(Error::other(format!("Submission queue is full: {e:?}")))
-            })?;
-        };
-
-        sq.sync();
-        submitter.submit().map_err(AsyncIoError::PunchHole)?;
-
-        Ok(())
-    }
-
-    fn write_zeroes(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        let (submitter, mut sq, _) = self.io_uring.split();
-
-        let mode = FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: The file descriptor is known to be valid.
-        unsafe {
-            sq.push(
-                &opcode::Fallocate::new(types::Fd(self.fd), length)
-                    .offset(offset)
-                    .mode(mode)
-                    .build()
-                    .user_data(user_data),
-            )
-            .map_err(|e| {
-                AsyncIoError::WriteZeroes(Error::other(format!("Submission queue is full: {e:?}")))
-            })?;
-        };
-
-        sq.sync();
-        submitter.submit().map_err(AsyncIoError::WriteZeroes)?;
-
-        Ok(())
-    }
-}

block/src/raw_async_aio.rs

@@ -1,218 +0,0 @@
-// Copyright © 2023 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-//
-// Copyright © 2023 Crusoe Energy Systems LLC
-//
-
-use std::collections::VecDeque;
-use std::os::unix::io::{AsRawFd, RawFd};
-
-use libc::{FALLOC_FL_KEEP_SIZE, FALLOC_FL_PUNCH_HOLE, FALLOC_FL_ZERO_RANGE};
-use vmm_sys_util::aio;
-use vmm_sys_util::eventfd::EventFd;
-
-use crate::SECTOR_SIZE;
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
-use crate::error::{BlockError, BlockErrorKind, BlockResult};
-
-pub struct RawFileAsyncAio {
-    fd: RawFd,
-    ctx: aio::IoContext,
-    eventfd: EventFd,
-    alignment: u64,
-    completion_list: VecDeque<(u64, i32)>,
-}
-
-impl RawFileAsyncAio {
-    pub fn new(fd: RawFd, queue_depth: u32) -> BlockResult<Self> {
-        let eventfd =
-            EventFd::new(libc::EFD_NONBLOCK).map_err(|e| BlockError::new(BlockErrorKind::Io, e))?;
-        let ctx =
-            aio::IoContext::new(queue_depth).map_err(|e| BlockError::new(BlockErrorKind::Io, e))?;
-
-        Ok(RawFileAsyncAio {
-            fd,
-            ctx,
-            eventfd,
-            alignment: SECTOR_SIZE,
-            completion_list: VecDeque::new(),
-        })
-    }
-}
-
-impl AsyncIo for RawFileAsyncAio {
-    fn notifier(&self) -> &EventFd {
-        &self.eventfd
-    }
-
-    fn alignment(&self) -> u64 {
-        self.alignment
-    }
-
-    fn read_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        let iocbs = [&mut aio::IoControlBlock {
-            aio_fildes: self.fd.as_raw_fd() as u32,
-            aio_lio_opcode: aio::IOCB_CMD_PREADV as u16,
-            aio_buf: iovecs.as_ptr() as u64,
-            aio_nbytes: iovecs.len() as u64,
-            aio_offset: offset,
-            aio_data: user_data,
-            aio_flags: aio::IOCB_FLAG_RESFD,
-            aio_resfd: self.eventfd.as_raw_fd() as u32,
-            ..Default::default()
-        }];
-        let _ = self
-            .ctx
-            .submit(&iocbs[..])
-            .map_err(AsyncIoError::ReadVectored)?;
-
-        Ok(())
-    }
-
-    fn write_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        let iocbs = [&mut aio::IoControlBlock {
-            aio_fildes: self.fd.as_raw_fd() as u32,
-            aio_lio_opcode: aio::IOCB_CMD_PWRITEV as u16,
-            aio_buf: iovecs.as_ptr() as u64,
-            aio_nbytes: iovecs.len() as u64,
-            aio_offset: offset,
-            aio_data: user_data,
-            aio_flags: aio::IOCB_FLAG_RESFD,
-            aio_resfd: self.eventfd.as_raw_fd() as u32,
-            ..Default::default()
-        }];
-        let _ = self
-            .ctx
-            .submit(&iocbs[..])
-            .map_err(AsyncIoError::WriteVectored)?;
-
-        Ok(())
-    }
-
-    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
-        if let Some(user_data) = user_data {
-            let iocbs = [&mut aio::IoControlBlock {
-                aio_fildes: self.fd.as_raw_fd() as u32,
-                aio_lio_opcode: aio::IOCB_CMD_FSYNC as u16,
-                aio_data: user_data,
-                aio_flags: aio::IOCB_FLAG_RESFD,
-                aio_resfd: self.eventfd.as_raw_fd() as u32,
-                ..Default::default()
-            }];
-            let _ = self.ctx.submit(&iocbs[..]).map_err(AsyncIoError::Fsync)?;
-        } else {
-            // SAFETY: FFI call with a valid fd
-            unsafe { libc::fsync(self.fd) };
-        }
-
-        Ok(())
-    }
-
-    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
-        if self.completion_list.is_empty() {
-            // Drain pending AIO completions batched into the same queue.
-            let mut events = [aio::IoEvent::default(); 32];
-            let rc = self.ctx.get_events(0, &mut events, None).unwrap();
-            for event in &events[..rc] {
-                self.completion_list
-                    .push_back((event.data, event.res as i32));
-            }
-        }
-        self.completion_list.pop_front()
-    }
-
-    fn punch_hole(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        // Linux AIO has no IOCB command for fallocate, so perform the operation
-        // synchronously and signal completion via the completion list, matching
-        // the pattern used by the sync backend (RawFileSync).
-        let mode = FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::fallocate(
-                self.fd as libc::c_int,
-                mode,
-                offset as libc::off_t,
-                length as libc::off_t,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::PunchHole(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-
-    fn write_zeroes(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        // Linux AIO has no IOCB command for fallocate, so perform the operation
-        // synchronously and signal completion via the completion list, matching
-        // the pattern used by the sync backend (RawFileSync).
-        let mode = FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::fallocate(
-                self.fd as libc::c_int,
-                mode,
-                offset as libc::off_t,
-                length as libc::off_t,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::WriteZeroes(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use std::os::unix::io::AsRawFd;
-
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::raw_async_io_tests;
-
-    #[test]
-    fn test_punch_hole() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileAsyncAio::new(file.as_raw_fd(), 128).unwrap();
-        raw_async_io_tests::test_punch_hole(&mut async_io, &mut file);
-    }
-
-    #[test]
-    fn test_write_zeroes() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileAsyncAio::new(file.as_raw_fd(), 128).unwrap();
-        raw_async_io_tests::test_write_zeroes(&mut async_io, &mut file);
-    }
-
-    #[test]
-    fn test_punch_hole_multiple_operations() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileAsyncAio::new(file.as_raw_fd(), 128).unwrap();
-        raw_async_io_tests::test_punch_hole_multiple_operations(&mut async_io, &mut file);
-    }
-}

block/src/raw_async_io_tests.rs

@@ -1,162 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-//! Shared test helpers for [`AsyncIo`] backends.
-//!
-//! Each helper takes a `&mut dyn AsyncIo` together with the [`File`] handle
-//! that backs the I/O object, so the same logic exercises every backend with
-//! only the constructor differing.
-
-use std::fs::File;
-use std::io::{Read, Seek, SeekFrom, Write};
-
-use crate::async_io::{AsyncIo, AsyncIoError};
-
-/// Tests punching a hole in the middle of a 4 MB file and verifying data
-/// integrity around the hole.
-pub fn test_punch_hole(async_io: &mut dyn AsyncIo, file: &mut File) {
-    // Write 4MB of data
-    let data = vec![0xAA; 4 * 1024 * 1024];
-    file.write_all(&data).unwrap();
-    file.sync_all().unwrap();
-
-    // Punch hole in the middle (1MB at offset 1MB)
-    let offset = 1024 * 1024;
-    let length = 1024 * 1024;
-    async_io.punch_hole(offset, length, 1).unwrap();
-
-    // Check completion
-    let (user_data, result) = async_io.next_completed_request().unwrap();
-    assert_eq!(user_data, 1);
-    assert_eq!(result, 0);
-
-    // Verify the hole reads as zeros
-    file.seek(SeekFrom::Start(offset)).unwrap();
-    let mut read_buf = vec![0; length as usize];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0),
-        "Punched hole should read as zeros"
-    );
-
-    // Verify data before hole is intact
-    file.seek(SeekFrom::Start(0)).unwrap();
-    let mut read_buf = vec![0; 1024];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0xAA),
-        "Data before hole should be intact"
-    );
-
-    // Verify data after hole is intact
-    file.seek(SeekFrom::Start(offset + length)).unwrap();
-    let mut read_buf = vec![0; 1024];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0xAA),
-        "Data after hole should be intact"
-    );
-}
-
-/// Tests writing zeroes to a 512 KB region inside a 4 MB file and verifying
-/// surrounding data is preserved.  Gracefully skips when the filesystem does
-/// not support `FALLOC_FL_ZERO_RANGE`.
-pub fn test_write_zeroes(async_io: &mut dyn AsyncIo, file: &mut File) {
-    // Write 4MB of data
-    let data = vec![0xBB; 4 * 1024 * 1024];
-    file.write_all(&data).unwrap();
-    file.sync_all().unwrap();
-
-    // Write zeros in the middle (512KB at offset 2MB)
-    let offset = 2 * 1024 * 1024;
-    let length = 512 * 1024;
-    let write_zeroes_result = async_io.write_zeroes(offset, length, 2);
-
-    // FALLOC_FL_ZERO_RANGE might not be supported on all filesystems (e.g., tmpfs)
-    // If it fails with ENOTSUP, skip the test
-    if let Err(AsyncIoError::WriteZeroes(ref e)) = write_zeroes_result
-        && (e.raw_os_error() == Some(libc::EOPNOTSUPP) || e.raw_os_error() == Some(libc::ENOTSUP))
-    {
-        eprintln!("Skipping test_write_zeroes: filesystem doesn't support FALLOC_FL_ZERO_RANGE");
-        return;
-    }
-    write_zeroes_result.unwrap();
-
-    // Check completion
-    let (user_data, result) = async_io.next_completed_request().unwrap();
-    assert_eq!(user_data, 2);
-    assert_eq!(result, 0);
-
-    // Verify the zeroed region reads as zeros
-    file.seek(SeekFrom::Start(offset)).unwrap();
-    let mut read_buf = vec![0; length as usize];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0),
-        "Zeroed region should read as zeros"
-    );
-
-    // Verify data before zeroed region is intact
-    file.seek(SeekFrom::Start(offset - 1024)).unwrap();
-    let mut read_buf = vec![0; 1024];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0xBB),
-        "Data before zeroed region should be intact"
-    );
-
-    // Verify data after zeroed region is intact
-    file.seek(SeekFrom::Start(offset + length)).unwrap();
-    let mut read_buf = vec![0; 1024];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(
-        read_buf.iter().all(|&b| b == 0xBB),
-        "Data after zeroed region should be intact"
-    );
-}
-
-/// Tests punching multiple holes in an 8 MB file and verifying each hole
-/// independently reads as zeroes.
-pub fn test_punch_hole_multiple_operations(async_io: &mut dyn AsyncIo, file: &mut File) {
-    // Write 8MB of data
-    let data = vec![0xCC; 8 * 1024 * 1024];
-    file.write_all(&data).unwrap();
-    file.sync_all().unwrap();
-
-    // Punch multiple holes
-    async_io.punch_hole(1024 * 1024, 512 * 1024, 10).unwrap();
-    async_io
-        .punch_hole(3 * 1024 * 1024, 512 * 1024, 11)
-        .unwrap();
-    async_io
-        .punch_hole(5 * 1024 * 1024, 512 * 1024, 12)
-        .unwrap();
-
-    // Check all completions
-    let (user_data, result) = async_io.next_completed_request().unwrap();
-    assert_eq!(user_data, 10);
-    assert_eq!(result, 0);
-
-    let (user_data, result) = async_io.next_completed_request().unwrap();
-    assert_eq!(user_data, 11);
-    assert_eq!(result, 0);
-
-    let (user_data, result) = async_io.next_completed_request().unwrap();
-    assert_eq!(user_data, 12);
-    assert_eq!(result, 0);
-
-    // Verify all holes read as zeros
-    file.seek(SeekFrom::Start(1024 * 1024)).unwrap();
-    let mut read_buf = vec![0; 512 * 1024];
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(read_buf.iter().all(|&b| b == 0));
-
-    file.seek(SeekFrom::Start(3 * 1024 * 1024)).unwrap();
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(read_buf.iter().all(|&b| b == 0));
-
-    file.seek(SeekFrom::Start(5 * 1024 * 1024)).unwrap();
-    file.read_exact(&mut read_buf).unwrap();
-    assert!(read_buf.iter().all(|&b| b == 0));
-}

block/src/raw_disk.rs

@@ -1,265 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::fs::File;
-use std::io;
-use std::os::unix::fs::FileTypeExt;
-use std::os::unix::io::AsRawFd;
-
-use log::warn;
-
-use crate::async_io::{AsyncIo, BorrowedDiskFd, DiskFileError};
-use crate::error::{BlockError, BlockErrorKind, BlockResult};
-#[cfg(feature = "io_uring")]
-use crate::raw_async::RawFileAsync;
-use crate::raw_async_aio::RawFileAsyncAio;
-use crate::raw_sync::RawFileSync;
-use crate::{DiskTopology, disk_file, probe_sparse_support, query_device_size};
-
-/// Selects which async I/O backend a `RawDisk` uses.
-#[derive(Clone, Copy, Debug, PartialEq)]
-pub enum RawBackend {
-    /// Blocking I/O where the caller waits for completion.
-    Sync,
-    /// Modern asynchronous I/O using shared submission and completion
-    /// rings for lower overhead operation dispatch and completion handling.
-    #[cfg(feature = "io_uring")]
-    IoUring,
-    /// Legacy asynchronous I/O where requests are handed to the kernel
-    /// and completions are collected later.
-    Aio,
-}
-
-/// Unified DiskFile wrapper for raw disk images.
-///
-/// Owns the underlying file and delegates async I/O creation to the
-/// backend selected at construction time via [`RawBackend`].
-#[derive(Debug)]
-pub struct RawDisk {
-    file: File,
-    backend: RawBackend,
-}
-
-impl RawDisk {
-    pub fn new(file: File, backend: RawBackend) -> Self {
-        Self { file, backend }
-    }
-}
-
-impl disk_file::DiskSize for RawDisk {
-    fn logical_size(&self) -> BlockResult<u64> {
-        query_device_size(&self.file)
-            .map(|(logical_size, _)| logical_size)
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::Size(e)))
-    }
-}
-
-impl disk_file::PhysicalSize for RawDisk {
-    fn physical_size(&self) -> BlockResult<u64> {
-        query_device_size(&self.file)
-            .map(|(_, physical_size)| physical_size)
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::Size(e)))
-    }
-}
-
-impl disk_file::DiskFd for RawDisk {
-    fn fd(&self) -> BorrowedDiskFd<'_> {
-        BorrowedDiskFd::new(self.file.as_raw_fd())
-    }
-}
-
-impl disk_file::Geometry for RawDisk {
-    fn topology(&self) -> DiskTopology {
-        DiskTopology::probe(&self.file).unwrap_or_else(|_| {
-            warn!("Unable to get device topology. Using default topology");
-            DiskTopology::default()
-        })
-    }
-}
-
-impl disk_file::SparseCapable for RawDisk {
-    fn supports_sparse_operations(&self) -> bool {
-        probe_sparse_support(&self.file)
-    }
-}
-
-impl disk_file::Resizable for RawDisk {
-    fn resize(&mut self, size: u64) -> BlockResult<()> {
-        let fd_metadata = self
-            .file
-            .metadata()
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::ResizeError(e)))?;
-
-        if fd_metadata.file_type().is_block_device() {
-            // Block devices cannot be resized via ftruncate; they are resized
-            // externally (LVM, losetup, etc.). Verify the size matches.
-            let (actual_size, _) = query_device_size(&self.file)
-                .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::ResizeError(e)))?;
-            if actual_size != size {
-                return Err(BlockError::new(
-                    BlockErrorKind::Io,
-                    DiskFileError::ResizeError(io::Error::other(format!(
-                        "Block device size {actual_size} does not match requested size {size}"
-                    ))),
-                ));
-            }
-            Ok(())
-        } else {
-            self.file
-                .set_len(size)
-                .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::ResizeError(e)))
-        }
-    }
-}
-
-impl disk_file::DiskFile for RawDisk {}
-
-impl disk_file::AsyncDiskFile for RawDisk {
-    fn try_clone(&self) -> BlockResult<Box<dyn disk_file::AsyncDiskFile>> {
-        let file = self
-            .file
-            .try_clone()
-            .map_err(|e| BlockError::new(BlockErrorKind::Io, DiskFileError::Clone(e)))?;
-        Ok(Box::new(RawDisk {
-            file,
-            backend: self.backend,
-        }))
-    }
-
-    fn create_async_io(&self, ring_depth: u32) -> BlockResult<Box<dyn AsyncIo>> {
-        match self.backend {
-            RawBackend::Sync => Ok(Box::new(RawFileSync::new(self.file.as_raw_fd()))),
-            #[cfg(feature = "io_uring")]
-            RawBackend::IoUring => Ok(Box::new(RawFileAsync::new(
-                self.file.as_raw_fd(),
-                ring_depth,
-            )?)),
-            RawBackend::Aio => Ok(Box::new(RawFileAsyncAio::new(
-                self.file.as_raw_fd(),
-                ring_depth,
-            )?)),
-        }
-    }
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use std::fs::File;
-
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::async_io::AsyncIo;
-    use crate::disk_file::{AsyncDiskFile, DiskSize, PhysicalSize, Resizable};
-
-    const TEST_SIZE: u64 = 0x1122_3344;
-
-    fn make_raw_file() -> File {
-        let file: File = TempFile::new().unwrap().into_file();
-        file.set_len(TEST_SIZE).unwrap();
-        file
-    }
-
-    #[test]
-    fn new_sync_returns_correct_size() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Sync);
-        assert_eq!(disk.logical_size().unwrap(), TEST_SIZE);
-    }
-
-    fn assert_async_io_from_dyn(disk: &dyn AsyncDiskFile, expect_backend: RawBackend) {
-        let io: Box<dyn AsyncIo> = disk.create_async_io(128).unwrap();
-        cfg_if::cfg_if! {
-            if #[cfg(feature = "io_uring")] {
-                let expected_batch_requests = expect_backend == RawBackend::IoUring;
-            } else {
-                let _ = expect_backend;
-                let expected_batch_requests = false;
-            }
-        }
-        assert_eq!(io.batch_requests_enabled(), expected_batch_requests);
-    }
-
-    fn assert_sync_backend(disk: &RawDisk) {
-        assert_eq!(disk.backend, RawBackend::Sync);
-        assert_async_io_from_dyn(disk, RawBackend::Sync);
-    }
-
-    fn assert_aio_backend(disk: &RawDisk) {
-        assert_eq!(disk.backend, RawBackend::Aio);
-        assert_async_io_from_dyn(disk, RawBackend::Aio);
-    }
-
-    #[cfg(feature = "io_uring")]
-    fn assert_io_uring_backend(disk: &RawDisk) {
-        assert_eq!(disk.backend, RawBackend::IoUring);
-        assert_async_io_from_dyn(disk, RawBackend::IoUring);
-    }
-
-    #[test]
-    fn sync_backend_disables_batch_requests() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Sync);
-        assert_sync_backend(&disk);
-    }
-
-    #[test]
-    fn aio_backend_disables_batch_requests() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Aio);
-        assert_aio_backend(&disk);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn io_uring_backend_enables_batch_requests() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::IoUring);
-        assert_io_uring_backend(&disk);
-    }
-
-    fn assert_try_clone(disk: &RawDisk, expect_backend: RawBackend) {
-        let cloned = disk.try_clone().unwrap();
-        assert_async_io_from_dyn(cloned.as_ref(), expect_backend);
-    }
-
-    #[test]
-    fn try_clone_preserves_sync_backend() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Sync);
-        assert_try_clone(&disk, RawBackend::Sync);
-    }
-
-    #[test]
-    fn try_clone_preserves_aio_backend() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Aio);
-        assert_try_clone(&disk, RawBackend::Aio);
-    }
-
-    #[cfg(feature = "io_uring")]
-    #[test]
-    fn try_clone_preserves_io_uring_backend() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::IoUring);
-        assert_try_clone(&disk, RawBackend::IoUring);
-    }
-
-    #[test]
-    fn resize_changes_file_size() {
-        let file = make_raw_file();
-        let mut disk = RawDisk::new(file, RawBackend::Aio);
-        let new_size = TEST_SIZE * 2;
-        disk.resize(new_size).unwrap();
-        assert_eq!(disk.logical_size().unwrap(), new_size);
-    }
-
-    #[test]
-    fn physical_size_reports_allocated_blocks() {
-        let file = make_raw_file();
-        let disk = RawDisk::new(file, RawBackend::Aio);
-        // Sparse file: physical size is less than logical size.
-        assert!(disk.physical_size().unwrap() < disk.logical_size().unwrap());
-    }
-}

block/src/raw_sync.rs

@@ -1,187 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::collections::VecDeque;
-use std::os::unix::io::RawFd;
-
-use libc::{FALLOC_FL_KEEP_SIZE, FALLOC_FL_PUNCH_HOLE, FALLOC_FL_ZERO_RANGE};
-use vmm_sys_util::eventfd::EventFd;
-
-use crate::SECTOR_SIZE;
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
-
-pub struct RawFileSync {
-    fd: RawFd,
-    eventfd: EventFd,
-    completion_list: VecDeque<(u64, i32)>,
-    alignment: u64,
-}
-
-impl RawFileSync {
-    pub fn new(fd: RawFd) -> Self {
-        RawFileSync {
-            fd,
-            eventfd: EventFd::new(libc::EFD_NONBLOCK).expect("Failed creating EventFd for RawFile"),
-            completion_list: VecDeque::new(),
-            alignment: SECTOR_SIZE,
-        }
-    }
-}
-
-impl AsyncIo for RawFileSync {
-    fn notifier(&self) -> &EventFd {
-        &self.eventfd
-    }
-
-    fn alignment(&self) -> u64 {
-        self.alignment
-    }
-
-    fn read_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::preadv(
-                self.fd as libc::c_int,
-                iovecs.as_ptr(),
-                iovecs.len() as libc::c_int,
-                offset,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::ReadVectored(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result as i32));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-
-    fn write_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::pwritev(
-                self.fd as libc::c_int,
-                iovecs.as_ptr(),
-                iovecs.len() as libc::c_int,
-                offset,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::WriteVectored(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result as i32));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-
-    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
-        // SAFETY: FFI call
-        let result = unsafe { libc::fsync(self.fd as libc::c_int) };
-        if result < 0 {
-            return Err(AsyncIoError::Fsync(std::io::Error::last_os_error()));
-        }
-
-        if let Some(user_data) = user_data {
-            self.completion_list.push_back((user_data, result));
-            self.eventfd.write(1).unwrap();
-        }
-
-        Ok(())
-    }
-
-    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
-        self.completion_list.pop_front()
-    }
-
-    fn punch_hole(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        let mode = FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::fallocate(
-                self.fd as libc::c_int,
-                mode,
-                offset as libc::off_t,
-                length as libc::off_t,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::PunchHole(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-
-    fn write_zeroes(&mut self, offset: u64, length: u64, user_data: u64) -> AsyncIoResult<()> {
-        let mode = FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE;
-
-        // SAFETY: FFI call with valid arguments
-        let result = unsafe {
-            libc::fallocate(
-                self.fd as libc::c_int,
-                mode,
-                offset as libc::off_t,
-                length as libc::off_t,
-            )
-        };
-        if result < 0 {
-            return Err(AsyncIoError::WriteZeroes(std::io::Error::last_os_error()));
-        }
-
-        self.completion_list.push_back((user_data, result));
-        self.eventfd.write(1).unwrap();
-
-        Ok(())
-    }
-}
-
-#[cfg(test)]
-mod unit_tests {
-    use std::os::unix::io::AsRawFd;
-
-    use vmm_sys_util::tempfile::TempFile;
-
-    use super::*;
-    use crate::raw_async_io_tests;
-
-    #[test]
-    fn test_punch_hole() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileSync::new(file.as_raw_fd());
-        raw_async_io_tests::test_punch_hole(&mut async_io, &mut file);
-    }
-
-    #[test]
-    fn test_write_zeroes() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileSync::new(file.as_raw_fd());
-        raw_async_io_tests::test_write_zeroes(&mut async_io, &mut file);
-    }
-
-    #[test]
-    fn test_punch_hole_multiple_operations() {
-        let temp_file = TempFile::new().unwrap();
-        let mut file = temp_file.into_file();
-        let mut async_io = RawFileSync::new(file.as_raw_fd());
-        raw_async_io_tests::test_punch_hole_multiple_operations(&mut async_io, &mut file);
-    }
-}

block/src/request.rs

@@ -1,573 +0,0 @@
-// 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 LICENSE-BSD-3-Clause file.
-//
-// Copyright © 2020 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
-
-use std::io::{Read, Seek, SeekFrom, Write};
-use std::mem;
-use std::time::Instant;
-
-use log::{error, warn};
-use smallvec::SmallVec;
-use virtio_bindings::virtio_blk::{
-    VIRTIO_BLK_T_DISCARD, VIRTIO_BLK_T_WRITE_ZEROES, VIRTIO_BLK_WRITE_ZEROES_FLAG_UNMAP,
-    virtio_blk_discard_write_zeroes,
-};
-use virtio_queue::DescriptorChain;
-use vm_memory::bitmap::Bitmap;
-use vm_memory::{
-    Address as _, Bytes as _, GuestAddress, GuestMemory as _, GuestMemoryError,
-    GuestMemoryLoadGuard,
-};
-use vm_virtio::{AccessPlatform, Translatable as _};
-
-use crate::aligned_operation::AlignedOperation;
-use crate::async_io::AsyncIo;
-use crate::{Error, ExecuteError, request_type, sector};
-
-const SECTOR_SHIFT: u8 = 9;
-pub const SECTOR_SIZE: u64 = 0x01 << SECTOR_SHIFT;
-
-/// Maximum number of segments per DISCARD or WRITE_ZEROES request.
-pub const MAX_DISCARD_WRITE_ZEROES_SEG: u32 = 1;
-/// Size and field offsets within `struct virtio_blk_discard_write_zeroes`.
-const DISCARD_WZ_SEG_SIZE: u32 = mem::size_of::<virtio_blk_discard_write_zeroes>() as u32;
-const DISCARD_WZ_MAX_PAYLOAD: u32 = DISCARD_WZ_SEG_SIZE * MAX_DISCARD_WRITE_ZEROES_SEG;
-const DISCARD_WZ_SECTOR_OFFSET: u64 =
-    mem::offset_of!(virtio_blk_discard_write_zeroes, sector) as u64;
-const DISCARD_WZ_NUM_SECTORS_OFFSET: u64 =
-    mem::offset_of!(virtio_blk_discard_write_zeroes, num_sectors) as u64;
-const DISCARD_WZ_FLAGS_OFFSET: u64 = mem::offset_of!(virtio_blk_discard_write_zeroes, flags) as u64;
-
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-pub enum RequestType {
-    In,
-    Out,
-    Flush,
-    GetDeviceId,
-    Discard,
-    WriteZeroes,
-    Unsupported(u32),
-}
-
-pub const DEFAULT_DESCRIPTOR_VEC_SIZE: usize = 32;
-pub struct BatchRequest {
-    pub offset: libc::off_t,
-    pub iovecs: SmallVec<[libc::iovec; DEFAULT_DESCRIPTOR_VEC_SIZE]>,
-    pub user_data: u64,
-    pub request_type: RequestType,
-}
-
-pub struct ExecuteAsync {
-    // `true` if the execution will complete asynchronously
-    pub async_complete: bool,
-    // request need to be batched for submission if any
-    pub batch_request: Option<BatchRequest>,
-}
-
-#[derive(Debug)]
-pub struct Request {
-    request_type: RequestType,
-    sector: u64,
-    data_descriptors: SmallVec<[(GuestAddress, u32); DEFAULT_DESCRIPTOR_VEC_SIZE]>,
-    status_addr: GuestAddress,
-    pub writeback: bool,
-    aligned_operations: SmallVec<[AlignedOperation; DEFAULT_DESCRIPTOR_VEC_SIZE]>,
-    start: Instant,
-}
-
-impl Request {
-    pub fn parse<B: Bitmap + 'static>(
-        desc_chain: &mut DescriptorChain<GuestMemoryLoadGuard<vm_memory::GuestMemoryMmap<B>>>,
-        access_platform: Option<&dyn AccessPlatform>,
-    ) -> Result<Request, Error> {
-        let hdr_desc = desc_chain
-            .next()
-            .ok_or(Error::DescriptorChainTooShort)
-            .inspect_err(|_| {
-                error!("Missing head descriptor");
-            })?;
-
-        // The head contains the request type which MUST be readable.
-        if hdr_desc.is_write_only() {
-            return Err(Error::UnexpectedWriteOnlyDescriptor);
-        }
-
-        let hdr_desc_addr = hdr_desc
-            .addr()
-            .translate_gva(access_platform, hdr_desc.len() as usize)
-            .map_err(|e| Error::GuestMemory(GuestMemoryError::IOError(e)))?;
-
-        let mut req = Request {
-            request_type: request_type(desc_chain.memory(), hdr_desc_addr)?,
-            sector: sector(desc_chain.memory(), hdr_desc_addr)?,
-            data_descriptors: SmallVec::with_capacity(DEFAULT_DESCRIPTOR_VEC_SIZE),
-            status_addr: GuestAddress(0),
-            writeback: true,
-            aligned_operations: SmallVec::with_capacity(DEFAULT_DESCRIPTOR_VEC_SIZE),
-            start: Instant::now(),
-        };
-
-        let status_desc;
-        let mut desc = desc_chain
-            .next()
-            .ok_or(Error::DescriptorChainTooShort)
-            .inspect_err(|_| {
-                error!("Only head descriptor present: request = {req:?}");
-            })?;
-
-        if desc.has_next() {
-            req.data_descriptors.reserve_exact(1);
-            while desc.has_next() {
-                if desc.is_write_only() && req.request_type == RequestType::Out {
-                    return Err(Error::UnexpectedWriteOnlyDescriptor);
-                }
-                if desc.is_write_only() && req.request_type == RequestType::Discard {
-                    return Err(Error::UnexpectedWriteOnlyDescriptor);
-                }
-                if desc.is_write_only() && req.request_type == RequestType::WriteZeroes {
-                    return Err(Error::UnexpectedWriteOnlyDescriptor);
-                }
-                if !desc.is_write_only() && req.request_type == RequestType::In {
-                    return Err(Error::UnexpectedReadOnlyDescriptor);
-                }
-                if !desc.is_write_only() && req.request_type == RequestType::GetDeviceId {
-                    return Err(Error::UnexpectedReadOnlyDescriptor);
-                }
-
-                req.data_descriptors.push((
-                    desc.addr()
-                        .translate_gva(access_platform, desc.len() as usize)
-                        .map_err(|e| Error::GuestMemory(GuestMemoryError::IOError(e)))?,
-                    desc.len(),
-                ));
-                desc = desc_chain
-                    .next()
-                    .ok_or(Error::DescriptorChainTooShort)
-                    .inspect_err(|_| {
-                        error!("DescriptorChain corrupted: request = {req:?}");
-                    })?;
-            }
-            status_desc = desc;
-        } else {
-            status_desc = desc;
-            // Only flush requests are allowed to skip the data descriptor.
-            if req.request_type != RequestType::Flush {
-                error!("Need a data descriptor: request = {req:?}");
-                return Err(Error::DescriptorChainTooShort);
-            }
-        }
-
-        // The status MUST always be writable.
-        if !status_desc.is_write_only() {
-            return Err(Error::UnexpectedReadOnlyDescriptor);
-        }
-
-        if status_desc.len() < 1 {
-            return Err(Error::DescriptorLengthTooSmall);
-        }
-
-        req.status_addr = status_desc
-            .addr()
-            .translate_gva(access_platform, status_desc.len() as usize)
-            .map_err(|e| Error::GuestMemory(GuestMemoryError::IOError(e)))?;
-
-        Ok(req)
-    }
-
-    pub fn execute<T: Seek + Read + Write, B: Bitmap + 'static>(
-        &self,
-        disk: &mut T,
-        disk_nsectors: u64,
-        mem: &vm_memory::GuestMemoryMmap<B>,
-        serial: &[u8],
-    ) -> Result<u32, ExecuteError> {
-        self.check_data_bounds(disk_nsectors)?;
-
-        disk.seek(SeekFrom::Start(self.sector << SECTOR_SHIFT))
-            .map_err(ExecuteError::Seek)?;
-        let mut len = 0;
-        for (data_addr, data_len) in &self.data_descriptors {
-            match self.request_type {
-                RequestType::In => {
-                    let mut buf = vec![0u8; *data_len as usize];
-                    disk.read_exact(&mut buf).map_err(ExecuteError::ReadExact)?;
-                    mem.read_exact_volatile_from(
-                        *data_addr,
-                        &mut buf.as_slice(),
-                        *data_len as usize,
-                    )
-                    .map_err(ExecuteError::Read)?;
-                    len += data_len;
-                }
-                RequestType::Out => {
-                    let mut buf: Vec<u8> = Vec::new();
-                    mem.write_all_volatile_to(*data_addr, &mut buf, *data_len as usize)
-                        .map_err(ExecuteError::Write)?;
-                    disk.write_all(&buf).map_err(ExecuteError::WriteAll)?;
-                    if !self.writeback {
-                        disk.flush().map_err(ExecuteError::Flush)?;
-                    }
-                }
-                RequestType::Flush => disk.flush().map_err(ExecuteError::Flush)?,
-                RequestType::GetDeviceId => {
-                    if (*data_len as usize) < serial.len() {
-                        return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                    }
-                    mem.write_slice(serial, *data_addr)
-                        .map_err(ExecuteError::Write)?;
-                }
-                RequestType::Discard => {
-                    return Err(ExecuteError::Unsupported(VIRTIO_BLK_T_DISCARD));
-                }
-                RequestType::WriteZeroes => {
-                    return Err(ExecuteError::Unsupported(VIRTIO_BLK_T_WRITE_ZEROES));
-                }
-                RequestType::Unsupported(t) => return Err(ExecuteError::Unsupported(t)),
-            }
-        }
-        Ok(len)
-    }
-
-    pub fn execute_async<B: Bitmap + 'static>(
-        &mut self,
-        mem: &vm_memory::GuestMemoryMmap<B>,
-        disk_nsectors: u64,
-        disk_image: &mut dyn AsyncIo,
-        serial: &[u8],
-        disable_sector0_writes: bool,
-        user_data: u64,
-    ) -> Result<ExecuteAsync, ExecuteError> {
-        let sector = self.sector;
-        let request_type = self.request_type;
-        let offset = (sector << SECTOR_SHIFT) as libc::off_t;
-        let alignment = disk_image.alignment();
-
-        self.check_data_bounds(disk_nsectors)?;
-
-        let mut iovecs: SmallVec<[libc::iovec; DEFAULT_DESCRIPTOR_VEC_SIZE]> =
-            SmallVec::with_capacity(self.data_descriptors.len());
-        for &(data_addr, data_len) in &self.data_descriptors {
-            let _: u32 = data_len; // compiler-checked documentation
-            const _: () = assert!(
-                core::mem::size_of::<u32>() <= core::mem::size_of::<usize>(),
-                "unsupported platform"
-            );
-            if data_len == 0 {
-                continue;
-            }
-            let data_len = data_len as usize;
-
-            let origin_ptr = mem
-                .get_slice(data_addr, data_len)
-                .map_err(ExecuteError::GetHostAddress)?;
-            assert!(origin_ptr.len() >= data_len);
-            let origin_ptr = origin_ptr.ptr_guard_mut();
-
-            // O_DIRECT requires buffer addresses to be aligned to the
-            // backend device's logical block size. In case it's not properly
-            // aligned, an intermediate buffer is created with the correct
-            // alignment, and a copy from/to the origin buffer is performed,
-            // depending on the type of operation.
-            let iov_base = if (origin_ptr.as_ptr() as u64).is_multiple_of(alignment) {
-                origin_ptr.as_ptr().cast()
-            } else {
-                let mut aligned_op = AlignedOperation::new(data_addr, data_len, alignment as usize)
-                    .map_err(ExecuteError::TemporaryBufferAllocation)?;
-
-                // We need to perform the copy beforehand in case we're writing
-                // data out.
-                if request_type == RequestType::Out {
-                    mem.read_slice(aligned_op.as_bytes_mut(), data_addr)
-                        .map_err(ExecuteError::Read)?;
-                }
-
-                let aligned_ptr = aligned_op.as_mut_ptr();
-                self.aligned_operations.push(aligned_op);
-
-                aligned_ptr.cast()
-            };
-
-            let iovec = libc::iovec {
-                iov_base,
-                iov_len: data_len as libc::size_t,
-            };
-            iovecs.push(iovec);
-        }
-
-        let mut ret = ExecuteAsync {
-            async_complete: true,
-            batch_request: None,
-        };
-        // Queue operations expected to be submitted.
-        match request_type {
-            RequestType::In => {
-                for (data_addr, data_len) in &self.data_descriptors {
-                    mem.get_slice(*data_addr, *data_len as usize)
-                        .map_err(ExecuteError::GetHostAddress)?
-                        .bitmap()
-                        .mark_dirty(0, *data_len as usize);
-                }
-                if disk_image.batch_requests_enabled() {
-                    ret.batch_request = Some(BatchRequest {
-                        offset,
-                        iovecs,
-                        user_data,
-                        request_type,
-                    });
-                } else {
-                    disk_image
-                        .read_vectored(offset, &iovecs, user_data)
-                        .map_err(ExecuteError::AsyncRead)?;
-                }
-            }
-            RequestType::Out => {
-                if disk_image.batch_requests_enabled() {
-                    ret.batch_request = Some(BatchRequest {
-                        offset,
-                        iovecs,
-                        user_data,
-                        request_type,
-                    });
-                } else {
-                    disk_image
-                        .write_vectored(offset, &iovecs, user_data)
-                        .map_err(ExecuteError::AsyncWrite)?;
-                }
-            }
-            RequestType::Flush => {
-                disk_image
-                    .fsync(Some(user_data))
-                    .map_err(ExecuteError::AsyncFlush)?;
-            }
-            RequestType::GetDeviceId => {
-                let (data_addr, data_len) = if self.data_descriptors.len() == 1 {
-                    (self.data_descriptors[0].0, self.data_descriptors[0].1)
-                } else {
-                    return Err(ExecuteError::BadRequest(Error::TooManyDescriptors));
-                };
-                if (data_len as usize) < serial.len() {
-                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                }
-                mem.write_slice(serial, data_addr)
-                    .map_err(ExecuteError::Write)?;
-                ret.async_complete = false;
-                return Ok(ret);
-            }
-            RequestType::Discard => {
-                let (data_addr, data_len) = if self.data_descriptors.len() == 1 {
-                    (self.data_descriptors[0].0, self.data_descriptors[0].1)
-                } else {
-                    return Err(ExecuteError::BadRequest(Error::TooManyDescriptors));
-                };
-
-                if data_len < DISCARD_WZ_SEG_SIZE {
-                    return Err(ExecuteError::BadRequest(Error::DescriptorLengthTooSmall));
-                }
-                if data_len > DISCARD_WZ_MAX_PAYLOAD {
-                    return Err(ExecuteError::BadRequest(Error::TooManySegments(
-                        data_len.div_ceil(DISCARD_WZ_SEG_SIZE),
-                    )));
-                }
-
-                let mut discard_sector = [0u8; 8];
-                let mut discard_num_sectors = [0u8; 4];
-                let mut discard_flags = [0u8; 4];
-
-                let sector_addr = data_addr.checked_add(DISCARD_WZ_SECTOR_OFFSET).unwrap();
-                mem.read_slice(&mut discard_sector, sector_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let num_sectors_addr = data_addr
-                    .checked_add(DISCARD_WZ_NUM_SECTORS_OFFSET)
-                    .unwrap();
-                mem.read_slice(&mut discard_num_sectors, num_sectors_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let flags_addr = data_addr.checked_add(DISCARD_WZ_FLAGS_OFFSET).unwrap();
-                mem.read_slice(&mut discard_flags, flags_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let discard_flags = u32::from_le_bytes(discard_flags);
-                // Per virtio spec v1.2 reject discard if any flag is set, including unmap.
-                if discard_flags != 0 {
-                    warn!("Unsupported flags {discard_flags:#x} in discard request");
-                    return Err(ExecuteError::UnsupportedFlags {
-                        request_type: VIRTIO_BLK_T_DISCARD,
-                        flags: discard_flags,
-                    });
-                }
-
-                let discard_sector = u64::from_le_bytes(discard_sector);
-
-                if discard_sector == 0 && disable_sector0_writes {
-                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                }
-
-                let discard_num_sectors = u32::from_le_bytes(discard_num_sectors);
-
-                let top = discard_sector
-                    .checked_add(discard_num_sectors as u64)
-                    .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
-                if top > disk_nsectors {
-                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                }
-
-                let discard_offset = discard_sector * SECTOR_SIZE;
-                let discard_length = (discard_num_sectors as u64) * SECTOR_SIZE;
-
-                disk_image
-                    .punch_hole(discard_offset, discard_length, user_data)
-                    .map_err(ExecuteError::AsyncPunchHole)?;
-            }
-            RequestType::WriteZeroes => {
-                let (data_addr, data_len) = if self.data_descriptors.len() == 1 {
-                    (self.data_descriptors[0].0, self.data_descriptors[0].1)
-                } else {
-                    return Err(ExecuteError::BadRequest(Error::TooManyDescriptors));
-                };
-
-                if data_len < DISCARD_WZ_SEG_SIZE {
-                    return Err(ExecuteError::BadRequest(Error::DescriptorLengthTooSmall));
-                }
-                if data_len > DISCARD_WZ_MAX_PAYLOAD {
-                    return Err(ExecuteError::BadRequest(Error::TooManySegments(
-                        data_len.div_ceil(DISCARD_WZ_SEG_SIZE),
-                    )));
-                }
-
-                let mut wz_sector = [0u8; 8];
-                let mut wz_num_sectors = [0u8; 4];
-                let mut wz_flags = [0u8; 4];
-
-                let sector_addr = data_addr.checked_add(DISCARD_WZ_SECTOR_OFFSET).unwrap();
-                mem.read_slice(&mut wz_sector, sector_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let num_sectors_addr = data_addr
-                    .checked_add(DISCARD_WZ_NUM_SECTORS_OFFSET)
-                    .unwrap();
-                mem.read_slice(&mut wz_num_sectors, num_sectors_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let flags_addr = data_addr.checked_add(DISCARD_WZ_FLAGS_OFFSET).unwrap();
-                mem.read_slice(&mut wz_flags, flags_addr)
-                    .map_err(ExecuteError::Read)?;
-
-                let wz_sector = u64::from_le_bytes(wz_sector);
-                let wz_num_sectors = u32::from_le_bytes(wz_num_sectors);
-
-                let wz_flags = u32::from_le_bytes(wz_flags);
-                // Per virtio spec v1.2 reject write zeroes if any unknown flag is set.
-                if (wz_flags & !VIRTIO_BLK_WRITE_ZEROES_FLAG_UNMAP) != 0 {
-                    warn!("Unsupported flags {wz_flags:#x} in write zeroes request");
-                    return Err(ExecuteError::UnsupportedFlags {
-                        request_type: VIRTIO_BLK_T_WRITE_ZEROES,
-                        flags: wz_flags,
-                    });
-                }
-
-                let wz_offset = wz_sector * SECTOR_SIZE;
-                if wz_offset == 0 && disable_sector0_writes {
-                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                }
-
-                let top = wz_sector
-                    .checked_add(wz_num_sectors as u64)
-                    .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
-                if top > disk_nsectors {
-                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-                }
-
-                let wz_length = (wz_num_sectors as u64) * SECTOR_SIZE;
-
-                if wz_flags & VIRTIO_BLK_WRITE_ZEROES_FLAG_UNMAP != 0 {
-                    disk_image
-                        .punch_hole(wz_offset, wz_length, user_data)
-                        .map_err(ExecuteError::AsyncPunchHole)?;
-                } else {
-                    disk_image
-                        .write_zeroes(wz_offset, wz_length, user_data)
-                        .map_err(ExecuteError::AsyncWriteZeroes)?;
-                }
-            }
-            RequestType::Unsupported(t) => return Err(ExecuteError::Unsupported(t)),
-        }
-
-        Ok(ret)
-    }
-
-    pub fn complete_async<B: Bitmap + 'static>(
-        &mut self,
-        mem: &vm_memory::GuestMemoryMmap<B>,
-    ) -> Result<(), Error> {
-        for aligned_op in self.aligned_operations.drain(..) {
-            // We need to perform the copy after the data has been read inside
-            // the aligned buffer in case we're reading data in.
-            if self.request_type == RequestType::In {
-                mem.write_slice(aligned_op.as_bytes(), aligned_op.data_addr())
-                    .map_err(Error::GuestMemory)?;
-            }
-        }
-
-        Ok(())
-    }
-
-    #[inline]
-    pub fn data_descriptors(
-        &self,
-    ) -> &SmallVec<[(GuestAddress, u32); DEFAULT_DESCRIPTOR_VEC_SIZE]> {
-        &self.data_descriptors
-    }
-
-    #[inline]
-    pub fn status_addr(&self) -> GuestAddress {
-        self.status_addr
-    }
-
-    #[inline]
-    pub fn start(&self) -> Instant {
-        self.start
-    }
-
-    #[inline]
-    pub fn sector(&self) -> u64 {
-        self.sector
-    }
-
-    #[inline]
-    pub fn request_type(&self) -> RequestType {
-        self.request_type
-    }
-
-    /// For In and Out requests, checks that the descriptors collectively fit in a backing disk of
-    /// the given size. Returns `Ok(())` if they fit, or `ExecuteError::BadRequest` otherwise.
-    fn check_data_bounds(&self, disk_nsectors: u64) -> Result<(), ExecuteError> {
-        if !matches!(self.request_type, RequestType::In | RequestType::Out) {
-            return Ok(());
-        }
-        let mut total_bytes: u64 = 0;
-        for (_, data_len) in &self.data_descriptors {
-            total_bytes = total_bytes
-                .checked_add(u64::from(*data_len))
-                .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
-        }
-        if total_bytes == 0 {
-            return Ok(());
-        }
-        let total_sectors = total_bytes.div_ceil(SECTOR_SIZE);
-        let end_sector = self
-            .sector
-            .checked_add(total_sectors)
-            .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
-        if end_sector > disk_nsectors {
-            return Err(ExecuteError::BadRequest(Error::InvalidOffset));
-        }
-        Ok(())
-    }
-}

block/src/vhdx/vhdx_io.rs

@@ -1,219 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::fs::File;
-use std::io::{self, Read, Seek, SeekFrom, Write};
-
-use remain::sorted;
-use thiserror::Error;
-
-use crate::vhdx::vhdx_bat::{self, BatEntry, VhdxBatError};
-use crate::vhdx::vhdx_metadata::{self, DiskSpec};
-
-const SECTOR_SIZE: u64 = 512;
-
-#[sorted]
-#[derive(Error, Debug)]
-pub enum VhdxIoError {
-    #[error("Invalid BAT entry state")]
-    InvalidBatEntryState,
-    #[error("Invalid BAT entry count")]
-    InvalidBatIndex,
-    #[error("Invalid disk size")]
-    InvalidDiskSize,
-    #[error("Failed reading sector blocks from file {0}")]
-    ReadSectorBlock(#[source] io::Error),
-    #[error("Failed changing file length {0}")]
-    ResizeFile(#[source] io::Error),
-    #[error("Differencing mode is not supported yet")]
-    UnsupportedMode,
-    #[error("Failed writing BAT to file {0}")]
-    WriteBat(#[source] VhdxBatError),
-}
-
-pub type Result<T> = std::result::Result<T, VhdxIoError>;
-
-macro_rules! align {
-    ($n:expr, $align:expr) => {{ $n.div_ceil($align) * $align }};
-}
-
-#[derive(Default)]
-struct Sector {
-    bat_index: u64,
-    free_sectors: u64,
-    free_bytes: u64,
-    file_offset: u64,
-    block_offset: u64,
-}
-
-impl Sector {
-    /// Translate sector index and count of data in file to actual offsets and
-    /// BAT index.
-    pub fn new(
-        disk_spec: &DiskSpec,
-        bat: &[BatEntry],
-        sector_index: u64,
-        sector_count: u64,
-    ) -> Result<Sector> {
-        let mut sector = Sector::default();
-
-        sector.bat_index = sector_index / disk_spec.sectors_per_block as u64;
-        sector.block_offset = sector_index % disk_spec.sectors_per_block as u64;
-        sector.free_sectors = disk_spec.sectors_per_block as u64 - sector.block_offset;
-        if sector.free_sectors > sector_count {
-            sector.free_sectors = sector_count;
-        }
-
-        sector.free_bytes = sector.free_sectors * disk_spec.logical_sector_size as u64;
-        sector.block_offset *= disk_spec.logical_sector_size as u64;
-
-        let bat_entry = match bat.get(sector.bat_index as usize) {
-            Some(entry) => entry.0,
-            None => {
-                return Err(VhdxIoError::InvalidBatIndex);
-            }
-        };
-        sector.file_offset = bat_entry & vhdx_bat::BAT_FILE_OFF_MASK;
-        if sector.file_offset != 0 {
-            sector.file_offset += sector.block_offset;
-        }
-
-        Ok(sector)
-    }
-}
-
-/// VHDx IO read routine: requires relative sector index and count for the
-/// requested data.
-pub fn read(
-    f: &mut File,
-    buf: &mut [u8],
-    disk_spec: &DiskSpec,
-    bat: &[BatEntry],
-    mut sector_index: u64,
-    mut sector_count: u64,
-) -> Result<usize> {
-    if disk_spec.has_parent {
-        return Err(VhdxIoError::UnsupportedMode);
-    }
-
-    let mut read_count: usize = 0;
-    while sector_count > 0 {
-        let sector = Sector::new(disk_spec, bat, sector_index, sector_count)?;
-
-        let bat_entry = match bat.get(sector.bat_index as usize) {
-            Some(entry) => entry.0,
-            None => {
-                return Err(VhdxIoError::InvalidBatIndex);
-            }
-        };
-
-        match bat_entry & vhdx_bat::BAT_STATE_BIT_MASK {
-            vhdx_bat::PAYLOAD_BLOCK_NOT_PRESENT
-            | vhdx_bat::PAYLOAD_BLOCK_UNDEFINED
-            | vhdx_bat::PAYLOAD_BLOCK_UNMAPPED
-            | vhdx_bat::PAYLOAD_BLOCK_ZERO => {}
-            vhdx_bat::PAYLOAD_BLOCK_FULLY_PRESENT => {
-                f.seek(SeekFrom::Start(sector.file_offset))
-                    .map_err(VhdxIoError::ReadSectorBlock)?;
-                f.read_exact(
-                    &mut buf
-                        [read_count..(read_count + (sector.free_sectors * SECTOR_SIZE) as usize)],
-                )
-                .map_err(VhdxIoError::ReadSectorBlock)?;
-            }
-            vhdx_bat::PAYLOAD_BLOCK_PARTIALLY_PRESENT => {
-                return Err(VhdxIoError::UnsupportedMode);
-            }
-            _ => {
-                return Err(VhdxIoError::InvalidBatEntryState);
-            }
-        }
-        sector_count -= sector.free_sectors;
-        sector_index += sector.free_sectors;
-        read_count += sector.free_bytes as usize;
-    }
-    Ok(read_count)
-}
-
-/// VHDx IO write routine: requires relative sector index and count for the
-/// requested data.
-pub fn write(
-    f: &mut File,
-    buf: &[u8],
-    disk_spec: &mut DiskSpec,
-    bat_offset: u64,
-    bat: &mut [BatEntry],
-    mut sector_index: u64,
-    mut sector_count: u64,
-) -> Result<usize> {
-    if disk_spec.has_parent {
-        return Err(VhdxIoError::UnsupportedMode);
-    }
-
-    let mut write_count: usize = 0;
-    while sector_count > 0 {
-        let sector = Sector::new(disk_spec, bat, sector_index, sector_count)?;
-
-        let bat_entry = match bat.get(sector.bat_index as usize) {
-            Some(entry) => entry.0,
-            None => {
-                return Err(VhdxIoError::InvalidBatIndex);
-            }
-        };
-
-        match bat_entry & vhdx_bat::BAT_STATE_BIT_MASK {
-            vhdx_bat::PAYLOAD_BLOCK_NOT_PRESENT
-            | vhdx_bat::PAYLOAD_BLOCK_UNDEFINED
-            | vhdx_bat::PAYLOAD_BLOCK_UNMAPPED
-            | vhdx_bat::PAYLOAD_BLOCK_ZERO => {
-                let file_offset =
-                    align!(disk_spec.image_size, vhdx_metadata::BLOCK_SIZE_MIN as u64);
-                let new_size = file_offset
-                    .checked_add(disk_spec.block_size as u64)
-                    .ok_or(VhdxIoError::InvalidDiskSize)?;
-
-                f.set_len(new_size).map_err(VhdxIoError::ResizeFile)?;
-                disk_spec.image_size = new_size;
-
-                let new_bat_entry = file_offset
-                    | (vhdx_bat::PAYLOAD_BLOCK_FULLY_PRESENT & vhdx_bat::BAT_STATE_BIT_MASK);
-                bat[sector.bat_index as usize] = BatEntry(new_bat_entry);
-                BatEntry::write_bat_entries(f, bat_offset, bat).map_err(VhdxIoError::WriteBat)?;
-
-                if file_offset < vhdx_metadata::BLOCK_SIZE_MIN as u64 {
-                    break;
-                }
-
-                f.seek(SeekFrom::Start(file_offset))
-                    .map_err(VhdxIoError::ReadSectorBlock)?;
-                f.write_all(
-                    &buf[write_count..(write_count + (sector.free_sectors * SECTOR_SIZE) as usize)],
-                )
-                .map_err(VhdxIoError::ReadSectorBlock)?;
-            }
-            vhdx_bat::PAYLOAD_BLOCK_FULLY_PRESENT => {
-                if sector.file_offset < vhdx_metadata::BLOCK_SIZE_MIN as u64 {
-                    break;
-                }
-
-                f.seek(SeekFrom::Start(sector.file_offset))
-                    .map_err(VhdxIoError::ReadSectorBlock)?;
-                f.write_all(
-                    &buf[write_count..(write_count + (sector.free_sectors * SECTOR_SIZE) as usize)],
-                )
-                .map_err(VhdxIoError::ReadSectorBlock)?;
-            }
-            vhdx_bat::PAYLOAD_BLOCK_PARTIALLY_PRESENT => {
-                return Err(VhdxIoError::UnsupportedMode);
-            }
-            _ => {
-                return Err(VhdxIoError::InvalidBatEntryState);
-            }
-        }
-        sector_count -= sector.free_sectors;
-        sector_index += sector.free_sectors;
-        write_count += sector.free_bytes as usize;
-    }
-    Ok(write_count)
-}

block/src/vhdx_sync.rs

@@ -1,180 +0,0 @@
-// Copyright © 2021 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::collections::VecDeque;
-use std::fs::File;
-use std::os::fd::AsRawFd;
-use std::sync::{Arc, Mutex};
-
-use vmm_sys_util::eventfd::EventFd;
-
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult, BorrowedDiskFd, DiskFileError};
-use crate::error::{BlockError, BlockErrorKind, BlockResult, ErrorOp};
-use crate::vhdx::{Vhdx, VhdxError};
-use crate::{AsyncAdaptor, BlockBackend, Error, disk_file};
-
-#[derive(Debug)]
-pub struct VhdxDiskSync {
-    // FIXME: The Mutex serializes all VHDX I/O operations across queues, which
-    // is necessary for correctness but eliminates any parallelism benefit from
-    // multiqueue. Vhdx::clone() shares the underlying file description across
-    // threads, so concurrent I/O from multiple queues races on the file offset
-    // causing data corruption.
-    //
-    // A proper fix would require restructuring the VHDX I/O path so that data
-    // operations can proceed in parallel with independent file descriptors.
-    vhdx_file: Arc<Mutex<Vhdx>>,
-}
-
-impl VhdxDiskSync {
-    pub fn new(f: File) -> BlockResult<Self> {
-        Ok(VhdxDiskSync {
-            vhdx_file: Arc::new(Mutex::new(Vhdx::new(f).map_err(|e| {
-                let kind = match &e {
-                    VhdxError::NotVhdx(_)
-                    | VhdxError::ParseVhdxHeader(_)
-                    | VhdxError::ParseVhdxMetadata(_)
-                    | VhdxError::ParseVhdxRegionEntry(_) => BlockErrorKind::InvalidFormat,
-                    VhdxError::ReadBatEntry(_) => BlockErrorKind::CorruptImage,
-                    VhdxError::ReadFailed(_) | VhdxError::WriteFailed(_) => BlockErrorKind::Io,
-                };
-                BlockError::new(kind, e).with_op(ErrorOp::Open)
-            })?)),
-        })
-    }
-}
-
-impl disk_file::DiskSize for VhdxDiskSync {
-    fn logical_size(&self) -> BlockResult<u64> {
-        Ok(self.vhdx_file.lock().unwrap().virtual_disk_size())
-    }
-}
-
-impl disk_file::PhysicalSize for VhdxDiskSync {
-    fn physical_size(&self) -> BlockResult<u64> {
-        self.vhdx_file
-            .lock()
-            .unwrap()
-            .physical_size()
-            .map_err(|e| match e {
-                Error::GetFileMetadata(io) => {
-                    BlockError::new(BlockErrorKind::Io, Error::GetFileMetadata(io))
-                }
-                _ => unreachable!("unexpected error from Vhdx::physical_size(): {e}"),
-            })
-    }
-}
-
-impl disk_file::DiskFd for VhdxDiskSync {
-    fn fd(&self) -> BorrowedDiskFd<'_> {
-        BorrowedDiskFd::new(self.vhdx_file.lock().unwrap().as_raw_fd())
-    }
-}
-
-impl disk_file::Geometry for VhdxDiskSync {}
-
-impl disk_file::SparseCapable for VhdxDiskSync {}
-
-impl disk_file::Resizable for VhdxDiskSync {
-    fn resize(&mut self, _size: u64) -> BlockResult<()> {
-        Err(BlockError::new(
-            BlockErrorKind::UnsupportedFeature,
-            DiskFileError::ResizeError(std::io::Error::other("resize not supported for VHDX")),
-        )
-        .with_op(ErrorOp::Resize))
-    }
-}
-
-impl disk_file::DiskFile for VhdxDiskSync {}
-
-impl disk_file::AsyncDiskFile for VhdxDiskSync {
-    fn try_clone(&self) -> BlockResult<Box<dyn disk_file::AsyncDiskFile>> {
-        Ok(Box::new(VhdxDiskSync {
-            vhdx_file: Arc::clone(&self.vhdx_file),
-        }))
-    }
-
-    fn create_async_io(&self, _ring_depth: u32) -> BlockResult<Box<dyn AsyncIo>> {
-        Ok(Box::new(VhdxSync::new(Arc::clone(&self.vhdx_file))))
-    }
-}
-
-pub struct VhdxSync {
-    vhdx_file: Arc<Mutex<Vhdx>>,
-    eventfd: EventFd,
-    completion_list: VecDeque<(u64, i32)>,
-}
-
-impl VhdxSync {
-    pub fn new(vhdx_file: Arc<Mutex<Vhdx>>) -> Self {
-        VhdxSync {
-            vhdx_file,
-            eventfd: EventFd::new(libc::EFD_NONBLOCK)
-                .expect("Failed creating EventFd for VhdxSync"),
-            completion_list: VecDeque::new(),
-        }
-    }
-}
-
-impl AsyncAdaptor for Vhdx {}
-
-impl AsyncIo for VhdxSync {
-    fn notifier(&self) -> &EventFd {
-        &self.eventfd
-    }
-
-    fn read_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        self.vhdx_file.lock().unwrap().read_vectored_sync(
-            offset,
-            iovecs,
-            user_data,
-            &self.eventfd,
-            &mut self.completion_list,
-        )
-    }
-
-    fn write_vectored(
-        &mut self,
-        offset: libc::off_t,
-        iovecs: &[libc::iovec],
-        user_data: u64,
-    ) -> AsyncIoResult<()> {
-        self.vhdx_file.lock().unwrap().write_vectored_sync(
-            offset,
-            iovecs,
-            user_data,
-            &self.eventfd,
-            &mut self.completion_list,
-        )
-    }
-
-    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
-        self.vhdx_file.lock().unwrap().fsync_sync(
-            user_data,
-            &self.eventfd,
-            &mut self.completion_list,
-        )
-    }
-
-    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
-        self.completion_list.pop_front()
-    }
-
-    fn punch_hole(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::PunchHole(std::io::Error::other(
-            "punch_hole not supported for VHDX",
-        )))
-    }
-
-    fn write_zeroes(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::WriteZeroes(std::io::Error::other(
-            "write_zeroes not supported for VHDX",
-        )))
-    }
-}

block_util/Cargo.toml

@@ -0,0 +1,25 @@
+[package]
+name = "block_util"
+version = "0.1.0"
+authors = ["The Cloud Hypervisor Authors"]
+edition = "2021"
+
+[features]
+default = []
+
+[dependencies]
+io-uring = "0.5.13"
+libc = "0.2.139"
+log = "0.4.17"
+qcow = { path = "../qcow" }
+smallvec = "1.10.0"
+thiserror = "1.0.39"
+versionize = "0.1.10"
+versionize_derive = "0.1.4"
+vhdx = { path = "../vhdx" }
+virtio-bindings = { version = "0.2.0", features = ["virtio-v5_0_0"] }
+virtio-queue = "0.7.1"
+vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
+vm-virtio = { path = "../vm-virtio" }
+vmm-sys-util = "0.11.0"
+

block_util/src/async_io.rs

@@ -0,0 +1,145 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+use libc::{ioctl, S_IFBLK, S_IFMT};
+use std::fs::File;
+use std::os::unix::io::AsRawFd;
+use thiserror::Error;
+use vmm_sys_util::eventfd::EventFd;
+use vmm_sys_util::{ioctl_io_nr, ioctl_ioc_nr};
+
+#[derive(Error, Debug)]
+pub enum DiskFileError {
+    /// Failed getting disk file size.
+    #[error("Failed getting disk file size: {0}")]
+    Size(#[source] std::io::Error),
+    /// Failed creating a new AsyncIo.
+    #[error("Failed creating a new AsyncIo: {0}")]
+    NewAsyncIo(#[source] std::io::Error),
+}
+
+#[derive(Debug)]
+pub struct DiskTopology {
+    pub logical_block_size: u64,
+    pub physical_block_size: u64,
+    pub minimum_io_size: u64,
+    pub optimal_io_size: u64,
+}
+
+impl Default for DiskTopology {
+    fn default() -> Self {
+        Self {
+            logical_block_size: 512,
+            physical_block_size: 512,
+            minimum_io_size: 512,
+            optimal_io_size: 0,
+        }
+    }
+}
+
+ioctl_io_nr!(BLKSSZGET, 0x12, 104);
+ioctl_io_nr!(BLKPBSZGET, 0x12, 123);
+ioctl_io_nr!(BLKIOMIN, 0x12, 120);
+ioctl_io_nr!(BLKIOOPT, 0x12, 121);
+
+enum BlockSize {
+    LogicalBlock,
+    PhysicalBlock,
+    MinimumIo,
+    OptimalIo,
+}
+
+impl DiskTopology {
+    fn is_block_device(f: &mut File) -> std::io::Result<bool> {
+        let mut stat = std::mem::MaybeUninit::<libc::stat>::uninit();
+        // SAFETY: FFI call with a valid fd and buffer
+        let ret = unsafe { libc::fstat(f.as_raw_fd(), stat.as_mut_ptr()) };
+        if ret != 0 {
+            return Err(std::io::Error::last_os_error());
+        }
+
+        // SAFETY: stat is valid at this point
+        let is_block = unsafe { (*stat.as_ptr()).st_mode & S_IFMT == S_IFBLK };
+        Ok(is_block)
+    }
+
+    // libc::ioctl() takes different types on different architectures
+    fn query_block_size(f: &mut File, block_size_type: BlockSize) -> std::io::Result<u64> {
+        let mut block_size = 0;
+        // SAFETY: FFI call with correct arguments
+        let ret = unsafe {
+            ioctl(
+                f.as_raw_fd(),
+                match block_size_type {
+                    BlockSize::LogicalBlock => BLKSSZGET(),
+                    BlockSize::PhysicalBlock => BLKPBSZGET(),
+                    BlockSize::MinimumIo => BLKIOMIN(),
+                    BlockSize::OptimalIo => BLKIOOPT(),
+                } as _,
+                &mut block_size,
+            )
+        };
+        if ret != 0 {
+            return Err(std::io::Error::last_os_error());
+        };
+
+        Ok(block_size)
+    }
+
+    pub fn probe(f: &mut File) -> std::io::Result<Self> {
+        if !Self::is_block_device(f)? {
+            return Ok(DiskTopology::default());
+        }
+
+        Ok(DiskTopology {
+            logical_block_size: Self::query_block_size(f, BlockSize::LogicalBlock)?,
+            physical_block_size: Self::query_block_size(f, BlockSize::PhysicalBlock)?,
+            minimum_io_size: Self::query_block_size(f, BlockSize::MinimumIo)?,
+            optimal_io_size: Self::query_block_size(f, BlockSize::OptimalIo)?,
+        })
+    }
+}
+
+pub type DiskFileResult<T> = std::result::Result<T, DiskFileError>;
+
+pub trait DiskFile: Send {
+    fn size(&mut self) -> DiskFileResult<u64>;
+    fn new_async_io(&self, ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>>;
+    fn topology(&mut self) -> DiskTopology {
+        DiskTopology::default()
+    }
+}
+
+#[derive(Error, Debug)]
+pub enum AsyncIoError {
+    /// Failed vectored reading from file.
+    #[error("Failed vectored reading from file: {0}")]
+    ReadVectored(#[source] std::io::Error),
+    /// Failed vectored writing to file.
+    #[error("Failed vectored writing to file: {0}")]
+    WriteVectored(#[source] std::io::Error),
+    /// Failed synchronizing file.
+    #[error("Failed synchronizing file: {0}")]
+    Fsync(#[source] std::io::Error),
+}
+
+pub type AsyncIoResult<T> = std::result::Result<T, AsyncIoError>;
+
+pub trait AsyncIo: Send {
+    fn notifier(&self) -> &EventFd;
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()>;
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()>;
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()>;
+    fn next_completed_request(&mut self) -> Option<(u64, i32)>;
+}

block_util/src/fixed_vhd_async.rs

@@ -2,14 +2,43 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use std::os::unix::io::RawFd;
-
+use crate::async_io::{
+    AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult,
+};
+use crate::raw_async::RawFileAsync;
+use crate::vhd::VhdFooter;
+use std::fs::File;
+use std::os::unix::io::{AsRawFd, RawFd};
 use vmm_sys_util::eventfd::EventFd;
 
-use crate::BatchRequest;
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
-use crate::error::BlockResult;
-use crate::raw_async::RawFileAsync;
+pub struct FixedVhdDiskAsync {
+    file: File,
+    size: u64,
+}
+
+impl FixedVhdDiskAsync {
+    pub fn new(mut file: File) -> std::io::Result<Self> {
+        let footer = VhdFooter::new(&mut file)?;
+
+        Ok(FixedVhdDiskAsync {
+            file,
+            size: footer.current_size(),
+        })
+    }
+}
+
+impl DiskFile for FixedVhdDiskAsync {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        Ok(self.size)
+    }
+
+    fn new_async_io(&self, ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(Box::new(
+            FixedVhdAsync::new(self.file.as_raw_fd(), ring_depth, self.size)
+                .map_err(DiskFileError::NewAsyncIo)?,
+        ) as Box<dyn AsyncIo>)
+    }
+}
 
 pub struct FixedVhdAsync {
     raw_file_async: RawFileAsync,
@@ -17,7 +46,7 @@ pub struct FixedVhdAsync {
 }
 
 impl FixedVhdAsync {
-    pub fn new(fd: RawFd, ring_depth: u32, size: u64) -> BlockResult<Self> {
+    pub fn new(fd: RawFd, ring_depth: u32, size: u64) -> std::io::Result<Self> {
         let raw_file_async = RawFileAsync::new(fd, ring_depth)?;
 
         Ok(FixedVhdAsync {
@@ -78,24 +107,4 @@ impl AsyncIo for FixedVhdAsync {
     fn next_completed_request(&mut self) -> Option<(u64, i32)> {
         self.raw_file_async.next_completed_request()
     }
-
-    fn punch_hole(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::PunchHole(std::io::Error::other(
-            "punch_hole not supported for fixed VHD",
-        )))
-    }
-
-    fn write_zeroes(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::WriteZeroes(std::io::Error::other(
-            "write_zeroes not supported for fixed VHD",
-        )))
-    }
-
-    fn batch_requests_enabled(&self) -> bool {
-        true
-    }
-
-    fn submit_batch_requests(&mut self, batch_request: &[BatchRequest]) -> AsyncIoResult<()> {
-        self.raw_file_async.submit_batch_requests(batch_request)
-    }
 }

block_util/src/fixed_vhd_sync.rs

@@ -2,12 +2,43 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use std::os::unix::io::RawFd;
-
+use crate::async_io::{
+    AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult,
+};
+use crate::raw_sync::RawFileSync;
+use crate::vhd::VhdFooter;
+use std::fs::File;
+use std::os::unix::io::{AsRawFd, RawFd};
 use vmm_sys_util::eventfd::EventFd;
 
-use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
-use crate::raw_sync::RawFileSync;
+pub struct FixedVhdDiskSync {
+    file: File,
+    size: u64,
+}
+
+impl FixedVhdDiskSync {
+    pub fn new(mut file: File) -> std::io::Result<Self> {
+        let footer = VhdFooter::new(&mut file)?;
+
+        Ok(FixedVhdDiskSync {
+            file,
+            size: footer.current_size(),
+        })
+    }
+}
+
+impl DiskFile for FixedVhdDiskSync {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        Ok(self.size)
+    }
+
+    fn new_async_io(&self, _ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(Box::new(
+            FixedVhdSync::new(self.file.as_raw_fd(), self.size)
+                .map_err(DiskFileError::NewAsyncIo)?,
+        ) as Box<dyn AsyncIo>)
+    }
+}
 
 pub struct FixedVhdSync {
     raw_file_sync: RawFileSync,
@@ -73,16 +104,4 @@ impl AsyncIo for FixedVhdSync {
     fn next_completed_request(&mut self) -> Option<(u64, i32)> {
         self.raw_file_sync.next_completed_request()
     }
-
-    fn punch_hole(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::PunchHole(std::io::Error::other(
-            "punch_hole not supported for fixed VHD",
-        )))
-    }
-
-    fn write_zeroes(&mut self, _offset: u64, _length: u64, _user_data: u64) -> AsyncIoResult<()> {
-        Err(AsyncIoError::WriteZeroes(std::io::Error::other(
-            "write_zeroes not supported for fixed VHD",
-        )))
-    }
 }

block_util/src/lib.rs

@@ -0,0 +1,722 @@
+// 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 LICENSE-BSD-3-Clause file.
+//
+// Copyright © 2020 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+#[macro_use]
+extern crate log;
+
+pub mod async_io;
+pub mod fixed_vhd_async;
+pub mod fixed_vhd_sync;
+pub mod qcow_sync;
+pub mod raw_async;
+pub mod raw_sync;
+pub mod vhd;
+pub mod vhdx_sync;
+
+use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
+use io_uring::{opcode, IoUring, Probe};
+use smallvec::SmallVec;
+use std::alloc::{alloc_zeroed, dealloc, Layout};
+use std::cmp;
+use std::collections::VecDeque;
+use std::convert::TryInto;
+use std::fs::File;
+use std::io::{self, IoSlice, IoSliceMut, Read, Seek, SeekFrom, Write};
+use std::os::linux::fs::MetadataExt;
+use std::path::Path;
+use std::result;
+use std::sync::Arc;
+use std::sync::MutexGuard;
+use std::time::Instant;
+use thiserror::Error;
+use versionize::{VersionMap, Versionize, VersionizeResult};
+use versionize_derive::Versionize;
+use virtio_bindings::virtio_blk::*;
+use virtio_queue::DescriptorChain;
+use vm_memory::{
+    bitmap::AtomicBitmap, bitmap::Bitmap, ByteValued, Bytes, GuestAddress, GuestMemory,
+    GuestMemoryError, GuestMemoryLoadGuard,
+};
+use vm_virtio::{AccessPlatform, Translatable};
+use vmm_sys_util::eventfd::EventFd;
+
+type GuestMemoryMmap = vm_memory::GuestMemoryMmap<AtomicBitmap>;
+
+const SECTOR_SHIFT: u8 = 9;
+pub const SECTOR_SIZE: u64 = 0x01 << SECTOR_SHIFT;
+
+#[derive(Error, Debug)]
+pub enum Error {
+    #[error("Guest gave us bad memory addresses")]
+    GuestMemory(GuestMemoryError),
+    #[error("Guest gave us offsets that would have overflowed a usize")]
+    CheckedOffset(GuestAddress, usize),
+    #[error("Guest gave us a write only descriptor that protocol says to read from")]
+    UnexpectedWriteOnlyDescriptor,
+    #[error("Guest gave us a read only descriptor that protocol says to write to")]
+    UnexpectedReadOnlyDescriptor,
+    #[error("Guest gave us too few descriptors in a descriptor chain")]
+    DescriptorChainTooShort,
+    #[error("Guest gave us a descriptor that was too short to use")]
+    DescriptorLengthTooSmall,
+    #[error("Getting a block's metadata fails for any reason")]
+    GetFileMetadata,
+    #[error("The requested operation would cause a seek beyond disk end")]
+    InvalidOffset,
+    #[error("The requested operation does not support multiple descriptors")]
+    TooManyDescriptors,
+}
+
+fn build_device_id(disk_path: &Path) -> result::Result<String, Error> {
+    let blk_metadata = match disk_path.metadata() {
+        Err(_) => return Err(Error::GetFileMetadata),
+        Ok(m) => m,
+    };
+    // This is how kvmtool does it.
+    let device_id = format!(
+        "{}{}{}",
+        blk_metadata.st_dev(),
+        blk_metadata.st_rdev(),
+        blk_metadata.st_ino()
+    );
+    Ok(device_id)
+}
+
+pub fn build_disk_image_id(disk_path: &Path) -> Vec<u8> {
+    let mut default_disk_image_id = vec![0; VIRTIO_BLK_ID_BYTES as usize];
+    match build_device_id(disk_path) {
+        Err(_) => {
+            warn!("Could not generate device id. We'll use a default.");
+        }
+        Ok(m) => {
+            // The kernel only knows to read a maximum of VIRTIO_BLK_ID_BYTES.
+            // This will also zero out any leftover bytes.
+            let disk_id = m.as_bytes();
+            let bytes_to_copy = cmp::min(disk_id.len(), VIRTIO_BLK_ID_BYTES as usize);
+            default_disk_image_id[..bytes_to_copy].clone_from_slice(&disk_id[..bytes_to_copy])
+        }
+    }
+    default_disk_image_id
+}
+
+#[derive(Error, Debug)]
+pub enum ExecuteError {
+    #[error("Bad request: {0}")]
+    BadRequest(Error),
+    #[error("Falied to flush: {0}")]
+    Flush(io::Error),
+    #[error("Failed to read: {0}")]
+    Read(GuestMemoryError),
+    #[error("Failed to seek: {0}")]
+    Seek(io::Error),
+    #[error("Failed to write: {0}")]
+    Write(GuestMemoryError),
+    #[error("Unsupported request: {0}")]
+    Unsupported(u32),
+    #[error("Failed to submit io uring: {0}")]
+    SubmitIoUring(io::Error),
+    #[error("Failed to get guest address: {0}")]
+    GetHostAddress(GuestMemoryError),
+    #[error("Failed to async read: {0}")]
+    AsyncRead(AsyncIoError),
+    #[error("Failed to async write: {0}")]
+    AsyncWrite(AsyncIoError),
+    #[error("failed to async flush: {0}")]
+    AsyncFlush(AsyncIoError),
+    #[error("Failed allocating a temporary buffer: {0}")]
+    TemporaryBufferAllocation(io::Error),
+}
+
+impl ExecuteError {
+    pub fn status(&self) -> u32 {
+        match *self {
+            ExecuteError::BadRequest(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::Flush(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::Read(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::Seek(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::Write(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::Unsupported(_) => VIRTIO_BLK_S_UNSUPP,
+            ExecuteError::SubmitIoUring(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::GetHostAddress(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::AsyncRead(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::AsyncWrite(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::AsyncFlush(_) => VIRTIO_BLK_S_IOERR,
+            ExecuteError::TemporaryBufferAllocation(_) => VIRTIO_BLK_S_IOERR,
+        }
+    }
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum RequestType {
+    In,
+    Out,
+    Flush,
+    GetDeviceId,
+    Unsupported(u32),
+}
+
+pub fn request_type(
+    mem: &GuestMemoryMmap,
+    desc_addr: GuestAddress,
+) -> result::Result<RequestType, Error> {
+    let type_ = mem.read_obj(desc_addr).map_err(Error::GuestMemory)?;
+    match type_ {
+        VIRTIO_BLK_T_IN => Ok(RequestType::In),
+        VIRTIO_BLK_T_OUT => Ok(RequestType::Out),
+        VIRTIO_BLK_T_FLUSH => Ok(RequestType::Flush),
+        VIRTIO_BLK_T_GET_ID => Ok(RequestType::GetDeviceId),
+        t => Ok(RequestType::Unsupported(t)),
+    }
+}
+
+fn sector(mem: &GuestMemoryMmap, desc_addr: GuestAddress) -> result::Result<u64, Error> {
+    const SECTOR_OFFSET: usize = 8;
+    let addr = match mem.checked_offset(desc_addr, SECTOR_OFFSET) {
+        Some(v) => v,
+        None => return Err(Error::CheckedOffset(desc_addr, SECTOR_OFFSET)),
+    };
+
+    mem.read_obj(addr).map_err(Error::GuestMemory)
+}
+
+#[derive(Debug)]
+pub struct AlignedOperation {
+    origin_ptr: u64,
+    aligned_ptr: u64,
+    size: usize,
+    layout: Layout,
+}
+
+#[derive(Debug)]
+pub struct Request {
+    pub request_type: RequestType,
+    pub sector: u64,
+    pub data_descriptors: SmallVec<[(GuestAddress, u32); 1]>,
+    pub status_addr: GuestAddress,
+    pub writeback: bool,
+    pub aligned_operations: SmallVec<[AlignedOperation; 1]>,
+    pub start: Instant,
+}
+
+impl Request {
+    pub fn parse(
+        desc_chain: &mut DescriptorChain<GuestMemoryLoadGuard<GuestMemoryMmap>>,
+        access_platform: Option<&Arc<dyn AccessPlatform>>,
+    ) -> result::Result<Request, Error> {
+        let hdr_desc = desc_chain
+            .next()
+            .ok_or(Error::DescriptorChainTooShort)
+            .map_err(|e| {
+                error!("Missing head descriptor");
+                e
+            })?;
+
+        // The head contains the request type which MUST be readable.
+        if hdr_desc.is_write_only() {
+            return Err(Error::UnexpectedWriteOnlyDescriptor);
+        }
+
+        let hdr_desc_addr = hdr_desc
+            .addr()
+            .translate_gva(access_platform, hdr_desc.len() as usize);
+
+        let mut req = Request {
+            request_type: request_type(desc_chain.memory(), hdr_desc_addr)?,
+            sector: sector(desc_chain.memory(), hdr_desc_addr)?,
+            data_descriptors: SmallVec::with_capacity(1),
+            status_addr: GuestAddress(0),
+            writeback: true,
+            aligned_operations: SmallVec::with_capacity(1),
+            start: Instant::now(),
+        };
+
+        let status_desc;
+        let mut desc = desc_chain
+            .next()
+            .ok_or(Error::DescriptorChainTooShort)
+            .map_err(|e| {
+                error!("Only head descriptor present: request = {:?}", req);
+                e
+            })?;
+
+        if !desc.has_next() {
+            status_desc = desc;
+            // Only flush requests are allowed to skip the data descriptor.
+            if req.request_type != RequestType::Flush {
+                error!("Need a data descriptor: request = {:?}", req);
+                return Err(Error::DescriptorChainTooShort);
+            }
+        } else {
+            req.data_descriptors.reserve_exact(1);
+            while desc.has_next() {
+                if desc.is_write_only() && req.request_type == RequestType::Out {
+                    return Err(Error::UnexpectedWriteOnlyDescriptor);
+                }
+                if !desc.is_write_only() && req.request_type == RequestType::In {
+                    return Err(Error::UnexpectedReadOnlyDescriptor);
+                }
+                if !desc.is_write_only() && req.request_type == RequestType::GetDeviceId {
+                    return Err(Error::UnexpectedReadOnlyDescriptor);
+                }
+
+                req.data_descriptors.push((
+                    desc.addr()
+                        .translate_gva(access_platform, desc.len() as usize),
+                    desc.len(),
+                ));
+                desc = desc_chain
+                    .next()
+                    .ok_or(Error::DescriptorChainTooShort)
+                    .map_err(|e| {
+                        error!("DescriptorChain corrupted: request = {:?}", req);
+                        e
+                    })?;
+            }
+            status_desc = desc;
+        }
+
+        // The status MUST always be writable.
+        if !status_desc.is_write_only() {
+            return Err(Error::UnexpectedReadOnlyDescriptor);
+        }
+
+        if status_desc.len() < 1 {
+            return Err(Error::DescriptorLengthTooSmall);
+        }
+
+        req.status_addr = status_desc
+            .addr()
+            .translate_gva(access_platform, status_desc.len() as usize);
+
+        Ok(req)
+    }
+
+    pub fn execute<T: Seek + Read + Write>(
+        &self,
+        disk: &mut T,
+        disk_nsectors: u64,
+        mem: &GuestMemoryMmap,
+        disk_id: &[u8],
+    ) -> result::Result<u32, ExecuteError> {
+        disk.seek(SeekFrom::Start(self.sector << SECTOR_SHIFT))
+            .map_err(ExecuteError::Seek)?;
+        let mut len = 0;
+        for (data_addr, data_len) in &self.data_descriptors {
+            let mut top: u64 = u64::from(*data_len) / SECTOR_SIZE;
+            if u64::from(*data_len) % SECTOR_SIZE != 0 {
+                top += 1;
+            }
+            top = top
+                .checked_add(self.sector)
+                .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
+            if top > disk_nsectors {
+                return Err(ExecuteError::BadRequest(Error::InvalidOffset));
+            }
+
+            match self.request_type {
+                RequestType::In => {
+                    mem.read_exact_from(*data_addr, disk, *data_len as usize)
+                        .map_err(ExecuteError::Read)?;
+                    len += data_len;
+                }
+                RequestType::Out => {
+                    mem.write_all_to(*data_addr, disk, *data_len as usize)
+                        .map_err(ExecuteError::Write)?;
+                    if !self.writeback {
+                        disk.flush().map_err(ExecuteError::Flush)?;
+                    }
+                }
+                RequestType::Flush => disk.flush().map_err(ExecuteError::Flush)?,
+                RequestType::GetDeviceId => {
+                    if (*data_len as usize) < disk_id.len() {
+                        return Err(ExecuteError::BadRequest(Error::InvalidOffset));
+                    }
+                    mem.write_slice(disk_id, *data_addr)
+                        .map_err(ExecuteError::Write)?;
+                }
+                RequestType::Unsupported(t) => return Err(ExecuteError::Unsupported(t)),
+            };
+        }
+        Ok(len)
+    }
+
+    pub fn execute_async(
+        &mut self,
+        mem: &GuestMemoryMmap,
+        disk_nsectors: u64,
+        disk_image: &mut dyn AsyncIo,
+        disk_id: &[u8],
+        user_data: u64,
+    ) -> result::Result<bool, ExecuteError> {
+        let sector = self.sector;
+        let request_type = self.request_type;
+        let offset = (sector << SECTOR_SHIFT) as libc::off_t;
+
+        let mut iovecs: SmallVec<[libc::iovec; 1]> =
+            SmallVec::with_capacity(self.data_descriptors.len());
+        for (data_addr, data_len) in &self.data_descriptors {
+            if *data_len == 0 {
+                continue;
+            }
+            let mut top: u64 = u64::from(*data_len) / SECTOR_SIZE;
+            if u64::from(*data_len) % SECTOR_SIZE != 0 {
+                top += 1;
+            }
+            top = top
+                .checked_add(sector)
+                .ok_or(ExecuteError::BadRequest(Error::InvalidOffset))?;
+            if top > disk_nsectors {
+                return Err(ExecuteError::BadRequest(Error::InvalidOffset));
+            }
+
+            let origin_ptr = mem
+                .get_slice(*data_addr, *data_len as usize)
+                .map_err(ExecuteError::GetHostAddress)?
+                .as_ptr();
+
+            // Verify the buffer alignment.
+            // In case it's not properly aligned, an intermediate buffer is
+            // created with the correct alignment, and a copy from/to the
+            // origin buffer is performed, depending on the type of operation.
+            let iov_base = if (origin_ptr as u64) % SECTOR_SIZE != 0 {
+                let layout =
+                    Layout::from_size_align(*data_len as usize, SECTOR_SIZE as usize).unwrap();
+                // SAFETY: layout has non-zero size
+                let aligned_ptr = unsafe { alloc_zeroed(layout) };
+                if aligned_ptr.is_null() {
+                    return Err(ExecuteError::TemporaryBufferAllocation(
+                        io::Error::last_os_error(),
+                    ));
+                }
+
+                // We need to perform the copy beforehand in case we're writing
+                // data out.
+                if request_type == RequestType::Out {
+                    // SAFETY: destination buffer has been allocated with
+                    // the proper size.
+                    unsafe {
+                        std::ptr::copy(origin_ptr as *const u8, aligned_ptr, *data_len as usize)
+                    };
+                }
+
+                // Store both origin and aligned pointers for complete_async()
+                // to process them.
+                self.aligned_operations.push(AlignedOperation {
+                    origin_ptr: origin_ptr as u64,
+                    aligned_ptr: aligned_ptr as u64,
+                    size: *data_len as usize,
+                    layout,
+                });
+
+                aligned_ptr as *mut libc::c_void
+            } else {
+                origin_ptr as *mut libc::c_void
+            };
+
+            let iovec = libc::iovec {
+                iov_base,
+                iov_len: *data_len as libc::size_t,
+            };
+            iovecs.push(iovec);
+        }
+
+        // Queue operations expected to be submitted.
+        match request_type {
+            RequestType::In => {
+                for (data_addr, data_len) in &self.data_descriptors {
+                    mem.get_slice(*data_addr, *data_len as usize)
+                        .map_err(ExecuteError::GetHostAddress)?
+                        .bitmap()
+                        .mark_dirty(0, *data_len as usize);
+                }
+                disk_image
+                    .read_vectored(offset, &iovecs, user_data)
+                    .map_err(ExecuteError::AsyncRead)?;
+            }
+            RequestType::Out => {
+                disk_image
+                    .write_vectored(offset, &iovecs, user_data)
+                    .map_err(ExecuteError::AsyncWrite)?;
+            }
+            RequestType::Flush => {
+                disk_image
+                    .fsync(Some(user_data))
+                    .map_err(ExecuteError::AsyncFlush)?;
+            }
+            RequestType::GetDeviceId => {
+                let (data_addr, data_len) = if self.data_descriptors.len() == 1 {
+                    (self.data_descriptors[0].0, self.data_descriptors[0].1)
+                } else {
+                    return Err(ExecuteError::BadRequest(Error::TooManyDescriptors));
+                };
+                if (data_len as usize) < disk_id.len() {
+                    return Err(ExecuteError::BadRequest(Error::InvalidOffset));
+                }
+                mem.write_slice(disk_id, data_addr)
+                    .map_err(ExecuteError::Write)?;
+                return Ok(false);
+            }
+            RequestType::Unsupported(t) => return Err(ExecuteError::Unsupported(t)),
+        }
+
+        Ok(true)
+    }
+
+    pub fn complete_async(&mut self) -> result::Result<(), Error> {
+        for aligned_operation in self.aligned_operations.drain(..) {
+            // We need to perform the copy after the data has been read inside
+            // the aligned buffer in case we're reading data in.
+            if self.request_type == RequestType::In {
+                // SAFETY: origin buffer has been allocated with the
+                // proper size.
+                unsafe {
+                    std::ptr::copy(
+                        aligned_operation.aligned_ptr as *const u8,
+                        aligned_operation.origin_ptr as *mut u8,
+                        aligned_operation.size,
+                    )
+                };
+            }
+
+            // Free the temporary aligned buffer.
+            // SAFETY: aligned_ptr was allocated by alloc_zeroed with the same
+            // layout
+            unsafe {
+                dealloc(
+                    aligned_operation.aligned_ptr as *mut u8,
+                    aligned_operation.layout,
+                )
+            };
+        }
+
+        Ok(())
+    }
+
+    pub fn set_writeback(&mut self, writeback: bool) {
+        self.writeback = writeback
+    }
+}
+
+#[derive(Copy, Clone, Debug, Default, Versionize)]
+#[repr(C, packed)]
+pub struct VirtioBlockConfig {
+    pub capacity: u64,
+    pub size_max: u32,
+    pub seg_max: u32,
+    pub geometry: VirtioBlockGeometry,
+    pub blk_size: u32,
+    pub physical_block_exp: u8,
+    pub alignment_offset: u8,
+    pub min_io_size: u16,
+    pub opt_io_size: u32,
+    pub writeback: u8,
+    pub unused: u8,
+    pub num_queues: u16,
+    pub max_discard_sectors: u32,
+    pub max_discard_seg: u32,
+    pub discard_sector_alignment: u32,
+    pub max_write_zeroes_sectors: u32,
+    pub max_write_zeroes_seg: u32,
+    pub write_zeroes_may_unmap: u8,
+    pub unused1: [u8; 3],
+}
+#[derive(Copy, Clone, Debug, Default, Versionize)]
+#[repr(C, packed)]
+pub struct VirtioBlockGeometry {
+    pub cylinders: u16,
+    pub heads: u8,
+    pub sectors: u8,
+}
+
+// SAFETY: data structure only contain a series of integers
+unsafe impl ByteValued for VirtioBlockConfig {}
+// SAFETY: data structure only contain a series of integers
+unsafe impl ByteValued for VirtioBlockGeometry {}
+
+/// Check if io_uring for block device can be used on the current system, as
+/// it correctly supports the expected io_uring features.
+pub fn block_io_uring_is_supported() -> bool {
+    let error_msg = "io_uring not supported:";
+
+    // Check we can create an io_uring instance, which effectively verifies
+    // that io_uring_setup() syscall is supported.
+    let io_uring = match IoUring::new(1) {
+        Ok(io_uring) => io_uring,
+        Err(e) => {
+            info!("{} failed to create io_uring instance: {}", error_msg, e);
+            return false;
+        }
+    };
+
+    let submitter = io_uring.submitter();
+
+    let mut probe = Probe::new();
+
+    // Check we can register a probe to validate supported operations.
+    match submitter.register_probe(&mut probe) {
+        Ok(_) => {}
+        Err(e) => {
+            info!("{} failed to register a probe: {}", error_msg, e);
+            return false;
+        }
+    }
+
+    // Check IORING_OP_FSYNC is supported
+    if !probe.is_supported(opcode::Fsync::CODE) {
+        info!("{} IORING_OP_FSYNC operation not supported", error_msg);
+        return false;
+    }
+
+    // Check IORING_OP_READV is supported
+    if !probe.is_supported(opcode::Readv::CODE) {
+        info!("{} IORING_OP_READV operation not supported", error_msg);
+        return false;
+    }
+
+    // Check IORING_OP_WRITEV is supported
+    if !probe.is_supported(opcode::Writev::CODE) {
+        info!("{} IORING_OP_WRITEV operation not supported", error_msg);
+        return false;
+    }
+
+    true
+}
+
+pub trait AsyncAdaptor<F>
+where
+    F: Read + Write + Seek,
+{
+    fn read_vectored_sync(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+        eventfd: &EventFd,
+        completion_list: &mut VecDeque<(u64, i32)>,
+    ) -> AsyncIoResult<()> {
+        // Convert libc::iovec into IoSliceMut
+        let mut slices: SmallVec<[IoSliceMut; 1]> = SmallVec::with_capacity(iovecs.len());
+        for iovec in iovecs.iter() {
+            // SAFETY: on Linux IoSliceMut wraps around libc::iovec
+            slices.push(IoSliceMut::new(unsafe { std::mem::transmute(*iovec) }));
+        }
+
+        let result = {
+            let mut file = self.file();
+
+            // Move the cursor to the right offset
+            file.seek(SeekFrom::Start(offset as u64))
+                .map_err(AsyncIoError::ReadVectored)?;
+
+            // Read vectored
+            file.read_vectored(slices.as_mut_slice())
+                .map_err(AsyncIoError::ReadVectored)?
+        };
+
+        completion_list.push_back((user_data, result as i32));
+        eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn write_vectored_sync(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+        eventfd: &EventFd,
+        completion_list: &mut VecDeque<(u64, i32)>,
+    ) -> AsyncIoResult<()> {
+        // Convert libc::iovec into IoSlice
+        let mut slices: SmallVec<[IoSlice; 1]> = SmallVec::with_capacity(iovecs.len());
+        for iovec in iovecs.iter() {
+            // SAFETY: on Linux IoSlice wraps around libc::iovec
+            slices.push(IoSlice::new(unsafe { std::mem::transmute(*iovec) }));
+        }
+
+        let result = {
+            let mut file = self.file();
+
+            // Move the cursor to the right offset
+            file.seek(SeekFrom::Start(offset as u64))
+                .map_err(AsyncIoError::WriteVectored)?;
+
+            // Write vectored
+            file.write_vectored(slices.as_slice())
+                .map_err(AsyncIoError::WriteVectored)?
+        };
+
+        completion_list.push_back((user_data, result as i32));
+        eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn fsync_sync(
+        &mut self,
+        user_data: Option<u64>,
+        eventfd: &EventFd,
+        completion_list: &mut VecDeque<(u64, i32)>,
+    ) -> AsyncIoResult<()> {
+        let result: i32 = {
+            let mut file = self.file();
+
+            // Flush
+            file.flush().map_err(AsyncIoError::Fsync)?;
+
+            0
+        };
+
+        if let Some(user_data) = user_data {
+            completion_list.push_back((user_data, result));
+            eventfd.write(1).unwrap();
+        }
+
+        Ok(())
+    }
+
+    fn file(&mut self) -> MutexGuard<F>;
+}
+
+pub enum ImageType {
+    FixedVhd,
+    Qcow2,
+    Raw,
+    Vhdx,
+}
+
+const QCOW_MAGIC: u32 = 0x5146_49fb;
+const VHDX_SIGN: u64 = 0x656C_6966_7864_6876;
+
+/// Determine image type through file parsing.
+pub fn detect_image_type(f: &mut File) -> std::io::Result<ImageType> {
+    // We must create a buffer aligned on 512 bytes with a size being a
+    // multiple of 512 bytes as the file might be opened with O_DIRECT flag.
+    #[repr(align(512))]
+    struct Sector {
+        data: [u8; 512],
+    }
+    let mut s = Sector { data: [0; 512] };
+
+    f.read_exact(&mut s.data)?;
+
+    // Check 4 first bytes to get the header value and determine the image type
+    let image_type = if u32::from_be_bytes(s.data[0..4].try_into().unwrap()) == QCOW_MAGIC {
+        ImageType::Qcow2
+    } else if vhd::is_fixed_vhd(f)? {
+        ImageType::FixedVhd
+    } else if u64::from_le_bytes(s.data[0..8].try_into().unwrap()) == VHDX_SIGN {
+        ImageType::Vhdx
+    } else {
+        ImageType::Raw
+    };
+
+    Ok(image_type)
+}

block_util/src/qcow_sync.rs

@@ -0,0 +1,104 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+use crate::async_io::{AsyncIo, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult};
+use crate::AsyncAdaptor;
+use qcow::{QcowFile, RawFile, Result as QcowResult};
+use std::collections::VecDeque;
+use std::fs::File;
+use std::io::{Seek, SeekFrom};
+use std::sync::{Arc, Mutex, MutexGuard};
+use vmm_sys_util::eventfd::EventFd;
+
+pub struct QcowDiskSync {
+    qcow_file: Arc<Mutex<QcowFile>>,
+}
+
+impl QcowDiskSync {
+    pub fn new(file: File, direct_io: bool) -> QcowResult<Self> {
+        Ok(QcowDiskSync {
+            qcow_file: Arc::new(Mutex::new(QcowFile::from(RawFile::new(file, direct_io))?)),
+        })
+    }
+}
+
+impl DiskFile for QcowDiskSync {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        let mut file = self.qcow_file.lock().unwrap();
+
+        file.seek(SeekFrom::End(0)).map_err(DiskFileError::Size)
+    }
+
+    fn new_async_io(&self, _ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(Box::new(QcowSync::new(self.qcow_file.clone())) as Box<dyn AsyncIo>)
+    }
+}
+
+pub struct QcowSync {
+    qcow_file: Arc<Mutex<QcowFile>>,
+    eventfd: EventFd,
+    completion_list: VecDeque<(u64, i32)>,
+}
+
+impl QcowSync {
+    pub fn new(qcow_file: Arc<Mutex<QcowFile>>) -> Self {
+        QcowSync {
+            qcow_file,
+            eventfd: EventFd::new(libc::EFD_NONBLOCK)
+                .expect("Failed creating EventFd for QcowSync"),
+            completion_list: VecDeque::new(),
+        }
+    }
+}
+
+impl AsyncAdaptor<QcowFile> for Arc<Mutex<QcowFile>> {
+    fn file(&mut self) -> MutexGuard<QcowFile> {
+        self.lock().unwrap()
+    }
+}
+
+impl AsyncIo for QcowSync {
+    fn notifier(&self) -> &EventFd {
+        &self.eventfd
+    }
+
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        self.qcow_file.read_vectored_sync(
+            offset,
+            iovecs,
+            user_data,
+            &self.eventfd,
+            &mut self.completion_list,
+        )
+    }
+
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        self.qcow_file.write_vectored_sync(
+            offset,
+            iovecs,
+            user_data,
+            &self.eventfd,
+            &mut self.completion_list,
+        )
+    }
+
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
+        self.qcow_file
+            .fsync_sync(user_data, &self.eventfd, &mut self.completion_list)
+    }
+
+    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
+        self.completion_list.pop_front()
+    }
+}

block_util/src/raw_async.rs

@@ -0,0 +1,164 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+use crate::async_io::{
+    AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult, DiskTopology,
+};
+use io_uring::{opcode, squeue, types, IoUring};
+use std::fs::File;
+use std::io::{Seek, SeekFrom};
+use std::os::unix::io::{AsRawFd, RawFd};
+use vmm_sys_util::eventfd::EventFd;
+
+pub struct RawFileDisk {
+    file: File,
+}
+
+impl RawFileDisk {
+    pub fn new(file: File) -> Self {
+        RawFileDisk { file }
+    }
+}
+
+impl DiskFile for RawFileDisk {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        self.file
+            .seek(SeekFrom::End(0))
+            .map_err(DiskFileError::Size)
+    }
+
+    fn new_async_io(&self, ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(Box::new(
+            RawFileAsync::new(self.file.as_raw_fd(), ring_depth)
+                .map_err(DiskFileError::NewAsyncIo)?,
+        ) as Box<dyn AsyncIo>)
+    }
+
+    fn topology(&mut self) -> DiskTopology {
+        if let Ok(topology) = DiskTopology::probe(&mut self.file) {
+            topology
+        } else {
+            warn!("Unable to get device topology. Using default topology");
+            DiskTopology::default()
+        }
+    }
+}
+
+pub struct RawFileAsync {
+    fd: RawFd,
+    io_uring: IoUring,
+    eventfd: EventFd,
+}
+
+impl RawFileAsync {
+    pub fn new(fd: RawFd, ring_depth: u32) -> std::io::Result<Self> {
+        let io_uring = IoUring::new(ring_depth)?;
+        let eventfd = EventFd::new(libc::EFD_NONBLOCK)?;
+
+        // Register the io_uring eventfd that will notify when something in
+        // the completion queue is ready.
+        io_uring.submitter().register_eventfd(eventfd.as_raw_fd())?;
+
+        Ok(RawFileAsync {
+            fd,
+            io_uring,
+            eventfd,
+        })
+    }
+}
+
+impl AsyncIo for RawFileAsync {
+    fn notifier(&self) -> &EventFd {
+        &self.eventfd
+    }
+
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        let (submitter, mut sq, _) = self.io_uring.split();
+
+        // SAFETY: we know the file descriptor is valid and we
+        // relied on vm-memory to provide the buffer address.
+        let _ = unsafe {
+            sq.push(
+                &opcode::Readv::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
+                    .offset(offset)
+                    .build()
+                    .flags(squeue::Flags::ASYNC)
+                    .user_data(user_data),
+            )
+        };
+
+        // Update the submission queue and submit new operations to the
+        // io_uring instance.
+        sq.sync();
+        submitter.submit().map_err(AsyncIoError::ReadVectored)?;
+
+        Ok(())
+    }
+
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        let (submitter, mut sq, _) = self.io_uring.split();
+
+        // SAFETY: we know the file descriptor is valid and we
+        // relied on vm-memory to provide the buffer address.
+        let _ = unsafe {
+            sq.push(
+                &opcode::Writev::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
+                    .offset(offset)
+                    .build()
+                    .flags(squeue::Flags::ASYNC)
+                    .user_data(user_data),
+            )
+        };
+
+        // Update the submission queue and submit new operations to the
+        // io_uring instance.
+        sq.sync();
+        submitter.submit().map_err(AsyncIoError::WriteVectored)?;
+
+        Ok(())
+    }
+
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
+        if let Some(user_data) = user_data {
+            let (submitter, mut sq, _) = self.io_uring.split();
+
+            // SAFETY: we know the file descriptor is valid.
+            let _ = unsafe {
+                sq.push(
+                    &opcode::Fsync::new(types::Fd(self.fd))
+                        .build()
+                        .flags(squeue::Flags::ASYNC)
+                        .user_data(user_data),
+                )
+            };
+
+            // Update the submission queue and submit new operations to the
+            // io_uring instance.
+            sq.sync();
+            submitter.submit().map_err(AsyncIoError::Fsync)?;
+        } else {
+            // SAFETY: FFI call with a valid fd
+            unsafe { libc::fsync(self.fd) };
+        }
+
+        Ok(())
+    }
+
+    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
+        self.io_uring
+            .completion()
+            .next()
+            .map(|entry| (entry.user_data(), entry.result()))
+    }
+}

block_util/src/raw_sync.rs

@@ -0,0 +1,134 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+use crate::async_io::{
+    AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult, DiskTopology,
+};
+use std::collections::VecDeque;
+use std::fs::File;
+use std::io::{Seek, SeekFrom};
+use std::os::unix::io::{AsRawFd, RawFd};
+use vmm_sys_util::eventfd::EventFd;
+
+pub struct RawFileDiskSync {
+    file: File,
+}
+
+impl RawFileDiskSync {
+    pub fn new(file: File) -> Self {
+        RawFileDiskSync { file }
+    }
+}
+
+impl DiskFile for RawFileDiskSync {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        self.file
+            .seek(SeekFrom::End(0))
+            .map_err(DiskFileError::Size)
+    }
+
+    fn new_async_io(&self, _ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(Box::new(RawFileSync::new(self.file.as_raw_fd())) as Box<dyn AsyncIo>)
+    }
+
+    fn topology(&mut self) -> DiskTopology {
+        if let Ok(topology) = DiskTopology::probe(&mut self.file) {
+            topology
+        } else {
+            warn!("Unable to get device topology. Using default topology");
+            DiskTopology::default()
+        }
+    }
+}
+
+pub struct RawFileSync {
+    fd: RawFd,
+    eventfd: EventFd,
+    completion_list: VecDeque<(u64, i32)>,
+}
+
+impl RawFileSync {
+    pub fn new(fd: RawFd) -> Self {
+        RawFileSync {
+            fd,
+            eventfd: EventFd::new(libc::EFD_NONBLOCK).expect("Failed creating EventFd for RawFile"),
+            completion_list: VecDeque::new(),
+        }
+    }
+}
+
+impl AsyncIo for RawFileSync {
+    fn notifier(&self) -> &EventFd {
+        &self.eventfd
+    }
+
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        // SAFETY: FFI call with valid arguments
+        let result = unsafe {
+            libc::preadv(
+                self.fd as libc::c_int,
+                iovecs.as_ptr() as *const libc::iovec,
+                iovecs.len() as libc::c_int,
+                offset,
+            )
+        };
+        if result < 0 {
+            return Err(AsyncIoError::ReadVectored(std::io::Error::last_os_error()));
+        }
+
+        self.completion_list.push_back((user_data, result as i32));
+        self.eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        // SAFETY: FFI call with valid arguments
+        let result = unsafe {
+            libc::pwritev(
+                self.fd as libc::c_int,
+                iovecs.as_ptr() as *const libc::iovec,
+                iovecs.len() as libc::c_int,
+                offset,
+            )
+        };
+        if result < 0 {
+            return Err(AsyncIoError::WriteVectored(std::io::Error::last_os_error()));
+        }
+
+        self.completion_list.push_back((user_data, result as i32));
+        self.eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
+        // SAFETY: FFI call
+        let result = unsafe { libc::fsync(self.fd as libc::c_int) };
+        if result < 0 {
+            return Err(AsyncIoError::Fsync(std::io::Error::last_os_error()));
+        }
+
+        if let Some(user_data) = user_data {
+            self.completion_list.push_back((user_data, result));
+            self.eventfd.write(1).unwrap();
+        }
+
+        Ok(())
+    }
+
+    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
+        self.completion_list.pop_front()
+    }
+}

block_util/src/vhd.rs

@@ -2,10 +2,9 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
+use std::convert::TryInto;
 use std::fs::File;
-use std::io::{Seek, SeekFrom};
-
-use crate::{DiskTopology, read_aligned_block_size};
+use std::io::{Read, Seek, SeekFrom};
 
 #[derive(Clone, Copy)]
 pub struct VhdFooter {
@@ -28,33 +27,37 @@ pub struct VhdFooter {
 
 impl VhdFooter {
     pub fn new(file: &mut File) -> std::io::Result<VhdFooter> {
-        let blocksize = DiskTopology::probe(file)?.logical_block_size as usize;
+        // We must create a buffer aligned on 512 bytes with a size being a
+        // multiple of 512 bytes as the file might be opened with O_DIRECT flag.
+        #[repr(align(512))]
+        struct Sector {
+            data: [u8; 512],
+        }
+        let mut s = Sector { data: [0; 512] };
 
-        // Place the cursor in the last block of the file
-        file.seek(SeekFrom::End(0 - (blocksize as i64)))?;
-        // Read in the last block
-        let data = read_aligned_block_size(file)?;
+        // Place the cursor 512 bytes before the end of the file, as this is
+        // where the footer starts.
+        file.seek(SeekFrom::End(-512))?;
 
-        // We only care about the last sector
-        let offset = blocksize - 512;
-        let sector = &data[offset..];
+        // Fill in the VhdFooter structure
+        file.read_exact(&mut s.data)?;
 
         Ok(VhdFooter {
-            cookie: u64::from_be_bytes(sector[0..8].try_into().unwrap()),
-            features: u32::from_be_bytes(sector[8..12].try_into().unwrap()),
-            file_format_version: u32::from_be_bytes(sector[12..16].try_into().unwrap()),
-            data_offset: u64::from_be_bytes(sector[16..24].try_into().unwrap()),
-            time_stamp: u32::from_be_bytes(sector[24..28].try_into().unwrap()),
-            creator_application: u32::from_be_bytes(sector[28..32].try_into().unwrap()),
-            creator_version: u32::from_be_bytes(sector[32..36].try_into().unwrap()),
-            creator_host_os: u32::from_be_bytes(sector[36..40].try_into().unwrap()),
-            original_size: u64::from_be_bytes(sector[40..48].try_into().unwrap()),
-            current_size: u64::from_be_bytes(sector[48..56].try_into().unwrap()),
-            disk_geometry: u32::from_be_bytes(sector[56..60].try_into().unwrap()),
-            disk_type: u32::from_be_bytes(sector[60..64].try_into().unwrap()),
-            checksum: u32::from_be_bytes(sector[64..68].try_into().unwrap()),
-            unique_id: u128::from_be_bytes(sector[68..84].try_into().unwrap()),
-            saved_state: u8::from_be_bytes(sector[84..85].try_into().unwrap()),
+            cookie: u64::from_be_bytes(s.data[0..8].try_into().unwrap()),
+            features: u32::from_be_bytes(s.data[8..12].try_into().unwrap()),
+            file_format_version: u32::from_be_bytes(s.data[12..16].try_into().unwrap()),
+            data_offset: u64::from_be_bytes(s.data[16..24].try_into().unwrap()),
+            time_stamp: u32::from_be_bytes(s.data[24..28].try_into().unwrap()),
+            creator_application: u32::from_be_bytes(s.data[28..32].try_into().unwrap()),
+            creator_version: u32::from_be_bytes(s.data[32..36].try_into().unwrap()),
+            creator_host_os: u32::from_be_bytes(s.data[36..40].try_into().unwrap()),
+            original_size: u64::from_be_bytes(s.data[40..48].try_into().unwrap()),
+            current_size: u64::from_be_bytes(s.data[48..56].try_into().unwrap()),
+            disk_geometry: u32::from_be_bytes(s.data[56..60].try_into().unwrap()),
+            disk_type: u32::from_be_bytes(s.data[60..64].try_into().unwrap()),
+            checksum: u32::from_be_bytes(s.data[64..68].try_into().unwrap()),
+            unique_id: u128::from_be_bytes(s.data[68..84].try_into().unwrap()),
+            saved_state: u8::from_be_bytes(s.data[84..85].try_into().unwrap()),
         })
     }
 
@@ -117,14 +120,12 @@ pub fn is_fixed_vhd(f: &mut File) -> std::io::Result<bool> {
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
+    use super::{is_fixed_vhd, VhdFooter};
     use std::fs::File;
     use std::io::{Seek, SeekFrom, Write};
-
     use vmm_sys_util::tempfile::TempFile;
 
-    use super::{VhdFooter, is_fixed_vhd};
-
     fn valid_fixed_vhd_footer() -> Vec<u8> {
         vec![
             0x63, 0x6f, 0x6e, 0x65, 0x63, 0x74, 0x69, 0x78, // cookie

block_util/src/vhdx_sync.rs

@@ -0,0 +1,103 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+
+use crate::async_io::{AsyncIo, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult};
+use crate::AsyncAdaptor;
+use std::collections::VecDeque;
+use std::fs::File;
+use std::sync::{Arc, Mutex, MutexGuard};
+use vhdx::vhdx::{Result as VhdxResult, Vhdx};
+use vmm_sys_util::eventfd::EventFd;
+
+pub struct VhdxDiskSync {
+    vhdx_file: Arc<Mutex<Vhdx>>,
+}
+
+impl VhdxDiskSync {
+    pub fn new(f: File) -> VhdxResult<Self> {
+        Ok(VhdxDiskSync {
+            vhdx_file: Arc::new(Mutex::new(Vhdx::new(f)?)),
+        })
+    }
+}
+
+impl DiskFile for VhdxDiskSync {
+    fn size(&mut self) -> DiskFileResult<u64> {
+        Ok(self.vhdx_file.lock().unwrap().virtual_disk_size())
+    }
+
+    fn new_async_io(&self, _ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
+        Ok(
+            Box::new(VhdxSync::new(self.vhdx_file.clone()).map_err(DiskFileError::NewAsyncIo)?)
+                as Box<dyn AsyncIo>,
+        )
+    }
+}
+
+pub struct VhdxSync {
+    vhdx_file: Arc<Mutex<Vhdx>>,
+    eventfd: EventFd,
+    completion_list: VecDeque<(u64, i32)>,
+}
+
+impl VhdxSync {
+    pub fn new(vhdx_file: Arc<Mutex<Vhdx>>) -> std::io::Result<Self> {
+        Ok(VhdxSync {
+            vhdx_file,
+            eventfd: EventFd::new(libc::EFD_NONBLOCK)?,
+            completion_list: VecDeque::new(),
+        })
+    }
+}
+
+impl AsyncAdaptor<Vhdx> for Arc<Mutex<Vhdx>> {
+    fn file(&mut self) -> MutexGuard<Vhdx> {
+        self.lock().unwrap()
+    }
+}
+
+impl AsyncIo for VhdxSync {
+    fn notifier(&self) -> &EventFd {
+        &self.eventfd
+    }
+
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        self.vhdx_file.read_vectored_sync(
+            offset,
+            iovecs,
+            user_data,
+            &self.eventfd,
+            &mut self.completion_list,
+        )
+    }
+
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: &[libc::iovec],
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        self.vhdx_file.write_vectored_sync(
+            offset,
+            iovecs,
+            user_data,
+            &self.eventfd,
+            &mut self.completion_list,
+        )
+    }
+
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()> {
+        self.vhdx_file
+            .fsync_sync(user_data, &self.eventfd, &mut self.completion_list)
+    }
+
+    fn next_completed_request(&mut self) -> Option<(u64, i32)> {
+        self.completion_list.pop_front()
+    }
+}

build.rs

@@ -0,0 +1,24 @@
+// Copyright © 2020 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+use std::process::Command;
+
+fn main() {
+    let mut version = "v".to_owned() + env!("CARGO_PKG_VERSION");
+
+    if let Ok(git_out) = Command::new("git").args(["describe", "--dirty"]).output() {
+        if git_out.status.success() {
+            if let Ok(git_out_str) = String::from_utf8(git_out.stdout) {
+                version = git_out_str;
+            }
+        }
+    }
+
+    // This println!() has a special behavior, as it will set the environment
+    // variable BUILT_VERSION, so that it can be reused from the binary.
+    // Particularly, this is used from src/main.rs to display the exact
+    // version.
+    println!("cargo:rustc-env=BUILT_VERSION={version}");
+}

cloud-hypervisor/Cargo.toml

@@ -1,64 +0,0 @@
-[package]
-authors = ["The Cloud Hypervisor Authors"]
-build = "build.rs"
-default-run = "cloud-hypervisor"
-description = "Open source Virtual Machine Monitor (VMM) that runs on top of KVM & MSHV"
-edition = "2024"
-homepage = "https://github.com/cloud-hypervisor/cloud-hypervisor"
-license = "Apache-2.0 AND BSD-3-Clause"
-name = "cloud-hypervisor"
-rust-version.workspace = true
-version = "52.0.0"
-
-[dependencies]
-anyhow = { workspace = true }
-api_client = { path = "../api_client" }
-clap = { workspace = true, features = ["string"] }
-dhat = { workspace = true, optional = true }
-env_logger = { workspace = true }
-epoll = { workspace = true }
-event_monitor = { path = "../event_monitor" }
-hypervisor = { path = "../hypervisor" }
-jiff = { workspace = true }
-libc = { workspace = true }
-log = { workspace = true, features = ["std"] }
-option_parser = { path = "../option_parser" }
-seccompiler = { workspace = true }
-serde_json = { workspace = true }
-signal-hook = { workspace = true }
-thiserror = { workspace = true }
-tpm = { path = "../tpm" }
-tracer = { path = "../tracer" }
-vm-memory = { workspace = true }
-vmm = { path = "../vmm" }
-vmm-sys-util = { workspace = true }
-zbus = { version = "5.15.0", optional = true }
-
-[dev-dependencies]
-block = { path = "../block" }
-dirs = { workspace = true }
-net_util = { path = "../net_util" }
-serde_json = { workspace = true }
-test_infra = { path = "../test_infra" }
-wait-timeout = { workspace = true }
-
-# Please adjust `vmm::feature_list()` accordingly when changing the
-# feature list below
-[features]
-dbus_api = ["vmm/dbus_api", "zbus"]
-default = ["io_uring", "kvm"]
-dhat-heap = ["dhat", "vmm/dhat-heap"]       # For heap profiling
-fw_cfg = ["vmm/fw_cfg"]
-guest_debug = ["vmm/guest_debug"]
-igvm = ["mshv", "vmm/igvm"]
-io_uring = ["vmm/io_uring"]
-ivshmem = ["vmm/ivshmem"]
-kvm = ["vmm/kvm"]
-mshv = ["vmm/mshv"]
-pvmemcontrol = ["vmm/pvmemcontrol"]
-sev_snp = ["igvm", "mshv", "vmm/sev_snp"]
-tdx = ["vmm/tdx"]
-tracing = ["tracer/tracing", "vmm/tracing"]
-
-[lints]
-workspace = true

cloud-hypervisor/build.rs

@@ -1,32 +0,0 @@
-// Copyright © 2020 Intel Corporation
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-
-use std::env;
-use std::process::Command;
-
-fn main() {
-    let mut version = "v".to_owned() + env!("CARGO_PKG_VERSION");
-
-    if let Ok(git_out) = Command::new("git").args(["describe", "--dirty"]).output()
-        && git_out.status.success()
-        && let Ok(git_out_str) = String::from_utf8(git_out.stdout)
-    {
-        version = git_out_str;
-        // Pop the trailing newline.
-        version.pop();
-    }
-
-    // Append CH_EXTRA_VERSION to version if it is set.
-    if let Ok(extra_version) = env::var("CH_EXTRA_VERSION") {
-        println!("cargo:rerun-if-env-changed=CH_EXTRA_VERSION");
-        version.push_str(&format!("-{extra_version}"));
-    }
-
-    // This println!() has a special behavior, as it will set the environment
-    // variable BUILD_VERSION, so that it can be reused from the binary.
-    // Particularly, this is used from src/main.rs to display the exact
-    // version.
-    println!("cargo:rustc-env=BUILD_VERSION={version}");
-}

cloud-hypervisor/src/lib.rs

@@ -1,46 +0,0 @@
-// Copyright © 2025 Cyberus Technology GmbH
-//
-// SPDX-License-Identifier: Apache-2.0
-
-use std::error::Error;
-
-use log::error;
-
-/// Prints a chain of errors to the user in a consistent manner.
-/// The user will see a clear chain of errors, followed by debug output
-/// for opening issues.
-pub fn cli_print_error_chain<'a>(
-    top_error: &'a (dyn Error + 'static),
-    component: &str,
-    // Function optionally returning the display representation of an error.
-    display_modifier: impl Fn(
-        /* level */ usize,
-        /*indention */ usize,
-        &'a (dyn Error + 'static),
-    ) -> Option<String>,
-) {
-    // Debug info.
-    error!("Fatal error: {top_error:?}");
-
-    eprint!("Error: {component} exited with the following ");
-    if top_error.source().is_none() {
-        eprintln!("error:");
-        eprintln!("  {top_error}");
-    } else {
-        eprintln!("chain of errors:");
-        std::iter::successors(Some(top_error), |sub_error| {
-            // Dereference necessary to mitigate rustc compiler bug.
-            // See <https://github.com/rust-lang/rust/issues/141673>
-            (*sub_error).source()
-        })
-        .enumerate()
-        .for_each(|(level, error)| {
-            // Special case: handling of HTTP Server responses in ch-remote
-            if let Some(message) = display_modifier(level, 2, error) {
-                eprintln!("{message}");
-            } else {
-                eprintln!("  {level}: {error}");
-            }
-        });
-    }
-}

cloud-hypervisor/src/logger.rs

@@ -1,436 +0,0 @@
-// Copyright © 2026 Cloud Hypervisor Contributors
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-
-use std::io::Write;
-use std::str::FromStr;
-use std::sync::Mutex;
-use std::time::Instant;
-
-use thiserror::Error;
-
-#[derive(Debug, Error)]
-pub enum Error {
-    #[error("Unterminated '{{' in format string")]
-    UnterminatedBrace,
-    #[error("Unmatched '}}' in format string")]
-    UnmatchedBrace,
-    #[error("Unknown format token '{{{0}}}'")]
-    UnknownToken(String),
-}
-
-enum Token {
-    Literal(String),
-    BootTime,
-    /// Wallclock using RFC 3339 formatting.
-    WallClock,
-    Pid,
-    Tid,
-    Thread,
-    Level,
-    Location,
-    Msg,
-}
-
-impl FromStr for Token {
-    type Err = Error;
-
-    fn from_str(s: &str) -> Result<Self, Self::Err> {
-        match s {
-            "boottime" => Ok(Self::BootTime),
-            "wallclock" => Ok(Self::WallClock),
-            "pid" => Ok(Self::Pid),
-            "tid" => Ok(Self::Tid),
-            "thread" => Ok(Self::Thread),
-            "level" => Ok(Self::Level),
-            "location" => Ok(Self::Location),
-            "msg" => Ok(Self::Msg),
-            _ => Err(Error::UnknownToken(s.to_string())),
-        }
-    }
-}
-
-fn parse_format(fmt: &str) -> Result<Vec<Token>, Error> {
-    let mut tokens = Vec::new();
-    let mut literal = String::new();
-    let mut chars = fmt.chars().peekable();
-
-    while let Some(c) = chars.next() {
-        match c {
-            '{' => {
-                if chars.peek() == Some(&'{') {
-                    chars.next();
-                    literal.push('{');
-                    continue;
-                }
-
-                if !literal.is_empty() {
-                    tokens.push(Token::Literal(std::mem::take(&mut literal)));
-                }
-
-                let mut name = String::new();
-                loop {
-                    match chars.next() {
-                        Some('}') => break,
-                        Some(ch) => name.push(ch),
-                        None => return Err(Error::UnterminatedBrace),
-                    }
-                }
-
-                tokens.push(name.parse()?);
-            }
-            '}' => {
-                if chars.peek() == Some(&'}') {
-                    chars.next();
-                    literal.push('}');
-                } else {
-                    return Err(Error::UnmatchedBrace);
-                }
-            }
-            _ => literal.push(c),
-        }
-    }
-    if !literal.is_empty() {
-        tokens.push(Token::Literal(literal));
-    }
-    Ok(tokens)
-}
-
-pub const DEFAULT_FORMAT: &str =
-    "cloud-hypervisor: {boottime}s: <{thread}> {level}:{location} -- {msg}";
-
-pub struct Logger {
-    output: Mutex<Box<dyn Write + Send>>,
-    start: Instant,
-    pid: u32,
-    tokens: Vec<Token>,
-}
-
-impl Logger {
-    pub fn new(output: Box<dyn Write + Send>, format: &str) -> Result<Self, Error> {
-        Ok(Self {
-            output: Mutex::new(output),
-            start: Instant::now(),
-            pid: std::process::id(),
-            tokens: parse_format(format)?,
-        })
-    }
-}
-
-impl log::Log for Logger {
-    fn enabled(&self, _metadata: &log::Metadata) -> bool {
-        true
-    }
-
-    fn log(&self, record: &log::Record) {
-        if !self.enabled(record.metadata()) {
-            return;
-        }
-
-        let duration_s = Instant::now().duration_since(self.start).as_secs_f32();
-        let mut out = self.output.lock().unwrap();
-        for token in &self.tokens {
-            let _ = match token {
-                Token::Literal(s) => out.write_all(s.as_bytes()),
-                // 10: 6 decimal places + sep => whole seconds in range `0..=999` properly aligned
-                Token::BootTime => write!(&mut *out, "{duration_s:>10.6?}"),
-                Token::WallClock => {
-                    write!(out, "{:.6}", jiff::Timestamp::now())
-                }
-                Token::Pid => write!(&mut *out, "{}", self.pid),
-                // SAFETY: gettid(2) always succeeds
-                Token::Tid => write!(&mut *out, "{}", unsafe { libc::gettid() }),
-                Token::Thread => write!(
-                    &mut *out,
-                    "{}",
-                    std::thread::current().name().unwrap_or("anonymous")
-                ),
-                Token::Level => write!(&mut *out, "{}", record.level()),
-                Token::Location => match (record.file(), record.line()) {
-                    (Some(file), Some(line)) => write!(&mut *out, "{file}:{line}"),
-                    _ => write!(&mut *out, "{}", record.target()),
-                },
-                Token::Msg => write!(&mut *out, "{}", record.args()),
-            };
-        }
-        let _ = out.write_all(b"\r\n");
-    }
-
-    fn flush(&self) {}
-}
-
-#[cfg(test)]
-mod tests {
-    use std::io;
-    use std::sync::Arc;
-
-    use log::Log;
-
-    use super::*;
-
-    /// A `Write` sink that appends to a shared byte buffer so tests can
-    /// inspect what the logger wrote.
-    #[derive(Clone, Default)]
-    struct SharedBuffer(Arc<Mutex<Vec<u8>>>);
-
-    impl SharedBuffer {
-        fn contents(&self) -> String {
-            String::from_utf8(self.0.lock().unwrap().clone()).unwrap()
-        }
-    }
-
-    impl Write for SharedBuffer {
-        fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
-            self.0.lock().unwrap().extend_from_slice(buf);
-            Ok(buf.len())
-        }
-
-        fn flush(&mut self) -> io::Result<()> {
-            Ok(())
-        }
-    }
-
-    fn render(tokens: &[Token]) -> String {
-        tokens
-            .iter()
-            .map(|t| match t {
-                Token::Literal(s) => format!("L({s})"),
-                Token::BootTime => "B".to_string(),
-                Token::WallClock => "W".to_string(),
-                Token::Pid => "P".to_string(),
-                Token::Tid => "I".to_string(),
-                Token::Thread => "T".to_string(),
-                Token::Level => "V".to_string(),
-                Token::Location => "O".to_string(),
-                Token::Msg => "M".to_string(),
-            })
-            .collect::<Vec<_>>()
-            .join("|")
-    }
-
-    #[test]
-    fn parse_plain_literal() {
-        let tokens = parse_format("hello world").unwrap();
-        assert_eq!(render(&tokens), "L(hello world)");
-    }
-
-    #[test]
-    fn parse_empty_string() {
-        let tokens = parse_format("").unwrap();
-        assert!(tokens.is_empty());
-    }
-
-    #[test]
-    fn parse_all_known_tokens() {
-        let tokens = parse_format(
-            "[{boottime}] {wallclock} {pid}/{tid} <{thread}> {level} {location} -- {msg}",
-        )
-        .unwrap();
-        assert_eq!(
-            render(&tokens),
-            "L([)|B|L(] )|W|L( )|P|L(/)|I|L( <)|T|L(> )|V|L( )|O|L( -- )|M"
-        );
-    }
-
-    #[test]
-    fn parse_default_format_succeeds() {
-        let tokens = parse_format(DEFAULT_FORMAT).unwrap();
-        // Default format has 5 tokens interleaved with literals.
-        assert!(tokens.iter().any(|t| matches!(t, Token::BootTime)));
-        assert!(tokens.iter().any(|t| matches!(t, Token::Thread)));
-        assert!(tokens.iter().any(|t| matches!(t, Token::Level)));
-        assert!(tokens.iter().any(|t| matches!(t, Token::Location)));
-        assert!(tokens.iter().any(|t| matches!(t, Token::Msg)));
-    }
-
-    #[test]
-    fn parse_escaped_braces() {
-        let tokens = parse_format("{{not-a-token}}").unwrap();
-        assert_eq!(render(&tokens), "L({not-a-token})");
-    }
-
-    #[test]
-    fn parse_escaped_braces_around_token() {
-        let tokens = parse_format("{{{level}}}").unwrap();
-        assert_eq!(render(&tokens), "L({)|V|L(})");
-    }
-
-    #[test]
-    fn parse_unterminated_brace_errors() {
-        match parse_format("hello {level") {
-            Err(Error::UnterminatedBrace) => {}
-            Err(other) => panic!("unexpected error: {other:?}"),
-            Ok(_) => panic!("expected error"),
-        }
-    }
-
-    #[test]
-    fn parse_unmatched_close_brace_errors() {
-        match parse_format("hello }") {
-            Err(Error::UnmatchedBrace) => {}
-            Err(other) => panic!("unexpected error: {other:?}"),
-            Ok(_) => panic!("expected error"),
-        }
-    }
-
-    #[test]
-    fn parse_unknown_token_errors() {
-        match parse_format("{nope}") {
-            Err(Error::UnknownToken(name)) => assert_eq!(name, "nope"),
-            Err(other) => panic!("unexpected error: {other:?}"),
-            Ok(_) => panic!("expected error"),
-        }
-    }
-
-    #[test]
-    fn logger_new_uses_default_format() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), DEFAULT_FORMAT).unwrap();
-        // The default format has all 5 dynamic tokens.
-        assert_eq!(
-            logger
-                .tokens
-                .iter()
-                .filter(|t| !matches!(t, Token::Literal(_)))
-                .count(),
-            5
-        );
-    }
-
-    #[test]
-    fn logger_enabled_always_true() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf), DEFAULT_FORMAT).unwrap();
-        let metadata = log::Metadata::builder()
-            .level(log::Level::Trace)
-            .target("anything")
-            .build();
-        assert!(logger.enabled(&metadata));
-    }
-
-    #[test]
-    fn logger_writes_expected_fields() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), DEFAULT_FORMAT).unwrap();
-
-        logger.log(
-            &log::Record::builder()
-                .args(format_args!("hello {}", "world"))
-                .level(log::Level::Info)
-                .target("unit_test_target")
-                .file(Some("foo.rs"))
-                .line(Some(42))
-                .build(),
-        );
-
-        let out = buf.contents();
-        assert!(out.starts_with("cloud-hypervisor: "), "got: {out}");
-        assert!(out.contains("INFO"), "got: {out}");
-        assert!(out.contains("foo.rs:42"), "got: {out}");
-        assert!(out.contains("hello world"), "got: {out}");
-        assert!(out.ends_with("\r\n"), "got: {out}");
-    }
-
-    #[test]
-    fn logger_uses_target_when_no_file() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), DEFAULT_FORMAT).unwrap();
-
-        logger.log(
-            &log::Record::builder()
-                .args(format_args!("no location"))
-                .level(log::Level::Warn)
-                .target("my_target")
-                .file(None)
-                .line(None)
-                .build(),
-        );
-
-        let out = buf.contents();
-        assert!(out.contains("my_target"), "got: {out}");
-        assert!(!out.contains("foo.rs"), "got: {out}");
-    }
-
-    #[test]
-    fn logger_wallclock_is_rfc3339() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), "{wallclock}").unwrap();
-
-        logger.log(
-            &log::Record::builder()
-                .args(format_args!(""))
-                .level(log::Level::Info)
-                .target("t")
-                .build(),
-        );
-
-        let out = buf.contents();
-        let out = out.trim();
-        assert_eq!(out.len(), 27, "got: {out}");
-        assert_eq!(&out[4..5], "-", "got: {out}");
-        assert_eq!(&out[7..8], "-", "got: {out}");
-        assert_eq!(&out[10..11], "T", "got: {out}");
-        assert_eq!(&out[13..14], ":", "got: {out}");
-        assert_eq!(&out[16..17], ":", "got: {out}");
-        assert_eq!(&out[19..20], ".", "got: {out}");
-        assert!(out.ends_with('Z'), "got: {out}");
-    }
-
-    #[test]
-    fn logger_pid_token() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), "{pid}").unwrap();
-
-        logger.log(
-            &log::Record::builder()
-                .args(format_args!(""))
-                .level(log::Level::Info)
-                .target("t")
-                .build(),
-        );
-
-        let out = buf.contents();
-        let out = out.trim();
-        assert_eq!(out, std::process::id().to_string(), "got: {out}");
-    }
-
-    #[test]
-    fn logger_tid_token() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), "{tid}").unwrap();
-
-        logger.log(
-            &log::Record::builder()
-                .args(format_args!(""))
-                .level(log::Level::Info)
-                .target("t")
-                .build(),
-        );
-
-        let out = buf.contents();
-        let out = out.trim();
-        let tid: i64 = out.parse().expect("tid should be numeric");
-        assert!(tid > 0, "got: {tid}");
-    }
-
-    #[test]
-    fn logger_appends_each_record() {
-        let buf = SharedBuffer::default();
-        let logger = Logger::new(Box::new(buf.clone()), DEFAULT_FORMAT).unwrap();
-
-        for i in 0..3 {
-            logger.log(
-                &log::Record::builder()
-                    .args(format_args!("entry-{i}"))
-                    .level(log::Level::Debug)
-                    .target("t")
-                    .build(),
-            );
-        }
-
-        let out = buf.contents();
-        assert_eq!(out.matches("entry-").count(), 3, "got: {out}");
-        assert_eq!(out.matches("\r\n").count(), 3, "got: {out}");
-    }
-}

cloud-hypervisor/src/test_util.rs

@@ -1,24 +0,0 @@
-// Copyright © 2025 Cyberus Technology GmbH
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-
-//! Test utilities.
-
-use std::cmp::Ordering;
-
-use clap::Arg;
-
-/// Ensures that all [`Arg`]s are sorted alphabetically.
-pub fn assert_args_sorted<'a, F: Fn() -> R, R: Iterator<Item = &'a Arg>>(get_base_iter: F) {
-    let iter = get_base_iter().zip(get_base_iter().skip(1));
-    for (arg, next) in iter {
-        assert_ne!(
-            arg.get_id().cmp(next.get_id()),
-            Ordering::Greater,
-            "args not alphabetically sorted: arg={}, next={}",
-            arg.get_id(),
-            next.get_id()
-        );
-    }
-}

cloud-hypervisor/tests/common/mod.rs

@@ -1,6 +0,0 @@
-// Copyright 2025 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-
-pub(crate) mod tests_wrappers;
-pub(crate) mod utils;

cloud-hypervisor/tests/integration_cvm.rs

@@ -1,327 +0,0 @@
-// Copyright 2026 The Cloud Hypervisor Authors. All rights reserved.
-//
-// SPDX-License-Identifier: Apache-2.0
-//
-#![cfg(any(devcli_testenv, clippy))]
-#![allow(clippy::undocumented_unsafe_blocks)]
-// When enabling the `mshv` feature, we skip quite some tests and
-// hence have known dead-code. This annotation silences dead-code
-// related warnings for our quality workflow to pass.
-#![allow(dead_code)]
-mod common;
-
-#[cfg(all(feature = "sev_snp", target_arch = "x86_64"))]
-mod common_cvm {
-    use block::ImageType;
-    use common::tests_wrappers::*;
-    use common::utils::*;
-    use test_infra::*;
-    const NUM_PCI_SEGMENTS: u16 = 8;
-
-    use super::*;
-    macro_rules! basic_cvm_guest {
-        ($image_name:expr) => {{
-            let disk_config = UbuntuDiskConfig::new($image_name.to_string());
-            GuestFactory::new_confidential_guest_factory().create_guest(Box::new(disk_config))
-        }};
-    }
-
-    #[test]
-    fn test_jammy_simple_launch() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-
-        _test_simple_launch(&guest);
-    }
-
-    #[test]
-    fn test_api_http_create_boot() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(4);
-        let target_api = TargetApi::new_http_api(&guest.tmp_dir);
-        _test_api_create_boot(&target_api, &guest);
-    }
-
-    #[test]
-    fn test_api_http_shutdown() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(4);
-
-        let target_api = TargetApi::new_http_api(&guest.tmp_dir);
-        _test_api_shutdown(&target_api, &guest);
-    }
-
-    #[test]
-    fn test_api_http_delete() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        let target_api = TargetApi::new_http_api(&guest.tmp_dir);
-        _test_api_delete(&target_api, &guest);
-    }
-
-    #[test]
-    fn test_power_button() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_power_button(&guest);
-    }
-
-    #[test]
-    fn test_virtio_vsock() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_virtio_vsock(&guest, false);
-    }
-
-    #[test]
-    fn test_multi_cpu() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_multi_cpu(&guest);
-    }
-
-    #[test]
-    fn test_cpu_affinity() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(2);
-        _test_cpu_affinity(&guest);
-    }
-
-    #[test]
-    fn test_virtio_queue_affinity() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(4);
-        _test_virtio_queue_affinity(&guest);
-    }
-
-    #[test]
-    fn test_pci_msi() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_pci_msi(&guest);
-    }
-
-    #[test]
-    fn test_virtio_net_ctrl_queue() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_virtio_net_ctrl_queue(&guest);
-    }
-
-    #[test]
-    fn test_pci_multiple_segments() {
-        // Use 8 segments to test the multiple segment support since it's more than the default 6
-        //  supported by Linux
-        // IGVM file used by Sev-Snp Guest now support up to 8 segments, so we can use 8 segments for testing.
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_pci_multiple_segments(&guest, NUM_PCI_SEGMENTS, 5);
-    }
-
-    #[test]
-    fn test_direct_kernel_boot() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_direct_kernel_boot(&guest);
-    }
-
-    #[test]
-    fn test_virtio_block_io_uring() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME,
-        );
-        _test_virtio_block(&guest, false, true, false, false, ImageType::Raw);
-    }
-
-    #[test]
-    fn test_virtio_block_aio() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME,
-        );
-        _test_virtio_block(&guest, true, false, false, false, ImageType::Raw);
-    }
-
-    #[test]
-    fn test_virtio_block_sync() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME,
-        );
-        _test_virtio_block(&guest, true, true, false, false, ImageType::Raw);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2,
-        );
-        _test_virtio_block(&guest, false, false, true, false, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2_zlib() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2_ZLIB,
-        );
-        _test_virtio_block(&guest, false, false, true, false, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2_zstd() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2_ZSTD,
-        );
-        _test_virtio_block(&guest, false, false, true, false, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2_backing_zstd_file() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2_BACKING_ZSTD_FILE,
-        );
-
-        _test_virtio_block(&guest, false, false, true, true, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2_backing_uncompressed_file() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2_BACKING_UNCOMPRESSED_FILE,
-        );
-
-        _test_virtio_block(&guest, false, false, true, true, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_qcow2_backing_raw_file() {
-        let guest = make_virtio_block_guest(
-            &GuestFactory::new_confidential_guest_factory(),
-            JAMMY_IMAGE_NAME_QCOW2_BACKING_RAW_FILE,
-        );
-        _test_virtio_block(&guest, false, false, true, true, ImageType::Qcow2);
-    }
-
-    #[test]
-    fn test_virtio_block_dynamic_vhdx_expand() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_virtio_block_dynamic_vhdx_expand(&guest);
-    }
-
-    #[test]
-    fn test_split_irqchip() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_split_irqchip(&guest);
-    }
-
-    #[test]
-    fn test_dmi_uuid() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_dmi_uuid(&guest);
-    }
-
-    #[test]
-    fn test_dmi_oem_strings() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_dmi_oem_strings(&guest);
-    }
-
-    #[test]
-    fn test_multiple_network_interfaces() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_multiple_network_interfaces(&guest);
-    }
-
-    #[test]
-    fn test_serial_off() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_serial_off(&guest);
-    }
-
-    #[test]
-    fn test_virtio_console() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_virtio_console(&guest);
-    }
-
-    #[test]
-    fn test_console_file() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_console_file(&guest);
-    }
-
-    #[test]
-    fn test_direct_kernel_boot_noacpi() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_direct_kernel_boot_noacpi(&guest);
-    }
-
-    #[test]
-    fn test_pci_bar_reprogramming() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_pci_bar_reprogramming(&guest);
-    }
-
-    #[test]
-    fn test_memory_overhead() {
-        let guest_memory_size_kb: u32 = 512 * 1024;
-        let guest =
-            basic_cvm_guest!(JAMMY_IMAGE_NAME).with_memory(&format!("{guest_memory_size_kb}K"));
-        _test_memory_overhead(&guest, guest_memory_size_kb);
-    }
-
-    #[test]
-    fn test_landlock() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_landlock(&guest);
-    }
-
-    #[test]
-    fn test_disk_hotplug() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_disk_hotplug(&guest, false);
-    }
-
-    #[test]
-    fn test_net_hotplug() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_net_hotplug(&guest, NUM_PCI_SEGMENTS, None);
-    }
-
-    #[test]
-    fn test_counters() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_counters(&guest);
-    }
-
-    #[test]
-    fn test_watchdog() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_watchdog(&guest);
-    }
-
-    #[test]
-    fn test_pvpanic() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME);
-        _test_pvpanic(&guest);
-    }
-
-    #[test]
-    fn test_tap_from_fd() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(2);
-        _test_tap_from_fd(&guest);
-    }
-
-    #[test]
-    fn test_macvtap() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(2);
-        _test_macvtap(&guest, false, "guestmacvtap0", "hostmacvtap0");
-    }
-
-    #[test]
-    fn test_macvtap_hotplug() {
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(2);
-        _test_macvtap(&guest, true, "guestmacvtap1", "hostmacvtap1");
-    }
-
-    #[test]
-    fn test_vdpa_block() {
-        assert!(exec_host_command_status("lsmod | grep vdpa_sim_blk").success());
-
-        let guest = basic_cvm_guest!(JAMMY_IMAGE_NAME).with_cpu(2);
-        _test_vdpa_block(&guest);
-    }
-}