build: Release v23.1 (bug fix release)

Signed-off-by: Rob Bradford <robert.bradford@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 +1,63 @@
+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
+          - 1.56
+        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: ${{ matrix.target }}
+            override: true
+
+      - name: Debug Check (default features)
+        run: |
+          git rev-list origin/main..$GITHUB_SHA | xargs -t -I % sh -c 'git checkout %; cargo check --tests --all --target=${{ matrix.target }}'
+          git checkout $GITHUB_SHA
+
+      - name: Build (default features)
+        run: cargo rustc --bin cloud-hypervisor -- -D warnings
+
+      - name: Build (common + kvm)
+        run: cargo rustc --bin cloud-hypervisor --no-default-features --features "common,kvm"  -- -D warnings
+
+      - name: Build (default features + tdx)
+        run: cargo rustc --bin cloud-hypervisor --features "tdx" -- -D warnings
+
+      - name: Build (default features + amx)
+        run: cargo rustc --bin cloud-hypervisor --features "amx" -- -D warnings
+
+      - name: Build (default features + gdb)
+        run: cargo rustc --bin cloud-hypervisor --features "gdb" -- -D warnings
+
+      - name: Build (common + mshv)
+        run: cargo rustc --bin cloud-hypervisor --no-default-features --features "common,mshv"  -- -D warnings
+
+      - name: Release Build (default features)
+        run: cargo build --all --release --target=${{ 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

@@ -1,52 +1,46 @@
 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
-  IMAGE_NAME: ${{ github.repository }}
 
 jobs:
   main:
     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
+      - name: Login to DockerHub
+        if: ${{ github.event_name == 'push' }}
+        uses: docker/login-action@v1 
         with:
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
 
       - name: Docker meta
         id: meta
-        uses: docker/metadata-action@v6
+        uses: docker/metadata-action@v3
         with:
-          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          # list of Docker images to use as base name for tags
+          images: cloudhypervisor/dev
           # 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 +49,7 @@ 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-aarch64.yaml

@@ -0,0 +1,47 @@
+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: ${{ matrix.experimental }}
+    strategy:
+      fail-fast: false
+      matrix:
+        rust:
+          - stable
+        target:
+          - aarch64-unknown-linux-gnu
+        experimental: [false]
+        include:
+          - rust: beta
+            target: aarch64-unknown-linux-gnu
+            experimental: true
+    steps:
+      - name: Code checkout
+        uses: actions/checkout@v2
+      - name: Install Rust toolchain (${{ matrix.rust }})
+        uses: actions-rs/toolchain@v1
+        with:
+            toolchain: ${{ matrix.rust }}
+            target: ${{ matrix.target }}
+            override: true
+            components: rustfmt, clippy
+      - name: Formatting (rustfmt)
+        run: cargo fmt -- --check
+
+      - name: Clippy (common + kvm)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: true
+          command: clippy
+          args: --target=${{ matrix.target }} --tests --all --no-default-features --features "common,kvm" -- -D warnings
+
+      - name: Clippy (default features)
+        uses: actions-rs/cargo@v1
+        with:
+          use-cross: true
+          command: clippy
+          args: --target=${{ matrix.target }} --tests --all -- -D warnings

.github/workflows/quality.yaml

@@ -0,0 +1,52 @@
+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: ${{ matrix.experimental }}
+    strategy:
+      fail-fast: false
+      matrix:
+        rust:
+          - stable
+        target:
+          - x86_64-unknown-linux-gnu
+        experimental: [false]
+        include:
+          - rust: beta
+            target: x86_64-unknown-linux-gnu
+            experimental: true
+    steps:
+      - name: Code checkout
+        uses: actions/checkout@v2
+      - name: Install Rust toolchain (${{ matrix.rust }})
+        uses: actions-rs/toolchain@v1
+        with:
+          toolchain: ${{ matrix.rust }}
+          target: ${{ matrix.target }}
+          override: true
+          components: rustfmt, clippy
+
+      - name: Formatting (rustfmt)
+        run: cargo fmt -- --check
+
+      - name: Clippy (common + kvm)
+        run: cargo clippy --all --all-targets --no-default-features --tests --features "common,kvm" -- -D warnings
+
+      - name: Clippy (default features)
+        run: cargo clippy --all --all-targets --tests -- -D warnings
+
+      - name: Clippy (default features + amx)
+        run: cargo clippy --all --all-targets --tests --features "amx" -- -D warnings
+
+      - name: Clippy (default features + gdb)
+        run: cargo clippy --all --all-targets --tests --features "gdb" -- -D warnings
+
+      - name: Clippy (common + mshv)
+        run: cargo clippy --all --all-targets --no-default-features --tests --features "common,mshv" -- -D warnings
+
+      - name: Check build did not modify any files
+        run: test -z "$(git status --porcelain)"

.github/workflows/release.yaml

@@ -1,95 +1,141 @@
 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: [create]
 
 jobs:
   release:
-    if: (github.event_name == 'create' && github.event.ref_type == 'tag') || github.event_name == 'merge_group'
-    name: Release ${{ matrix.platform.target }}
-    strategy:
-      fail-fast: false
-      matrix:
-        platform:
-        - target: x86_64-unknown-linux-gnu
-          args: --all --release --features mshv
-          name_ch: cloud-hypervisor
-          name_ch_remote: ch-remote
-        - target: x86_64-unknown-linux-musl
-          args: --all --release --features mshv
-          name_ch: cloud-hypervisor-static
-          name_ch_remote: ch-remote-static
-        - target: aarch64-unknown-linux-musl
-          args:  --all --release
-          name_ch: cloud-hypervisor-static-aarch64
-          name_ch_remote: ch-remote-static-aarch64
+    if: github.event_name == 'create' && github.event.ref_type == 'tag'
+    name: Release
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@v2
       - name: Install musl-gcc
-        if: contains(matrix.platform.target, 'musl')
-        run: sudo apt install -y musl-tools
+        run: sudo apt install -y musl-tools      
       - name: Create release directory
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
         run: rsync -rv --exclude=.git . ../cloud-hypervisor-${{ 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:
-          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
+            toolchain: 1.58
+            target: x86_64-unknown-linux-gnu
+      - name: Install Rust toolchain (x86_64-unknown-linux-musl)
+        uses: actions-rs/toolchain@v1
         with:
-          name: Artifacts for ${{ matrix.platform.target }}
-          path: |
-            ./${{ matrix.platform.name_ch }}
-            ./${{ matrix.platform.name_ch_remote }}
+            toolchain: 1.58
+            target: x86_64-unknown-linux-musl
+      - name: Build
+        uses: actions-rs/cargo@v1
+        with:
+            toolchain: 1.58
+            command: build
+            args: --all --release --target=x86_64-unknown-linux-gnu
+      - name: Static Build
+        uses: actions-rs/cargo@v1
+        with:
+            toolchain: 1.58
+            command: build
+            args: --all --release --target=x86_64-unknown-linux-musl
+      - name: Strip cloud-hypervisor binaries
+        run: strip target/*/release/cloud-hypervisor
+      - name: Install Rust toolchain (aarch64-unknown-linux-musl)
+        uses: actions-rs/toolchain@v1
+        with:
+            toolchain: 1.58
+            target: aarch64-unknown-linux-musl
+            override: true
+      - name: Static Build (AArch64)
+        uses: actions-rs/cargo@v1
+        with:
+            use-cross: true
+            command: build
+            args: --all --release --target=aarch64-unknown-linux-musl
       - name: Vendor
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
         working-directory: ../cloud-hypervisor-${{ github.event.ref }}
         run: |
           mkdir ../vendor-cargo-home
           export CARGO_HOME=$(realpath ../vendor-cargo-home)
           mkdir .cargo
           cargo vendor > .cargo/config.toml
-      - name: Create vendored source archive
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
-        run: tar cJf cloud-hypervisor-${{ github.event.ref }}.tar.xz ../cloud-hypervisor-${{ github.event.ref }}
-      - name: Upload cloud-hypervisor vendored source archive
-        if: |
-          github.event_name == 'create' && github.event.ref_type == 'tag' &&
-          matrix.platform.target == 'x86_64-unknown-linux-gnu'
-        id: upload-release-cloud-hypervisor-vendored-sources
-        uses: actions/upload-artifact@v7
-        with:
-          path: cloud-hypervisor-${{ 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
+      - name: Create Release
+        id: create_release
+        uses: actions/create-release@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
+          tag_name: ${{ github.ref }}
+          release_name: ${{ github.ref }}
           draft: true
-          files: |
-            ./${{ matrix.platform.name_ch }}
-            ./${{ matrix.platform.name_ch_remote }}
-            ./cloud-hypervisor-${{ github.event.ref }}.tar.xz
+          prerelease: true
+      - name: Create vendored source archive
+        working-directory: ../
+        run:  tar cJf cloud-hypervisor-${{ github.event.ref }}.tar.xz cloud-hypervisor-${{ github.event.ref }}
+      - name: Upload cloud-hypervisor vendored source archive
+        id: upload-release-cloud-hypervisor-vendored-sources
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          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
+      - name: Upload cloud-hypervisor
+        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
+        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
+        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
+        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: Upload static AArch64 cloud-hypervisor
+        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
+        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

.gitignore

@@ -1,13 +1,6 @@
+/build
+/.cargo
+/target
 **/*.rs.bk
 **/Cargo.lock
 **/rusty-tags.vi
-/.agents
-/.cargo
-/.claude
-/.codex
-/.vscode
-/build
-/rpm/SOURCES
-/target
-/vendor
-__pycache__

.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.
-

CODEOWNERS

@@ -1,2 +0,0 @@
-# Add the list of code owners here (using their GitHub username)
-* @cloud-hypervisor/cloud-hypervisor-reviewers

CONTRIBUTING.md

@@ -2,140 +2,57 @@
 
 Cloud Hypervisor is an open source project licensed under the [Apache v2
 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
-license of those projects.
+Clause](https://opensource.org/licenses/BSD-3-Clause) license. Contributions
+can be made under either license or both. 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 license of those
+projects.
 
-New code should be under the [Apache v2
-License](https://opensource.org/licenses/Apache-2.0).
+## Coding Style
 
-## Coding Style & Code Comments
-
-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
-
-## 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.
-
-```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
-```
-
-You will need to `chmod +x .git/hooks/pre-commit` to have it run on every
-commit you make.
+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).
 
 ## 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 +60,14 @@ 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. Add reviewers to your pull request and then work with your reviewers to address
+   any comments and obtain minimum of 2 [maintainers](MAINTAINERS.md) approvals.
+   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, one of the maintainers will merge it.
 
 ## Issue tracking
 
@@ -169,83 +84,16 @@ 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 +1,92 @@
-# Cloud Hypervisor Workspace
-#
-# The main crate producing the binaries is in `./cloud-hypervisor`.
+[package]
+name = "cloud-hypervisor"
+version = "23.1.0"
+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
+rust-version = "1.56"
 
 [profile.release]
-codegen-units = 1
 lto = true
-opt-level = "s"
-strip = true
 
-[profile.profiling]
-debug = true
-inherits = "release"
-strip = false
+[dependencies]
+anyhow = "1.0.56"
+api_client = { path = "api_client" }
+clap = { version = "3.1.8", features = ["wrap_help","cargo"] }
+epoll = "4.3.1"
+event_monitor = { path = "event_monitor" }
+hypervisor = { path = "hypervisor" }
+libc = "0.2.123"
+log = { version = "0.4.16", features = ["std"] }
+option_parser = { path = "option_parser" }
+seccompiler = "0.2.0"
+serde_json = "1.0.79"
+signal-hook = "0.3.13"
+thiserror = "1.0.30"
+vmm = { path = "vmm" }
+vmm-sys-util = "0.9.0"
+vm-memory = "0.7.0"
+
+[build-dependencies]
+clap = { version = "3.1.8", features = ["cargo"] }
+
+# List of patched crates
+[patch.crates-io]
+kvm-bindings = { git = "https://github.com/cloud-hypervisor/kvm-bindings", branch = "ch-v0.5.0-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"
+lazy_static= "1.4.0"
+net_util = { path = "net_util" }
+serde_json = "1.0.79"
+test_infra = { path = "test_infra" }
+wait-timeout = "0.2.0"
+
+[features]
+default = ["common", "kvm"]
+# Common features for all hypervisors
+common = ["fwdebug"]
+amx = ["vmm/amx"]
+cmos = ["vmm/cmos"]
+fwdebug = ["vmm/fwdebug"]
+gdb = ["vmm/gdb"]
+kvm = ["vmm/kvm"]
+mshv = ["vmm/mshv"]
+tdx = ["vmm/tdx"]
 
 [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",
+    "acpi_tables",
+    "api_client",
+    "arch",
+    "block_util",
+    "devices",
+    "event_monitor",
+    "hypervisor",
+    "net_gen",
+    "net_util",
+    "option_parser",
+    "pci",
+    "performance-metrics",
+    "qcow",
+    "rate_limiter",
+    "test_infra",
+    "vfio_user",
+    "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 +1,404 @@
+def runWorkers = true
+pipeline{
+	agent none
+	stages {
+		stage ('Early checks') {
+			agent { node { label 'built-in' } }
+			stages {
+				stage ('Checkout') {
+					steps {
+						checkout scm
+					}
+				}
+				stage ('Check for documentation only changes') {
+					when {
+						expression {
+							return docsFileOnly()
+						}
+					}
+					steps {
+						script {
+							runWorkers = false
+							echo "Documentation only changes, no need to run the CI"
+						}
+					}
+				}
+				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 'focal' } }
+					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 ('AArch64 worker build') {
+					agent { node { label 'bionic-arm64' } }
+					when {
+						beforeAgent true
+						expression {
+							return runWorkers
+						}
+					}
+					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"
+							}
+						}
+					}
+					post {
+						always {
+							sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+							deleteDir()
+						}
+					}
+				}
+				stage ('Worker build (musl)') {
+					agent { node { label 'focal' } }
+					when {
+						beforeAgent true
+						expression {
+							return runWorkers
+						}
+					}
+					stages {
+						stage ('Checkout') {
+							steps {
+								checkout scm
+							}
+						}
+						stage ('Prepare environment') {
+							steps {
+								sh "scripts/prepare_vdpa.sh"
+							}
+						}
+						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 ('Worker build SGX') {
+					agent { node { label 'bionic-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 'bionic-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()
+						}
+					}
+				}
+				stage ('Worker build - Windows guest') {
+					agent { node { label 'focal' } }
+					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()
+							}
+						}
+						stage ('Download assets') {
+							steps {
+								sh "mkdir ${env.HOME}/workloads"
+								sh 'az storage blob download --container-name private-images --file "$HOME/workloads/windows-server-2019.raw" --name windows-server-2019.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 - Live Migration') {
+					agent { node { label 'focal-small' } }
+					when {
+						beforeAgent true
+						expression {
+							return runWorkers
+						}
+					}
+					stages {
+						stage ('Checkout') {
+							steps {
+								checkout scm
+							}
+						}
+						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 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 ('Worker build - Metrics') {
+					agent { node { label 'focal-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'
+							}
+						}
+					}
+				}
+			}
+		}
+	}
+	post {
+		regression {
+			script {
+				if (env.BRANCH_NAME == 'main') {
+					slackSend (color: '#ff0000', message: '"main" branch build is now failing')
+				}
+			}
+		}
+		fixed {
+			script {
+				if (env.BRANCH_NAME == 'main') {
+					slackSend (color: '#00ff00', message: '"main" branch build is now fixed')
+				}
+			}
+		}
+	}
+}
+
+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: ${build.number}"
+				build.doStop()
+			}
+		}
+}
+
+def installAzureCli() {
+	sh "sudo apt install -y ca-certificates curl apt-transport-https lsb-release gnupg"
+	sh "curl -sL https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor | sudo tee /etc/apt/trusted.gpg.d/microsoft.gpg > /dev/null"
+	sh "echo \"deb [arch=amd64] https://packages.microsoft.com/repos/azure-cli/ focal main\" | sudo tee /etc/apt/sources.list.d/azure-cli.list"
+	sh "sudo apt update"
+	sh "sudo apt install -y azure-cli"
+}
+
+def boolean docsFileOnly() {
+    if (env.CHANGE_TARGET == null) {
+        return false;
+    }
+
+    return sh(
+        returnStatus: true,
+        script: "git diff --name-only origin/${env.CHANGE_TARGET}... | grep -v '\\.md'"
+    ) != 0
+}

MAINTAINERS.md

@@ -2,7 +2,6 @@
 
 - Sebastien Boeuf - @sboeuf
 - Robert Bradford - @rbradford
-- Bo Chen - @likebreath
 - Samuel Ortiz - @sameo
 - Wei Liu - @liuw
 - Michael Zhao - @michael2012z

README.md

@@ -4,43 +4,42 @@
     - [Architectures](#architectures)
     - [Guest OS](#guest-os)
 - [2. Getting Started](#2-getting-started)
-  - [Host OS](#host-os)
-  - [Use Pre-built Binaries](#use-pre-built-binaries)
-  - [Packages](#packages)
-  - [Building from Source](#building-from-source)
-  - [Booting Linux](#booting-linux)
-    - [Firmware Booting](#firmware-booting)
-    - [Custom Kernel and Disk Image](#custom-kernel-and-disk-image)
-      - [Building your Kernel](#building-your-kernel)
+  - [Preparation](#preparation)
+  - [Install prerequisites](#install-prerequisites)
+  - [Clone and build](#clone-and-build)
+    - [Containerized builds and tests](#containerized-builds-and-tests)
+  - [Run](#run)
+    - [Cloud image](#cloud-image)
+    - [Custom kernel and disk image](#custom-kernel-and-disk-image)
+      - [Building your kernel](#building-your-kernel)
       - [Disk image](#disk-image)
       - [Booting the guest VM](#booting-the-guest-vm)
 - [3. Status](#3-status)
   - [Hot Plug](#hot-plug)
   - [Device Model](#device-model)
-  - [Roadmap](#roadmap)
-- [4. Relationship with _Rust VMM_ Project](#4-relationship-with-rust-vmm-project)
-  - [Differences with Firecracker and crosvm](#differences-with-firecracker-and-crosvm)
+  - [TODO](#todo)
+- [4. `rust-vmm` project dependency](#4-rust-vmm-project-dependency)
+  - [Firecracker and crosvm](#firecracker-and-crosvm)
 - [5. Community](#5-community)
   - [Contribute](#contribute)
-  - [Slack](#slack)
-  - [Mailing list](#mailing-list)
+  - [Join us](#join-us)
   - [Security issues](#security-issues)
 
 # 1. What is Cloud Hypervisor?
 
 Cloud Hypervisor is an open source Virtual Machine Monitor (VMM) that runs on
-top of the [KVM](https://www.kernel.org/doc/Documentation/virtual/kvm/api.txt)
-hypervisor and the Microsoft Hypervisor (MSHV).
+top of [KVM](https://www.kernel.org/doc/Documentation/virtual/kvm/api.txt)
+hypervisor and Microsoft Hypervisor (MSHV).
 
-The project focuses on running modern, _Cloud Workloads_, on specific, common,
-hardware architectures. In this case _Cloud Workloads_ refers to those that are
-run by customers inside a Cloud Service Provider. This means modern operating
-systems with most I/O handled by
-paravirtualised devices (e.g. _virtio_), no requirement for legacy devices, and
+The project focuses on exclusively running modern, cloud workloads, on top of
+a limited set of hardware architectures and platforms. Cloud workloads refers
+to those that are usually run by customers inside a cloud provider. For our
+purposes this means modern operating systems with most I/O handled by
+paravirtualised devices (i.e. virtio), no requirement for legacy devices, and
 64-bit CPUs.
 
 Cloud Hypervisor is implemented in [Rust](https://www.rust-lang.org/) and is
-based on the [Rust VMM](https://github.com/rust-vmm) crates.
+based on the [rust-vmm](https://github.com/rust-vmm) crates.
 
 ## Objectives
 
@@ -59,13 +58,9 @@ 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
+some small differences in functionality between the two architectures
+(see [#1125](https://github.com/cloud-hypervisor/cloud-hypervisor/issues/1125)).
 
 ### Guest OS
 
@@ -73,202 +68,189 @@ Cloud Hypervisor supports `64-bit Linux` and Windows 10/Windows Server 2019.
 
 # 2. Getting Started
 
-The following sections describe how to build and run Cloud Hypervisor.
+Below sections describe how to build and run Cloud Hypervisor on the `x86_64`
+platform. For getting started on the `AArch64` platform, please refer to the
+[Arm64 documentation](docs/arm64.md).
 
-## Prerequisites for AArch64
+## Preparation
 
-- AArch64 servers (recommended) or development boards equipped with the GICv3
-  interrupt controller.
-
-## 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.
-
-## Use Pre-built Binaries
-
-The recommended approach to getting started with Cloud Hypervisor is by using a
-pre-built binary. Binaries are available for the [latest
-release](https://github.com/cloud-hypervisor/cloud-hypervisor/releases/latest).
-Use `cloud-hypervisor-static` for `x86-64` or `cloud-hypervisor-static-aarch64`
-for `AArch64` platform.
-
-## Packages
-
-For convenience, packages are also available targeting some popular Linux
-distributions. This is thanks to the [Open Build
-Service](https://build.opensuse.org). The [OBS
-README](https://github.com/cloud-hypervisor/obs-packaging) explains how to
-enable the repository in a supported Linux distribution and install Cloud Hypervisor
-and accompanying packages. Please report any packaging issues in the
-[obs-packaging](https://github.com/cloud-hypervisor/obs-packaging) repository.
-
-## Building from Source
-
-Please see the [instructions for building from source](docs/building.md) if you
-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.
-
-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
-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.
-
-### Firmware Booting
-
-Cloud Hypervisor supports booting disk images containing all needed components
-to run cloud workloads, a.k.a. cloud images.
-
-The following sample commands will download an Ubuntu Cloud image, converting
-it into a format that Cloud Hypervisor can use and a firmware to boot the image
-with.
+We create a folder to build and run `cloud-hypervisor` at `$HOME/cloud-hypervisor`
 
 ```shell
-$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img
-$ qemu-img convert -p -f qcow2 -O raw focal-server-cloudimg-amd64.img focal-server-cloudimg-amd64.raw
-$ wget https://github.com/cloud-hypervisor/rust-hypervisor-firmware/releases/download/0.4.2/hypervisor-fw
+$ export CLOUDH=$HOME/cloud-hypervisor
+$ mkdir $CLOUDH
 ```
 
-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.
+## Install prerequisites
+
+You need to install some prerequisite packages in order to build and test Cloud
+Hypervisor. Here, all the steps are based on Ubuntu, for other Linux
+distributions please replace the package manager and package name.
 
 ```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 \
+# Install git
+$ sudo apt install git
+# Install rust tool chain
+$ curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
+# Install build-essential
+$ sudo apt install build-essential
+# If you want to build statically linked binary please add musl target
+$ rustup target add x86_64-unknown-linux-musl
+```
+
+## Clone and build
+
+First you need to clone and build the cloud-hypervisor repo:
+
+```shell
+$ pushd $CLOUDH
+$ git clone https://github.com/cloud-hypervisor/cloud-hypervisor.git
+$ cd cloud-hypervisor
+$ cargo build --release
+
+# We need to give the cloud-hypervisor binary the NET_ADMIN capabilities for it to set TAP interfaces up on the host.
+$ sudo setcap cap_net_admin+ep ./target/release/cloud-hypervisor
+
+# If you want to build statically linked binary
+$ cargo build --release --target=x86_64-unknown-linux-musl --all
+$ popd
+```
+
+This will build a `cloud-hypervisor` binary under
+`$CLOUDH/cloud-hypervisor/target/release/cloud-hypervisor`.
+
+### Containerized builds and tests
+
+If you want to build and test Cloud Hypervisor without having to install all the
+required dependencies (The rust toolchain, cargo tools, etc), you can also use
+Cloud Hypervisor's development script: `dev_cli.sh`. Please note that upon its
+first invocation, this script will pull a fairly large container image.
+
+For example, to build the Cloud Hypervisor release binary:
+
+```shell
+$ pushd $CLOUDH
+$ cd cloud-hypervisor
+$ ./scripts/dev_cli.sh build --release
+```
+
+With `dev_cli.sh`, one can also run the Cloud Hypervisor CI locally. This can be
+very convenient for debugging CI errors without having to fully rely on the
+Cloud Hypervisor CI infrastructure.
+
+For example, to run the Cloud Hypervisor unit tests:
+
+```shell
+$ ./scripts/dev_cli.sh tests --unit
+```
+
+Run the `./scripts/dev_cli.sh --help` command to view all the supported
+development script commands and their related options.
+
+## Run
+
+You can run a guest VM by either using an existing cloud image or booting into
+your own kernel and disk image.
+
+### Cloud image
+
+Cloud Hypervisor supports booting disk images containing all needed
+components to run cloud workloads, a.k.a. cloud images. To do that we rely on
+the [Rust Hypervisor
+Firmware](https://github.com/cloud-hypervisor/rust-hypervisor-firmware) project
+to provide an ELF formatted KVM firmware for `cloud-hypervisor` to directly
+boot into.
+
+We need to get the latest `rust-hypervisor-firmware` release and also a working
+cloud image. Here we will use a Ubuntu image:
+
+```shell
+$ pushd $CLOUDH
+$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img
+$ qemu-img convert -p -f qcow2 -O raw focal-server-cloudimg-amd64.img focal-server-cloudimg-amd64.raw
+$ wget https://github.com/cloud-hypervisor/rust-hypervisor-firmware/releases/download/0.3.2/hypervisor-fw
+$ popd
+```
+
+```shell
+$ pushd $CLOUDH
+$ sudo setcap cap_net_admin+ep ./cloud-hypervisor/target/release/cloud-hypervisor
+$ ./cloud-hypervisor/target/release/cloud-hypervisor \
+	--kernel ./hypervisor-fw \
+	--disk path=focal-server-cloudimg-amd64.raw \
 	--cpus boot=4 \
 	--memory size=1024M \
 	--net "tap=,mac=,ip=,mask="
+$ popd
 ```
 
-If access to the firmware messages or interaction with the boot loader (e.g.
-GRUB) is required then it necessary to switch to the serial console instead of
-`virtio-console`.
+Multiple arguments can be given to the `--disk` parameter.
 
-```shell
-$ ./cloud-hypervisor \
-	--kernel ./hypervisor-fw \
-	--disk path=focal-server-cloudimg-amd64.raw path=/tmp/ubuntu-cloudinit.img \
-	--cpus boot=4 \
-	--memory size=1024M \
-	--net "tap=,mac=,ip=,mask=" \
-	--serial tty \
-	--console off
-```
+### Custom kernel and disk image
 
-## Booting: `--firmware` vs `--kernel`
+#### Building your 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 into a `vmlinux` ELF kernel.
+In order to support virtio-watchdog we have our own development branch. You are
+of course able to use your own kernel but these instructions will continue with
+the version that we develop and test against.
 
 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
+$ pushd $CLOUDH
+$ git clone --depth 1 https://github.com/cloud-hypervisor/linux.git -b ch-5.15.12 linux-cloud-hypervisor
 $ pushd linux-cloud-hypervisor
-$ make ch_defconfig
-# Do native build of the x86-64 kernel
+
+# Use the cloud-hypervisor kernel config to build your kernel
+$ cp $CLOUDH/cloud-hypervisor/resources/linux-config-x86_64 .config
 $ KCFLAGS="-Wa,-mx86-used-note=no" make bzImage -j `nproc`
-# Do native build of the AArch64 kernel
-$ make -j `nproc`
 $ popd
 ```
 
-For x86-64, the `vmlinux` kernel image will then be located at
+The `vmlinux` kernel image will then be located at
 `linux-cloud-hypervisor/arch/x86/boot/compressed/vmlinux.bin`.
-For AArch64, the `Image` kernel image will then be located at
-`linux-cloud-hypervisor/arch/arm64/boot/Image`.
 
 #### Disk image
 
-For the disk image the same Ubuntu image as before can be used. This contains
-an `ext4` root filesystem.
+For the disk image, we will use a Ubuntu cloud image that contains a root
+partition:
 
 ```shell
-$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img # x86-64
-$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-arm64.img # AArch64
-$ qemu-img convert -p -f qcow2 -O raw focal-server-cloudimg-amd64.img focal-server-cloudimg-amd64.raw # x86-64
-$ qemu-img convert -p -f qcow2 -O raw focal-server-cloudimg-arm64.img focal-server-cloudimg-arm64.raw # AArch64
+$ pushd $CLOUDH
+$ wget https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img
+$ qemu-img convert -p -f qcow2 -O raw focal-server-cloudimg-amd64.img focal-server-cloudimg-amd64.raw
+$ popd
 ```
 
 #### Booting the guest VM
 
-These sample commands boot the disk image using the custom kernel whilst also
-supplying the desired kernel command line.
-
-- x86-64
+Now we can directly boot into our custom kernel and make it use the Ubuntu root
+partition. If we want to have 4 vCPUs and 1024 MBytes of memory:
 
 ```shell
-$ sudo setcap cap_net_admin+ep ./cloud-hypervisor
-$ ./create-cloud-init.sh
-$ ./cloud-hypervisor \
+$ pushd $CLOUDH
+$ sudo setcap cap_net_admin+ep ./cloud-hypervisor/target/release/cloud-hypervisor
+$ ./cloud-hypervisor/target/release/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 \
 	--cmdline "console=hvc0 root=/dev/vda1 rw" \
 	--cpus boot=4 \
 	--memory size=1024M \
 	--net "tap=,mac=,ip=,mask="
 ```
 
-- AArch64
+The above example use the `virtio-console` device as the guest console, and this
+device may not be enabled soon enough by the guest kernel to get early kernel
+debug messages.
+
+When in need for earlier debug messages, using the legacy serial device based
+console is preferred:
 
-```shell
-$ 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 \
-	--cmdline "console=hvc0 root=/dev/vda1 rw" \
-	--cpus boot=4 \
-	--memory size=1024M \
-	--net "tap=,mac=,ip=,mask="
 ```
-
-If earlier kernel messages are required the serial console should be used instead of `virtio-console`.
-
-- x86-64
-
-```shell
-$ ./cloud-hypervisor \
+$ ./cloud-hypervisor/target/release/cloud-hypervisor \
 	--kernel ./linux-cloud-hypervisor/arch/x86/boot/compressed/vmlinux.bin \
 	--console off \
 	--serial tty \
@@ -279,24 +261,10 @@ $ ./cloud-hypervisor \
 	--net "tap=,mac=,ip=,mask="
 ```
 
-- AArch64
-
-```shell
-$ ./cloud-hypervisor \
-	--kernel ./linux-cloud-hypervisor/arch/arm64/boot/Image \
-	--console off \
-	--serial tty \
-	--disk path=focal-server-cloudimg-arm64.raw \
-	--cmdline "console=ttyAMA0 root=/dev/vda1 rw" \
-	--cpus boot=4 \
-	--memory size=1024M \
-	--net "tap=,mac=,ip=,mask="
-```
-
 # 3. Status
 
-Cloud Hypervisor is under active development. The following stability
-guarantees are currently made:
+Cloud Hypervisor is under active development. The following stability guarantees
+are currently made:
 
 * The API (including command line options) will not be removed or changed in a
   breaking way without a minimum of 2 major releases notice. Where possible
@@ -314,18 +282,14 @@ Currently the following items are **not** guaranteed across updates:
 * The following features are considered experimental and may change
   substantially between releases: TDX, vfio-user, vDPA.
 
-Further details can be found in the [release documentation](docs/releases.md).
+As of 2022-04-05, the following cloud images are supported:
 
-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 Bionic](https://cloud-images.ubuntu.com/bionic/current/) (cloudimg)
+- [Ubuntu Focal](https://cloud-images.ubuntu.com/focal/current/) (cloudimg)
+- [Ubuntu Jammy](https://cloud-images.ubuntu.com/jammy/current/) (cloudimg)
 
 Direct kernel boot to userspace should work with a rootfs from most
-distributions although you may need to enable exotic filesystem types in the
-reference kernel configuration (e.g. XFS or btrfs.)
+distributions.
 
 ## Hot Plug
 
@@ -338,12 +302,14 @@ Cloud Hypervisor supports hotplug of CPUs, passthrough devices (VFIO),
 Details of the device model can be found in this
 [documentation](docs/device_model.md).
 
-## Roadmap
+## TODO
 
-The project roadmap is tracked through a [GitHub
-project](https://github.com/orgs/cloud-hypervisor/projects/6).
+We are not tracking the Cloud Hypervisor TODO list from a specific git tracked
+file but through
+[github issues](https://github.com/cloud-hypervisor/cloud-hypervisor/issues/new)
+instead.
 
-# 4. Relationship with _Rust VMM_ Project
+# 4. `rust-vmm` project dependency
 
 In order to satisfy the design goal of having a high-performance,
 security-focused hypervisor the decision was made to use the
@@ -352,26 +318,39 @@ focus on memory and thread safety makes it an ideal candidate for implementing
 VMMs.
 
 Instead of implementing the VMM components from scratch, Cloud Hypervisor is
-importing the [Rust VMM](https://github.com/rust-vmm) crates, and sharing code
+importing the [rust-vmm](https://github.com/rust-vmm) crates, and sharing code
 and architecture together with other VMMs like e.g. Amazon's
 [Firecracker](https://firecracker-microvm.github.io/) and Google's
 [crosvm](https://chromium.googlesource.com/chromiumos/platform/crosvm/).
 
-Cloud Hypervisor embraces the _Rust VMM_ project's goals, which is to be able
-to share and re-use as many virtualization crates as possible.
+Cloud Hypervisor embraces the rust-vmm project goals, which is to be able to
+share and re-use as many virtualization crates as possible. As such, the Cloud
+Hypervisor relationship with the rust-vmm project is twofold:
 
-## Differences with Firecracker and crosvm
+1. It will use as much of the rust-vmm code as possible. Any new rust-vmm crate
+   that's relevant to the project goals will be integrated as soon as possible.
+2. As it is likely that the rust-vmm project will lack some of the features that
+   Cloud Hypervisor needs (e.g. ACPI, VFIO, vhost-user, etc), we will be using
+   the Cloud Hypervisor VMM to implement and test them, and contribute them back
+   to the rust-vmm project.
+
+## Firecracker and crosvm
 
 A large part of the Cloud Hypervisor code is based on either the Firecracker or
-the crosvm project's implementations. Both of these are VMMs written in Rust
-with a focus on safety and security, like Cloud Hypervisor.
+the crosvm projects implementations. Both of these are VMMs written in Rust with
+a focus on safety and security, like Cloud Hypervisor.
 
-The goal of the Cloud Hypervisor project differs from the aforementioned
-projects in that it aims to be a general purpose VMM for _Cloud Workloads_ and
-not limited to container/serverless or client workloads.
+However we want to emphasize that the Cloud Hypervisor project is neither a fork
+nor a reimplementation of any of those projects. The goals and use cases we're
+trying to meet are different. We're aiming at supporting cloud workloads, i.e.
+those modern, full Linux distribution images currently being run by Cloud
+Service Provider (CSP) tenants.
 
-The Cloud Hypervisor community thanks the communities of both the Firecracker
-and crosvm projects for their excellent work.
+Our primary target is not to support client or serverless use cases, and as such
+our code base already diverges from the crosvm and Firecracker ones. As we add
+more features to support our use cases, we believe that the divergence will
+increase while at the same time sharing as much of the fundamental
+virtualization code through the rust-vmm project crates as possible.
 
 # 5. Community
 
@@ -381,27 +360,20 @@ repository.
 
 ## Contribute
 
-The project strongly believes in building a global, diverse and collaborative
-community around the Cloud Hypervisor project. Anyone who is interested in
+We are working on building a global, diverse and collaborative community around
+the Cloud Hypervisor project. Anyone who is interested in
 [contributing](CONTRIBUTING.md) to the project is welcome to participate.
 
-Contributing to a open source project like Cloud Hypervisor covers a lot more
-than just sending code. Testing, documentation, pull request
+We believe that contributing to a open source project like Cloud Hypervisor
+covers a lot more than just sending code. Testing, documentation, pull request
 reviews, bug reports, feature requests, project improvement suggestions, etc,
 are all equal and welcome means of contribution. See the
 [CONTRIBUTING](CONTRIBUTING.md) document for more details.
 
-## Slack
+## Join us
 
-Get an [invite to our Slack channel](https://join.slack.com/t/cloud-hypervisor/shared_invite/enQtNjY3MTE3MDkwNDQ4LWQ1MTA1ZDVmODkwMWQ1MTRhYzk4ZGNlN2UwNTI3ZmFlODU0OTcwOWZjMTkwZDExYWE3YjFmNzgzY2FmNDAyMjI),
- [join us on Slack](https://cloud-hypervisor.slack.com/), and [participate in our community activities](https://cloud-hypervisor.slack.com/archives/C04R5DUQVBN).
-
-## Mailing list
-
-Please report bugs using the [GitHub issue
-tracker](https://github.com/cloud-hypervisor/cloud-hypervisor/issues) but for
-broader community discussions you may use our [mailing
-list](https://lists.cloudhypervisor.org/g/dev/).
+Get an [invite to our Slack channel](https://join.slack.com/t/cloud-hypervisor/shared_invite/enQtNjY3MTE3MDkwNDQ4LWQ1MTA1ZDVmODkwMWQ1MTRhYzk4ZGNlN2UwNTI3ZmFlODU0OTcwOWZjMTkwZDExYWE3YjFmNzgzY2FmNDAyMjI)
+and [join us on Slack](https://cloud-hypervisor.slack.com/).
 
 ## Security issues
 

acpi_tables/Cargo.toml

@@ -0,0 +1,8 @@
+[package]
+name = "acpi_tables"
+version = "0.1.0"
+authors = ["The Cloud Hypervisor Authors"]
+edition = "2021"
+
+[dependencies]
+vm-memory = "0.7.0"

acpi_tables/src/lib.rs

@@ -0,0 +1,12 @@
+// Copyright © 2019 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+pub mod aml;
+pub mod rsdp;
+pub mod sdt;
+
+fn generate_checksum(data: &[u8]) -> u8 {
+    (255 - data.iter().fold(0u8, |acc, x| acc.wrapping_add(*x))).wrapping_add(1)
+}

acpi_tables/src/rsdp.rs

@@ -0,0 +1,68 @@
+// Copyright © 2019 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+use vm_memory::ByteValued;
+
+#[repr(packed)]
+#[derive(Clone, Copy, Default)]
+pub struct Rsdp {
+    pub signature: [u8; 8],
+    pub checksum: u8,
+    pub oem_id: [u8; 6],
+    pub revision: u8,
+    _rsdt_addr: u32,
+    pub length: u32,
+    pub xsdt_addr: u64,
+    pub extended_checksum: u8,
+    _reserved: [u8; 3],
+}
+
+// SAFETY: Rsdp only contains a series of integers
+unsafe impl ByteValued for Rsdp {}
+
+impl Rsdp {
+    pub fn new(oem_id: [u8; 6], xsdt_addr: u64) -> Self {
+        let mut rsdp = Rsdp {
+            signature: *b"RSD PTR ",
+            checksum: 0,
+            oem_id,
+            revision: 2,
+            _rsdt_addr: 0,
+            length: std::mem::size_of::<Rsdp>() as u32,
+            xsdt_addr,
+            extended_checksum: 0,
+            _reserved: [0; 3],
+        };
+
+        rsdp.checksum = super::generate_checksum(&rsdp.as_slice()[0..19]);
+        rsdp.extended_checksum = super::generate_checksum(rsdp.as_slice());
+        rsdp
+    }
+
+    pub fn len() -> usize {
+        std::mem::size_of::<Rsdp>()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::Rsdp;
+    use vm_memory::bytes::ByteValued;
+
+    #[test]
+    fn test_rsdp() {
+        let rsdp = Rsdp::new(*b"CHYPER", 0xdead_beef);
+        let sum = rsdp
+            .as_slice()
+            .iter()
+            .fold(0u8, |acc, x| acc.wrapping_add(*x));
+        assert_eq!(sum, 0);
+        let sum: u8 = rsdp
+            .as_slice()
+            .iter()
+            .fold(0u8, |acc, x| acc.wrapping_add(*x));
+        assert_eq!(sum, 0);
+    }
+}

acpi_tables/src/sdt.rs

@@ -0,0 +1,146 @@
+// Copyright © 2019 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#[repr(packed)]
+pub struct GenericAddress {
+    pub address_space_id: u8,
+    pub register_bit_width: u8,
+    pub register_bit_offset: u8,
+    pub access_size: u8,
+    pub address: u64,
+}
+
+impl GenericAddress {
+    pub fn io_port_address<T>(address: u16) -> Self {
+        GenericAddress {
+            address_space_id: 1,
+            register_bit_width: 8 * std::mem::size_of::<T>() as u8,
+            register_bit_offset: 0,
+            access_size: std::mem::size_of::<T>() as u8,
+            address: u64::from(address),
+        }
+    }
+    pub fn mmio_address<T>(address: u64) -> Self {
+        GenericAddress {
+            address_space_id: 0,
+            register_bit_width: 8 * std::mem::size_of::<T>() as u8,
+            register_bit_offset: 0,
+            access_size: std::mem::size_of::<T>() as u8,
+            address,
+        }
+    }
+}
+
+pub struct Sdt {
+    data: Vec<u8>,
+}
+
+#[allow(clippy::len_without_is_empty)]
+impl Sdt {
+    pub fn new(
+        signature: [u8; 4],
+        length: u32,
+        revision: u8,
+        oem_id: [u8; 6],
+        oem_table: [u8; 8],
+        oem_revision: u32,
+    ) -> Self {
+        assert!(length >= 36);
+
+        let mut data = Vec::with_capacity(length as usize);
+        data.extend_from_slice(&signature);
+        data.extend_from_slice(&length.to_le_bytes());
+        data.push(revision);
+        data.push(0); // checksum
+        data.extend_from_slice(&oem_id);
+        data.extend_from_slice(&oem_table);
+        data.extend_from_slice(&oem_revision.to_le_bytes());
+        data.extend_from_slice(b"CLDH");
+        data.extend_from_slice(&0u32.to_le_bytes());
+        assert_eq!(data.len(), 36);
+
+        data.resize(length as usize, 0);
+        let mut sdt = Sdt { data };
+
+        sdt.update_checksum();
+        sdt
+    }
+
+    pub fn update_checksum(&mut self) {
+        self.data[9] = 0;
+        let checksum = super::generate_checksum(self.data.as_slice());
+        self.data[9] = checksum
+    }
+
+    pub fn as_slice(&self) -> &[u8] {
+        self.data.as_slice()
+    }
+
+    pub fn append<T>(&mut self, value: T) {
+        let orig_length = self.data.len();
+        let new_length = orig_length + std::mem::size_of::<T>();
+        self.data.resize(new_length, 0);
+        self.write_u32(4, new_length as u32);
+        self.write(orig_length, value);
+    }
+
+    pub fn append_slice(&mut self, data: &[u8]) {
+        let orig_length = self.data.len();
+        let new_length = orig_length + data.len();
+        self.write_u32(4, new_length as u32);
+        self.data.extend_from_slice(data);
+        self.update_checksum();
+    }
+
+    /// Write a value at the given offset
+    pub fn write<T>(&mut self, offset: usize, value: T) {
+        assert!((offset + (std::mem::size_of::<T>() - 1)) < self.data.len());
+        unsafe {
+            *(((self.data.as_mut_ptr() as usize) + offset) as *mut T) = value;
+        }
+        self.update_checksum();
+    }
+
+    pub fn write_u8(&mut self, offset: usize, val: u8) {
+        self.write(offset, val);
+    }
+
+    pub fn write_u16(&mut self, offset: usize, val: u16) {
+        self.write(offset, val);
+    }
+
+    pub fn write_u32(&mut self, offset: usize, val: u32) {
+        self.write(offset, val);
+    }
+
+    pub fn write_u64(&mut self, offset: usize, val: u64) {
+        self.write(offset, val);
+    }
+
+    pub fn len(&self) -> usize {
+        self.data.len()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::Sdt;
+
+    #[test]
+    fn test_sdt() {
+        let mut sdt = Sdt::new(*b"TEST", 40, 1, *b"CLOUDH", *b"TESTTEST", 1);
+        let sum: u8 = sdt
+            .as_slice()
+            .iter()
+            .fold(0u8, |acc, x| acc.wrapping_add(*x));
+        assert_eq!(sum, 0);
+        sdt.write_u32(36, 0x12345678);
+        let sum: u8 = sdt
+            .as_slice()
+            .iter()
+            .fold(0u8, |acc, x| acc.wrapping_add(*x));
+        assert_eq!(sum, 0);
+    }
+}

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 = "0.9.0"

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,
@@ -74,7 +79,7 @@ impl StatusCode {
 }
 
 fn get_header<'a>(res: &'a str, header: &'a str) -> Option<&'a str> {
-    let header_str = format!("{header}: ");
+    let header_str = format!("{}: ", header);
     res.find(&header_str)
         .map(|o| &res[o + header_str.len()..o + res[o..].find('\r').unwrap()])
 }
@@ -96,10 +101,6 @@ fn parse_http_response(socket: &mut dyn Read) -> Result<Option<String>, Error> {
     loop {
         let mut bytes = vec![0; 256];
         let count = socket.read(&mut bytes).map_err(Error::Socket)?;
-        // If the return value is 0, the peer has performed an orderly shutdown.
-        if count == 0 {
-            break;
-        }
         res.push_str(std::str::from_utf8(&bytes[0..count]).unwrap());
 
         // End of headers
@@ -118,14 +119,15 @@ 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..])));
+    let body_string = content_length.and(Some(String::from(&res[body_offset.unwrap()..])));
     let status_code = get_status_code(&res)?;
 
     if status_code.is_server_error() {
@@ -135,22 +137,21 @@ fn parse_http_response(socket: &mut dyn Read) -> Result<Option<String>, Error> {
     }
 }
 
-/// Make an API request using the fully qualified command name.
-/// For example, full_command could be "vm.create" or "vmm.ping".
-pub fn simple_api_full_command_with_fds_and_response<T: Read + Write + ScmSocket>(
+pub fn simple_api_command_with_fds<T: Read + Write + ScmSocket>(
     socket: &mut T,
     method: &str,
-    full_command: &str,
+    c: &str,
     request_body: Option<&str>,
-    request_fds: &[RawFd],
-) -> Result<Option<String>, Error> {
+    request_fds: Vec<RawFd>,
+) -> Result<(), Error> {
     socket
         .send_with_fds(
             &[format!(
-                "{method} /api/v1/{full_command} HTTP/1.1\r\nHost: localhost\r\nAccept: */*\r\n"
+                "{} /api/v1/vm.{} HTTP/1.1\r\nHost: localhost\r\nAccept: */*\r\n",
+                method, c
             )
             .as_bytes()],
-            request_fds,
+            &request_fds,
         )
         .map_err(Error::SocketSendFds)?;
 
@@ -170,68 +171,17 @@ pub fn simple_api_full_command_with_fds_and_response<T: Read + Write + ScmSocket
 
     socket.flush().map_err(Error::Socket)?;
 
-    parse_http_response(socket)
-}
-
-pub fn simple_api_full_command_with_fds<T: Read + Write + ScmSocket>(
-    socket: &mut T,
-    method: &str,
-    full_command: &str,
-    request_body: Option<&str>,
-    request_fds: &[RawFd],
-) -> Result<(), Error> {
-    let response = simple_api_full_command_with_fds_and_response(
-        socket,
-        method,
-        full_command,
-        request_body,
-        request_fds,
-    )?;
-
-    if let Some(response) = response {
-        println!("{response}");
+    if let Some(body) = parse_http_response(socket)? {
+        println!("{}", body);
     }
-
     Ok(())
 }
 
-pub fn simple_api_full_command<T: Read + Write + ScmSocket>(
-    socket: &mut T,
-    method: &str,
-    full_command: &str,
-    request_body: Option<&str>,
-) -> Result<(), Error> {
-    simple_api_full_command_with_fds(socket, method, full_command, request_body, &[])
-}
-
-pub fn simple_api_full_command_and_response<T: Read + Write + ScmSocket>(
-    socket: &mut T,
-    method: &str,
-    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, &[])
-}
-
-pub fn simple_api_command_with_fds<T: Read + Write + ScmSocket>(
-    socket: &mut T,
-    method: &str,
-    c: &str,
-    request_body: Option<&str>,
-    request_fds: &[RawFd],
-) -> Result<(), Error> {
-    // Create the full VM command. For VMM commands, use
-    // simple_api_full_command().
-    let full_command = format!("vm.{c}");
-
-    simple_api_full_command_with_fds(socket, method, &full_command, request_body, request_fds)
-}
-
 pub fn simple_api_command<T: Read + Write + ScmSocket>(
     socket: &mut T,
     method: &str,
     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 +1,30 @@
 [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 }
+acpi_tables = { path = "../acpi_tables" }
+anyhow = "1.0.56"
+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 = "0.2.123"
+linux-loader = { version = "0.4.0", features = ["elf", "bzimage", "pe"] }
+log = "0.4.16"
+serde = { version = "1.0.136", features = ["rc"] }
+serde_derive = "1.0.136"
+thiserror = "1.0.30"
+versionize = "0.1.6"
+versionize_derive = "0.1.4"
+vm-memory = { version = "0.7.0", features = ["backend-mmap", "backend-bitmap"] }
+vm-migration = { path = "../vm-migration" }
+vmm-sys-util = { version = "0.9.0", 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 = "0.1.3", package = 'fdt'}
+vm-fdt = { git = "https://github.com/rust-vmm/vm-fdt", branch = "main" }

arch/src/aarch64/gic/dist_regs.rs

@@ -1,17 +1,16 @@
 // Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-use kvm_ioctls::DeviceFd;
-
-use crate::arch::aarch64::gic::{Error, Result};
-use crate::device::HypervisorDeviceError;
-use crate::kvm::kvm_bindings::{
-    KVM_DEV_ARM_VGIC_GRP_DIST_REGS, KVM_DEV_ARM_VGIC_GRP_NR_IRQS, kvm_device_attr,
+use super::{Error, Result};
+use crate::layout::IRQ_BASE;
+use hypervisor::kvm::kvm_bindings::{
+    kvm_device_attr, KVM_DEV_ARM_VGIC_GRP_DIST_REGS, KVM_DEV_ARM_VGIC_GRP_NR_IRQS,
 };
+use std::sync::Arc;
 
 /*
  Distributor registers as detailed at page 456 from
- https://developer.arm.com/documentation/ihi0069/c/?lang=en.
+ https://static.docs.arm.com/ihi0069/c/IHI0069C_gic_architecture_specification.pdf.
  Address offsets are relative to the Distributor base address defined
 by the system memory map. Unless otherwise stated in the register description,
 all GIC registers are 32-bits wide.
@@ -78,71 +77,55 @@ static VGIC_DIST_REGS: &[DistReg] = &[
     VGIC_DIST_REG!(GICD_IPRIORITYR, 8, 0),
 ];
 
-fn dist_attr_set(gic: &DeviceFd, offset: u32, val: u32) -> Result<()> {
-    let gic_dist_attr = kvm_device_attr {
-        group: KVM_DEV_ARM_VGIC_GRP_DIST_REGS,
-        attr: offset as u64,
-        addr: &raw const val as u64,
-        flags: 0,
-    };
-
-    gic.set_device_attr(&gic_dist_attr).map_err(|e| {
-        Error::SetDeviceAttribute(HypervisorDeviceError::SetDeviceAttribute(e.into()))
-    })?;
-
-    Ok(())
-}
-
-fn dist_attr_get(gic: &DeviceFd, offset: u32) -> Result<u32> {
-    let mut val = 0;
-
+fn dist_attr_access(
+    gic: &Arc<dyn hypervisor::Device>,
+    offset: u32,
+    val: &u32,
+    set: bool,
+) -> Result<()> {
     let mut gic_dist_attr = kvm_device_attr {
         group: KVM_DEV_ARM_VGIC_GRP_DIST_REGS,
         attr: offset as u64,
-        addr: &raw mut val as u64,
+        addr: val as *const u32 as u64,
         flags: 0,
     };
-
-    // SAFETY: gic_dist_attr.addr is safe to write to.
-    unsafe { gic.get_device_attr(&mut gic_dist_attr) }.map_err(|e| {
-        Error::GetDeviceAttribute(HypervisorDeviceError::GetDeviceAttribute(e.into()))
-    })?;
-
-    Ok(val)
+    if set {
+        gic.set_device_attr(&gic_dist_attr)
+            .map_err(Error::SetDeviceAttribute)?;
+    } else {
+        gic.get_device_attr(&mut gic_dist_attr)
+            .map_err(Error::GetDeviceAttribute)?;
+    }
+    Ok(())
 }
 
 /// Get the distributor control register.
-pub fn read_ctlr(gic: &DeviceFd) -> Result<u32> {
-    dist_attr_get(gic, GICD_CTLR)
+pub fn read_ctlr(gic: &Arc<dyn hypervisor::Device>) -> Result<u32> {
+    let val: u32 = 0;
+    dist_attr_access(gic, GICD_CTLR, &val, false)?;
+    Ok(val)
 }
 
 /// Set the distributor control register.
-pub fn write_ctlr(gic: &DeviceFd, val: u32) -> Result<()> {
-    dist_attr_set(gic, GICD_CTLR, val)
+pub fn write_ctlr(gic: &Arc<dyn hypervisor::Device>, val: u32) -> Result<()> {
+    dist_attr_access(gic, GICD_CTLR, &val, true)
 }
 
-fn get_interrupts_num(gic: &DeviceFd) -> Result<u32> {
-    let mut num_irq = 0;
+fn get_interrupts_num(gic: &Arc<dyn hypervisor::Device>) -> Result<u32> {
+    let num_irq = 0;
 
     let mut nr_irqs_attr = kvm_device_attr {
         group: KVM_DEV_ARM_VGIC_GRP_NR_IRQS,
         attr: 0,
-        addr: &raw mut num_irq as u64,
+        addr: &num_irq as *const u32 as u64,
         flags: 0,
     };
-    // SAFETY: nr_irqs_attr.addr is safe to write to.
-    unsafe { gic.get_device_attr(&mut nr_irqs_attr) }.map_err(|e| {
-        Error::GetDeviceAttribute(HypervisorDeviceError::GetDeviceAttribute(e.into()))
-    })?;
+    gic.get_device_attr(&mut nr_irqs_attr)
+        .map_err(Error::GetDeviceAttribute)?;
     Ok(num_irq)
 }
 
-fn compute_reg_len(gic: &DeviceFd, reg: &DistReg, base: u32) -> Result<u32> {
-    // FIXME:
-    // Redefine some GIC constants to avoid the dependency on `layout` crate.
-    // This is temporary solution, will be fixed in future refactoring.
-    const LAYOUT_IRQ_BASE: u32 = 32;
-
+fn compute_reg_len(gic: &Arc<dyn hypervisor::Device>, reg: &DistReg, base: u32) -> Result<u32> {
     let mut end = base;
     let num_irq = get_interrupts_num(gic)?;
     if reg.length > 0 {
@@ -155,8 +138,8 @@ fn compute_reg_len(gic: &DeviceFd, reg: &DistReg, base: u32) -> Result<u32> {
         // This is the type of register that takes into account the number of interrupts
         // that the model has. It is also the type of register where
         // a register relates to multiple interrupts.
-        end = base + (reg.bpi as u32 * (num_irq - LAYOUT_IRQ_BASE) / 8);
-        if !(reg.bpi as u32 * (num_irq - LAYOUT_IRQ_BASE)).is_multiple_of(8) {
+        end = base + (reg.bpi as u32 * (num_irq - IRQ_BASE) / 8);
+        if reg.bpi as u32 * (num_irq - IRQ_BASE) % 8 > 0 {
             end += REG_SIZE as u32;
         }
     }
@@ -164,7 +147,7 @@ fn compute_reg_len(gic: &DeviceFd, reg: &DistReg, base: u32) -> Result<u32> {
 }
 
 /// Set distributor registers of the GIC.
-pub fn set_dist_regs(gic: &DeviceFd, state: &[u32]) -> Result<()> {
+pub fn set_dist_regs(gic: &Arc<dyn hypervisor::Device>, state: &[u32]) -> Result<()> {
     let mut idx = 0;
 
     for dreg in VGIC_DIST_REGS {
@@ -172,7 +155,8 @@ pub fn set_dist_regs(gic: &DeviceFd, state: &[u32]) -> Result<()> {
         let end = compute_reg_len(gic, dreg, base)?;
 
         while base < end {
-            dist_attr_set(gic, base, state[idx])?;
+            let val = state[idx];
+            dist_attr_access(gic, base, &val, true)?;
             idx += 1;
             base += REG_SIZE as u32;
         }
@@ -180,7 +164,7 @@ pub fn set_dist_regs(gic: &DeviceFd, state: &[u32]) -> Result<()> {
     Ok(())
 }
 /// Get distributor registers of the GIC.
-pub fn get_dist_regs(gic: &DeviceFd) -> Result<Vec<u32>> {
+pub fn get_dist_regs(gic: &Arc<dyn hypervisor::Device>) -> Result<Vec<u32>> {
     let mut state = Vec::new();
 
     for dreg in VGIC_DIST_REGS {
@@ -188,7 +172,9 @@ pub fn get_dist_regs(gic: &DeviceFd) -> Result<Vec<u32>> {
         let end = compute_reg_len(gic, dreg, base)?;
 
         while base < end {
-            state.push(dist_attr_get(gic, base)?);
+            let val: u32 = 0;
+            dist_attr_access(gic, base, &val, false)?;
+            state.push(val);
             base += REG_SIZE as u32;
         }
     }

arch/src/aarch64/gic/gicv3.rs

@@ -0,0 +1,262 @@
+// Copyright 2021 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
+//
+// This file implements the GicV3 device.
+
+pub mod kvm {
+    use crate::aarch64::gic::dist_regs::{get_dist_regs, read_ctlr, set_dist_regs, write_ctlr};
+    use crate::aarch64::gic::icc_regs::{get_icc_regs, set_icc_regs};
+    use crate::aarch64::gic::kvm::{save_pending_tables, KvmGicDevice};
+    use crate::aarch64::gic::redist_regs::{
+        construct_gicr_typers, get_redist_regs, set_redist_regs,
+    };
+    use crate::aarch64::gic::GicDevice;
+    use crate::layout;
+    use anyhow::anyhow;
+    use hypervisor::kvm::kvm_bindings;
+    use hypervisor::CpuState;
+    use std::any::Any;
+    use std::convert::TryInto;
+    use std::sync::Arc;
+    use std::{boxed::Box, result};
+    use versionize::{VersionMap, Versionize, VersionizeResult};
+    use versionize_derive::Versionize;
+    use vm_memory::Address;
+    use vm_migration::{
+        Migratable, MigratableError, Pausable, Snapshot, Snapshottable, Transportable,
+        VersionMapped,
+    };
+
+    /// Errors thrown while saving/restoring the GICv3.
+    #[derive(Debug)]
+    pub enum Error {
+        /// Error in saving RDIST pending tables into guest RAM.
+        SavePendingTables(crate::aarch64::gic::Error),
+        /// Error in saving GIC distributor registers.
+        SaveDistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC distributor registers.
+        RestoreDistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC distributor control registers.
+        SaveDistributorCtrlRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC distributor control registers.
+        RestoreDistributorCtrlRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC redistributor registers.
+        SaveRedistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC redistributor registers.
+        RestoreRedistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC CPU interface registers.
+        SaveIccRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC CPU interface registers.
+        RestoreIccRegisters(crate::aarch64::gic::Error),
+    }
+
+    type Result<T> = result::Result<T, Error>;
+
+    pub struct KvmGicV3 {
+        /// The hypervisor agnostic device for the GicV3
+        device: Arc<dyn hypervisor::Device>,
+
+        /// Vector holding values of GICR_TYPER for each vCPU
+        gicr_typers: Vec<u64>,
+
+        /// GIC device properties, to be used for setting up the fdt entry
+        properties: [u64; 4],
+
+        /// Number of CPUs handled by the device
+        vcpu_count: u64,
+    }
+
+    #[derive(Versionize)]
+    pub struct Gicv3State {
+        dist: Vec<u32>,
+        rdist: Vec<u32>,
+        icc: Vec<u32>,
+        // special register that enables interrupts and affinity routing
+        gicd_ctlr: u32,
+    }
+
+    impl VersionMapped for Gicv3State {}
+
+    impl KvmGicV3 {
+        // Device trees specific constants
+        pub const ARCH_GIC_V3_MAINT_IRQ: u32 = 9;
+
+        /// Get the address of the GIC distributor.
+        pub fn get_dist_addr() -> u64 {
+            layout::GIC_V3_DIST_START.raw_value()
+        }
+
+        /// Get the size of the GIC distributor.
+        pub fn get_dist_size() -> u64 {
+            layout::GIC_V3_DIST_SIZE
+        }
+
+        /// Get the address of the GIC redistributors.
+        pub fn get_redists_addr(vcpu_count: u64) -> u64 {
+            KvmGicV3::get_dist_addr() - KvmGicV3::get_redists_size(vcpu_count)
+        }
+
+        /// Get the size of the GIC redistributors.
+        pub fn get_redists_size(vcpu_count: u64) -> u64 {
+            vcpu_count * layout::GIC_V3_REDIST_SIZE
+        }
+
+        /// Save the state of GIC.
+        fn state(&self, gicr_typers: &[u64]) -> Result<Gicv3State> {
+            let gicd_ctlr =
+                read_ctlr(self.device()).map_err(Error::SaveDistributorCtrlRegisters)?;
+
+            let dist_state =
+                get_dist_regs(self.device()).map_err(Error::SaveDistributorRegisters)?;
+
+            let rdist_state = get_redist_regs(self.device(), gicr_typers)
+                .map_err(Error::SaveRedistributorRegisters)?;
+
+            let icc_state =
+                get_icc_regs(self.device(), gicr_typers).map_err(Error::SaveIccRegisters)?;
+
+            Ok(Gicv3State {
+                dist: dist_state,
+                rdist: rdist_state,
+                icc: icc_state,
+                gicd_ctlr,
+            })
+        }
+
+        /// Restore the state of GIC.
+        fn set_state(&mut self, gicr_typers: &[u64], state: &Gicv3State) -> Result<()> {
+            write_ctlr(self.device(), state.gicd_ctlr)
+                .map_err(Error::RestoreDistributorCtrlRegisters)?;
+
+            set_dist_regs(self.device(), &state.dist)
+                .map_err(Error::RestoreDistributorRegisters)?;
+
+            set_redist_regs(self.device(), gicr_typers, &state.rdist)
+                .map_err(Error::RestoreRedistributorRegisters)?;
+
+            set_icc_regs(self.device(), gicr_typers, &state.icc)
+                .map_err(Error::RestoreIccRegisters)?;
+
+            Ok(())
+        }
+    }
+
+    impl GicDevice for KvmGicV3 {
+        fn device(&self) -> &Arc<dyn hypervisor::Device> {
+            &self.device
+        }
+
+        fn fdt_compatibility(&self) -> &str {
+            "arm,gic-v3"
+        }
+
+        fn fdt_maint_irq(&self) -> u32 {
+            KvmGicV3::ARCH_GIC_V3_MAINT_IRQ
+        }
+
+        fn device_properties(&self) -> &[u64] {
+            &self.properties
+        }
+
+        fn vcpu_count(&self) -> u64 {
+            self.vcpu_count
+        }
+
+        fn set_its_device(&mut self, _its_device: Option<Arc<dyn hypervisor::Device>>) {}
+
+        fn set_gicr_typers(&mut self, vcpu_states: &[CpuState]) {
+            let gicr_typers = construct_gicr_typers(vcpu_states);
+            self.gicr_typers = gicr_typers;
+        }
+
+        fn as_any_concrete_mut(&mut self) -> &mut dyn Any {
+            self
+        }
+    }
+
+    impl KvmGicDevice for KvmGicV3 {
+        fn version() -> u32 {
+            kvm_bindings::kvm_device_type_KVM_DEV_TYPE_ARM_VGIC_V3
+        }
+
+        fn create_device(
+            device: Arc<dyn hypervisor::Device>,
+            vcpu_count: u64,
+        ) -> Box<dyn GicDevice> {
+            Box::new(KvmGicV3 {
+                device,
+                gicr_typers: vec![0; vcpu_count.try_into().unwrap()],
+                properties: [
+                    KvmGicV3::get_dist_addr(),
+                    KvmGicV3::get_dist_size(),
+                    KvmGicV3::get_redists_addr(vcpu_count),
+                    KvmGicV3::get_redists_size(vcpu_count),
+                ],
+                vcpu_count,
+            })
+        }
+
+        fn init_device_attributes(
+            _vm: &Arc<dyn hypervisor::Vm>,
+            gic_device: &mut dyn GicDevice,
+        ) -> crate::aarch64::gic::Result<()> {
+            /* Setting up the distributor attribute.
+             We are placing the GIC below 1GB so we need to substract the size of the distributor.
+            */
+            Self::set_device_attribute(
+                gic_device.device(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ADDR,
+                u64::from(kvm_bindings::KVM_VGIC_V3_ADDR_TYPE_DIST),
+                &KvmGicV3::get_dist_addr() as *const u64 as u64,
+                0,
+            )?;
+
+            /* Setting up the redistributors' attribute.
+            We are calculating here the start of the redistributors address. We have one per CPU.
+            */
+            Self::set_device_attribute(
+                gic_device.device(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ADDR,
+                u64::from(kvm_bindings::KVM_VGIC_V3_ADDR_TYPE_REDIST),
+                &KvmGicV3::get_redists_addr(gic_device.vcpu_count()) as *const u64 as u64,
+                0,
+            )?;
+
+            Ok(())
+        }
+    }
+
+    pub const GIC_V3_SNAPSHOT_ID: &str = "gic-v3";
+    impl Snapshottable for KvmGicV3 {
+        fn id(&self) -> String {
+            GIC_V3_SNAPSHOT_ID.to_string()
+        }
+
+        fn snapshot(&mut self) -> std::result::Result<Snapshot, MigratableError> {
+            let gicr_typers = self.gicr_typers.clone();
+            Snapshot::new_from_versioned_state(&self.id(), &self.state(&gicr_typers).unwrap())
+        }
+
+        fn restore(&mut self, snapshot: Snapshot) -> std::result::Result<(), MigratableError> {
+            let gicr_typers = self.gicr_typers.clone();
+            self.set_state(&gicr_typers, &snapshot.to_versioned_state(&self.id())?)
+                .map_err(|e| {
+                    MigratableError::Restore(anyhow!("Could not restore GICv3 state {:?}", e))
+                })
+        }
+    }
+
+    impl Pausable for KvmGicV3 {
+        fn pause(&mut self) -> std::result::Result<(), MigratableError> {
+            // Flush redistributors pending tables to guest RAM.
+            save_pending_tables(self.device()).map_err(|e| {
+                MigratableError::Pause(anyhow!("Could not save GICv3 GIC pending tables {:?}", e))
+            })?;
+
+            Ok(())
+        }
+    }
+    impl Transportable for KvmGicV3 {}
+    impl Migratable for KvmGicV3 {}
+}

arch/src/aarch64/gic/gicv3_its.rs

@@ -0,0 +1,514 @@
+// Copyright 2021 Arm Limited (or its affiliates). All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+//
+// This file implements the GicV3 device with ITS (Virtual Interrupt Translation Service).
+
+pub mod kvm {
+    use crate::aarch64::gic::dist_regs::{get_dist_regs, read_ctlr, set_dist_regs, write_ctlr};
+    use crate::aarch64::gic::icc_regs::{get_icc_regs, set_icc_regs};
+    use crate::aarch64::gic::redist_regs::{
+        construct_gicr_typers, get_redist_regs, set_redist_regs,
+    };
+
+    use crate::aarch64::gic::gicv3::kvm::KvmGicV3;
+    use crate::aarch64::gic::kvm::{save_pending_tables, KvmGicDevice};
+    use crate::aarch64::gic::GicDevice;
+    use crate::layout;
+    use anyhow::anyhow;
+    use hypervisor::kvm::kvm_bindings;
+    use hypervisor::CpuState;
+    use std::any::Any;
+    use std::convert::TryInto;
+    use std::sync::Arc;
+    use std::{boxed::Box, result};
+    use versionize::{VersionMap, Versionize, VersionizeResult};
+    use versionize_derive::Versionize;
+    use vm_migration::{
+        Migratable, MigratableError, Pausable, Snapshot, Snapshottable, Transportable,
+        VersionMapped,
+    };
+
+    const GITS_CTLR: u32 = 0x0000;
+    const GITS_IIDR: u32 = 0x0004;
+    const GITS_CBASER: u32 = 0x0080;
+    const GITS_CWRITER: u32 = 0x0088;
+    const GITS_CREADR: u32 = 0x0090;
+    const GITS_BASER: u32 = 0x0100;
+
+    /// Errors thrown while saving/restoring the GICv3ITS.
+    #[derive(Debug)]
+    pub enum Error {
+        /// Error in saving RDIST pending tables into guest RAM.
+        SavePendingTables(crate::aarch64::gic::Error),
+        /// Error in saving GIC distributor registers.
+        SaveDistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC distributor registers.
+        RestoreDistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC distributor control registers.
+        SaveDistributorCtrlRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC distributor control registers.
+        RestoreDistributorCtrlRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC redistributor registers.
+        SaveRedistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC redistributor registers.
+        RestoreRedistributorRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GIC CPU interface registers.
+        SaveIccRegisters(crate::aarch64::gic::Error),
+        /// Error in restoring GIC CPU interface registers.
+        RestoreIccRegisters(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS IIDR register.
+        SaveITSIIDR(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS IIDR register.
+        RestoreITSIIDR(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS CBASER register.
+        SaveITSCBASER(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS CBASER register.
+        RestoreITSCBASER(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS CREADR register.
+        SaveITSCREADR(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS CREADR register.
+        RestoreITSCREADR(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS CWRITER register.
+        SaveITSCWRITER(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS CWRITER register.
+        RestoreITSCWRITER(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS BASER register.
+        SaveITSBASER(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS BASER register.
+        RestoreITSBASER(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS CTLR register.
+        SaveITSCTLR(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS CTLR register.
+        RestoreITSCTLR(crate::aarch64::gic::Error),
+        /// Error in saving GICv3ITS restore tables.
+        SaveITSTables(crate::aarch64::gic::Error),
+        /// Error in restoring GICv3ITS restore tables.
+        RestoreITSTables(crate::aarch64::gic::Error),
+    }
+
+    type Result<T> = result::Result<T, Error>;
+
+    /// Access an ITS device attribute.
+    ///
+    /// This is a helper function to get/set the ITS device attribute depending
+    /// the bool parameter `set` provided.
+    pub fn gicv3_its_attr_access(
+        its_device: &Arc<dyn hypervisor::Device>,
+        group: u32,
+        attr: u32,
+        val: &u64,
+        set: bool,
+    ) -> crate::aarch64::gic::Result<()> {
+        let mut gicv3_its_attr = kvm_bindings::kvm_device_attr {
+            group,
+            attr: attr as u64,
+            addr: val as *const u64 as u64,
+            flags: 0,
+        };
+        if set {
+            its_device
+                .set_device_attr(&gicv3_its_attr)
+                .map_err(crate::aarch64::gic::Error::SetDeviceAttribute)?;
+        } else {
+            its_device
+                .get_device_attr(&mut gicv3_its_attr)
+                .map_err(crate::aarch64::gic::Error::GetDeviceAttribute)?;
+        }
+        Ok(())
+    }
+
+    /// Function that saves/restores ITS tables into guest RAM.
+    ///
+    /// The tables get flushed to guest RAM whenever the VM gets stopped.
+    pub fn gicv3_its_tables_access(
+        its_device: &Arc<dyn hypervisor::Device>,
+        save: bool,
+    ) -> crate::aarch64::gic::Result<()> {
+        let attr = if save {
+            u64::from(kvm_bindings::KVM_DEV_ARM_ITS_SAVE_TABLES)
+        } else {
+            u64::from(kvm_bindings::KVM_DEV_ARM_ITS_RESTORE_TABLES)
+        };
+
+        let init_gic_attr = kvm_bindings::kvm_device_attr {
+            group: kvm_bindings::KVM_DEV_ARM_VGIC_GRP_CTRL,
+            attr,
+            addr: 0,
+            flags: 0,
+        };
+        its_device
+            .set_device_attr(&init_gic_attr)
+            .map_err(crate::aarch64::gic::Error::SetDeviceAttribute)
+    }
+
+    pub struct KvmGicV3Its {
+        /// The hypervisor agnostic device for the GicV3
+        device: Arc<dyn hypervisor::Device>,
+
+        /// The hypervisor agnostic device for the Its device
+        its_device: Option<Arc<dyn hypervisor::Device>>,
+
+        /// Vector holding values of GICR_TYPER for each vCPU
+        gicr_typers: Vec<u64>,
+
+        /// GIC device properties, to be used for setting up the fdt entry
+        gic_properties: [u64; 4],
+
+        /// MSI device properties, to be used for setting up the fdt entry
+        msi_properties: [u64; 2],
+
+        /// Number of CPUs handled by the device
+        vcpu_count: u64,
+    }
+
+    #[derive(Versionize)]
+    pub struct Gicv3ItsState {
+        dist: Vec<u32>,
+        rdist: Vec<u32>,
+        icc: Vec<u32>,
+        // special register that enables interrupts and affinity routing
+        gicd_ctlr: u32,
+        its_ctlr: u64,
+        its_iidr: u64,
+        its_cbaser: u64,
+        its_cwriter: u64,
+        its_creadr: u64,
+        its_baser: [u64; 8],
+    }
+
+    impl VersionMapped for Gicv3ItsState {}
+
+    impl KvmGicV3Its {
+        fn get_msi_size() -> u64 {
+            layout::GIC_V3_ITS_SIZE
+        }
+
+        fn get_msi_addr(vcpu_count: u64) -> u64 {
+            KvmGicV3::get_redists_addr(vcpu_count) - KvmGicV3Its::get_msi_size()
+        }
+
+        /// Save the state of GICv3ITS.
+        fn state(&self, gicr_typers: &[u64]) -> Result<Gicv3ItsState> {
+            let gicd_ctlr =
+                read_ctlr(self.device()).map_err(Error::SaveDistributorCtrlRegisters)?;
+
+            let dist_state =
+                get_dist_regs(self.device()).map_err(Error::SaveDistributorRegisters)?;
+
+            let rdist_state = get_redist_regs(self.device(), gicr_typers)
+                .map_err(Error::SaveRedistributorRegisters)?;
+
+            let icc_state =
+                get_icc_regs(self.device(), gicr_typers).map_err(Error::SaveIccRegisters)?;
+
+            let its_baser_state: [u64; 8] = [0; 8];
+            for i in 0..8 {
+                gicv3_its_attr_access(
+                    self.its_device().unwrap(),
+                    kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                    GITS_BASER + i * 8,
+                    &its_baser_state[i as usize],
+                    false,
+                )
+                .map_err(Error::SaveITSBASER)?;
+            }
+
+            let its_ctlr_state: u64 = 0;
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CTLR,
+                &its_ctlr_state,
+                false,
+            )
+            .map_err(Error::SaveITSCTLR)?;
+
+            let its_cbaser_state: u64 = 0;
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CBASER,
+                &its_cbaser_state,
+                false,
+            )
+            .map_err(Error::SaveITSCBASER)?;
+
+            let its_creadr_state: u64 = 0;
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CREADR,
+                &its_creadr_state,
+                false,
+            )
+            .map_err(Error::SaveITSCREADR)?;
+
+            let its_cwriter_state: u64 = 0;
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CWRITER,
+                &its_cwriter_state,
+                false,
+            )
+            .map_err(Error::SaveITSCWRITER)?;
+
+            let its_iidr_state: u64 = 0;
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_IIDR,
+                &its_iidr_state,
+                false,
+            )
+            .map_err(Error::SaveITSIIDR)?;
+
+            Ok(Gicv3ItsState {
+                dist: dist_state,
+                rdist: rdist_state,
+                icc: icc_state,
+                gicd_ctlr,
+                its_ctlr: its_ctlr_state,
+                its_iidr: its_iidr_state,
+                its_cbaser: its_cbaser_state,
+                its_cwriter: its_cwriter_state,
+                its_creadr: its_creadr_state,
+                its_baser: its_baser_state,
+            })
+        }
+
+        /// Restore the state of GICv3ITS.
+        fn set_state(&mut self, gicr_typers: &[u64], state: &Gicv3ItsState) -> Result<()> {
+            write_ctlr(self.device(), state.gicd_ctlr)
+                .map_err(Error::RestoreDistributorCtrlRegisters)?;
+
+            set_dist_regs(self.device(), &state.dist)
+                .map_err(Error::RestoreDistributorRegisters)?;
+
+            set_redist_regs(self.device(), gicr_typers, &state.rdist)
+                .map_err(Error::RestoreRedistributorRegisters)?;
+
+            set_icc_regs(self.device(), gicr_typers, &state.icc)
+                .map_err(Error::RestoreIccRegisters)?;
+
+            //Restore GICv3ITS registers
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_IIDR,
+                &state.its_iidr,
+                true,
+            )
+            .map_err(Error::RestoreITSIIDR)?;
+
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CBASER,
+                &state.its_cbaser,
+                true,
+            )
+            .map_err(Error::RestoreITSCBASER)?;
+
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CREADR,
+                &state.its_creadr,
+                true,
+            )
+            .map_err(Error::RestoreITSCREADR)?;
+
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CWRITER,
+                &state.its_cwriter,
+                true,
+            )
+            .map_err(Error::RestoreITSCWRITER)?;
+
+            for i in 0..8 {
+                gicv3_its_attr_access(
+                    self.its_device().unwrap(),
+                    kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                    GITS_BASER + i * 8,
+                    &state.its_baser[i as usize],
+                    true,
+                )
+                .map_err(Error::RestoreITSBASER)?;
+            }
+
+            // Restore ITS tables
+            gicv3_its_tables_access(self.its_device().unwrap(), false)
+                .map_err(Error::RestoreITSTables)?;
+
+            gicv3_its_attr_access(
+                self.its_device().unwrap(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ITS_REGS,
+                GITS_CTLR,
+                &state.its_ctlr,
+                true,
+            )
+            .map_err(Error::RestoreITSCTLR)?;
+
+            Ok(())
+        }
+    }
+
+    impl GicDevice for KvmGicV3Its {
+        fn device(&self) -> &Arc<dyn hypervisor::Device> {
+            &self.device
+        }
+
+        fn its_device(&self) -> Option<&Arc<dyn hypervisor::Device>> {
+            self.its_device.as_ref()
+        }
+
+        fn fdt_compatibility(&self) -> &str {
+            "arm,gic-v3"
+        }
+
+        fn msi_compatible(&self) -> bool {
+            true
+        }
+
+        fn msi_compatibility(&self) -> &str {
+            "arm,gic-v3-its"
+        }
+
+        fn fdt_maint_irq(&self) -> u32 {
+            KvmGicV3::ARCH_GIC_V3_MAINT_IRQ
+        }
+
+        fn msi_properties(&self) -> &[u64] {
+            &self.msi_properties
+        }
+
+        fn device_properties(&self) -> &[u64] {
+            &self.gic_properties
+        }
+
+        fn vcpu_count(&self) -> u64 {
+            self.vcpu_count
+        }
+
+        fn set_its_device(&mut self, its_device: Option<Arc<dyn hypervisor::Device>>) {
+            self.its_device = its_device;
+        }
+
+        fn set_gicr_typers(&mut self, vcpu_states: &[CpuState]) {
+            let gicr_typers = construct_gicr_typers(vcpu_states);
+            self.gicr_typers = gicr_typers;
+        }
+
+        fn as_any_concrete_mut(&mut self) -> &mut dyn Any {
+            self
+        }
+    }
+
+    impl KvmGicDevice for KvmGicV3Its {
+        fn version() -> u32 {
+            KvmGicV3::version()
+        }
+
+        fn create_device(
+            device: Arc<dyn hypervisor::Device>,
+            vcpu_count: u64,
+        ) -> Box<dyn GicDevice> {
+            Box::new(KvmGicV3Its {
+                device,
+                its_device: None,
+                gicr_typers: vec![0; vcpu_count.try_into().unwrap()],
+                gic_properties: [
+                    KvmGicV3::get_dist_addr(),
+                    KvmGicV3::get_dist_size(),
+                    KvmGicV3::get_redists_addr(vcpu_count),
+                    KvmGicV3::get_redists_size(vcpu_count),
+                ],
+                msi_properties: [
+                    KvmGicV3Its::get_msi_addr(vcpu_count),
+                    KvmGicV3Its::get_msi_size(),
+                ],
+                vcpu_count,
+            })
+        }
+
+        fn init_device_attributes(
+            vm: &Arc<dyn hypervisor::Vm>,
+            gic_device: &mut dyn GicDevice,
+        ) -> crate::aarch64::gic::Result<()> {
+            KvmGicV3::init_device_attributes(vm, gic_device)?;
+
+            let mut its_device = kvm_bindings::kvm_create_device {
+                type_: kvm_bindings::kvm_device_type_KVM_DEV_TYPE_ARM_VGIC_ITS,
+                fd: 0,
+                flags: 0,
+            };
+
+            let its_fd = vm
+                .create_device(&mut its_device)
+                .map_err(crate::aarch64::gic::Error::CreateGic)?;
+
+            Self::set_device_attribute(
+                &its_fd,
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_ADDR,
+                u64::from(kvm_bindings::KVM_VGIC_ITS_ADDR_TYPE),
+                &KvmGicV3Its::get_msi_addr(gic_device.vcpu_count()) as *const u64 as u64,
+                0,
+            )?;
+
+            Self::set_device_attribute(
+                &its_fd,
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_CTRL,
+                u64::from(kvm_bindings::KVM_DEV_ARM_VGIC_CTRL_INIT),
+                0,
+                0,
+            )?;
+
+            gic_device.set_its_device(Some(its_fd));
+
+            Ok(())
+        }
+    }
+
+    pub const GIC_V3_ITS_SNAPSHOT_ID: &str = "gic-v3-its";
+    impl Snapshottable for KvmGicV3Its {
+        fn id(&self) -> String {
+            GIC_V3_ITS_SNAPSHOT_ID.to_string()
+        }
+
+        fn snapshot(&mut self) -> std::result::Result<Snapshot, MigratableError> {
+            let gicr_typers = self.gicr_typers.clone();
+            Snapshot::new_from_versioned_state(&self.id(), &self.state(&gicr_typers).unwrap())
+        }
+
+        fn restore(&mut self, snapshot: Snapshot) -> std::result::Result<(), MigratableError> {
+            let gicr_typers = self.gicr_typers.clone();
+            self.set_state(&gicr_typers, &snapshot.to_versioned_state(&self.id())?)
+                .map_err(|e| {
+                    MigratableError::Restore(anyhow!("Could not restore GICv3ITS state {:?}", e))
+                })
+        }
+    }
+
+    impl Pausable for KvmGicV3Its {
+        fn pause(&mut self) -> std::result::Result<(), MigratableError> {
+            // Flush redistributors pending tables to guest RAM.
+            save_pending_tables(self.device()).map_err(|e| {
+                MigratableError::Pause(anyhow!(
+                    "Could not save GICv3ITS GIC pending tables {:?}",
+                    e
+                ))
+            })?;
+            // Flush ITS tables to guest RAM.
+            gicv3_its_tables_access(self.its_device().unwrap(), true).map_err(|e| {
+                MigratableError::Pause(anyhow!("Could not save GICv3ITS ITS tables {:?}", e))
+            })?;
+
+            Ok(())
+        }
+    }
+    impl Transportable for KvmGicV3Its {}
+    impl Migratable for KvmGicV3Its {}
+}

arch/src/aarch64/gic/icc_regs.rs

@@ -1,18 +1,14 @@
-// Copyright 2022 Arm Limited (or its affiliates). All rights reserved.
 // Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-use kvm_ioctls::DeviceFd;
-
-use crate::arch::aarch64::gic::{Error, Result};
-use crate::device::HypervisorDeviceError;
-use crate::kvm::kvm_bindings::{
-    KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS, KVM_REG_ARM64_SYSREG_CRM_MASK,
+use super::{Error, Result};
+use hypervisor::kvm::kvm_bindings::{
+    kvm_device_attr, KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS, KVM_REG_ARM64_SYSREG_CRM_MASK,
     KVM_REG_ARM64_SYSREG_CRM_SHIFT, KVM_REG_ARM64_SYSREG_CRN_MASK, KVM_REG_ARM64_SYSREG_CRN_SHIFT,
     KVM_REG_ARM64_SYSREG_OP0_MASK, KVM_REG_ARM64_SYSREG_OP0_SHIFT, KVM_REG_ARM64_SYSREG_OP1_MASK,
     KVM_REG_ARM64_SYSREG_OP1_SHIFT, KVM_REG_ARM64_SYSREG_OP2_MASK, KVM_REG_ARM64_SYSREG_OP2_SHIFT,
-    kvm_device_attr,
 };
+use std::sync::Arc;
 
 const KVM_DEV_ARM_VGIC_V3_MPIDR_SHIFT: u32 = 32;
 const KVM_DEV_ARM_VGIC_V3_MPIDR_MASK: u64 = 0xffffffff << KVM_DEV_ARM_VGIC_V3_MPIDR_SHIFT as u64;
@@ -81,41 +77,31 @@ static VGIC_ICC_REGS: &[u64] = &[
     SYS_ICC_AP1R3_EL1,
 ];
 
-fn icc_attr_set(gic: &DeviceFd, offset: u64, typer: u64, val: u32) -> Result<()> {
-    let gic_icc_attr = kvm_device_attr {
-        group: KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS,
-        attr: ((typer & KVM_DEV_ARM_VGIC_V3_MPIDR_MASK) | offset), // this needs the mpidr
-        addr: &raw const val as u64,
-        flags: 0,
-    };
-
-    gic.set_device_attr(&gic_icc_attr).map_err(|e| {
-        Error::SetDeviceAttribute(HypervisorDeviceError::SetDeviceAttribute(e.into()))
-    })?;
-
-    Ok(())
-}
-
-fn icc_attr_get(gic: &DeviceFd, offset: u64, typer: u64) -> Result<u32> {
-    let mut val = 0;
-
+fn icc_attr_access(
+    gic: &Arc<dyn hypervisor::Device>,
+    offset: u64,
+    typer: u64,
+    val: &u32,
+    set: bool,
+) -> Result<()> {
     let mut gic_icc_attr = kvm_device_attr {
         group: KVM_DEV_ARM_VGIC_GRP_CPU_SYSREGS,
         attr: ((typer & KVM_DEV_ARM_VGIC_V3_MPIDR_MASK) | offset), // this needs the mpidr
-        addr: &raw mut val as u64,
+        addr: val as *const u32 as u64,
         flags: 0,
     };
-
-    // SAFETY: gic_icc_attr.addr is safe to write to.
-    unsafe { gic.get_device_attr(&mut gic_icc_attr) }.map_err(|e| {
-        Error::GetDeviceAttribute(HypervisorDeviceError::GetDeviceAttribute(e.into()))
-    })?;
-
-    Ok(val)
+    if set {
+        gic.set_device_attr(&gic_icc_attr)
+            .map_err(Error::SetDeviceAttribute)?;
+    } else {
+        gic.get_device_attr(&mut gic_icc_attr)
+            .map_err(Error::GetDeviceAttribute)?;
+    }
+    Ok(())
 }
 
 /// Get ICC registers.
-pub fn get_icc_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
+pub fn get_icc_regs(gic: &Arc<dyn hypervisor::Device>, gicr_typer: &[u64]) -> Result<Vec<u32>> {
     let mut state: Vec<u32> = Vec::new();
     // We need this for the ICC_AP<m>R<n>_EL1 registers.
     let mut num_priority_bits = 0;
@@ -123,9 +109,10 @@ pub fn get_icc_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
     for ix in gicr_typer {
         let i = *ix;
         for icc_offset in VGIC_ICC_REGS {
+            let val = 0;
             if *icc_offset == SYS_ICC_CTLR_EL1 {
                 // calculate priority bits by reading the ctrl_el1 register.
-                let val = icc_attr_get(gic, *icc_offset, i)?;
+                icc_attr_access(gic, *icc_offset, i, &val, false)?;
                 // The priority bits are found in the ICC_CTLR_EL1 register (bits from  10:8).
                 // See page 194 from https://static.docs.arm.com/ihi0069/c/IHI0069C_gic_
                 // architecture_specification.pdf.
@@ -145,7 +132,8 @@ pub fn get_icc_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
             // 7 bits of priority.
             else if *icc_offset == SYS_ICC_AP0R1_EL1 || *icc_offset == SYS_ICC_AP1R1_EL1 {
                 if num_priority_bits >= 6 {
-                    state.push(icc_attr_get(gic, *icc_offset, i)?);
+                    icc_attr_access(gic, *icc_offset, i, &val, false)?;
+                    state.push(val);
                 }
             } else if *icc_offset == SYS_ICC_AP0R2_EL1
                 || *icc_offset == SYS_ICC_AP0R3_EL1
@@ -153,10 +141,12 @@ pub fn get_icc_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
                 || *icc_offset == SYS_ICC_AP1R3_EL1
             {
                 if num_priority_bits == 7 {
-                    state.push(icc_attr_get(gic, *icc_offset, i)?);
+                    icc_attr_access(gic, *icc_offset, i, &val, false)?;
+                    state.push(val);
                 }
             } else {
-                state.push(icc_attr_get(gic, *icc_offset, i)?);
+                icc_attr_access(gic, *icc_offset, i, &val, false)?;
+                state.push(val);
             }
         }
     }
@@ -164,7 +154,11 @@ pub fn get_icc_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
 }
 
 /// Set ICC registers.
-pub fn set_icc_regs(gic: &DeviceFd, gicr_typer: &[u64], state: &[u32]) -> Result<()> {
+pub fn set_icc_regs(
+    gic: &Arc<dyn hypervisor::Device>,
+    gicr_typer: &[u64],
+    state: &[u32],
+) -> Result<()> {
     let mut num_priority_bits = 0;
     let mut idx = 0;
     for ix in gicr_typer {
@@ -177,7 +171,7 @@ pub fn set_icc_regs(gic: &DeviceFd, gicr_typer: &[u64], state: &[u32]) -> Result
             }
             if *icc_offset == SYS_ICC_AP0R1_EL1 || *icc_offset == SYS_ICC_AP1R1_EL1 {
                 if num_priority_bits >= 6 {
-                    icc_attr_set(gic, *icc_offset, i, state[idx])?;
+                    icc_attr_access(gic, *icc_offset, i, &state[idx], true)?;
                     idx += 1;
                 }
                 continue;
@@ -188,12 +182,12 @@ pub fn set_icc_regs(gic: &DeviceFd, gicr_typer: &[u64], state: &[u32]) -> Result
                 || *icc_offset == SYS_ICC_AP1R3_EL1
             {
                 if num_priority_bits == 7 {
-                    icc_attr_set(gic, *icc_offset, i, state[idx])?;
+                    icc_attr_access(gic, *icc_offset, i, &state[idx], true)?;
                     idx += 1;
                 }
                 continue;
             }
-            icc_attr_set(gic, *icc_offset, i, state[idx])?;
+            icc_attr_access(gic, *icc_offset, i, &state[idx], true)?;
             idx += 1;
         }
     }

arch/src/aarch64/gic/mod.rs

@@ -0,0 +1,220 @@
+// Copyright 2021 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
+
+pub mod dist_regs;
+pub mod gicv3;
+pub mod gicv3_its;
+pub mod icc_regs;
+pub mod redist_regs;
+
+pub use self::dist_regs::{get_dist_regs, read_ctlr, set_dist_regs, write_ctlr};
+pub use self::icc_regs::{get_icc_regs, set_icc_regs};
+pub use self::redist_regs::{get_redist_regs, set_redist_regs};
+use hypervisor::CpuState;
+use std::any::Any;
+use std::result;
+use std::sync::Arc;
+
+/// Errors thrown while setting up the GIC.
+#[derive(Debug)]
+pub enum Error {
+    /// Error while calling KVM ioctl for setting up the global interrupt controller.
+    CreateGic(hypervisor::HypervisorVmError),
+    /// Error while setting device attributes for the GIC.
+    SetDeviceAttribute(hypervisor::HypervisorDeviceError),
+    /// Error while getting device attributes for the GIC.
+    GetDeviceAttribute(hypervisor::HypervisorDeviceError),
+}
+type Result<T> = result::Result<T, Error>;
+
+pub trait GicDevice: Send {
+    /// Returns the hypervisor agnostic Device of the GIC device
+    fn device(&self) -> &Arc<dyn hypervisor::Device>;
+
+    /// Returns the hypervisor agnostic Device of the ITS device
+    fn its_device(&self) -> Option<&Arc<dyn hypervisor::Device>> {
+        None
+    }
+
+    /// Returns the fdt compatibility property of the device
+    fn fdt_compatibility(&self) -> &str;
+
+    /// Returns the maint_irq fdt property of the device
+    fn fdt_maint_irq(&self) -> u32;
+
+    /// Returns an array with GIC device properties
+    fn device_properties(&self) -> &[u64];
+
+    /// Returns the number of vCPUs this GIC handles
+    fn vcpu_count(&self) -> u64;
+
+    /// Returns whether the GIC device is MSI compatible or not
+    fn msi_compatible(&self) -> bool {
+        false
+    }
+
+    /// Returns the MSI compatibility property of the device
+    fn msi_compatibility(&self) -> &str {
+        ""
+    }
+
+    /// Returns the MSI reg property of the device
+    fn msi_properties(&self) -> &[u64] {
+        &[]
+    }
+
+    fn set_its_device(&mut self, its_device: Option<Arc<dyn hypervisor::Device>>);
+
+    /// Get the values of GICR_TYPER for each vCPU.
+    fn set_gicr_typers(&mut self, vcpu_states: &[CpuState]);
+
+    /// Downcast the trait object to its concrete type.
+    fn as_any_concrete_mut(&mut self) -> &mut dyn Any;
+}
+
+pub mod kvm {
+    use super::GicDevice;
+    use super::Result;
+    use crate::aarch64::gic::gicv3_its::kvm::KvmGicV3Its;
+    use crate::layout;
+    use hypervisor::kvm::kvm_bindings;
+    use std::boxed::Box;
+    use std::sync::Arc;
+
+    /// Trait for GIC devices.
+    pub trait KvmGicDevice: Send + Sync + GicDevice {
+        /// Returns the GIC version of the device
+        fn version() -> u32;
+
+        /// Create the GIC device object
+        fn create_device(
+            device: Arc<dyn hypervisor::Device>,
+            vcpu_count: u64,
+        ) -> Box<dyn GicDevice>;
+
+        /// Setup the device-specific attributes
+        fn init_device_attributes(
+            vm: &Arc<dyn hypervisor::Vm>,
+            gic_device: &mut dyn GicDevice,
+        ) -> Result<()>;
+
+        /// Initialize a GIC device
+        fn init_device(vm: &Arc<dyn hypervisor::Vm>) -> Result<Arc<dyn hypervisor::Device>> {
+            let mut gic_device = kvm_bindings::kvm_create_device {
+                type_: Self::version(),
+                fd: 0,
+                flags: 0,
+            };
+
+            vm.create_device(&mut gic_device)
+                .map_err(super::Error::CreateGic)
+        }
+
+        /// Set a GIC device attribute
+        fn set_device_attribute(
+            device: &Arc<dyn hypervisor::Device>,
+            group: u32,
+            attr: u64,
+            addr: u64,
+            flags: u32,
+        ) -> Result<()> {
+            let attr = kvm_bindings::kvm_device_attr {
+                flags,
+                group,
+                attr,
+                addr,
+            };
+            device
+                .set_device_attr(&attr)
+                .map_err(super::Error::SetDeviceAttribute)?;
+
+            Ok(())
+        }
+
+        /// Get a GIC device attribute
+        fn get_device_attribute(
+            device: &Arc<dyn hypervisor::Device>,
+            group: u32,
+            attr: u64,
+            addr: u64,
+            flags: u32,
+        ) -> Result<()> {
+            let mut attr = kvm_bindings::kvm_device_attr {
+                flags,
+                group,
+                attr,
+                addr,
+            };
+            device
+                .get_device_attr(&mut attr)
+                .map_err(super::Error::GetDeviceAttribute)?;
+
+            Ok(())
+        }
+
+        /// Finalize the setup of a GIC device
+        fn finalize_device(gic_device: &dyn GicDevice) -> Result<()> {
+            /* We need to tell the kernel how many irqs to support with this vgic.
+             * See the `layout` module for details.
+             */
+            let nr_irqs: u32 = layout::IRQ_NUM;
+            let nr_irqs_ptr = &nr_irqs as *const u32;
+            Self::set_device_attribute(
+                gic_device.device(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_NR_IRQS,
+                0,
+                nr_irqs_ptr as u64,
+                0,
+            )?;
+
+            /* Finalize the GIC.
+             * See https://code.woboq.org/linux/linux/virt/kvm/arm/vgic/vgic-kvm-device.c.html#211.
+             */
+            Self::set_device_attribute(
+                gic_device.device(),
+                kvm_bindings::KVM_DEV_ARM_VGIC_GRP_CTRL,
+                u64::from(kvm_bindings::KVM_DEV_ARM_VGIC_CTRL_INIT),
+                0,
+                0,
+            )?;
+
+            Ok(())
+        }
+
+        /// Method to initialize the GIC device
+        #[allow(clippy::new_ret_no_self)]
+        fn new(vm: &Arc<dyn hypervisor::Vm>, vcpu_count: u64) -> Result<Box<dyn GicDevice>> {
+            let vgic_fd = Self::init_device(vm)?;
+
+            let mut device = Self::create_device(vgic_fd, vcpu_count);
+
+            Self::init_device_attributes(vm, &mut *device)?;
+
+            Self::finalize_device(&*device)?;
+
+            Ok(device)
+        }
+    }
+
+    /// Create a GICv3-ITS device.
+    ///
+    pub fn create_gic(vm: &Arc<dyn hypervisor::Vm>, vcpu_count: u64) -> Result<Box<dyn GicDevice>> {
+        debug!("creating a GICv3-ITS");
+        KvmGicV3Its::new(vm, vcpu_count)
+    }
+
+    /// Function that saves RDIST pending tables into guest RAM.
+    ///
+    /// The tables get flushed to guest RAM whenever the VM gets stopped.
+    pub fn save_pending_tables(gic: &Arc<dyn hypervisor::Device>) -> Result<()> {
+        let init_gic_attr = kvm_bindings::kvm_device_attr {
+            group: kvm_bindings::KVM_DEV_ARM_VGIC_GRP_CTRL,
+            attr: u64::from(kvm_bindings::KVM_DEV_ARM_VGIC_SAVE_PENDING_TABLES),
+            addr: 0,
+            flags: 0,
+        };
+        gic.set_device_attr(&init_gic_attr)
+            .map_err(super::Error::SetDeviceAttribute)
+    }
+}

arch/src/aarch64/gic/redist_regs.rs

@@ -1,18 +1,10 @@
-// Copyright 2022 Arm Limited (or its affiliates). All rights reserved.
 // Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-use kvm_ioctls::DeviceFd;
-
-use crate::CpuState;
-use crate::arch::aarch64::gic::{Error, Result};
-use crate::device::HypervisorDeviceError;
-use crate::kvm::VcpuKvmState;
-use crate::kvm::kvm_bindings::{
-    KVM_DEV_ARM_VGIC_GRP_REDIST_REGS, KVM_REG_ARM64, KVM_REG_ARM64_SYSREG,
-    KVM_REG_ARM64_SYSREG_OP0_MASK, KVM_REG_ARM64_SYSREG_OP0_SHIFT, KVM_REG_ARM64_SYSREG_OP2_MASK,
-    KVM_REG_ARM64_SYSREG_OP2_SHIFT, KVM_REG_SIZE_U64, kvm_device_attr, kvm_one_reg,
-};
+use super::{Error, Result};
+use hypervisor::kvm::kvm_bindings::{kvm_device_attr, KVM_DEV_ARM_VGIC_GRP_REDIST_REGS};
+use hypervisor::CpuState;
+use std::sync::Arc;
 
 // Relevant redistributor registers that we want to save/restore.
 const GICR_CTLR: u32 = 0x0000;
@@ -39,14 +31,8 @@ const GICR_ICFGR0: u32 = GICR_SGI_OFFSET + 0x0C00;
 const KVM_DEV_ARM_VGIC_V3_MPIDR_SHIFT: u32 = 32;
 const KVM_DEV_ARM_VGIC_V3_MPIDR_MASK: u64 = 0xffffffff << KVM_DEV_ARM_VGIC_V3_MPIDR_SHIFT as u64;
 
-const KVM_ARM64_SYSREG_MPIDR_EL1: u64 = KVM_REG_ARM64
-    | KVM_REG_SIZE_U64
-    | KVM_REG_ARM64_SYSREG as u64
-    | (((3_u64) << KVM_REG_ARM64_SYSREG_OP0_SHIFT) & KVM_REG_ARM64_SYSREG_OP0_MASK as u64)
-    | (((5_u64) << KVM_REG_ARM64_SYSREG_OP2_SHIFT) & KVM_REG_ARM64_SYSREG_OP2_MASK as u64);
-
 /// This is how we represent the registers of a distributor.
-/// It is relevant their offset from the base address of the
+/// It is relrvant their offset from the base address of the
 /// distributor.
 /// Each register has a different number
 /// of bits_per_irq and is therefore variable length.
@@ -96,41 +82,31 @@ static VGIC_SGI_REGS: &[RdistReg] = &[
     VGIC_RDIST_REG!(GICR_IPRIORITYR0, 32),
 ];
 
-fn redist_attr_set(gic: &DeviceFd, offset: u32, typer: u64, val: u32) -> Result<()> {
-    let gic_redist_attr = kvm_device_attr {
+fn redist_attr_access(
+    gic: &Arc<dyn hypervisor::Device>,
+    offset: u32,
+    typer: u64,
+    val: &u32,
+    set: bool,
+) -> Result<()> {
+    let mut gic_dist_attr = kvm_device_attr {
         group: KVM_DEV_ARM_VGIC_GRP_REDIST_REGS,
         attr: (typer & KVM_DEV_ARM_VGIC_V3_MPIDR_MASK) | (offset as u64), // this needs the mpidr
-        addr: &raw const val as u64,
+        addr: val as *const u32 as u64,
         flags: 0,
     };
-
-    gic.set_device_attr(&gic_redist_attr).map_err(|e| {
-        Error::SetDeviceAttribute(HypervisorDeviceError::SetDeviceAttribute(e.into()))
-    })?;
-
+    if set {
+        gic.set_device_attr(&gic_dist_attr)
+            .map_err(Error::SetDeviceAttribute)?;
+    } else {
+        gic.get_device_attr(&mut gic_dist_attr)
+            .map_err(Error::GetDeviceAttribute)?;
+    }
     Ok(())
 }
 
-fn redist_attr_get(gic: &DeviceFd, offset: u32, typer: u64) -> Result<u32> {
-    let mut val = 0;
-
-    let mut gic_redist_attr = kvm_device_attr {
-        group: KVM_DEV_ARM_VGIC_GRP_REDIST_REGS,
-        attr: (typer & KVM_DEV_ARM_VGIC_V3_MPIDR_MASK) | (offset as u64), // this needs the mpidr
-        addr: &raw mut val as u64,
-        flags: 0,
-    };
-
-    // SAFETY: gic_redist_attr.addr is safe to write to.
-    unsafe { gic.get_device_attr(&mut gic_redist_attr) }.map_err(|e| {
-        Error::GetDeviceAttribute(HypervisorDeviceError::GetDeviceAttribute(e.into()))
-    })?;
-
-    Ok(val)
-}
-
 fn access_redists_aux(
-    gic: &DeviceFd,
+    gic: &Arc<dyn hypervisor::Device>,
     gicr_typer: &[u64],
     state: &mut Vec<u32>,
     reg_list: &[RdistReg],
@@ -143,11 +119,14 @@ fn access_redists_aux(
             let end = base + rdreg.length as u32;
 
             while base < end {
+                let mut val = 0;
                 if set {
-                    redist_attr_set(gic, base, *i, state[*idx])?;
+                    val = state[*idx];
+                    redist_attr_access(gic, base, *i, &val, true)?;
                     *idx += 1;
                 } else {
-                    state.push(redist_attr_get(gic, base, *i)?);
+                    redist_attr_access(gic, base, *i, &val, false)?;
+                    state.push(val);
                 }
                 base += REG_SIZE as u32;
             }
@@ -157,7 +136,7 @@ fn access_redists_aux(
 }
 
 /// Get redistributor registers.
-pub fn get_redist_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
+pub fn get_redist_regs(gic: &Arc<dyn hypervisor::Device>, gicr_typer: &[u64]) -> Result<Vec<u32>> {
     let mut state = Vec::new();
     let mut idx: usize = 0;
     access_redists_aux(
@@ -174,7 +153,11 @@ pub fn get_redist_regs(gic: &DeviceFd, gicr_typer: &[u64]) -> Result<Vec<u32>> {
 }
 
 /// Set redistributor registers.
-pub fn set_redist_regs(gic: &DeviceFd, gicr_typer: &[u64], state: &[u32]) -> Result<()> {
+pub fn set_redist_regs(
+    gic: &Arc<dyn hypervisor::Device>,
+    gicr_typer: &[u64],
+    state: &[u32],
+) -> Result<()> {
     let mut idx: usize = 0;
     let mut mut_state = state.to_owned();
     access_redists_aux(
@@ -210,18 +193,17 @@ pub fn construct_gicr_typers(vcpu_states: &[CpuState]) -> Vec<u64> {
      */
     let mut gicr_typers: Vec<u64> = Vec::new();
     for (index, state) in vcpu_states.iter().enumerate() {
-        let state: VcpuKvmState = state.clone().into();
-        let last = (index == vcpu_states.len() - 1) as u64;
-        // state.sys_regs is a big collection of system registers, including MIPDR_EL1
-        let mpidr: Vec<kvm_one_reg> = state
-            .sys_regs
-            .into_iter()
-            .filter(|reg| reg.id == KVM_ARM64_SYSREG_MPIDR_EL1)
-            .collect();
+        let last = {
+            if index == vcpu_states.len() - 1 {
+                1
+            } else {
+                0
+            }
+        };
         //calculate affinity
-        let mut cpu_affid = mpidr[0].addr & 1095233437695;
+        let mut cpu_affid = state.mpidr & 1095233437695;
         cpu_affid = ((cpu_affid & 0xFF00000000) >> 8) | (cpu_affid & 0xFFFFFF);
-        gicr_typers.push((cpu_affid << 32) | (1 << 24) | ((index as u64) << 8) | (last << 4));
+        gicr_typers.push((cpu_affid << 32) | (1 << 24) | (index as u64) << 8 | (last << 4));
     }
 
     gicr_typers

arch/src/aarch64/layout.rs

@@ -59,7 +59,7 @@ pub const UEFI_START: GuestAddress = GuestAddress(0);
 pub const UEFI_SIZE: u64 = 0x040_0000;
 
 /// Below this address will reside the GIC, above this address will reside the MMIO devices.
-const MAPPED_IO_START: GuestAddress = GuestAddress(0x0900_0000);
+pub const MAPPED_IO_START: GuestAddress = GuestAddress(0x0900_0000);
 
 /// See kernel file arch/arm64/include/uapi/asm/kvm.h for the GIC related definitions.
 /// 0x08ff_0000 ~ 0x0900_0000 is reserved for GICv3 Distributor
@@ -73,7 +73,7 @@ pub const GIC_V3_REDIST_SIZE: u64 = 0x02_0000;
 pub const GIC_V3_ITS_SIZE: u64 = 0x02_0000;
 
 /// Space 0x0900_0000 ~ 0x0905_0000 is reserved for legacy devices.
-pub const LEGACY_SERIAL_MAPPED_IO_START: GuestAddress = MAPPED_IO_START;
+pub const LEGACY_SERIAL_MAPPED_IO_START: GuestAddress = GuestAddress(0x0900_0000);
 pub const LEGACY_RTC_MAPPED_IO_START: GuestAddress = GuestAddress(0x0901_0000);
 pub const LEGACY_GPIO_MAPPED_IO_START: GuestAddress = GuestAddress(0x0902_0000);
 
@@ -98,11 +98,6 @@ pub const RAM_START: GuestAddress = GuestAddress(0x4000_0000);
 pub const MEM_32BIT_RESERVED_START: GuestAddress = GuestAddress(0xfc00_0000);
 pub const MEM_32BIT_RESERVED_SIZE: u64 = 0x0400_0000;
 
-/// TPM Address Range
-/// This Address range is specific to CRB Interface
-pub const TPM_START: GuestAddress = GuestAddress(0xfed4_0000);
-pub const TPM_SIZE: u64 = 0x1000;
-
 /// Start of 64-bit RAM.
 pub const RAM_64BIT_START: GuestAddress = GuestAddress(0x1_0000_0000);
 
@@ -111,9 +106,8 @@ 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 +132,3 @@ 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

@@ -4,58 +4,56 @@
 
 /// Module for the flattened device tree.
 pub mod fdt;
+/// Module for the global interrupt controller configuration.
+pub mod gic;
 /// Layout for this aarch64 system.
 pub mod layout;
+/// Logic for configuring aarch64 registers.
+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};
-
-pub const _NSIG: i32 = 65;
+use gic::GicDevice;
+use log::{log_enabled, Level};
+use std::collections::HashMap;
+use std::convert::TryInto;
+use std::fmt::Debug;
+use std::sync::Arc;
+use vm_memory::{Address, GuestAddress, GuestMemory, GuestUsize};
 
 /// 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,
+    SetupGic(gic::Error),
 
     /// 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(regs::Error),
 
     /// 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::AArch64Setup(e)
+    }
+}
+
 #[derive(Debug, Copy, Clone)]
 /// Specifies the entry point address where the guest must start
 /// executing code.
@@ -66,25 +64,21 @@ pub struct EntryPoint {
 
 /// Configure the specified VCPU, and return its MPIDR.
 pub fn configure_vcpu(
-    vcpu: &dyn hypervisor::Vcpu,
-    id: u32,
-    boot_setup: Option<(EntryPoint, &GuestMemoryAtomic<GuestMemoryMmap>)>,
+    fd: &Arc<dyn hypervisor::Vcpu>,
+    id: u8,
+    kernel_entry_point: Option<EntryPoint>,
 ) -> super::Result<u64> {
-    if let Some((kernel_entry_point, _guest_memory)) = boot_setup {
-        vcpu.setup_regs(
-            id,
-            kernel_entry_point.entry_addr.raw_value(),
-            super::layout::FDT_START.raw_value(),
-        )
-        .map_err(Error::RegsConfiguration)?;
+    if let Some(kernel_entry_point) = kernel_entry_point {
+        regs::setup_regs(fd, id, kernel_entry_point.entry_addr.raw_value())
+            .map_err(Error::RegsConfiguration)?;
     }
 
-    let mpidr = vcpu.get_sys_reg(MPIDR_EL1).map_err(Error::VcpuRegMpidr)?;
+    let mpidr = fd.read_mpidr().map_err(Error::VcpuRegMpidr)?;
     Ok(mpidr)
 }
 
-pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
-    vec![
+pub fn arch_memory_regions(size: GuestUsize) -> Vec<(GuestAddress, usize, RegionType)> {
+    let mut regions = vec![
         // 0 MiB ~ 256 MiB: UEFI, GIC and legacy devices
         (
             GuestAddress(0),
@@ -103,21 +97,39 @@ pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
             layout::PCI_MMCONFIG_SIZE as usize,
             RegionType::Reserved,
         ),
-        // 1GiB ~ 4032 MiB: RAM before the gap
-        (
+    ];
+
+    let ram_32bit_space_size =
+        layout::MEM_32BIT_RESERVED_START.unchecked_offset_from(layout::RAM_START);
+
+    // RAM space
+    // Case1: guest memory fits before the gap
+    if size as u64 <= ram_32bit_space_size {
+        regions.push((layout::RAM_START, size as usize, RegionType::Ram));
+    // Case2: guest memory extends beyond the gap
+    } else {
+        // Push memory before the gap
+        regions.push((
             layout::RAM_START,
-            layout::MEM_32BIT_RESERVED_START.unchecked_offset_from(layout::RAM_START) as usize,
+            ram_32bit_space_size as usize,
             RegionType::Ram,
-        ),
-        // 4GiB ~ inf: RAM after the gap
-        (layout::RAM_64BIT_START, usize::MAX, RegionType::Ram),
-        // Add the 32-bit reserved memory hole as a reserved region
-        (
-            layout::MEM_32BIT_RESERVED_START,
-            layout::MEM_32BIT_RESERVED_SIZE as usize,
-            RegionType::Reserved,
-        ),
-    ]
+        ));
+        // Other memory is placed after 4GiB
+        regions.push((
+            layout::RAM_64BIT_START,
+            (size - ram_32bit_space_size) as usize,
+            RegionType::Ram,
+        ));
+    }
+
+    // Add the 32-bit reserved memory hole as a reserved region
+    regions.push((
+        layout::MEM_32BIT_RESERVED_START,
+        layout::MEM_32BIT_RESERVED_SIZE as usize,
+        RegionType::Reserved,
+    ));
+
+    regions
 }
 
 /// Configures the system and should be called once per vm before starting vcpu threads.
@@ -125,13 +137,13 @@ 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],
     virtio_iommu_bdf: Option<u32>,
-    gic_device: &Arc<Mutex<dyn Vgic>>,
+    gic_device: &dyn GicDevice,
     numa_nodes: &NumaNodes,
     pmu_supported: bool,
 ) -> super::Result<()> {
@@ -154,7 +166,7 @@ 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(())
 }
@@ -173,15 +185,18 @@ pub fn initramfs_load_addr(
             if guest_mem.address_in_range(offset) {
                 Ok(offset.raw_value())
             } else {
-                Err(super::Error::PlatformSpecific(Error::InitramfsAddress))
+                Err(super::Error::AArch64Setup(Error::InitramfsAddress))
             }
         }
-        None => Err(super::Error::PlatformSpecific(Error::InitramfsAddress)),
+        None => Err(super::Error::AArch64Setup(Error::InitramfsAddress)),
     }
 }
 
-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.
@@ -192,16 +207,30 @@ pub fn get_host_cpu_phys_bits(hypervisor: &dyn hypervisor::Hypervisor) -> u8 {
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]
-    fn test_arch_memory_regions_dram() {
-        let regions = arch_memory_regions();
+    fn test_arch_memory_regions_dram_2gb() {
+        let regions = arch_memory_regions((1usize << 31) as u64); //2GB
+        assert_eq!(5, regions.len());
+        assert_eq!(layout::RAM_START, regions[3].0);
+        assert_eq!((1usize << 31), regions[3].1);
+        assert_eq!(RegionType::Ram, regions[3].2);
+        assert_eq!(RegionType::Reserved, regions[4].2);
+    }
+
+    #[test]
+    fn test_arch_memory_regions_dram_4gb() {
+        let regions = arch_memory_regions((1usize << 32) as u64); //4GB
+        let ram_32bit_space_size =
+            layout::MEM_32BIT_RESERVED_START.unchecked_offset_from(layout::RAM_START) as usize;
         assert_eq!(6, regions.len());
         assert_eq!(layout::RAM_START, regions[3].0);
+        assert_eq!(ram_32bit_space_size as usize, regions[3].1);
         assert_eq!(RegionType::Ram, regions[3].2);
         assert_eq!(RegionType::Reserved, regions[5].2);
         assert_eq!(RegionType::Ram, regions[4].2);
+        assert_eq!(((1usize << 32) - ram_32bit_space_size), regions[4].1);
     }
 }

arch/src/aarch64/regs.rs

@@ -0,0 +1,75 @@
+// 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 hypervisor::kvm::kvm_bindings::{
+    kvm_regs, user_pt_regs, KVM_REG_ARM64, KVM_REG_ARM_CORE, KVM_REG_SIZE_U64,
+};
+use hypervisor::{arm64_core_reg_id, offset__of};
+use std::sync::Arc;
+use std::{mem, result};
+use vm_memory::Address;
+
+/// Errors thrown while setting aarch64 registers.
+#[derive(Debug)]
+pub enum Error {
+    /// Failed to set core register (PC, PSTATE or general purpose ones).
+    SetCoreRegister(hypervisor::HypervisorCpuError),
+    /// Failed to get a system register.
+    GetSysRegister(hypervisor::HypervisorCpuError),
+}
+type Result<T> = result::Result<T, Error>;
+
+#[allow(non_upper_case_globals)]
+// PSR (Processor State Register) bits.
+// Taken from arch/arm64/include/uapi/asm/ptrace.h.
+const PSR_MODE_EL1h: u64 = 0x0000_0005;
+const PSR_F_BIT: u64 = 0x0000_0040;
+const PSR_I_BIT: u64 = 0x0000_0080;
+const PSR_A_BIT: u64 = 0x0000_0100;
+const PSR_D_BIT: u64 = 0x0000_0200;
+// Taken from arch/arm64/kvm/inject_fault.c.
+const PSTATE_FAULT_BITS_64: u64 = PSR_MODE_EL1h | PSR_A_BIT | PSR_F_BIT | PSR_I_BIT | PSR_D_BIT;
+
+/// Configure core registers for a given CPU.
+///
+/// # Arguments
+///
+/// * `vcpu` - Structure for the VCPU that holds the VCPU's fd.
+/// * `cpu_id` - Index of current vcpu.
+/// * `boot_ip` - Starting instruction pointer.
+/// * `mem` - Reserved DRAM for current VM.
+pub fn setup_regs(vcpu: &Arc<dyn hypervisor::Vcpu>, cpu_id: u8, boot_ip: u64) -> Result<()> {
+    let kreg_off = offset__of!(kvm_regs, regs);
+
+    // Get the register index of the PSTATE (Processor State) register.
+    let pstate = offset__of!(user_pt_regs, pstate) + kreg_off;
+    vcpu.set_reg(
+        arm64_core_reg_id!(KVM_REG_SIZE_U64, pstate),
+        PSTATE_FAULT_BITS_64,
+    )
+    .map_err(Error::SetCoreRegister)?;
+
+    // Other vCPUs are powered off initially awaiting PSCI wakeup.
+    if cpu_id == 0 {
+        // Setting the PC (Processor Counter) to the current program address (kernel address).
+        let pc = offset__of!(user_pt_regs, pc) + kreg_off;
+        vcpu.set_reg(arm64_core_reg_id!(KVM_REG_SIZE_U64, pc), boot_ip as u64)
+            .map_err(Error::SetCoreRegister)?;
+
+        // Last mandatory thing to set -> the address pointing to the FDT (also called DTB).
+        // "The device tree blob (dtb) must be placed on an 8-byte boundary and must
+        // not exceed 2 megabytes in size." -> https://www.kernel.org/doc/Documentation/arm64/booting.txt.
+        // We are choosing to place it the end of DRAM. See `get_fdt_addr`.
+        let regs0 = offset__of!(user_pt_regs, regs) + kreg_off;
+        vcpu.set_reg(
+            arm64_core_reg_id!(KVM_REG_SIZE_U64, regs0),
+            super::layout::FDT_START.raw_value(),
+        )
+        .map_err(Error::SetCoreRegister)?;
+    }
+    Ok(())
+}

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))
@@ -43,8 +34,10 @@ where
     if uefi_size > 0x300000 {
         return Err(Error::UefiTooBig);
     }
-    uefi_image.rewind().map_err(|_| Error::SeekUefiStart)?;
+    uefi_image
+        .seek(SeekFrom::Start(0))
+        .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.
 //
@@ -6,55 +5,60 @@
 // SPDX-License-Identifier: Apache-2.0
 
 //! Implements platform specific functionality.
-//! Supported platforms: x86_64, aarch64, riscv64.
+//! Supported platforms: x86_64, aarch64.
 
+#[macro_use]
+extern crate log;
+#[macro_use]
+extern crate serde_derive;
+
+#[cfg(target_arch = "x86_64")]
+use crate::x86_64::SgxEpcSection;
 use std::collections::BTreeMap;
+use std::fmt;
+use std::result;
 use std::sync::Arc;
-use std::{fmt, result};
-
-use serde::{Deserialize, Serialize};
-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>;
 
 /// Type for returning error code.
-#[derive(Debug, Error)]
+#[derive(Debug)]
 pub enum Error {
     #[cfg(target_arch = "x86_64")]
-    #[error("Platform specific error (x86_64)")]
-    PlatformSpecific(#[from] x86_64::Error),
+    /// X86_64 specific error triggered during system configuration.
+    X86_64Setup(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("The memory map table extends past the end of guest memory")]
-    MemmapTablePastRamEnd,
-    #[error("Error writing memory map table to guest memory")]
-    MemmapTableSetup,
-    #[error("The hvm_start_info structure extends past the end of guest memory")]
-    StartInfoPastRamEnd,
-    #[error("Error writing hvm_start_info to guest memory")]
-    StartInfoSetup,
-    #[error("Failed to compute initramfs address")]
-    InitramfsAddress,
-    #[error("Error writing module entry to guest memory")]
-    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")]
+    /// AArch64 specific error triggered during system configuration.
+    AArch64Setup(aarch64::Error),
+    /// The zero page extends past the end of guest_mem.
     ZeroPagePastRamEnd,
+    /// Error writing the zero page of guest memory.
+    ZeroPageSetup(vm_memory::GuestMemoryError),
+    /// The memory map table extends past the end of guest memory.
+    MemmapTablePastRamEnd,
+    /// Error writing memory map table to guest memory.
+    MemmapTableSetup,
+    /// The hvm_start_info structure extends past the end of guest memory.
+    StartInfoPastRamEnd,
+    /// Error writing hvm_start_info to guest memory.
+    StartInfoSetup,
+    /// Failed to compute initramfs address.
+    InitramfsAddress,
+    /// Error writing module entry to guest memory.
+    ModlistSetup(vm_memory::GuestMemoryError),
+    /// RSDP Beyond Guest Memory
+    RsdpPastRamEnd,
 }
 
 /// 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, Debug, Serialize, Deserialize, Versionize)]
 pub enum RegionType {
     /// RAM type
     Ram,
@@ -72,26 +76,17 @@ 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,
 };
 
 #[cfg(target_arch = "x86_64")]
@@ -99,16 +94,16 @@ 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,
 };
 
 /// Safe wrapper for `sysconf(_SC_PAGESIZE)`.
 #[cfg(target_arch = "x86_64")]
 #[inline(always)]
 fn pagesize() -> usize {
-    // SAFETY: Trivially safe
+    // Trivially safe
     unsafe { libc::sysconf(libc::_SC_PAGESIZE) as usize }
 }
 
@@ -116,11 +111,11 @@ 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 +134,7 @@ 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 +142,6 @@ 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.
@@ -157,13 +149,13 @@ pub const PAGE_SIZE: usize = 4096;
 
 impl fmt::Display for DeviceType {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        write!(f, "{self:?}")
+        write!(f, "{:?}", self)
     }
 }
 
 /// 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 +164,7 @@ 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 +172,7 @@ 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

@@ -5,7 +5,12 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE-BSD-3-Clause file.
 
+use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
+use hypervisor::x86_64::LapicState;
+use std::io::Cursor;
+use std::mem;
 use std::result;
+use std::sync::Arc;
 
 pub type Result<T> = result::Result<T, hypervisor::HypervisorCpuError>;
 
@@ -15,6 +20,32 @@ pub const APIC_LVT1: usize = 0x360;
 pub const APIC_MODE_NMI: u32 = 0x4;
 pub const APIC_MODE_EXTINT: u32 = 0x7;
 
+pub fn get_klapic_reg(klapic: &LapicState, reg_offset: usize) -> u32 {
+    let sliceu8 = unsafe {
+        // This array is only accessed as parts of a u32 word, so interpret it as a u8 array.
+        // Cursors are only readable on arrays of u8, not i8(c_char).
+        mem::transmute::<&[i8], &[u8]>(&klapic.regs[reg_offset..])
+    };
+    let mut reader = Cursor::new(sliceu8);
+    // Following call can't fail if the offsets defined above are correct.
+    reader
+        .read_u32::<LittleEndian>()
+        .expect("Failed to read klapic register")
+}
+
+pub fn set_klapic_reg(klapic: &mut LapicState, reg_offset: usize, value: u32) {
+    let sliceu8 = unsafe {
+        // This array is only accessed as parts of a u32 word, so interpret it as a u8 array.
+        // Cursors are only readable on arrays of u8, not i8(c_char).
+        mem::transmute::<&mut [i8], &mut [u8]>(&mut klapic.regs[reg_offset..])
+    };
+    let mut writer = Cursor::new(sliceu8);
+    // Following call can't fail if the offsets defined above are correct.
+    writer
+        .write_u32::<LittleEndian>(value)
+        .expect("Failed to write klapic register")
+}
+
 pub fn set_apic_delivery_mode(reg: u32, mode: u32) -> u32 {
     ((reg) & !0x700) | ((mode) << 8)
 }
@@ -23,16 +54,45 @@ 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);
-    klapic.set_klapic_reg(
+    let lvt_lint0 = get_klapic_reg(&klapic, APIC_LVT0);
+    set_klapic_reg(
+        &mut klapic,
         APIC_LVT0,
         set_apic_delivery_mode(lvt_lint0, APIC_MODE_EXTINT),
     );
-    let lvt_lint1 = klapic.get_klapic_reg(APIC_LVT1);
-    klapic.set_klapic_reg(APIC_LVT1, set_apic_delivery_mode(lvt_lint1, APIC_MODE_NMI));
+    let lvt_lint1 = get_klapic_reg(&klapic, APIC_LVT1);
+    set_klapic_reg(
+        &mut klapic,
+        APIC_LVT1,
+        set_apic_delivery_mode(lvt_lint1, APIC_MODE_NMI),
+    );
 
     vcpu.set_lapic(&klapic)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    const KVM_APIC_REG_SIZE: usize = 0x400;
+
+    #[test]
+    fn test_set_and_get_klapic_reg() {
+        let reg_offset = 0x340;
+        let mut klapic = LapicState::default();
+        set_klapic_reg(&mut klapic, reg_offset, 3);
+        let value = get_klapic_reg(&klapic, reg_offset);
+        assert_eq!(value, 3);
+    }
+
+    #[test]
+    #[should_panic]
+    fn test_set_and_get_klapic_out_of_bounds() {
+        let reg_offset = KVM_APIC_REG_SIZE + 10;
+        let mut klapic = LapicState::default();
+        set_klapic_reg(&mut klapic, reg_offset, 3);
+    }
+}

arch/src/x86_64/layout.rs

@@ -107,11 +107,6 @@ pub const KVM_TSS_SIZE: u64 = (3 * 4) << 10;
 pub const KVM_IDENTITY_MAP_START: GuestAddress = GuestAddress(KVM_TSS_START.0 + KVM_TSS_SIZE);
 pub const KVM_IDENTITY_MAP_SIZE: u64 = 4 << 10;
 
-/// TPM Address Range
-/// This Address range is specific to CRB Interface
-pub const TPM_START: GuestAddress = GuestAddress(0xfed4_0000);
-pub const TPM_SIZE: u64 = 0x1000;
-
 // IOAPIC
 pub const IOAPIC_START: GuestAddress = GuestAddress(0xfec0_0000);
 pub const IOAPIC_SIZE: u64 = 0x20;

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:
@@ -39,71 +37,66 @@ struct MpfIntelWrapper(mpspec::mpf_intel);
 
 // SAFETY: These `mpspec` wrapper types are only data, reading them from data is a safe initialization.
 unsafe impl ByteValued for MpcBusWrapper {}
-// SAFETY: see above
 unsafe impl ByteValued for MpcCpuWrapper {}
-// SAFETY: see above
 unsafe impl ByteValued for MpcIntsrcWrapper {}
-// SAFETY: see above
 unsafe impl ByteValued for MpcIoapicWrapper {}
-// SAFETY: see above
 unsafe impl ByteValued for MpcTableWrapper {}
-// SAFETY: see above
 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;
-    // 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>()) };
+fn compute_checksum<T: Copy>(v: &T) -> u8 {
+    // Safe because we are only reading the bytes within the size of the `T` reference `v`.
+    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 +109,7 @@ 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 +120,9 @@ 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 +136,7 @@ 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 +148,13 @@ 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 +174,7 @@ 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 +195,7 @@ 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 +266,14 @@ 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 +282,10 @@ 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 {
@@ -313,7 +294,7 @@ mod unit_tests {
             mpspec::MP_IOAPIC => mem::size_of::<MpcIoapicWrapper>(),
             mpspec::MP_INTSRC => mem::size_of::<MpcIntsrcWrapper>(),
             mpspec::MP_LINTSRC => mem::size_of::<MpcLintsrcWrapper>(),
-            _ => panic!("unrecognized mpc table entry type: {type_}"),
+            _ => panic!("unrecognized mpc table entry type: {}", type_),
         }
     }
 
@@ -323,7 +304,7 @@ 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 +313,7 @@ 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 +322,7 @@ 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 +338,27 @@ 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 +367,12 @@ 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 +401,11 @@ 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::regs::*;
+use hypervisor::x86_64::{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(&hypervisor::x86_64::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)
 }
 
@@ -151,9 +123,8 @@ 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] = {
+    let gdt_table: [u64; BOOT_GDT_MAX as usize] = {
         // Configure GDT entries as specified by PVH boot protocol
         [
             gdt_entry(0, 0, 0),               // NULL
@@ -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 +6,46 @@
 //
 // 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 vm_memory::ByteValued;
+use vm_memory::{Address, Bytes, GuestAddress};
 
-#[derive(Debug, Error)]
+#[allow(unused_variables)]
+#[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),
+}
+
+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",
+            AddressOverflow => "The SMBIOS table has too little address space to be stored",
+            Clear => "Failure while zeroing out the memory for the SMBIOS table",
+            WriteSmbiosEp => "Failure to write SMBIOS entrypoint structure",
+            WriteData => "Failure to write additional data to memory",
+        };
+
+        write!(f, "SMBIOS error: {}", description)
+    }
 }
 
 pub type Result<T> = result::Result<T, Error>;
@@ -43,15 +54,13 @@ pub type Result<T> = result::Result<T, Error>;
 const SM3_MAGIC_IDENT: &[u8; 5usize] = b"_SM3_";
 const BIOS_INFORMATION: u8 = 0;
 const SYSTEM_INFORMATION: u8 = 1;
-const OEM_STRINGS: u8 = 11;
 const END_OF_TABLE: u8 = 127;
 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>()) };
+    // Safe because we are only reading the bytes within the size of the `T` reference `v`.
+    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);
@@ -59,85 +68,75 @@ fn compute_checksum<T: Copy>(v: &T) -> u8 {
     (!checksum).wrapping_add(1)
 }
 
-#[repr(C)]
 #[repr(packed)]
-#[derive(Default, Copy, Clone)]
-struct Smbios30Entrypoint {
-    signature: [u8; 5usize],
-    checksum: u8,
-    length: u8,
-    majorver: u8,
-    minorver: u8,
-    docrev: u8,
-    revision: u8,
-    reserved: u8,
-    max_size: u32,
-    physptr: u64,
+#[derive(Default, Copy)]
+pub struct Smbios30Entrypoint {
+    pub signature: [u8; 5usize],
+    pub checksum: u8,
+    pub length: u8,
+    pub majorver: u8,
+    pub minorver: u8,
+    pub docrev: u8,
+    pub revision: u8,
+    pub reserved: u8,
+    pub max_size: u32,
+    pub physptr: u64,
 }
 
-#[repr(C)]
-#[repr(packed)]
-#[derive(Default, Copy, Clone)]
-struct SmbiosBiosInfo {
-    r#type: u8,
-    length: u8,
-    handle: u16,
-    vendor: u8,
-    version: u8,
-    start_addr: u16,
-    release_date: u8,
-    rom_size: u8,
-    characteristics: u64,
-    characteristics_ext1: u8,
-    characteristics_ext2: u8,
+impl Clone for Smbios30Entrypoint {
+    fn clone(&self) -> Self {
+        *self
+    }
 }
 
-#[repr(C)]
 #[repr(packed)]
-#[derive(Default, Copy, Clone)]
-struct SmbiosSysInfo {
-    r#type: u8,
-    length: u8,
-    handle: u16,
-    manufacturer: u8,
-    product_name: u8,
-    version: u8,
-    serial_number: u8,
-    uuid: [u8; 16usize],
-    wake_up_type: u8,
-    sku: u8,
-    family: u8,
+#[derive(Default, Copy)]
+pub struct SmbiosBiosInfo {
+    pub typ: u8,
+    pub length: u8,
+    pub handle: u16,
+    pub vendor: u8,
+    pub version: u8,
+    pub start_addr: u16,
+    pub release_date: u8,
+    pub rom_size: u8,
+    pub characteristics: u64,
+    pub characteristics_ext1: u8,
+    pub characteristics_ext2: u8,
 }
 
-#[repr(C)]
-#[repr(packed)]
-#[derive(Default, Copy, Clone)]
-struct SmbiosOemStrings {
-    r#type: u8,
-    length: u8,
-    handle: u16,
-    count: u8,
+impl Clone for SmbiosBiosInfo {
+    fn clone(&self) -> Self {
+        *self
+    }
 }
 
-#[repr(C)]
 #[repr(packed)]
-#[derive(Default, Copy, Clone)]
-struct SmbiosEndOfTable {
-    r#type: u8,
-    length: u8,
-    handle: u16,
+#[derive(Default, Copy)]
+pub struct SmbiosSysInfo {
+    pub typ: u8,
+    pub length: u8,
+    pub handle: u16,
+    pub manufacturer: u8,
+    pub product_name: u8,
+    pub version: u8,
+    pub serial_number: u8,
+    pub uuid: [u8; 16usize],
+    pub wake_up_type: u8,
+    pub sku: u8,
+    pub family: u8,
 }
 
-// SAFETY: data structure only contain a series of integers
+impl Clone for SmbiosSysInfo {
+    fn clone(&self) -> Self {
+        *self
+    }
+}
+
+// SAFETY: These data structures only contain a series of integers
 unsafe impl ByteValued for Smbios30Entrypoint {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for SmbiosBiosInfo {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for SmbiosSysInfo {}
-// SAFETY: data structure only contain a series of integers
-unsafe impl ByteValued for SmbiosOemStrings {}
-// SAFETY: data structure only contain a series of integers
-unsafe impl ByteValued for SmbiosEndOfTable {}
 
 fn write_and_incr<T: ByteValued>(
     mem: &GuestMemoryMmap,
@@ -163,12 +162,7 @@ fn write_string(
     Ok(curptr)
 }
 
-pub fn setup_smbios(
-    mem: &GuestMemoryMmap,
-    serial_number: Option<&str>,
-    uuid: Option<&str>,
-    oem_strings: Option<&[&str]>,
-) -> Result<u64> {
+pub fn setup_smbios(mem: &GuestMemoryMmap) -> Result<u64> {
     let physptr = GuestAddress(SMBIOS_START)
         .checked_add(mem::size_of::<Smbios30Entrypoint>() as u64)
         .ok_or(Error::NotEnoughMemory)?;
@@ -178,7 +172,7 @@ pub fn setup_smbios(
     {
         handle += 1;
         let smbios_biosinfo = SmbiosBiosInfo {
-            r#type: BIOS_INFORMATION,
+            typ: BIOS_INFORMATION,
             length: mem::size_of::<SmbiosBiosInfo>() as u8,
             handle,
             vendor: 1,  // First string written in this section
@@ -195,59 +189,29 @@ pub fn setup_smbios(
 
     {
         handle += 1;
-
-        let uuid_number = uuid
-            .map(Uuid::parse_str)
-            .transpose()
-            .map_err(|e| Error::ParseUuid(e, uuid.unwrap().to_string()))?
-            .unwrap_or(Uuid::nil());
         let smbios_sysinfo = SmbiosSysInfo {
-            r#type: SYSTEM_INFORMATION,
+            typ: SYSTEM_INFORMATION,
             length: mem::size_of::<SmbiosSysInfo>() as u8,
             handle,
             manufacturer: 1, // First string written in this section
             product_name: 2, // Second string written in this section
-            serial_number: serial_number.map(|_| 3).unwrap_or_default(), // 3rd string
-            uuid: uuid_number.to_bytes_le(), // set uuid
             ..Default::default()
         };
         curptr = write_and_incr(mem, smbios_sysinfo, curptr)?;
         curptr = write_string(mem, "Cloud Hypervisor", curptr)?;
         curptr = write_string(mem, "cloud-hypervisor", curptr)?;
-        if let Some(serial_number) = serial_number {
-            curptr = write_string(mem, serial_number, curptr)?;
-        }
-        curptr = write_and_incr(mem, 0u8, curptr)?;
-    }
-
-    if let Some(oem_strings) = oem_strings {
-        handle += 1;
-
-        let smbios_oemstrings = SmbiosOemStrings {
-            r#type: OEM_STRINGS,
-            length: mem::size_of::<SmbiosOemStrings>() as u8,
-            handle,
-            count: oem_strings.len() as u8,
-        };
-
-        curptr = write_and_incr(mem, smbios_oemstrings, curptr)?;
-
-        for s in oem_strings {
-            curptr = write_string(mem, s, curptr)?;
-        }
-
         curptr = write_and_incr(mem, 0u8, curptr)?;
     }
 
     {
         handle += 1;
-        let smbios_end = SmbiosEndOfTable {
-            r#type: END_OF_TABLE,
-            length: mem::size_of::<SmbiosEndOfTable>() as u8,
+        let smbios_sysinfo = SmbiosSysInfo {
+            typ: END_OF_TABLE,
+            length: mem::size_of::<SmbiosSysInfo>() as u8,
             handle,
+            ..Default::default()
         };
-        curptr = write_and_incr(mem, smbios_end, curptr)?;
-        curptr = write_and_incr(mem, 0u8, curptr)?;
+        curptr = write_and_incr(mem, smbios_sysinfo, curptr)?;
         curptr = write_and_incr(mem, 0u8, curptr)?;
     }
 
@@ -269,11 +233,11 @@ pub fn setup_smbios(
             .map_err(|_| Error::WriteSmbiosEp)?;
     }
 
-    Ok(curptr.unchecked_offset_from(physptr) + std::mem::size_of::<Smbios30Entrypoint>() as u64)
+    Ok(curptr.unchecked_offset_from(physptr))
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]
@@ -299,7 +263,7 @@ mod unit_tests {
     fn entrypoint_checksum() {
         let mem = GuestMemoryMmap::from_ranges(&[(GuestAddress(SMBIOS_START), 4096)]).unwrap();
 
-        setup_smbios(&mem, None, None, None).unwrap();
+        setup_smbios(&mem).unwrap();
 
         let smbios_ep: Smbios30Entrypoint = mem.read_obj(GuestAddress(SMBIOS_START)).unwrap();
 

arch/src/x86_64/tdx/mod.rs

@@ -1,42 +1,30 @@
 // 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")]
-    ReadGuidTable(#[source] std::io::Error),
     #[error("Invalid descriptor signature")]
     InvalidDescriptorSignature,
     #[error("Invalid descriptor size")]
     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")]
-    UuidCreation(#[source] uuid::Error),
 }
 
-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 +34,7 @@ pub struct TdvfDescriptor {
 }
 
 // TDVF_SECTION
-#[repr(C, packed)]
+#[repr(packed)]
 #[derive(Clone, Copy, Default, Debug)]
 pub struct TdvfSection {
     pub data_offset: u32,
@@ -58,7 +46,7 @@ pub struct TdvfSection {
 }
 
 #[repr(u32)]
-#[derive(Clone, Copy, Debug, Default)]
+#[derive(Clone, Copy, Debug)]
 pub enum TdvfSectionType {
     Bfv,
     Cfv,
@@ -67,76 +55,16 @@ pub enum TdvfSectionType {
     PermMem,
     Payload,
     PayloadParam,
-    #[default]
     Reserved = 0xffffffff,
 }
 
-fn tdvf_descriptor_offset(file: &mut File) -> Result<(SeekFrom, bool), TdvfError> {
-    // Let's first try to identify the presence of the table footer GUID
-    file.seek(SeekFrom::End(-0x30))
-        .map_err(TdvfError::ReadGuidTable)?;
-    let mut table_footer_guid: [u8; 16] = [0; 16];
-    file.read_exact(&mut table_footer_guid)
-        .map_err(TdvfError::ReadGuidTable)?;
-    let uuid =
-        Uuid::from_slice_le(table_footer_guid.as_slice()).map_err(TdvfError::UuidCreation)?;
-    let expected_uuid = Uuid::from_str(TABLE_FOOTER_GUID).map_err(TdvfError::UuidCreation)?;
-    if uuid == expected_uuid {
-        // Retrieve the table size
-        file.seek(SeekFrom::End(-0x32))
-            .map_err(TdvfError::ReadGuidTable)?;
-        let mut table_size: [u8; 2] = [0; 2];
-        file.read_exact(&mut table_size)
-            .map_err(TdvfError::ReadGuidTable)?;
-        let table_size = u16::from_le_bytes(table_size) as usize;
-        let mut table: Vec<u8> = vec![0; table_size];
-
-        // Read the entire table
-        file.seek(SeekFrom::End(-(table_size as i64 + 0x20)))
-            .map_err(TdvfError::ReadGuidTable)?;
-        file.read_exact(table.as_mut_slice())
-            .map_err(TdvfError::ReadGuidTable)?;
-
-        // Let's start from the top and go backward down the table.
-        // We start after the footer GUID and the table length.
-        let mut offset = table_size - 18;
-
-        debug!("Parsing GUID structure");
-        while offset >= 18 {
-            let entry_uuid = Uuid::from_slice_le(&table[offset - 16..offset])
-                .map_err(TdvfError::UuidCreation)?;
-            let entry_size =
-                u16::from_le_bytes(table[offset - 18..offset - 16].try_into().unwrap()) as usize;
-            debug!(
-                "Entry GUID = {}, size = {}",
-                entry_uuid.hyphenated(),
-                entry_size
-            );
-
-            // Avoid going through an infinite loop if the entry size is 0
-            if entry_size == 0 {
-                break;
-            }
-
-            offset -= entry_size;
-
-            let expected_uuid =
-                Uuid::from_str(TDVF_METADATA_OFFSET_GUID).map_err(TdvfError::UuidCreation)?;
-            if entry_uuid == expected_uuid && entry_size == 22 {
-                return Ok((
-                    SeekFrom::End(
-                        -(u32::from_le_bytes(table[offset..offset + 4].try_into().unwrap()) as i64),
-                    ),
-                    true,
-                ));
-            }
-        }
+impl Default for TdvfSectionType {
+    fn default() -> Self {
+        TdvfSectionType::Reserved
     }
+}
 
-    // If we end up here, this means the firmware doesn't support the new way
-    // of exposing the TDVF descriptor offset through the table of GUIDs.
-    // That's why we fallback onto the deprecated method.
-
+pub fn parse_tdvf_sections(file: &mut File) -> Result<Vec<TdvfSection>, TdvfError> {
     // The 32-bit offset to the TDVF metadata is located 32 bytes from
     // the end of the file.
     // See "TDVF Metadata Pointer" in "TDX Virtual Firmware Design Guide
@@ -146,24 +74,16 @@ fn tdvf_descriptor_offset(file: &mut File) -> Result<(SeekFrom, bool), TdvfError
     let mut descriptor_offset: [u8; 4] = [0; 4];
     file.read_exact(&mut descriptor_offset)
         .map_err(TdvfError::ReadDescriptorOffset)?;
+    let descriptor_offset = u32::from_le_bytes(descriptor_offset) as u64;
 
-    Ok((
-        SeekFrom::Start(u32::from_le_bytes(descriptor_offset) as u64),
-        false,
-    ))
-}
-
-pub fn parse_tdvf_sections(file: &mut File) -> Result<(Vec<TdvfSection>, bool), TdvfError> {
-    let (descriptor_offset, guid_found) = tdvf_descriptor_offset(file)?;
-
-    file.seek(descriptor_offset)
+    file.seek(SeekFrom::Start(descriptor_offset))
         .map_err(TdvfError::ReadDescriptor)?;
 
     let mut descriptor: TdvfDescriptor = Default::default();
-    // SAFETY: we read exactly the size of the descriptor header
+    // Safe as 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>(),
         )
     })
@@ -187,30 +107,35 @@ pub fn parse_tdvf_sections(file: &mut File) -> Result<(Vec<TdvfSection>, bool),
     let mut sections = Vec::new();
     sections.resize_with(descriptor.num_sections as usize, TdvfSection::default);
 
-    // SAFETY: we read exactly the advertised sections
+    // Safe as 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>(),
         )
     })
     .map_err(TdvfError::ReadDescriptor)?;
 
-    Ok((sections, guid_found))
+    Ok(sections)
 }
 
 #[repr(u16)]
-#[derive(Copy, Clone, Debug, Default)]
+#[derive(Copy, Clone, Debug)]
 enum HobType {
     Handoff = 0x1,
     ResourceDescriptor = 0x3,
     GuidExtension = 0x4,
-    #[default]
     Unused = 0xfffe,
     EndOfHobList = 0xffff,
 }
 
-#[repr(C, packed)]
+impl Default for HobType {
+    fn default() -> Self {
+        HobType::Unused
+    }
+}
+
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobHeader {
     r#type: HobType,
@@ -218,7 +143,7 @@ struct HobHeader {
     reserved: u32,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobHandoffInfoTable {
     header: HobHeader,
@@ -231,7 +156,7 @@ struct HobHandoffInfoTable {
     efi_end_of_hob_list: u64,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct EfiGuid {
     data1: u32,
@@ -240,7 +165,7 @@ struct EfiGuid {
     data4: [u8; 8],
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobResourceDescriptor {
     header: HobHeader,
@@ -251,7 +176,7 @@ struct HobResourceDescriptor {
     resource_length: u64,
 }
 
-#[repr(C, packed)]
+#[repr(C)]
 #[derive(Copy, Clone, Default, Debug)]
 struct HobGuidType {
     header: HobHeader,
@@ -259,39 +184,39 @@ struct HobGuidType {
 }
 
 #[repr(u32)]
-#[derive(Clone, Copy, Debug, Default)]
+#[derive(Clone, Copy, Debug)]
 pub enum PayloadImageType {
-    #[default]
     ExecutablePayload,
     BzImage,
     RawVmLinux,
 }
 
-#[repr(C, packed)]
+impl Default for PayloadImageType {
+    fn default() -> Self {
+        PayloadImageType::ExecutablePayload
+    }
+}
+
+#[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,
     payload_info: PayloadInfo,
 }
 
-// SAFETY: data structure only contain a series of integers
+// SAFETY: These data structures only contain a series of integers
 unsafe impl ByteValued for HobHeader {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for HobHandoffInfoTable {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for HobResourceDescriptor {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for HobGuidType {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for PayloadInfo {}
-// SAFETY: data structure only contain a series of integers
 unsafe impl ByteValued for TdPayload {}
 
 pub struct TdHob {
@@ -300,12 +225,12 @@ 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 {
@@ -387,19 +312,12 @@ impl TdHob {
         physical_start: u64,
         resource_length: u64,
         ram: bool,
-        guid_found: bool,
     ) -> Result<(), TdvfError> {
         self.add_resource(
             mem,
             physical_start,
             resource_length,
             if ram {
-                if guid_found {
-                    0x7 /* EFI_RESOURCE_MEMORY_UNACCEPTED */
-                } else {
-                    0 /* EFI_RESOURCE_SYSTEM_MEMORY */
-                }
-            } else if guid_found {
                 0 /* EFI_RESOURCE_SYSTEM_MEMORY */
             } else {
                 0x5 /*EFI_RESOURCE_MEMORY_RESERVED */
@@ -520,16 +438,16 @@ impl TdHob {
 }
 
 #[cfg(test)]
-mod unit_tests {
+mod tests {
     use super::*;
 
     #[test]
     #[ignore]
     fn test_parse_tdvf_sections() {
         let mut f = std::fs::File::open("tdvf.fd").unwrap();
-        let (sections, _) = parse_tdvf_sections(&mut f).unwrap();
+        let sections = parse_tdvf_sections(&mut f).unwrap();
         for section in sections {
-            eprintln!("{section:x?}");
+            eprintln!("{:x?}", section)
         }
     }
 }

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 +1,24 @@
+[package]
+name = "block_util"
+version = "0.1.0"
+authors = ["The Cloud Hypervisor Authors"]
+edition = "2021"
+
+[features]
+default = []
+
+[dependencies]
+io-uring = "0.5.2"
+libc = "0.2.123"
+log = "0.4.16"
+qcow = { path = "../qcow" }
+thiserror = "1.0.30"
+versionize = "0.1.6"
+versionize_derive = "0.1.4"
+vhdx = { path = "../vhdx" }
+virtio-bindings = { version = "0.1.0", features = ["virtio-v5_0_0"] }
+virtio-queue = "0.2.0"
+vm-memory = { version = "0.7.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
+vm-virtio = { path = "../vm-virtio" }
+vmm-sys-util = "0.9.0"
+

block_util/src/async_io.rs

@@ -0,0 +1,146 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
+
+use libc::{ioctl, S_IFBLK, S_IFMT};
+use std::convert::TryInto;
+use std::fs::File;
+use std::os::unix::io::AsRawFd;
+use thiserror::Error;
+use vmm_sys_util::eventfd::EventFd;
+use vmm_sys_util::{ioctl_expr, 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();
+        let ret = unsafe { libc::fstat(f.as_raw_fd(), stat.as_mut_ptr()) };
+        if ret != 0 {
+            return Err(std::io::Error::last_os_error());
+        }
+
+        let is_block = unsafe { (*stat.as_ptr()).st_mode & S_IFMT == S_IFBLK };
+        Ok(is_block)
+    }
+
+    // libc::ioctl() takes different types on different architectures
+    #[allow(clippy::useless_conversion)]
+    fn query_block_size(f: &mut File, block_size_type: BlockSize) -> std::io::Result<u64> {
+        let mut block_size = 0;
+        let ret = unsafe {
+            ioctl(
+                f.as_raw_fd(),
+                match block_size_type {
+                    BlockSize::LogicalBlock => BLKSSZGET(),
+                    BlockSize::PhysicalBlock => BLKPBSZGET(),
+                    BlockSize::MinimumIo => BLKIOMIN(),
+                    BlockSize::OptimalIo => BLKIOOPT(),
+                }
+                .try_into()
+                .unwrap(),
+                &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 + Sync {
+    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 + Sync {
+    fn notifier(&self) -> &EventFd;
+    fn read_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: Vec<libc::iovec>,
+        user_data: u64,
+    ) -> AsyncIoResult<()>;
+    fn write_vectored(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: Vec<libc::iovec>,
+        user_data: u64,
+    ) -> AsyncIoResult<()>;
+    fn fsync(&mut self, user_data: Option<u64>) -> AsyncIoResult<()>;
+    fn complete(&mut self) -> Vec<(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 {
@@ -35,7 +64,7 @@ impl AsyncIo for FixedVhdAsync {
     fn read_vectored(
         &mut self,
         offset: libc::off_t,
-        iovecs: &[libc::iovec],
+        iovecs: Vec<libc::iovec>,
         user_data: u64,
     ) -> AsyncIoResult<()> {
         if offset as u64 >= self.size {
@@ -54,7 +83,7 @@ impl AsyncIo for FixedVhdAsync {
     fn write_vectored(
         &mut self,
         offset: libc::off_t,
-        iovecs: &[libc::iovec],
+        iovecs: Vec<libc::iovec>,
         user_data: u64,
     ) -> AsyncIoResult<()> {
         if offset as u64 >= self.size {
@@ -75,27 +104,7 @@ impl AsyncIo for FixedVhdAsync {
         self.raw_file_async.fsync(user_data)
     }
 
-    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)
+    fn complete(&mut self) -> Vec<(u64, i32)> {
+        self.raw_file_async.complete()
     }
 }

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,
@@ -31,7 +62,7 @@ impl AsyncIo for FixedVhdSync {
     fn read_vectored(
         &mut self,
         offset: libc::off_t,
-        iovecs: &[libc::iovec],
+        iovecs: Vec<libc::iovec>,
         user_data: u64,
     ) -> AsyncIoResult<()> {
         if offset as u64 >= self.size {
@@ -50,7 +81,7 @@ impl AsyncIo for FixedVhdSync {
     fn write_vectored(
         &mut self,
         offset: libc::off_t,
-        iovecs: &[libc::iovec],
+        iovecs: Vec<libc::iovec>,
         user_data: u64,
     ) -> AsyncIoResult<()> {
         if offset as u64 >= self.size {
@@ -70,19 +101,7 @@ impl AsyncIo for FixedVhdSync {
         self.raw_file_sync.fsync(user_data)
     }
 
-    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",
-        )))
+    fn complete(&mut self) -> Vec<(u64, i32)> {
+        self.raw_file_sync.complete()
     }
 }

block_util/src/lib.rs

@@ -0,0 +1,700 @@
+// 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 std::alloc::{alloc_zeroed, dealloc, Layout};
+use std::cmp;
+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 versionize::{VersionMap, Versionize, VersionizeResult};
+use versionize_derive::Versionize;
+use virtio_bindings::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(Debug)]
+pub enum Error {
+    /// Guest gave us bad memory addresses.
+    GuestMemory(GuestMemoryError),
+    /// Guest gave us offsets that would have overflowed a usize.
+    CheckedOffset(GuestAddress, usize),
+    /// Guest gave us a write only descriptor that protocol says to read from.
+    UnexpectedWriteOnlyDescriptor,
+    /// Guest gave us a read only descriptor that protocol says to write to.
+    UnexpectedReadOnlyDescriptor,
+    /// Guest gave us too few descriptors in a descriptor chain.
+    DescriptorChainTooShort,
+    /// Guest gave us a descriptor that was too short to use.
+    DescriptorLengthTooSmall,
+    /// Getting a block's metadata fails for any reason.
+    GetFileMetadata,
+    /// The requested operation would cause a seek beyond disk end.
+    InvalidOffset,
+    /// 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(Debug)]
+pub enum ExecuteError {
+    BadRequest(Error),
+    Flush(io::Error),
+    Read(GuestMemoryError),
+    Seek(io::Error),
+    Write(GuestMemoryError),
+    Unsupported(u32),
+    SubmitIoUring(io::Error),
+    GetHostAddress(GuestMemoryError),
+    AsyncRead(AsyncIoError),
+    AsyncWrite(AsyncIoError),
+    AsyncFlush(AsyncIoError),
+    /// Failed allocating a temporary buffer.
+    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)]
+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: Vec<(GuestAddress, u32)>,
+    pub status_addr: GuestAddress,
+    pub writeback: bool,
+    pub aligned_operations: Vec<AlignedOperation>,
+}
+
+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: Vec::new(),
+            status_addr: GuestAddress(0),
+            writeback: true,
+            aligned_operations: Vec::new(),
+        };
+
+        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 {
+            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 = Vec::new();
+        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();
+                // Safe because 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 {
+                    // Safe because 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 {
+                // Safe because 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.
+            // Safe because 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: these data structures only contain a series of integers
+unsafe impl ByteValued for VirtioBlockConfig {}
+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_READ is supported
+    if !probe.is_supported(opcode::Read::CODE) {
+        info!("{} IORING_OP_READ operation not supported", error_msg);
+        return false;
+    }
+
+    // Check IORING_OP_WRITE is supported
+    if !probe.is_supported(opcode::Write::CODE) {
+        info!("{} IORING_OP_WRITE 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: Vec<libc::iovec>,
+        user_data: u64,
+        eventfd: &EventFd,
+        completion_list: &mut Vec<(u64, i32)>,
+    ) -> AsyncIoResult<()> {
+        // Convert libc::iovec into IoSliceMut
+        let mut slices = Vec::new();
+        for iovec in iovecs.iter() {
+            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((user_data, result as i32));
+        eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn write_vectored_sync(
+        &mut self,
+        offset: libc::off_t,
+        iovecs: Vec<libc::iovec>,
+        user_data: u64,
+        eventfd: &EventFd,
+        completion_list: &mut Vec<(u64, i32)>,
+    ) -> AsyncIoResult<()> {
+        // Convert libc::iovec into IoSlice
+        let mut slices = Vec::new();
+        for iovec in iovecs.iter() {
+            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((user_data, result as i32));
+        eventfd.write(1).unwrap();
+
+        Ok(())
+    }
+
+    fn fsync_sync(
+        &mut self,
+        user_data: Option<u64>,
+        eventfd: &EventFd,
+        completion_list: &mut Vec<(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((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 +1,103 @@
+// 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::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();
+
+        Ok(file.seek(SeekFrom::End(0)).map_err(DiskFileError::Size)? as u64)
+    }
+
+    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: Vec<(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: Vec::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: Vec<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: Vec<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 complete(&mut self) -> Vec<(u64, i32)> {
+        self.completion_list.drain(..).collect()
+    }
+}

block_util/src/raw_async.rs

@@ -0,0 +1,168 @@
+// 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> {
+        Ok(self
+            .file
+            .seek(SeekFrom::End(0))
+            .map_err(DiskFileError::Size)? as u64)
+    }
+
+    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: Vec<libc::iovec>,
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        let (submitter, mut sq, _) = self.io_uring.split();
+
+        // Safe because 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: Vec<libc::iovec>,
+        user_data: u64,
+    ) -> AsyncIoResult<()> {
+        let (submitter, mut sq, _) = self.io_uring.split();
+
+        // Safe because 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();
+
+            // Safe because 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 {
+            unsafe { libc::fsync(self.fd) };
+        }
+
+        Ok(())
+    }
+
+    fn complete(&mut self) -> Vec<(u64, i32)> {
+        let mut completion_list = Vec::new();
+
+        let cq = self.io_uring.completion();
+        for cq_entry in cq {
+            completion_list.push((cq_entry.user_data(), cq_entry.result()));
+        }
+
+        completion_list
+    }
+}