banger/internal/daemon/vm_test.go

package daemon

import (
	"context"
	"errors"
	"fmt"
	"os"
	"os/exec"
	"path/filepath"
	"strconv"
	"strings"
	"testing"
	"time"

	"banger/internal/api"
	"banger/internal/model"
	"banger/internal/paths"
	"banger/internal/store"
	"banger/internal/vmdns"
)

func TestFindVMPrefixResolution(t *testing.T) {
	t.Parallel()

	ctx := context.Background()
	db := openDaemonStore(t)
	d := &Daemon{store: db}

	for _, vm := range []model.VMRecord{
		testVM("alpha", "image-alpha", "172.16.0.2"),
		testVM("alpine", "image-alpha", "172.16.0.3"),
		testVM("bravo", "image-alpha", "172.16.0.4"),
	} {
		if err := db.UpsertVM(ctx, vm); err != nil {
			t.Fatalf("UpsertVM(%s): %v", vm.Name, err)
		}
	}

	vm, err := d.FindVM(ctx, "alpha")
	if err != nil || vm.Name != "alpha" {
		t.Fatalf("FindVM(alpha) = %+v, %v", vm, err)
	}

	vm, err = d.FindVM(ctx, "br")
	if err != nil || vm.Name != "bravo" {
		t.Fatalf("FindVM(br) = %+v, %v", vm, err)
	}

	_, err = d.FindVM(ctx, "al")
	if err == nil || !strings.Contains(err.Error(), "multiple VMs match") {
		t.Fatalf("FindVM(al) error = %v, want ambiguity", err)
	}

	_, err = d.FindVM(ctx, "missing")
	if err == nil || !strings.Contains(err.Error(), `vm "missing" not found`) {
		t.Fatalf("FindVM(missing) error = %v, want not-found", err)
	}
}

func TestFindImagePrefixResolution(t *testing.T) {
	t.Parallel()

	ctx := context.Background()
	db := openDaemonStore(t)
	d := &Daemon{store: db}

	for _, image := range []model.Image{
		testImage("base"),
		testImage("basic"),
		testImage("docker"),
	} {
		if err := db.UpsertImage(ctx, image); err != nil {
			t.Fatalf("UpsertImage(%s): %v", image.Name, err)
		}
	}

	image, err := d.FindImage(ctx, "base")
	if err != nil || image.Name != "base" {
		t.Fatalf("FindImage(base) = %+v, %v", image, err)
	}

	image, err = d.FindImage(ctx, "dock")
	if err != nil || image.Name != "docker" {
		t.Fatalf("FindImage(dock) = %+v, %v", image, err)
	}

	_, err = d.FindImage(ctx, "ba")
	if err == nil || !strings.Contains(err.Error(), "multiple images match") {
		t.Fatalf("FindImage(ba) error = %v, want ambiguity", err)
	}

	_, err = d.FindImage(ctx, "missing")
	if err == nil || !strings.Contains(err.Error(), `image "missing" not found`) {
		t.Fatalf("FindImage(missing) error = %v, want not-found", err)
	}
}

func TestReconcileStopsStaleRunningVMAndClearsRuntimeHandles(t *testing.T) {
	t.Parallel()

	ctx := context.Background()
	db := openDaemonStore(t)
	apiSock := filepath.Join(t.TempDir(), "fc.sock")
	if err := os.WriteFile(apiSock, []byte{}, 0o644); err != nil {
		t.Fatalf("WriteFile(api sock): %v", err)
	}
	vm := testVM("stale", "image-stale", "172.16.0.9")
	vm.State = model.VMStateRunning
	vm.Runtime.State = model.VMStateRunning
	vm.Runtime.PID = 999999
	vm.Runtime.APISockPath = apiSock
	vm.Runtime.DMName = "fc-rootfs-stale"
	vm.Runtime.DMDev = "/dev/mapper/fc-rootfs-stale"
	vm.Runtime.COWLoop = "/dev/loop11"
	vm.Runtime.BaseLoop = "/dev/loop10"
	vm.Runtime.DNSName = ""
	if err := db.UpsertVM(ctx, vm); err != nil {
		t.Fatalf("UpsertVM: %v", err)
	}

	runner := &scriptedRunner{
		t: t,
		steps: []runnerStep{
			{call: runnerCall{name: "pgrep", args: []string{"-n", "-f", apiSock}}, err: errors.New("exit status 1")},
			sudoStep("", nil, "dmsetup", "remove", "fc-rootfs-stale"),
			sudoStep("", nil, "losetup", "-d", "/dev/loop11"),
			sudoStep("", nil, "losetup", "-d", "/dev/loop10"),
		},
	}
	d := &Daemon{store: db, runner: runner}

	if err := d.reconcile(ctx); err != nil {
		t.Fatalf("reconcile: %v", err)
	}
	runner.assertExhausted()

	got, err := db.GetVM(ctx, vm.ID)
	if err != nil {
		t.Fatalf("GetVM: %v", err)
	}
	if got.State != model.VMStateStopped || got.Runtime.State != model.VMStateStopped {
		t.Fatalf("vm state after reconcile = %s/%s, want stopped", got.State, got.Runtime.State)
	}
	if got.Runtime.PID != 0 || got.Runtime.APISockPath != "" || got.Runtime.DMName != "" || got.Runtime.COWLoop != "" || got.Runtime.BaseLoop != "" {
		t.Fatalf("runtime handles not cleared after reconcile: %+v", got.Runtime)
	}
}

func TestRebuildDNSIncludesOnlyLiveRunningVMs(t *testing.T) {
	t.Parallel()

	ctx := context.Background()
	db := openDaemonStore(t)

	liveSock := filepath.Join(t.TempDir(), "live.sock")
	liveCmd := startFakeFirecrackerProcess(t, liveSock)
	t.Cleanup(func() {
		_ = liveCmd.Process.Kill()
		_ = liveCmd.Wait()
	})

	live := testVM("live", "image-live", "172.16.0.21")
	live.State = model.VMStateRunning
	live.Runtime.State = model.VMStateRunning
	live.Runtime.PID = liveCmd.Process.Pid
	live.Runtime.APISockPath = liveSock

	stale := testVM("stale", "image-stale", "172.16.0.22")
	stale.State = model.VMStateRunning
	stale.Runtime.State = model.VMStateRunning
	stale.Runtime.PID = 999999
	stale.Runtime.APISockPath = filepath.Join(t.TempDir(), "stale.sock")

	stopped := testVM("stopped", "image-stopped", "172.16.0.23")

	for _, vm := range []model.VMRecord{live, stale, stopped} {
		if err := db.UpsertVM(ctx, vm); err != nil {
			t.Fatalf("UpsertVM(%s): %v", vm.Name, err)
		}
	}

	server, err := vmdns.New("127.0.0.1:0", nil)
	if err != nil {
		t.Fatalf("vmdns.New: %v", err)
	}
	t.Cleanup(func() {
		if err := server.Close(); err != nil {
			t.Fatalf("server.Close: %v", err)
		}
	})

	d := &Daemon{store: db, vmDNS: server}
	if err := d.rebuildDNS(ctx); err != nil {
		t.Fatalf("rebuildDNS: %v", err)
	}

	if _, ok := server.Lookup("live.vm"); !ok {
		t.Fatal("live.vm missing after rebuildDNS")
	}
	if _, ok := server.Lookup("stale.vm"); ok {
		t.Fatal("stale.vm should not be published")
	}
	if _, ok := server.Lookup("stopped.vm"); ok {
		t.Fatal("stopped.vm should not be published")
	}
}

func TestSetVMRejectsStoppedOnlyChangesForRunningVM(t *testing.T) {
	t.Parallel()

	ctx := context.Background()
	db := openDaemonStore(t)
	apiSock := filepath.Join(t.TempDir(), "running.sock")
	cmd := startFakeFirecrackerProcess(t, apiSock)
	t.Cleanup(func() {
		_ = cmd.Process.Kill()
		_ = cmd.Wait()
	})

	vm := testVM("running", "image-run", "172.16.0.10")
	vm.State = model.VMStateRunning
	vm.Runtime.State = model.VMStateRunning
	vm.Runtime.PID = cmd.Process.Pid
	vm.Runtime.APISockPath = apiSock
	if err := db.UpsertVM(ctx, vm); err != nil {
		t.Fatalf("UpsertVM: %v", err)
	}

	d := &Daemon{store: db}
	tests := []struct {
		name   string
		params api.VMSetParams
		want   string
	}{
		{
			name:   "vcpu",
			params: api.VMSetParams{IDOrName: vm.ID, VCPUCount: ptr(4)},
			want:   "vcpu changes require the VM to be stopped",
		},
		{
			name:   "memory",
			params: api.VMSetParams{IDOrName: vm.ID, MemoryMiB: ptr(2048)},
			want:   "memory changes require the VM to be stopped",
		},
		{
			name:   "disk",
			params: api.VMSetParams{IDOrName: vm.ID, WorkDiskSize: "16G"},
			want:   "disk changes require the VM to be stopped",
		},
	}

	for _, tt := range tests {
		t.Run(tt.name, func(t *testing.T) {
			_, err := d.SetVM(ctx, tt.params)
			if err == nil || !strings.Contains(err.Error(), tt.want) {
				t.Fatalf("SetVM(%s) error = %v, want %q", tt.name, err, tt.want)
			}
		})
	}
}

func TestSetVMDiskResizeFailsPreflightWhenToolsMissing(t *testing.T) {
	ctx := context.Background()
	db := openDaemonStore(t)
	workDisk := filepath.Join(t.TempDir(), "root.ext4")
	if err := os.WriteFile(workDisk, []byte("disk"), 0o644); err != nil {
		t.Fatalf("WriteFile: %v", err)
	}
	vm := testVM("resize", "image-resize", "172.16.0.11")
	vm.Runtime.WorkDiskPath = workDisk
	vm.Spec.WorkDiskSizeBytes = 8 * 1024 * 1024 * 1024
	if err := db.UpsertVM(ctx, vm); err != nil {
		t.Fatalf("UpsertVM: %v", err)
	}

	t.Setenv("PATH", t.TempDir())
	d := &Daemon{store: db}
	_, err := d.SetVM(ctx, api.VMSetParams{IDOrName: vm.ID, WorkDiskSize: "16G"})
	if err == nil || !strings.Contains(err.Error(), "work disk resize preflight failed") {
		t.Fatalf("SetVM() error = %v, want preflight failure", err)
	}
}

func TestFlattenNestedWorkHomeCopiesEntriesIndividually(t *testing.T) {
	t.Parallel()

	workMount := t.TempDir()
	nestedHome := filepath.Join(workMount, "root")
	if err := os.MkdirAll(filepath.Join(nestedHome, ".ssh"), 0o755); err != nil {
		t.Fatalf("MkdirAll(.ssh): %v", err)
	}
	if err := os.WriteFile(filepath.Join(nestedHome, "notes.txt"), []byte("seed"), 0o644); err != nil {
		t.Fatalf("WriteFile(notes.txt): %v", err)
	}

	runner := &scriptedRunner{
		t: t,
		steps: []runnerStep{
			sudoStep("", nil, "chmod", "755", nestedHome),
			sudoStep("", nil, "cp", "-a", filepath.Join(nestedHome, ".ssh"), workMount+"/"),
			sudoStep("", nil, "cp", "-a", filepath.Join(nestedHome, "notes.txt"), workMount+"/"),
			sudoStep("", nil, "rm", "-rf", nestedHome),
		},
	}
	d := &Daemon{runner: runner}

	if err := d.flattenNestedWorkHome(context.Background(), workMount); err != nil {
		t.Fatalf("flattenNestedWorkHome: %v", err)
	}
	runner.assertExhausted()
}

func TestCreateVMRejectsNonPositiveCPUAndMemory(t *testing.T) {
	d := &Daemon{}
	if _, err := d.CreateVM(context.Background(), api.VMCreateParams{VCPUCount: ptr(0)}); err == nil || !strings.Contains(err.Error(), "vcpu must be a positive integer") {
		t.Fatalf("CreateVM(vcpu=0) error = %v", err)
	}
	if _, err := d.CreateVM(context.Background(), api.VMCreateParams{MemoryMiB: ptr(-1)}); err == nil || !strings.Contains(err.Error(), "memory must be a positive integer") {
		t.Fatalf("CreateVM(memory=-1) error = %v", err)
	}
}

func TestCreateVMUsesDefaultsWhenCPUAndMemoryOmitted(t *testing.T) {
	ctx := context.Background()
	db := openDaemonStore(t)
	image := testImage("default")
	if err := db.UpsertImage(ctx, image); err != nil {
		t.Fatalf("UpsertImage: %v", err)
	}
	d := &Daemon{
		store: db,
		layout: paths.Layout{
			VMsDir: t.TempDir(),
		},
		config: model.DaemonConfig{
			DefaultImageName: image.Name,
			BridgeIP:         model.DefaultBridgeIP,
		},
	}

	vm, err := d.CreateVM(ctx, api.VMCreateParams{Name: "defaults", ImageName: image.Name, NoStart: true})
	if err != nil {
		t.Fatalf("CreateVM: %v", err)
	}
	if vm.Spec.VCPUCount != model.DefaultVCPUCount {
		t.Fatalf("VCPUCount = %d, want %d", vm.Spec.VCPUCount, model.DefaultVCPUCount)
	}
	if vm.Spec.MemoryMiB != model.DefaultMemoryMiB {
		t.Fatalf("MemoryMiB = %d, want %d", vm.Spec.MemoryMiB, model.DefaultMemoryMiB)
	}
}

func TestSetVMRejectsNonPositiveCPUAndMemory(t *testing.T) {
	ctx := context.Background()
	db := openDaemonStore(t)
	vm := testVM("validate", "image-validate", "172.16.0.13")
	if err := db.UpsertVM(ctx, vm); err != nil {
		t.Fatalf("UpsertVM: %v", err)
	}
	d := &Daemon{store: db}

	if _, err := d.SetVM(ctx, api.VMSetParams{IDOrName: vm.ID, VCPUCount: ptr(0)}); err == nil || !strings.Contains(err.Error(), "vcpu must be a positive integer") {
		t.Fatalf("SetVM(vcpu=0) error = %v", err)
	}
	if _, err := d.SetVM(ctx, api.VMSetParams{IDOrName: vm.ID, MemoryMiB: ptr(0)}); err == nil || !strings.Contains(err.Error(), "memory must be a positive integer") {
		t.Fatalf("SetVM(memory=0) error = %v", err)
	}
}

func TestCollectStatsIgnoresMalformedMetricsFile(t *testing.T) {
	t.Parallel()

	overlay := filepath.Join(t.TempDir(), "system.cow")
	workDisk := filepath.Join(t.TempDir(), "root.ext4")
	metrics := filepath.Join(t.TempDir(), "metrics.json")
	for _, path := range []string{overlay, workDisk} {
		if err := os.WriteFile(path, []byte("allocated"), 0o644); err != nil {
			t.Fatalf("WriteFile(%s): %v", path, err)
		}
	}
	if err := os.WriteFile(metrics, []byte("{not-json}\n"), 0o644); err != nil {
		t.Fatalf("WriteFile(metrics): %v", err)
	}

	d := &Daemon{}
	stats, err := d.collectStats(context.Background(), model.VMRecord{
		Runtime: model.VMRuntime{
			SystemOverlay: overlay,
			WorkDiskPath:  workDisk,
			MetricsPath:   metrics,
		},
	})
	if err != nil {
		t.Fatalf("collectStats: %v", err)
	}
	if stats.MetricsRaw != nil {
		t.Fatalf("MetricsRaw = %v, want nil for malformed metrics", stats.MetricsRaw)
	}
	if stats.SystemOverlayBytes == 0 || stats.WorkDiskBytes == 0 {
		t.Fatalf("allocated bytes not captured: %+v", stats)
	}
}

func TestValidateStartPrereqsReportsNATUplinkFailure(t *testing.T) {
	ctx := context.Background()
	binDir := t.TempDir()
	for _, name := range []string{
		"sudo", "ip", "dmsetup", "losetup", "blockdev", "truncate", "pgrep", "ps",
		"chown", "chmod", "kill", "e2cp", "e2rm", "debugfs", "mkfs.ext4", "mount",
		"umount", "cp", "iptables", "sysctl",
	} {
		writeFakeExecutable(t, filepath.Join(binDir, name))
	}
	t.Setenv("PATH", binDir)

	firecrackerBin := filepath.Join(t.TempDir(), "firecracker")
	rootfsPath := filepath.Join(t.TempDir(), "rootfs.ext4")
	kernelPath := filepath.Join(t.TempDir(), "vmlinux")
	for _, path := range []string{firecrackerBin, rootfsPath, kernelPath} {
		writeFakeExecutable(t, path)
	}

	runner := &scriptedRunner{
		t: t,
		steps: []runnerStep{
			{call: runnerCall{name: "ip", args: []string{"route", "show", "default"}}, out: []byte("10.0.0.0/24 dev br-fc\n")},
		},
	}
	d := &Daemon{
		runner: runner,
		config: model.DaemonConfig{
			FirecrackerBin: firecrackerBin,
		},
	}
	vm := testVM("nat", "image-nat", "172.16.0.12")
	vm.Spec.NATEnabled = true
	vm.Runtime.WorkDiskPath = filepath.Join(t.TempDir(), "missing-root.ext4")
	image := testImage("image-nat")
	image.RootfsPath = rootfsPath
	image.KernelPath = kernelPath

	err := d.validateStartPrereqs(ctx, vm, image)
	if err == nil || !strings.Contains(err.Error(), "uplink interface for NAT") {
		t.Fatalf("validateStartPrereqs() error = %v, want NAT uplink failure", err)
	}
	runner.assertExhausted()
}

func TestCleanupRuntimeRediscoversLiveFirecrackerPID(t *testing.T) {
	apiSock := filepath.Join(t.TempDir(), "fc.sock")
	fake := startFakeFirecrackerProcess(t, apiSock)
	t.Cleanup(func() {
		if fake.ProcessState == nil || !fake.ProcessState.Exited() {
			_ = fake.Process.Kill()
			_ = fake.Wait()
		}
	})

	runner := &processKillingRunner{
		scriptedRunner: &scriptedRunner{
			t: t,
			steps: []runnerStep{
				{call: runnerCall{name: "pgrep", args: []string{"-n", "-f", apiSock}}, out: []byte(strconv.Itoa(fake.Process.Pid) + "\n")},
				sudoStep("", nil, "kill", "-KILL", strconv.Itoa(fake.Process.Pid)),
			},
		},
		proc: fake,
	}
	d := &Daemon{runner: runner}
	vm := testVM("cleanup", "image-cleanup", "172.16.0.22")
	vm.Runtime.PID = fake.Process.Pid + 999
	vm.Runtime.APISockPath = apiSock

	if err := d.cleanupRuntime(context.Background(), vm, true); err != nil {
		t.Fatalf("cleanupRuntime returned error: %v", err)
	}
	runner.assertExhausted()
	if systemProcessRunning(fake.Process.Pid, apiSock) {
		t.Fatalf("fake firecracker pid %d still looks running", fake.Process.Pid)
	}
}

func openDaemonStore(t *testing.T) *store.Store {
	t.Helper()
	db, err := store.Open(filepath.Join(t.TempDir(), "state.db"))
	if err != nil {
		t.Fatalf("store.Open: %v", err)
	}
	t.Cleanup(func() {
		_ = db.Close()
	})
	return db
}

func testVM(name, imageID, guestIP string) model.VMRecord {
	now := time.Date(2026, time.March, 16, 12, 0, 0, 0, time.UTC)
	return model.VMRecord{
		ID:            name + "-id",
		Name:          name,
		ImageID:       imageID,
		State:         model.VMStateStopped,
		CreatedAt:     now,
		UpdatedAt:     now,
		LastTouchedAt: now,
		Spec: model.VMSpec{
			VCPUCount:             2,
			MemoryMiB:             1024,
			SystemOverlaySizeByte: model.DefaultSystemOverlaySize,
			WorkDiskSizeBytes:     model.DefaultWorkDiskSize,
		},
		Runtime: model.VMRuntime{
			State:         model.VMStateStopped,
			GuestIP:       guestIP,
			DNSName:       name + ".vm",
			VMDir:         filepath.Join("/state", name),
			SystemOverlay: filepath.Join("/state", name, "system.cow"),
			WorkDiskPath:  filepath.Join("/state", name, "root.ext4"),
		},
	}
}

func testImage(name string) model.Image {
	now := time.Date(2026, time.March, 16, 12, 0, 0, 0, time.UTC)
	return model.Image{
		ID:         name + "-id",
		Name:       name,
		RootfsPath: filepath.Join("/images", name+".ext4"),
		KernelPath: filepath.Join("/kernels", name),
		CreatedAt:  now,
		UpdatedAt:  now,
	}
}

func startFakeFirecrackerProcess(t *testing.T, apiSock string) *exec.Cmd {
	t.Helper()

	cmd := exec.Command("bash", "-lc", fmt.Sprintf("exec -a %q sleep 30", "firecracker --api-sock "+apiSock))
	if err := cmd.Start(); err != nil {
		t.Fatalf("start fake firecracker: %v", err)
	}

	deadline := time.Now().Add(2 * time.Second)
	for time.Now().Before(deadline) {
		if cmd.Process != nil && cmd.Process.Pid > 0 && systemProcessRunning(cmd.Process.Pid, apiSock) {
			return cmd
		}
		time.Sleep(20 * time.Millisecond)
	}
	_ = cmd.Process.Kill()
	_ = cmd.Wait()
	t.Fatalf("fake firecracker process never looked running for %s", apiSock)
	return nil
}

type processKillingRunner struct {
	*scriptedRunner
	proc *exec.Cmd
}

func (r *processKillingRunner) Run(ctx context.Context, name string, args ...string) ([]byte, error) {
	return r.scriptedRunner.Run(ctx, name, args...)
}

func (r *processKillingRunner) RunSudo(ctx context.Context, args ...string) ([]byte, error) {
	out, err := r.scriptedRunner.RunSudo(ctx, args...)
	if err != nil {
		return out, err
	}
	if len(args) >= 3 && args[0] == "kill" && args[1] == "-KILL" && r.proc != nil && (r.proc.ProcessState == nil || !r.proc.ProcessState.Exited()) {
		_ = r.proc.Process.Kill()
		_ = r.proc.Wait()
	}
	return out, nil
}

func systemProcessRunning(pid int, apiSock string) bool {
	data, err := os.ReadFile(filepath.Join("/proc", strconv.Itoa(pid), "cmdline"))
	if err != nil {
		return false
	}
	cmdline := strings.ReplaceAll(string(data), "\x00", " ")
	return strings.Contains(cmdline, "firecracker") && strings.Contains(cmdline, apiSock)
}

func writeFakeExecutable(t *testing.T, path string) {
	t.Helper()
	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
		t.Fatalf("MkdirAll(%s): %v", filepath.Dir(path), err)
	}
	if err := os.WriteFile(path, []byte("#!/bin/sh\nexit 0\n"), 0o755); err != nil {
		t.Fatalf("WriteFile(%s): %v", path, err)
	}
}

func ptr[T any](value T) *T {
	return &value
}