build: Release v34.0

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>

.github/workflows/build.yaml

@@ -13,7 +13,7 @@ jobs:
           - stable
           - beta
           - nightly
-          - "1.62"
+          - "1.64"
         target:
           - x86_64-unknown-linux-gnu
           - x86_64-unknown-linux-musl
@@ -42,6 +42,9 @@ jobs:
       - name: Build (default features + tdx)
         run: cargo rustc --locked --bin cloud-hypervisor --features "tdx" -- -D warnings -D clippy::undocumented_unsafe_blocks
 
+      - name: Build (default features + dbus_api)
+        run: cargo rustc --locked --bin cloud-hypervisor --features "dbus_api" -- -D warnings -D clippy::undocumented_unsafe_blocks
+
       - name: Build (default features + guest_debug)
         run: cargo rustc --locked --bin cloud-hypervisor --features "guest_debug" -- -D warnings -D clippy::undocumented_unsafe_blocks
 

.github/workflows/hadolint.yaml

@@ -0,0 +1,26 @@
+name: Lint Dockerfile
+
+on:
+  push:
+    paths:
+    - resources/Dockerfile
+  pull_request:
+    paths:
+    - resources/Dockerfile
+
+jobs:
+  hadolint:
+    name: Run Hadolint Dockerfile Linter
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v3
+
+      - name: Lint Dockerfile
+        uses: hadolint/hadolint-action@master
+        with:
+          dockerfile: ./resources/Dockerfile
+          format: tty
+          no-fail: false
+          verbose: true
+          failure-threshold: info

.github/workflows/release.yaml

@@ -28,13 +28,13 @@ jobs:
         with:
           toolchain: "1.67.1"
           command: build
-          args: --all --release --no-default-features --features "kvm,mshv" --target=x86_64-unknown-linux-gnu
+          args: --all --release --features mshv --target=x86_64-unknown-linux-gnu
       - name: Static Build
         uses: actions-rs/cargo@v1
         with:
           toolchain: "1.67.1"
           command: build
-          args: --all --release --no-default-features --features "kvm,mshv" --target=x86_64-unknown-linux-musl
+          args: --all --release --features mshv --target=x86_64-unknown-linux-musl
       - name: Install Rust toolchain (aarch64-unknown-linux-musl)
         uses: actions-rs/toolchain@v1
         with:

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "cloud-hypervisor"
-version = "31.0.0"
+version = "34.0.0"
 authors = ["The Cloud Hypervisor Authors"]
 edition = "2021"
 default-run = "cloud-hypervisor"
@@ -15,7 +15,7 @@ homepage = "https://github.com/cloud-hypervisor/cloud-hypervisor"
 # a.) A dependency requires it,
 # b.) If we want to use a new feature and that MSRV is at least 6 months old,
 # c.) There is a security issue that is addressed by the toolchain update.
-rust-version = "1.62"
+rust-version = "1.64"
 
 [profile.release]
 lto = true
@@ -29,43 +29,48 @@ strip = false
 debug = true
 
 [dependencies]
-anyhow = "1.0.69"
+anyhow = "1.0.71"
 api_client = { path = "api_client" }
 argh = "0.1.9"
 dhat = { version = "0.3.2", optional = true }
-epoll = "4.3.1"
+epoll = "4.3.3"
 event_monitor = { path = "event_monitor" }
 hypervisor = { path = "hypervisor" }
 libc = "0.2.139"
 log = { version = "0.4.17", features = ["std"] }
 option_parser = { path = "option_parser" }
 seccompiler = "0.3.0"
-serde_json = "1.0.95"
-signal-hook = "0.3.15"
-thiserror = "1.0.39"
+serde_json = "1.0.96"
+signal-hook = "0.3.17"
+thiserror = "1.0.40"
 tpm = { path = "tpm"}
 tracer = { path = "tracer" }
 vmm = { path = "vmm" }
 vmm-sys-util = "0.11.0"
-vm-memory = "0.10.0"
+vm-memory = "0.11.0"
+zbus = { version = "3.11.1", optional = true }
 
 # List of patched crates
 [patch.crates-io]
 kvm-bindings = { git = "https://github.com/cloud-hypervisor/kvm-bindings", branch = "ch-v0.6.0-tdx" }
+kvm-ioctls = { git = "https://github.com/rust-vmm/kvm-ioctls", branch = "main" }
 versionize_derive = { git = "https://github.com/cloud-hypervisor/versionize_derive", branch = "ch" }
+vhost = { git = "https://github.com/rust-vmm/vhost", branch = "main" }
 
 [dev-dependencies]
-dirs = "4.0.0"
+dirs = "5.0.0"
 net_util = { path = "net_util" }
-once_cell = "1.17.1"
-serde_json = "1.0.95"
+once_cell = "1.18.0"
+serde_json = "1.0.96"
 test_infra = { path = "test_infra" }
 wait-timeout = "0.2.0"
 
 [features]
-default = ["kvm"]
+default = ["kvm", "io_uring"]
+dbus_api = ["zbus", "vmm/dbus_api"]
 dhat-heap = ["dhat"] # For heap profiling
 guest_debug = ["vmm/guest_debug"]
+io_uring = ["vmm/io_uring"]
 kvm = ["vmm/kvm"]
 mshv = ["vmm/mshv"]
 tdx = ["vmm/tdx"]
@@ -75,7 +80,7 @@ tracing = ["vmm/tracing", "tracer/tracing"]
 members = [
     "api_client",
     "arch",
-    "block_util",
+    "block",
     "devices",
     "event_monitor",
     "hypervisor",
@@ -84,12 +89,10 @@ members = [
     "option_parser",
     "pci",
     "performance-metrics",
-    "qcow",
     "rate_limiter",
     "serial_buffer",
     "test_infra",
     "tracer",
-    "vhdx",
     "vhost_user_block",
     "vhost_user_net",
     "virtio-devices",

Jenkinsfile

@@ -117,6 +117,63 @@ pipeline {
                         }
                     }
                 }
+                stage('Worker build - AMD') {
+                    agent { node { label 'jammy-amd' } }
+                    when {
+                        beforeAgent true
+                        expression {
+                            return runWorkers
+                        }
+                    }
+                    stages {
+                        stage('Checkout') {
+                            steps {
+                                checkout scm
+                            }
+                        }
+                        stage('Prepare environment') {
+                            steps {
+                                sh 'scripts/prepare_vdpa.sh'
+                            }
+                        }
+                        stage('Run integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration'
+                            }
+                        }
+                        stage('Run live-migration integration tests') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration-live-migration --integration -- -- --skip live_migration::live_migration_parallel::test_live_upgrade_watchdog --skip live_migration::live_migration_parallel::test_live_upgrade_watchdog_local'
+                            }
+                        }
+                        stage('Run integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration --libc musl'
+                            }
+                        }
+                        stage('Run live-migration integration tests for musl') {
+                            options {
+                                timeout(time: 1, unit: 'HOURS')
+                            }
+                            steps {
+                                sh 'sudo modprobe openvswitch'
+                                sh 'scripts/dev_cli.sh tests --integration-live-migration --libc musl -- -- --skip live_migration::live_migration_parallel::test_live_upgrade_watchdog --skip live_migration::live_migration_parallel::test_live_upgrade_watchdog_local'
+                            }
+                        }
+                    }
+                }
                 stage('AArch64 worker build') {
                     agent { node { label 'bionic-arm64' } }
                     when {
@@ -238,146 +295,146 @@ pipeline {
                         }
                     }
                 }
-                stage('Worker build - Metrics') {
-                    agent { node { label 'jammy-metrics' } }
-                    when {
-                        branch 'main'
-                        beforeAgent true
-                        expression {
-                            return runWorkers
-                        }
-                    }
-                    environment {
-                        METRICS_PUBLISH_KEY = credentials('52e0945f-ce7a-43d1-87af-67d1d87cc40f')
-                    }
-                    stages {
-                        stage('Checkout') {
-                            steps {
-                                checkout scm
-                            }
-                        }
-                        stage('Run metrics tests') {
-                            options {
-                                timeout(time: 1, unit: 'HOURS')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --metrics -- -- --report-file /root/workloads/metrics.json'
-                            }
-                        }
-                        stage('Upload metrics report') {
-                            steps {
-                                sh 'curl -X PUT https://cloud-hypervisor-metrics.azurewebsites.net/api/publishmetrics -H "x-functions-key: $METRICS_PUBLISH_KEY" -T ~/workloads/metrics.json'
-                            }
-                        }
-                    }
-                }
-                stage('Worker build - Rate Limiter') {
-                    agent { node { label 'focal-metrics' } }
-                    when {
-                        branch 'main'
-                        beforeAgent true
-                        expression {
-                            return runWorkers
-                        }
-                    }
-                    stages {
-                        stage('Checkout') {
-                            steps {
-                                checkout scm
-                            }
-                        }
-                        stage('Run rate-limiter integration tests') {
-                            options {
-                                timeout(time: 10, unit: 'MINUTES')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --integration-rate-limiter'
-                            }
-                        }
-                    }
-                }
-                stage('Worker build - SGX') {
-                    agent { node { label 'jammy-sgx' } }
-                    when {
-                        beforeAgent true
-                        allOf {
-                            branch 'main'
-                            expression {
-                                return runWorkers
-                            }
-                        }
-                    }
-                    stages {
-                        stage('Checkout') {
-                            steps {
-                                checkout scm
-                            }
-                        }
-                        stage('Run SGX integration tests') {
-                            options {
-                                timeout(time: 1, unit: 'HOURS')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --integration-sgx'
-                            }
-                        }
-                        stage('Run SGX integration tests for musl') {
-                            options {
-                                timeout(time: 1, unit: 'HOURS')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --integration-sgx --libc musl'
-                            }
-                        }
-                    }
-                    post {
-                        always {
-                            sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
-                            deleteDir()
-                        }
-                    }
-                }
-                stage('Worker build - VFIO') {
-                    agent { node { label 'jammy-vfio' } }
-                    when {
-                        beforeAgent true
-                        allOf {
-                            branch 'main'
-                            expression {
-                                return runWorkers
-                            }
-                        }
-                    }
-                    stages {
-                        stage('Checkout') {
-                            steps {
-                                checkout scm
-                            }
-                        }
-                        stage('Run VFIO integration tests') {
-                            options {
-                                timeout(time: 1, unit: 'HOURS')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --integration-vfio'
-                            }
-                        }
-                        stage('Run VFIO integration tests for musl') {
-                            options {
-                                timeout(time: 1, unit: 'HOURS')
-                            }
-                            steps {
-                                sh 'scripts/dev_cli.sh tests --integration-vfio --libc musl'
-                            }
-                        }
-                    }
-                    post {
-                        always {
-                            sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
-                            deleteDir()
-                        }
-                    }
-                }
+                // stage('Worker build - Metrics') {
+                //     agent { node { label 'jammy-metrics' } }
+                //     when {
+                //         branch 'main'
+                //         beforeAgent true
+                //         expression {
+                //             return runWorkers
+                //         }
+                //     }
+                //     environment {
+                //         METRICS_PUBLISH_KEY = credentials('52e0945f-ce7a-43d1-87af-67d1d87cc40f')
+                //     }
+                //     stages {
+                //         stage('Checkout') {
+                //             steps {
+                //                 checkout scm
+                //             }
+                //         }
+                //         stage('Run metrics tests') {
+                //             options {
+                //                 timeout(time: 1, unit: 'HOURS')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --metrics -- -- --report-file /root/workloads/metrics.json'
+                //             }
+                //         }
+                //         stage('Upload metrics report') {
+                //             steps {
+                //                 sh 'curl -X PUT https://cloud-hypervisor-metrics.azurewebsites.net/api/publishmetrics -H "x-functions-key: $METRICS_PUBLISH_KEY" -T ~/workloads/metrics.json'
+                //             }
+                //         }
+                //     }
+                // }
+                // stage('Worker build - Rate Limiter') {
+                //     agent { node { label 'focal-metrics' } }
+                //     when {
+                //         branch 'main'
+                //         beforeAgent true
+                //         expression {
+                //             return runWorkers
+                //         }
+                //     }
+                //     stages {
+                //         stage('Checkout') {
+                //             steps {
+                //                 checkout scm
+                //             }
+                //         }
+                //         stage('Run rate-limiter integration tests') {
+                //             options {
+                //                 timeout(time: 10, unit: 'MINUTES')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --integration-rate-limiter'
+                //             }
+                //         }
+                //     }
+                // }
+                // stage('Worker build - SGX') {
+                //     agent { node { label 'jammy-sgx' } }
+                //     when {
+                //         beforeAgent true
+                //         allOf {
+                //             branch 'main'
+                //             expression {
+                //                 return runWorkers
+                //             }
+                //         }
+                //     }
+                //     stages {
+                //         stage('Checkout') {
+                //             steps {
+                //                 checkout scm
+                //             }
+                //         }
+                //         stage('Run SGX integration tests') {
+                //             options {
+                //                 timeout(time: 1, unit: 'HOURS')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --integration-sgx'
+                //             }
+                //         }
+                //         stage('Run SGX integration tests for musl') {
+                //             options {
+                //                 timeout(time: 1, unit: 'HOURS')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --integration-sgx --libc musl'
+                //             }
+                //         }
+                //     }
+                //     post {
+                //         always {
+                //             sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+                //             deleteDir()
+                //         }
+                //     }
+                // }
+                // stage('Worker build - VFIO') {
+                //     agent { node { label 'jammy-vfio' } }
+                //     when {
+                //         beforeAgent true
+                //         allOf {
+                //             branch 'main'
+                //             expression {
+                //                 return runWorkers
+                //             }
+                //         }
+                //     }
+                //     stages {
+                //         stage('Checkout') {
+                //             steps {
+                //                 checkout scm
+                //             }
+                //         }
+                //         stage('Run VFIO integration tests') {
+                //             options {
+                //                 timeout(time: 1, unit: 'HOURS')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --integration-vfio'
+                //             }
+                //         }
+                //         stage('Run VFIO integration tests for musl') {
+                //             options {
+                //                 timeout(time: 1, unit: 'HOURS')
+                //             }
+                //             steps {
+                //                 sh 'scripts/dev_cli.sh tests --integration-vfio --libc musl'
+                //             }
+                //         }
+                //     }
+                //     post {
+                //         always {
+                //             sh "sudo chown -R jenkins.jenkins ${WORKSPACE}"
+                //             deleteDir()
+                //         }
+                //     }
+                // }
             }
         }
     }

README.md

@@ -78,9 +78,9 @@ The following sections describe how to build and run Cloud Hypervisor.
 
 ## Host OS
 
-For required KVM functionality the minimum host kernel version is 4.11.  For
-adequate performance the minimum recommended host kernel version is 5.6. The
-majority of the CI currently tests with kernel version 5.15.
+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
 
@@ -140,7 +140,10 @@ 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.
+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.
 
 ```shell
 $ sudo setcap cap_net_admin+ep ./cloud-hypervisor

arch/Cargo.toml

@@ -9,18 +9,18 @@ default = []
 tdx = []
 
 [dependencies]
-anyhow = "1.0.69"
+anyhow = "1.0.71"
 byteorder = "1.4.3"
 hypervisor = { path = "../hypervisor" }
 libc = "0.2.139"
-linux-loader = { version = "0.8.1", features = ["elf", "bzimage", "pe"] }
+linux-loader = { version = "0.9.0", features = ["elf", "bzimage", "pe"] }
 log = "0.4.17"
-serde = { version = "1.0.151", features = ["rc", "derive"] }
-thiserror = "1.0.39"
-uuid = "1.3.0"
+serde = { version = "1.0.164", features = ["rc", "derive"] }
+thiserror = "1.0.40"
+uuid = "1.3.4"
 versionize = "0.1.10"
 versionize_derive = "0.1.4"
-vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-bitmap"] }
+vm-memory = { version = "0.11.0", features = ["backend-mmap", "backend-bitmap"] }
 vm-migration = { path = "../vm-migration" }
 vmm-sys-util = { version = "0.11.0", features = ["with-serde"] }
 

arch/src/aarch64/fdt.rs

@@ -24,6 +24,8 @@ 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 std::fs;
+use std::path::Path;
 use vm_fdt::{FdtWriter, FdtWriterResult};
 use vm_memory::{Address, Bytes, GuestMemory, GuestMemoryError, GuestMemoryRegion};
 
@@ -40,8 +42,12 @@ const VIRTIO_IOMMU_PHANDLE: u32 = 5;
 // NOTE: Keep FIRST_VCPU_PHANDLE the last PHANDLE defined.
 // This is a value for uniquely identifying the FDT node containing the first vCPU.
 // The last number of vCPU phandle depends on the number of vCPUs.
-const FIRST_VCPU_PHANDLE: u32 = 6;
+const FIRST_VCPU_PHANDLE: u32 = 8;
 
+// This is a value for uniquely identifying the FDT node containing the L2 cache info
+const L2_CACHE_PHANDLE: u32 = 6;
+// This is a value for uniquely identifying the FDT node containing the L3 cache info
+const L3_CACHE_PHANDLE: u32 = 7;
 // Read the documentation specified when appending the root node to the FDT.
 const ADDRESS_CELLS: u32 = 0x2;
 const SIZE_CELLS: u32 = 0x2;
@@ -81,6 +87,97 @@ pub enum Error {
 }
 type Result<T> = result::Result<T, Error>;
 
+pub enum CacheLevel {
+    /// L1 data cache
+    L1D = 0,
+    /// L1 instruction cache
+    L1I = 1,
+    /// L2 cache
+    L2 = 2,
+    /// L3 cache
+    L3 = 3,
+}
+
+/// NOTE: CACHE SIZE file directory example,
+/// "/sys/devices/system/cpu/cpu0/cache/index0/size".
+pub fn get_cache_size(cache_level: CacheLevel) -> u32 {
+    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
+    match cache_level {
+        CacheLevel::L1D => file_directory += "/index0/size",
+        CacheLevel::L1I => file_directory += "/index1/size",
+        CacheLevel::L2 => file_directory += "/index2/size",
+        CacheLevel::L3 => file_directory += "/index3/size",
+    }
+
+    let file_path = Path::new(&file_directory);
+    if !file_path.exists() {
+        error!("File: {} not exist.", file_directory);
+        0
+    } else {
+        info!("File: {} exist.", file_directory);
+
+        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
+        // The content of the file is as simple as a size, like: "32K"
+        let src = src.trim();
+        let src_digits: u32 = src[0..src.len() - 1].parse().unwrap();
+        let src_unit = &src[src.len() - 1..];
+
+        src_digits
+            * match src_unit {
+                "K" => 1024,
+                "M" => 1024u32.pow(2),
+                "G" => 1024u32.pow(3),
+                _ => 1,
+            }
+    }
+}
+
+/// NOTE: CACHE COHERENCY LINE SIZE file directory example,
+/// "/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size".
+pub fn get_cache_coherency_line_size(cache_level: CacheLevel) -> u32 {
+    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
+    match cache_level {
+        CacheLevel::L1D => file_directory += "/index0/coherency_line_size",
+        CacheLevel::L1I => file_directory += "/index1/coherency_line_size",
+        CacheLevel::L2 => file_directory += "/index2/coherency_line_size",
+        CacheLevel::L3 => file_directory += "/index3/coherency_line_size",
+    }
+
+    let file_path = Path::new(&file_directory);
+    if !file_path.exists() {
+        error!("File: {} not exist.", file_directory);
+        0
+    } else {
+        info!("File: {} exist.", file_directory);
+
+        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
+        src.trim().parse::<u32>().unwrap()
+    }
+}
+
+/// NOTE: CACHE NUMBER OF SETS file directory example,
+/// "/sys/devices/system/cpu/cpu0/cache/index0/number_of_sets".
+pub fn get_cache_number_of_sets(cache_level: CacheLevel) -> u32 {
+    let mut file_directory: String = "/sys/devices/system/cpu/cpu0/cache".to_string();
+    match cache_level {
+        CacheLevel::L1D => file_directory += "/index0/number_of_sets",
+        CacheLevel::L1I => file_directory += "/index1/number_of_sets",
+        CacheLevel::L2 => file_directory += "/index2/number_of_sets",
+        CacheLevel::L3 => file_directory += "/index3/number_of_sets",
+    }
+
+    let file_path = Path::new(&file_directory);
+    if !file_path.exists() {
+        error!("File: {} not exist.", file_directory);
+        0
+    } else {
+        info!("File: {} exist.", file_directory);
+
+        let src = fs::read_to_string(file_directory).expect("File not exists or file corrupted.");
+        src.trim().parse::<u32>().unwrap()
+    }
+}
+
 /// Creates the flattened device tree for this aarch64 VM.
 #[allow(clippy::too_many_arguments)]
 pub fn create_fdt<T: DeviceInfoForFdt + Clone + Debug, S: ::std::hash::BuildHasher>(
@@ -159,6 +256,54 @@ fn create_cpu_nodes(
 
     let num_cpus = vcpu_mpidr.len();
 
+    // Add cache info.
+    // L1 Data Cache Info.
+    let mut l1_d_cache_size: u32 = 0;
+    let mut l1_d_cache_line_size: u32 = 0;
+    let mut l1_d_cache_sets: u32 = 0;
+
+    // L1 Instruction Cache Info.
+    let mut l1_i_cache_size: u32 = 0;
+    let mut l1_i_cache_line_size: u32 = 0;
+    let mut l1_i_cache_sets: u32 = 0;
+
+    // L2 Cache Info.
+    let mut l2_cache_size: u32 = 0;
+    let mut l2_cache_line_size: u32 = 0;
+    let mut l2_cache_sets: u32 = 0;
+
+    // L3 Cache Info.
+    let mut l3_cache_size: u32 = 0;
+    let mut l3_cache_line_size: u32 = 0;
+    let mut l3_cache_sets: u32 = 0;
+
+    let cache_path = Path::new("/sys/devices/system/cpu/cpu0/cache");
+    let cache_exist: bool = cache_path.exists();
+    if !cache_exist {
+        error!("cache sysfs system does not exist.");
+    } else {
+        info!("cache sysfs system exists.");
+        // L1 Data Cache Info.
+        l1_d_cache_size = get_cache_size(CacheLevel::L1D);
+        l1_d_cache_line_size = get_cache_coherency_line_size(CacheLevel::L1D);
+        l1_d_cache_sets = get_cache_number_of_sets(CacheLevel::L1D);
+
+        // L1 Instruction Cache Info.
+        l1_i_cache_size = get_cache_size(CacheLevel::L1I);
+        l1_i_cache_line_size = get_cache_coherency_line_size(CacheLevel::L1I);
+        l1_i_cache_sets = get_cache_number_of_sets(CacheLevel::L1I);
+
+        // L2 Cache Info.
+        l2_cache_size = get_cache_size(CacheLevel::L2);
+        l2_cache_line_size = get_cache_coherency_line_size(CacheLevel::L2);
+        l2_cache_sets = get_cache_number_of_sets(CacheLevel::L2);
+
+        // L3 Cache Info.
+        l3_cache_size = get_cache_size(CacheLevel::L3);
+        l3_cache_line_size = get_cache_coherency_line_size(CacheLevel::L3);
+        l3_cache_sets = get_cache_number_of_sets(CacheLevel::L3);
+    }
+
     for (cpu_id, mpidr) in vcpu_mpidr.iter().enumerate().take(num_cpus) {
         let cpu_name = format!("cpu@{cpu_id:x}");
         let cpu_node = fdt.begin_node(&cpu_name)?;
@@ -173,6 +318,21 @@ fn create_cpu_nodes(
         fdt.property_u32("reg", (mpidr & 0x7FFFFF) as u32)?;
         fdt.property_u32("phandle", cpu_id as u32 + FIRST_VCPU_PHANDLE)?;
 
+        if cache_exist && l1_d_cache_size != 0 && l1_i_cache_size != 0 {
+            // Add cache info.
+            fdt.property_u32("d-cache-size", l1_d_cache_size)?;
+            fdt.property_u32("d-cache-line-size", l1_d_cache_line_size)?;
+            fdt.property_u32("d-cache-sets", l1_d_cache_sets)?;
+
+            fdt.property_u32("i-cache-size", l1_i_cache_size)?;
+            fdt.property_u32("i-cache-line-size", l1_i_cache_line_size)?;
+            fdt.property_u32("i-cache-sets", l1_i_cache_sets)?;
+
+            if l2_cache_size != 0 {
+                fdt.property_u32("next-level-cache", L2_CACHE_PHANDLE)?;
+            }
+        }
+
         // Add `numa-node-id` property if there is any numa config.
         if numa_nodes.len() > 1 {
             for numa_node_idx in 0..numa_nodes.len() {
@@ -186,6 +346,36 @@ fn create_cpu_nodes(
         fdt.end_node(cpu_node)?;
     }
 
+    if cache_exist && l2_cache_size != 0 {
+        let l2_cache_name = "l2-cache0";
+        let l2_cache_node = fdt.begin_node(l2_cache_name)?;
+        fdt.property_u32("phandle", L2_CACHE_PHANDLE)?;
+        fdt.property_string("compatible", "cache")?;
+        fdt.property_u32("cache-size", l2_cache_size)?;
+        fdt.property_u32("cache-line-size", l2_cache_line_size)?;
+        fdt.property_u32("cache-sets", l2_cache_sets)?;
+        fdt.property_u32("cache-level", 2)?;
+
+        if l3_cache_size != 0 {
+            fdt.property_u32("next-level-cache", L3_CACHE_PHANDLE)?;
+        }
+
+        fdt.end_node(l2_cache_node)?;
+    }
+
+    if cache_exist && l3_cache_size != 0 {
+        let l3_cache_name = "l3-cache0";
+        let l3_cache_node = fdt.begin_node(l3_cache_name)?;
+        fdt.property_u32("phandle", L3_CACHE_PHANDLE)?;
+        fdt.property_string("compatible", "cache")?;
+        fdt.property_null("cache-unified")?;
+        fdt.property_u32("cache-size", l3_cache_size)?;
+        fdt.property_u32("cache-line-size", l3_cache_line_size)?;
+        fdt.property_u32("cache-sets", l3_cache_sets)?;
+        fdt.property_u32("cache-level", 3)?;
+        fdt.end_node(l3_cache_node)?;
+    }
+
     if let Some(topology) = vcpu_topology {
         let (threads_per_core, cores_per_package, packages) = topology;
         let cpu_map_node = fdt.begin_node("cpu-map")?;
@@ -260,32 +450,86 @@ fn create_memory_node(
             fdt.end_node(memory_node)?;
         }
     } else {
-        let last_addr = guest_mem.last_addr().raw_value();
-        if last_addr < super::layout::MEM_32BIT_RESERVED_START.raw_value() {
-            // Case 1: all RAM is under the hole
-            let mem_size = last_addr - super::layout::RAM_START.raw_value() + 1;
-            let mem_reg_prop = [super::layout::RAM_START.raw_value(), mem_size];
-            let memory_node = fdt.begin_node("memory")?;
-            fdt.property_string("device_type", "memory")?;
-            fdt.property_array_u64("reg", &mem_reg_prop)?;
-            fdt.end_node(memory_node)?;
-        } else {
-            // Case 2: RAM is split by the hole
-            // Region 1: RAM before the hole
-            let mem_size = super::layout::MEM_32BIT_RESERVED_START.raw_value()
-                - super::layout::RAM_START.raw_value();
-            let mem_reg_prop = [super::layout::RAM_START.raw_value(), mem_size];
-            let memory_node_name = format!("memory@{:x}", super::layout::RAM_START.raw_value());
+        // 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
+        };
+
+        if ram_regions.len() > 2 {
+            panic!(
+                "There should be up to two non-continuous regions, devidided by the
+                    gap at the end of 32bit address space."
+            );
+        }
+
+        // Create the memory node for memory region before the gap
+        {
+            let (first_region_start, first_region_end) = ram_regions
+                .first()
+                .expect("There should be at last one memory region");
+            let ram_start = super::layout::RAM_START.raw_value();
+            let mem_32bit_reserved_start = super::layout::MEM_32BIT_RESERVED_START.raw_value();
+
+            if !((first_region_start <= &ram_start)
+                && (first_region_end > &ram_start)
+                && (first_region_end <= &mem_32bit_reserved_start))
+            {
+                panic!(
+                    "Unexpected first memory region layout: (start: 0x{:08x}, end: 0x{:08x}).
+                    ram_start: 0x{:08x}, mem_32bit_reserved_start: 0x{:08x}",
+                    first_region_start, first_region_end, ram_start, mem_32bit_reserved_start
+                );
+            }
+
+            let mem_size = first_region_end - ram_start;
+            let mem_reg_prop = [ram_start, mem_size];
+            let memory_node_name = format!("memory@{:x}", ram_start);
             let memory_node = fdt.begin_node(&memory_node_name)?;
             fdt.property_string("device_type", "memory")?;
             fdt.property_array_u64("reg", &mem_reg_prop)?;
             fdt.end_node(memory_node)?;
+        }
 
-            // Region 2: RAM after the hole
-            let mem_size = last_addr - super::layout::RAM_64BIT_START.raw_value() + 1;
-            let mem_reg_prop = [super::layout::RAM_64BIT_START.raw_value(), mem_size];
-            let memory_node_name =
-                format!("memory@{:x}", super::layout::RAM_64BIT_START.raw_value());
+        // Create the memory map entry for memory region after the gap if any
+        if let Some((second_region_start, second_region_end)) = ram_regions.get(1) {
+            let ram_64bit_start = super::layout::RAM_64BIT_START.raw_value();
+
+            if second_region_start != &ram_64bit_start {
+                panic!(
+                    "Unexpected second memory region layout: start: 0x{:08x}, ram_64bit_start: 0x{:08x}",
+                    second_region_start, ram_64bit_start
+                );
+            }
+
+            let mem_size = second_region_end - ram_64bit_start;
+            let mem_reg_prop = [ram_64bit_start, mem_size];
+            let memory_node_name = format!("memory@{:x}", ram_64bit_start);
             let memory_node = fdt.begin_node(&memory_node_name)?;
             fdt.property_string("device_type", "memory")?;
             fdt.property_array_u64("reg", &mem_reg_prop)?;

arch/src/aarch64/mod.rs

@@ -19,7 +19,7 @@ use std::collections::HashMap;
 use std::convert::TryInto;
 use std::fmt::Debug;
 use std::sync::{Arc, Mutex};
-use vm_memory::{Address, GuestAddress, GuestMemory, GuestMemoryAtomic, GuestUsize};
+use vm_memory::{Address, GuestAddress, GuestMemory, GuestMemoryAtomic};
 
 pub const _NSIG: i32 = 65;
 
@@ -83,8 +83,8 @@ pub fn configure_vcpu(
     Ok(mpidr)
 }
 
-pub fn arch_memory_regions(size: GuestUsize) -> Vec<(GuestAddress, usize, RegionType)> {
-    let mut regions = vec![
+pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
+    vec![
         // 0 MiB ~ 256 MiB: UEFI, GIC and legacy devices
         (
             GuestAddress(0),
@@ -103,39 +103,21 @@ pub fn arch_memory_regions(size: GuestUsize) -> Vec<(GuestAddress, usize, Region
             layout::PCI_MMCONFIG_SIZE as usize,
             RegionType::Reserved,
         ),
-    ];
-
-    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 <= 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((
+        // 1GiB ~ 4032 MiB: RAM before the gap
+        (
             layout::RAM_START,
-            ram_32bit_space_size as usize,
+            layout::MEM_32BIT_RESERVED_START.unchecked_offset_from(layout::RAM_START) as usize,
             RegionType::Ram,
-        ));
-        // 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
+        ),
+        // 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,
+        ),
+    ]
 }
 
 /// Configures the system and should be called once per vm before starting vcpu threads.
@@ -198,11 +180,8 @@ pub fn initramfs_load_addr(
     }
 }
 
-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();
+pub fn get_host_cpu_phys_bits(hypervisor: &Arc<dyn hypervisor::Hypervisor>) -> u8 {
+    let host_cpu_phys_bits = hypervisor.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.
@@ -217,26 +196,12 @@ mod tests {
     use super::*;
 
     #[test]
-    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;
+    fn test_arch_memory_regions_dram() {
+        let regions = arch_memory_regions();
         assert_eq!(6, regions.len());
         assert_eq!(layout::RAM_START, regions[3].0);
-        assert_eq!(ram_32bit_space_size, 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/x86_64/mod.rs

@@ -16,7 +16,7 @@ use crate::GuestMemoryMmap;
 use crate::InitramfsConfig;
 use crate::RegionType;
 use hypervisor::arch::x86::{CpuIdEntry, CPUID_FLAG_VALID_INDEX};
-use hypervisor::{HypervisorCpuError, HypervisorError};
+use hypervisor::{CpuVendor, HypervisorCpuError, HypervisorError};
 use linux_loader::loader::bootparam::boot_params;
 use linux_loader::loader::elf::start_info::{
     hvm_memmap_table_entry, hvm_modlist_entry, hvm_start_info,
@@ -39,7 +39,6 @@ const MTRR_EDX_BIT: u8 = 12; // Hypervisor ecx bit.
 const INVARIANT_TSC_EDX_BIT: u8 = 8; // Invariant TSC bit on 0x8000_0007 EDX
 
 // KVM feature bits
-const KVM_FEATURE_ASYNC_PF_INT_BIT: u8 = 14;
 #[cfg(feature = "tdx")]
 const KVM_FEATURE_CLOCKSOURCE_BIT: u8 = 0;
 #[cfg(feature = "tdx")]
@@ -210,7 +209,6 @@ impl From<Error> for super::Error {
     }
 }
 
-#[allow(clippy::upper_case_acronyms)]
 #[derive(Copy, Clone, Debug)]
 pub enum CpuidReg {
     EAX,
@@ -676,14 +674,7 @@ pub fn generate_common_cpuid(
             0x8000_0008 => {
                 entry.eax = (entry.eax & 0xffff_ff00) | (phys_bits as u32 & 0xff);
             }
-            // Disable KVM_FEATURE_ASYNC_PF_INT
-            // This is required until we find out why the asynchronous page
-            // fault is generating unexpected behavior when using interrupt
-            // mechanism.
-            // TODO: Re-enable KVM_FEATURE_ASYNC_PF_INT (#2277)
             0x4000_0001 => {
-                entry.eax &= !(1 << KVM_FEATURE_ASYNC_PF_INT_BIT);
-
                 // These features are not supported by TDX
                 #[cfg(feature = "tdx")]
                 if tdx_enabled {
@@ -776,6 +767,13 @@ pub fn configure_vcpu(
     CpuidPatch::set_cpuid_reg(&mut cpuid, 0xb, None, CpuidReg::EDX, u32::from(id));
     CpuidPatch::set_cpuid_reg(&mut cpuid, 0x1f, None, CpuidReg::EDX, u32::from(id));
 
+    // Set ApicId in cpuid for each vcpu
+    // SAFETY: get host cpuid when eax=1
+    let mut cpu_ebx = unsafe { core::arch::x86_64::__cpuid(1) }.ebx;
+    cpu_ebx &= 0xffffff;
+    cpu_ebx |= (id as u32) << 24;
+    CpuidPatch::set_cpuid_reg(&mut cpuid, 0x1, None, CpuidReg::EBX, cpu_ebx);
+
     // The TSC frequency CPUID leaf should not be included when running with HyperV emulation
     if !kvm_hyperv {
         if let Some(tsc_khz) = vcpu.tsc_khz().map_err(Error::GetTscFrequency)? {
@@ -828,47 +826,29 @@ pub fn configure_vcpu(
 /// These should be used to configure the GuestMemory structure for the platform.
 /// For x86_64 all addresses are valid from the start of the kernel except a
 /// carve out at the end of 32bit address space.
-pub fn arch_memory_regions(size: GuestUsize) -> Vec<(GuestAddress, usize, RegionType)> {
-    let reserved_memory_gap_start = layout::MEM_32BIT_RESERVED_START
-        .checked_add(layout::MEM_32BIT_DEVICES_SIZE)
-        .expect("32-bit reserved region is too large");
-
-    let requested_memory_size = GuestAddress(size);
-    let mut regions = Vec::new();
-
-    // case1: guest memory fits before the gap
-    if size <= layout::MEM_32BIT_RESERVED_START.raw_value() {
-        regions.push((GuestAddress(0), size as usize, RegionType::Ram));
-    // case2: guest memory extends beyond the gap
-    } else {
-        // push memory before the gap
-        regions.push((
+pub fn arch_memory_regions() -> Vec<(GuestAddress, usize, RegionType)> {
+    vec![
+        // 0 GiB ~ 3GiB: memory before the gap
+        (
             GuestAddress(0),
             layout::MEM_32BIT_RESERVED_START.raw_value() as usize,
             RegionType::Ram,
-        ));
-        regions.push((
-            layout::RAM_64BIT_START,
-            requested_memory_size.unchecked_offset_from(layout::MEM_32BIT_RESERVED_START) as usize,
-            RegionType::Ram,
-        ));
-    }
-
-    // Add the 32-bit device memory hole as a sub region.
-    regions.push((
-        layout::MEM_32BIT_RESERVED_START,
-        layout::MEM_32BIT_DEVICES_SIZE as usize,
-        RegionType::SubRegion,
-    ));
-
-    // Add the 32-bit reserved memory hole as a sub region.
-    regions.push((
-        reserved_memory_gap_start,
-        (layout::MEM_32BIT_RESERVED_SIZE - layout::MEM_32BIT_DEVICES_SIZE) as usize,
-        RegionType::Reserved,
-    ));
-
-    regions
+        ),
+        // 4 GiB ~ inf: memory after the gap
+        (layout::RAM_64BIT_START, usize::MAX, RegionType::Ram),
+        // 3 GiB ~ 3712 MiB: 32-bit device memory hole
+        (
+            layout::MEM_32BIT_RESERVED_START,
+            layout::MEM_32BIT_DEVICES_SIZE as usize,
+            RegionType::SubRegion,
+        ),
+        // 3712 MiB ~ 3968 MiB: 32-bit reserved memory hole
+        (
+            layout::MEM_32BIT_RESERVED_START.unchecked_add(layout::MEM_32BIT_DEVICES_SIZE),
+            (layout::MEM_32BIT_RESERVED_SIZE - layout::MEM_32BIT_DEVICES_SIZE) as usize,
+            RegionType::Reserved,
+        ),
+    ]
 }
 
 /// Configures the system and should be called once per vm before starting vcpu threads.
@@ -966,30 +946,102 @@ fn configure_pvh(
     // Create the memory map entries.
     add_memmap_entry(&mut memmap, 0, layout::EBDA_START.raw_value(), E820_RAM);
 
-    let mem_end = guest_mem.last_addr();
+    // Merge continuous memory regions into one region.
+    // 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;
 
-    if mem_end < layout::MEM_32BIT_RESERVED_START {
-        add_memmap_entry(
-            &mut memmap,
-            layout::HIGH_RAM_START.raw_value(),
-            mem_end.unchecked_offset_from(layout::HIGH_RAM_START) + 1,
-            E820_RAM,
-        );
-    } else {
-        add_memmap_entry(
-            &mut memmap,
-            layout::HIGH_RAM_START.raw_value(),
-            layout::MEM_32BIT_RESERVED_START.unchecked_offset_from(layout::HIGH_RAM_START),
-            E820_RAM,
-        );
-        if mem_end > layout::RAM_64BIT_START {
-            add_memmap_entry(
-                &mut memmap,
-                layout::RAM_64BIT_START.raw_value(),
-                mem_end.unchecked_offset_from(layout::RAM_64BIT_START) + 1,
-                E820_RAM,
-            );
+        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
+    };
+
+    if ram_regions.len() > 2 {
+        error!(
+            "There should be up to two non-continuous regions, devidided by the
+            gap at the end of 32bit address space (e.g. between 3G and 4G)."
+        );
+        return Err(super::Error::MemmapTableSetup);
+    }
+
+    // Create the memory map entry for memory region before the gap
+    {
+        let (first_region_start, first_region_end) =
+            ram_regions.first().ok_or(super::Error::MemmapTableSetup)?;
+        let high_ram_start = layout::HIGH_RAM_START.raw_value();
+        let mem_32bit_reserved_start = layout::MEM_32BIT_RESERVED_START.raw_value();
+
+        if !((first_region_start <= &high_ram_start)
+            && (first_region_end > &high_ram_start)
+            && (first_region_end <= &mem_32bit_reserved_start))
+        {
+            error!(
+                "Unexpected first memory region layout: (start: 0x{:08x}, end: 0x{:08x}).
+                high_ram_start: 0x{:08x}, mem_32bit_reserved_start: 0x{:08x}",
+                first_region_start, first_region_end, high_ram_start, mem_32bit_reserved_start
+            );
+
+            return Err(super::Error::MemmapTableSetup);
+        }
+
+        info!(
+            "create_memmap_entry, start: 0x{:08x}, end: 0x{:08x}",
+            high_ram_start, first_region_end
+        );
+
+        add_memmap_entry(
+            &mut memmap,
+            high_ram_start,
+            first_region_end - high_ram_start,
+            E820_RAM,
+        );
+    }
+
+    // Create the memory map entry for memory region after the gap if any
+    if let Some((second_region_start, second_region_end)) = ram_regions.get(1) {
+        let ram_64bit_start = layout::RAM_64BIT_START.raw_value();
+
+        if second_region_start != &ram_64bit_start {
+            error!(
+                "Unexpected second memory region layout: start: 0x{:08x}, ram_64bit_start: 0x{:08x}",
+                second_region_start, ram_64bit_start
+            );
+
+            return Err(super::Error::MemmapTableSetup);
+        }
+
+        info!(
+            "create_memmap_entry, start: 0x{:08x}, end: 0x{:08x}",
+            ram_64bit_start, second_region_end
+        );
+        add_memmap_entry(
+            &mut memmap,
+            ram_64bit_start,
+            second_region_end - ram_64bit_start,
+            E820_RAM,
+        );
     }
 
     add_memmap_entry(
@@ -1079,7 +1131,7 @@ pub fn initramfs_load_addr(
     Ok(aligned_addr)
 }
 
-pub fn get_host_cpu_phys_bits() -> u8 {
+pub fn get_host_cpu_phys_bits(hypervisor: &Arc<dyn hypervisor::Hypervisor>) -> u8 {
     // SAFETY: call cpuid with valid leaves
     unsafe {
         let leaf = x86_64::__cpuid(0x8000_0000);
@@ -1088,9 +1140,7 @@ pub fn get_host_cpu_phys_bits() -> u8 {
         // Some physical address bits may become reserved when the feature is enabled.
         // See AMD64 Architecture Programmer's Manual Volume 2, Section 7.10.1
         let reduced = if leaf.eax >= 0x8000_001f
-            && leaf.ebx == 0x6874_7541    // Vendor ID: AuthenticAMD
-            && leaf.ecx == 0x444d_4163
-            && leaf.edx == 0x6974_6e65
+            && matches!(hypervisor.get_cpu_vendor(), CpuVendor::AMD)
             && x86_64::__cpuid(0x8000_001f).eax & 0x1 != 0
         {
             (x86_64::__cpuid(0x8000_001f).ebx >> 6) & 0x3f
@@ -1225,16 +1275,8 @@ mod tests {
     use super::*;
 
     #[test]
-    fn regions_lt_4gb() {
-        let regions = arch_memory_regions(1 << 29);
-        assert_eq!(3, regions.len());
-        assert_eq!(GuestAddress(0), regions[0].0);
-        assert_eq!(1usize << 29, regions[0].1);
-    }
-
-    #[test]
-    fn regions_gt_4gb() {
-        let regions = arch_memory_regions((1 << 32) + 0x8000);
+    fn regions_base_addr() {
+        let regions = arch_memory_regions();
         assert_eq!(4, regions.len());
         assert_eq!(GuestAddress(0), regions[0].0);
         assert_eq!(GuestAddress(1 << 32), regions[1].0);
@@ -1258,11 +1300,10 @@ mod tests {
         assert!(config_err.is_err());
 
         // Now assigning some memory that falls before the 32bit memory hole.
-        let mem_size = 128 << 20;
-        let arch_mem_regions = arch_memory_regions(mem_size);
+        let arch_mem_regions = arch_memory_regions();
         let ram_regions: Vec<(GuestAddress, usize)> = arch_mem_regions
             .iter()
-            .filter(|r| r.2 == RegionType::Ram)
+            .filter(|r| r.2 == RegionType::Ram && r.1 != usize::MAX)
             .map(|r| (r.0, r.1))
             .collect();
         let gm = GuestMemoryMmap::from_ranges(&ram_regions).unwrap();
@@ -1280,48 +1321,18 @@ mod tests {
         )
         .unwrap();
 
-        // Now assigning some memory that is equal to the start of the 32bit memory hole.
-        let mem_size = 3328 << 20;
-        let arch_mem_regions = arch_memory_regions(mem_size);
-        let ram_regions: Vec<(GuestAddress, usize)> = arch_mem_regions
-            .iter()
-            .filter(|r| r.2 == RegionType::Ram)
-            .map(|r| (r.0, r.1))
-            .collect();
-        let gm = GuestMemoryMmap::from_ranges(&ram_regions).unwrap();
-        configure_system(
-            &gm,
-            GuestAddress(0),
-            &None,
-            no_vcpus,
-            None,
-            None,
-            None,
-            None,
-            None,
-        )
-        .unwrap();
-
-        configure_system(
-            &gm,
-            GuestAddress(0),
-            &None,
-            no_vcpus,
-            None,
-            None,
-            None,
-            None,
-            None,
-        )
-        .unwrap();
-
         // Now assigning some memory that falls after the 32bit memory hole.
-        let mem_size = 3330 << 20;
-        let arch_mem_regions = arch_memory_regions(mem_size);
+        let arch_mem_regions = arch_memory_regions();
         let ram_regions: Vec<(GuestAddress, usize)> = arch_mem_regions
             .iter()
             .filter(|r| r.2 == RegionType::Ram)
-            .map(|r| (r.0, r.1))
+            .map(|r| {
+                if r.1 == usize::MAX {
+                    (r.0, 128 << 20)
+                } else {
+                    (r.0, r.1)
+                }
+            })
             .collect();
         let gm = GuestMemoryMmap::from_ranges(&ram_regions).unwrap();
         configure_system(

block/Cargo.toml

@@ -0,0 +1,27 @@
+[package]
+name = "block"
+version = "0.1.0"
+edition = "2021"
+authors = ["The Cloud Hypervisor Authors", "The Chromium OS Authors"]
+
+[features]
+default = []
+io_uring = ["dep:io-uring"]
+
+[dependencies]
+byteorder = "1.4.3"
+crc32c = "0.6.3"
+io-uring = { version = "0.6.0", optional = true }
+libc = "0.2.139"
+log = "0.4.17"
+remain = "0.2.11"
+smallvec = "1.10.0"
+thiserror = "1.0.40"
+uuid = { version = "1.3.4", features = ["v4"] }
+versionize = "0.1.10"
+versionize_derive = "0.1.4"
+virtio-bindings = { version = "0.2.0", features = ["virtio-v5_0_0"] }
+virtio-queue = "0.8.0"
+vm-memory = { version = "0.11.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
+vm-virtio = { path = "../vm-virtio" }
+vmm-sys-util = "0.11.0"

block/src/fixed_vhd.rs

@@ -0,0 +1,90 @@
+// Copyright © 2021 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+
+use crate::vhd::VhdFooter;
+use crate::BlockBackend;
+use std::fs::File;
+use std::io::{Read, Seek, SeekFrom, Write};
+use std::os::unix::io::{AsRawFd, RawFd};
+
+#[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 size(&self) -> std::result::Result<u64, crate::Error> {
+        Ok(self.size)
+    }
+}
+
+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_async.rs

@@ -5,36 +5,29 @@
 use crate::async_io::{
     AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult,
 };
+use crate::fixed_vhd::FixedVhd;
 use crate::raw_async::RawFileAsync;
-use crate::vhd::VhdFooter;
+use crate::BlockBackend;
 use std::fs::File;
 use std::os::unix::io::{AsRawFd, RawFd};
 use vmm_sys_util::eventfd::EventFd;
 
-pub struct FixedVhdDiskAsync {
-    file: File,
-    size: u64,
-}
+pub struct FixedVhdDiskAsync(FixedVhd);
 
 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(),
-        })
+    pub fn new(file: File) -> std::io::Result<Self> {
+        Ok(Self(FixedVhd::new(file)?))
     }
 }
 
 impl DiskFile for FixedVhdDiskAsync {
     fn size(&mut self) -> DiskFileResult<u64> {
-        Ok(self.size)
+        Ok(self.0.size().unwrap())
     }
 
     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)
+            FixedVhdAsync::new(self.0.as_raw_fd(), ring_depth, self.0.size().unwrap())
                 .map_err(DiskFileError::NewAsyncIo)?,
         ) as Box<dyn AsyncIo>)
     }

block/src/fixed_vhd_sync.rs

@@ -5,36 +5,29 @@
 use crate::async_io::{
     AsyncIo, AsyncIoError, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult,
 };
+use crate::fixed_vhd::FixedVhd;
 use crate::raw_sync::RawFileSync;
-use crate::vhd::VhdFooter;
+use crate::BlockBackend;
 use std::fs::File;
 use std::os::unix::io::{AsRawFd, RawFd};
 use vmm_sys_util::eventfd::EventFd;
 
-pub struct FixedVhdDiskSync {
-    file: File,
-    size: u64,
-}
+pub struct FixedVhdDiskSync(FixedVhd);
 
 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(),
-        })
+    pub fn new(file: File) -> std::io::Result<Self> {
+        Ok(Self(FixedVhd::new(file)?))
     }
 }
 
 impl DiskFile for FixedVhdDiskSync {
     fn size(&mut self) -> DiskFileResult<u64> {
-        Ok(self.size)
+        Ok(self.0.size().unwrap())
     }
 
     fn new_async_io(&self, _ring_depth: u32) -> DiskFileResult<Box<dyn AsyncIo>> {
         Ok(Box::new(
-            FixedVhdSync::new(self.file.as_raw_fd(), self.size)
+            FixedVhdSync::new(self.0.as_raw_fd(), self.0.size().unwrap())
                 .map_err(DiskFileError::NewAsyncIo)?,
         ) as Box<dyn AsyncIo>)
     }

block/src/lib.rs

@@ -12,21 +12,35 @@
 extern crate log;
 
 pub mod async_io;
+pub mod fixed_vhd;
+#[cfg(feature = "io_uring")]
+/// Enabled with the `"io_uring"` feature
 pub mod fixed_vhd_async;
 pub mod fixed_vhd_sync;
+pub mod qcow;
 pub mod qcow_sync;
+#[cfg(feature = "io_uring")]
+/// Async primitives based on `io-uring`
+///
+/// Enabled with the `"io_uring"` feature
 pub mod raw_async;
 pub mod raw_sync;
 pub mod vhd;
+pub mod vhdx;
 pub mod vhdx_sync;
 
 use crate::async_io::{AsyncIo, AsyncIoError, AsyncIoResult};
+use crate::fixed_vhd::FixedVhd;
+use crate::qcow::{QcowFile, RawFile};
+use crate::vhdx::{Vhdx, VhdxError};
+#[cfg(feature = "io_uring")]
 use io_uring::{opcode, IoUring, Probe};
 use smallvec::SmallVec;
 use std::alloc::{alloc_zeroed, dealloc, Layout};
 use std::cmp;
 use std::collections::VecDeque;
 use std::convert::TryInto;
+use std::fmt::Debug;
 use std::fs::File;
 use std::io::{self, IoSlice, IoSliceMut, Read, Seek, SeekFrom, Write};
 use std::os::linux::fs::MetadataExt;
@@ -66,12 +80,22 @@ pub enum Error {
     DescriptorChainTooShort,
     #[error("Guest gave us a descriptor that was too short to use")]
     DescriptorLengthTooSmall,
+    #[error("Failed to detect image type: {0}")]
+    DetectImageType(std::io::Error),
+    #[error("Failure in fixed vhd: {0}")]
+    FixedVhdError(std::io::Error),
     #[error("Getting a block's metadata fails for any reason")]
     GetFileMetadata,
     #[error("The requested operation would cause a seek beyond disk end")]
     InvalidOffset,
+    #[error("Failure in qcow: {0}")]
+    QcowError(qcow::Error),
+    #[error("Failure in raw file: {0}")]
+    RawFileError(std::io::Error),
     #[error("The requested operation does not support multiple descriptors")]
     TooManyDescriptors,
+    #[error("Failure in vhdx: {0}")]
+    VhdxError(VhdxError),
 }
 
 fn build_device_id(disk_path: &Path) -> result::Result<String, Error> {
@@ -543,50 +567,59 @@ 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:";
+    #[cfg(not(feature = "io_uring"))]
+    {
+        info!("io_uring is disabled by crate features");
+        false
+    }
 
-    // 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);
+    #[cfg(feature = "io_uring")]
+    {
+        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;
         }
-    };
 
-    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);
+        // Check IORING_OP_READV is supported
+        if !probe.is_supported(opcode::Readv::CODE) {
+            info!("{} IORING_OP_READV operation not supported", error_msg);
             return false;
         }
-    }
 
-    // Check IORING_OP_FSYNC is supported
-    if !probe.is_supported(opcode::Fsync::CODE) {
-        info!("{} IORING_OP_FSYNC operation not supported", error_msg);
-        return false;
-    }
+        // Check IORING_OP_WRITEV is supported
+        if !probe.is_supported(opcode::Writev::CODE) {
+            info!("{} IORING_OP_WRITEV operation not supported", error_msg);
+            return false;
+        }
 
-    // Check IORING_OP_READV is supported
-    if !probe.is_supported(opcode::Readv::CODE) {
-        info!("{} IORING_OP_READV operation not supported", error_msg);
-        return false;
+        true
     }
-
-    // Check IORING_OP_WRITEV is supported
-    if !probe.is_supported(opcode::Writev::CODE) {
-        info!("{} IORING_OP_WRITEV operation not supported", error_msg);
-        return false;
-    }
-
-    true
 }
 
 pub trait AsyncAdaptor<F>
@@ -720,3 +753,26 @@ pub fn detect_image_type(f: &mut File) -> std::io::Result<ImageType> {
 
     Ok(image_type)
 }
+
+pub trait BlockBackend: Read + Write + Seek + Send + Debug {
+    fn size(&self) -> Result<u64, Error>;
+}
+
+/// Inspect the image file type and create an appropriate disk file to match it.
+pub fn create_disk_file(mut file: File, direct_io: bool) -> Result<Box<dyn BlockBackend>, Error> {
+    let image_type = detect_image_type(&mut file).map_err(Error::DetectImageType)?;
+
+    Ok(match image_type {
+        ImageType::Qcow2 => {
+            Box::new(QcowFile::from(RawFile::new(file, direct_io)).map_err(Error::QcowError)?)
+                as Box<dyn BlockBackend>
+        }
+        ImageType::FixedVhd => {
+            Box::new(FixedVhd::new(file).map_err(Error::FixedVhdError)?) as Box<dyn BlockBackend>
+        }
+        ImageType::Vhdx => {
+            Box::new(Vhdx::new(file).map_err(Error::VhdxError)?) as Box<dyn BlockBackend>
+        }
+        ImageType::Raw => Box::new(RawFile::new(file, direct_io)) as Box<dyn BlockBackend>,
+    })
+}

block/src/qcow/mod.rs

@@ -2,40 +2,45 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE-BSD-3-Clause file.
 
-#[macro_use]
-extern crate log;
-
 mod qcow_raw_file;
 mod raw_file;
 mod refcount;
 mod vec_cache;
 
-use crate::qcow_raw_file::QcowRawFile;
-use crate::refcount::RefCount;
-use crate::vec_cache::{CacheMap, Cacheable, VecCache};
+use crate::qcow::{
+    qcow_raw_file::QcowRawFile,
+    refcount::RefCount,
+    vec_cache::{CacheMap, Cacheable, VecCache},
+};
+use crate::BlockBackend;
 use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
 use libc::{EINVAL, ENOSPC, ENOTSUP};
 use remain::sorted;
 use std::cmp::{max, min};
 use std::fmt::{self, Display};
+use std::fs::OpenOptions;
 use std::io::{self, Read, Seek, SeekFrom, Write};
 use std::mem::size_of;
+use std::str;
 use vmm_sys_util::{
     file_traits::FileSetLen, file_traits::FileSync, seek_hole::SeekHole, write_zeroes::PunchHole,
     write_zeroes::WriteZeroesAt,
 };
 
-pub use crate::raw_file::RawFile;
+pub use crate::qcow::raw_file::RawFile;
 
 #[sorted]
 #[derive(Debug)]
 pub enum Error {
-    BackingFilesNotSupported,
+    BackingFileIo(io::Error),
+    BackingFileOpen(Box<crate::Error>),
+    BackingFileTooLong(usize),
     CompressedBlocksNotSupported,
     EvictingCache(io::Error),
     FileTooBig(u64),
     GettingFileSize(io::Error),
     GettingRefcount(refcount::Error),
+    InvalidBackingFileName(str::Utf8Error),
     InvalidClusterIndex,
     InvalidClusterSize,
     InvalidIndex,
@@ -78,12 +83,17 @@ impl Display for Error {
 
         #[sorted]
         match self {
-            BackingFilesNotSupported => write!(f, "backing files not supported"),
+            BackingFileIo(e) => write!(f, "backing file io error: {}", e),
+            BackingFileOpen(e) => write!(f, "backing file open error: {}", *e),
+            BackingFileTooLong(len) => {
+                write!(f, "backing file name is too long: {} bytes over", len)
+            }
             CompressedBlocksNotSupported => write!(f, "compressed blocks not supported"),
             EvictingCache(e) => write!(f, "failed to evict cache: {e}"),
             FileTooBig(size) => write!(f, "file larger than max of {MAX_QCOW_FILE_SIZE}: {size}"),
             GettingFileSize(e) => write!(f, "failed to get file size: {e}"),
             GettingRefcount(e) => write!(f, "failed to get refcount: {e}"),
+            InvalidBackingFileName(e) => write!(f, "failed to parse filename: {}", e),
             InvalidClusterIndex => write!(f, "invalid cluster index"),
             InvalidClusterSize => write!(f, "invalid cluster size"),
             InvalidIndex => write!(f, "invalid index"),
@@ -153,8 +163,14 @@ const COMPRESSED_FLAG: u64 = 1 << 62;
 const CLUSTER_USED_FLAG: u64 = 1 << 63;
 const COMPATIBLE_FEATURES_LAZY_REFCOUNTS: u64 = 1;
 
+// 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(Copy, Clone, Debug)]
+#[derive(Clone, Debug)]
 pub struct QcowHeader {
     pub magic: u32,
     pub version: u32,
@@ -181,6 +197,9 @@ pub struct QcowHeader {
     pub autoclear_features: u64,
     pub refcount_order: u32,
     pub header_size: u32,
+
+    // Post-header entries
+    pub backing_file_path: Option<String>,
 }
 
 impl QcowHeader {
@@ -204,7 +223,7 @@ impl QcowHeader {
 
         let version = read_u32_from_file(f)?;
 
-        Ok(QcowHeader {
+        let mut header = QcowHeader {
             magic,
             version,
             backing_file_offset: read_u64_from_file(f)?,
@@ -243,24 +262,58 @@ impl QcowHeader {
             } else {
                 read_u32_from_file(f)?
             },
-        })
+            backing_file_path: None,
+        };
+        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)?;
+            header.backing_file_path = Some(
+                String::from_utf8(backing_file_name_bytes)
+                    .map_err(|err| Error::InvalidBackingFileName(err.utf8_error()))?,
+            );
+        }
+        Ok(header)
     }
 
-    /// Create a header for the given `size`.
-    pub fn create_for_size(version: u32, size: u64) -> QcowHeader {
+    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 {
+            if 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);
-        QcowHeader {
+        Ok(QcowHeader {
             magic: QCOW_MAGIC,
             version,
-            backing_file_offset: 0,
-            backing_file_size: 0,
+            backing_file_offset: (if backing_file.is_none() {
+                0
+            } else {
+                header_size
+            }) as u64,
+            backing_file_size: backing_file.map_or(0, |x| x.len()) as u32,
             cluster_bits: DEFAULT_CLUSTER_BITS,
             size,
             crypt_method: 0,
@@ -289,12 +342,9 @@ impl QcowHeader {
             compatible_features: 0,
             autoclear_features: 0,
             refcount_order: DEFAULT_REFCOUNT_ORDER,
-            header_size: if version == 2 {
-                V2_BARE_HEADER_SIZE
-            } else {
-                V3_BARE_HEADER_SIZE
-            },
-        }
+            header_size,
+            backing_file_path: backing_file.map(String::from),
+        })
     }
 
     /// Write the header to `file`.
@@ -324,11 +374,20 @@ impl QcowHeader {
         write_u32_to_file(file, self.refcount_table_clusters)?;
         write_u32_to_file(file, self.nb_snapshots)?;
         write_u64_to_file(file, self.snapshots_offset)?;
-        write_u64_to_file(file, self.incompatible_features)?;
-        write_u64_to_file(file, self.compatible_features)?;
-        write_u64_to_file(file, self.autoclear_features)?;
-        write_u32_to_file(file, self.refcount_order)?;
-        write_u32_to_file(file, self.header_size)?;
+
+        if self.version == 3 {
+            write_u64_to_file(file, self.incompatible_features)?;
+            write_u64_to_file(file, self.compatible_features)?;
+            write_u64_to_file(file, self.autoclear_features)?;
+            write_u32_to_file(file, self.refcount_order)?;
+            write_u32_to_file(file, self.header_size)?;
+            write_u32_to_file(file, 0)?; // header extension type: end of header extension area
+            write_u32_to_file(file, 0)?; // length of header extension data: 0
+        }
+
+        if let Some(backing_file_path) = self.backing_file_path.as_ref() {
+            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.
@@ -362,8 +421,8 @@ fn max_refcount_clusters(refcount_order: u32, cluster_size: u32, num_clusters: u
 /// # Example
 ///
 /// ```
+/// # use block::qcow::{self, QcowFile, RawFile};
 /// # use std::io::{Read, Seek, SeekFrom};
-/// # use qcow::{self, QcowFile, RawFile};
 /// # fn test(file: std::fs::File) -> std::io::Result<()> {
 ///     let mut raw_img = RawFile::new(file, false);
 ///     let mut q = QcowFile::from(raw_img).expect("Can't open qcow file");
@@ -373,7 +432,7 @@ fn max_refcount_clusters(refcount_order: u32, cluster_size: u32, num_clusters: u
 /// #   Ok(())
 /// # }
 /// ```
-#[derive(Clone, Debug)]
+#[derive(Debug)]
 pub struct QcowFile {
     raw_file: QcowRawFile,
     header: QcowHeader,
@@ -386,7 +445,7 @@ pub struct QcowFile {
     // List of unreferenced clusters available to be used. unref clusters become available once the
     // removal of references to them have been synced to disk.
     avail_clusters: Vec<u64>,
-    //TODO(dgreid) Add support for backing files. - backing_file: Option<Box<QcowFile<T>>>,
+    backing_file: Option<Box<dyn BlockBackend>>,
 }
 
 impl QcowFile {
@@ -415,10 +474,20 @@ impl QcowFile {
             return Err(Error::FileTooBig(header.size));
         }
 
-        // No current support for backing files.
-        if header.backing_file_offset != 0 {
-            return Err(Error::BackingFilesNotSupported);
-        }
+        let direct_io = file.is_direct();
+
+        let backing_file = if let Some(backing_file_path) = header.backing_file_path.as_ref() {
+            let path = backing_file_path.clone();
+            let backing_raw_file = OpenOptions::new()
+                .read(true)
+                .open(path)
+                .map_err(Error::BackingFileIo)?;
+            let backing_file = crate::create_disk_file(backing_raw_file, direct_io)
+                .map_err(|e| Error::BackingFileOpen(Box::new(e)))?;
+            Some(backing_file)
+        } else {
+            None
+        };
 
         // Only support two byte refcounts.
         let refcount_bits: u64 = 0x01u64
@@ -433,7 +502,6 @@ impl QcowFile {
         if header.refcount_table_clusters == 0 {
             return Err(Error::NoRefcountClusters);
         }
-        offset_is_cluster_boundary(header.backing_file_offset, header.cluster_bits)?;
         offset_is_cluster_boundary(header.l1_table_offset, header.cluster_bits)?;
         offset_is_cluster_boundary(header.snapshots_offset, header.cluster_bits)?;
         // refcount table must be a cluster boundary, and within the file's virtual or actual size.
@@ -466,7 +534,7 @@ impl QcowFile {
         let mut raw_file =
             QcowRawFile::from(file, cluster_size).ok_or(Error::InvalidClusterSize)?;
         if refcount_rebuild_required {
-            QcowFile::rebuild_refcounts(&mut raw_file, header)?;
+            QcowFile::rebuild_refcounts(&mut raw_file, header.clone())?;
         }
 
         let entries_per_cluster = cluster_size / size_of::<u64>() as u64;
@@ -535,6 +603,7 @@ impl QcowFile {
             current_offset: 0,
             unref_clusters: Vec::new(),
             avail_clusters: Vec::new(),
+            backing_file,
         };
 
         // Check that the L1 and refcount tables fit in a 64bit address space.
@@ -553,8 +622,34 @@ impl QcowFile {
     }
 
     /// Creates a new QcowFile at the given path.
-    pub fn new(mut file: RawFile, version: u32, virtual_size: u64) -> Result<QcowFile> {
-        let header = QcowHeader::create_for_size(version, virtual_size);
+    pub fn new(file: RawFile, version: u32, virtual_size: u64) -> Result<QcowFile> {
+        let header = QcowHeader::create_for_size_and_path(version, virtual_size, None)?;
+        QcowFile::new_from_header(file, header)
+    }
+
+    /// Creates a new QcowFile at the given path.
+    pub fn new_from_backing(
+        file: RawFile,
+        version: u32,
+        backing_file_name: &str,
+    ) -> Result<QcowFile> {
+        let direct_io = file.is_direct();
+        let backing_raw_file = OpenOptions::new()
+            .read(true)
+            .open(backing_file_name)
+            .map_err(Error::BackingFileIo)?;
+        let backing_file = crate::create_disk_file(backing_raw_file, direct_io)
+            .map_err(|e| Error::BackingFileOpen(Box::new(e)))?;
+        let size = backing_file
+            .size()
+            .map_err(|e| Error::BackingFileOpen(Box::new(e)))?;
+        let header = QcowHeader::create_for_size_and_path(version, size, Some(backing_file_name))?;
+        let mut result = QcowFile::new_from_header(file, header)?;
+        result.backing_file = Some(backing_file);
+        Ok(result)
+    }
+
+    fn new_from_header(mut file: RawFile, header: QcowHeader) -> Result<QcowFile> {
         file.rewind().map_err(Error::SeekingFile)?;
         header.write_to(&mut file)?;
 
@@ -578,6 +673,10 @@ impl QcowFile {
         Ok(qcow)
     }
 
+    pub fn set_backing_file(&mut self, backing: Option<Box<dyn BlockBackend>>) {
+        self.backing_file = backing;
+    }
+
     /// Returns the `QcowHeader` for this file.
     pub fn header(&self) -> &QcowHeader {
         &self.header
@@ -890,9 +989,9 @@ impl QcowFile {
 
         // Find all references clusters and rebuild refcounts.
         set_header_refcount(&mut refcounts, cluster_size)?;
-        set_l1_refcounts(&mut refcounts, header, cluster_size)?;
-        set_data_refcounts(&mut refcounts, header, cluster_size, raw_file)?;
-        set_refcount_table_refcounts(&mut refcounts, header, cluster_size)?;
+        set_l1_refcounts(&mut refcounts, header.clone(), cluster_size)?;
+        set_data_refcounts(&mut refcounts, header.clone(), cluster_size, raw_file)?;
+        set_refcount_table_refcounts(&mut refcounts, header.clone(), cluster_size)?;
 
         // Allocate clusters to store the new reference count blocks.
         let ref_table = alloc_refblocks(&mut refcounts, cluster_size, refblock_clusters)?;
@@ -1007,7 +1106,7 @@ impl QcowFile {
             let l2_table = if l2_addr_disk == 0 {
                 // Allocate a new cluster to store the L2 table and update the L1 table to point
                 // to the new table.
-                let new_addr: u64 = self.get_new_cluster()?;
+                let new_addr: u64 = self.get_new_cluster(None)?;
                 // The cluster refcount starts at one meaning it is used but doesn't need COW.
                 set_refcounts.push((new_addr, 1));
                 self.l1_table[l1_index] = new_addr;
@@ -1028,8 +1127,18 @@ impl QcowFile {
 
         let cluster_addr = match self.l2_cache.get(l1_index).unwrap()[l2_index] {
             0 => {
+                let initial_data = if let Some(backing) = self.backing_file.as_mut() {
+                    let cluster_size = self.raw_file.cluster_size();
+                    let cluster_begin = address - (address % cluster_size);
+                    let mut cluster_data = vec![0u8; cluster_size as usize];
+                    backing.seek(SeekFrom::Start(cluster_begin))?;
+                    backing.read_exact(&mut cluster_data)?;
+                    Some(cluster_data)
+                } else {
+                    None
+                };
                 // Need to allocate a data cluster
-                let cluster_addr = self.append_data_cluster()?;
+                let cluster_addr = self.append_data_cluster(initial_data)?;
                 self.update_cluster_addr(l1_index, l2_index, cluster_addr, &mut set_refcounts)?;
                 cluster_addr
             }
@@ -1066,7 +1175,7 @@ impl QcowFile {
             // Allocate a new cluster to store the L2 table and update the L1 table to point
             // to the new table. The cluster will be written when the cache is flushed, no
             // need to copy the data now.
-            let new_addr: u64 = self.get_new_cluster()?;
+            let new_addr: u64 = self.get_new_cluster(None)?;
             // The cluster refcount starts at one indicating it is used but doesn't need
             // COW.
             set_refcounts.push((new_addr, 1));
@@ -1078,15 +1187,22 @@ impl QcowFile {
     }
 
     // Allocate a new cluster and return its offset within the raw file.
-    fn get_new_cluster(&mut self) -> std::io::Result<u64> {
+    fn get_new_cluster(&mut self, initial_data: Option<Vec<u8>>) -> std::io::Result<u64> {
         // First use a pre allocated cluster if one is available.
         if let Some(free_cluster) = self.avail_clusters.pop() {
-            self.raw_file.zero_cluster(free_cluster)?;
+            if let Some(initial_data) = initial_data {
+                self.raw_file.write_cluster(free_cluster, initial_data)?;
+            } else {
+                self.raw_file.zero_cluster(free_cluster)?;
+            }
             return Ok(free_cluster);
         }
 
         let max_valid_cluster_offset = self.refcounts.max_valid_cluster_offset();
         if let Some(new_cluster) = self.raw_file.add_cluster_end(max_valid_cluster_offset)? {
+            if let Some(initial_data) = initial_data {
+                self.raw_file.write_cluster(new_cluster, initial_data)?;
+            }
             Ok(new_cluster)
         } else {
             error!("No free clusters in get_new_cluster()");
@@ -1096,8 +1212,8 @@ impl QcowFile {
 
     // Allocate and initialize a new data cluster. Returns the offset of the
     // cluster in to the file on success.
-    fn append_data_cluster(&mut self) -> std::io::Result<u64> {
-        let new_addr: u64 = self.get_new_cluster()?;
+    fn append_data_cluster(&mut self, initial_data: Option<Vec<u8>>) -> std::io::Result<u64> {
+        let new_addr: u64 = self.get_new_cluster(initial_data)?;
         // The cluster refcount starts at one indicating it is used but doesn't need COW.
         let mut newly_unref = self.set_cluster_refcount(new_addr, 1)?;
         self.unref_clusters.append(&mut newly_unref);
@@ -1331,7 +1447,7 @@ impl QcowFile {
                 }
                 Err(refcount::Error::NeedNewCluster) => {
                     // Allocate the cluster and call set_cluster_refcount again.
-                    let addr = self.get_new_cluster()?;
+                    let addr = self.get_new_cluster(None)?;
                     added_clusters.push(addr);
                     new_cluster = Some((
                         addr,
@@ -1414,6 +1530,9 @@ impl Read for QcowFile {
                 self.raw_file
                     .file_mut()
                     .read_exact(&mut buf[nread..(nread + count)])?;
+            } else if let Some(backing) = self.backing_file.as_mut() {
+                backing.seek(SeekFrom::Start(curr_addr))?;
+                backing.read_exact(&mut buf[nread..(nread + count)])?;
             } else {
                 // Previously unwritten region, return zeros
                 for b in &mut buf[nread..(nread + count)] {
@@ -1556,6 +1675,12 @@ impl SeekHole for QcowFile {
     }
 }
 
+impl BlockBackend for QcowFile {
+    fn size(&self) -> std::result::Result<u64, crate::Error> {
+        Ok(self.virtual_size())
+    }
+}
+
 // Returns an Error if the given offset doesn't align to a cluster boundary.
 fn offset_is_cluster_boundary(offset: u64, cluster_bits: u32) -> Result<()> {
     if offset & ((0x01 << cluster_bits) - 1) != 0 {
@@ -1766,16 +1891,19 @@ mod tests {
         ]
     }
 
-    fn with_basic_file<F>(header: &[u8], mut testfn: F)
-    where
-        F: FnMut(RawFile),
-    {
+    fn basic_file(header: &[u8]) -> RawFile {
         let mut disk_file: RawFile = RawFile::new(TempFile::new().unwrap().into_file(), false);
         disk_file.write_all(header).unwrap();
         disk_file.set_len(0x1_0000_0000).unwrap();
         disk_file.rewind().unwrap();
+        disk_file
+    }
 
-        testfn(disk_file); // File closed when the function exits.
+    fn with_basic_file<F>(header: &[u8], mut testfn: F)
+    where
+        F: FnMut(RawFile),
+    {
+        testfn(basic_file(header)); // File closed when the function exits.
     }
 
     fn with_default_file<F>(file_size: u64, direct: bool, mut testfn: F)
@@ -1788,11 +1916,44 @@ mod tests {
         testfn(qcow_file); // File closed when the function exits.
     }
 
+    #[test]
+    fn write_read_start_backing_v2() {
+        let disk_file = basic_file(&valid_header_v2());
+        let mut backing = QcowFile::from(disk_file).unwrap();
+        backing
+            .write_all(b"test first bytes")
+            .expect("Failed to write test string.");
+        let mut buf = [0u8; 4];
+        let wrapping_disk_file = basic_file(&valid_header_v2());
+        let mut wrapping = QcowFile::from(wrapping_disk_file).unwrap();
+        wrapping.set_backing_file(Some(Box::new(backing)));
+        wrapping.seek(SeekFrom::Start(0)).expect("Failed to seek.");
+        wrapping.read_exact(&mut buf).expect("Failed to read.");
+        assert_eq!(&buf, b"test");
+    }
+
+    #[test]
+    fn write_read_start_backing_v3() {
+        let disk_file = basic_file(&valid_header_v3());
+        let mut backing = QcowFile::from(disk_file).unwrap();
+        backing
+            .write_all(b"test first bytes")
+            .expect("Failed to write test string.");
+        let mut buf = [0u8; 4];
+        let wrapping_disk_file = basic_file(&valid_header_v3());
+        let mut wrapping = QcowFile::from(wrapping_disk_file).unwrap();
+        wrapping.set_backing_file(Some(Box::new(backing)));
+        wrapping.seek(SeekFrom::Start(0)).expect("Failed to seek.");
+        wrapping.read_exact(&mut buf).expect("Failed to read.");
+        assert_eq!(&buf, b"test");
+    }
+
     #[test]
     fn default_header_v2() {
-        let header = QcowHeader::create_for_size(2, 0x10_0000);
+        let header = QcowHeader::create_for_size_and_path(2, 0x10_0000, None);
         let mut disk_file: RawFile = RawFile::new(TempFile::new().unwrap().into_file(), false);
         header
+            .expect("Failed to create header.")
             .write_to(&mut disk_file)
             .expect("Failed to write header to temporary file.");
         disk_file.rewind().unwrap();
@@ -1801,9 +1962,10 @@ mod tests {
 
     #[test]
     fn default_header_v3() {
-        let header = QcowHeader::create_for_size(3, 0x10_0000);
+        let header = QcowHeader::create_for_size_and_path(3, 0x10_0000, None);
         let mut disk_file: RawFile = RawFile::new(TempFile::new().unwrap().into_file(), false);
         header
+            .expect("Failed to create header.")
             .write_to(&mut disk_file)
             .expect("Failed to write header to temporary file.");
         disk_file.rewind().unwrap();
@@ -1826,6 +1988,40 @@ mod tests {
         });
     }
 
+    #[test]
+    fn header_v2_with_backing() {
+        let header = QcowHeader::create_for_size_and_path(2, 0x10_0000, Some("/my/path/to/a/file"))
+            .expect("Failed to create header.");
+        let mut disk_file: RawFile = RawFile::new(TempFile::new().unwrap().into_file(), false);
+        header
+            .write_to(&mut disk_file)
+            .expect("Failed to write header to shm.");
+        disk_file.rewind().unwrap();
+        let read_header = QcowHeader::new(&mut disk_file).expect("Failed to create header.");
+        assert_eq!(
+            header.backing_file_path,
+            Some(String::from("/my/path/to/a/file"))
+        );
+        assert_eq!(read_header.backing_file_path, header.backing_file_path);
+    }
+
+    #[test]
+    fn header_v3_with_backing() {
+        let header = QcowHeader::create_for_size_and_path(3, 0x10_0000, Some("/my/path/to/a/file"))
+            .expect("Failed to create header.");
+        let mut disk_file: RawFile = RawFile::new(TempFile::new().unwrap().into_file(), false);
+        header
+            .write_to(&mut disk_file)
+            .expect("Failed to write header to shm.");
+        disk_file.rewind().unwrap();
+        let read_header = QcowHeader::new(&mut disk_file).expect("Failed to create header.");
+        assert_eq!(
+            header.backing_file_path,
+            Some(String::from("/my/path/to/a/file"))
+        );
+        assert_eq!(read_header.backing_file_path, header.backing_file_path);
+    }
+
     #[test]
     fn invalid_magic() {
         let invalid_header = vec![0x51u8, 0x46, 0x4a, 0xfb];
@@ -1940,6 +2136,26 @@ mod tests {
         });
     }
 
+    #[test]
+    fn write_read_start_backing_overlap() {
+        let disk_file = basic_file(&valid_header_v3());
+        let mut backing = QcowFile::from(disk_file).unwrap();
+        backing
+            .write_all(b"test first bytes")
+            .expect("Failed to write test string.");
+        let wrapping_disk_file = basic_file(&valid_header_v3());
+        let mut wrapping = QcowFile::from(wrapping_disk_file).unwrap();
+        wrapping.set_backing_file(Some(Box::new(backing)));
+        wrapping.seek(SeekFrom::Start(0)).expect("Failed to seek.");
+        wrapping
+            .write_all(b"TEST")
+            .expect("Failed to write second test string.");
+        let mut buf = [0u8; 10];
+        wrapping.seek(SeekFrom::Start(0)).expect("Failed to seek.");
+        wrapping.read_exact(&mut buf).expect("Failed to read.");
+        assert_eq!(&buf, b"TEST first");
+    }
+
     #[test]
     fn offset_write_read() {
         with_basic_file(&valid_header_v3(), |disk_file: RawFile| {

block/src/qcow/qcow_raw_file.rs

@@ -4,7 +4,7 @@
 
 use super::RawFile;
 use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
-use std::io::{self, BufWriter, Seek, SeekFrom};
+use std::io::{self, BufWriter, Seek, SeekFrom, Write};
 use std::mem::size_of;
 use vmm_sys_util::write_zeroes::WriteZeroes;
 
@@ -66,7 +66,7 @@ impl QcowRawFile {
         non_zero_flags: u64,
     ) -> io::Result<()> {
         self.file.seek(SeekFrom::Start(offset))?;
-        let mut buffer = BufWriter::with_capacity(table.len() * size_of::<u64>(), &mut self.file);
+        let mut buffer = BufWriter::with_capacity(std::mem::size_of_val(table), &mut self.file);
         for addr in table {
             let val = if *addr == 0 {
                 0
@@ -91,7 +91,8 @@ impl QcowRawFile {
     /// Writes a refcount block to the file.
     pub fn write_refcount_block(&mut self, offset: u64, table: &[u16]) -> io::Result<()> {
         self.file.seek(SeekFrom::Start(offset))?;
-        let mut buffer = BufWriter::with_capacity(table.len() * size_of::<u16>(), &mut self.file);
+        let mut buffer = BufWriter::with_capacity(std::mem::size_of_val(table), &mut self.file);
+
         for count in table {
             buffer.write_u16::<BigEndian>(*count)?;
         }
@@ -136,6 +137,13 @@ impl QcowRawFile {
         self.file.write_zeroes(cluster_size)?;
         Ok(())
     }
+
+    /// Writes
+    pub fn write_cluster(&mut self, address: u64, data: Vec<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])
+    }
 }
 
 impl Clone for QcowRawFile {

block/src/qcow/raw_file.rs

@@ -8,6 +8,7 @@
 //
 // SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
 
+use crate::BlockBackend;
 use libc::c_void;
 use std::alloc::{alloc_zeroed, dealloc, Layout};
 use std::convert::TryInto;
@@ -22,6 +23,7 @@ pub struct RawFile {
     file: File,
     alignment: usize,
     position: u64,
+    direct_io: bool,
 }
 
 const BLK_ALIGNMENTS: [usize; 2] = [512, 4096];
@@ -64,6 +66,7 @@ impl RawFile {
             file,
             alignment,
             position: 0,
+            direct_io,
         }
     }
 
@@ -102,6 +105,7 @@ impl RawFile {
             file: self.file.try_clone().expect("RawFile cloning failed"),
             alignment: self.alignment,
             position: self.position,
+            direct_io: self.direct_io,
         })
     }
 
@@ -112,6 +116,10 @@ impl RawFile {
     pub fn sync_data(&self) -> std::io::Result<()> {
         self.file.sync_data()
     }
+
+    pub fn is_direct(&self) -> bool {
+        self.direct_io
+    }
 }
 
 impl Read for RawFile {
@@ -343,12 +351,19 @@ impl SeekHole for RawFile {
     }
 }
 
+impl BlockBackend for RawFile {
+    fn size(&self) -> std::result::Result<u64, crate::Error> {
+        Ok(self.metadata().map_err(crate::Error::RawFileError)?.len())
+    }
+}
+
 impl Clone for RawFile {
     fn clone(&self) -> Self {
         RawFile {
             file: self.file.try_clone().expect("RawFile cloning failed"),
             alignment: self.alignment,
             position: self.position,
+            direct_io: self.direct_io,
         }
     }
 }

block/src/qcow/refcount.rs

@@ -7,8 +7,10 @@ use std::io;
 
 use libc::EINVAL;
 
-use crate::qcow_raw_file::QcowRawFile;
-use crate::vec_cache::{CacheMap, Cacheable, VecCache};
+use crate::qcow::{
+    qcow_raw_file::QcowRawFile,
+    vec_cache::{CacheMap, Cacheable, VecCache},
+};
 
 #[derive(Debug)]
 pub enum Error {

block/src/qcow_sync.rs

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause
 
 use crate::async_io::{AsyncIo, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult};
+use crate::qcow::{QcowFile, RawFile, Result as QcowResult};
 use crate::AsyncAdaptor;
-use qcow::{QcowFile, RawFile, Result as QcowResult};
 use std::collections::VecDeque;
 use std::fs::File;
 use std::io::{Seek, SeekFrom};

block/src/raw_async.rs

@@ -86,7 +86,7 @@ impl AsyncIo for RawFileAsync {
         let _ = unsafe {
             sq.push(
                 &opcode::Readv::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
-                    .offset(offset)
+                    .offset(offset.try_into().unwrap())
                     .build()
                     .flags(squeue::Flags::ASYNC)
                     .user_data(user_data),
@@ -114,7 +114,7 @@ impl AsyncIo for RawFileAsync {
         let _ = unsafe {
             sq.push(
                 &opcode::Writev::new(types::Fd(self.fd), iovecs.as_ptr(), iovecs.len() as u32)
-                    .offset(offset)
+                    .offset(offset.try_into().unwrap())
                     .build()
                     .flags(squeue::Flags::ASYNC)
                     .user_data(user_data),

block/src/raw_sync.rs

@@ -73,7 +73,7 @@ impl AsyncIo for RawFileSync {
         let result = unsafe {
             libc::preadv(
                 self.fd as libc::c_int,
-                iovecs.as_ptr() as *const libc::iovec,
+                iovecs.as_ptr(),
                 iovecs.len() as libc::c_int,
                 offset,
             )
@@ -98,7 +98,7 @@ impl AsyncIo for RawFileSync {
         let result = unsafe {
             libc::pwritev(
                 self.fd as libc::c_int,
-                iovecs.as_ptr() as *const libc::iovec,
+                iovecs.as_ptr(),
                 iovecs.len() as libc::c_int,
                 offset,
             )

block/src/vhdx/mod.rs

@@ -2,15 +2,31 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use crate::vhdx_bat::{BatEntry, VhdxBatError};
-use crate::vhdx_header::{self, RegionInfo, RegionTableEntry, VhdxHeader, VhdxHeaderError};
-use crate::vhdx_io::{self, VhdxIoError};
-use crate::vhdx_metadata::{DiskSpec, VhdxMetadataError};
+use crate::vhdx::{
+    vhdx_bat::{BatEntry, VhdxBatError},
+    vhdx_header::{RegionInfo, RegionTableEntry, VhdxHeader, VhdxHeaderError},
+    vhdx_io::VhdxIoError,
+    vhdx_metadata::{DiskSpec, VhdxMetadataError},
+};
+use crate::BlockBackend;
+use byteorder::{BigEndian, ByteOrder};
 use remain::sorted;
 use std::collections::btree_map::BTreeMap;
 use std::fs::File;
 use std::io::{Read, Seek, SeekFrom, Write};
 use thiserror::Error;
+use uuid::Uuid;
+
+macro_rules! div_round_up {
+    ($n:expr,$d:expr) => {
+        ($n + $d - 1) / $d
+    };
+}
+
+mod vhdx_bat;
+mod vhdx_header;
+mod vhdx_io;
+mod vhdx_metadata;
 
 #[sorted]
 #[derive(Error, Debug)]
@@ -193,6 +209,12 @@ impl Seek for Vhdx {
     }
 }
 
+impl BlockBackend for Vhdx {
+    fn size(&self) -> std::result::Result<u64, crate::Error> {
+        Ok(self.virtual_disk_size())
+    }
+}
+
 impl Clone for Vhdx {
     fn clone(&self) -> Self {
         Vhdx {
@@ -208,3 +230,15 @@ impl Clone for Vhdx {
         }
     }
 }
+
+pub(crate) fn uuid_from_guid(buf: &[u8]) -> Uuid {
+    // The first 3 fields of UUID are stored in Big Endian format, and
+    // the last 8 bytes are stored as byte array. Therefore, we read the
+    // first 3 fields in Big Endian format instead of Little Endian.
+    Uuid::from_fields_le(
+        BigEndian::read_u32(&buf[0..4]),
+        BigEndian::read_u16(&buf[4..6]),
+        BigEndian::read_u16(&buf[6..8]),
+        buf[8..16].try_into().unwrap(),
+    )
+}

block/src/vhdx/vhdx_bat.rs

@@ -2,8 +2,7 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use crate::vhdx_header::RegionTableEntry;
-use crate::vhdx_metadata::DiskSpec;
+use crate::vhdx::{vhdx_header::RegionTableEntry, vhdx_metadata::DiskSpec};
 use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
 use remain::sorted;
 use std::fs::File;

block/src/vhdx/vhdx_header.rs

@@ -344,7 +344,7 @@ impl RegionTableEntry {
         // SAFETY: the assertion above makes sure the buffer size is correct.
         let mut region_table_entry = unsafe { *(buffer.as_ptr() as *mut RegionTableEntry) };
 
-        let uuid = crate::uuid_from_guid(buffer);
+        let uuid = crate::vhdx::uuid_from_guid(buffer);
         region_table_entry.guid = uuid;
 
         Ok(region_table_entry)

block/src/vhdx/vhdx_io.rs

@@ -2,8 +2,10 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use crate::vhdx_bat::{self, BatEntry, VhdxBatError};
-use crate::vhdx_metadata::{self, DiskSpec};
+use crate::vhdx::{
+    vhdx_bat::{self, BatEntry, VhdxBatError},
+    vhdx_metadata::{self, DiskSpec},
+};
 use remain::sorted;
 use std::fs::File;
 use std::io::{self, Read, Seek, SeekFrom, Write};

block/src/vhdx/vhdx_metadata.rs

@@ -2,7 +2,7 @@
 //
 // SPDX-License-Identifier: Apache-2.0
 
-use crate::vhdx_header::RegionTableEntry;
+use crate::vhdx::vhdx_header::RegionTableEntry;
 use byteorder::{LittleEndian, ReadBytesExt};
 use remain::sorted;
 use std::fs::File;
@@ -145,11 +145,7 @@ impl DiskSpec {
                 let bits = f
                     .read_u32::<LittleEndian>()
                     .map_err(VhdxMetadataError::ReadMetadata)?;
-                if bits & BLOCK_HAS_PARENT != 0 {
-                    disk_spec.has_parent = true;
-                } else {
-                    disk_spec.has_parent = false;
-                }
+                disk_spec.has_parent = bits & BLOCK_HAS_PARENT != 0;
 
                 metadata_presence |= METADATA_FILE_PARAMETER_PRESENT;
             } else if metadata_entry.item_id
@@ -307,7 +303,7 @@ impl MetadataTableEntry {
         // SAFETY: the assertion above makes sure the buffer size is correct.
         let mut metadata_table_entry = unsafe { *(buffer.as_ptr() as *mut MetadataTableEntry) };
 
-        let uuid = crate::uuid_from_guid(buffer);
+        let uuid = crate::vhdx::uuid_from_guid(buffer);
         metadata_table_entry.item_id = uuid;
 
         if metadata_table_entry.length > METADATA_LENGTH_MAX {

block/src/vhdx_sync.rs

@@ -3,11 +3,11 @@
 // SPDX-License-Identifier: Apache-2.0
 
 use crate::async_io::{AsyncIo, AsyncIoResult, DiskFile, DiskFileError, DiskFileResult};
+use crate::vhdx::{Result as VhdxResult, Vhdx};
 use crate::AsyncAdaptor;
 use std::collections::VecDeque;
 use std::fs::File;
 use std::sync::{Arc, Mutex, MutexGuard};
-use vhdx::vhdx::{Result as VhdxResult, Vhdx};
 use vmm_sys_util::eventfd::EventFd;
 
 pub struct VhdxDiskSync {

block_util/Cargo.toml

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

build.rs

@@ -17,8 +17,8 @@ fn main() {
     }
 
     // This println!() has a special behavior, as it will set the environment
-    // variable BUILT_VERSION, so that it can be reused from the binary.
+    // variable BUILD_VERSION, so that it can be reused from the binary.
     // Particularly, this is used from src/main.rs to display the exact
     // version.
-    println!("cargo:rustc-env=BUILT_VERSION={version}");
+    println!("cargo:rustc-env=BUILD_VERSION={version}");
 }

devices/Cargo.toml

@@ -6,19 +6,22 @@ edition = "2021"
 
 [dependencies]
 acpi_tables = { git = "https://github.com/rust-vmm/acpi_tables", branch = "main"  }
-anyhow = "1.0.69"
+anyhow = "1.0.71"
 arch = { path = "../arch" }
-bitflags = "1.3.2"
+bitflags = "2.3.3"
 byteorder = "1.4.3"
+event_monitor = { path = "../event_monitor" }
 hypervisor = { path = "../hypervisor" }
 libc = "0.2.139"
 log = "0.4.17"
-thiserror = "1.0.39"
+pci = { path = "../pci" }
+thiserror = "1.0.40"
 tpm = { path = "../tpm" }
 versionize = "0.1.10"
 versionize_derive = "0.1.4"
+vm-allocator = { path = "../vm-allocator" }
 vm-device = { path = "../vm-device" }
-vm-memory = "0.10.0"
+vm-memory = "0.11.0"
 vm-migration = { path = "../vm-migration" }
 vmm-sys-util = "0.11.0"
 

devices/src/acpi.rs

@@ -5,7 +5,9 @@
 
 use super::AcpiNotificationFlags;
 use acpi_tables::{aml, Aml, AmlSink};
+use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Barrier};
+use std::thread;
 use std::time::Instant;
 use vm_device::interrupt::InterruptSourceGroup;
 use vm_device::BusDevice;
@@ -18,14 +20,20 @@ pub const GED_DEVICE_ACPI_SIZE: usize = 0x1;
 pub struct AcpiShutdownDevice {
     exit_evt: EventFd,
     reset_evt: EventFd,
+    vcpus_kill_signalled: Arc<AtomicBool>,
 }
 
 impl AcpiShutdownDevice {
     /// Constructs a device that will signal the given event when the guest requests it.
-    pub fn new(exit_evt: EventFd, reset_evt: EventFd) -> AcpiShutdownDevice {
+    pub fn new(
+        exit_evt: EventFd,
+        reset_evt: EventFd,
+        vcpus_kill_signalled: Arc<AtomicBool>,
+    ) -> AcpiShutdownDevice {
         AcpiShutdownDevice {
             exit_evt,
             reset_evt,
+            vcpus_kill_signalled,
         }
     }
 }
@@ -43,6 +51,13 @@ impl BusDevice for AcpiShutdownDevice {
             if let Err(e) = self.reset_evt.write(1) {
                 error!("Error triggering ACPI reset event: {}", e);
             }
+            // Spin until we are sure the reset_evt has been handled and that when
+            // we return from the KVM_RUN we will exit rather than re-enter the guest.
+            while !self.vcpus_kill_signalled.load(Ordering::SeqCst) {
+                // This is more effective than thread::yield_now() at
+                // avoiding a priority inversion with the VMM thread
+                thread::sleep(std::time::Duration::from_millis(1));
+            }
         }
         // The ACPI DSDT table specifies the S5 sleep state (shutdown) as value 5
         const S5_SLEEP_VALUE: u8 = 5;
@@ -53,6 +68,13 @@ impl BusDevice for AcpiShutdownDevice {
             if let Err(e) = self.exit_evt.write(1) {
                 error!("Error triggering ACPI shutdown event: {}", e);
             }
+            // Spin until we are sure the reset_evt has been handled and that when
+            // we return from the KVM_RUN we will exit rather than re-enter the guest.
+            while !self.vcpus_kill_signalled.load(Ordering::SeqCst) {
+                // This is more effective than thread::yield_now() at
+                // avoiding a priority inversion with the VMM thread
+                thread::sleep(std::time::Duration::from_millis(1));
+            }
         }
         None
     }
@@ -116,6 +138,7 @@ impl Aml for AcpiGedDevice {
                         true,
                         self.address.0,
                         self.address.0 + GED_DEVICE_ACPI_SIZE as u64 - 1,
+                        None,
                     )]),
                 ),
                 &aml::OpRegion::new(

devices/src/gic.rs

@@ -98,9 +98,14 @@ impl Gic {
                     i as InterruptIndex,
                     InterruptSourceConfig::LegacyIrq(config),
                     false,
+                    false,
                 )
                 .map_err(Error::EnableInterrupt)?;
         }
+
+        self.interrupt_source_group
+            .set_gsi()
+            .map_err(Error::EnableInterrupt)?;
         Ok(())
     }
 

devices/src/ioapic.rs

@@ -237,9 +237,14 @@ impl Ioapic {
         if state.is_some() {
             for (irq, entry) in ioapic.used_entries.iter().enumerate() {
                 if *entry {
-                    ioapic.update_entry(irq)?;
+                    ioapic.update_entry(irq, false)?;
                 }
             }
+
+            ioapic
+                .interrupt_source_group
+                .set_gsi()
+                .map_err(Error::UpdateInterrupt)?;
         }
 
         Ok(ioapic)
@@ -278,7 +283,7 @@ impl Ioapic {
                 }
                 // The entry must be updated through the interrupt source
                 // group.
-                if let Err(e) = self.update_entry(index) {
+                if let Err(e) = self.update_entry(index, true) {
                     error!("Failed updating IOAPIC entry: {:?}", e);
                 }
                 // Store the information this IRQ is now being used.
@@ -329,7 +334,7 @@ impl Ioapic {
         }
     }
 
-    fn update_entry(&self, irq: usize) -> Result<()> {
+    fn update_entry(&self, irq: usize, set_gsi: bool) -> Result<()> {
         let entry = self.reg_entries[irq];
 
         // Validate Destination Mode value, and retrieve Destination ID
@@ -386,6 +391,7 @@ impl Ioapic {
                 irq as InterruptIndex,
                 InterruptSourceConfig::MsiIrq(config),
                 interrupt_mask(entry) == 1,
+                set_gsi,
             )
             .map_err(Error::UpdateInterrupt)?;
 

devices/src/legacy/cmos.rs

@@ -5,7 +5,9 @@
 use libc::{clock_gettime, gmtime_r, timespec, tm, CLOCK_REALTIME};
 use std::cmp::min;
 use std::mem;
+use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Barrier};
+use std::thread;
 use vm_device::BusDevice;
 use vmm_sys_util::eventfd::EventFd;
 
@@ -23,13 +25,19 @@ pub struct Cmos {
     index: u8,
     data: [u8; DATA_LEN],
     reset_evt: EventFd,
+    vcpus_kill_signalled: Option<Arc<AtomicBool>>,
 }
 
 impl Cmos {
     /// Constructs a CMOS/RTC device with initial data.
     /// `mem_below_4g` is the size of memory in bytes below the 32-bit gap.
     /// `mem_above_4g` is the size of memory in bytes above the 32-bit gap.
-    pub fn new(mem_below_4g: u64, mem_above_4g: u64, reset_evt: EventFd) -> Cmos {
+    pub fn new(
+        mem_below_4g: u64,
+        mem_above_4g: u64,
+        reset_evt: EventFd,
+        vcpus_kill_signalled: Option<Arc<AtomicBool>>,
+    ) -> Cmos {
         let mut data = [0u8; DATA_LEN];
 
         // Extended memory from 16 MB to 4 GB in units of 64 KB
@@ -50,6 +58,7 @@ impl Cmos {
             index: 0,
             data,
             reset_evt,
+            vcpus_kill_signalled,
         }
     }
 }
@@ -67,6 +76,15 @@ impl BusDevice for Cmos {
                 if self.index == 0x8f && data[0] == 0 {
                     info!("CMOS reset");
                     self.reset_evt.write(1).unwrap();
+                    if let Some(vcpus_kill_signalled) = self.vcpus_kill_signalled.take() {
+                        // Spin until we are sure the reset_evt has been handled and that when
+                        // we return from the KVM_RUN we will exit rather than re-enter the guest.
+                        while !vcpus_kill_signalled.load(Ordering::SeqCst) {
+                            // This is more effective than thread::yield_now() at
+                            // avoiding a priority inversion with the VMM thread
+                            thread::sleep(std::time::Duration::from_millis(1));
+                        }
+                    }
                 } else {
                     self.data[(self.index & INDEX_MASK) as usize] = data[0]
                 }

devices/src/legacy/gpio_pl061.rs

@@ -361,10 +361,15 @@ mod tests {
             _index: InterruptIndex,
             _config: InterruptSourceConfig,
             _masked: bool,
+            _set_gsi: bool,
         ) -> result::Result<(), std::io::Error> {
             Ok(())
         }
 
+        fn set_gsi(&self) -> result::Result<(), std::io::Error> {
+            Ok(())
+        }
+
         fn notifier(&self, _index: InterruptIndex) -> Option<EventFd> {
             Some(self.event_fd.try_clone().unwrap())
         }

devices/src/legacy/i8042.rs

@@ -2,19 +2,27 @@
 // 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::sync::{Arc, Barrier};
+use std::sync::{
+    atomic::{AtomicBool, Ordering},
+    Arc, Barrier,
+};
+use std::thread;
 use vm_device::BusDevice;
 use vmm_sys_util::eventfd::EventFd;
 
 /// A i8042 PS/2 controller that emulates just enough to shutdown the machine.
 pub struct I8042Device {
     reset_evt: EventFd,
+    vcpus_kill_signalled: Arc<AtomicBool>,
 }
 
 impl I8042Device {
     /// Constructs a i8042 device that will signal the given event when the guest requests it.
-    pub fn new(reset_evt: EventFd) -> I8042Device {
-        I8042Device { reset_evt }
+    pub fn new(reset_evt: EventFd, vcpus_kill_signalled: Arc<AtomicBool>) -> I8042Device {
+        I8042Device {
+            reset_evt,
+            vcpus_kill_signalled,
+        }
     }
 }
 
@@ -38,6 +46,13 @@ impl BusDevice for I8042Device {
             if let Err(e) = self.reset_evt.write(1) {
                 error!("Error triggering i8042 reset event: {}", e);
             }
+            // Spin until we are sure the reset_evt has been handled and that when
+            // we return from the KVM_RUN we will exit rather than re-enter the guest.
+            while !self.vcpus_kill_signalled.load(Ordering::SeqCst) {
+                // This is more effective than thread::yield_now() at
+                // avoiding a priority inversion with the VMM thread
+                thread::sleep(std::time::Duration::from_millis(1));
+            }
         }
 
         None

devices/src/legacy/rtc_pl031.rs

@@ -413,10 +413,15 @@ mod tests {
             _index: InterruptIndex,
             _config: InterruptSourceConfig,
             _masked: bool,
+            _set_gsi: bool,
         ) -> result::Result<(), std::io::Error> {
             Ok(())
         }
 
+        fn set_gsi(&self) -> result::Result<(), std::io::Error> {
+            Ok(())
+        }
+
         fn notifier(&self, _index: InterruptIndex) -> Option<EventFd> {
             Some(self.event_fd.try_clone().unwrap())
         }

devices/src/legacy/serial.rs

@@ -365,9 +365,13 @@ mod tests {
             _index: InterruptIndex,
             _config: InterruptSourceConfig,
             _masked: bool,
+            _set_gsi: bool,
         ) -> result::Result<(), std::io::Error> {
             Ok(())
         }
+        fn set_gsi(&self) -> result::Result<(), std::io::Error> {
+            Ok(())
+        }
         fn notifier(&self, _index: InterruptIndex) -> Option<EventFd> {
             Some(self.event_fd.try_clone().unwrap())
         }

devices/src/legacy/uart_pl011.rs

@@ -485,9 +485,13 @@ mod tests {
             _index: InterruptIndex,
             _config: InterruptSourceConfig,
             _masked: bool,
+            _set_gsi: bool,
         ) -> result::Result<(), std::io::Error> {
             Ok(())
         }
+        fn set_gsi(&self) -> result::Result<(), std::io::Error> {
+            Ok(())
+        }
         fn notifier(&self, _index: InterruptIndex) -> Option<EventFd> {
             Some(self.event_fd.try_clone().unwrap())
         }

devices/src/lib.rs

@@ -10,6 +10,8 @@
 #[macro_use]
 extern crate bitflags;
 #[macro_use]
+extern crate event_monitor;
+#[macro_use]
 extern crate log;
 
 pub mod acpi;
@@ -19,9 +21,11 @@ pub mod interrupt_controller;
 #[cfg(target_arch = "x86_64")]
 pub mod ioapic;
 pub mod legacy;
+pub mod pvpanic;
 pub mod tpm;
 
 pub use self::acpi::{AcpiGedDevice, AcpiPmTimerDevice, AcpiShutdownDevice};
+pub use self::pvpanic::{PvPanicDevice, PVPANIC_DEVICE_MMIO_SIZE};
 
 bitflags! {
     pub struct AcpiNotificationFlags: u8 {

devices/src/pvpanic.rs

@@ -0,0 +1,272 @@
+// Copyright © 2023 Tencent Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+use anyhow::anyhow;
+use pci::{
+    BarReprogrammingParams, PciBarConfiguration, PciBarPrefetchable, PciBarRegionType,
+    PciClassCode, PciConfiguration, PciDevice, PciDeviceError, PciHeaderType, PciSubclass,
+    PCI_CONFIGURATION_ID,
+};
+use std::any::Any;
+use std::result;
+use std::sync::{Arc, Barrier, Mutex};
+use thiserror::Error;
+use versionize::{VersionMap, Versionize, VersionizeResult};
+use versionize_derive::Versionize;
+use vm_allocator::{AddressAllocator, SystemAllocator};
+use vm_device::{BusDevice, Resource};
+use vm_memory::{Address, GuestAddress};
+use vm_migration::{
+    Migratable, MigratableError, Pausable, Snapshot, Snapshottable, Transportable, VersionMapped,
+};
+
+const PVPANIC_VENDOR_ID: u16 = 0x1b36;
+const PVPANIC_DEVICE_ID: u16 = 0x0011;
+
+pub const PVPANIC_DEVICE_MMIO_SIZE: u64 = 0x2;
+
+const PVPANIC_PANICKED: u8 = 1 << 0;
+const PVPANIC_CRASH_LOADED: u8 = 1 << 1;
+
+#[derive(Debug, Error)]
+pub enum PvPanicError {
+    #[error("Failed creating PvPanicDevice: {0}")]
+    CreatePvPanicDevice(#[source] anyhow::Error),
+    #[error("Failed to retrieve PciConfigurationState: {0}")]
+    RetrievePciConfigurationState(#[source] anyhow::Error),
+}
+
+#[allow(dead_code)]
+#[derive(Copy, Clone)]
+pub enum PvPanicSubclass {
+    Other = 0x80,
+}
+
+impl PciSubclass for PvPanicSubclass {
+    fn get_register_value(&self) -> u8 {
+        *self as u8
+    }
+}
+
+/// A device for handling guest panic event
+pub struct PvPanicDevice {
+    id: String,
+    events: u8,
+
+    // PCI configuration registers.
+    configuration: PciConfiguration,
+    bar_regions: Vec<PciBarConfiguration>,
+}
+
+#[derive(Versionize)]
+pub struct PvPanicDeviceState {
+    events: u8,
+}
+
+impl VersionMapped for PvPanicDeviceState {}
+
+impl PvPanicDevice {
+    pub fn new(id: String, snapshot: Option<Snapshot>) -> Result<Self, PvPanicError> {
+        let pci_configuration_state =
+            vm_migration::versioned_state_from_id(snapshot.as_ref(), PCI_CONFIGURATION_ID)
+                .map_err(|e| {
+                    PvPanicError::RetrievePciConfigurationState(anyhow!(
+                        "Failed to get PciConfigurationState from Snapshot: {}",
+                        e
+                    ))
+                })?;
+
+        let mut configuration = PciConfiguration::new(
+            PVPANIC_VENDOR_ID,
+            PVPANIC_DEVICE_ID,
+            0x1, // modern pci devices
+            PciClassCode::BaseSystemPeripheral,
+            &PvPanicSubclass::Other,
+            None,
+            PciHeaderType::Device,
+            0,
+            0,
+            None,
+            pci_configuration_state,
+        );
+
+        let command: [u8; 2] = [0x03, 0x01];
+        configuration.write_config_register(1, 0, &command);
+
+        let state: Option<PvPanicDeviceState> = snapshot
+            .as_ref()
+            .map(|s| s.to_versioned_state())
+            .transpose()
+            .map_err(|e| {
+                PvPanicError::CreatePvPanicDevice(anyhow!(
+                    "Failed to get PvPanicDeviceState from Snapshot: {}",
+                    e
+                ))
+            })?;
+        let events = if let Some(state) = state {
+            state.events
+        } else {
+            PVPANIC_PANICKED | PVPANIC_CRASH_LOADED
+        };
+
+        let pvpanic_device = PvPanicDevice {
+            id,
+            events,
+            configuration,
+            bar_regions: vec![],
+        };
+
+        Ok(pvpanic_device)
+    }
+
+    pub fn event_to_string(&self, event: u8) -> String {
+        if event == PVPANIC_PANICKED {
+            "panic".to_string()
+        } else if event == PVPANIC_CRASH_LOADED {
+            "crash_loaded".to_string()
+        } else {
+            "unknown_event".to_string()
+        }
+    }
+
+    fn state(&self) -> PvPanicDeviceState {
+        PvPanicDeviceState {
+            events: self.events,
+        }
+    }
+
+    pub fn config_bar_addr(&self) -> u64 {
+        self.configuration.get_bar_addr(0)
+    }
+}
+
+impl BusDevice for PvPanicDevice {
+    fn read(&mut self, base: u64, offset: u64, data: &mut [u8]) {
+        self.read_bar(base, offset, data)
+    }
+
+    fn write(&mut self, _base: u64, _offset: u64, data: &[u8]) -> Option<Arc<Barrier>> {
+        let event = self.event_to_string(data[0]);
+        info!("pvpanic got guest event {}", event);
+        event!("guest", "panic", "event", &event);
+        None
+    }
+}
+
+impl PciDevice for PvPanicDevice {
+    fn write_config_register(
+        &mut self,
+        reg_idx: usize,
+        offset: u64,
+        data: &[u8],
+    ) -> Option<Arc<Barrier>> {
+        self.configuration
+            .write_config_register(reg_idx, offset, data);
+        None
+    }
+
+    fn read_config_register(&mut self, reg_idx: usize) -> u32 {
+        self.configuration.read_reg(reg_idx)
+    }
+
+    fn detect_bar_reprogramming(
+        &mut self,
+        reg_idx: usize,
+        data: &[u8],
+    ) -> Option<BarReprogrammingParams> {
+        self.configuration.detect_bar_reprogramming(reg_idx, data)
+    }
+
+    fn allocate_bars(
+        &mut self,
+        allocator: &Arc<Mutex<SystemAllocator>>,
+        _mmio_allocator: &mut AddressAllocator,
+        resources: Option<Vec<Resource>>,
+    ) -> std::result::Result<Vec<PciBarConfiguration>, PciDeviceError> {
+        let mut bars = Vec::new();
+        let region_type = PciBarRegionType::Memory32BitRegion;
+        let bar_id = 0;
+        let region_size = PVPANIC_DEVICE_MMIO_SIZE;
+        let restoring = resources.is_some();
+        let bar_addr = allocator
+            .lock()
+            .unwrap()
+            .allocate_mmio_hole_addresses(None, region_size, None)
+            .ok_or(PciDeviceError::IoAllocationFailed(region_size))?;
+
+        let bar = PciBarConfiguration::default()
+            .set_index(bar_id as usize)
+            .set_address(bar_addr.raw_value())
+            .set_size(region_size)
+            .set_region_type(region_type)
+            .set_prefetchable(PciBarPrefetchable::NotPrefetchable);
+
+        debug!("pvpanic bar address 0x{:x}", bar_addr.0);
+        if !restoring {
+            self.configuration
+                .add_pci_bar(&bar)
+                .map_err(|e| PciDeviceError::IoRegistrationFailed(bar_addr.raw_value(), e))?;
+        }
+
+        bars.push(bar);
+        self.bar_regions = bars.clone();
+
+        Ok(bars)
+    }
+
+    fn free_bars(
+        &mut self,
+        allocator: &mut SystemAllocator,
+        _mmio_allocator: &mut AddressAllocator,
+    ) -> std::result::Result<(), PciDeviceError> {
+        for bar in self.bar_regions.drain(..) {
+            allocator.free_mmio_hole_addresses(GuestAddress(bar.addr()), bar.size());
+        }
+
+        Ok(())
+    }
+
+    fn move_bar(&mut self, old_base: u64, new_base: u64) -> result::Result<(), std::io::Error> {
+        for bar in self.bar_regions.iter_mut() {
+            if bar.addr() == old_base {
+                *bar = bar.set_address(new_base);
+            }
+        }
+
+        Ok(())
+    }
+
+    fn read_bar(&mut self, _base: u64, _offset: u64, data: &mut [u8]) {
+        data[0] = self.events;
+    }
+
+    fn as_any(&mut self) -> &mut dyn Any {
+        self
+    }
+
+    fn id(&self) -> Option<String> {
+        Some(self.id.clone())
+    }
+}
+
+impl Pausable for PvPanicDevice {}
+
+impl Snapshottable for PvPanicDevice {
+    fn id(&self) -> String {
+        self.id.clone()
+    }
+
+    fn snapshot(&mut self) -> std::result::Result<Snapshot, MigratableError> {
+        let mut snapshot = Snapshot::new_from_versioned_state(&self.state())?;
+
+        // Snapshot PciConfiguration
+        snapshot.add_snapshot(self.configuration.id(), self.configuration.snapshot()?);
+
+        Ok(snapshot)
+    }
+}
+
+impl Transportable for PvPanicDevice {}
+impl Migratable for PvPanicDevice {}

docs/api.md

@@ -1,18 +1,21 @@
 - [Cloud Hypervisor API](#cloud-hypervisor-api)
   - [External API](#external-api)
     - [REST API](#rest-api)
-    - [Location and availability](#location-and-availability)
-    - [Endpoints](#endpoints)
-      - [Virtual Machine Manager (VMM) Actions](#virtual-machine-manager-vmm-actions)
-      - [Virtual Machine (VM) Actions](#virtual-machine-vm-actions)
-    - [REST API Examples](#rest-api-examples)
-      - [Create a Virtual Machine](#create-a-virtual-machine)
-      - [Boot a Virtual Machine](#boot-a-virtual-machine)
-      - [Dump a Virtual Machine Information](#dump-a-virtual-machine-information)
-      - [Reboot a Virtual Machine](#reboot-a-virtual-machine)
-      - [Shut a Virtual Machine Down](#shut-a-virtual-machine-down)
+      - [REST API Location and availability](#rest-api-location-and-availability)
+      - [REST API Endpoints](#rest-api-endpoints)
+        - [Virtual Machine Manager (VMM) Actions](#virtual-machine-manager-vmm-actions)
+        - [Virtual Machine (VM) Actions](#virtual-machine-vm-actions)
+      - [REST API Examples](#rest-api-examples)
+        - [Create a Virtual Machine](#create-a-virtual-machine)
+        - [Boot a Virtual Machine](#boot-a-virtual-machine)
+        - [Dump a Virtual Machine Information](#dump-a-virtual-machine-information)
+        - [Reboot a Virtual Machine](#reboot-a-virtual-machine)
+        - [Shut a Virtual Machine Down](#shut-a-virtual-machine-down)
+    - [D-Bus API](#d-bus-api)
+      - [D-Bus API Location and availability](#d-bus-api-location-and-availability)
+      - [D-Bus API Interface](#d-bus-api-interface)
     - [Command Line Interface](#command-line-interface)
-    - [REST API and CLI Architectural Relationship](#rest-api-and-cli-architectural-relationship)
+    - [REST API, D-Bus API and CLI Architectural Relationship](#rest-api-and-cli-architectural-relationship)
   - [Internal API](#internal-api)
     - [Goals and Design](#goals-and-design)
   - [End to End Example](#end-to-end-example)
@@ -21,9 +24,11 @@
 
 The Cloud Hypervisor API is made of 2 distinct interfaces:
 
-1. **The external API**. This is the user facing API. Users and operators can
-   control and manage Cloud Hypervisor through either a REST API or a Command
-   Line Interface (CLI).
+1. **The External API** This is the user facing API. Users and operators
+   can control and manage the Cloud Hypervisor through various options
+   including a REST API, a Command Line Interface (CLI) or a D-Bus based API,
+   which is not compiled into Cloud Hypervisor by default.
+
 1. **The internal API**, based on [rust's Multi-Producer, Single-Consumer (MPSC)](https://doc.rust-lang.org/std/sync/mpsc/)
    module. This API is used internally by the Cloud Hypervisor threads to
    communicate between each others.
@@ -40,10 +45,10 @@ API triggers VM and VMM specific actions, and as such it is designed as a
 collection of RPC-style, static methods.
 
 The API is [OpenAPI 3.0](https://github.com/OAI/OpenAPI-Specification/blob/master/versions/3.0.0.md)
-compliant. Please consult the [Cloud Hypervisor API](https://raw.githubusercontent.com/cloud-hypervisor/cloud-hypervisor/master/vmm/src/api/openapi/cloud-hypervisor.yaml)
-document for more details about the API payloads and responses.
+compliant. Please consult the [Cloud Hypervisor OpenAPI Document](https://raw.githubusercontent.com/cloud-hypervisor/cloud-hypervisor/master/vmm/src/api/openapi/cloud-hypervisor.yaml)
+for more details about the API payloads and responses.
 
-### Location and availability
+#### REST API Location and availability
 
 The REST API is available as soon as the Cloud Hypervisor binary is started,
 through a local UNIX socket.
@@ -65,18 +70,18 @@ Cloud Hypervisor Guest
     Disk(s): None
 ```
 
-### Endpoints
+#### REST API Endpoints
 
 The Cloud Hypervisor API exposes the following actions through its endpoints:
 
-#### Virtual Machine Manager (VMM) Actions
+##### Virtual Machine Manager (VMM) Actions
 
 | Action                              | Endpoint        | Request Body | Response Body              | Prerequisites      |
 | ----------------------------------- | --------------- | ------------ | -------------------------- | ------------------ |
 | Check for the REST API availability | `/vmm.ping`     | N/A          | `/schemas/VmmPingResponse` | N/A                |
 | Shut the VMM down                   | `/vmm.shutdown` | N/A          | N/A                        | The VMM is running |
 
-#### Virtual Machine (VM) Actions
+##### Virtual Machine (VM) Actions
 
 | Action                             | Endpoint                | Request Body                    | Response Body            | Prerequisites                                          |
 | ---------------------------------- | ----------------------- | ------------------------------- | ------------------------ | ------------------------------------------------------ |
@@ -110,10 +115,10 @@ The Cloud Hypervisor API exposes the following actions through its endpoints:
 
 * The `vmcoredump` action is available exclusively for the `x86_64`
 architecture and can be executed only when the `guest_debug` feature is
-enabled. Without this feature, the corresponding REST API endpoint is not
-available.
+enabled. Without this feature, the corresponding [REST API](#rest-api) or
+[D-Bus API](#d-bus-api) endpoints are not available.
 
-### REST API Examples
+#### REST API Examples
 
 For the following set of examples, we assume Cloud Hypervisor is started with
 the REST API available at `/tmp/cloud-hypervisor.sock`:
@@ -129,7 +134,7 @@ Cloud Hypervisor Guest
     Disk(s): None
 ```
 
-#### Create a Virtual Machine
+##### Create a Virtual Machine
 
 We want to create a virtual machine with the following characteristics:
 
@@ -157,7 +162,7 @@ curl --unix-socket /tmp/cloud-hypervisor.sock -i \
          }'
 ```
 
-#### Boot a Virtual Machine
+##### Boot a Virtual Machine
 
 Once the VM is created, we can boot it:
 
@@ -167,7 +172,7 @@ Once the VM is created, we can boot it:
 curl --unix-socket /tmp/cloud-hypervisor.sock -i -X PUT 'http://localhost/api/v1/vm.boot'
 ```
 
-#### Dump a Virtual Machine Information
+##### Dump a Virtual Machine Information
 
 We can fetch information about any VM, as soon as it's created:
 
@@ -179,7 +184,7 @@ curl --unix-socket /tmp/cloud-hypervisor.sock -i \
      -H 'Accept: application/json'
 ```
 
-#### Reboot a Virtual Machine
+##### Reboot a Virtual Machine
 
 We can reboot a VM that's already booted:
 
@@ -189,7 +194,7 @@ We can reboot a VM that's already booted:
 curl --unix-socket /tmp/cloud-hypervisor.sock -i -X PUT 'http://localhost/api/v1/vm.reboot'
 ```
 
-#### Shut a Virtual Machine Down
+##### Shut a Virtual Machine Down
 
 Once booted, we can shut a VM down from the REST API:
 
@@ -199,6 +204,50 @@ Once booted, we can shut a VM down from the REST API:
 curl --unix-socket /tmp/cloud-hypervisor.sock -i -X PUT 'http://localhost/api/v1/vm.shutdown'
 ```
 
+### D-Bus API
+
+Cloud Hypervisor offers a D-Bus API as an alternative to its REST API. As of
+writing this document, the D-Bus API mirrors the functionality of the REST
+API and shares the same set of endpoints, meaning that it supports every call
+that is supported by the REST API and can be a drop-in replacement since it
+also consumes/produces JSON.
+
+#### D-Bus API Location and availability
+
+This feature is not compiled into Cloud Hypervisor by default. Users who
+wish to use the D-Bus API, must explicitly enable it with the `dbus_api`
+feature flag when compiling Cloud Hypervisor.
+
+```sh
+$ ./scripts/dev_cli.sh build --release --libc musl -- --features dbus_api
+```
+
+Once this feature is enabled, it can be configured with the following
+CLI options:
+
+```
+  --dbus-service-name
+                    well known name of the service
+  --dbus-object-path
+                    object path to serve the dbus interface
+  --dbus-system-bus use the system bus instead of a session bus
+```
+
+Example invocation:
+
+```sh
+$ ./cloud-hypervisor --dbus-service-name "org.cloudhypervisor.DBusApi" \
+                     --dbus-object-path "/org/cloudhypervisor/DBusApi"
+```
+
+This will start serving a service with the name `org.cloudhypervisor.DBusApi1`
+which in turn can be used to control and manage Cloud Hypervisor.
+
+#### D-Bus API Interface
+
+Please refer to the [REST API](#rest-api) documentation. As previously
+mentioned, the D-Bus API currently mirrors the behaviour of the REST API.
+
 ### Command Line Interface
 
 The Cloud Hypervisor Command Line Interface (CLI) can only be used for launching
@@ -206,22 +255,26 @@ the Cloud Hypervisor binary, i.e. it can not be used for controlling the VMM or
 the launched VM once they're up and running.
 
 If you want to inspect the VMM, or control the VM after launching Cloud
-Hypervisor from the CLI, you must use the [REST API](#rest-api).
+Hypervisor from the CLI, you must use either the [REST API](#rest-api)
+or the [D-Bus API](#d-bus-api).
 
-From the CLI, one can either:
+From the CLI, one can:
 
 1. Create and boot a complete virtual machine by using the CLI options to build
    the VM config. Run `cloud-hypervisor --help` for a complete list of CLI
-   options. As soon as the `cloud-hypervisor` binary is launched, the
-   [REST API](#rest-api) is available for controlling and managing the VM.
-1. Start the [REST API](#rest-api) server only, by not passing any VM
-   configuration options. The VM can then be asynchronously created and booted
-   by sending HTTP commands to the [REST API](#rest-api). Check the
-   [REST API examples](#rest-api-examples) section for more details.
+   options. As soon as the `cloud-hypervisor` binary is launched, contrary
+   to the [D-Bus API](#d-bus-api), the [REST API](#rest-api) is available
+   for controlling and managing the VM. The [D-Bus API](#d-bus-api) doesn't start
+   automatically and needs to be explicitly configured in order to be run.
+1. Start either the REST API, D-Bus API or both simultaneously without passing
+   any VM configuration options. The VM can then be asynchronously created and
+   booted by calling API methods of choice. It should be noted that one external
+   API does not exclude another; it is possible to have both the REST and D-Bus
+   APIs running simultaneously.
 
-### REST API and CLI Architectural Relationship
+### REST API, D-Bus API and CLI Architectural Relationship
 
-The REST API and the CLI both rely on a common, [internal API](#internal-api).
+The REST API, D-Bus API and the CLI all rely on a common, [internal API](#internal-api).
 
 The CLI options are parsed by the
 [argh crate](https://docs.rs/argh/latest/argh/) and then translated into
@@ -232,7 +285,11 @@ The REST API is processed by an HTTP thread using the
 crate. As with the CLI, the HTTP requests eventually get translated into
 [internal API](#internal-api) commands.
 
-As a summary, the REST API and the CLI are essentially frontends for the
+The D-Bus API is implemented using the [zbus](https://github.com/dbus2/zbus)
+crate and runs in its own thread. Whenever it needs to call the [internal API](#internal-api),
+the [blocking](https://github.com/smol-rs/blocking) crate is used perform the call in zbus' async context.
+
+As a summary, the REST API, the D-Bus API and the CLI are essentially frontends for the
 [internal API](#internal-api):
 
 ```
@@ -243,11 +300,11 @@ As a summary, the REST API and the CLI are essentially frontends for the
                        |          +------------------+        |
                        |                                      |      +------------------------+
                        |                                      |      |                        |
-+------------+         |                                      |      |                        |
-|            |         |                                      |      | +--------------+       |
-|    User    +---------+                                      +------> | Internal API |       |
-|            |         |                                      |      | +--------------+       |
-+------------+         |                                      |      |                        |
++------------+         |            +----------+              |      |                        |
+|            |         |  D-Bus API |          |              |      | +--------------+       |
+|    User    +---------+----------->+   zbus   +--------------+------> | Internal API |       |
+|            |         |            |          |              |      | +--------------+       |
++------------+         |            +----------+              |      |                        |
                        |                                      |      |                        |
                        |                                      |      +------------------------+
                        |            +----------+              |                 VMM
@@ -262,22 +319,23 @@ As a summary, the REST API and the CLI are essentially frontends for the
 ## Internal API
 
 The Cloud Hypervisor internal API, as its name suggests, is used internally
-by the different Cloud Hypervisor threads (VMM, HTTP, control loop, etc) to
-send commands and responses to each others.
+by the different Cloud Hypervisor threads (VMM, HTTP, D-Bus, control loop,
+etc) to send commands and responses to each others.
 
 It is based on [rust's Multi-Producer, Single-Consumer (MPSC)](https://doc.rust-lang.org/std/sync/mpsc/),
 and the single consumer (a.k.a. the API receiver) is the Cloud Hypervisor
 control loop.
 
-API producers are the HTTP thread handling the [REST API](#rest-api) and the
-main thread that initially parses the [CLI](#command-line-interface).
+API producers are the HTTP thread handling the [REST API](#rest-api), the
+D-Bus thread handling the [D-Bus API](#d-bus-api) and the main thread that
+initially parses the [CLI](#command-line-interface).
 
 ### Goals and Design
 
 The internal API is designed for controlling, managing and inspecting a Cloud
 Hypervisor VMM and its guest. It is a backend for handling external, user
-visible requests through either the [REST API](#rest-api) or the
-[CLI](#command-line-interface) interfaces.
+visible requests through the [REST API](#rest-api), the [D-Bus API](#d-bus-api)
+or the [CLI](#command-line-interface) interfaces.
 
 The API follows a command-response scheme that closely maps the [REST API](#rest-api).
 Any command must be replied to with a response.

docs/custom-image.md

@@ -225,7 +225,7 @@ Number  Start   End     Size    File system  Name  Flags
 
 ### Create a macvtap interface
 
-Rely on the following [documentation](docs/macvtap-bridge.md) to set up a
+Rely on the following [documentation](macvtap-bridge.md) to set up a
 macvtap interface to provide your VM with proper connectivity.
 
 ### Boot the image

event_monitor/Cargo.toml

@@ -5,6 +5,7 @@ authors = ["The Cloud Hypervisor Authors"]
 edition = "2021"
 
 [dependencies]
+flume = "0.10.14"
 libc = "0.2.139"
-serde = { version = "1.0.151", features = ["rc", "derive"] }
-serde_json = "1.0.95"
+serde = { version = "1.0.164", features = ["rc", "derive"] }
+serde_json = "1.0.96"

event_monitor/src/lib.rs

@@ -7,33 +7,11 @@ use serde::Serialize;
 use std::borrow::Cow;
 use std::collections::HashMap;
 use std::fs::File;
-use std::io::Write;
+use std::io;
 use std::os::unix::io::AsRawFd;
 use std::time::{Duration, Instant};
 
-static mut MONITOR: Option<(File, Instant)> = None;
-
-/// This function must only be called once from the main process before any threads
-/// are created to avoid race conditions
-pub fn set_monitor(file: File) -> Result<(), std::io::Error> {
-    // SAFETY: there is only one caller of this function, so MONITOR is written to only once
-    assert!(unsafe { MONITOR.is_none() });
-    let fd = file.as_raw_fd();
-    // SAFETY: FFI call to configure the fd
-    let ret = unsafe {
-        let mut flags = libc::fcntl(fd, libc::F_GETFL);
-        flags |= libc::O_NONBLOCK;
-        libc::fcntl(fd, libc::F_SETFL, flags)
-    };
-    if ret < 0 {
-        return Err(std::io::Error::last_os_error());
-    }
-    // SAFETY: MONITOR is None. Nobody else can hold a reference to it.
-    unsafe {
-        MONITOR = Some((file, Instant::now()));
-    };
-    Ok(())
-}
+static mut MONITOR: Option<MonitorHandle> = None;
 
 #[derive(Serialize)]
 struct Event<'a> {
@@ -43,19 +21,71 @@ struct Event<'a> {
     properties: Option<&'a HashMap<Cow<'a, str>, Cow<'a, str>>>,
 }
 
+pub struct Monitor {
+    pub rx: flume::Receiver<String>,
+    pub file: File,
+}
+
+struct MonitorHandle {
+    tx: flume::Sender<String>,
+    start: Instant,
+}
+
+fn set_file_nonblocking(file: &File) -> io::Result<()> {
+    let fd = file.as_raw_fd();
+
+    // SAFETY: FFI call to configure the fd
+    let ret = unsafe {
+        let mut flags = libc::fcntl(fd, libc::F_GETFL);
+        flags |= libc::O_NONBLOCK;
+        libc::fcntl(fd, libc::F_SETFL, flags)
+    };
+
+    if ret < 0 {
+        Err(io::Error::last_os_error())
+    } else {
+        Ok(())
+    }
+}
+
+/// This function must only be called once from the main thread before any threads
+/// are created to avoid race conditions.
+pub fn set_monitor(file: File) -> io::Result<Monitor> {
+    // SAFETY: there is only one caller of this function, so MONITOR is written to only once
+    assert!(unsafe { MONITOR.is_none() });
+
+    set_file_nonblocking(&file)?;
+    let (tx, rx) = flume::unbounded();
+    let monitor = Monitor { rx, file };
+
+    // SAFETY: MONITOR is None. Nobody else can hold a reference to it.
+    unsafe {
+        MONITOR = Some(MonitorHandle {
+            tx,
+            start: Instant::now(),
+        });
+    };
+
+    Ok(monitor)
+}
+
 pub fn event_log(source: &str, event: &str, properties: Option<&HashMap<Cow<str>, Cow<str>>>) {
-    // SAFETY: MONITOR is always in a valid state (None or Some).
-    if let Some((file, start)) = unsafe { MONITOR.as_ref() } {
-        let e = Event {
-            timestamp: start.elapsed(),
+    // SAFETY: `MONITOR` is always in a valid state (None or Some), because it
+    // is set only once before any threads are spawned, and it's not mutated
+    // afterwards. This function only creates immutable references to `MONITOR`.
+    // Because `MONITOR.tx` is `Sync`, it's safe to share `MONITOR` across
+    // threads, making this function thread-safe.
+    if let Some(monitor_handle) = unsafe { MONITOR.as_ref() } {
+        let event = Event {
+            timestamp: monitor_handle.start.elapsed(),
             source,
             event,
             properties,
         };
-        serde_json::to_writer_pretty(file, &e).ok();
 
-        let mut file = file;
-        file.write_all(b"\n\n").ok();
+        if let Ok(event) = serde_json::to_string_pretty(&event) {
+            monitor_handle.tx.send(event).ok();
+        }
     }
 }
 
@@ -79,5 +109,4 @@ macro_rules! event {
             $crate::event_log($source, $event, Some(&properties))
         }
      };
-
 }

fuzz/Cargo.lock

@@ -5,16 +5,16 @@ version = 3
 [[package]]
 name = "acpi_tables"
 version = "0.1.0"
-source = "git+https://github.com/rust-vmm/acpi_tables?branch=main#0b892e2c2053c4ecfac8d9e5a52babae75114702"
+source = "git+https://github.com/rust-vmm/acpi_tables?branch=main#05a609136387cc1cc9b499cee4320020325c263f"
 dependencies = [
  "zerocopy",
 ]
 
 [[package]]
 name = "anyhow"
-version = "1.0.70"
+version = "1.0.72"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7de8ce5e0f9f8d88245311066a578d72b7af3e7088f32783804676302df237e4"
+checksum = "3b13c32d80ecc7ab747b80c3784bce54ee8a7a0cc4fbda9bf4cda2cf6fe90854"
 
 [[package]]
 name = "api_client"
@@ -85,6 +85,12 @@ version = "0.1.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "64cb94155d965e3d37ffbbe7cc5b82c3dd79dd33bd48e536f73d2cfb8d85506f"
 
+[[package]]
+name = "autocfg"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d468802bab17cbc0cc575e9b053f41e72aa36bfa6b7f55e3529ffa43161b97fa"
+
 [[package]]
 name = "bincode"
 version = "1.3.3"
@@ -101,25 +107,39 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
 
 [[package]]
-name = "block_util"
+name = "bitflags"
+version = "2.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "630be753d4e58660abd17930c71b647fe46c27ea6b63cc59e1e3851406972e42"
+
+[[package]]
+name = "block"
 version = "0.1.0"
 dependencies = [
+ "byteorder",
+ "crc32c",
  "io-uring",
  "libc",
  "log",
- "qcow",
+ "remain",
  "smallvec",
  "thiserror",
+ "uuid",
  "versionize",
  "versionize_derive",
- "vhdx",
- "virtio-bindings 0.2.0",
+ "virtio-bindings",
  "virtio-queue",
  "vm-memory",
  "vm-virtio",
  "vmm-sys-util",
 ]
 
+[[package]]
+name = "bumpalo"
+version = "3.13.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a3e2c3daef883ecc1b5d58c15adae93470a91d425f3532ba1695849656af3fc1"
+
 [[package]]
 name = "byteorder"
 version = "1.4.3"
@@ -128,11 +148,12 @@ checksum = "14c189c53d098945499cdfa7ecc63567cf3886b3332b312a5b4585d8d3a6a610"
 
 [[package]]
 name = "cc"
-version = "1.0.79"
+version = "1.0.82"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "50d30906286121d95be3d479533b458f87493b30a4b5f79a607db8f5d11aa91f"
+checksum = "305fe645edc1442a0fa8b6726ba61d422798d37a52e12eaecf4b022ebbb88f01"
 dependencies = [
  "jobserver",
+ "libc",
 ]
 
 [[package]]
@@ -143,7 +164,7 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
 [[package]]
 name = "cloud-hypervisor"
-version = "30.0.0"
+version = "33.0.0"
 dependencies = [
  "anyhow",
  "api_client",
@@ -169,7 +190,7 @@ dependencies = [
 name = "cloud-hypervisor-fuzz"
 version = "0.0.0"
 dependencies = [
- "block_util",
+ "block",
  "cloud-hypervisor",
  "devices",
  "epoll",
@@ -179,9 +200,7 @@ dependencies = [
  "micro_http",
  "net_util",
  "once_cell",
- "qcow",
  "seccompiler",
- "vhdx",
  "virtio-devices",
  "virtio-queue",
  "vm-device",
@@ -193,9 +212,9 @@ dependencies = [
 
 [[package]]
 name = "crc32c"
-version = "0.6.3"
+version = "0.6.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3dfea2db42e9927a3845fb268a10a72faed6d416065f77873f05e411457c363e"
+checksum = "d8f48d60e5b4d2c53d5c2b1d8a58c849a70ae5e5509b08a48d047e3b65714a74"
 dependencies = [
  "rustc_version",
 ]
@@ -208,9 +227,9 @@ checksum = "55626594feae15d266d52440b26ff77de0e22230cf0c113abe619084c1ddc910"
 
 [[package]]
 name = "darling"
-version = "0.14.4"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7b750cb3417fd1b327431a470f388520309479ab0bf5e323505daf0290cd3850"
+checksum = "0209d94da627ab5605dcccf08bb18afa5009cfbef48d8a8b7d7bdbc79be25c5e"
 dependencies = [
  "darling_core",
  "darling_macro",
@@ -218,27 +237,27 @@ dependencies = [
 
 [[package]]
 name = "darling_core"
-version = "0.14.4"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "109c1ca6e6b7f82cc233a97004ea8ed7ca123a9af07a8230878fcfda9b158bf0"
+checksum = "177e3443818124b357d8e76f53be906d60937f0d3a90773a664fa63fa253e621"
 dependencies = [
  "fnv",
  "ident_case",
  "proc-macro2",
  "quote",
  "strsim",
- "syn 1.0.109",
+ "syn 2.0.23",
 ]
 
 [[package]]
 name = "darling_macro"
-version = "0.14.4"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a4aab4dbc9f7611d8b55048a3a16d2d010c2c8334e46304b40ac1cc14bf3b48e"
+checksum = "836a9bbc7ad63342d6d6e7b815ccab164bc77a2d95d84bc3117a8c0d5c98e2d5"
 dependencies = [
  "darling_core",
  "quote",
- "syn 1.0.109",
+ "syn 2.0.23",
 ]
 
 [[package]]
@@ -248,15 +267,18 @@ dependencies = [
  "acpi_tables",
  "anyhow",
  "arch",
- "bitflags",
+ "bitflags 2.3.3",
  "byteorder",
+ "event_monitor",
  "hypervisor",
  "libc",
  "log",
+ "pci",
  "thiserror",
  "tpm",
  "versionize",
  "versionize_derive",
+ "vm-allocator",
  "vm-device",
  "vm-memory",
  "vm-migration",
@@ -265,11 +287,11 @@ dependencies = [
 
 [[package]]
 name = "epoll"
-version = "4.3.1"
+version = "4.3.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "20df693c700404f7e19d4d6fae6b15215d2913c27955d2b9d6f2c0f537511cd0"
+checksum = "74351c3392ea1ff6cd2628e0042d268ac2371cb613252ff383b6dfa50d22fa79"
 dependencies = [
- "bitflags",
+ "bitflags 2.3.3",
  "libc",
 ]
 
@@ -277,6 +299,7 @@ dependencies = [
 name = "event_monitor"
 version = "0.1.0"
 dependencies = [
+ "flume",
  "libc",
  "serde",
  "serde_json",
@@ -288,6 +311,19 @@ version = "0.1.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "784a4df722dc6267a04af36895398f59d21d07dce47232adf31ec0ff2fa45e67"
 
+[[package]]
+name = "flume"
+version = "0.10.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1657b4441c3403d9f7b3409e47575237dac27b1b5726df654a6ecbf92f0f7577"
+dependencies = [
+ "futures-core",
+ "futures-sink",
+ "nanorand",
+ "pin-project",
+ "spin",
+]
+
 [[package]]
 name = "fnv"
 version = "1.0.7"
@@ -295,14 +331,28 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
 
 [[package]]
-name = "getrandom"
-version = "0.2.8"
+name = "futures-core"
+version = "0.3.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c05aeb6a22b8f62540c194aac980f2115af067bfe15a0734d7277a768d396b31"
+checksum = "4bca583b7e26f571124fe5b7561d49cb2868d79116cfa0eefce955557c6fee8c"
+
+[[package]]
+name = "futures-sink"
+version = "0.3.28"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f43be4fe21a13b9781a69afa4985b0f6ee0e1afab2c6f454a8cf30e2b2237b6e"
+
+[[package]]
+name = "getrandom"
+version = "0.2.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "be4136b2a15dd319360be1c07d9933517ccf0be8f16bf62a3bee4f0d618df427"
 dependencies = [
  "cfg-if",
+ "js-sys",
  "libc",
  "wasi",
+ "wasm-bindgen",
 ]
 
 [[package]]
@@ -311,7 +361,6 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "byteorder",
- "iced-x86",
  "kvm-bindings",
  "kvm-ioctls",
  "libc",
@@ -324,15 +373,6 @@ dependencies = [
  "vmm-sys-util",
 ]
 
-[[package]]
-name = "iced-x86"
-version = "1.18.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1dd04b950d75b3498320253b17fb92745b2cc79ead8814aede2f7c1bab858bec"
-dependencies = [
- "lazy_static",
-]
-
 [[package]]
 name = "ident_case"
 version = "1.0.1"
@@ -341,19 +381,19 @@ checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
 
 [[package]]
 name = "io-uring"
-version = "0.5.13"
+version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dd1e1a01cfb924fd8c5c43b6827965db394f5a3a16c599ce03452266e1cf984c"
+checksum = "8b7b36074613a723279637061b40db993208908a94f10ccb14436ce735bc0f57"
 dependencies = [
- "bitflags",
+ "bitflags 1.3.2",
  "libc",
 ]
 
 [[package]]
 name = "itoa"
-version = "1.0.6"
+version = "1.0.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "453ad9f582a441959e5f0d088b02ce04cfe8d51a8eaf077f12ac6d3e94164ca6"
+checksum = "af150ab688ff2122fcef229be89cb50dd66af9e01a4ff320cc137eecc9bacc38"
 
 [[package]]
 name = "jobserver"
@@ -364,6 +404,15 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "js-sys"
+version = "0.3.64"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c5f195fe497f702db0f318b07fdd68edb16955aed830df8363d837542f8f935a"
+dependencies = [
+ "wasm-bindgen",
+]
+
 [[package]]
 name = "kvm-bindings"
 version = "0.6.0"
@@ -385,17 +434,11 @@ dependencies = [
  "vmm-sys-util",
 ]
 
-[[package]]
-name = "lazy_static"
-version = "1.4.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
-
 [[package]]
 name = "libc"
-version = "0.2.141"
+version = "0.2.147"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3304a64d199bb964be99741b7a14d26972741915b3649639149b2479bb46f4b5"
+checksum = "b4668fb0ea861c1df094127ac5f1da3409a82116a4ba74fca2e58ef927159bb3"
 
 [[package]]
 name = "libfuzzer-sys"
@@ -410,22 +453,29 @@ dependencies = [
 
 [[package]]
 name = "linux-loader"
-version = "0.8.1"
+version = "0.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b9259ddbfbb52cc918f6bbc60390004ddd0228cf1d85f402009ff2b3d95de83f"
+checksum = "8d3adb7b28e189741eca3b1a4a27de0bf15e0907c9d4b0c74bd2d7d84ef72e08"
 dependencies = [
  "vm-memory",
 ]
 
 [[package]]
-name = "log"
-version = "0.4.17"
+name = "lock_api"
+version = "0.4.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "abb12e687cfb44aa40f41fc3978ef76448f9b6038cad6aef4259d3c095a2382e"
+checksum = "c1cc9717a20b1bb222f333e6a92fd32f7d8a18ddc5a3191a11af45dcbf4dcd16"
 dependencies = [
- "cfg-if",
+ "autocfg",
+ "scopeguard",
 ]
 
+[[package]]
+name = "log"
+version = "0.4.19"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b06a4cde4c0f271a446782e3eff8de789548ce57dbc8eca9292c27f4a42004b4"
+
 [[package]]
 name = "micro_http"
 version = "0.1.0"
@@ -435,6 +485,15 @@ dependencies = [
  "vmm-sys-util",
 ]
 
+[[package]]
+name = "nanorand"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6a51313c5820b0b02bd422f4b44776fbf47961755c74ce64afc73bfad10226c3"
+dependencies = [
+ "getrandom",
+]
+
 [[package]]
 name = "net_gen"
 version = "0.1.0"
@@ -456,7 +515,7 @@ dependencies = [
  "thiserror",
  "versionize",
  "versionize_derive",
- "virtio-bindings 0.2.0",
+ "virtio-bindings",
  "virtio-queue",
  "vm-memory",
  "vm-virtio",
@@ -465,9 +524,9 @@ dependencies = [
 
 [[package]]
 name = "once_cell"
-version = "1.17.1"
+version = "1.18.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b7e5500299e16ebb147ae15a00a942af264cf3688f47923b8fc2cd5858f23ad3"
+checksum = "dd8b5dd2ae5ed71462c540258bedcb51965123ad7e7ccf4b9a8cafaa4a63576d"
 
 [[package]]
 name = "option_parser"
@@ -497,30 +556,39 @@ dependencies = [
 ]
 
 [[package]]
-name = "proc-macro2"
-version = "1.0.56"
+name = "pin-project"
+version = "1.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2b63bdb0cd06f1f4dedf69b254734f9b45af66e4a031e42a7480257d9898b435"
+checksum = "fda4ed1c6c173e3fc7a83629421152e01d7b1f9b7f65fb301e490e8cfc656422"
+dependencies = [
+ "pin-project-internal",
+]
+
+[[package]]
+name = "pin-project-internal"
+version = "1.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4359fd9c9171ec6e8c62926d6faaf553a8dc3f64e1507e76da7911b4f6a04405"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.23",
+]
+
+[[package]]
+name = "proc-macro2"
+version = "1.0.66"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "18fb31db3f9bddb2ea821cde30a9f70117e3f119938b5ee630b7403aa6e2ead9"
 dependencies = [
  "unicode-ident",
 ]
 
-[[package]]
-name = "qcow"
-version = "0.1.0"
-dependencies = [
- "byteorder",
- "libc",
- "log",
- "remain",
- "vmm-sys-util",
-]
-
 [[package]]
 name = "quote"
-version = "1.0.26"
+version = "1.0.32"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4424af4bf778aae2051a77b60283332f386554255d722233d09fbfc7e30da2fc"
+checksum = "50f3b39ccfb720540debaa0164757101c08ecb8d326b15358ce76a62c7e85965"
 dependencies = [
  "proc-macro2",
 ]
@@ -536,13 +604,13 @@ dependencies = [
 
 [[package]]
 name = "remain"
-version = "0.2.8"
+version = "0.2.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e13cca257d068dd3a390d04b2c3009a3fad2ee5048dfa8f239d048372810470c"
+checksum = "bce3a7139d2ee67d07538ee5dba997364fbc243e7e7143e96eb830c74bfaa082"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.4",
+ "syn 2.0.23",
 ]
 
 [[package]]
@@ -556,9 +624,15 @@ dependencies = [
 
 [[package]]
 name = "ryu"
-version = "1.0.13"
+version = "1.0.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f91339c0467de62360649f8d3e185ca8de4224ff281f66000de5eb2a77a79041"
+checksum = "1ad4cc8da4ef723ed60bced201181d83791ad433213d8c24efffda1eec85d741"
+
+[[package]]
+name = "scopeguard"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
 
 [[package]]
 name = "seccompiler"
@@ -571,35 +645,35 @@ dependencies = [
 
 [[package]]
 name = "semver"
-version = "1.0.17"
+version = "1.0.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bebd363326d05ec3e2f532ab7660680f3b02130d780c299bca73469d521bc0ed"
+checksum = "b0293b4b29daaf487284529cc2f5675b8e57c61f70167ba415a463651fd6a918"
 
 [[package]]
 name = "serde"
-version = "1.0.159"
+version = "1.0.168"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3c04e8343c3daeec41f58990b9d77068df31209f2af111e059e9fe9646693065"
+checksum = "d614f89548720367ded108b3c843be93f3a341e22d5674ca0dd5cd57f34926af"
 dependencies = [
  "serde_derive",
 ]
 
 [[package]]
 name = "serde_derive"
-version = "1.0.159"
+version = "1.0.168"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4c614d17805b093df4b147b51339e7e44bf05ef59fba1e45d83500bcfb4d8585"
+checksum = "d4fe589678c688e44177da4f27152ee2d190757271dc7f1d5b6b9f68d869d641"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.4",
+ "syn 2.0.23",
 ]
 
 [[package]]
 name = "serde_json"
-version = "1.0.95"
+version = "1.0.104"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d721eca97ac802aa7777b701877c8004d950fc142651367300d21c1cc0194744"
+checksum = "076066c5f1078eac5b722a31827a8832fe108bed65dfa75e233c89f8206e976c"
 dependencies = [
  "itoa",
  "ryu",
@@ -608,9 +682,9 @@ dependencies = [
 
 [[package]]
 name = "serde_with"
-version = "2.3.2"
+version = "3.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "331bb8c3bf9b92457ab7abecf07078c13f7d270ba490103e84e8b014490cd0b0"
+checksum = "1402f54f9a3b9e2efe71c1cea24e648acce55887983553eeb858cf3115acfd49"
 dependencies = [
  "serde",
  "serde_with_macros",
@@ -618,14 +692,14 @@ dependencies = [
 
 [[package]]
 name = "serde_with_macros"
-version = "2.3.2"
+version = "3.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "859011bddcc11f289f07f467cc1fe01c7a941daa4d8f6c40d4d1c92eb6d9319c"
+checksum = "9197f1ad0e3c173a0222d3c4404fb04c3afe87e962bcb327af73e8301fa203c7"
 dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn 1.0.109",
+ "syn 2.0.23",
 ]
 
 [[package]]
@@ -634,9 +708,9 @@ version = "0.1.0"
 
 [[package]]
 name = "signal-hook"
-version = "0.3.15"
+version = "0.3.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "732768f1176d21d09e076c23a93123d40bba92d50c4058da34d45c8de8e682b9"
+checksum = "8621587d4798caf8eb44879d42e56b9a93ea5dcd315a6487c357130095b62801"
 dependencies = [
  "libc",
  "signal-hook-registry",
@@ -653,9 +727,18 @@ dependencies = [
 
 [[package]]
 name = "smallvec"
-version = "1.10.0"
+version = "1.11.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a507befe795404456341dfab10cef66ead4c041f62b8b11bbb92bffe5d0953e0"
+checksum = "62bb4feee49fdd9f707ef802e22365a35de4b7b299de4763d44bfea899442ff9"
+
+[[package]]
+name = "spin"
+version = "0.9.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6980e8d7511241f8acf4aebddbb1ff938df5eebe98691418c4468d0b72a96a67"
+dependencies = [
+ "lock_api",
+]
 
 [[package]]
 name = "strsim"
@@ -676,9 +759,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.4"
+version = "2.0.23"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2c622ae390c9302e214c31013517c2061ecb2699935882c60a9b37f82f8625ae"
+checksum = "59fb7d6d8281a51045d62b8eb3a7d1ce347b76f312af50cd3dc0af39c87c1737"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -687,22 +770,22 @@ dependencies = [
 
 [[package]]
 name = "thiserror"
-version = "1.0.40"
+version = "1.0.44"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "978c9a314bd8dc99be594bc3c175faaa9794be04a5a5e153caba6915336cebac"
+checksum = "611040a08a0439f8248d1990b111c95baa9c704c805fa1f62104b39655fd7f90"
 dependencies = [
  "thiserror-impl",
 ]
 
 [[package]]
 name = "thiserror-impl"
-version = "1.0.40"
+version = "1.0.44"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f9456a42c5b0d803c8cd86e73dd7cc9edd429499f37a3550d286d5e86720569f"
+checksum = "090198534930841fab3a5d1bb637cde49e339654e606195f8d9c76eeb081dc96"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.4",
+ "syn 2.0.23",
 ]
 
 [[package]]
@@ -731,15 +814,15 @@ dependencies = [
 
 [[package]]
 name = "unicode-ident"
-version = "1.0.8"
+version = "1.0.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e5464a87b239f13a63a501f2701565754bae92d243d4bb7eb12f6d57d2269bf4"
+checksum = "301abaae475aa91687eb82514b328ab47a211a533026cb25fc3e519b86adfc3c"
 
 [[package]]
 name = "uuid"
-version = "1.3.0"
+version = "1.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1674845326ee10d37ca60470760d4288a6f80f304007d92e5c53bab78c9cfd79"
+checksum = "79daa5ed5740825c40b389c5e50312b9c86df53fccd33f281df655642b43869d"
 dependencies = [
  "getrandom",
 ]
@@ -774,7 +857,7 @@ dependencies = [
 [[package]]
 name = "vfio-bindings"
 version = "0.4.0"
-source = "git+https://github.com/rust-vmm/vfio?branch=main#43439e056ddfa84a4f7906ee7f2f58be70505c08"
+source = "git+https://github.com/rust-vmm/vfio?branch=main#89f8e77dd1a2829197ecde65b686bafcc8a1def4"
 dependencies = [
  "vmm-sys-util",
 ]
@@ -782,7 +865,7 @@ dependencies = [
 [[package]]
 name = "vfio-ioctls"
 version = "0.2.0"
-source = "git+https://github.com/rust-vmm/vfio?branch=main#43439e056ddfa84a4f7906ee7f2f58be70505c08"
+source = "git+https://github.com/rust-vmm/vfio?branch=main#89f8e77dd1a2829197ecde65b686bafcc8a1def4"
 dependencies = [
  "byteorder",
  "kvm-bindings",
@@ -798,9 +881,9 @@ dependencies = [
 [[package]]
 name = "vfio_user"
 version = "0.1.0"
-source = "git+https://github.com/rust-vmm/vfio-user?branch=main#afbbd5722885e961ce12baea12efe01d52ce14b0"
+source = "git+https://github.com/rust-vmm/vfio-user?branch=main#eef6bec4d421f08ed1688fe67c5ea33aabbf5069"
 dependencies = [
- "bitflags",
+ "bitflags 1.3.2",
  "libc",
  "log",
  "serde",
@@ -812,26 +895,13 @@ dependencies = [
  "vmm-sys-util",
 ]
 
-[[package]]
-name = "vhdx"
-version = "0.1.0"
-dependencies = [
- "byteorder",
- "crc32c",
- "libc",
- "log",
- "remain",
- "thiserror",
- "uuid",
-]
-
 [[package]]
 name = "vhost"
-version = "0.6.0"
+version = "0.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c9b791c5b0717a0558888a4cf7240cea836f39a99cb342e12ce633dcaa078072"
+checksum = "84f81f436bca4541f4d33172e1202882c9d437db34ed17fc6d84c8ff2bde21f5"
 dependencies = [
- "bitflags",
+ "bitflags 1.3.2",
  "libc",
  "vm-memory",
  "vmm-sys-util",
@@ -839,15 +909,9 @@ dependencies = [
 
 [[package]]
 name = "virtio-bindings"
-version = "0.1.0"
+version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3ff512178285488516ed85f15b5d0113a7cdb89e9e8a760b269ae4f02b84bd6b"
-
-[[package]]
-name = "virtio-bindings"
-version = "0.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0b9084faf91b9aa9676ae2cac8f1432df2839d9566e6f19f29dbc13a8b831dff"
+checksum = "c18d7b74098a946470ea265b5bacbbf877abc3373021388454de0d47735a5b98"
 
 [[package]]
 name = "virtio-devices"
@@ -855,11 +919,10 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "arc-swap",
- "block_util",
+ "block",
  "byteorder",
  "epoll",
  "event_monitor",
- "io-uring",
  "libc",
  "log",
  "net_gen",
@@ -874,7 +937,7 @@ dependencies = [
  "versionize",
  "versionize_derive",
  "vhost",
- "virtio-bindings 0.2.0",
+ "virtio-bindings",
  "virtio-queue",
  "vm-allocator",
  "vm-device",
@@ -886,12 +949,12 @@ dependencies = [
 
 [[package]]
 name = "virtio-queue"
-version = "0.7.1"
+version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3ba81e2bcc21c0d2fc5e6683e79367e26ad219197423a498df801d79d5ba77bd"
+checksum = "91aebb1df33db33cbf04d4c2445e4f78d0b0c8e65acfd16a4ee95ef63ca252f8"
 dependencies = [
  "log",
- "virtio-bindings 0.1.0",
+ "virtio-bindings",
  "vm-memory",
  "vmm-sys-util",
 ]
@@ -925,9 +988,9 @@ source = "git+https://github.com/rust-vmm/vm-fdt?branch=main#c5a99ab71b130435927
 
 [[package]]
 name = "vm-memory"
-version = "0.10.0"
+version = "0.11.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "688a70366615b45575a424d9c665561c1b5ab2224d494f706b6a6812911a827c"
+checksum = "9d6ea57fe00f9086c59eeeb68e102dd611686bc3c28520fa465996d4d4bdce07"
 dependencies = [
  "arc-swap",
  "libc",
@@ -964,8 +1027,8 @@ dependencies = [
  "anyhow",
  "arc-swap",
  "arch",
- "bitflags",
- "block_util",
+ "bitflags 2.3.3",
+ "block",
  "devices",
  "epoll",
  "event_monitor",
@@ -978,7 +1041,6 @@ dependencies = [
  "once_cell",
  "option_parser",
  "pci",
- "qcow",
  "seccompiler",
  "serde",
  "serde_json",
@@ -991,7 +1053,6 @@ dependencies = [
  "versionize_derive",
  "vfio-ioctls",
  "vfio_user",
- "vhdx",
  "virtio-devices",
  "virtio-queue",
  "vm-allocator",
@@ -1009,7 +1070,7 @@ version = "0.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dd64fe09d8e880e600c324e7d664760a17f56e9672b7495a86381b49e4f72f46"
 dependencies = [
- "bitflags",
+ "bitflags 1.3.2",
  "libc",
  "serde",
  "serde_derive",
@@ -1021,6 +1082,60 @@ version = "0.11.0+wasi-snapshot-preview1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9c8d87e72b64a3b4db28d11ce29237c246188f4f51057d65a7eab63b7987e423"
 
+[[package]]
+name = "wasm-bindgen"
+version = "0.2.87"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7706a72ab36d8cb1f80ffbf0e071533974a60d0a308d01a5d0375bf60499a342"
+dependencies = [
+ "cfg-if",
+ "wasm-bindgen-macro",
+]
+
+[[package]]
+name = "wasm-bindgen-backend"
+version = "0.2.87"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5ef2b6d3c510e9625e5fe6f509ab07d66a760f0885d858736483c32ed7809abd"
+dependencies = [
+ "bumpalo",
+ "log",
+ "once_cell",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.23",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-macro"
+version = "0.2.87"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dee495e55982a3bd48105a7b947fd2a9b4a8ae3010041b9e0faab3f9cd028f1d"
+dependencies = [
+ "quote",
+ "wasm-bindgen-macro-support",
+]
+
+[[package]]
+name = "wasm-bindgen-macro-support"
+version = "0.2.87"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "54681b18a46765f095758388f2d0cf16eb8d4169b639ab575a8f5693af210c7b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.23",
+ "wasm-bindgen-backend",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-shared"
+version = "0.2.87"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca6ad05a4870b2bf5fe995117d3728437bd27d7cd5f06f13c17443ef369775a1"
+
 [[package]]
 name = "winapi"
 version = "0.3.9"
@@ -1045,9 +1160,9 @@ checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
 
 [[package]]
 name = "zerocopy"
-version = "0.6.1"
+version = "0.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "332f188cc1bcf1fe1064b8c58d150f497e697f49774aa846f2dc949d9a25f236"
+checksum = "f3b9c234616391070b0b173963ebc65a9195068e7ed3731c6edac2ec45ebe106"
 dependencies = [
  "byteorder",
  "zerocopy-derive",
@@ -1055,11 +1170,11 @@ dependencies = [
 
 [[package]]
 name = "zerocopy-derive"
-version = "0.3.2"
+version = "0.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6505e6815af7de1746a08f69c69606bb45695a17149517680f3b2149713b19a3"
+checksum = "8f7f3a471f98d0a61c34322fbbfd10c384b07687f680d4119813713f72308d91"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 1.0.109",
+ "syn 2.0.23",
 ]

fuzz/Cargo.toml

@@ -9,23 +9,21 @@ edition = "2021"
 cargo-fuzz = true
 
 [dependencies]
-block_util = { path = "../block_util" }
+block = { path = "../block" }
 devices = { path = "../devices" }
 epoll = "4.3.1"
-libc = "0.2.141"
+libc = "0.2.147"
 libfuzzer-sys = "0.4.6"
-linux-loader = { version = "0.8.1", features = ["elf", "bzimage", "pe"] }
+linux-loader = { version = "0.9.0", features = ["elf", "bzimage", "pe"] }
 micro_http = { git = "https://github.com/firecracker-microvm/micro-http", branch = "main" }
 net_util = { path = "../net_util" }
-once_cell = "1.17.1"
-qcow = { path = "../qcow" }
+once_cell = "1.18.0"
 seccompiler = "0.3.0"
-vhdx = { path = "../vhdx" }
 virtio-devices = { path = "../virtio-devices" }
-virtio-queue = "0.7.1"
+virtio-queue = "0.8.0"
 vmm = { path = "../vmm" }
 vmm-sys-util = "0.11.1"
-vm-memory = "0.10.0"
+vm-memory = "0.11.0"
 vm-device = { path = "../vm-device" }
 vm-virtio = { path = "../vm-virtio" }
 

fuzz/fuzz_targets/block.rs

@@ -8,7 +8,7 @@
 
 #![no_main]
 
-use block_util::{async_io::DiskFile, raw_sync::RawFileDiskSync};
+use block::{async_io::DiskFile, raw_sync::RawFileDiskSync};
 use libfuzzer_sys::fuzz_target;
 use seccompiler::SeccompAction;
 use std::ffi;

fuzz/fuzz_targets/cmos.rs

@@ -6,6 +6,8 @@
 use devices::legacy::Cmos;
 use libc::EFD_NONBLOCK;
 use libfuzzer_sys::fuzz_target;
+use std::sync::atomic::AtomicBool;
+use std::sync::Arc;
 use vm_device::BusDevice;
 use vmm_sys_util::eventfd::EventFd;
 
@@ -25,6 +27,7 @@ fuzz_target!(|bytes| {
         u64::from_le_bytes(below_4g),
         u64::from_le_bytes(above_4g),
         EventFd::new(EFD_NONBLOCK).unwrap(),
+        None,
     );
 
     let mut i = 16;

fuzz/fuzz_targets/qcow.rs

@@ -4,7 +4,7 @@
 
 #![no_main]
 use libfuzzer_sys::fuzz_target;
-use qcow::{QcowFile, RawFile};
+use block::qcow::{QcowFile, RawFile};
 use std::ffi;
 use std::fs::File;
 use std::io::{self, Cursor, Read, Seek, SeekFrom, Write};

fuzz/fuzz_targets/serial.rs

@@ -59,9 +59,13 @@ impl InterruptSourceGroup for TestInterrupt {
         _index: InterruptIndex,
         _config: InterruptSourceConfig,
         _masked: bool,
+        _set_gsi: bool,
     ) -> Result<(), std::io::Error> {
         Ok(())
     }
+    fn set_gsi(&self) -> Result<(), std::io::Error> {
+        Ok(())
+    }
     fn notifier(&self, _index: InterruptIndex) -> Option<EventFd> {
         Some(self.event_fd.try_clone().unwrap())
     }

fuzz/fuzz_targets/vhdx.rs

@@ -8,7 +8,7 @@ use std::ffi;
 use std::fs::File;
 use std::io::{self, Read, Seek, SeekFrom, Write};
 use std::os::unix::io::{FromRawFd, RawFd};
-use vhdx::vhdx::Vhdx;
+use block::vhdx::Vhdx;
 
 // Populate the corpus directory with a test file:
 // truncate -s 16M /tmp/source

hypervisor/Cargo.toml

@@ -7,27 +7,28 @@ license = "Apache-2.0 OR BSD-3-Clause"
 
 [features]
 kvm = ["kvm-ioctls", "kvm-bindings", "vfio-ioctls/kvm"]
-mshv = ["mshv-ioctls", "mshv-bindings", "vfio-ioctls/mshv"]
+mshv = ["mshv-ioctls", "mshv-bindings", "vfio-ioctls/mshv", "iced-x86"]
 tdx = []
 
 [dependencies]
-anyhow = "1.0.69"
+anyhow = "1.0.71"
 byteorder = "1.4.3"
-thiserror = "1.0.39"
+thiserror = "1.0.40"
 libc = "0.2.139"
 log = "0.4.17"
 kvm-ioctls = { version = "0.13.0", optional = true }
 kvm-bindings = { git = "https://github.com/cloud-hypervisor/kvm-bindings", branch = "ch-v0.6.0-tdx", features = ["with-serde", "fam-wrappers"], optional  = true }
 mshv-bindings = { git = "https://github.com/rust-vmm/mshv", branch = "main", features = ["with-serde", "fam-wrappers"], optional  = true }
 mshv-ioctls = { git = "https://github.com/rust-vmm/mshv", branch = "main", optional  = true}
-serde = { version = "1.0.151", features = ["rc", "derive"] }
-serde_with = { version = "2.3.1", default-features = false, features = ["macros"] }
+serde = { version = "1.0.164", features = ["rc", "derive"] }
+serde_with = { version = "3.0.0", default-features = false, features = ["macros"] }
 vfio-ioctls = { git = "https://github.com/rust-vmm/vfio", branch = "main", default-features = false }
-vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-atomic"] }
+vm-memory = { version = "0.11.0", features = ["backend-mmap", "backend-atomic"] }
 vmm-sys-util = { version = "0.11.0", features = ["with-serde"] }
 
 [target.'cfg(target_arch = "x86_64")'.dependencies.iced-x86]
-version = "1.18.0"
+optional = true
+version = "1.19.0"
 default-features = false
 features = ["std", "decoder", "op_code_info", "instr_info", "fast_fmt"]
 

hypervisor/src/arch/x86/emulator/instructions/mov.rs

@@ -239,6 +239,36 @@ impl<T: CpuStateManager> InstructionHandler<T> for Movzx_r64_rm16 {
     movzx!(u64, u16);
 }
 
+pub struct Mov_moffs16_AX;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_moffs16_AX {
+    movzx!(u16, u16);
+}
+
+pub struct Mov_AX_moffs16;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_AX_moffs16 {
+    movzx!(u16, u16);
+}
+
+pub struct Mov_moffs32_EAX;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_moffs32_EAX {
+    movzx!(u32, u32);
+}
+
+pub struct Mov_EAX_moffs32;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_EAX_moffs32 {
+    movzx!(u32, u32);
+}
+
+pub struct Mov_moffs64_RAX;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_moffs64_RAX {
+    movzx!(u64, u64);
+}
+
+pub struct Mov_RAX_moffs64;
+impl<T: CpuStateManager> InstructionHandler<T> for Mov_RAX_moffs64 {
+    movzx!(u64, u64);
+}
+
 #[cfg(test)]
 mod tests {
     #![allow(unused_mut)]
@@ -617,4 +647,123 @@ mod tests {
             .unwrap();
         assert_eq!(eax, value as u64);
     }
+
+    #[test]
+    // movabs ax, ds:0x1337
+    // movabs eax, ds:0x1337
+    // movabs rax, ds:0x1337
+    fn test_mov_memoff_ax() {
+        let test_inputs: [(Register, &[u8]); 3] = [
+            (Register::AX, &[0x66, 0xa1]),
+            (Register::EAX, &[0xa1]),
+            (Register::RAX, &[0x48, 0xa1]),
+        ];
+
+        // Constructs the instruction with the provided inputs and emulates it.
+        fn helper(register: Register, instruction_prefix: &[u8]) {
+            let mem_addr: u64 = 0x1337;
+            let mem_value: u64 = 0x13371337deadbeef;
+            let ip: u64 = 0x1000;
+            let cpu_id = 0;
+
+            let mut instruction_bytes = Vec::new();
+            // instruction prefix with specified register
+            instruction_bytes.extend(instruction_prefix);
+            // 64-bit memory operand
+            instruction_bytes.extend([
+                mem_addr.to_le_bytes()[0],
+                mem_addr.to_le_bytes()[1],
+                0,
+                0,
+                0,
+                0,
+                0,
+                0,
+            ]);
+
+            let memory: [u8; 8] = mem_value.to_le_bytes();
+            let mut vmm = MockVmm::new(ip, vec![], Some((mem_addr, &memory)));
+            assert!(vmm.emulate_first_insn(cpu_id, &instruction_bytes).is_ok());
+
+            let ax: u64 = vmm.cpu_state(cpu_id).unwrap().read_reg(register).unwrap();
+
+            match register {
+                Register::AX => {
+                    assert_eq!(ax as u16, mem_value as u16);
+                }
+                Register::EAX => {
+                    assert_eq!(ax as u32, mem_value as u32);
+                }
+                Register::RAX => {
+                    assert_eq!(ax, mem_value);
+                }
+                _ => panic!(),
+            }
+        }
+
+        for (register, instruction_prefix) in test_inputs {
+            helper(register, instruction_prefix)
+        }
+    }
+
+    #[test]
+    // movabs ds:0x1337, ax
+    // movabs ds:0x1337, eax
+    // movabs ds:0x1337, rax
+    fn test_mov_ax_memoff() {
+        let test_inputs: [(Register, &[u8]); 3] = [
+            (Register::AX, &[0x66, 0xa3]),
+            (Register::EAX, &[0xa3]),
+            (Register::RAX, &[0x48, 0xa3]),
+        ];
+
+        // Constructs the instruction with the provided inputs and emulates it.
+        fn helper(register: Register, instruction_prefix: &[u8]) {
+            let mem_addr: u64 = 0x1337;
+            let ax: u64 = 0x13371337deadbeef;
+            let ip: u64 = 0x1000;
+            let cpu_id = 0;
+
+            let mut instruction_bytes = Vec::new();
+            // instruction prefix with specified register
+            instruction_bytes.extend(instruction_prefix);
+            // 64-bit memory operand
+            instruction_bytes.extend([
+                mem_addr.to_le_bytes()[0],
+                mem_addr.to_le_bytes()[1],
+                0,
+                0,
+                0,
+                0,
+                0,
+                0,
+            ]);
+
+            let mut vmm = MockVmm::new(ip, vec![(Register::RAX, ax)], None);
+            assert!(vmm.emulate_first_insn(cpu_id, &instruction_bytes).is_ok());
+
+            match register {
+                Register::AX => {
+                    let mut memory: [u8; 2] = [0; 2];
+                    vmm.read_memory(mem_addr, &mut memory).unwrap();
+                    assert_eq!(u16::from_le_bytes(memory), ax as u16);
+                }
+                Register::EAX => {
+                    let mut memory: [u8; 4] = [0; 4];
+                    vmm.read_memory(mem_addr, &mut memory).unwrap();
+                    assert_eq!(u32::from_le_bytes(memory), ax as u32);
+                }
+                Register::RAX => {
+                    let mut memory: [u8; 8] = [0; 8];
+                    vmm.read_memory(mem_addr, &mut memory).unwrap();
+                    assert_eq!(u64::from_le_bytes(memory), ax);
+                }
+                _ => panic!(),
+            }
+        }
+
+        for (register, instruction_prefix) in test_inputs {
+            helper(register, instruction_prefix)
+        }
+    }
 }

hypervisor/src/arch/x86/emulator/mod.rs

@@ -524,6 +524,13 @@ impl<'a, T: CpuStateManager> Emulator<'a, T> {
             (mov, Movzx_r64_rm8),
             (mov, Movzx_r32_rm16),
             (mov, Movzx_r64_rm16),
+            // MOV MOFFS
+            (mov, Mov_moffs16_AX),
+            (mov, Mov_AX_moffs16),
+            (mov, Mov_moffs32_EAX),
+            (mov, Mov_EAX_moffs32),
+            (mov, Mov_moffs64_RAX),
+            (mov, Mov_RAX_moffs64),
             // MOVS
             (movs, Movsd_m32_m32),
             (movs, Movsw_m16_m16),

hypervisor/src/arch/x86/mod.rs

@@ -11,6 +11,7 @@
 // Copyright © 2020, Microsoft Corporation
 //
 
+#[cfg(all(feature = "mshv", target_arch = "x86_64"))]
 pub mod emulator;
 pub mod gdt;
 #[allow(non_camel_case_types)]
@@ -26,7 +27,6 @@ pub const MTRR_MEM_TYPE_WB: u64 = 0x6;
 pub const NUM_IOAPIC_PINS: usize = 24;
 
 // X86 Exceptions
-#[allow(clippy::upper_case_acronyms)]
 #[derive(Clone, Debug)]
 pub enum Exception {
     DE = 0,  // Divide Error

hypervisor/src/cpu.rs

@@ -21,6 +21,15 @@ use crate::MpState;
 use thiserror::Error;
 use vm_memory::GuestAddress;
 
+#[cfg(target_arch = "x86_64")]
+#[derive(Copy, Clone, Default)]
+pub enum CpuVendor {
+    #[default]
+    Unknown,
+    Intel,
+    AMD,
+}
+
 #[derive(Error, Debug)]
 ///
 /// Enum for CPU error

hypervisor/src/hypervisor.rs

@@ -9,10 +9,14 @@
 //
 #[cfg(target_arch = "x86_64")]
 use crate::arch::x86::CpuIdEntry;
+#[cfg(target_arch = "x86_64")]
+use crate::cpu::CpuVendor;
 #[cfg(feature = "tdx")]
 use crate::kvm::TdxCapabilities;
 use crate::vm::Vm;
 use crate::HypervisorType;
+#[cfg(target_arch = "x86_64")]
+use std::arch::x86_64;
 use std::sync::Arc;
 use thiserror::Error;
 
@@ -75,6 +79,11 @@ pub enum HypervisorError {
     ///
     #[error("Failed to set partition property:{0}")]
     SetPartitionProperty(#[source] anyhow::Error),
+    ///
+    /// Running on an unsupported CPU
+    ///
+    #[error("Unsupported CPU:{0}")]
+    UnsupportedCpu(#[source] anyhow::Error),
 }
 
 ///
@@ -133,4 +142,30 @@ pub trait Hypervisor: Send + Sync {
     fn get_guest_debug_hw_bps(&self) -> usize {
         unimplemented!()
     }
+
+    /// Get maximum number of vCPUs
+    fn get_max_vcpus(&self) -> u32;
+    #[cfg(target_arch = "x86_64")]
+    ///
+    /// Determine CPU vendor
+    ///
+    fn get_cpu_vendor(&self) -> CpuVendor {
+        // SAFETY: call cpuid with valid leaves
+        unsafe {
+            let leaf = x86_64::__cpuid(0x0);
+
+            if leaf.ebx == 0x756e_6547 && leaf.ecx == 0x6c65_746e && leaf.edx == 0x4965_6e69 {
+                // Vendor string GenuineIntel
+                CpuVendor::Intel
+            } else if leaf.ebx == 0x6874_7541 && leaf.ecx == 0x444d_4163 && leaf.edx == 0x6974_6e65
+            {
+                // Vendor string AuthenticAMD
+                CpuVendor::AMD
+            } else {
+                // Not known yet, the corresponding manufacturer manual should contain the
+                // necesssary info. See also https://wiki.osdev.org/CPUID#CPU_Vendor_ID_String
+                CpuVendor::default()
+            }
+        }
+    }
 }

hypervisor/src/kvm/aarch64/gic/mod.rs

@@ -243,7 +243,6 @@ impl KvmGicV3Its {
     }
 
     /// Method to initialize the GIC device
-    #[allow(clippy::new_ret_no_self)]
     pub fn new(vm: &dyn Vm, config: VgicConfig) -> Result<KvmGicV3Its> {
         // This is inside KVM module
         let vm = vm.as_any().downcast_ref::<KvmVm>().expect("Wrong VM type?");

hypervisor/src/kvm/mod.rs

@@ -330,7 +330,7 @@ impl KvmVm {
         Ok(VfioDeviceFd::new_from_kvm(device_fd))
     }
     /// Checks if a particular `Cap` is available.
-    fn check_extension(&self, c: Cap) -> bool {
+    pub fn check_extension(&self, c: Cap) -> bool {
         self.fd.check_extension(c)
     }
 }
@@ -1084,6 +1084,11 @@ impl hypervisor::Hypervisor for KvmHypervisor {
             self.kvm.get_guest_debug_hw_bps() as usize
         }
     }
+
+    /// Get maximum number of vCPUs
+    fn get_max_vcpus(&self) -> u32 {
+        self.kvm.get_max_vcpus().min(u32::MAX as usize) as u32
+    }
 }
 /// Vcpu struct for KVM
 pub struct KvmVcpu {

hypervisor/src/lib.rs

@@ -36,7 +36,7 @@ pub mod kvm;
 #[cfg(all(feature = "mshv", target_arch = "x86_64"))]
 pub mod mshv;
 
-/// Hypevisor related module
+/// Hypervisor related module
 mod hypervisor;
 
 /// Vm related module
@@ -49,6 +49,8 @@ mod cpu;
 mod device;
 
 pub use crate::hypervisor::{Hypervisor, HypervisorError};
+#[cfg(target_arch = "x86_64")]
+pub use cpu::CpuVendor;
 pub use cpu::{HypervisorCpuError, Vcpu, VmExit};
 pub use device::HypervisorDeviceError;
 #[cfg(all(feature = "kvm", target_arch = "aarch64"))]

hypervisor/src/mshv/mod.rs

@@ -38,9 +38,7 @@ use std::fs::File;
 use std::os::unix::io::AsRawFd;
 
 #[cfg(target_arch = "x86_64")]
-use crate::arch::x86::{
-    CpuIdEntry, FpuState, LapicState, MsrEntry, SpecialRegisters, StandardRegisters,
-};
+use crate::arch::x86::{CpuIdEntry, FpuState, MsrEntry};
 
 const DIRTY_BITMAP_CLEAR_DIRTY: u64 = 0x4;
 const DIRTY_BITMAP_SET_DIRTY: u64 = 0x8;
@@ -279,6 +277,13 @@ impl hypervisor::Hypervisor for MshvHypervisor {
     fn get_supported_cpuid(&self) -> hypervisor::Result<Vec<CpuIdEntry>> {
         Ok(Vec::new())
     }
+
+    /// Get maximum number of vCPUs
+    fn get_max_vcpus(&self) -> u32 {
+        // TODO: Using HV_MAXIMUM_PROCESSORS would be better
+        // but the ioctl API is limited to u8
+        256
+    }
 }
 
 /// Vcpu struct for Microsoft Hypervisor
@@ -307,7 +312,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Returns the vCPU general purpose registers.
     ///
-    fn get_regs(&self) -> cpu::Result<StandardRegisters> {
+    fn get_regs(&self) -> cpu::Result<crate::arch::x86::StandardRegisters> {
         Ok(self
             .fd
             .get_regs()
@@ -318,7 +323,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Sets the vCPU general purpose registers.
     ///
-    fn set_regs(&self, regs: &StandardRegisters) -> cpu::Result<()> {
+    fn set_regs(&self, regs: &crate::arch::x86::StandardRegisters) -> cpu::Result<()> {
         let regs = (*regs).into();
         self.fd
             .set_regs(&regs)
@@ -328,7 +333,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Returns the vCPU special registers.
     ///
-    fn get_sregs(&self) -> cpu::Result<SpecialRegisters> {
+    fn get_sregs(&self) -> cpu::Result<crate::arch::x86::SpecialRegisters> {
         Ok(self
             .fd
             .get_sregs()
@@ -339,7 +344,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Sets the vCPU special registers.
     ///
-    fn set_sregs(&self, sregs: &SpecialRegisters) -> cpu::Result<()> {
+    fn set_sregs(&self, sregs: &crate::arch::x86::SpecialRegisters) -> cpu::Result<()> {
         let sregs = (*sregs).into();
         self.fd
             .set_sregs(&sregs)
@@ -454,10 +459,10 @@ impl cpu::Vcpu for MshvVcpu {
                             /* Advance RIP and update RAX */
                             let arr_reg_name_value = [
                                 (
-                                    hv_x64_register_name_HV_X64_REGISTER_RIP,
+                                    hv_register_name_HV_X64_REGISTER_RIP,
                                     info.header.rip + insn_len,
                                 ),
-                                (hv_x64_register_name_HV_X64_REGISTER_RAX, ret_rax),
+                                (hv_register_name_HV_X64_REGISTER_RAX, ret_rax),
                             ];
                             set_registers_64!(self.fd, arr_reg_name_value)
                                 .map_err(|e| cpu::HypervisorCpuError::SetRegister(e.into()))?;
@@ -503,10 +508,10 @@ impl cpu::Vcpu for MshvVcpu {
                     /* Advance RIP and update RAX */
                     let arr_reg_name_value = [
                         (
-                            hv_x64_register_name_HV_X64_REGISTER_RIP,
+                            hv_register_name_HV_X64_REGISTER_RIP,
                             info.header.rip + insn_len,
                         ),
-                        (hv_x64_register_name_HV_X64_REGISTER_RAX, ret_rax),
+                        (hv_register_name_HV_X64_REGISTER_RAX, ret_rax),
                     ];
                     set_registers_64!(self.fd, arr_reg_name_value)
                         .map_err(|e| cpu::HypervisorCpuError::SetRegister(e.into()))?;
@@ -576,8 +581,14 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// X86 specific call to setup the CPUID registers.
     ///
-    fn set_cpuid2(&self, _cpuid: &[CpuIdEntry]) -> cpu::Result<()> {
-        Ok(())
+    fn set_cpuid2(&self, cpuid: &[CpuIdEntry]) -> cpu::Result<()> {
+        let cpuid: Vec<mshv_bindings::hv_cpuid_entry> = cpuid.iter().map(|e| (*e).into()).collect();
+        let mshv_cpuid = <CpuId>::from_entries(&cpuid)
+            .map_err(|_| cpu::HypervisorCpuError::SetCpuid(anyhow!("failed to create CpuId")))?;
+
+        self.fd
+            .register_intercept_result_cpuid(&mshv_cpuid)
+            .map_err(|e| cpu::HypervisorCpuError::SetCpuid(e.into()))
     }
     #[cfg(target_arch = "x86_64")]
     ///
@@ -590,7 +601,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Returns the state of the LAPIC (Local Advanced Programmable Interrupt Controller).
     ///
-    fn get_lapic(&self) -> cpu::Result<LapicState> {
+    fn get_lapic(&self) -> cpu::Result<crate::arch::x86::LapicState> {
         Ok(self
             .fd
             .get_lapic()
@@ -601,7 +612,7 @@ impl cpu::Vcpu for MshvVcpu {
     ///
     /// Sets the state of the LAPIC (Local Advanced Programmable Interrupt Controller).
     ///
-    fn set_lapic(&self, lapic: &LapicState) -> cpu::Result<()> {
+    fn set_lapic(&self, lapic: &crate::arch::x86::LapicState) -> cpu::Result<()> {
         let lapic: mshv_bindings::LapicState = (*lapic).clone().into();
         self.fd
             .set_lapic(&lapic)

net_gen/src/if_tun.rs

@@ -10,6 +10,7 @@
 #[repr(C)]
 #[derive(Default)]
 pub struct __IncompleteArrayField<T>(::std::marker::PhantomData<T>, [T; 0]);
+#[allow(clippy::missing_safety_doc)]
 impl<T> __IncompleteArrayField<T> {
     #[inline]
     pub const fn new() -> Self {

net_gen/src/iff.rs

@@ -10,6 +10,7 @@
 #[repr(C)]
 #[derive(Default)]
 pub struct __IncompleteArrayField<T>(::std::marker::PhantomData<T>, [T; 0]);
+#[allow(clippy::missing_safety_doc)]
 impl<T> __IncompleteArrayField<T> {
     #[inline]
     pub const fn new() -> Self {

net_gen/src/lib.rs

@@ -2,7 +2,6 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the THIRD-PARTY file.
 
-#![allow(clippy::all)]
 #![allow(non_upper_case_globals)]
 #![allow(non_camel_case_types)]
 #![allow(non_snake_case)]
@@ -28,10 +27,12 @@ pub mod inn;
 // generated with bindgen /usr/include/linux/sockios.h --no-unstable-rust
 // --constified-enum '*' --with-derive-default
 pub mod sockios;
-pub use if_tun::*;
-pub use iff::*;
-pub use inn::*;
-pub use sockios::*;
+pub use if_tun::{
+    sock_fprog, IFF_MULTI_QUEUE, IFF_NO_PI, IFF_TAP, IFF_VNET_HDR, TUN_F_CSUM, TUN_F_TSO4,
+    TUN_F_TSO6, TUN_F_TSO_ECN, TUN_F_UFO,
+};
+pub use iff::{ifreq, net_device_flags_IFF_UP, setsockopt, sockaddr, AF_INET};
+pub use inn::sockaddr_in;
 
 pub const TUNTAP: ::std::os::raw::c_uint = 84;
 

net_util/Cargo.toml

@@ -5,24 +5,24 @@ authors = ["The Chromium OS Authors"]
 edition = "2021"
 
 [dependencies]
-epoll = "4.3.1"
-getrandom = "0.2.8"
+epoll = "4.3.3"
+getrandom = "0.2.10"
 libc = "0.2.139"
 log = "0.4.17"
 net_gen = { path = "../net_gen" }
 rate_limiter = { path = "../rate_limiter" }
-serde = "1.0.151"
-thiserror = "1.0.39"
+serde = "1.0.164"
+thiserror = "1.0.40"
 versionize = "0.1.10"
 versionize_derive = "0.1.4"
 virtio-bindings = "0.2.0"
-virtio-queue = "0.7.1"
-vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
+virtio-queue = "0.8.0"
+vm-memory = { version = "0.11.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
 vm-virtio = { path = "../vm-virtio" }
 vmm-sys-util = "0.11.0"
 
 [dev-dependencies]
-once_cell = "1.17.1"
+once_cell = "1.18.0"
 pnet = "0.33.0"
 pnet_datalink = "0.33.0"
-serde_json = "1.0.95"
+serde_json = "1.0.96"

net_util/src/queue_pair.rs

@@ -87,7 +87,7 @@ impl TxVirtio {
                 let result = unsafe {
                     libc::writev(
                         tap.as_raw_fd() as libc::c_int,
-                        iovecs.as_ptr() as *const libc::iovec,
+                        iovecs.as_ptr(),
                         iovecs.len() as libc::c_int,
                     )
                 };
@@ -226,7 +226,7 @@ impl RxVirtio {
                 let result = unsafe {
                     libc::readv(
                         tap.as_raw_fd() as libc::c_int,
-                        iovecs.as_ptr() as *const libc::iovec,
+                        iovecs.as_ptr(),
                         iovecs.len() as libc::c_int,
                     )
                 };

pci/Cargo.toml

@@ -10,7 +10,7 @@ kvm = ["vfio-ioctls/kvm"]
 mshv = ["vfio-ioctls/mshv"]
 
 [dependencies]
-anyhow = "1.0.69"
+anyhow = "1.0.71"
 byteorder = "1.4.3"
 hypervisor = { path = "../hypervisor" }
 vfio-bindings = { git = "https://github.com/rust-vmm/vfio", branch = "main", features = ["fam-wrappers"] }
@@ -20,11 +20,11 @@ vfio_user = { git = "https://github.com/rust-vmm/vfio-user", branch = "main" }
 vmm-sys-util = "0.11.0"
 libc = "0.2.139"
 log = "0.4.17"
-serde = { version = "1.0.151", features = ["derive"] }
-thiserror = "1.0.39"
+serde = { version = "1.0.164", features = ["derive"] }
+thiserror = "1.0.40"
 versionize = "0.1.10"
 versionize_derive = "0.1.4"
 vm-allocator = { path = "../vm-allocator" }
 vm-device = { path = "../vm-device" }
-vm-memory = { version = "0.10.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
+vm-memory = { version = "0.11.0", features = ["backend-mmap", "backend-atomic", "backend-bitmap"] }
 vm-migration = { path = "../vm-migration" }

pci/src/configuration.rs

@@ -452,10 +452,9 @@ impl From<PciBarType> for PciBarRegionType {
     }
 }
 
-#[allow(clippy::from_over_into)]
-impl Into<PciBarType> for PciBarRegionType {
-    fn into(self) -> PciBarType {
-        match self {
+impl From<PciBarRegionType> for PciBarType {
+    fn from(val: PciBarRegionType) -> Self {
+        match val {
             PciBarRegionType::IoRegion => PciBarType::Io,
             PciBarRegionType::Memory32BitRegion => PciBarType::Mmio32,
             PciBarRegionType::Memory64BitRegion => PciBarType::Mmio64,
@@ -469,10 +468,9 @@ pub enum PciBarPrefetchable {
     Prefetchable = 0x08,
 }
 
-#[allow(clippy::from_over_into)]
-impl Into<bool> for PciBarPrefetchable {
-    fn into(self) -> bool {
-        match self {
+impl From<PciBarPrefetchable> for bool {
+    fn from(val: PciBarPrefetchable) -> Self {
+        match val {
             PciBarPrefetchable::NotPrefetchable => false,
             PciBarPrefetchable::Prefetchable => true,
         }

pci/src/msi.rs

@@ -205,10 +205,15 @@ impl MsiConfig {
                             idx as InterruptIndex,
                             InterruptSourceConfig::MsiIrq(config),
                             state.cap.vector_masked(idx),
+                            false,
                         )
                         .map_err(Error::UpdateInterruptRoute)?;
                 }
 
+                interrupt_source_group
+                    .set_gsi()
+                    .map_err(Error::EnableInterruptRoute)?;
+
                 interrupt_source_group
                     .enable()
                     .map_err(Error::EnableInterruptRoute)?;
@@ -262,6 +267,7 @@ impl MsiConfig {
                     idx as InterruptIndex,
                     InterruptSourceConfig::MsiIrq(config),
                     self.cap.vector_masked(idx),
+                    true,
                 ) {
                     error!("Failed updating vector: {:?}", e);
                 }

pci/src/msix.rs

@@ -107,6 +107,7 @@ impl MsixConfig {
                             idx as InterruptIndex,
                             InterruptSourceConfig::MsiIrq(config),
                             state.masked,
+                            true,
                         )
                         .map_err(Error::UpdateInterruptRoute)?;
 
@@ -182,6 +183,7 @@ impl MsixConfig {
                         idx as InterruptIndex,
                         InterruptSourceConfig::MsiIrq(config),
                         table_entry.masked(),
+                        true,
                     ) {
                         error!("Failed updating vector: {:?}", e);
                     }
@@ -320,6 +322,7 @@ impl MsixConfig {
                 index as InterruptIndex,
                 InterruptSourceConfig::MsiIrq(config),
                 table_entry.masked(),
+                true,
             ) {
                 error!("Failed updating vector: {:?}", e);
             }
@@ -511,6 +514,16 @@ impl MsixCap {
         self.pba & 0xffff_fff8
     }
 
+    pub fn table_set_offset(&mut self, addr: u32) {
+        self.table &= 0x7;
+        self.table += addr;
+    }
+
+    pub fn pba_set_offset(&mut self, addr: u32) {
+        self.pba &= 0x7;
+        self.pba += addr;
+    }
+
     pub fn table_bir(&self) -> u32 {
         self.table & 0x7
     }

pci/src/vfio.rs

@@ -14,6 +14,7 @@ use crate::{
 use anyhow::anyhow;
 use byteorder::{ByteOrder, LittleEndian};
 use hypervisor::HypervisorVmError;
+use libc::{sysconf, _SC_PAGESIZE};
 use std::any::Any;
 use std::collections::{BTreeMap, HashMap};
 use std::io;
@@ -27,6 +28,9 @@ use vfio_bindings::bindings::vfio::*;
 use vfio_ioctls::{
     VfioContainer, VfioDevice, VfioIrq, VfioRegionInfoCap, VfioRegionSparseMmapArea,
 };
+use vm_allocator::page_size::{
+    align_page_size_down, align_page_size_up, is_4k_aligned, is_4k_multiple, is_page_size_aligned,
+};
 use vm_allocator::{AddressAllocator, SystemAllocator};
 use vm_device::interrupt::{
     InterruptIndex, InterruptManager, InterruptSourceGroup, MsiIrqGroupConfig,
@@ -498,6 +502,33 @@ impl VfioCommon {
         Ok(vfio_common)
     }
 
+    /// In case msix table offset is not page size aligned, we need do some fixup to achive it.
+    /// Becuse we don't want the MMIO RW region and trap region overlap each other.
+    fn fixup_msix_region(&mut self, bar_id: u32, region_size: u64) -> u64 {
+        if let Some(msix) = self.interrupt.msix.as_mut() {
+            let msix_cap = &mut msix.cap;
+
+            // Suppose table_bir equals to pba_bir here. Am I right?
+            let (table_offset, table_size) = msix_cap.table_range();
+            if is_page_size_aligned(table_offset) || msix_cap.table_bir() != bar_id {
+                return region_size;
+            }
+
+            let (pba_offset, pba_size) = msix_cap.pba_range();
+            let msix_sz = align_page_size_up(table_size + pba_size);
+            // Expand region to hold RW and trap region which both page size aligned
+            let size = std::cmp::max(region_size * 2, msix_sz * 2);
+            // let table starts from the middle of the region
+            msix_cap.table_set_offset((size / 2) as u32);
+            msix_cap.pba_set_offset((size / 2 + pba_offset - table_offset) as u32);
+
+            size
+        } else {
+            // MSI-X not supported for this device
+            region_size
+        }
+    }
+
     pub(crate) fn allocate_bars(
         &mut self,
         allocator: &Arc<Mutex<SystemAllocator>>,
@@ -661,9 +692,16 @@ impl VfioCommon {
                         .ok_or(PciDeviceError::IoAllocationFailed(region_size))?
                 }
                 PciBarRegionType::Memory64BitRegion => {
-                    // BAR allocation must be naturally aligned
+                    // We need do some fixup to keep MMIO RW region and msix cap region page size
+                    // aligned.
+                    region_size = self.fixup_msix_region(bar_id, region_size);
                     mmio_allocator
-                        .allocate(restored_bar_addr, region_size, Some(region_size))
+                        .allocate(
+                            restored_bar_addr,
+                            region_size,
+                            // SAFETY: FFI call. Trivially safe.
+                            Some(unsafe { sysconf(_SC_PAGESIZE) as GuestUsize }),
+                        )
                         .ok_or(PciDeviceError::IoAllocationFailed(region_size))?
                 }
             };
@@ -800,6 +838,23 @@ impl VfioCommon {
         });
     }
 
+    pub(crate) fn get_msix_cap_idx(&self) -> Option<usize> {
+        let mut cap_next = self
+            .vfio_wrapper
+            .read_config_byte(PCI_CONFIG_CAPABILITY_OFFSET);
+
+        while cap_next != 0 {
+            let cap_id = self.vfio_wrapper.read_config_byte(cap_next.into());
+            if PciCapabilityId::from(cap_id) == PciCapabilityId::MsiX {
+                return Some(cap_next as usize);
+            } else {
+                cap_next = self.vfio_wrapper.read_config_byte((cap_next + 1).into());
+            }
+        }
+
+        None
+    }
+
     pub(crate) fn parse_capabilities(&mut self, bdf: PciBdf) {
         let mut cap_next = self
             .vfio_wrapper
@@ -1154,6 +1209,15 @@ impl VfioCommon {
             return self.configuration.read_reg(reg_idx);
         }
 
+        if let Some(id) = self.get_msix_cap_idx() {
+            let msix = self.interrupt.msix.as_mut().unwrap();
+            if reg_idx * 4 == id + 4 {
+                return msix.cap.table;
+            } else if reg_idx * 4 == id + 8 {
+                return msix.cap.pba;
+            }
+        }
+
         // Since we don't support passing multi-functions devices, we should
         // mask the multi-function bit, bit 7 of the Header Type byte on the
         // register 3.
@@ -1316,18 +1380,6 @@ impl VfioPciDevice {
         self.iommu_attached
     }
 
-    fn align_4k(address: u64) -> u64 {
-        (address + 0xfff) & 0xffff_ffff_ffff_f000
-    }
-
-    fn is_4k_aligned(address: u64) -> bool {
-        (address & 0xfff) == 0
-    }
-
-    fn is_4k_multiple(size: u64) -> bool {
-        (size & 0xfff) == 0
-    }
-
     fn generate_sparse_areas(
         caps: &[VfioRegionInfoCap],
         region_index: u32,
@@ -1339,14 +1391,14 @@ impl VfioPciDevice {
             match cap {
                 VfioRegionInfoCap::SparseMmap(sparse_mmap) => return Ok(sparse_mmap.areas.clone()),
                 VfioRegionInfoCap::MsixMappable => {
-                    if !Self::is_4k_aligned(region_start) {
+                    if !is_4k_aligned(region_start) {
                         error!(
                             "Region start address 0x{:x} must be at least aligned on 4KiB",
                             region_start
                         );
                         return Err(VfioPciError::RegionAlignment);
                     }
-                    if !Self::is_4k_multiple(region_size) {
+                    if !is_4k_multiple(region_size) {
                         error!(
                             "Region size 0x{:x} must be at least a multiple of 4KiB",
                             region_size
@@ -1358,7 +1410,8 @@ impl VfioPciDevice {
                     // the MSI-X PBA table, we must calculate the subregions
                     // around them, leading to a list of sparse areas.
                     // We want to make sure we will still trap MMIO accesses
-                    // to these MSI-X specific ranges.
+                    // to these MSI-X specific ranges. If these region don't align
+                    // with pagesize, we can achive it by enlarging its range.
                     //
                     // Using a BtreeMap as the list provided through the iterator is sorted
                     // by key. This ensures proper split of the whole region.
@@ -1366,10 +1419,14 @@ impl VfioPciDevice {
                     if let Some(msix) = vfio_msix {
                         if region_index == msix.cap.table_bir() {
                             let (offset, size) = msix.cap.table_range();
+                            let offset = align_page_size_down(offset);
+                            let size = align_page_size_up(size);
                             inter_ranges.insert(offset, size);
                         }
                         if region_index == msix.cap.pba_bir() {
                             let (offset, size) = msix.cap.pba_range();
+                            let offset = align_page_size_down(offset);
+                            let size = align_page_size_up(size);
                             inter_ranges.insert(offset, size);
                         }
                     }
@@ -1383,8 +1440,7 @@ impl VfioPciDevice {
                                 size: range_offset - current_offset,
                             });
                         }
-
-                        current_offset = Self::align_4k(range_offset + range_size);
+                        current_offset = align_page_size_down(range_offset + range_size);
                     }
 
                     if region_size > current_offset {
@@ -1482,6 +1538,15 @@ impl VfioPciDevice {
                         return Err(VfioPciError::MmapArea);
                     }
 
+                    if !is_page_size_aligned(area.size) || !is_page_size_aligned(area.offset) {
+                        warn!(
+                            "Could not mmap sparse area that is not page size aligned (offset = 0x{:x}, size = 0x{:x})",
+                            area.offset,
+                            area.size,
+                            );
+                        return Ok(());
+                    }
+
                     let user_memory_region = UserMemoryRegion {
                         slot: (self.memory_slot)(),
                         start: region.start.0 + area.offset,

performance-metrics/Cargo.toml

@@ -7,9 +7,9 @@ build = "../build.rs"
 
 [dependencies]
 argh = "0.1.9"
-dirs = "4.0.0"
-serde = { version = "1.0.151", features = ["rc", "derive"] }
-serde_json = "1.0.95"
+dirs = "5.0.0"
+serde = { version = "1.0.164", features = ["rc", "derive"] }
+serde_json = "1.0.96"
 test_infra = { path = "../test_infra" }
-thiserror = "1.0.39"
+thiserror = "1.0.40"
 wait-timeout = "0.2.0"

performance-metrics/src/main.rs

@@ -671,7 +671,7 @@ fn main() {
     let opts: Options = argh::from_env();
 
     if opts.version {
-        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILT_VERSION"));
+        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILD_VERSION"));
         return;
     }
 

qcow/Cargo.toml

@@ -1,16 +0,0 @@
-[package]
-name = "qcow"
-version = "0.1.0"
-authors = ["The Chromium OS Authors"]
-edition = "2021"
-license = "BSD-3-Clause"
-
-[lib]
-path = "src/qcow.rs"
-
-[dependencies]
-byteorder = "1.4.3"
-libc = "0.2.139"
-log = "0.4.17"
-remain = "0.2.6"
-vmm-sys-util = "0.11.0"

release-notes.md

@@ -1,18 +1,36 @@
+- [v34.0](#v340)
+    - [Paravirtualised Panic Device Support](#paravirtualised-panic-device-support)
+    - [Improvements to VM Core Dump](#improvements-to-vm-core-dump)
+    - [QCOW2 Support for Backing Files](#qcow2-support-for-backing-files)
+    - [Minimum Host Kernel Bump](#minimum-host-kernel-bump)
+    - [Notable Bug Fixes](#notable-bug-fixes)
+    - [Contributors](#contributors)
+- [v33.0](#v330)
+    - [D-Bus based API](#d-bus-based-api)
+    - [Expose Host CPU Cache Details for AArch64](#expose-host-cpu-cache-details-for-aarch64)
+    - [Notable Bug Fixes](#notable-bug-fixes-1)
+    - [Contributors](#contributors-1)
+- [v32.0](#v320)
+    - [Increased PCI Segment Limit](#increased-pci-segment-limit)
+    - [API Changes](#api-changes)
+    - [Notable Bug Fixes](#notable-bug-fixes-2)
+    - [Contributors](#contributors-2)
+- [v31.1](#v311)
 - [v31.0](#v310)
     - [Update to Latest `acpi_tables`](#update-to-latest-acpi_tables)
     - [Update Reference Kernel to 6.2](#update-reference-kernel-to-62)
     - [Improvements on Console `SIGWINCH` Handler](#improvements-on-console-sigwinch-handler)
     - [Remove Directory Support from `MemoryZoneConfig::file`](#remove-directory-support-from-memoryzoneconfigfile)
     - [Documentation Improvements](#documentation-improvements)
-    - [Notable Bug Fixes](#notable-bug-fixes)
-    - [Contributors](#contributors)
+    - [Notable Bug Fixes](#notable-bug-fixes-3)
+    - [Contributors](#contributors-3)
 - [v30.0](#v300)
     - [Command Line Changes for Reduced Binary Size](#command-line-changes-for-reduced-binary-size)
     - [Basic vfio-user Server Support](#basic-vfio-user-server-support)
     - [Heap Profiling Support](#heap-profiling-support)
     - [Documentation Improvements](#documentation-improvements-1)
-    - [Notable Bug Fixes](#notable-bug-fixes-1)
-    - [Contributors](#contributors-1)
+    - [Notable Bug Fixes](#notable-bug-fixes-4)
+    - [Contributors](#contributors-4)
 - [v28.2](#v282)
 - [v29.0](#v290)
     - [Release Binary Supports Both MSHV and KVM](#release-binary-supports-both-mshv-and-kvm)
@@ -22,10 +40,10 @@
     - [`AArch64` Documentation Integration](#aarch64-documentation-integration)
     - [`virtio-block` Counters Enhancement](#virtio-block-counters-enhancement)
     - [TCP Offload Control](#tcp-offload-control)
-    - [Notable Bug Fixes](#notable-bug-fixes-2)
+    - [Notable Bug Fixes](#notable-bug-fixes-5)
     - [Removals](#removals)
     - [Deprecations](#deprecations)
-    - [Contributors](#contributors-2)
+    - [Contributors](#contributors-5)
 - [v28.1](#v281)
 - [v28.0](#v280)
     - [Community Engagement (Reminder)](#community-engagement-reminder)
@@ -33,9 +51,9 @@
     - [Virtualised TPM Support](#virtualised-tpm-support)
     - [Transparent Huge Page Support](#transparent-huge-page-support)
     - [README Quick Start Improved](#readme-quick-start-improved)
-    - [Notable Bug Fixes](#notable-bug-fixes-3)
+    - [Notable Bug Fixes](#notable-bug-fixes-6)
     - [Removals](#removals-1)
-    - [Contributors](#contributors-3)
+    - [Contributors](#contributors-6)
 - [v27.0](#v270)
     - [Community Engagement](#community-engagement)
     - [Prebuilt Packages](#prebuilt-packages)
@@ -44,41 +62,41 @@
     - [Simplified Build Feature Flags](#simplified-build-feature-flags)
     - [Asynchronous Kernel Loading](#asynchronous-kernel-loading)
     - [GDB Support for AArch64](#gdb-support-for-aarch64)
-    - [Notable Bug Fixes](#notable-bug-fixes-4)
+    - [Notable Bug Fixes](#notable-bug-fixes-7)
     - [Deprecations](#deprecations-1)
-    - [Contributors](#contributors-4)
+    - [Contributors](#contributors-7)
 - [v26.0](#v260)
     - [SMBIOS Improvements via `--platform`](#smbios-improvements-via---platform)
     - [Unified Binary MSHV and KVM Support](#unified-binary-mshv-and-kvm-support)
-    - [Notable Bug Fixes](#notable-bug-fixes-5)
+    - [Notable Bug Fixes](#notable-bug-fixes-8)
     - [Deprecations](#deprecations-2)
     - [Removals](#removals-2)
-    - [Contributors](#contributors-5)
+    - [Contributors](#contributors-8)
 - [v25.0](#v250)
     - [`ch-remote` Improvements](#ch-remote-improvements-1)
     - [VM "Coredump" Support](#vm-coredump-support)
-    - [Notable Bug Fixes](#notable-bug-fixes-6)
+    - [Notable Bug Fixes](#notable-bug-fixes-9)
     - [Removals](#removals-3)
-    - [Contributors](#contributors-6)
+    - [Contributors](#contributors-9)
 - [v24.0](#v240)
     - [Bypass Mode for `virtio-iommu`](#bypass-mode-for-virtio-iommu)
     - [Ensure Identifiers Uniqueness](#ensure-identifiers-uniqueness)
     - [Sparse Mmap support](#sparse-mmap-support)
     - [Expose Platform Serial Number](#expose-platform-serial-number)
-    - [Notable Bug Fixes](#notable-bug-fixes-7)
+    - [Notable Bug Fixes](#notable-bug-fixes-10)
     - [Notable Improvements](#notable-improvements)
     - [Deprecations](#deprecations-3)
     - [New on the Website](#new-on-the-website)
-    - [Contributors](#contributors-7)
+    - [Contributors](#contributors-10)
 - [v23.1](#v231)
 - [v23.0](#v230)
     - [vDPA Support](#vdpa-support)
     - [Updated OS Support list](#updated-os-support-list)
     - [`AArch64` Memory Map Improvements](#aarch64-memory-map-improvements)
     - [`AMX` Support](#amx-support)
-    - [Notable Bug Fixes](#notable-bug-fixes-8)
+    - [Notable Bug Fixes](#notable-bug-fixes-11)
     - [Deprecations](#deprecations-4)
-    - [Contributors](#contributors-8)
+    - [Contributors](#contributors-11)
 - [v22.1](#v221)
 - [v22.0](#v220)
     - [GDB Debug Stub Support](#gdb-debug-stub-support)
@@ -89,13 +107,13 @@
     - [PMU Support for AArch64](#pmu-support-for-aarch64)
     - [Documentation Under CC-BY-4.0 License](#documentation-under-cc-by-40-license)
     - [Deprecation of "Classic" `virtiofsd`](#deprecation-of-classic-virtiofsd)
-    - [Notable Bug Fixes](#notable-bug-fixes-9)
-    - [Contributors](#contributors-9)
+    - [Notable Bug Fixes](#notable-bug-fixes-12)
+    - [Contributors](#contributors-12)
 - [v21.0](#v210)
     - [Efficient Local Live Migration (for Live Upgrade)](#efficient-local-live-migration-for-live-upgrade)
     - [Recommended Kernel is Now 5.15](#recommended-kernel-is-now-515)
-    - [Notable Bug fixes](#notable-bug-fixes-10)
-    - [Contributors](#contributors-10)
+    - [Notable Bug fixes](#notable-bug-fixes-13)
+    - [Contributors](#contributors-13)
 - [v20.2](#v202)
 - [v20.1](#v201)
 - [v20.0](#v200)
@@ -104,8 +122,8 @@
     - [Improved VFIO support](#improved-vfio-support)
     - [Safer code](#safer-code)
     - [Extended documentation](#extended-documentation)
-    - [Notable bug fixes](#notable-bug-fixes-11)
-    - [Contributors](#contributors-11)
+    - [Notable bug fixes](#notable-bug-fixes-14)
+    - [Contributors](#contributors-14)
 - [v19.0](#v190)
     - [Improved PTY handling for serial and `virtio-console`](#improved-pty-handling-for-serial-and-virtio-console)
     - [PCI boot time optimisations](#pci-boot-time-optimisations)
@@ -113,8 +131,8 @@
     - [Live migration enhancements](#live-migration-enhancements)
     - [`virtio-mem` support with `vfio-user`](#virtio-mem-support-with-vfio-user)
     - [AArch64 for `virtio-iommu`](#aarch64-for-virtio-iommu)
-    - [Notable bug fixes](#notable-bug-fixes-12)
-    - [Contributors](#contributors-12)
+    - [Notable bug fixes](#notable-bug-fixes-15)
+    - [Contributors](#contributors-15)
 - [v18.0](#v180)
     - [Experimental User Device (`vfio-user`) support](#experimental-user-device-vfio-user-support)
     - [Migration support for `vhost-user` devices](#migration-support-for-vhost-user-devices)
@@ -124,23 +142,23 @@
     - [Live migration on MSHV hypervisor](#live-migration-on-mshv-hypervisor)
     - [AArch64 CPU topology support](#aarch64-cpu-topology-support)
     - [Power button support on AArch64](#power-button-support-on-aarch64)
-    - [Notable bug fixes](#notable-bug-fixes-13)
-    - [Contributors](#contributors-13)
+    - [Notable bug fixes](#notable-bug-fixes-16)
+    - [Contributors](#contributors-16)
 - [v17.0](#v170)
     - [ARM64 NUMA support using ACPI](#arm64-numa-support-using-acpi)
     - [`Seccomp` support for MSHV backend](#seccomp-support-for-mshv-backend)
     - [Hotplug of `macvtap` devices](#hotplug-of-macvtap-devices)
     - [Improved SGX support](#improved-sgx-support)
     - [Inflight tracking for `vhost-user` devices](#inflight-tracking-for-vhost-user-devices)
-    - [Notable bug fixes](#notable-bug-fixes-14)
-    - [Contributors](#contributors-14)
+    - [Notable bug fixes](#notable-bug-fixes-17)
+    - [Contributors](#contributors-17)
 - [v16.0](#v160)
     - [Improved live migration support](#improved-live-migration-support)
     - [Improved `vhost-user` support](#improved-vhost-user-support)
     - [ARM64 ACPI and UEFI support](#arm64-acpi-and-uefi-support)
-    - [Notable bug fixes](#notable-bug-fixes-15)
+    - [Notable bug fixes](#notable-bug-fixes-18)
     - [Removed functionality](#removed-functionality)
-    - [Contributors](#contributors-15)
+    - [Contributors](#contributors-18)
 - [v15.0](#v150)
     - [Version numbering and stability guarantees](#version-numbering-and-stability-guarantees)
     - [Network device rate limiting](#network-device-rate-limiting)
@@ -148,7 +166,7 @@
     - [`--api-socket` supports file descriptor parameter](#--api-socket-supports-file-descriptor-parameter)
     - [Bug fixes](#bug-fixes)
     - [Deprecations](#deprecations-5)
-    - [Contributors](#contributors-16)
+    - [Contributors](#contributors-19)
 - [v0.14.1](#v0141)
 - [v0.14.0](#v0140)
     - [Structured event monitoring](#structured-event-monitoring)
@@ -158,7 +176,7 @@
     - [PTY control for serial and `virtio-console`](#pty-control-for-serial-and-virtio-console)
     - [Block device rate limiting](#block-device-rate-limiting)
     - [Deprecations](#deprecations-6)
-    - [Contributors](#contributors-17)
+    - [Contributors](#contributors-20)
 - [v0.13.0](#v0130)
     - [Wider VFIO device support](#wider-vfio-device-support)
     - [Improved huge page support](#improved-huge-page-support)
@@ -166,13 +184,13 @@
     - [VHD disk image support](#vhd-disk-image-support)
     - [Improved Virtio device threading](#improved-virtio-device-threading)
     - [Clean shutdown support via synthetic power button](#clean-shutdown-support-via-synthetic-power-button)
-    - [Contributors](#contributors-18)
+    - [Contributors](#contributors-21)
 - [v0.12.0](#v0120)
     - [ARM64 enhancements](#arm64-enhancements)
     - [Removal of `vhost-user-net` and `vhost-user-block` self spawning](#removal-of-vhost-user-net-and-vhost-user-block-self-spawning)
     - [Migration of `vhost-user-fs` backend](#migration-of-vhost-user-fs-backend)
     - [Enhanced "info" API](#enhanced-info-api)
-    - [Contributors](#contributors-19)
+    - [Contributors](#contributors-22)
 - [v0.11.0](#v0110)
     - [`io_uring` support by default for `virtio-block`](#io_uring-support-by-default-for-virtio-block)
     - [Windows Guest Support](#windows-guest-support)
@@ -184,15 +202,15 @@
     - [Default Log Level Changed](#default-log-level-changed)
     - [New `--balloon` Parameter Added](#new---balloon-parameter-added)
     - [Experimental `virtio-watchdog` Support](#experimental-virtio-watchdog-support)
-    - [Notable Bug Fixes](#notable-bug-fixes-16)
-    - [Contributors](#contributors-20)
+    - [Notable Bug Fixes](#notable-bug-fixes-19)
+    - [Contributors](#contributors-23)
 - [v0.10.0](#v0100)
     - [`virtio-block` Support for Multiple Descriptors](#virtio-block-support-for-multiple-descriptors)
     - [Memory Zones](#memory-zones)
     - [`Seccomp` Sandbox Improvements](#seccomp-sandbox-improvements)
     - [Preliminary KVM HyperV Emulation Control](#preliminary-kvm-hyperv-emulation-control)
-    - [Notable Bug Fixes](#notable-bug-fixes-17)
-    - [Contributors](#contributors-21)
+    - [Notable Bug Fixes](#notable-bug-fixes-20)
+    - [Contributors](#contributors-24)
 - [v0.9.0](#v090)
     - [`io_uring` Based Block Device Support](#io_uring-based-block-device-support)
     - [Block and Network Device Statistics](#block-and-network-device-statistics)
@@ -205,17 +223,17 @@
     - [Enhancements to ARM64 Support](#enhancements-to-arm64-support)
     - [Intel SGX Support](#intel-sgx-support)
     - [`Seccomp` Sandbox Improvements](#seccomp-sandbox-improvements-1)
-    - [Notable Bug Fixes](#notable-bug-fixes-18)
-    - [Contributors](#contributors-22)
+    - [Notable Bug Fixes](#notable-bug-fixes-21)
+    - [Contributors](#contributors-25)
 - [v0.8.0](#v080)
     - [Experimental Snapshot and Restore Support](#experimental-snapshot-and-restore-support)
     - [Experimental ARM64 Support](#experimental-arm64-support)
     - [Support for Using 5-level Paging in Guests](#support-for-using-5-level-paging-in-guests)
     - [Virtio Device Interrupt Suppression for Network Devices](#virtio-device-interrupt-suppression-for-network-devices)
     - [`vhost_user_fs` Improvements](#vhost_user_fs-improvements)
-    - [Notable Bug Fixes](#notable-bug-fixes-19)
+    - [Notable Bug Fixes](#notable-bug-fixes-22)
     - [Command Line and API Changes](#command-line-and-api-changes)
-    - [Contributors](#contributors-23)
+    - [Contributors](#contributors-26)
 - [v0.7.0](#v070)
     - [Block, Network, Persistent Memory (PMEM), VirtioFS and Vsock hotplug](#block-network-persistent-memory-pmem-virtiofs-and-vsock-hotplug)
     - [Alternative `libc` Support](#alternative-libc-support)
@@ -225,14 +243,14 @@
     - [`Seccomp` Sandboxing](#seccomp-sandboxing)
     - [Updated Distribution Support](#updated-distribution-support)
     - [Command Line and API Changes](#command-line-and-api-changes-1)
-    - [Contributors](#contributors-24)
+    - [Contributors](#contributors-27)
 - [v0.6.0](#v060)
     - [Directly Assigned Devices Hotplug](#directly-assigned-devices-hotplug)
     - [Shared Filesystem Improvements](#shared-filesystem-improvements)
     - [Block and Networking IO Self Offloading](#block-and-networking-io-self-offloading)
     - [Command Line Interface](#command-line-interface)
     - [PVH Boot](#pvh-boot)
-    - [Contributors](#contributors-25)
+    - [Contributors](#contributors-28)
 - [v0.5.1](#v051)
 - [v0.5.0](#v050)
     - [Virtual Machine Dynamic Resizing](#virtual-machine-dynamic-resizing)
@@ -240,7 +258,7 @@
     - [New Interrupt Management Framework](#new-interrupt-management-framework)
     - [Development Tools](#development-tools)
     - [Kata Containers Integration](#kata-containers-integration)
-    - [Contributors](#contributors-26)
+    - [Contributors](#contributors-29)
 - [v0.4.0](#v040)
     - [Dynamic virtual CPUs addition](#dynamic-virtual-cpus-addition)
     - [Programmatic firmware tables generation](#programmatic-firmware-tables-generation)
@@ -249,7 +267,7 @@
     - [Userspace IOAPIC by default](#userspace-ioapic-by-default)
     - [PCI BAR reprogramming](#pci-bar-reprogramming)
     - [New `cloud-hypervisor` organization](#new-cloud-hypervisor-organization)
-    - [Contributors](#contributors-27)
+    - [Contributors](#contributors-30)
 - [v0.3.0](#v030)
     - [Block device offloading](#block-device-offloading)
     - [Network device backend](#network-device-backend)
@@ -276,6 +294,163 @@
     - [Unit testing](#unit-testing)
     - [Integration tests parallelization](#integration-tests-parallelization)
 
+# v34.0
+
+This release has been tracked in our [roadmap
+project](https://github.com/orgs/cloud-hypervisor/projects/6) as iteration
+v34.0. The following user visible changes have been made:
+
+### Paravirtualised Panic Device Support
+
+A new device has been added that can communicate when the guest kernel has
+panicked and share those details with the VMM. This is controlled with a new
+`--pvpanic` command line option and JSON API change equivalent. (#5526)
+
+### Improvements to VM Core Dump
+
+Requesting to dump the guest memory as core dump will now transparently pause
+the VM if required; returning to the original state after. (#5604)
+
+### QCOW2 Support for Backing Files
+
+The support for QCOW2 files has been enhanced to include support for using
+backing files. (#5573)
+
+### Minimum Host Kernel Bump
+
+The minimum supported host kernel is now 5.13 in order to incorporate a bug fix
+for `KVM_FEATURE_ASYNC_PF_INT` functionality. (#5626)
+
+### Notable Bug Fixes
+
+* The x86 emulator is only compiled in if MSHV is compiled in (the kernel
+  carries out this job with KVM) (#5561).
+* A regression has been fixed in VFIO support for devices that use MSI rather
+  than MSI-X (#5658).
+* When triggering a VM shutdown or reset via I/O the vCPU thread will be
+  blocked until that asynchronous event has been received (#5645).
+* Pausing a VM is now a synchronous action with the request only completing
+  when all vCPUs are paused (#5611).
+* Event monitor support now correctly supports concurrent access (#5633).
+* Bug fixes for the QCOW2 file support (#5573).
+
+### Contributors
+
+Many thanks to everyone who has contributed to our release:
+
+* Alyssa Ross <hi@alyssa.is>
+* Anatol Belski <anbelski@linux.microsoft.com>
+* Bo Chen <chen.bo@intel.com>
+* Changyuan Lyu <changyuanl@google.com>
+* Christian Blichmann <cblichmann@google.com>
+* Manish Goregaokar <manishsmail@gmail.com>
+* Omer Faruk Bayram <omer.faruk@sartura.hr>
+* Philipp Schuster <philipp.schuster@cyberus-technology.de>
+* Praveen K Paladugu <prapal@linux.microsoft.com>
+* Rob Bradford <rbradford@rivosinc.com>
+* Ruslan Mstoi <ruslan.mstoi@intel.com>
+* Yi Wang <foxywang@tencent.com>
+* Yong He <alexyonghe@tencent.com>
+* Yu Li <liyu.yukiteru@bytedance.com>
+* dom.song <dom.song@amperecomputing.com>
+
+# v33.0
+
+This release has been tracked in our [roadmap
+project](https://github.com/orgs/cloud-hypervisor/projects/6) as iteration
+v33.0. The following user visible changes have been made:
+
+### D-Bus based API
+
+A D-Bus based API has been added as an alternative to the existing REST
+API. This feature is gated by the `dbus_api` feature. Details can be
+found in the [API documentation](docs/api.md).
+
+### Expose Host CPU Cache Details for AArch64
+
+Now the CPU cache information on the host is properly exposed to the
+guest on AArch64.
+
+### Notable Bug Fixes
+
+* Report errors explicitly to users when VM failed to boot (#5453)
+* Fix VFIO on platforms with non-4k page size (#5450, #5469)
+* Fix TDX initialization (#5454)
+* Ensure all guest memory regions are page-size aligned (#5496)
+* Fix seccomp filter lists related to virtio-console, serial and pty
+  (#5506, #5524)
+* Populate APIC ID properly (#5512)
+* Ignore and warn TAP FDs in more situations (#5522)
+
+### Contributors
+
+Many thanks to everyone who has contributed to our release:
+
+* Alyssa Ross <hi@alyssa.is>
+* Anatol Belski <anbelski@linux.microsoft.com>
+* Bo Chen <chen.bo@intel.com>
+* Jianyong Wu <jianyong.wu@arm.com>
+* Omer Faruk Bayram <omer.faruk@sartura.hr>
+* Rafael Mendonca <rafaelmendsr@gmail.com>
+* Ravi kumar Veeramally <ravikumar.veeramally@intel.com>
+* Rob Bradford <rbradford@rivosinc.com>
+* Ruslan Mstoi <ruslan.mstoi@intel.com>
+* Yu Li <liyu.yukiteru@bytedance.com>
+* zhongbingnan <zhongbingnan@bytedance.com>
+
+# v32.0
+
+This release has been tracked in our [roadmap
+project](https://github.com/orgs/cloud-hypervisor/projects/6) as iteration
+v32.0. The following user visible changes have been made:
+
+### Increased PCI Segment Limit
+
+The maximum number of PCI segments that can be used is now 96 (up from 16).
+
+### API Changes
+
+* The VmmPingResponse now includes the PID as well as the build details.
+  (#5348)
+
+### Notable Bug Fixes
+
+* Ignore and warn TAP FDs sent via the HTTP request body (#5350)
+* Properly preserve and close valid FDs for TAP devices (#5373)
+* Only use `KVM_ARM_VCPU_PMU_V3` if available (#5360)
+* Only touch the tty flags if it's being used (#5343)
+* Fix seccomp filter lists for vhost-user devices (#5361)
+* The number of vCPUs is capped at the hypervisor maximum (#5357)
+* Fixes for TTY reset (#5414)
+* CPU topology fixes on MSHV (#5325)
+* Seccomp fixes for older distributions (#5397)
+
+### Contributors
+
+Many thanks to everyone who has contributed to our release:
+
+* Alyssa Ross <hi@alyssa.is>
+* Anatol Belski <anbelski@linux.microsoft.com>
+* Bo Chen <chen.bo@intel.com>
+* Hao Xu <howeyxu@tencent.com>
+* Muminul Islam <muislam@microsoft.com>
+* Omer Faruk Bayram <omer.faruk@sartura.hr>
+* Rafael Mendonca <rafaelmendsr@gmail.com>
+* Rob Bradford <rbradford@rivosinc.com>
+* Ruslan Mstoi <ruslan.mstoi@intel.com>
+* Smit Gardhariya <gardhariya.smit@gmail.com>
+* Wei Liu <liuwe@microsoft.com>
+
+# v31.1
+
+This is a bug fix release. The following issues have been addressed:
+
+* Ignore and warn TAP FDs sent via the HTTP request body (#5350)
+* Properly preserve and close valid FDs for TAP devices (#5373)
+* Only use `KVM_ARM_VCPU_PMU_V3` if available (#5360)
+* Only touch the tty flags if it's being used (#5343)
+* Fix seccomp filter lists for vhost-user devices (#5361)
+
 # v31.0
 
 This release has been tracked in our [roadmap

resources/Dockerfile

@@ -15,58 +15,67 @@ ARG CARGO_GIT_REGISTRY_DIR="$CLH_BUILD_DIR/cargo_git_registry"
 ENV CARGO_HOME=/usr/local/rust
 ENV RUSTUP_HOME=$CARGO_HOME
 ENV PATH="$PATH:$CARGO_HOME/bin"
+ENV DEBIAN_FRONTEND=noninteractive
 
 # Install all CI dependencies
+# DL3015 ignored cause not installing openvswitch-switch-dpdk recommended packages breaks ovs_dpdk test
+# hadolint ignore=DL3008,DL3015
 RUN apt-get update \
-	&& apt-get -yq upgrade \
-	&& DEBIAN_FRONTEND=noninteractive apt-get install -yq \
-	build-essential \
-	bc \
-	curl \
-	wget \
-	sudo \
-	mtools \
-	musl-tools \
-	libssl-dev \
-	pkg-config \
-	flex \
-	bison \
-	libelf-dev \
-	qemu-utils \
-	libglib2.0-dev \
-	libpixman-1-dev \
-	libseccomp-dev \
-	libcap-ng-dev \
-	socat \
-	dosfstools \
-	cpio \
-	python \
-	python3 \
-	python3-setuptools \
-	ntfs-3g \
-	openvswitch-switch-dpdk \
-	python3-distutils \
-	uuid-dev \
-	iperf3 \
-	zip \
-	git-core \
-	dnsmasq \
-	dmsetup \
+        && apt-get -yq upgrade \
+        && apt-get install --no-install-recommends -yq \
+        build-essential \
+        bc \
+        curl \
+        wget \
+        sudo \
+        mtools \
+        musl-tools \
+        libssl-dev \
+        pkg-config \
+        flex \
+        bison \
+        libelf-dev \
+        qemu-utils \
+        libglib2.0-dev \
+        libpixman-1-dev \
+        libseccomp-dev \
+        libcap-ng-dev \
+        socat \
+        dosfstools \
+        cpio \
+        python \
+        python3 \
+        python3-setuptools \
+        ntfs-3g \
+        python3-distutils \
+        uuid-dev \
+        iperf3 \
+        zip \
+        git-core \
+        dnsmasq \
+        dmsetup \
+        ca-certificates \
+        unzip \
+        iproute2 \
+        dbus \
+    && apt-get install openvswitch-switch-dpdk -yq \
     && apt-get clean \
-    && rm -rf /var/lib/apt/lists/*
+    && rm -rf /var/lib/apt/lists/* /var/log/*log  /var/log/apt/* /var/lib/dpkg/*-old /var/cache/debconf/*-old
 
 RUN update-alternatives --set ovs-vswitchd /usr/lib/openvswitch-switch-dpdk/ovs-vswitchd-dpdk
 
+# hadolint ignore=DL3008
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
-	apt-get update \
-	&& apt-get -yq upgrade \
-	&& DEBIAN_FRONTEND=noninteractive apt-get install -yq gcc-multilib gawk \
-	libtool expect gnutls-dev gnutls-bin libfuse-dev \
-	libjson-glib-dev libgmp-dev libtasn1-dev python3-twisted \
-	net-tools softhsm2 \
-	&& apt-get clean \
-    && rm -rf /var/lib/apt/lists/*;	fi
+        apt-get update \
+        && apt-get -yq upgrade \
+        && apt-get install --no-install-recommends -yq gcc-multilib gawk \
+        libtool expect gnutls-dev gnutls-bin libfuse-dev \
+        libjson-glib-dev libgmp-dev libtasn1-dev python3-twisted \
+        net-tools softhsm2 \
+        && apt-get clean \
+        && rm -rf /var/lib/apt/lists/* /var/log/*log  /var/log/apt/* /var/lib/dpkg/*-old /var/cache/debconf/*-old; fi
 
+# hadolint ignore=DL3008
 RUN if [ "$TARGETARCH" = "arm64" ]; then \
         # On AArch64, `setcap` binary should be installed via `libcap2-bin`.
         # The `setcap` binary is used in integration tests.
@@ -74,7 +83,7 @@ RUN if [ "$TARGETARCH" = "arm64" ]; then \
         # kernel (any version) image in `/boot` and modules in `/lib/modules`.
         apt-get update \
         && apt-get -yq upgrade \
-        && DEBIAN_FRONTEND=noninteractive apt-get install -yq \
+        && apt-get install --no-install-recommends -yq \
         libcap2-bin \
         libguestfs-tools \
         linux-image-generic \
@@ -84,11 +93,12 @@ RUN if [ "$TARGETARCH" = "arm64" ]; then \
         perl \
         texinfo \
         && apt-get clean \
-        && rm -rf /var/lib/apt/lists/*;	fi
+        && rm -rf /var/lib/apt/lists/* /var/log/*log  /var/log/apt/* /var/lib/dpkg/*-old /var/cache/debconf/*-old; fi
 
 # Fix the libssl-dev install
+# hadolint ignore=SC2155
 RUN export ARCH="$(uname -m)" \
-    && cp /usr/include/$ARCH-linux-gnu/openssl/opensslconf.h /usr/include/openssl/
+    && cp /usr/include/"$ARCH"-linux-gnu/openssl/opensslconf.h /usr/include/openssl/
 ENV X86_64_UNKNOWN_LINUX_GNU_OPENSSL_LIB_DIR=/usr/lib/x86_64-linux-gnu/
 ENV X86_64_UNKNOWN_LINUX_MUSL_OPENSSL_LIB_DIR=/usr/lib/x86_64-linux-gnu/
 ENV AARCH64_UNKNOWN_LINUX_GNU_OPENSSL_LIB_DIR=/usr/lib/aarch64-linux-gnu/
@@ -96,9 +106,10 @@ ENV AARCH64_UNKNOWN_LINUX_MUSL_OPENSSL_LIB_DIR=/usr/lib/aarch64-linux-gnu/
 ENV OPENSSL_INCLUDE_DIR=/usr/include/
 
 # Install the rust toolchain
+# hadolint ignore=DL4006,SC2155
 RUN export ARCH="$(uname -m)" \
     && nohup curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal --default-toolchain "$RUST_TOOLCHAIN" \
-    && rustup target add $ARCH-unknown-linux-musl --toolchain "$RUST_TOOLCHAIN" \
+    && rustup target add "$ARCH"-unknown-linux-musl --toolchain "$RUST_TOOLCHAIN" \
     && if [ "$TARGETARCH" = "amd64" ]; then rustup toolchain add --profile minimal $RUST_TOOLCHAIN-x86_64-unknown-linux-musl; fi \
     && if [ "$TARGETARCH" = "amd64" ]; then rustup component add rustfmt; fi \
     && if [ "$TARGETARCH" = "amd64" ]; then rustup component add clippy; fi \
@@ -108,12 +119,14 @@ RUN export ARCH="$(uname -m)" \
     && ln -s "$CARGO_GIT_REGISTRY_DIR" "$CARGO_HOME/git"
 
 # Set the rust environment
-RUN echo 'source $CARGO_HOME/env' >> $HOME/.bashrc \
-    && mkdir $HOME/.cargo \
-    && ln -s $CARGO_HOME/env $HOME/.cargo/env
+# hadolint ignore=SC2016
+RUN echo 'source $CARGO_HOME/env' >> "$HOME"/.bashrc \
+    && mkdir "$HOME"/.cargo \
+    && ln -s $CARGO_HOME/env "$HOME"/.cargo/env
 
 # install SPDK NVMe
 # only for 'x86_64' platform images as 'docker buildx' can't build 'spdk'
+# hadolint ignore=DL3003,SC2046
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
        git clone https://github.com/spdk/spdk \
        && cd spdk \
@@ -123,7 +136,7 @@ RUN if [ "$TARGETARCH" = "amd64" ]; then \
        && ./scripts/pkgdep.sh \
        && apt-get clean \
        && ./configure --with-vfio-user \
-       && make -j `nproc` \
+       && make -j $(nproc) TARGET_ARCHITECTURE=skylake \
        && mkdir /usr/local/bin/spdk-nvme \
        && cp ./build/bin/nvmf_tgt /usr/local/bin/spdk-nvme \
        && cp ./scripts/rpc.py /usr/local/bin/spdk-nvme \
@@ -131,6 +144,7 @@ RUN if [ "$TARGETARCH" = "amd64" ]; then \
        && cd .. && rm -rf spdk; fi
 
 # install swtpm only for x86_64 arch
+# hadolint ignore=DL3003
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
     git clone https://github.com/stefanberger/libtpms libtpms_build \
     && cd libtpms_build \
@@ -151,7 +165,7 @@ RUN if [ "$TARGETARCH" = "amd64" ]; then \
 
 # install ethr tool for performance tests
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
-    wget https://github.com/microsoft/ethr/releases/latest/download/ethr_linux.zip \
+    wget -nv https://github.com/microsoft/ethr/releases/latest/download/ethr_linux.zip \
     && unzip ethr_linux.zip \
     && cp ethr /usr/local/bin \
     && rm ethr_linux.zip; fi

resources/linux-config-aarch64

@@ -1402,7 +1402,8 @@ CONFIG_NVME_MULTIPATH=y
 # CONFIG_MISC_RTSX_PCI is not set
 # CONFIG_HABANA_AI is not set
 # CONFIG_UACCE is not set
-# CONFIG_PVPANIC is not set
+CONFIG_PVPANIC=y
+CONFIG_PVPANIC_PCI=y
 # CONFIG_GP_PCI1XXXX is not set
 # end of Misc devices
 
@@ -3050,23 +3051,16 @@ CONFIG_SYMBOLIC_ERRNAME=y
 # end of printk and dmesg options
 
 CONFIG_DEBUG_KERNEL=y
-CONFIG_DEBUG_MISC=y
+# CONFIG_DEBUG_MISC is not set
 
 #
 # Compile-time checks and compiler options
 #
-CONFIG_DEBUG_INFO=y
 CONFIG_AS_HAS_NON_CONST_LEB128=y
-# CONFIG_DEBUG_INFO_NONE is not set
-CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y
+CONFIG_DEBUG_INFO_NONE=y
+# CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT is not set
 # CONFIG_DEBUG_INFO_DWARF4 is not set
 # CONFIG_DEBUG_INFO_DWARF5 is not set
-# CONFIG_DEBUG_INFO_REDUCED is not set
-CONFIG_DEBUG_INFO_COMPRESSED_NONE=y
-# CONFIG_DEBUG_INFO_COMPRESSED_ZLIB is not set
-# CONFIG_DEBUG_INFO_SPLIT is not set
-# CONFIG_DEBUG_INFO_BTF is not set
-# CONFIG_GDB_SCRIPTS is not set
 CONFIG_FRAME_WARN=2048
 CONFIG_STRIP_ASM_SYMS=y
 # CONFIG_READABLE_ASM is not set

resources/linux-config-x86_64

@@ -1378,7 +1378,8 @@ CONFIG_NVME_MULTIPATH=y
 # CONFIG_MISC_RTSX_PCI is not set
 # CONFIG_HABANA_AI is not set
 # CONFIG_UACCE is not set
-# CONFIG_PVPANIC is not set
+CONFIG_PVPANIC=y
+CONFIG_PVPANIC_PCI=y
 # CONFIG_GP_PCI1XXXX is not set
 # end of Misc devices
 
@@ -2924,23 +2925,16 @@ CONFIG_SYMBOLIC_ERRNAME=y
 # end of printk and dmesg options
 
 CONFIG_DEBUG_KERNEL=y
-CONFIG_DEBUG_MISC=y
+# CONFIG_DEBUG_MISC is not set
 
 #
 # Compile-time checks and compiler options
 #
-CONFIG_DEBUG_INFO=y
 CONFIG_AS_HAS_NON_CONST_LEB128=y
-# CONFIG_DEBUG_INFO_NONE is not set
-CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y
+CONFIG_DEBUG_INFO_NONE=y
+# CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT is not set
 # CONFIG_DEBUG_INFO_DWARF4 is not set
 # CONFIG_DEBUG_INFO_DWARF5 is not set
-# CONFIG_DEBUG_INFO_REDUCED is not set
-CONFIG_DEBUG_INFO_COMPRESSED_NONE=y
-# CONFIG_DEBUG_INFO_COMPRESSED_ZLIB is not set
-# CONFIG_DEBUG_INFO_SPLIT is not set
-# CONFIG_DEBUG_INFO_BTF is not set
-# CONFIG_GDB_SCRIPTS is not set
 CONFIG_FRAME_WARN=2048
 CONFIG_STRIP_ASM_SYMS=y
 # CONFIG_READABLE_ASM is not set

scripts/create-cloud-init.sh

@@ -3,7 +3,7 @@ set -x
 
 rm -f /tmp/ubuntu-cloudinit.img
 mkdosfs -n CIDATA -C /tmp/ubuntu-cloudinit.img 8192
-mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/user-data ::
-mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/meta-data ::
-mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/network-config ::
+mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/local/user-data ::
+mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/local/meta-data ::
+mcopy -oi /tmp/ubuntu-cloudinit.img -s test_data/cloud-init/ubuntu/local/network-config ::
 

scripts/dev_cli.sh

@@ -7,8 +7,8 @@
 CLI_NAME="Cloud Hypervisor"
 
 CTR_IMAGE_TAG="ghcr.io/cloud-hypervisor/cloud-hypervisor"
-CTR_IMAGE_VERSION="20230316-0"
-CTR_IMAGE="${CTR_IMAGE_TAG}:${CTR_IMAGE_VERSION}"
+CTR_IMAGE_VERSION="20230804-0"
+: "${CTR_IMAGE:=${CTR_IMAGE_TAG}:${CTR_IMAGE_VERSION}}"
 
 DOCKER_RUNTIME="docker"
 
@@ -426,7 +426,7 @@ cmd_tests() {
             --env USER="root" \
             --env CH_LIBC="${libc}" \
             "$CTR_IMAGE" \
-            ./scripts/run_integration_tests_"$(uname -m)".sh "$@" || fix_dir_perms $? || exit $?
+            dbus-run-session ./scripts/run_integration_tests_"$(uname -m)".sh "$@" || fix_dir_perms $? || exit $?
     fi
 
     if [ "$integration_sgx" = true ]; then
@@ -539,6 +539,7 @@ cmd_tests() {
             --volume "$CLH_INTEGRATION_WORKLOADS:$CTR_CLH_INTEGRATION_WORKLOADS" \
             --env USER="root" \
             --env CH_LIBC="${libc}" \
+            --env RUST_BACKTRACE="${RUST_BACKTRACE}" \
             "$CTR_IMAGE" \
             ./scripts/run_metrics.sh "$@" || fix_dir_perms $? || exit $?
     fi

scripts/run_integration_tests_aarch64.sh

@@ -95,6 +95,14 @@ update_workloads() {
         popd
     fi
 
+    FOCAL_OS_QCOW2_IMAGE_BACKING_FILE_NAME="focal-server-cloudimg-arm64-custom-20210929-0-backing.qcow2"
+    FOCAL_OS_QCOW2_BACKING_FILE_IMAGE="$WORKLOADS_DIR/$FOCAL_OS_QCOW2_IMAGE_BACKING_FILE_NAME"
+    if [ ! -f "$FOCAL_OS_QCOW2_BACKING_FILE_IMAGE" ]; then
+        pushd $WORKLOADS_DIR
+        time qemu-img create -f qcow2 -b $FOCAL_OS_QCOW2_UNCOMPRESSED_IMAGE -F qcow2 $FOCAL_OS_QCOW2_IMAGE_BACKING_FILE_NAME
+        popd
+    fi
+
     JAMMY_OS_RAW_IMAGE_NAME="jammy-server-cloudimg-arm64-custom-20220329-0.raw"
     JAMMY_OS_RAW_IMAGE_DOWNLOAD_URL="https://cloud-hypervisor.azureedge.net/$JAMMY_OS_RAW_IMAGE_NAME"
     JAMMY_OS_RAW_IMAGE="$WORKLOADS_DIR/$JAMMY_OS_RAW_IMAGE_NAME"
@@ -225,8 +233,8 @@ fi
 
 BUILD_TARGET="aarch64-unknown-linux-${CH_LIBC}"
 if [[ "${BUILD_TARGET}" == "aarch64-unknown-linux-musl" ]]; then
-export TARGET_CC="musl-gcc"
-export RUSTFLAGS="-C link-arg=-lgcc -C link_arg=-specs -C link_arg=/usr/lib/aarch64-linux-musl/musl-gcc.specs"
+    export TARGET_CC="musl-gcc"
+    export RUSTFLAGS="-C link-arg=-lgcc -C link_arg=-specs -C link_arg=/usr/lib/aarch64-linux-musl/musl-gcc.specs"
 fi
 
 export RUST_BACKTRACE=1
@@ -281,4 +289,12 @@ else
     exit $RES
 fi
 
+# Run tests on dbus_api
+if [ $RES -eq 0 ]; then
+    cargo build --features "dbus_api" --all --release --target $BUILD_TARGET
+    export RUST_BACKTRACE=1
+    time cargo test "dbus_api::$test_filter" --target $BUILD_TARGET -- ${test_binary_args[*]}
+    RES=$?
+fi
+
 exit $RES

scripts/run_integration_tests_live_migration.sh

@@ -15,7 +15,7 @@ process_common_args "$@"
 test_features=""
 
 if [ "$hypervisor" = "mshv" ] ;  then
-    test_features="--no-default-features --features mshv"
+    test_features="--features mshv"
 fi
 
 cp scripts/sha1sums-x86_64 $WORKLOADS_DIR
@@ -68,7 +68,7 @@ if [[ "${BUILD_TARGET}" == "x86_64-unknown-linux-musl" ]]; then
     CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
 fi
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 # Test ovs-dpdk relies on hugepages
 HUGEPAGESIZE=`grep Hugepagesize /proc/meminfo | awk '{print $2}'`

scripts/run_integration_tests_rate_limiter.sh

@@ -15,7 +15,7 @@ process_common_args "$@"
 test_features=""
 
 if [ "$hypervisor" = "mshv" ] ;  then
-    test_features="--no-default-features --features mshv"
+    test_features="--features mshv"
 fi
 
 cp scripts/sha1sums-x86_64 $WORKLOADS_DIR
@@ -55,7 +55,7 @@ if [[ "${BUILD_TARGET}" == "x86_64-unknown-linux-musl" ]]; then
     CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
 fi
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 export RUST_BACKTRACE=1
 time cargo test $test_features "rate_limiter::$test_filter" -- --test-threads=1 ${test_binary_args[*]}

scripts/run_integration_tests_sgx.sh

@@ -42,11 +42,11 @@ BUILD_TARGET="$(uname -m)-unknown-linux-${CH_LIBC}"
 CFLAGS=""
 TARGET_CC=""
 if [[ "${BUILD_TARGET}" == "x86_64-unknown-linux-musl" ]]; then
-TARGET_CC="musl-gcc"
-CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
+    TARGET_CC="musl-gcc"
+    CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
 fi
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 export RUST_BACKTRACE=1
 

scripts/run_integration_tests_vfio.sh

@@ -7,9 +7,42 @@ source $(dirname "$0")/test-util.sh
 process_common_args "$@"
 
 WORKLOADS_DIR="$HOME/workloads"
-FOCAL_OS_IMAGE_NAME="focal-server-cloudimg-amd64-custom-20210609-0.raw"
-FOCAL_OS_IMAGE="$WORKLOADS_DIR/$FOCAL_OS_IMAGE_NAME"
+mkdir -p "$WORKLOADS_DIR"
+
+cp scripts/sha1sums-x86_64 $WORKLOADS_DIR
+
+FW_URL=$(curl --silent https://api.github.com/repos/cloud-hypervisor/rust-hypervisor-firmware/releases/latest | grep "browser_download_url" | grep -o 'https://.*[^ "]')
 FW="$WORKLOADS_DIR/hypervisor-fw"
+if [ ! -f "$FW" ]; then
+    pushd $WORKLOADS_DIR
+    time wget --quiet $FW_URL || exit 1
+    popd
+fi
+
+FOCAL_OS_IMAGE_NAME="focal-server-cloudimg-amd64-custom-20210609-0.qcow2"
+FOCAL_OS_IMAGE_URL="https://cloud-hypervisor.azureedge.net/$FOCAL_OS_IMAGE_NAME"
+FOCAL_OS_IMAGE="$WORKLOADS_DIR/$FOCAL_OS_IMAGE_NAME"
+if [ ! -f "$FOCAL_OS_IMAGE" ]; then
+    pushd $WORKLOADS_DIR
+    time wget --quiet $FOCAL_OS_IMAGE_URL || exit 1
+    popd
+fi
+
+FOCAL_OS_RAW_IMAGE_NAME="focal-server-cloudimg-amd64-custom-20210609-0.raw"
+FOCAL_OS_RAW_IMAGE="$WORKLOADS_DIR/$FOCAL_OS_RAW_IMAGE_NAME"
+if [ ! -f "$FOCAL_OS_RAW_IMAGE" ]; then
+    pushd $WORKLOADS_DIR
+    time qemu-img convert -p -f qcow2 -O raw $FOCAL_OS_IMAGE_NAME $FOCAL_OS_RAW_IMAGE_NAME || exit 1
+    popd
+fi
+
+pushd $WORKLOADS_DIR
+sha1sum sha1sums-x86_64 --check --ignore-missing
+if [ $? -ne 0 ]; then
+    echo "sha1sum validation of images failed, remove invalid images to fix the issue."
+    exit 1
+fi
+popd
 
 VMLINUX_IMAGE="$WORKLOADS_DIR/vmlinux"
 build_custom_linux
@@ -31,7 +64,7 @@ VFIO_DIR="$WORKLOADS_DIR/vfio"
 VFIO_DISK_IMAGE="$WORKLOADS_DIR/vfio.img"
 rm -rf $VFIO_DIR $VFIO_DISK_IMAGE
 mkdir -p $VFIO_DIR
-cp $FOCAL_OS_IMAGE $VFIO_DIR
+cp $FOCAL_OS_RAW_IMAGE $VFIO_DIR
 cp $FW $VFIO_DIR
 cp $VMLINUX_IMAGE $VFIO_DIR || exit 1
 
@@ -43,7 +76,7 @@ TARGET_CC="musl-gcc"
 CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
 fi
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 # We always copy a fresh version of our binary for our L2 guest.
 cp target/$BUILD_TARGET/release/cloud-hypervisor $VFIO_DIR

scripts/run_integration_tests_windows_aarch64.sh

@@ -24,8 +24,8 @@ BUILD_TARGET="$(uname -m)-unknown-linux-${CH_LIBC}"
 CFLAGS=""
 TARGET_CC=""
 if [[ "${BUILD_TARGET}" == "aarch64-unknown-linux-musl" ]]; then
-export TARGET_CC="musl-gcc"
-export RUSTFLAGS="-C link-arg=-lgcc -C link_arg=-specs -C link_arg=/usr/lib/aarch64-linux-musl/musl-gcc.specs"
+    export TARGET_CC="musl-gcc"
+    export RUSTFLAGS="-C link-arg=-lgcc -C link_arg=-specs -C link_arg=/usr/lib/aarch64-linux-musl/musl-gcc.specs"
 fi
 
 # Check if the images are present

scripts/run_integration_tests_windows_x86_64.sh

@@ -9,7 +9,7 @@ process_common_args "$@"
 test_features=""
 
 if [ "$hypervisor" = "mshv" ] ;  then
-    test_features="--no-default-features --features mshv"
+    test_features="--features mshv"
 fi
 WIN_IMAGE_FILE="/root/workloads/windows-server-2022-amd64-2.raw"
 
@@ -26,8 +26,8 @@ BUILD_TARGET="$(uname -m)-unknown-linux-${CH_LIBC}"
 CFLAGS=""
 TARGET_CC=""
 if [[ "${BUILD_TARGET}" == "x86_64-unknown-linux-musl" ]]; then
-TARGET_CC="musl-gcc"
-CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
+    TARGET_CC="musl-gcc"
+    CFLAGS="-I /usr/include/x86_64-linux-musl/ -idirafter /usr/include/"
 fi
 
 # Check if the images are present
@@ -44,7 +44,7 @@ dmsetup mknodes
 dmsetup create windows-snapshot-base --table "0 $img_blk_size snapshot-origin /dev/mapper/windows-base"
 dmsetup mknodes
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 export RUST_BACKTRACE=1
 

scripts/run_integration_tests_x86_64.sh

@@ -15,7 +15,7 @@ process_common_args "$@"
 test_features=""
 
 if [ "$hypervisor" = "mshv" ] ;  then
-    test_features="--no-default-features --features mshv"
+    test_features="--features mshv"
 fi
 
 cp scripts/sha1sums-x86_64 $WORKLOADS_DIR
@@ -53,6 +53,14 @@ if [ ! -f "$FOCAL_OS_RAW_IMAGE" ]; then
     popd
 fi
 
+FOCAL_OS_QCOW_BACKING_FILE_IMAGE_NAME="focal-server-cloudimg-amd64-custom-20210609-0-backing.qcow2"
+FOCAL_OS_QCOW_BACKING_FILE_IMAGE="$WORKLOADS_DIR/$FOCAL_OS_QCOW_BACKING_FILE_IMAGE_NAME"
+if [ ! -f "$FOCAL_OS_QCOW_BACKING_FILE_IMAGE" ]; then
+    pushd $WORKLOADS_DIR
+    time qemu-img create -f qcow2 -b $FOCAL_OS_IMAGE -F qcow2 $FOCAL_OS_QCOW_BACKING_FILE_IMAGE_NAME
+    popd
+fi
+
 JAMMY_OS_IMAGE_NAME="jammy-server-cloudimg-amd64-custom-20230119-0.qcow2"
 JAMMY_OS_IMAGE_URL="https://cloud-hypervisor.azureedge.net/$JAMMY_OS_IMAGE_NAME"
 JAMMY_OS_IMAGE="$WORKLOADS_DIR/$JAMMY_OS_IMAGE_NAME"
@@ -156,7 +164,7 @@ cp $VMLINUX_IMAGE $VFIO_DIR || exit 1
 
 BUILD_TARGET="$(uname -m)-unknown-linux-${CH_LIBC}"
 
-cargo build --no-default-features --features "kvm,mshv" --all  --release --target $BUILD_TARGET
+cargo build --features mshv --all  --release --target $BUILD_TARGET
 
 # We always copy a fresh version of our binary for our L2 guest.
 cp target/$BUILD_TARGET/release/cloud-hypervisor $VFIO_DIR
@@ -177,6 +185,9 @@ sudo chmod a+rwX /dev/hugepages
 # Update max locked memory to 'unlimited' to avoid issues with vDPA
 ulimit -l unlimited
 
+# Set number of open descriptors high enough for VFIO tests to run
+ulimit -n 4096
+
 export RUST_BACKTRACE=1
 time cargo test $test_features "common_parallel::$test_filter" -- ${test_binary_args[*]}
 RES=$?
@@ -189,4 +200,13 @@ if [ $RES -eq 0 ]; then
     RES=$?
 fi
 
+# Run tests on dbus_api
+if [ $RES -eq 0 ]; then
+    cargo build --features "mshv,dbus_api" --all  --release --target $BUILD_TARGET
+    export RUST_BACKTRACE=1
+    # integration tests now do not reply on build feature "dbus_api"
+    time cargo test $test_features "dbus_api::$test_filter" -- ${test_binary_args[*]}
+    RES=$?
+fi
+
 exit $RES

scripts/run_metrics.sh

@@ -92,7 +92,7 @@ if [[ "${BUILD_TARGET}" == "${TEST_ARCH}-unknown-linux-musl" ]]; then
     CFLAGS="-I /usr/include/${TEST_ARCH}-linux-musl/ -idirafter /usr/include/"
 fi
 
-cargo build --no-default-features --features "kvm,mshv" --all --release --target $BUILD_TARGET
+cargo build --features mshv --all --release --target $BUILD_TARGET
 
 # setup hugepages
 HUGEPAGESIZE=`grep Hugepagesize /proc/meminfo | awk '{print $2}'`
@@ -107,7 +107,12 @@ fi
 # Ensure that git commands can be run in this directory (for metrics report)
 git config --global --add safe.directory $PWD
 
-export RUST_BACKTRACE=1
+RUST_BACKTRACE_VALUE=`echo $RUST_BACKTRACE`
+if [ -z $RUST_BACKTRACE_VALUE ];then
+	export RUST_BACKTRACE=1
+else
+	echo "RUST_BACKTRACE is set to: $RUST_BACKTRACE_VALUE"
+fi
 time target/$BUILD_TARGET/release/performance-metrics ${test_binary_args[*]}
 RES=$?
 

scripts/run_unit_tests.sh

@@ -9,7 +9,6 @@ BUILD_TARGET=${BUILD_TARGET-x86_64-unknown-linux-gnu}
 cargo_args=("")
 
 if [[ $hypervisor = "mshv" ]]; then
-    cargo_args+=("--no-default-features")
     cargo_args+=("--features $hypervisor")
 elif [[ $(uname -m) = "x86_64" ]]; then
     cargo_args+=("--features tdx")

scripts/sha1sums-x86_64

@@ -3,4 +3,3 @@ f1eccdc5e1b515dbad294426ab081b47ebfb97c0 focal-server-cloudimg-amd64-custom-2021
 7f5a8358243a96adf61f5c20139b29f308f2c0e3 focal-server-cloudimg-amd64-custom-20210609-0.raw
 864c074e2f1bd753667a35188b510d83a1d62793 jammy-server-cloudimg-amd64-custom-20230119-0.qcow2
 24358ee053f94e7f710ea4ce9e7a63eff2a1eb25 jammy-server-cloudimg-amd64-custom-20230119-0.raw
-

src/bin/ch-remote.rs

@@ -11,13 +11,19 @@ use argh::FromArgs;
 use option_parser::{ByteSized, ByteSizedParseError};
 use std::fmt;
 use std::io::Read;
+use std::marker::PhantomData;
 use std::os::unix::net::UnixStream;
 use std::process;
+#[cfg(feature = "dbus_api")]
+use zbus::{dbus_proxy, zvariant::Optional};
+
+type ApiResult = Result<(), Error>;
 
 #[derive(Debug)]
 enum Error {
-    Connect(std::io::Error),
-    ApiClient(ApiClientError),
+    HttpApiClient(ApiClientError),
+    #[cfg(feature = "dbus_api")]
+    DBusApiClient(zbus::Error),
     InvalidMemorySize(ByteSizedParseError),
     InvalidBalloonSize(ByteSizedParseError),
     AddDeviceConfig(vmm::config::Error),
@@ -37,8 +43,9 @@ impl fmt::Display for Error {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         use Error::*;
         match self {
-            ApiClient(e) => e.fmt(f),
-            Connect(e) => write!(f, "Error opening HTTP socket: {e}"),
+            HttpApiClient(e) => e.fmt(f),
+            #[cfg(feature = "dbus_api")]
+            DBusApiClient(e) => write!(f, "Error D-Bus proxy: {e}"),
             InvalidMemorySize(e) => write!(f, "Error parsing memory size: {e:?}"),
             InvalidBalloonSize(e) => write!(f, "Error parsing balloon size: {e:?}"),
             AddDeviceConfig(e) => write!(f, "Error parsing device syntax: {e}"),
@@ -56,12 +63,430 @@ impl fmt::Display for Error {
     }
 }
 
-fn resize_api_command(
-    socket: &mut UnixStream,
+enum TargetApi<'a> {
+    HttpApi(UnixStream, PhantomData<&'a ()>),
+    #[cfg(feature = "dbus_api")]
+    DBusApi(DBusApi1ProxyBlocking<'a>),
+}
+
+#[cfg(feature = "dbus_api")]
+#[dbus_proxy(name = "org.cloudhypervisor.DBusApi1", assume_defaults = false)]
+trait DBusApi1 {
+    fn vmm_ping(&self) -> zbus::Result<String>;
+    fn vmm_shutdown(&self) -> zbus::Result<()>;
+    fn vm_add_device(&self, device_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_disk(&self, disk_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_fs(&self, fs_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_net(&self, net_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_pmem(&self, pmem_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_user_device(&self, vm_add_user_device: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_vdpa(&self, vdpa_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_add_vsock(&self, vsock_config: &str) -> zbus::Result<Optional<String>>;
+    fn vm_boot(&self) -> zbus::Result<()>;
+    fn vm_coredump(&self, vm_coredump_data: &str) -> zbus::Result<()>;
+    fn vm_counters(&self) -> zbus::Result<Optional<String>>;
+    fn vm_create(&self, vm_config: &str) -> zbus::Result<()>;
+    fn vm_delete(&self) -> zbus::Result<()>;
+    fn vm_info(&self) -> zbus::Result<String>;
+    fn vm_pause(&self) -> zbus::Result<()>;
+    fn vm_power_button(&self) -> zbus::Result<()>;
+    fn vm_reboot(&self) -> zbus::Result<()>;
+    fn vm_remove_device(&self, vm_remove_device: &str) -> zbus::Result<()>;
+    fn vm_resize(&self, vm_resize: &str) -> zbus::Result<()>;
+    fn vm_resize_zone(&self, vm_resize_zone: &str) -> zbus::Result<()>;
+    fn vm_restore(&self, restore_config: &str) -> zbus::Result<()>;
+    fn vm_receive_migration(&self, receive_migration_data: &str) -> zbus::Result<()>;
+    fn vm_send_migration(&self, receive_migration_data: &str) -> zbus::Result<()>;
+    fn vm_resume(&self) -> zbus::Result<()>;
+    fn vm_shutdown(&self) -> zbus::Result<()>;
+    fn vm_snapshot(&self, vm_snapshot_config: &str) -> zbus::Result<()>;
+}
+
+#[cfg(feature = "dbus_api")]
+impl<'a> DBusApi1ProxyBlocking<'a> {
+    fn new_connection(name: &'a str, path: &'a str, system_bus: bool) -> Result<Self, zbus::Error> {
+        let connection = if system_bus {
+            zbus::blocking::Connection::system()?
+        } else {
+            zbus::blocking::Connection::session()?
+        };
+
+        Self::builder(&connection)
+            .destination(name)?
+            .path(path)?
+            .build()
+    }
+
+    fn print_response(&self, result: zbus::Result<Optional<String>>) -> ApiResult {
+        result
+            .map(|ret| {
+                if let Some(ref output) = *ret {
+                    println!("{output}");
+                }
+            })
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vmm_ping(&self) -> ApiResult {
+        self.vmm_ping()
+            .map(|ping| println!("{ping}"))
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vmm_shutdown(&self) -> ApiResult {
+        self.vmm_shutdown().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_add_device(&self, device_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_device(device_config))
+    }
+
+    fn api_vm_add_disk(&self, disk_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_disk(disk_config))
+    }
+
+    fn api_vm_add_fs(&self, fs_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_fs(fs_config))
+    }
+
+    fn api_vm_add_net(&self, net_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_net(net_config))
+    }
+
+    fn api_vm_add_pmem(&self, pmem_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_pmem(pmem_config))
+    }
+
+    fn api_vm_add_user_device(&self, vm_add_user_device: &str) -> ApiResult {
+        self.print_response(self.vm_add_user_device(vm_add_user_device))
+    }
+
+    fn api_vm_add_vdpa(&self, vdpa_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_vdpa(vdpa_config))
+    }
+
+    fn api_vm_add_vsock(&self, vsock_config: &str) -> ApiResult {
+        self.print_response(self.vm_add_vsock(vsock_config))
+    }
+
+    fn api_vm_boot(&self) -> ApiResult {
+        self.vm_boot().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_coredump(&self, vm_coredump_data: &str) -> ApiResult {
+        self.vm_coredump(vm_coredump_data)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_counters(&self) -> ApiResult {
+        self.print_response(self.vm_counters())
+    }
+
+    fn api_vm_create(&self, vm_config: &str) -> ApiResult {
+        self.vm_create(vm_config).map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_delete(&self) -> ApiResult {
+        self.vm_delete().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_info(&self) -> ApiResult {
+        self.vm_info()
+            .map(|info| println!("{info}"))
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_pause(&self) -> ApiResult {
+        self.vm_pause().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_power_button(&self) -> ApiResult {
+        self.vm_power_button().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_reboot(&self) -> ApiResult {
+        self.vm_reboot().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_remove_device(&self, vm_remove_device: &str) -> ApiResult {
+        self.vm_remove_device(vm_remove_device)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_resize(&self, vm_resize: &str) -> ApiResult {
+        self.vm_resize(vm_resize).map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_resize_zone(&self, vm_resize_zone: &str) -> ApiResult {
+        self.vm_resize_zone(vm_resize_zone)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_restore(&self, restore_config: &str) -> ApiResult {
+        self.vm_restore(restore_config)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_receive_migration(&self, receive_migration_data: &str) -> ApiResult {
+        self.vm_receive_migration(receive_migration_data)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_send_migration(&self, send_migration_data: &str) -> ApiResult {
+        self.vm_send_migration(send_migration_data)
+            .map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_resume(&self) -> ApiResult {
+        self.vm_resume().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_shutdown(&self) -> ApiResult {
+        self.vm_shutdown().map_err(Error::DBusApiClient)
+    }
+
+    fn api_vm_snapshot(&self, vm_snapshot_config: &str) -> ApiResult {
+        self.vm_snapshot(vm_snapshot_config)
+            .map_err(Error::DBusApiClient)
+    }
+}
+
+impl<'a> TargetApi<'a> {
+    fn do_command(&mut self, toplevel: &TopLevel) -> ApiResult {
+        match self {
+            Self::HttpApi(api_socket, _) => rest_api_do_command(toplevel, api_socket),
+            #[cfg(feature = "dbus_api")]
+            Self::DBusApi(proxy) => dbus_api_do_command(toplevel, proxy),
+        }
+    }
+}
+
+fn rest_api_do_command(toplevel: &TopLevel, socket: &mut UnixStream) -> ApiResult {
+    match toplevel.command {
+        SubCommandEnum::Boot(_) => {
+            simple_api_command(socket, "PUT", "boot", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Delete(_) => {
+            simple_api_command(socket, "PUT", "delete", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::ShutdownVmm(_) => {
+            simple_api_full_command(socket, "PUT", "vmm.shutdown", None)
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Resume(_) => {
+            simple_api_command(socket, "PUT", "resume", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::PowerButton(_) => {
+            simple_api_command(socket, "PUT", "power-button", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Reboot(_) => {
+            simple_api_command(socket, "PUT", "reboot", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Pause(_) => {
+            simple_api_command(socket, "PUT", "pause", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Info(_) => {
+            simple_api_command(socket, "GET", "info", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Counters(_) => {
+            simple_api_command(socket, "GET", "counters", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Ping(_) => {
+            simple_api_full_command(socket, "GET", "vmm.ping", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Shutdown(_) => {
+            simple_api_command(socket, "PUT", "shutdown", None).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Resize(ref config) => {
+            let resize = resize_config(config.cpus, &config.memory, &config.balloon)?;
+            simple_api_command(socket, "PUT", "resize", Some(&resize)).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::ResizeZone(ref config) => {
+            let resize_zone = resize_zone_config(&config.id, &config.size)?;
+            simple_api_command(socket, "PUT", "resize-zone", Some(&resize_zone))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddDevice(ref config) => {
+            let device_config = add_device_config(&config.device_config)?;
+            simple_api_command(socket, "PUT", "add-device", Some(&device_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::RemoveDevice(ref config) => {
+            let remove_device_data = remove_device_config(&config.device_config);
+            simple_api_command(socket, "PUT", "remove-device", Some(&remove_device_data))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddDisk(ref config) => {
+            let disk_config = add_disk_config(&config.disk_config)?;
+            simple_api_command(socket, "PUT", "add-disk", Some(&disk_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddFs(ref config) => {
+            let fs_config = add_fs_config(&config.fs_config)?;
+            simple_api_command(socket, "PUT", "add-fs", Some(&fs_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddPmem(ref config) => {
+            let pmem_config = add_pmem_config(&config.pmem_config)?;
+            simple_api_command(socket, "PUT", "add-pmem", Some(&pmem_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddNet(ref config) => {
+            let (net_config, fds) = add_net_config(&config.net_config)?;
+            simple_api_command_with_fds(socket, "PUT", "add-net", Some(&net_config), fds)
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddUserDevice(ref config) => {
+            let device_config = add_user_device_config(&config.device_config)?;
+            simple_api_command(socket, "PUT", "add-user-device", Some(&device_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddVdpa(ref config) => {
+            let vdpa_config = add_vdpa_config(&config.vdpa_config)?;
+            simple_api_command(socket, "PUT", "add-vdpa", Some(&vdpa_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::AddVsock(ref config) => {
+            let vsock_config = add_vsock_config(&config.vsock_config)?;
+            simple_api_command(socket, "PUT", "add-vsock", Some(&vsock_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Snapshot(ref config) => {
+            let snapshot_config = snapshot_api_config(&config.snapshot_config);
+            simple_api_command(socket, "PUT", "snapshot", Some(&snapshot_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Restore(ref config) => {
+            let restore_config = restore_config(&config.restore_config)?;
+            simple_api_command(socket, "PUT", "restore", Some(&restore_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Coredump(ref config) => {
+            let coredump_config = coredump_config(&config.coredump_config);
+            simple_api_command(socket, "PUT", "coredump", Some(&coredump_config))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::SendMigration(ref config) => {
+            let send_migration_data =
+                send_migration_data(&config.send_migration_config, config.send_migration_local);
+            simple_api_command(socket, "PUT", "send-migration", Some(&send_migration_data))
+                .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::ReceiveMigration(ref config) => {
+            let receive_migration_data = receive_migration_data(&config.receive_migration_config);
+            simple_api_command(
+                socket,
+                "PUT",
+                "receive-migration",
+                Some(&receive_migration_data),
+            )
+            .map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Create(ref config) => {
+            let data = create_data(&config.vm_config)?;
+            simple_api_command(socket, "PUT", "create", Some(&data)).map_err(Error::HttpApiClient)
+        }
+        SubCommandEnum::Version(_) => {
+            // Already handled outside of this function
+            panic!()
+        }
+    }
+}
+
+#[cfg(feature = "dbus_api")]
+fn dbus_api_do_command(toplevel: &TopLevel, proxy: &mut DBusApi1ProxyBlocking<'_>) -> ApiResult {
+    match toplevel.command {
+        SubCommandEnum::Boot(_) => proxy.api_vm_boot(),
+        SubCommandEnum::Delete(_) => proxy.api_vm_delete(),
+        SubCommandEnum::ShutdownVmm(_) => proxy.api_vmm_shutdown(),
+        SubCommandEnum::Resume(_) => proxy.api_vm_resume(),
+        SubCommandEnum::PowerButton(_) => proxy.api_vm_power_button(),
+        SubCommandEnum::Reboot(_) => proxy.api_vm_reboot(),
+        SubCommandEnum::Pause(_) => proxy.api_vm_pause(),
+        SubCommandEnum::Info(_) => proxy.api_vm_info(),
+        SubCommandEnum::Counters(_) => proxy.api_vm_counters(),
+        SubCommandEnum::Ping(_) => proxy.api_vmm_ping(),
+        SubCommandEnum::Shutdown(_) => proxy.api_vm_shutdown(),
+        SubCommandEnum::Resize(ref config) => {
+            let resize = resize_config(config.cpus, &config.memory, &config.balloon)?;
+            proxy.api_vm_resize(&resize)
+        }
+        SubCommandEnum::ResizeZone(ref config) => {
+            let resize_zone = resize_zone_config(&config.id, &config.size)?;
+            proxy.api_vm_resize_zone(&resize_zone)
+        }
+        SubCommandEnum::AddDevice(ref config) => {
+            let device_config = add_device_config(&config.device_config)?;
+            proxy.api_vm_add_device(&device_config)
+        }
+        SubCommandEnum::RemoveDevice(ref config) => {
+            let remove_device_data = remove_device_config(&config.device_config);
+            proxy.api_vm_remove_device(&remove_device_data)
+        }
+        SubCommandEnum::AddDisk(ref config) => {
+            let disk_config = add_disk_config(&config.disk_config)?;
+            proxy.api_vm_add_disk(&disk_config)
+        }
+        SubCommandEnum::AddFs(ref config) => {
+            let fs_config = add_fs_config(&config.fs_config)?;
+            proxy.api_vm_add_fs(&fs_config)
+        }
+        SubCommandEnum::AddPmem(ref config) => {
+            let pmem_config = add_pmem_config(&config.pmem_config)?;
+            proxy.api_vm_add_pmem(&pmem_config)
+        }
+        SubCommandEnum::AddNet(ref config) => {
+            let (net_config, _fds) = add_net_config(&config.net_config)?;
+            proxy.api_vm_add_net(&net_config)
+        }
+        SubCommandEnum::AddUserDevice(ref config) => {
+            let device_config = add_user_device_config(&config.device_config)?;
+            proxy.api_vm_add_user_device(&device_config)
+        }
+        SubCommandEnum::AddVdpa(ref config) => {
+            let vdpa_config = add_vdpa_config(&config.vdpa_config)?;
+            proxy.api_vm_add_vdpa(&vdpa_config)
+        }
+        SubCommandEnum::AddVsock(ref config) => {
+            let vsock_config = add_vsock_config(&config.vsock_config)?;
+            proxy.api_vm_add_vsock(&vsock_config)
+        }
+        SubCommandEnum::Snapshot(ref config) => {
+            let snapshot_config = snapshot_api_config(&config.snapshot_config);
+            proxy.api_vm_snapshot(&snapshot_config)
+        }
+        SubCommandEnum::Restore(ref config) => {
+            let restore_config = restore_config(&config.restore_config)?;
+            proxy.api_vm_restore(&restore_config)
+        }
+        SubCommandEnum::Coredump(ref config) => {
+            let coredump_config = coredump_config(&config.coredump_config);
+            proxy.api_vm_coredump(&coredump_config)
+        }
+        SubCommandEnum::SendMigration(ref config) => {
+            let send_migration_data =
+                send_migration_data(&config.send_migration_config, config.send_migration_local);
+            proxy.api_vm_send_migration(&send_migration_data)
+        }
+        SubCommandEnum::ReceiveMigration(ref config) => {
+            let receive_migration_data = receive_migration_data(&config.receive_migration_config);
+            proxy.api_vm_receive_migration(&receive_migration_data)
+        }
+        SubCommandEnum::Create(ref config) => {
+            let data = create_data(&config.vm_config)?;
+            proxy.api_vm_create(&data)
+        }
+        SubCommandEnum::Version(_) => {
+            // Already handled outside of this function
+            panic!()
+        }
+    }
+}
+fn resize_config(
     desired_vcpus: Option<u8>,
     memory: &Option<String>,
     balloon: &Option<String>,
-) -> Result<(), Error> {
+) -> Result<String, Error> {
     let desired_ram: Option<u64> = if let Some(memory) = memory {
         Some(
             memory
@@ -90,16 +515,10 @@ fn resize_api_command(
         desired_balloon,
     };
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "resize",
-        Some(&serde_json::to_string(&resize).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(serde_json::to_string(&resize).unwrap())
 }
 
-fn resize_zone_api_command(socket: &mut UnixStream, id: &str, size: &str) -> Result<(), Error> {
+fn resize_zone_config(id: &str, size: &str) -> Result<String, Error> {
     let resize_zone = vmm::api::VmResizeZoneData {
         id: id.to_owned(),
         desired_ram: size
@@ -108,89 +527,52 @@ fn resize_zone_api_command(socket: &mut UnixStream, id: &str, size: &str) -> Res
             .0,
     };
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "resize-zone",
-        Some(&serde_json::to_string(&resize_zone).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(serde_json::to_string(&resize_zone).unwrap())
 }
 
-fn add_device_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_device_config(config: &str) -> Result<String, Error> {
     let device_config = vmm::config::DeviceConfig::parse(config).map_err(Error::AddDeviceConfig)?;
+    let device_config = serde_json::to_string(&device_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-device",
-        Some(&serde_json::to_string(&device_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(device_config)
 }
 
-fn add_user_device_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_user_device_config(config: &str) -> Result<String, Error> {
     let device_config =
         vmm::config::UserDeviceConfig::parse(config).map_err(Error::AddUserDeviceConfig)?;
+    let device_config = serde_json::to_string(&device_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-user-device",
-        Some(&serde_json::to_string(&device_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(device_config)
 }
 
-fn remove_device_api_command(socket: &mut UnixStream, id: &str) -> Result<(), Error> {
+fn remove_device_config(id: &str) -> String {
     let remove_device_data = vmm::api::VmRemoveDeviceData { id: id.to_owned() };
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "remove-device",
-        Some(&serde_json::to_string(&remove_device_data).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    serde_json::to_string(&remove_device_data).unwrap()
 }
 
-fn add_disk_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_disk_config(config: &str) -> Result<String, Error> {
     let disk_config = vmm::config::DiskConfig::parse(config).map_err(Error::AddDiskConfig)?;
+    let disk_config = serde_json::to_string(&disk_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-disk",
-        Some(&serde_json::to_string(&disk_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(disk_config)
 }
 
-fn add_fs_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_fs_config(config: &str) -> Result<String, Error> {
     let fs_config = vmm::config::FsConfig::parse(config).map_err(Error::AddFsConfig)?;
+    let fs_config = serde_json::to_string(&fs_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-fs",
-        Some(&serde_json::to_string(&fs_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(fs_config)
 }
 
-fn add_pmem_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_pmem_config(config: &str) -> Result<String, Error> {
     let pmem_config = vmm::config::PmemConfig::parse(config).map_err(Error::AddPmemConfig)?;
+    let pmem_config = serde_json::to_string(&pmem_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-pmem",
-        Some(&serde_json::to_string(&pmem_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(pmem_config)
 }
 
-fn add_net_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_net_config(config: &str) -> Result<(String, Vec<i32>), Error> {
     let mut net_config = vmm::config::NetConfig::parse(config).map_err(Error::AddNetConfig)?;
 
     // NetConfig is modified on purpose here by taking the list of file
@@ -198,113 +580,66 @@ fn add_net_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Erro
     // process would not make any sense since the file descriptor may be
     // represented with different values.
     let fds = net_config.fds.take().unwrap_or_default();
+    let net_config = serde_json::to_string(&net_config).unwrap();
 
-    simple_api_command_with_fds(
-        socket,
-        "PUT",
-        "add-net",
-        Some(&serde_json::to_string(&net_config).unwrap()),
-        fds,
-    )
-    .map_err(Error::ApiClient)
+    Ok((net_config, fds))
 }
 
-fn add_vdpa_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_vdpa_config(config: &str) -> Result<String, Error> {
     let vdpa_config = vmm::config::VdpaConfig::parse(config).map_err(Error::AddVdpaConfig)?;
+    let vdpa_config = serde_json::to_string(&vdpa_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-vdpa",
-        Some(&serde_json::to_string(&vdpa_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(vdpa_config)
 }
 
-fn add_vsock_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn add_vsock_config(config: &str) -> Result<String, Error> {
     let vsock_config = vmm::config::VsockConfig::parse(config).map_err(Error::AddVsockConfig)?;
+    let vsock_config = serde_json::to_string(&vsock_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "add-vsock",
-        Some(&serde_json::to_string(&vsock_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(vsock_config)
 }
 
-fn snapshot_api_command(socket: &mut UnixStream, url: &str) -> Result<(), Error> {
+fn snapshot_api_config(url: &str) -> String {
     let snapshot_config = vmm::api::VmSnapshotConfig {
         destination_url: String::from(url),
     };
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "snapshot",
-        Some(&serde_json::to_string(&snapshot_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    serde_json::to_string(&snapshot_config).unwrap()
 }
 
-fn restore_api_command(socket: &mut UnixStream, config: &str) -> Result<(), Error> {
+fn restore_config(config: &str) -> Result<String, Error> {
     let restore_config = vmm::config::RestoreConfig::parse(config).map_err(Error::Restore)?;
+    let restore_config = serde_json::to_string(&restore_config).unwrap();
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "restore",
-        Some(&serde_json::to_string(&restore_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    Ok(restore_config)
 }
 
-fn coredump_api_command(socket: &mut UnixStream, destination_url: &str) -> Result<(), Error> {
+fn coredump_config(destination_url: &str) -> String {
     let coredump_config = vmm::api::VmCoredumpData {
         destination_url: String::from(destination_url),
     };
 
-    simple_api_command(
-        socket,
-        "PUT",
-        "coredump",
-        Some(&serde_json::to_string(&coredump_config).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+    serde_json::to_string(&coredump_config).unwrap()
 }
 
-fn receive_migration_api_command(socket: &mut UnixStream, url: &str) -> Result<(), Error> {
+fn receive_migration_data(url: &str) -> String {
     let receive_migration_data = vmm::api::VmReceiveMigrationData {
         receiver_url: url.to_owned(),
     };
-    simple_api_command(
-        socket,
-        "PUT",
-        "receive-migration",
-        Some(&serde_json::to_string(&receive_migration_data).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+
+    serde_json::to_string(&receive_migration_data).unwrap()
 }
 
-fn send_migration_api_command(
-    socket: &mut UnixStream,
-    url: &str,
-    local: bool,
-) -> Result<(), Error> {
+fn send_migration_data(url: &str, local: bool) -> String {
     let send_migration_data = vmm::api::VmSendMigrationData {
         destination_url: url.to_owned(),
         local,
     };
-    simple_api_command(
-        socket,
-        "PUT",
-        "send-migration",
-        Some(&serde_json::to_string(&send_migration_data).unwrap()),
-    )
-    .map_err(Error::ApiClient)
+
+    serde_json::to_string(&send_migration_data).unwrap()
 }
 
-fn create_api_command(socket: &mut UnixStream, path: &str) -> Result<(), Error> {
+fn create_data(path: &str) -> Result<String, Error> {
     let mut data = String::default();
     if path == "-" {
         std::io::stdin()
@@ -314,100 +649,7 @@ fn create_api_command(socket: &mut UnixStream, path: &str) -> Result<(), Error>
         data = std::fs::read_to_string(path).map_err(Error::ReadingFile)?;
     }
 
-    simple_api_command(socket, "PUT", "create", Some(&data)).map_err(Error::ApiClient)
-}
-
-fn do_command(toplevel: &TopLevel) -> Result<(), Error> {
-    let mut socket =
-        UnixStream::connect(toplevel.api_socket.as_deref().unwrap()).map_err(Error::Connect)?;
-
-    match toplevel.command {
-        SubCommandEnum::Boot(_) => {
-            simple_api_command(&mut socket, "PUT", "boot", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Delete(_) => {
-            simple_api_command(&mut socket, "PUT", "delete", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::ShutdownVmm(_) => {
-            simple_api_full_command(&mut socket, "PUT", "vmm.shutdown", None)
-                .map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Resume(_) => {
-            simple_api_command(&mut socket, "PUT", "resume", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::PowerButton(_) => {
-            simple_api_command(&mut socket, "PUT", "power-button", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Reboot(_) => {
-            simple_api_command(&mut socket, "PUT", "reboot", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Pause(_) => {
-            simple_api_command(&mut socket, "PUT", "pause", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Info(_) => {
-            simple_api_command(&mut socket, "GET", "info", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Counters(_) => {
-            simple_api_command(&mut socket, "GET", "counters", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Ping(_) => {
-            simple_api_full_command(&mut socket, "GET", "vmm.ping", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Shutdown(_) => {
-            simple_api_command(&mut socket, "PUT", "shutdown", None).map_err(Error::ApiClient)
-        }
-        SubCommandEnum::Resize(ref config) => {
-            resize_api_command(&mut socket, config.cpus, &config.memory, &config.balloon)
-        }
-        SubCommandEnum::ResizeZone(ref config) => {
-            resize_zone_api_command(&mut socket, &config.id, &config.size)
-        }
-        SubCommandEnum::AddDevice(ref config) => {
-            add_device_api_command(&mut socket, &config.device_config)
-        }
-        SubCommandEnum::RemoveDevice(ref config) => {
-            remove_device_api_command(&mut socket, &config.device_config)
-        }
-        SubCommandEnum::AddDisk(ref config) => {
-            add_disk_api_command(&mut socket, &config.disk_config)
-        }
-        SubCommandEnum::AddFs(ref config) => add_fs_api_command(&mut socket, &config.fs_config),
-        SubCommandEnum::AddPmem(ref config) => {
-            add_pmem_api_command(&mut socket, &config.pmem_config)
-        }
-        SubCommandEnum::AddNet(ref config) => add_net_api_command(&mut socket, &config.net_config),
-        SubCommandEnum::AddUserDevice(ref config) => {
-            add_user_device_api_command(&mut socket, &config.device_config)
-        }
-        SubCommandEnum::AddVdpa(ref config) => {
-            add_vdpa_api_command(&mut socket, &config.vdpa_config)
-        }
-        SubCommandEnum::AddVsock(ref config) => {
-            add_vsock_api_command(&mut socket, &config.vsock_config)
-        }
-        SubCommandEnum::Snapshot(ref config) => {
-            snapshot_api_command(&mut socket, &config.snapshot_config)
-        }
-        SubCommandEnum::Restore(ref config) => {
-            restore_api_command(&mut socket, &config.restore_config)
-        }
-        SubCommandEnum::Coredump(ref config) => {
-            coredump_api_command(&mut socket, &config.coredump_config)
-        }
-        SubCommandEnum::SendMigration(ref config) => send_migration_api_command(
-            &mut socket,
-            &config.send_migration_config,
-            config.send_migration_local,
-        ),
-        SubCommandEnum::ReceiveMigration(ref config) => {
-            receive_migration_api_command(&mut socket, &config.receive_migration_config)
-        }
-        SubCommandEnum::Create(ref config) => create_api_command(&mut socket, &config.vm_config),
-        SubCommandEnum::Version(_) => {
-            // Already handled outside of this function
-            panic!()
-        }
-    }
+    Ok(data)
 }
 
 #[derive(FromArgs, PartialEq, Debug)]
@@ -419,6 +661,21 @@ struct TopLevel {
     #[argh(option, long = "api-socket")]
     /// HTTP API socket path (UNIX domain socket)
     api_socket: Option<String>,
+
+    #[cfg(feature = "dbus_api")]
+    #[argh(option, long = "dbus-service-name")]
+    /// well known name of the dbus service
+    dbus_name: Option<String>,
+
+    #[cfg(feature = "dbus_api")]
+    #[argh(option, long = "dbus-object-path")]
+    /// object path which the interface is being served at
+    dbus_path: Option<String>,
+
+    #[cfg(feature = "dbus_api")]
+    #[argh(switch, long = "dbus-system-bus")]
+    /// use the system bus instead of a session bus
+    dbus_system_bus: bool,
 }
 
 #[derive(FromArgs, PartialEq, Debug)]
@@ -688,16 +945,56 @@ fn main() {
     let toplevel: TopLevel = argh::from_env();
 
     if matches!(toplevel.command, SubCommandEnum::Version(_)) {
-        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILT_VERSION"));
+        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILD_VERSION"));
         return;
     }
 
-    if toplevel.api_socket.is_none() {
-        println!("Please specify --api-socket");
-        process::exit(1)
-    }
+    let mut target_api = match (
+        &toplevel.api_socket,
+        #[cfg(feature = "dbus_api")]
+        &toplevel.dbus_name,
+        #[cfg(feature = "dbus_api")]
+        &toplevel.dbus_path,
+    ) {
+        #[cfg(not(feature = "dbus_api"))]
+        (Some(ref api_socket),) => TargetApi::HttpApi(
+            UnixStream::connect(api_socket).unwrap_or_else(|e| {
+                eprintln!("Error opening HTTP socket: {e}");
+                process::exit(1)
+            }),
+            PhantomData,
+        ),
+        #[cfg(feature = "dbus_api")]
+        (Some(ref api_socket), None, None) => TargetApi::HttpApi(
+            UnixStream::connect(api_socket).unwrap_or_else(|e| {
+                eprintln!("Error opening HTTP socket: {e}");
+                process::exit(1)
+            }),
+            PhantomData,
+        ),
+        #[cfg(feature = "dbus_api")]
+        (None, Some(ref dbus_name), Some(ref dbus_path)) => TargetApi::DBusApi(
+            DBusApi1ProxyBlocking::new_connection(dbus_name, dbus_path, toplevel.dbus_system_bus)
+                .map_err(Error::DBusApiClient)
+                .unwrap_or_else(|e| {
+                    eprintln!("Error creating D-Bus proxy: {e}");
+                    process::exit(1)
+                }),
+        ),
+        #[cfg(feature = "dbus_api")]
+        (Some(_), Some(_) | None, Some(_) | None) => {
+            println!(
+                "`api-socket` and (dbus-service-name or dbus-object-path) are mutually exclusive"
+            );
+            process::exit(1);
+        }
+        _ => {
+            println!("Please either provide the api-socket option or dbus-service-name and dbus-object-path options");
+            process::exit(1);
+        }
+    };
 
-    if let Err(e) = do_command(&toplevel) {
+    if let Err(e) = target_api.do_command(&toplevel) {
         eprintln!("Error running command: {e}");
         process::exit(1)
     };

src/main.rs

@@ -8,7 +8,7 @@ extern crate event_monitor;
 
 use argh::FromArgs;
 use libc::EFD_NONBLOCK;
-use log::LevelFilter;
+use log::{warn, LevelFilter};
 use option_parser::OptionParser;
 use seccompiler::SeccompAction;
 use signal_hook::consts::SIGSYS;
@@ -18,10 +18,11 @@ use std::os::unix::io::{FromRawFd, RawFd};
 use std::sync::mpsc::channel;
 use std::sync::{Arc, Mutex};
 use thiserror::Error;
+#[cfg(feature = "dbus_api")]
+use vmm::api::dbus::{dbus_api_graceful_shutdown, DBusApiOptions};
 use vmm::config;
 use vmm_sys_util::eventfd::EventFd;
 use vmm_sys_util::signal::block_signal;
-use vmm_sys_util::terminal::Terminal;
 
 #[cfg(feature = "dhat-heap")]
 #[global_allocator]
@@ -34,6 +35,8 @@ enum Error {
     #[cfg(feature = "guest_debug")]
     #[error("Failed to create Debug EventFd: {0}")]
     CreateDebugEventFd(#[source] std::io::Error),
+    #[error("Failed to create exit EventFd: {0}")]
+    CreateExitEventFd(#[source] std::io::Error),
     #[error("Failed to open hypervisor interface (is hypervisor interface available?): {0}")]
     CreateHypervisor(#[source] hypervisor::HypervisorError),
     #[error("Failed to start the VMM thread: {0}")]
@@ -56,10 +59,18 @@ enum Error {
     ParsingApiSocket(std::num::ParseIntError),
     #[error("Error parsing --event-monitor: {0}")]
     ParsingEventMonitor(option_parser::OptionParserError),
+    #[cfg(feature = "dbus_api")]
+    #[error("`--dbus-object-path` option isn't provided")]
+    MissingDBusObjectPath,
+    #[cfg(feature = "dbus_api")]
+    #[error("`--dbus-service-name` option isn't provided")]
+    MissingDBusServiceName,
     #[error("Error parsing --event-monitor: path or fd required")]
     BareEventMonitor,
     #[error("Error doing event monitor I/O: {0}")]
     EventMonitorIo(std::io::Error),
+    #[error("Event monitor thread failed: {0}")]
+    EventMonitorThread(#[source] vmm::Error),
     #[cfg(feature = "guest_debug")]
     #[error("Error parsing --gdb: {0}")]
     ParsingGdb(option_parser::OptionParserError),
@@ -91,9 +102,9 @@ impl log::Log for Logger {
         let duration = now.duration_since(self.start);
 
         if record.file().is_some() && record.line().is_some() {
-            writeln!(
+            write!(
                 *(*(self.output.lock().unwrap())),
-                "cloud-hypervisor: {:.6?}: <{}> {}:{}:{} -- {}",
+                "cloud-hypervisor: {:.6?}: <{}> {}:{}:{} -- {}\r\n",
                 duration,
                 std::thread::current().name().unwrap_or("anonymous"),
                 record.level(),
@@ -102,9 +113,9 @@ impl log::Log for Logger {
                 record.args()
             )
         } else {
-            writeln!(
+            write!(
                 *(*(self.output.lock().unwrap())),
-                "cloud-hypervisor: {:.6?}: <{}> {}:{} -- {}",
+                "cloud-hypervisor: {:.6?}: <{}> {}:{} -- {}\r\n",
                 duration,
                 std::thread::current().name().unwrap_or("anonymous"),
                 record.level(),
@@ -137,19 +148,19 @@ fn default_rng() -> String {
 /// Launch a cloud-hypervisor VMM.
 pub struct TopLevel {
     #[argh(option, long = "cpus", default = "default_vcpus()")]
-    /// boot=<boot_vcpus>,max=<max_vcpus>,topology=<threads_per_core>:<cores_per_die>:<dies_per_package>:<packages>,kvm_hyperv=on|off,max_phys_bits=<maximum_number_of_physical_bits>,affinity=<list_of_vcpus_with_their_associated_cpuset>,features=<list_of_features_to_enable>
+    /// boot=<boot_vcpus>, max=<max_vcpus>, topology=<threads_per_core>:<cores_per_die>:<dies_per_package>:<packages>, kvm_hyperv=on|off, max_phys_bits=<maximum_number_of_physical_bits>, affinity=<list_of_vcpus_with_their_associated_cpuset>, features=<list_of_features_to_enable>
     cpus: String,
 
     #[argh(option, long = "platform")]
-    /// num_pci_segments=<num_pci_segments>,iommu_segments=<list_of_segments>,serial_number=<dmi_device_serial_number>,uuid=<dmi_device_uuid>,oem_strings=<list_of_strings>
+    /// num_pci_segments=<num_pci_segments>, iommu_segments=<list_of_segments>, serial_number=<dmi_device_serial_number>, uuid=<dmi_device_uuid>, oem_strings=<list_of_strings>
     platform: Option<String>,
 
     #[argh(option, long = "memory", default = "default_memory()")]
-    /// size=<guest_memory_size>,mergeable=on|off,shared=on|off,hugepages=on|off,hugepage_size=<hugepage_size>,hotplug_method=acpi|virtio-mem,hotplug_size=<hotpluggable_memory_size>,hotplugged_size=<hotplugged_memory_size>,prefault=on|off,thp=on|off
+    /// size=<guest_memory_size>, mergeable=on|off, shared=on|off, hugepages=on|off, hugepage_size=<hugepage_size>, hotplug_method=acpi|virtio-mem, hotplug_size=<hotpluggable_memory_size>, hotplugged_size=<hotplugged_memory_size>, prefault=on|off, thp=on|off
     memory: String,
 
     #[argh(option, long = "memory-zone")]
-    /// size=<guest_memory_region_size>,file=<backing_file>,shared=on|off,hugepages=on|off,hugepage_size=<hugepage_size>,host_numa_node=<node_id>,id=<zone_identifier>,hotplug_size=<hotpluggable_memory_size>,hotplugged_size=<hotplugged_memory_size>,prefault=on|off
+    /// size=<guest_memory_region_size>, file=<backing_file>, shared=on|off, hugepages=on|off, hugepage_size=<hugepage_size>, host_numa_node=<node_id>, id=<zone_identifier>, hotplug_size=<hotpluggable_memory_size>, hotplugged_size=<hotplugged_memory_size>, prefault=on|off
     memory_zone: Vec<String>,
 
     #[argh(option, long = "firmware")]
@@ -169,27 +180,27 @@ pub struct TopLevel {
     cmdline: Option<String>,
 
     #[argh(option, long = "disk")]
-    /// path=<disk_image_path>,readonly=on|off,direct=on|off,iommu=on|off,num_queues=<number_of_queues>,queue_size=<size_of_each_queue>,vhost_user=on|off,socket=<vhost_user_socket_path>,bw_size=<bytes>,bw_one_time_burst=<bytes>,bw_refill_time=<ms>,ops_size=<io_ops>,ops_one_time_burst=<io_ops>,ops_refill_time=<ms>,id=<device_id>,pci_segment=<segment_id>
+    /// path=<disk_image_path>, readonly=on|off, direct=on|off, iommu=on|off, num_queues=<number_of_queues>, queue_size=<size_of_each_queue>, vhost_user=on|off, socket=<vhost_user_socket_path>, bw_size=<bytes>, bw_one_time_burst=<bytes>, bw_refill_time=<ms>, ops_size=<io_ops>, ops_one_time_burst=<io_ops>, ops_refill_time=<ms>, id=<device_id>, pci_segment=<segment_id>
     disk: Vec<String>,
 
     #[argh(option, long = "net")]
-    /// tap=<if_name>,ip=<ip_addr>,mask=<net_mask>,mac=<mac_addr>,fd=<fd1,fd2...>,iommu=on|off,num_queues=<number_of_queues>,queue_size=<size_of_each_queue>,id=<device_id>,vhost_user=<vhost_user_enable>,socket=<vhost_user_socket_path>,vhost_mode=client|server,bw_size=<bytes>,bw_one_time_burst=<bytes>,bw_refill_time=<ms>,ops_size=<io_ops>,ops_one_time_burst=<io_ops>,ops_refill_time=<ms>,pci_segment=<segment_id>offload_tso=on|off,offload_ufo=on|off,offload_csum=on|off
+    /// tap=<if_name>, ip=<ip_addr>, mask=<net_mask>, mac=<mac_addr>, fd=<fd1,fd2...>, iommu=on|off, num_queues=<number_of_queues>, queue_size=<size_of_each_queue>, id=<device_id>, vhost_user=<vhost_user_enable>, socket=<vhost_user_socket_path>, vhost_mode=client|server, bw_size=<bytes>, bw_one_time_burst=<bytes>, bw_refill_time=<ms>, ops_size=<io_ops>, ops_one_time_burst=<io_ops>, ops_refill_time=<ms>, pci_segment=<segment_id>, offload_tso=on|off, offload_ufo=on|off, offload_csum=on|off
     net: Vec<String>,
 
     #[argh(option, long = "rng", default = "default_rng()")]
-    /// src=<entropy_source_path>,iommu=on|off
+    /// src=<entropy_source_path>, iommu=on|off
     rng: String,
 
     #[argh(option, long = "balloon")]
-    /// size=<balloon_size>,deflate_on_oom=on|off,free_page_reporting=on|off
+    /// size=<balloon_size>, deflate_on_oom=on|off, free_page_reporting=on|off
     balloon: Option<String>,
 
     #[argh(option, long = "fs")]
-    /// tag=<tag_name>,socket=<socket_path>,num_queues=<number_of_queues>,queue_size=<size_of_each_queue>,id=<device_id>,pci_segment=<segment_id>
+    /// tag=<tag_name>, socket=<socket_path>, num_queues=<number_of_queues>, queue_size=<size_of_each_queue>, id=<device_id>, pci_segment=<segment_id>
     fs: Vec<String>,
 
     #[argh(option, long = "pmem")]
-    /// file=<backing_file_path>,size=<persistent_memory_size>,iommu=on|off,discard_writes=on|off,id=<device_id>,pci_segment=<segment_id>
+    /// file=<backing_file_path>, size=<persistent_memory_size>, iommu=on|off, discard_writes=on|off, id=<device_id>, pci_segment=<segment_id>
     pmem: Vec<String>,
 
     #[argh(option, long = "serial", default = "String::from(\"null\")")]
@@ -197,27 +208,31 @@ pub struct TopLevel {
     serial: String,
 
     #[argh(option, long = "console", default = "String::from(\"tty\")")]
-    /// off|null|pty|tty|file=/path/to/a/file,iommu=on|off
+    /// off|null|pty|tty|file=/path/to/a/file, iommu=on|off
     console: String,
 
     #[argh(option, long = "device")]
-    /// path=<device_path>,iommu=on|off,id=<device_id>,pci_segment=<segment_id>
+    /// path=<device_path>, iommu=on|off, id=<device_id>, pci_segment=<segment_id>
     device: Vec<String>,
 
     #[argh(option, long = "user-device")]
-    /// socket=<socket_path>,id=<device_id>,pci_segment=<segment_id>
+    /// socket=<socket_path>, id=<device_id>, pci_segment=<segment_id>
     user_device: Vec<String>,
 
     #[argh(option, long = "vdpa")]
-    /// path=<device_path>,num_queues=<number_of_queues>,iommu=on|off,id=<device_id>,pci_segment=<segment_id>
+    /// path=<device_path>, num_queues=<number_of_queues>, iommu=on|off, id=<device_id>, pci_segment=<segment_id>
     vdpa: Vec<String>,
 
     #[argh(option, long = "vsock")]
-    /// cid=<context_id>,socket=<socket_path>,iommu=on|off,id=<device_id>,pci_segment=<segment_id>
+    /// cid=<context_id>, socket=<socket_path>, iommu=on|off, id=<device_id>, pci_segment=<segment_id>
     vsock: Option<String>,
 
+    #[argh(switch, long = "pvpanic")]
+    /// enable pvpanic device
+    pvpanic: bool,
+
     #[argh(option, long = "numa")]
-    /// guest_numa_id=<node_id>,cpus=<cpus_id>,distances=<list_of_distances_to_destination_nodes>,memory_zones=<list_of_memory_zones>,sgx_epc_sections=<list_of_sgx_epc_sections>
+    /// guest_numa_id=<node_id>, cpus=<cpus_id>, distances=<list_of_distances_to_destination_nodes>, memory_zones=<list_of_memory_zones>, sgx_epc_sections=<list_of_sgx_epc_sections>
     numa: Vec<String>,
 
     #[argh(switch, long = "watchdog")]
@@ -236,12 +251,27 @@ pub struct TopLevel {
     /// path=<path/to/a/file>|fd=<fd>
     api_socket: Option<String>,
 
+    #[cfg(feature = "dbus_api")]
+    #[argh(option, long = "dbus-service-name")]
+    /// well known name of the service
+    dbus_name: Option<String>,
+
+    #[cfg(feature = "dbus_api")]
+    #[argh(option, long = "dbus-object-path")]
+    /// object path to serve the dbus interface
+    dbus_path: Option<String>,
+
+    #[cfg(feature = "dbus_api")]
+    #[argh(switch, long = "dbus-system-bus")]
+    /// use the system bus instead of a session bus
+    dbus_system_bus: bool,
+
     #[argh(option, long = "event-monitor")]
     /// path=<path/to/a/file>|fd=<fd>
     event_monitor: Option<String>,
 
     #[argh(option, long = "restore")]
-    /// source_url=<source_url>,prefault=on|off
+    /// source_url=<source_url>, prefault=on|off
     restore: Option<String>,
 
     #[argh(option, long = "seccomp", default = "String::from(\"true\")")]
@@ -254,7 +284,7 @@ pub struct TopLevel {
 
     #[cfg(target_arch = "x86_64")]
     #[argh(option, long = "sgx-epc")]
-    /// id=<epc_section_identifier>,size=<epc_section_size>,prefault=on|off
+    /// id=<epc_section_identifier>, size=<epc_section_size>, prefault=on|off
     sgx_epc: Vec<String>,
 
     #[cfg(feature = "guest_debug")]
@@ -325,6 +355,9 @@ impl TopLevel {
         };
 
         let vsock = self.vsock.as_deref();
+
+        let pvpanic = self.pvpanic;
+
         #[cfg(target_arch = "x86_64")]
         let sgx_epc = if !self.sgx_epc.is_empty() {
             Some(self.sgx_epc.iter().map(|x| x.as_str()).collect())
@@ -363,6 +396,7 @@ impl TopLevel {
             user_devices,
             vdpa,
             vsock,
+            pvpanic,
             #[cfg(target_arch = "x86_64")]
             sgx_epc,
             numa,
@@ -415,36 +449,22 @@ fn start_vmm(toplevel: TopLevel) -> Result<Option<String>, Error> {
         (None, None)
     };
 
-    if let Some(ref monitor_config) = toplevel.event_monitor {
-        let mut parser = OptionParser::new();
-        parser.add("path").add("fd");
-        parser
-            .parse(monitor_config)
-            .map_err(Error::ParsingEventMonitor)?;
-
-        let file = if parser.is_set("fd") {
-            let fd = parser
-                .convert("fd")
-                .map_err(Error::ParsingEventMonitor)?
-                .unwrap();
-            // SAFETY: fd is valid
-            unsafe { File::from_raw_fd(fd) }
-        } else if parser.is_set("path") {
-            std::fs::OpenOptions::new()
-                .write(true)
-                .create(true)
-                .open(parser.get("path").unwrap())
-                .map_err(Error::EventMonitorIo)?
-        } else {
-            return Err(Error::BareEventMonitor);
-        };
-        event_monitor::set_monitor(file).map_err(Error::EventMonitorIo)?;
-    }
+    #[cfg(feature = "dbus_api")]
+    let dbus_options = match (&toplevel.dbus_name, &toplevel.dbus_path) {
+        (Some(ref name), Some(ref path)) => Ok(Some(DBusApiOptions {
+            service_name: name.to_owned(),
+            object_path: path.to_owned(),
+            system_bus: toplevel.dbus_system_bus,
+        })),
+        (Some(_), None) => Err(Error::MissingDBusObjectPath),
+        (None, Some(_)) => Err(Error::MissingDBusServiceName),
+        (None, None) => Ok(None),
+    }?;
 
     let (api_request_sender, api_request_receiver) = channel();
     let api_evt = EventFd::new(EFD_NONBLOCK).map_err(Error::CreateApiEventFd)?;
 
-    let http_sender = api_request_sender.clone();
+    let api_request_sender_clone = api_request_sender.clone();
     let seccomp_action = match &toplevel.seccomp as &str {
         "true" => SeccompAction::Trap,
         "false" => SeccompAction::Allow,
@@ -487,8 +507,6 @@ fn start_vmm(toplevel: TopLevel) -> Result<Option<String>, Error> {
         }
     }
 
-    event!("vmm", "starting");
-
     let hypervisor = hypervisor::new().map_err(Error::CreateHypervisor)?;
 
     #[cfg(feature = "guest_debug")]
@@ -510,12 +528,52 @@ fn start_vmm(toplevel: TopLevel) -> Result<Option<String>, Error> {
     #[cfg(feature = "guest_debug")]
     let vm_debug_evt = EventFd::new(EFD_NONBLOCK).map_err(Error::CreateDebugEventFd)?;
 
-    let vmm_thread = vmm::start_vmm_thread(
-        env!("CARGO_PKG_VERSION").to_string(),
+    let exit_evt = EventFd::new(EFD_NONBLOCK).map_err(Error::CreateExitEventFd)?;
+
+    if let Some(ref monitor_config) = toplevel.event_monitor {
+        let mut parser = OptionParser::new();
+        parser.add("path").add("fd");
+        parser
+            .parse(monitor_config)
+            .map_err(Error::ParsingEventMonitor)?;
+
+        let file = if parser.is_set("fd") {
+            let fd = parser
+                .convert("fd")
+                .map_err(Error::ParsingEventMonitor)?
+                .unwrap();
+            // SAFETY: fd is valid
+            unsafe { File::from_raw_fd(fd) }
+        } else if parser.is_set("path") {
+            std::fs::OpenOptions::new()
+                .write(true)
+                .create(true)
+                .open(parser.get("path").unwrap())
+                .map_err(Error::EventMonitorIo)?
+        } else {
+            return Err(Error::BareEventMonitor);
+        };
+
+        let monitor = event_monitor::set_monitor(file).map_err(Error::EventMonitorIo)?;
+        vmm::start_event_monitor_thread(
+            monitor,
+            &seccomp_action,
+            hypervisor.hypervisor_type(),
+            exit_evt.try_clone().unwrap(),
+        )
+        .map_err(Error::EventMonitorThread)?;
+    }
+
+    event!("vmm", "starting");
+
+    let vmm_thread_handle = vmm::start_vmm_thread(
+        vmm::VmmVersionInfo::new(env!("BUILD_VERSION"), env!("CARGO_PKG_VERSION")),
         &api_socket_path,
         api_socket_fd,
+        #[cfg(feature = "dbus_api")]
+        dbus_options,
         api_evt.try_clone().unwrap(),
-        http_sender,
+        api_request_sender_clone,
         api_request_receiver,
         #[cfg(feature = "guest_debug")]
         gdb_socket_path,
@@ -523,41 +581,60 @@ fn start_vmm(toplevel: TopLevel) -> Result<Option<String>, Error> {
         debug_evt.try_clone().unwrap(),
         #[cfg(feature = "guest_debug")]
         vm_debug_evt.try_clone().unwrap(),
+        exit_evt.try_clone().unwrap(),
         &seccomp_action,
         hypervisor,
     )
     .map_err(Error::StartVmmThread)?;
 
-    let payload_present = toplevel.kernel.is_some() || toplevel.firmware.is_some();
+    let r: Result<(), Error> = (|| {
+        let payload_present = toplevel.kernel.is_some() || toplevel.firmware.is_some();
 
-    if payload_present {
-        let vm_params = toplevel.to_vm_params();
-        let vm_config = config::VmConfig::parse(vm_params).map_err(Error::ParsingConfig)?;
+        if payload_present {
+            let vm_params = toplevel.to_vm_params();
+            let vm_config = config::VmConfig::parse(vm_params).map_err(Error::ParsingConfig)?;
 
-        // Create and boot the VM based off the VM config we just built.
-        let sender = api_request_sender.clone();
-        vmm::api::vm_create(
-            api_evt.try_clone().unwrap(),
-            api_request_sender,
-            Arc::new(Mutex::new(vm_config)),
-        )
-        .map_err(Error::VmCreate)?;
-        vmm::api::vm_boot(api_evt.try_clone().unwrap(), sender).map_err(Error::VmBoot)?;
-    } else if let Some(restore_params) = toplevel.restore {
-        vmm::api::vm_restore(
-            api_evt.try_clone().unwrap(),
-            api_request_sender,
-            Arc::new(config::RestoreConfig::parse(&restore_params).map_err(Error::ParsingRestore)?),
-        )
-        .map_err(Error::VmRestore)?;
+            // Create and boot the VM based off the VM config we just built.
+            let sender = api_request_sender.clone();
+            vmm::api::vm_create(
+                api_evt.try_clone().unwrap(),
+                api_request_sender,
+                Arc::new(Mutex::new(vm_config)),
+            )
+            .map_err(Error::VmCreate)?;
+            vmm::api::vm_boot(api_evt.try_clone().unwrap(), sender).map_err(Error::VmBoot)?;
+        } else if let Some(restore_params) = toplevel.restore {
+            vmm::api::vm_restore(
+                api_evt.try_clone().unwrap(),
+                api_request_sender,
+                Arc::new(
+                    config::RestoreConfig::parse(&restore_params).map_err(Error::ParsingRestore)?,
+                ),
+            )
+            .map_err(Error::VmRestore)?;
+        }
+
+        Ok(())
+    })();
+
+    if r.is_err() {
+        if let Err(e) = exit_evt.write(1) {
+            warn!("writing to exit EventFd: {e}");
+        }
     }
 
-    vmm_thread
+    vmm_thread_handle
+        .thread_handle
         .join()
         .map_err(Error::ThreadJoin)?
         .map_err(Error::VmmThread)?;
 
-    Ok(api_socket_path)
+    #[cfg(feature = "dbus_api")]
+    if let Some(chs) = vmm_thread_handle.dbus_shutdown_chs {
+        dbus_api_graceful_shutdown(chs);
+    }
+
+    r.map(|_| api_socket_path)
 }
 
 fn main() {
@@ -571,7 +648,7 @@ fn main() {
     let toplevel: TopLevel = argh::from_env();
 
     if toplevel.version {
-        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILT_VERSION"));
+        println!("{} {}", env!("CARGO_BIN_NAME"), env!("BUILD_VERSION"));
         return;
     }
 
@@ -586,14 +663,6 @@ fn main() {
         }
     };
 
-    // SAFETY: trivially safe
-    let on_tty = unsafe { libc::isatty(libc::STDIN_FILENO) } != 0;
-    if on_tty {
-        // Don't forget to set the terminal in canonical mode
-        // before to exit.
-        std::io::stdin().lock().set_canon_mode().unwrap();
-    }
-
     #[cfg(feature = "dhat-heap")]
     drop(_profiler);
 
@@ -722,6 +791,7 @@ mod unit_tests {
             user_devices: None,
             vdpa: None,
             vsock: None,
+            pvpanic: false,
             iommu: false,
             #[cfg(target_arch = "x86_64")]
             sgx_epc: None,
@@ -731,6 +801,7 @@ mod unit_tests {
             gdb: false,
             platform: None,
             tpm: None,
+            preserved_fds: None,
         };
 
         assert_eq!(expected_vm_config, result_vm_config);
@@ -738,7 +809,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_cpus() {
-        vec![
+        [
             (
                 vec![
                     "cloud-hypervisor",
@@ -856,7 +927,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_kernel() {
-        vec![(
+        [(
             vec!["cloud-hypervisor", "--kernel", "/path/to/kernel"],
             r#"{
                 "payload": {"kernel": "/path/to/kernel"}
@@ -871,7 +942,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_cmdline() {
-        vec![(
+        [(
             vec![
                 "cloud-hypervisor",
                 "--kernel",
@@ -892,7 +963,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_disks() {
-        vec![
+        [
             (
                 vec![
                     "cloud-hypervisor",
@@ -1166,7 +1237,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_rng() {
-        vec![(
+        [(
             vec![
                 "cloud-hypervisor",
                 "--kernel",
@@ -1188,8 +1259,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_fs() {
-        vec![
-            (
+        [(
                 vec![
                     "cloud-hypervisor", "--kernel", "/path/to/kernel",
                     "--memory", "shared=true",
@@ -1259,8 +1329,7 @@ mod unit_tests {
                     ]
                 }"#,
                 true,
-            ),
-        ]
+            )]
         .iter()
         .for_each(|(cli, openapi, equal)| {
             compare_vm_config_cli_vs_json(cli, openapi, *equal);
@@ -1269,7 +1338,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_pmem() {
-        vec![
+        [
             (
                 vec![
                     "cloud-hypervisor",
@@ -1333,7 +1402,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_serial_console() {
-        vec![
+        [
             (
                 vec!["cloud-hypervisor", "--kernel", "/path/to/kernel"],
                 r#"{
@@ -1384,7 +1453,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_serial_pty_console_pty() {
-        vec![
+        [
             (
                 vec!["cloud-hypervisor", "--kernel", "/path/to/kernel"],
                 r#"{
@@ -1532,7 +1601,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_vdpa() {
-        vec![
+        [
             (
                 vec![
                     "cloud-hypervisor",
@@ -1579,7 +1648,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_vsock() {
-        vec![
+        [
             (
                 vec![
                     "cloud-hypervisor",
@@ -1662,7 +1731,7 @@ mod unit_tests {
 
     #[test]
     fn test_valid_vm_config_tpm_socket() {
-        vec![(
+        [(
             vec![
                 "cloud-hypervisor",
                 "--kernel",