banger/internal/daemon/vm.go

package daemon

import (
	"context"
	"errors"
	"fmt"
	"log/slog"
	"net"
	"os"
	"path/filepath"
	"strconv"
	"strings"
	"time"

	"banger/internal/api"
	"banger/internal/firecracker"
	"banger/internal/guest"
	"banger/internal/guestconfig"
	"banger/internal/guestnet"
	"banger/internal/model"
	"banger/internal/namegen"
	"banger/internal/system"
	"banger/internal/vmdns"
	"banger/internal/vsockagent"
)

var (
	errWaitForExitTimeout = errors.New("timed out waiting for VM to exit")
	gracefulShutdownWait  = 10 * time.Second
	vsockReadyWait        = 30 * time.Second
	vsockReadyPoll        = 200 * time.Millisecond
)

const (
	workDiskGitConfigRelativePath       = ".gitconfig"
	workDiskOpencodeAuthDirRelativePath = ".local/share/opencode"
	workDiskOpencodeAuthRelativePath    = workDiskOpencodeAuthDirRelativePath + "/auth.json"
	hostGlobalGitIdentitySource         = "git config --global"
	hostOpencodeAuthDefaultDisplayPath  = "~/" + workDiskOpencodeAuthRelativePath
)

type gitIdentity struct {
	Name  string
	Email string
}

func (d *Daemon) CreateVM(ctx context.Context, params api.VMCreateParams) (vm model.VMRecord, err error) {
	d.mu.Lock()
	defer d.mu.Unlock()
	op := d.beginOperation("vm.create")
	defer func() {
		if err != nil {
			op.fail(err)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	if err := validateOptionalPositiveSetting("vcpu", params.VCPUCount); err != nil {
		return model.VMRecord{}, err
	}
	if err := validateOptionalPositiveSetting("memory", params.MemoryMiB); err != nil {
		return model.VMRecord{}, err
	}

	imageName := params.ImageName
	if imageName == "" {
		imageName = d.config.DefaultImageName
	}
	vmCreateStage(ctx, "resolve_image", "resolving image")
	image, err := d.FindImage(ctx, imageName)
	if err != nil {
		return model.VMRecord{}, err
	}
	vmCreateStage(ctx, "resolve_image", "using image "+image.Name)
	op.stage("image_resolved", imageLogAttrs(image)...)
	name := strings.TrimSpace(params.Name)
	if name == "" {
		name, err = d.generateName(ctx)
		if err != nil {
			return model.VMRecord{}, err
		}
	}
	if _, err := d.FindVM(ctx, name); err == nil {
		return model.VMRecord{}, fmt.Errorf("vm name already exists: %s", name)
	}
	id, err := model.NewID()
	if err != nil {
		return model.VMRecord{}, err
	}
	unlockVM := d.lockVMID(id)
	defer unlockVM()
	guestIP, err := d.store.NextGuestIP(ctx, bridgePrefix(d.config.BridgeIP))
	if err != nil {
		return model.VMRecord{}, err
	}
	vmDir := filepath.Join(d.layout.VMsDir, id)
	if err := os.MkdirAll(vmDir, 0o755); err != nil {
		return model.VMRecord{}, err
	}
	vsockCID, err := defaultVSockCID(guestIP)
	if err != nil {
		return model.VMRecord{}, err
	}
	systemOverlaySize := int64(model.DefaultSystemOverlaySize)
	if params.SystemOverlaySize != "" {
		systemOverlaySize, err = model.ParseSize(params.SystemOverlaySize)
		if err != nil {
			return model.VMRecord{}, err
		}
	}
	workDiskSize := int64(model.DefaultWorkDiskSize)
	if params.WorkDiskSize != "" {
		workDiskSize, err = model.ParseSize(params.WorkDiskSize)
		if err != nil {
			return model.VMRecord{}, err
		}
	}
	now := model.Now()
	spec := model.VMSpec{
		VCPUCount:             optionalIntOrDefault(params.VCPUCount, model.DefaultVCPUCount),
		MemoryMiB:             optionalIntOrDefault(params.MemoryMiB, model.DefaultMemoryMiB),
		SystemOverlaySizeByte: systemOverlaySize,
		WorkDiskSizeBytes:     workDiskSize,
		NATEnabled:            params.NATEnabled,
	}
	vm = model.VMRecord{
		ID:            id,
		Name:          name,
		ImageID:       image.ID,
		State:         model.VMStateCreated,
		CreatedAt:     now,
		UpdatedAt:     now,
		LastTouchedAt: now,
		Spec:          spec,
		Runtime: model.VMRuntime{
			State:         model.VMStateCreated,
			GuestIP:       guestIP,
			DNSName:       vmdns.RecordName(name),
			VMDir:         vmDir,
			VSockPath:     defaultVSockPath(d.layout.RuntimeDir, id),
			VSockCID:      vsockCID,
			SystemOverlay: filepath.Join(vmDir, "system.cow"),
			WorkDiskPath:  filepath.Join(vmDir, "root.ext4"),
			LogPath:       filepath.Join(vmDir, "firecracker.log"),
			MetricsPath:   filepath.Join(vmDir, "metrics.json"),
		},
	}
	vmCreateBindVM(ctx, vm)
	vmCreateStage(ctx, "reserve_vm", fmt.Sprintf("allocated %s (%s)", vm.Name, vm.Runtime.GuestIP))
	if err := d.store.UpsertVM(ctx, vm); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("persisted", vmLogAttrs(vm)...)
	if params.NoStart {
		vm.State = model.VMStateStopped
		vm.Runtime.State = model.VMStateStopped
		if err := d.store.UpsertVM(ctx, vm); err != nil {
			return model.VMRecord{}, err
		}
		return vm, nil
	}
	return d.startVMLocked(ctx, vm, image)
}

func (d *Daemon) StartVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
	return d.withVMLockByRef(ctx, idOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		image, err := d.store.GetImageByID(ctx, vm.ImageID)
		if err != nil {
			return model.VMRecord{}, err
		}
		if vm.State == model.VMStateRunning && system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
			if d.logger != nil {
				d.logger.Info("vm already running", vmLogAttrs(vm)...)
			}
			return vm, nil
		}
		return d.startVMLocked(ctx, vm, image)
	})
}

func (d *Daemon) startVMLocked(ctx context.Context, vm model.VMRecord, image model.Image) (_ model.VMRecord, err error) {
	op := d.beginOperation("vm.start", append(vmLogAttrs(vm), imageLogAttrs(image)...)...)
	defer func() {
		if err != nil {
			err = annotateLogPath(err, vm.Runtime.LogPath)
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	op.stage("preflight")
	vmCreateStage(ctx, "preflight", "checking host prerequisites")
	if err := d.validateStartPrereqs(ctx, vm, image); err != nil {
		return model.VMRecord{}, err
	}
	if err := os.MkdirAll(vm.Runtime.VMDir, 0o755); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("cleanup_runtime")
	if err := d.cleanupRuntime(ctx, vm, true); err != nil {
		return model.VMRecord{}, err
	}
	clearRuntimeHandles(&vm)
	op.stage("bridge")
	if err := d.ensureBridge(ctx); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("socket_dir")
	if err := d.ensureSocketDir(); err != nil {
		return model.VMRecord{}, err
	}

	shortID := system.ShortID(vm.ID)
	apiSock := filepath.Join(d.layout.RuntimeDir, "fc-"+shortID+".sock")
	dmName := "fc-rootfs-" + shortID
	tapName := "tap-fc-" + shortID
	if strings.TrimSpace(vm.Runtime.VSockPath) == "" {
		vm.Runtime.VSockPath = defaultVSockPath(d.layout.RuntimeDir, vm.ID)
	}
	if vm.Runtime.VSockCID == 0 {
		vm.Runtime.VSockCID, err = defaultVSockCID(vm.Runtime.GuestIP)
		if err != nil {
			return model.VMRecord{}, err
		}
	}
	if err := os.RemoveAll(apiSock); err != nil && !os.IsNotExist(err) {
		return model.VMRecord{}, err
	}
	if err := os.RemoveAll(vm.Runtime.VSockPath); err != nil && !os.IsNotExist(err) {
		return model.VMRecord{}, err
	}

	op.stage("system_overlay", "overlay_path", vm.Runtime.SystemOverlay)
	vmCreateStage(ctx, "prepare_rootfs", "preparing system overlay")
	if err := d.ensureSystemOverlay(ctx, &vm); err != nil {
		return model.VMRecord{}, err
	}

	op.stage("dm_snapshot", "dm_name", dmName)
	vmCreateStage(ctx, "prepare_rootfs", "creating root filesystem snapshot")
	handles, err := d.createDMSnapshot(ctx, image.RootfsPath, vm.Runtime.SystemOverlay, dmName)
	if err != nil {
		return model.VMRecord{}, err
	}
	vm.Runtime.BaseLoop = handles.BaseLoop
	vm.Runtime.COWLoop = handles.COWLoop
	vm.Runtime.DMName = handles.DMName
	vm.Runtime.DMDev = handles.DMDev
	vm.Runtime.APISockPath = apiSock
	vm.Runtime.State = model.VMStateRunning
	vm.State = model.VMStateRunning
	vm.Runtime.LastError = ""

	cleanupOnErr := func(err error) (model.VMRecord, error) {
		vm.State = model.VMStateError
		vm.Runtime.State = model.VMStateError
		vm.Runtime.LastError = err.Error()
		op.stage("cleanup_after_failure", "error", err.Error())
		if cleanupErr := d.cleanupRuntime(context.Background(), vm, true); cleanupErr != nil {
			err = errors.Join(err, cleanupErr)
		}
		clearRuntimeHandles(&vm)
		_ = d.store.UpsertVM(context.Background(), vm)
		return model.VMRecord{}, err
	}

	op.stage("patch_root_overlay")
	vmCreateStage(ctx, "prepare_rootfs", "writing guest configuration")
	if err := d.patchRootOverlay(ctx, vm, image); err != nil {
		return cleanupOnErr(err)
	}
	op.stage("prepare_host_features")
	vmCreateStage(ctx, "prepare_host_features", "preparing host-side vm features")
	if err := d.prepareCapabilityHosts(ctx, &vm, image); err != nil {
		return cleanupOnErr(err)
	}
	op.stage("tap")
	tap, err := d.acquireTap(ctx, tapName)
	if err != nil {
		return cleanupOnErr(err)
	}
	vm.Runtime.TapDevice = tap
	op.stage("metrics_file", "metrics_path", vm.Runtime.MetricsPath)
	if err := os.WriteFile(vm.Runtime.MetricsPath, nil, 0o644); err != nil {
		return cleanupOnErr(err)
	}

	op.stage("firecracker_binary")
	fcPath, err := d.firecrackerBinary()
	if err != nil {
		return cleanupOnErr(err)
	}
	op.stage("firecracker_launch", "log_path", vm.Runtime.LogPath, "metrics_path", vm.Runtime.MetricsPath)
	vmCreateStage(ctx, "boot_firecracker", "starting firecracker")
	firecrackerCtx := context.Background()
	machineConfig := firecracker.MachineConfig{
		BinaryPath:      fcPath,
		VMID:            vm.ID,
		SocketPath:      apiSock,
		LogPath:         vm.Runtime.LogPath,
		MetricsPath:     vm.Runtime.MetricsPath,
		KernelImagePath: image.KernelPath,
		InitrdPath:      image.InitrdPath,
		KernelArgs:      system.BuildBootArgs(vm.Name),
		Drives: []firecracker.DriveConfig{{
			ID:       "rootfs",
			Path:     vm.Runtime.DMDev,
			ReadOnly: false,
			IsRoot:   true,
		}},
		TapDevice: tap,
		VSockPath: vm.Runtime.VSockPath,
		VSockCID:  vm.Runtime.VSockCID,
		VCPUCount: vm.Spec.VCPUCount,
		MemoryMiB: vm.Spec.MemoryMiB,
		Logger:    d.logger,
	}
	d.contributeMachineConfig(&machineConfig, vm, image)
	machine, err := firecracker.NewMachine(firecrackerCtx, machineConfig)
	if err != nil {
		return cleanupOnErr(err)
	}
	if err := machine.Start(firecrackerCtx); err != nil {
		vm.Runtime.PID = d.resolveFirecrackerPID(firecrackerCtx, machine, apiSock)
		return cleanupOnErr(err)
	}
	vm.Runtime.PID = d.resolveFirecrackerPID(firecrackerCtx, machine, apiSock)
	op.debugStage("firecracker_started", "pid", vm.Runtime.PID)
	op.stage("socket_access", "api_socket", apiSock)
	if err := d.ensureSocketAccess(ctx, apiSock, "firecracker api socket"); err != nil {
		return cleanupOnErr(err)
	}
	op.stage("vsock_access", "vsock_path", vm.Runtime.VSockPath, "vsock_cid", vm.Runtime.VSockCID)
	if err := d.ensureSocketAccess(ctx, vm.Runtime.VSockPath, "firecracker vsock socket"); err != nil {
		return cleanupOnErr(err)
	}
	vmCreateStage(ctx, "wait_vsock_agent", "waiting for guest vsock agent")
	if err := waitForGuestVSockAgent(ctx, d.logger, vm.Runtime.VSockPath, vsockReadyWait); err != nil {
		return cleanupOnErr(err)
	}
	op.stage("post_start_features")
	vmCreateStage(ctx, "wait_guest_ready", "waiting for guest services")
	if err := d.postStartCapabilities(ctx, vm, image); err != nil {
		return cleanupOnErr(err)
	}
	system.TouchNow(&vm)
	op.stage("persist")
	vmCreateStage(ctx, "finalize", "saving vm state")
	if err := d.store.UpsertVM(ctx, vm); err != nil {
		return cleanupOnErr(err)
	}
	return vm, nil
}

func (d *Daemon) StopVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
	return d.withVMLockByRef(ctx, idOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		return d.stopVMLocked(ctx, vm)
	})
}

func (d *Daemon) stopVMLocked(ctx context.Context, current model.VMRecord) (vm model.VMRecord, err error) {
	vm = current
	op := d.beginOperation("vm.stop", "vm_ref", vm.ID)
	defer func() {
		if err != nil {
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
		op.stage("cleanup_stale_runtime")
		if err := d.cleanupRuntime(ctx, vm, true); err != nil {
			return model.VMRecord{}, err
		}
		vm.State = model.VMStateStopped
		vm.Runtime.State = model.VMStateStopped
		clearRuntimeHandles(&vm)
		if err := d.store.UpsertVM(ctx, vm); err != nil {
			return model.VMRecord{}, err
		}
		return vm, nil
	}
	op.stage("graceful_shutdown")
	if err := d.sendCtrlAltDel(ctx, vm); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("wait_for_exit", "pid", vm.Runtime.PID)
	if err := d.waitForExit(ctx, vm.Runtime.PID, vm.Runtime.APISockPath, gracefulShutdownWait); err != nil {
		if !errors.Is(err, errWaitForExitTimeout) {
			return model.VMRecord{}, err
		}
		op.stage("graceful_shutdown_timeout", "pid", vm.Runtime.PID)
	}
	op.stage("cleanup_runtime")
	if err := d.cleanupRuntime(ctx, vm, true); err != nil {
		return model.VMRecord{}, err
	}
	vm.State = model.VMStateStopped
	vm.Runtime.State = model.VMStateStopped
	clearRuntimeHandles(&vm)
	system.TouchNow(&vm)
	if err := d.store.UpsertVM(ctx, vm); err != nil {
		return model.VMRecord{}, err
	}
	return vm, nil
}

func (d *Daemon) KillVM(ctx context.Context, params api.VMKillParams) (model.VMRecord, error) {
	return d.withVMLockByRef(ctx, params.IDOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		return d.killVMLocked(ctx, vm, params.Signal)
	})
}

func (d *Daemon) killVMLocked(ctx context.Context, current model.VMRecord, signalValue string) (vm model.VMRecord, err error) {
	vm = current
	op := d.beginOperation("vm.kill", "vm_ref", vm.ID, "signal", signalValue)
	defer func() {
		if err != nil {
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
		op.stage("cleanup_stale_runtime")
		if err := d.cleanupRuntime(ctx, vm, true); err != nil {
			return model.VMRecord{}, err
		}
		vm.State = model.VMStateStopped
		vm.Runtime.State = model.VMStateStopped
		clearRuntimeHandles(&vm)
		if err := d.store.UpsertVM(ctx, vm); err != nil {
			return model.VMRecord{}, err
		}
		return vm, nil
	}

	signal := strings.TrimSpace(signalValue)
	if signal == "" {
		signal = "TERM"
	}
	op.stage("send_signal", "pid", vm.Runtime.PID, "signal", signal)
	if _, err := d.runner.RunSudo(ctx, "kill", "-"+signal, strconv.Itoa(vm.Runtime.PID)); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("wait_for_exit", "pid", vm.Runtime.PID)
	if err := d.waitForExit(ctx, vm.Runtime.PID, vm.Runtime.APISockPath, 30*time.Second); err != nil {
		if !errors.Is(err, errWaitForExitTimeout) {
			return model.VMRecord{}, err
		}
		op.stage("signal_timeout", "pid", vm.Runtime.PID, "signal", signal)
	}
	op.stage("cleanup_runtime")
	if err := d.cleanupRuntime(ctx, vm, true); err != nil {
		return model.VMRecord{}, err
	}
	vm.State = model.VMStateStopped
	vm.Runtime.State = model.VMStateStopped
	clearRuntimeHandles(&vm)
	system.TouchNow(&vm)
	if err := d.store.UpsertVM(ctx, vm); err != nil {
		return model.VMRecord{}, err
	}
	return vm, nil
}

func (d *Daemon) RestartVM(ctx context.Context, idOrName string) (vm model.VMRecord, err error) {
	op := d.beginOperation("vm.restart", "vm_ref", idOrName)
	defer func() {
		if err != nil {
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	resolved, err := d.FindVM(ctx, idOrName)
	if err != nil {
		return model.VMRecord{}, err
	}
	return d.withVMLockByID(ctx, resolved.ID, func(vm model.VMRecord) (model.VMRecord, error) {
		op.stage("stop")
		vm, err = d.stopVMLocked(ctx, vm)
		if err != nil {
			return model.VMRecord{}, err
		}
		image, err := d.store.GetImageByID(ctx, vm.ImageID)
		if err != nil {
			return model.VMRecord{}, err
		}
		op.stage("start", vmLogAttrs(vm)...)
		return d.startVMLocked(ctx, vm, image)
	})
}

func (d *Daemon) DeleteVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
	return d.withVMLockByRef(ctx, idOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		return d.deleteVMLocked(ctx, vm)
	})
}

func (d *Daemon) deleteVMLocked(ctx context.Context, current model.VMRecord) (vm model.VMRecord, err error) {
	vm = current
	op := d.beginOperation("vm.delete", "vm_ref", vm.ID)
	defer func() {
		if err != nil {
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	if vm.State == model.VMStateRunning && system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
		op.stage("kill_running_vm", "pid", vm.Runtime.PID)
		_ = d.killVMProcess(ctx, vm.Runtime.PID)
	}
	op.stage("cleanup_runtime")
	if err := d.cleanupRuntime(ctx, vm, false); err != nil {
		return model.VMRecord{}, err
	}
	op.stage("delete_store_record")
	if err := d.store.DeleteVM(ctx, vm.ID); err != nil {
		return model.VMRecord{}, err
	}
	if vm.Runtime.VMDir != "" {
		op.stage("delete_vm_dir", "vm_dir", vm.Runtime.VMDir)
		if err := os.RemoveAll(vm.Runtime.VMDir); err != nil {
			return model.VMRecord{}, err
		}
	}
	return vm, nil
}

func (d *Daemon) SetVM(ctx context.Context, params api.VMSetParams) (model.VMRecord, error) {
	return d.withVMLockByRef(ctx, params.IDOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		return d.setVMLocked(ctx, vm, params)
	})
}

func (d *Daemon) setVMLocked(ctx context.Context, current model.VMRecord, params api.VMSetParams) (vm model.VMRecord, err error) {
	vm = current
	op := d.beginOperation("vm.set", "vm_ref", vm.ID)
	defer func() {
		if err != nil {
			op.fail(err, vmLogAttrs(vm)...)
			return
		}
		op.done(vmLogAttrs(vm)...)
	}()
	running := vm.State == model.VMStateRunning && system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath)
	if params.VCPUCount != nil {
		if err := validateOptionalPositiveSetting("vcpu", params.VCPUCount); err != nil {
			return model.VMRecord{}, err
		}
		if running {
			return model.VMRecord{}, errors.New("vcpu changes require the VM to be stopped")
		}
		op.stage("update_vcpu", "vcpu_count", *params.VCPUCount)
		vm.Spec.VCPUCount = *params.VCPUCount
	}
	if params.MemoryMiB != nil {
		if err := validateOptionalPositiveSetting("memory", params.MemoryMiB); err != nil {
			return model.VMRecord{}, err
		}
		if running {
			return model.VMRecord{}, errors.New("memory changes require the VM to be stopped")
		}
		op.stage("update_memory", "memory_mib", *params.MemoryMiB)
		vm.Spec.MemoryMiB = *params.MemoryMiB
	}
	if params.WorkDiskSize != "" {
		size, err := model.ParseSize(params.WorkDiskSize)
		if err != nil {
			return model.VMRecord{}, err
		}
		if running {
			return model.VMRecord{}, errors.New("disk changes require the VM to be stopped")
		}
		if size < vm.Spec.WorkDiskSizeBytes {
			return model.VMRecord{}, errors.New("disk size can only grow")
		}
		if size > vm.Spec.WorkDiskSizeBytes {
			if exists(vm.Runtime.WorkDiskPath) {
				op.stage("resize_work_disk", "from_bytes", vm.Spec.WorkDiskSizeBytes, "to_bytes", size)
				if err := d.validateWorkDiskResizePrereqs(); err != nil {
					return model.VMRecord{}, err
				}
				if err := system.ResizeExt4Image(ctx, d.runner, vm.Runtime.WorkDiskPath, size); err != nil {
					return model.VMRecord{}, err
				}
			}
			vm.Spec.WorkDiskSizeBytes = size
		}
	}
	if params.NATEnabled != nil {
		op.stage("update_nat", "nat_enabled", *params.NATEnabled)
		vm.Spec.NATEnabled = *params.NATEnabled
	}
	if running {
		if err := d.applyCapabilityConfigChanges(ctx, current, vm); err != nil {
			return model.VMRecord{}, err
		}
	}
	system.TouchNow(&vm)
	if err := d.store.UpsertVM(ctx, vm); err != nil {
		return model.VMRecord{}, err
	}
	return vm, nil
}

func (d *Daemon) GetVMStats(ctx context.Context, idOrName string) (model.VMRecord, model.VMStats, error) {
	vm, err := d.withVMLockByRef(ctx, idOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		return d.getVMStatsLocked(ctx, vm)
	})
	if err != nil {
		return model.VMRecord{}, model.VMStats{}, err
	}
	return vm, vm.Stats, nil
}

func (d *Daemon) HealthVM(ctx context.Context, idOrName string) (result api.VMHealthResult, err error) {
	_, err = d.withVMLockByRef(ctx, idOrName, func(vm model.VMRecord) (model.VMRecord, error) {
		result.Name = vm.Name
		if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
			result.Healthy = false
			return vm, nil
		}
		if strings.TrimSpace(vm.Runtime.VSockPath) == "" {
			return model.VMRecord{}, errors.New("vm has no vsock path")
		}
		if vm.Runtime.VSockCID == 0 {
			return model.VMRecord{}, errors.New("vm has no vsock cid")
		}
		if err := d.ensureSocketAccess(ctx, vm.Runtime.VSockPath, "firecracker vsock socket"); err != nil {
			return model.VMRecord{}, err
		}
		pingCtx, cancel := context.WithTimeout(ctx, 3*time.Second)
		defer cancel()
		if err := vsockagent.Health(pingCtx, d.logger, vm.Runtime.VSockPath); err != nil {
			return model.VMRecord{}, err
		}
		result.Healthy = true
		return vm, nil
	})
	return result, err
}

func (d *Daemon) PingVM(ctx context.Context, idOrName string) (result api.VMPingResult, err error) {
	health, err := d.HealthVM(ctx, idOrName)
	if err != nil {
		return api.VMPingResult{}, err
	}
	return api.VMPingResult{Name: health.Name, Alive: health.Healthy}, nil
}

func (d *Daemon) getVMStatsLocked(ctx context.Context, vm model.VMRecord) (model.VMRecord, error) {
	stats, err := d.collectStats(ctx, vm)
	if err == nil {
		vm.Stats = stats
		vm.UpdatedAt = model.Now()
		_ = d.store.UpsertVM(ctx, vm)
		if d.logger != nil {
			d.logger.Debug("vm stats collected", append(vmLogAttrs(vm), "rss_bytes", stats.RSSBytes, "vsz_bytes", stats.VSZBytes, "cpu_percent", stats.CPUPercent)...)
		}
	}
	return vm, nil
}

func (d *Daemon) pollStats(ctx context.Context) error {
	vms, err := d.store.ListVMs(ctx)
	if err != nil {
		return err
	}
	for _, vm := range vms {
		if err := d.withVMLockByIDErr(ctx, vm.ID, func(vm model.VMRecord) error {
			if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
				return nil
			}
			stats, err := d.collectStats(ctx, vm)
			if err != nil {
				if d.logger != nil {
					d.logger.Debug("vm stats collection failed", append(vmLogAttrs(vm), "error", err.Error())...)
				}
				return nil
			}
			vm.Stats = stats
			vm.UpdatedAt = model.Now()
			return d.store.UpsertVM(ctx, vm)
		}); err != nil {
			return err
		}
	}
	return nil
}

func (d *Daemon) stopStaleVMs(ctx context.Context) (err error) {
	if d.config.AutoStopStaleAfter <= 0 {
		return nil
	}
	op := d.beginOperation("vm.stop_stale")
	defer func() {
		if err != nil {
			op.fail(err)
			return
		}
		op.done()
	}()
	vms, err := d.store.ListVMs(ctx)
	if err != nil {
		return err
	}
	now := model.Now()
	for _, vm := range vms {
		if err := d.withVMLockByIDErr(ctx, vm.ID, func(vm model.VMRecord) error {
			if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
				return nil
			}
			if now.Sub(vm.LastTouchedAt) < d.config.AutoStopStaleAfter {
				return nil
			}
			op.stage("stopping_vm", vmLogAttrs(vm)...)
			_ = d.sendCtrlAltDel(ctx, vm)
			_ = d.waitForExit(ctx, vm.Runtime.PID, vm.Runtime.APISockPath, 10*time.Second)
			_ = d.cleanupRuntime(ctx, vm, true)
			vm.State = model.VMStateStopped
			vm.Runtime.State = model.VMStateStopped
			clearRuntimeHandles(&vm)
			vm.UpdatedAt = model.Now()
			return d.store.UpsertVM(ctx, vm)
		}); err != nil {
			return err
		}
	}
	return nil
}

func (d *Daemon) collectStats(ctx context.Context, vm model.VMRecord) (model.VMStats, error) {
	stats := model.VMStats{
		CollectedAt:        model.Now(),
		SystemOverlayBytes: system.AllocatedBytes(vm.Runtime.SystemOverlay),
		WorkDiskBytes:      system.AllocatedBytes(vm.Runtime.WorkDiskPath),
		MetricsRaw:         system.ParseMetricsFile(vm.Runtime.MetricsPath),
	}
	if vm.Runtime.PID > 0 && system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
		ps, err := system.ReadProcessStats(ctx, vm.Runtime.PID)
		if err == nil {
			stats.CPUPercent = ps.CPUPercent
			stats.RSSBytes = ps.RSSBytes
			stats.VSZBytes = ps.VSZBytes
		}
	}
	return stats, nil
}

func (d *Daemon) ensureSystemOverlay(ctx context.Context, vm *model.VMRecord) error {
	if exists(vm.Runtime.SystemOverlay) {
		return nil
	}
	_, err := d.runner.Run(ctx, "truncate", "-s", strconv.FormatInt(vm.Spec.SystemOverlaySizeByte, 10), vm.Runtime.SystemOverlay)
	return err
}

func (d *Daemon) patchRootOverlay(ctx context.Context, vm model.VMRecord, image model.Image) error {
	resolv := []byte(fmt.Sprintf("nameserver %s\n", d.config.DefaultDNS))
	hostname := []byte(vm.Name + "\n")
	hosts := []byte(fmt.Sprintf("127.0.0.1 localhost\n127.0.1.1 %s\n", vm.Name))
	sshdConfig := []byte(strings.Join([]string{
		"LogLevel DEBUG3",
		"PermitRootLogin yes",
		"PubkeyAuthentication yes",
		"AuthorizedKeysFile /root/.ssh/authorized_keys",
		"StrictModes no",
		"",
	}, "\n"))
	fstab, err := system.ReadDebugFSText(ctx, d.runner, vm.Runtime.DMDev, "/etc/fstab")
	if err != nil {
		fstab = ""
	}
	builder := guestconfig.NewBuilder()
	builder.WriteFile("/etc/resolv.conf", resolv)
	builder.WriteFile("/etc/hostname", hostname)
	builder.WriteFile("/etc/hosts", hosts)
	builder.WriteFile(guestnet.ConfigPath, guestnet.ConfigFile(vm.Runtime.GuestIP, d.config.BridgeIP, d.config.DefaultDNS))
	builder.WriteFile(guestnet.GuestScriptPath, []byte(guestnet.BootstrapScript()))
	builder.WriteFile("/etc/ssh/sshd_config.d/99-banger.conf", sshdConfig)
	builder.DropMountTarget("/home")
	builder.DropMountTarget("/var")
	builder.AddMount(guestconfig.MountSpec{
		Source:  "tmpfs",
		Target:  "/run",
		FSType:  "tmpfs",
		Options: []string{"defaults", "nodev", "nosuid", "mode=0755"},
		Dump:    0,
		Pass:    0,
	})
	builder.AddMount(guestconfig.MountSpec{
		Source:  "tmpfs",
		Target:  "/tmp",
		FSType:  "tmpfs",
		Options: []string{"defaults", "nodev", "nosuid", "mode=1777"},
		Dump:    0,
		Pass:    0,
	})
	d.contributeGuestConfig(builder, vm, image)
	builder.WriteFile("/etc/fstab", []byte(builder.RenderFSTab(fstab)))
	files := builder.Files()
	for _, guestPath := range builder.FilePaths() {
		data := files[guestPath]
		if guestPath == guestnet.GuestScriptPath {
			if err := system.WriteExt4FileMode(ctx, d.runner, vm.Runtime.DMDev, guestPath, 0o755, data); err != nil {
				return err
			}
			continue
		}
		if err := system.WriteExt4File(ctx, d.runner, vm.Runtime.DMDev, guestPath, data); err != nil {
			return err
		}
	}
	return nil
}

type workDiskPreparation struct {
	ClonedFromSeed bool
}

func (d *Daemon) ensureWorkDisk(ctx context.Context, vm *model.VMRecord, image model.Image) (workDiskPreparation, error) {
	if exists(vm.Runtime.WorkDiskPath) {
		return workDiskPreparation{}, nil
	}
	if exists(image.WorkSeedPath) {
		vmCreateStage(ctx, "prepare_work_disk", "cloning work seed")
		if err := system.CopyFilePreferClone(image.WorkSeedPath, vm.Runtime.WorkDiskPath); err != nil {
			return workDiskPreparation{}, err
		}
		seedInfo, err := os.Stat(image.WorkSeedPath)
		if err != nil {
			return workDiskPreparation{}, err
		}
		if vm.Spec.WorkDiskSizeBytes < seedInfo.Size() {
			return workDiskPreparation{}, fmt.Errorf("requested work disk size %d is smaller than seed image %d", vm.Spec.WorkDiskSizeBytes, seedInfo.Size())
		}
		if vm.Spec.WorkDiskSizeBytes > seedInfo.Size() {
			vmCreateStage(ctx, "prepare_work_disk", "resizing work disk")
			if err := system.ResizeExt4Image(ctx, d.runner, vm.Runtime.WorkDiskPath, vm.Spec.WorkDiskSizeBytes); err != nil {
				return workDiskPreparation{}, err
			}
		}
		return workDiskPreparation{ClonedFromSeed: true}, nil
	}
	vmCreateStage(ctx, "prepare_work_disk", "creating empty work disk")
	if _, err := d.runner.Run(ctx, "truncate", "-s", strconv.FormatInt(vm.Spec.WorkDiskSizeBytes, 10), vm.Runtime.WorkDiskPath); err != nil {
		return workDiskPreparation{}, err
	}
	if _, err := d.runner.Run(ctx, "mkfs.ext4", "-F", vm.Runtime.WorkDiskPath); err != nil {
		return workDiskPreparation{}, err
	}
	rootMount, cleanupRoot, err := system.MountTempDir(ctx, d.runner, vm.Runtime.DMDev, true)
	if err != nil {
		return workDiskPreparation{}, err
	}
	defer cleanupRoot()
	workMount, cleanupWork, err := system.MountTempDir(ctx, d.runner, vm.Runtime.WorkDiskPath, false)
	if err != nil {
		return workDiskPreparation{}, err
	}
	defer cleanupWork()
	vmCreateStage(ctx, "prepare_work_disk", "copying /root into work disk")
	if err := system.CopyDirContents(ctx, d.runner, filepath.Join(rootMount, "root"), workMount, true); err != nil {
		return workDiskPreparation{}, err
	}
	if err := d.flattenNestedWorkHome(ctx, workMount); err != nil {
		return workDiskPreparation{}, err
	}
	return workDiskPreparation{}, nil
}

func (d *Daemon) ensureAuthorizedKeyOnWorkDisk(ctx context.Context, vm *model.VMRecord, image model.Image, prep workDiskPreparation) error {
	fingerprint, err := guest.AuthorizedPublicKeyFingerprint(d.config.SSHKeyPath)
	if err != nil {
		return fmt.Errorf("derive authorized ssh key fingerprint: %w", err)
	}
	if prep.ClonedFromSeed && image.SeededSSHPublicKeyFingerprint != "" && image.SeededSSHPublicKeyFingerprint == fingerprint {
		vmCreateStage(ctx, "prepare_work_disk", "using seeded SSH access")
		return nil
	}
	publicKey, err := guest.AuthorizedPublicKey(d.config.SSHKeyPath)
	if err != nil {
		return fmt.Errorf("derive authorized ssh key: %w", err)
	}
	vmCreateStage(ctx, "prepare_work_disk", "repairing SSH access on work disk")
	workMount, cleanupWork, err := system.MountTempDir(ctx, d.runner, vm.Runtime.WorkDiskPath, false)
	if err != nil {
		return err
	}
	defer cleanupWork()

	if err := d.flattenNestedWorkHome(ctx, workMount); err != nil {
		return err
	}

	sshDir := filepath.Join(workMount, ".ssh")
	if _, err := d.runner.RunSudo(ctx, "mkdir", "-p", sshDir); err != nil {
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "chmod", "700", sshDir); err != nil {
		return err
	}

	authorizedKeysPath := filepath.Join(sshDir, "authorized_keys")
	existing, err := d.runner.RunSudo(ctx, "cat", authorizedKeysPath)
	if err != nil {
		existing = nil
	}
	merged := mergeAuthorizedKey(existing, publicKey)

	tmpFile, err := os.CreateTemp("", "banger-authorized-keys-*")
	if err != nil {
		return err
	}
	tmpPath := tmpFile.Name()
	if _, err := tmpFile.Write(merged); err != nil {
		_ = tmpFile.Close()
		_ = os.Remove(tmpPath)
		return err
	}
	if err := tmpFile.Close(); err != nil {
		_ = os.Remove(tmpPath)
		return err
	}
	defer os.Remove(tmpPath)

	if _, err := d.runner.RunSudo(ctx, "install", "-m", "600", tmpPath, authorizedKeysPath); err != nil {
		return err
	}
	if prep.ClonedFromSeed && image.Managed {
		vmCreateStage(ctx, "prepare_work_disk", "refreshing managed work seed")
		if err := d.refreshManagedWorkSeedFingerprint(ctx, image, fingerprint); err != nil {
			return err
		}
	}
	return nil
}

func (d *Daemon) ensureGitIdentityOnWorkDisk(ctx context.Context, vm *model.VMRecord) error {
	runner := d.runner
	if runner == nil {
		runner = system.NewRunner()
	}

	identity, err := resolveHostGlobalGitIdentity(ctx, runner)
	if err != nil {
		d.warnGitIdentitySyncSkipped(*vm, hostGlobalGitIdentitySource, err)
		return nil
	}

	vmCreateStage(ctx, "prepare_work_disk", "syncing git identity")
	workMount, cleanupWork, err := system.MountTempDir(ctx, runner, vm.Runtime.WorkDiskPath, false)
	if err != nil {
		return err
	}
	defer cleanupWork()

	if err := d.flattenNestedWorkHome(ctx, workMount); err != nil {
		return err
	}

	return writeGitIdentity(ctx, runner, filepath.Join(workMount, workDiskGitConfigRelativePath), identity)
}

func (d *Daemon) ensureOpencodeAuthOnWorkDisk(ctx context.Context, vm *model.VMRecord) error {
	hostAuthPath, err := resolveHostOpencodeAuthPath()
	if err != nil {
		d.warnOpencodeAuthSyncSkipped(*vm, hostOpencodeAuthDefaultDisplayPath, err)
		return nil
	}
	authData, err := os.ReadFile(hostAuthPath)
	if err != nil {
		d.warnOpencodeAuthSyncSkipped(*vm, hostAuthPath, err)
		return nil
	}

	vmCreateStage(ctx, "prepare_work_disk", "syncing opencode auth")
	workMount, cleanupWork, err := system.MountTempDir(ctx, d.runner, vm.Runtime.WorkDiskPath, false)
	if err != nil {
		return err
	}
	defer cleanupWork()

	if err := d.flattenNestedWorkHome(ctx, workMount); err != nil {
		return err
	}

	authDir := filepath.Join(workMount, workDiskOpencodeAuthDirRelativePath)
	if _, err := d.runner.RunSudo(ctx, "mkdir", "-p", authDir); err != nil {
		return err
	}
	authPath := filepath.Join(workMount, workDiskOpencodeAuthRelativePath)

	tmpFile, err := os.CreateTemp("", "banger-opencode-auth-*")
	if err != nil {
		return err
	}
	tmpPath := tmpFile.Name()
	if _, err := tmpFile.Write(authData); err != nil {
		_ = tmpFile.Close()
		_ = os.Remove(tmpPath)
		return err
	}
	if err := tmpFile.Close(); err != nil {
		_ = os.Remove(tmpPath)
		return err
	}
	defer os.Remove(tmpPath)

	_, err = d.runner.RunSudo(ctx, "install", "-m", "600", tmpPath, authPath)
	return err
}

func resolveHostOpencodeAuthPath() (string, error) {
	home, err := os.UserHomeDir()
	if err != nil {
		return "", err
	}
	return filepath.Join(home, workDiskOpencodeAuthRelativePath), nil
}

func resolveHostGlobalGitIdentity(ctx context.Context, runner system.CommandRunner) (gitIdentity, error) {
	name, err := gitConfigValue(ctx, runner, nil, "user.name")
	if err != nil {
		return gitIdentity{}, err
	}
	if name == "" {
		return gitIdentity{}, errors.New("host git user.name is empty")
	}

	email, err := gitConfigValue(ctx, runner, nil, "user.email")
	if err != nil {
		return gitIdentity{}, err
	}
	if email == "" {
		return gitIdentity{}, errors.New("host git user.email is empty")
	}

	return gitIdentity{Name: name, Email: email}, nil
}

func gitConfigValue(ctx context.Context, runner system.CommandRunner, extraArgs []string, key string) (string, error) {
	args := []string{"config"}
	args = append(args, extraArgs...)
	args = append(args, "--default", "", "--get", key)
	out, err := runner.Run(ctx, "git", args...)
	if err != nil {
		return "", err
	}
	return strings.TrimSpace(string(out)), nil
}

func writeGitIdentity(ctx context.Context, runner system.CommandRunner, gitConfigPath string, identity gitIdentity) error {
	existing, err := runner.RunSudo(ctx, "cat", gitConfigPath)
	if err != nil {
		existing = nil
	}

	tmpFile, err := os.CreateTemp("", "banger-gitconfig-*")
	if err != nil {
		return err
	}
	tmpPath := tmpFile.Name()
	if _, err := tmpFile.Write(existing); err != nil {
		_ = tmpFile.Close()
		_ = os.Remove(tmpPath)
		return err
	}
	if err := tmpFile.Close(); err != nil {
		_ = os.Remove(tmpPath)
		return err
	}
	defer os.Remove(tmpPath)

	if _, err := runner.Run(ctx, "git", "config", "--file", tmpPath, "user.name", identity.Name); err != nil {
		return err
	}
	if _, err := runner.Run(ctx, "git", "config", "--file", tmpPath, "user.email", identity.Email); err != nil {
		return err
	}
	_, err = runner.RunSudo(ctx, "install", "-m", "644", tmpPath, gitConfigPath)
	return err
}

func (d *Daemon) warnOpencodeAuthSyncSkipped(vm model.VMRecord, hostPath string, err error) {
	if d.logger == nil || err == nil {
		return
	}
	d.logger.Warn("guest opencode auth sync skipped", append(vmLogAttrs(vm), "host_path", hostPath, "error", err.Error())...)
}

func (d *Daemon) warnGitIdentitySyncSkipped(vm model.VMRecord, source string, err error) {
	if d.logger == nil || err == nil {
		return
	}
	d.logger.Warn("guest git identity sync skipped", append(vmLogAttrs(vm), "source", source, "error", err.Error())...)
}

func mergeAuthorizedKey(existing, managed []byte) []byte {
	managedLine := strings.TrimSpace(string(managed))
	if managedLine == "" {
		return append([]byte(nil), existing...)
	}

	lines := strings.Split(strings.ReplaceAll(string(existing), "\r\n", "\n"), "\n")
	out := make([]string, 0, len(lines)+1)
	found := false
	for _, line := range lines {
		line = strings.TrimRight(line, "\r")
		trimmed := strings.TrimSpace(line)
		if trimmed == "" {
			continue
		}
		if trimmed == managedLine {
			found = true
		}
		out = append(out, line)
	}
	if !found {
		out = append(out, managedLine)
	}
	return []byte(strings.Join(out, "\n") + "\n")
}

func (d *Daemon) flattenNestedWorkHome(ctx context.Context, workMount string) error {
	nestedHome := filepath.Join(workMount, "root")
	if !exists(nestedHome) {
		return nil
	}
	if _, err := d.runner.RunSudo(ctx, "chmod", "755", nestedHome); err != nil {
		return err
	}
	entries, err := os.ReadDir(nestedHome)
	if err != nil {
		return err
	}
	for _, entry := range entries {
		sourcePath := filepath.Join(nestedHome, entry.Name())
		if _, err := d.runner.RunSudo(ctx, "cp", "-a", sourcePath, workMount+"/"); err != nil {
			return err
		}
	}
	_, err = d.runner.RunSudo(ctx, "rm", "-rf", nestedHome)
	return err
}

func (d *Daemon) ensureBridge(ctx context.Context) error {
	if _, err := d.runner.Run(ctx, "ip", "link", "show", d.config.BridgeName); err == nil {
		_, err = d.runner.RunSudo(ctx, "ip", "link", "set", d.config.BridgeName, "up")
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "ip", "link", "add", "name", d.config.BridgeName, "type", "bridge"); err != nil {
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "ip", "addr", "add", fmt.Sprintf("%s/%s", d.config.BridgeIP, d.config.CIDR), "dev", d.config.BridgeName); err != nil {
		return err
	}
	_, err := d.runner.RunSudo(ctx, "ip", "link", "set", d.config.BridgeName, "up")
	return err
}

func (d *Daemon) ensureSocketDir() error {
	return os.MkdirAll(d.layout.RuntimeDir, 0o755)
}

func (d *Daemon) createTap(ctx context.Context, tap string) error {
	if _, err := d.runner.Run(ctx, "ip", "link", "show", tap); err == nil {
		_, _ = d.runner.RunSudo(ctx, "ip", "link", "del", tap)
	}
	if _, err := d.runner.RunSudo(ctx, "ip", "tuntap", "add", "dev", tap, "mode", "tap", "user", strconv.Itoa(os.Getuid()), "group", strconv.Itoa(os.Getgid())); err != nil {
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "ip", "link", "set", tap, "master", d.config.BridgeName); err != nil {
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "ip", "link", "set", tap, "up"); err != nil {
		return err
	}
	_, err := d.runner.RunSudo(ctx, "ip", "link", "set", d.config.BridgeName, "up")
	return err
}

func (d *Daemon) firecrackerBinary() (string, error) {
	if d.config.FirecrackerBin == "" {
		return "", fmt.Errorf("firecracker binary not configured; install firecracker or set firecracker_bin")
	}
	path := d.config.FirecrackerBin
	if strings.ContainsRune(path, os.PathSeparator) {
		if !exists(path) {
			return "", fmt.Errorf("firecracker binary not found at %s; install firecracker or set firecracker_bin", path)
		}
		return path, nil
	}
	resolved, err := system.LookupExecutable(path)
	if err != nil {
		return "", fmt.Errorf("firecracker binary %q not found in PATH; install firecracker or set firecracker_bin", path)
	}
	return resolved, nil
}

func (d *Daemon) ensureSocketAccess(ctx context.Context, socketPath, label string) error {
	if err := waitForPath(ctx, socketPath, 5*time.Second, label); err != nil {
		return err
	}
	if _, err := d.runner.RunSudo(ctx, "chown", fmt.Sprintf("%d:%d", os.Getuid(), os.Getgid()), socketPath); err != nil {
		return err
	}
	_, err := d.runner.RunSudo(ctx, "chmod", "600", socketPath)
	return err
}

func (d *Daemon) findFirecrackerPID(ctx context.Context, apiSock string) (int, error) {
	out, err := d.runner.Run(ctx, "pgrep", "-n", "-f", apiSock)
	if err != nil {
		return 0, err
	}
	return strconv.Atoi(strings.TrimSpace(string(out)))
}

func (d *Daemon) resolveFirecrackerPID(ctx context.Context, machine *firecracker.Machine, apiSock string) int {
	if pid, err := d.findFirecrackerPID(ctx, apiSock); err == nil && pid > 0 {
		return pid
	}
	if machine != nil {
		if pid, err := machine.PID(); err == nil && pid > 0 {
			return pid
		}
	}
	return 0
}

func (d *Daemon) sendCtrlAltDel(ctx context.Context, vm model.VMRecord) error {
	if err := d.ensureSocketAccess(ctx, vm.Runtime.APISockPath, "firecracker api socket"); err != nil {
		return err
	}
	client := firecracker.New(vm.Runtime.APISockPath, d.logger)
	return client.SendCtrlAltDel(ctx)
}

func (d *Daemon) waitForExit(ctx context.Context, pid int, apiSock string, timeout time.Duration) error {
	deadline := time.Now().Add(timeout)
	for {
		if !system.ProcessRunning(pid, apiSock) {
			return nil
		}
		if time.Now().After(deadline) {
			return errWaitForExitTimeout
		}
		select {
		case <-ctx.Done():
			return ctx.Err()
		case <-time.After(100 * time.Millisecond):
		}
	}
}

func (d *Daemon) cleanupRuntime(ctx context.Context, vm model.VMRecord, preserveDisks bool) error {
	if d.logger != nil {
		d.logger.Debug("cleanup runtime", append(vmLogAttrs(vm), "preserve_disks", preserveDisks)...)
	}
	cleanupPID := vm.Runtime.PID
	if vm.Runtime.APISockPath != "" {
		if pid, err := d.findFirecrackerPID(ctx, vm.Runtime.APISockPath); err == nil && pid > 0 {
			cleanupPID = pid
		}
	}
	if cleanupPID > 0 && system.ProcessRunning(cleanupPID, vm.Runtime.APISockPath) {
		_ = d.killVMProcess(ctx, cleanupPID)
		if err := d.waitForExit(ctx, cleanupPID, vm.Runtime.APISockPath, 30*time.Second); err != nil {
			return err
		}
	}
	snapshotErr := d.cleanupDMSnapshot(ctx, dmSnapshotHandles{
		BaseLoop: vm.Runtime.BaseLoop,
		COWLoop:  vm.Runtime.COWLoop,
		DMName:   vm.Runtime.DMName,
		DMDev:    vm.Runtime.DMDev,
	})
	featureErr := d.cleanupCapabilityState(ctx, vm)
	var tapErr error
	if vm.Runtime.TapDevice != "" {
		tapErr = d.releaseTap(ctx, vm.Runtime.TapDevice)
	}
	if vm.Runtime.APISockPath != "" {
		_ = os.Remove(vm.Runtime.APISockPath)
	}
	if vm.Runtime.VSockPath != "" {
		_ = os.Remove(vm.Runtime.VSockPath)
	}
	if !preserveDisks && vm.Runtime.VMDir != "" {
		return errors.Join(snapshotErr, featureErr, tapErr, os.RemoveAll(vm.Runtime.VMDir))
	}
	return errors.Join(snapshotErr, featureErr, tapErr)
}

func clearRuntimeHandles(vm *model.VMRecord) {
	vm.Runtime.PID = 0
	vm.Runtime.APISockPath = ""
	vm.Runtime.TapDevice = ""
	vm.Runtime.BaseLoop = ""
	vm.Runtime.COWLoop = ""
	vm.Runtime.DMName = ""
	vm.Runtime.DMDev = ""
}

func defaultVSockPath(runtimeDir, vmID string) string {
	return filepath.Join(runtimeDir, "fc-"+system.ShortID(vmID)+".vsock")
}

func defaultVSockCID(guestIP string) (uint32, error) {
	ip := net.ParseIP(strings.TrimSpace(guestIP)).To4()
	if ip == nil {
		return 0, fmt.Errorf("guest IP is not IPv4: %q", guestIP)
	}
	return 10000 + uint32(ip[3]), nil
}

func waitForPath(ctx context.Context, path string, timeout time.Duration, label string) error {
	deadline := time.Now().Add(timeout)
	for {
		if _, err := os.Stat(path); err == nil {
			return nil
		} else if err != nil && !os.IsNotExist(err) {
			return err
		}
		if time.Now().After(deadline) {
			return fmt.Errorf("%s not ready: %s: %w", label, path, context.DeadlineExceeded)
		}
		select {
		case <-ctx.Done():
			return ctx.Err()
		case <-time.After(100 * time.Millisecond):
		}
	}
}

func waitForGuestVSockAgent(ctx context.Context, logger *slog.Logger, socketPath string, timeout time.Duration) error {
	if strings.TrimSpace(socketPath) == "" {
		return errors.New("vsock path is required")
	}

	waitCtx, cancel := context.WithTimeout(ctx, timeout)
	defer cancel()

	ticker := time.NewTicker(vsockReadyPoll)
	defer ticker.Stop()

	var lastErr error
	for {
		pingCtx, pingCancel := context.WithTimeout(waitCtx, 3*time.Second)
		err := vsockagent.Health(pingCtx, logger, socketPath)
		pingCancel()
		if err == nil {
			return nil
		}
		lastErr = err

		select {
		case <-waitCtx.Done():
			if lastErr != nil {
				return fmt.Errorf("guest vsock agent not ready: %w", lastErr)
			}
			return errors.New("guest vsock agent not ready before timeout")
		case <-ticker.C:
		}
	}
}

func (d *Daemon) setDNS(ctx context.Context, vmName, guestIP string) error {
	if d.vmDNS == nil {
		return nil
	}
	if err := d.vmDNS.Set(vmdns.RecordName(vmName), guestIP); err != nil {
		return err
	}
	d.ensureVMDNSResolverRouting(ctx)
	return nil
}

func (d *Daemon) removeDNS(ctx context.Context, dnsName string) error {
	if dnsName == "" {
		return nil
	}
	if d.vmDNS == nil {
		return nil
	}
	return d.vmDNS.Remove(dnsName)
}

func (d *Daemon) rebuildDNS(ctx context.Context) error {
	if d.vmDNS == nil {
		return nil
	}
	vms, err := d.store.ListVMs(ctx)
	if err != nil {
		return err
	}
	records := make(map[string]string)
	for _, vm := range vms {
		if vm.State != model.VMStateRunning {
			continue
		}
		if !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
			continue
		}
		if strings.TrimSpace(vm.Runtime.GuestIP) == "" {
			continue
		}
		records[vmdns.RecordName(vm.Name)] = vm.Runtime.GuestIP
	}
	return d.vmDNS.Replace(records)
}

func (d *Daemon) killVMProcess(ctx context.Context, pid int) error {
	_, err := d.runner.RunSudo(ctx, "kill", "-KILL", strconv.Itoa(pid))
	return err
}

func (d *Daemon) generateName(ctx context.Context) (string, error) {
	_ = ctx
	if name := strings.TrimSpace(namegen.Generate()); name != "" {
		return name, nil
	}
	return "vm-" + strconv.FormatInt(time.Now().Unix(), 10), nil
}

func bridgePrefix(bridgeIP string) string {
	parts := strings.Split(bridgeIP, ".")
	if len(parts) < 3 {
		return bridgeIP
	}
	return strings.Join(parts[:3], ".")
}

func optionalIntOrDefault(value *int, fallback int) int {
	if value != nil {
		return *value
	}
	return fallback
}

func validateOptionalPositiveSetting(label string, value *int) error {
	if value == nil {
		return nil
	}
	if *value <= 0 {
		return fmt.Errorf("%s must be a positive integer", label)
	}
	return nil
}