banger/internal/daemon/doctor.go

package daemon

import (
	"context"
	"fmt"
	"os"
	"runtime"
	"strings"

	"banger/internal/config"
	"banger/internal/imagecat"
	"banger/internal/model"
	"banger/internal/paths"
	"banger/internal/store"
	"banger/internal/system"
)

func Doctor(ctx context.Context) (system.Report, error) {
	layout, err := paths.Resolve()
	if err != nil {
		return system.Report{}, err
	}
	cfg, err := config.Load(layout)
	if err != nil {
		return system.Report{}, err
	}
	// Doctor must be read-only: running it should never mutate the
	// state DB (no migrations, no WAL checkpoint, no pragma writes).
	// Skip OpenReadOnly entirely when the DB file doesn't exist —
	// that's a fresh install, not an error condition. The first
	// daemon start will create the file. storeMissing differentiates
	// "no DB yet" (pass) from "DB present but unreadable" (fail) in
	// the report.
	d := &Daemon{
		layout: layout,
		config: cfg,
		runner: system.NewRunner(),
	}
	var storeErr error
	storeMissing := false
	if _, statErr := os.Stat(layout.DBPath); statErr != nil {
		if os.IsNotExist(statErr) {
			storeMissing = true
		} else {
			storeErr = statErr
		}
	} else {
		db, err := store.OpenReadOnly(layout.DBPath)
		if err != nil {
			storeErr = err
		} else {
			defer db.Close()
			d.store = db
		}
	}
	wireServices(d)
	return d.doctorReport(ctx, storeErr, storeMissing), nil
}

func (d *Daemon) doctorReport(ctx context.Context, storeErr error, storeMissing bool) system.Report {
	report := system.Report{}

	addArchitectureCheck(&report)

	switch {
	case storeMissing:
		report.AddPass("state store", "will be created on first daemon start at "+d.layout.DBPath)
	case storeErr != nil:
		report.AddFail(
			"state store",
			fmt.Sprintf("open %s: %v", d.layout.DBPath, storeErr),
			"remove or restore the file if corrupt; otherwise check its permissions",
		)
	default:
		report.AddPass("state store", "readable at "+d.layout.DBPath)
	}

	report.AddPreflight("host runtime", d.runtimeChecks(), runtimeStatus(d.config))
	report.AddPreflight("core vm lifecycle", d.coreVMLifecycleChecks(), "required host tools available")
	report.AddPreflight("vsock guest agent", d.vsockChecks(), "vsock guest agent prerequisites available")
	d.addVMDefaultsCheck(&report)
	d.addSSHShortcutCheck(&report)
	d.addCapabilityDoctorChecks(ctx, &report)

	return report
}

// addSSHShortcutCheck surfaces a gentle warning when banger maintains
// an ssh_config file but the user hasn't wired it into ~/.ssh/config.
// This is intentionally a warn, not a fail — the shortcut is opt-in
// convenience and `banger vm ssh` works either way.
func (d *Daemon) addSSHShortcutCheck(report *system.Report) {
	bangerConfig := BangerSSHConfigPath(d.layout)
	if strings.TrimSpace(bangerConfig) == "" {
		return
	}
	if _, err := os.Stat(bangerConfig); err != nil {
		// No banger ssh_config rendered yet — nothing to include.
		return
	}
	installed, err := UserSSHIncludeInstalled()
	if err != nil {
		report.AddWarn("ssh shortcut", fmt.Sprintf("could not read ~/.ssh/config: %v", err))
		return
	}
	if installed {
		report.AddPass("ssh shortcut", "enabled — `ssh <name>.vm` routes through banger")
		return
	}
	report.AddWarn(
		"ssh shortcut",
		fmt.Sprintf("`ssh <name>.vm` not enabled (opt-in); run `banger ssh-config --install` or add `Include %s` to ~/.ssh/config", bangerConfig),
	)
}

// addArchitectureCheck surfaces a hard-fail when banger is running on
// a non-amd64 host. Companion binaries are pinned to amd64 in the
// Makefile, the published kernel catalog ships only x86_64 images, and
// OCI import pulls linux/amd64 layers. Letting users discover this
// through cryptic downstream failures is worse than saying it up front.
func addArchitectureCheck(report *system.Report) {
	if runtime.GOARCH == "amd64" {
		report.AddPass("host architecture", "amd64")
		return
	}
	report.AddFail(
		"host architecture",
		fmt.Sprintf("running on %s; banger today only supports amd64/x86_64 hosts", runtime.GOARCH),
		"companion build, kernel catalog, and OCI import all assume linux/amd64",
	)
}

// addVMDefaultsCheck surfaces the effective VM sizing that `vm run` /
// `vm create` will apply when the user omits the flags. Shown as a
// PASS check so it always renders, with per-field provenance
// (config|auto|builtin) so users can tell what's driving each number.
func (d *Daemon) addVMDefaultsCheck(report *system.Report) {
	host, err := system.ReadHostResources()
	var cpus int
	var memBytes int64
	if err == nil {
		cpus = host.CPUCount
		memBytes = host.TotalMemoryBytes
	}
	defaults := model.ResolveVMDefaults(d.config.VMDefaults, cpus, memBytes)
	details := []string{
		fmt.Sprintf("vcpu:   %d (%s)", defaults.VCPUCount, defaults.VCPUSource),
		fmt.Sprintf("memory: %d MiB (%s)", defaults.MemoryMiB, defaults.MemorySource),
		fmt.Sprintf("disk:   %s (%s)", model.FormatSizeBytes(defaults.WorkDiskSizeBytes), defaults.WorkDiskSource),
		"override any of these in ~/.config/banger/config.toml under [vm_defaults]",
	}
	report.AddPass("vm defaults", details...)
}

func (d *Daemon) runtimeChecks() *system.Preflight {
	checks := system.NewPreflight()
	checks.RequireExecutable(d.config.FirecrackerBin, "firecracker binary", `install firecracker or set "firecracker_bin"`)
	checks.RequireFile(d.config.SSHKeyPath, "ssh private key", `set "ssh_key_path" or let banger create its default key`)
	if helper, err := vsockAgentBinary(d.layout); err == nil {
		checks.RequireExecutable(helper, "vsock agent helper", `run 'make build' or reinstall banger`)
	} else {
		checks.Addf("%v", err)
	}
	if d.store != nil && strings.TrimSpace(d.config.DefaultImageName) != "" {
		name := d.config.DefaultImageName
		image, err := d.store.GetImageByName(context.Background(), name)
		if err == nil {
			checks.RequireFile(image.RootfsPath, "default image rootfs", `re-register or rebuild the default image`)
			checks.RequireFile(image.KernelPath, "default image kernel", `re-register or rebuild the default image`)
			if strings.TrimSpace(image.InitrdPath) != "" {
				checks.RequireFile(image.InitrdPath, "default image initrd", `re-register or rebuild the default image`)
			}
		} else if !defaultImageInCatalog(name) {
			checks.Addf("default image %q is not registered and not in the imagecat catalog", name)
		}
		// If the default image isn't local but is cataloged, vm create
		// will auto-pull it on first use — no error to surface.
	}
	return checks
}

func defaultImageInCatalog(name string) bool {
	catalog, err := imagecat.LoadEmbedded()
	if err != nil {
		return false
	}
	_, err = catalog.Lookup(name)
	return err == nil
}

func (d *Daemon) coreVMLifecycleChecks() *system.Preflight {
	checks := system.NewPreflight()
	d.vm.addBaseStartCommandPrereqs(checks)
	return checks
}

func (d *Daemon) vsockChecks() *system.Preflight {
	checks := system.NewPreflight()
	if helper, err := vsockAgentBinary(d.layout); err == nil {
		checks.RequireExecutable(helper, "vsock agent helper", `run 'make build' or reinstall banger`)
	} else {
		checks.Addf("%v", err)
	}
	checks.RequireFile(d.vm.vsockHostDevice, "vsock host device", "load the vhost_vsock kernel module on the host")
	return checks
}

func runtimeStatus(cfg model.DaemonConfig) string {
	if strings.TrimSpace(cfg.FirecrackerBin) == "" {
		return "firecracker not configured"
	}
	return "firecracker and ssh key resolved"
}