banger/internal/daemon/daemon.go

package daemon

import (
	"bufio"
	"context"
	"encoding/json"
	"errors"
	"fmt"
	"net"
	"os"
	"path/filepath"
	"strings"
	"sync"
	"time"

	"banger/internal/api"
	"banger/internal/config"
	"banger/internal/model"
	"banger/internal/paths"
	"banger/internal/rpc"
	"banger/internal/store"
	"banger/internal/system"
)

type Daemon struct {
	layout   paths.Layout
	config   model.DaemonConfig
	store    *store.Store
	runner   system.CommandRunner
	mu       sync.Mutex
	closing  chan struct{}
	once     sync.Once
	pid      int
	listener net.Listener
}

func Open(ctx context.Context) (*Daemon, error) {
	layout, err := paths.Resolve()
	if err != nil {
		return nil, err
	}
	if err := paths.Ensure(layout); err != nil {
		return nil, err
	}
	cfg, err := config.Load(layout)
	if err != nil {
		return nil, err
	}
	db, err := store.Open(layout.DBPath)
	if err != nil {
		return nil, err
	}
	d := &Daemon{
		layout:  layout,
		config:  cfg,
		store:   db,
		runner:  system.NewRunner(),
		closing: make(chan struct{}),
		pid:     os.Getpid(),
	}
	if err := d.ensureDefaultImage(ctx); err != nil {
		return nil, err
	}
	if err := d.reconcile(ctx); err != nil {
		return nil, err
	}
	return d, nil
}

func (d *Daemon) Close() error {
	var err error
	d.once.Do(func() {
		close(d.closing)
		if d.listener != nil {
			_ = d.listener.Close()
		}
		err = d.store.Close()
	})
	return err
}

func (d *Daemon) Serve(ctx context.Context) error {
	_ = os.Remove(d.layout.SocketPath)
	listener, err := net.Listen("unix", d.layout.SocketPath)
	if err != nil {
		return err
	}
	d.listener = listener
	defer listener.Close()
	defer os.Remove(d.layout.SocketPath)
	if err := os.Chmod(d.layout.SocketPath, 0o600); err != nil {
		return err
	}

	go d.backgroundLoop()

	for {
		conn, err := listener.Accept()
		if err != nil {
			select {
			case <-ctx.Done():
				return nil
			case <-d.closing:
				return nil
			default:
			}
			if ne, ok := err.(net.Error); ok && ne.Temporary() {
				time.Sleep(100 * time.Millisecond)
				continue
			}
			return err
		}
		go d.handleConn(conn)
	}
}

func (d *Daemon) handleConn(conn net.Conn) {
	defer conn.Close()
	var req rpc.Request
	if err := json.NewDecoder(bufio.NewReader(conn)).Decode(&req); err != nil {
		_ = json.NewEncoder(conn).Encode(rpc.NewError("bad_request", err.Error()))
		return
	}
	resp := d.dispatch(req)
	_ = json.NewEncoder(conn).Encode(resp)
}

func (d *Daemon) dispatch(req rpc.Request) rpc.Response {
	if req.Version != rpc.Version {
		return rpc.NewError("bad_version", fmt.Sprintf("unsupported version %d", req.Version))
	}
	ctx := context.Background()
	switch req.Method {
	case "ping":
		result, _ := rpc.NewResult(api.PingResult{Status: "ok", PID: d.pid})
		return result
	case "shutdown":
		go d.Close()
		result, _ := rpc.NewResult(api.ShutdownResult{Status: "stopping"})
		return result
	case "vm.create":
		params, err := rpc.DecodeParams[api.VMCreateParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.CreateVM(ctx, params)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.list":
		vms, err := d.store.ListVMs(ctx)
		return marshalResultOrError(api.VMListResult{VMs: vms}, err)
	case "vm.show":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.FindVM(ctx, params.IDOrName)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.start":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.StartVM(ctx, params.IDOrName)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.stop":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.StopVM(ctx, params.IDOrName)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.restart":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.RestartVM(ctx, params.IDOrName)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.delete":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.DeleteVM(ctx, params.IDOrName)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.set":
		params, err := rpc.DecodeParams[api.VMSetParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.SetVM(ctx, params)
		return marshalResultOrError(api.VMShowResult{VM: vm}, err)
	case "vm.stats":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, stats, err := d.GetVMStats(ctx, params.IDOrName)
		return marshalResultOrError(api.VMStatsResult{VM: vm, Stats: stats}, err)
	case "vm.logs":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.FindVM(ctx, params.IDOrName)
		if err != nil {
			return rpc.NewError("not_found", err.Error())
		}
		return marshalResultOrError(api.VMLogsResult{LogPath: vm.Runtime.LogPath}, nil)
	case "vm.ssh":
		params, err := rpc.DecodeParams[api.VMRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		vm, err := d.TouchVM(ctx, params.IDOrName)
		if err != nil {
			return rpc.NewError("not_found", err.Error())
		}
		if vm.State != model.VMStateRunning || !system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
			return rpc.NewError("not_running", fmt.Sprintf("vm %s is not running", vm.Name))
		}
		return marshalResultOrError(api.VMSSHResult{Name: vm.Name, GuestIP: vm.Runtime.GuestIP}, nil)
	case "image.list":
		images, err := d.store.ListImages(ctx)
		return marshalResultOrError(api.ImageListResult{Images: images}, err)
	case "image.show":
		params, err := rpc.DecodeParams[api.ImageRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		image, err := d.FindImage(ctx, params.IDOrName)
		return marshalResultOrError(api.ImageShowResult{Image: image}, err)
	case "image.build":
		params, err := rpc.DecodeParams[api.ImageBuildParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		image, err := d.BuildImage(ctx, params)
		return marshalResultOrError(api.ImageShowResult{Image: image}, err)
	case "image.delete":
		params, err := rpc.DecodeParams[api.ImageRefParams](req)
		if err != nil {
			return rpc.NewError("bad_request", err.Error())
		}
		image, err := d.DeleteImage(ctx, params.IDOrName)
		return marshalResultOrError(api.ImageShowResult{Image: image}, err)
	default:
		return rpc.NewError("unknown_method", req.Method)
	}
}

func (d *Daemon) backgroundLoop() {
	statsTicker := time.NewTicker(d.config.StatsPollInterval)
	staleTicker := time.NewTicker(model.DefaultStaleSweepInterval)
	defer statsTicker.Stop()
	defer staleTicker.Stop()
	for {
		select {
		case <-d.closing:
			return
		case <-statsTicker.C:
			_ = d.pollStats(context.Background())
		case <-staleTicker.C:
			_ = d.stopStaleVMs(context.Background())
		}
	}
}

func (d *Daemon) ensureDefaultImage(ctx context.Context) error {
	if d.config.DefaultImageName == "" {
		return nil
	}
	if _, err := d.store.GetImageByName(ctx, d.config.DefaultImageName); err == nil {
		return nil
	}
	rootfs := d.config.DefaultRootfs
	kernel := d.config.DefaultKernel
	initrd := d.config.DefaultInitrd
	if !exists(rootfs) || !exists(kernel) {
		return nil
	}
	id, err := model.NewID()
	if err != nil {
		return err
	}
	now := model.Now()
	image := model.Image{
		ID:           id,
		Name:         d.config.DefaultImageName,
		Managed:      false,
		RootfsPath:   rootfs,
		KernelPath:   kernel,
		InitrdPath:   initrd,
		ModulesDir:   d.config.DefaultModulesDir,
		PackagesPath: d.config.DefaultPackagesFile,
		Docker:       strings.Contains(filepath.Base(rootfs), "docker"),
		CreatedAt:    now,
		UpdatedAt:    now,
	}
	return d.store.UpsertImage(ctx, image)
}

func (d *Daemon) reconcile(ctx context.Context) error {
	vms, err := d.store.ListVMs(ctx)
	if err != nil {
		return err
	}
	for _, vm := range vms {
		if vm.State != model.VMStateRunning {
			continue
		}
		if system.ProcessRunning(vm.Runtime.PID, vm.Runtime.APISockPath) {
			continue
		}
		_ = d.cleanupRuntime(ctx, vm, true)
		vm.State = model.VMStateStopped
		vm.Runtime.State = model.VMStateStopped
		vm.Runtime.PID = 0
		vm.Runtime.TapDevice = ""
		vm.Runtime.APISockPath = ""
		vm.Runtime.BaseLoop = ""
		vm.Runtime.COWLoop = ""
		vm.Runtime.DMName = ""
		vm.Runtime.DMDev = ""
		vm.UpdatedAt = model.Now()
		if err := d.store.UpsertVM(ctx, vm); err != nil {
			return err
		}
	}
	return nil
}

func (d *Daemon) FindVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
	if idOrName == "" {
		return model.VMRecord{}, errors.New("vm id or name is required")
	}
	if vm, err := d.store.GetVM(ctx, idOrName); err == nil {
		return vm, nil
	}
	vms, err := d.store.ListVMs(ctx)
	if err != nil {
		return model.VMRecord{}, err
	}
	matchCount := 0
	var match model.VMRecord
	for _, vm := range vms {
		if strings.HasPrefix(vm.ID, idOrName) || strings.HasPrefix(vm.Name, idOrName) {
			match = vm
			matchCount++
		}
	}
	if matchCount == 1 {
		return match, nil
	}
	if matchCount > 1 {
		return model.VMRecord{}, fmt.Errorf("multiple VMs match %q", idOrName)
	}
	return model.VMRecord{}, fmt.Errorf("vm %q not found", idOrName)
}

func (d *Daemon) FindImage(ctx context.Context, idOrName string) (model.Image, error) {
	if idOrName == "" {
		return model.Image{}, errors.New("image id or name is required")
	}
	if image, err := d.store.GetImageByName(ctx, idOrName); err == nil {
		return image, nil
	}
	if image, err := d.store.GetImageByID(ctx, idOrName); err == nil {
		return image, nil
	}
	images, err := d.store.ListImages(ctx)
	if err != nil {
		return model.Image{}, err
	}
	matchCount := 0
	var match model.Image
	for _, image := range images {
		if strings.HasPrefix(image.ID, idOrName) || strings.HasPrefix(image.Name, idOrName) {
			match = image
			matchCount++
		}
	}
	if matchCount == 1 {
		return match, nil
	}
	if matchCount > 1 {
		return model.Image{}, fmt.Errorf("multiple images match %q", idOrName)
	}
	return model.Image{}, fmt.Errorf("image %q not found", idOrName)
}

func (d *Daemon) TouchVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
	d.mu.Lock()
	defer d.mu.Unlock()
	vm, err := d.FindVM(ctx, idOrName)
	if 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 marshalResultOrError(v any, err error) rpc.Response {
	if err != nil {
		return rpc.NewError("operation_failed", err.Error())
	}
	resp, marshalErr := rpc.NewResult(v)
	if marshalErr != nil {
		return rpc.NewError("marshal_failed", marshalErr.Error())
	}
	return resp
}

func exists(path string) bool {
	_, err := os.Stat(path)
	return err == nil
}