add Confidential Computing readiness check + collect nvidia-smi conf-compute -q

New read-only "Check" step reports whether this server can run NVIDIA
Confidential Computing: CPU TEE support (Intel TDX / AMD SEV-SNP, via
dmesg and kvm_amd sysfs params) and GPU firmware CC capability (via
`nvidia-smi conf-compute -q`). Also collect that command's output into
the techdump export bundle.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

audit/internal/app/app.go

@@ -134,6 +134,7 @@ type satRunner interface {
 	ResetNvidiaGPU(index int) (string, error)
 	RunMemoryAcceptancePack(ctx context.Context, baseDir string, sizeMB, passes int, logFunc func(string)) (string, error)
 	RunStorageAcceptancePack(ctx context.Context, baseDir string, extended bool, logFunc func(string)) (string, error)
+	RunConfidentialComputingCheckPack(ctx context.Context, baseDir string, logFunc func(string)) (string, error)
 	RunCPUAcceptancePack(ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error)
 	ListNvidiaGPUs() ([]platform.NvidiaGPU, error)
 	DetectGPUVendor() string

audit/internal/app/app_packs.go

@@ -206,6 +206,22 @@ func (a *App) RunStorageAcceptancePackResult(baseDir string) (ActionResult, erro
 	return ActionResult{Title: "Storage SAT", Body: satResultBody(path)}, err
 }
 
+func (a *App) RunConfidentialComputingCheckPackCtx(ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
+	if strings.TrimSpace(baseDir) == "" {
+		baseDir = DefaultSATBaseDir
+	}
+	return a.sat.RunConfidentialComputingCheckPack(ctx, baseDir, logFunc)
+}
+
+func (a *App) RunConfidentialComputingCheckPack(baseDir string, logFunc func(string)) (string, error) {
+	return a.RunConfidentialComputingCheckPackCtx(context.Background(), baseDir, logFunc)
+}
+
+func (a *App) RunConfidentialComputingCheckPackResult(baseDir string) (ActionResult, error) {
+	path, err := a.RunConfidentialComputingCheckPack(baseDir, nil)
+	return ActionResult{Title: "Confidential Computing Check", Body: satResultBody(path)}, err
+}
+
 func (a *App) DetectGPUVendor() string {
 	return a.sat.DetectGPUVendor()
 }

audit/internal/app/app_test.go

@@ -243,6 +243,10 @@ func (f fakeSAT) RunStorageAcceptancePack(_ context.Context, baseDir string, _ b
 	return f.runStorageFn(baseDir)
 }
 
+func (f fakeSAT) RunConfidentialComputingCheckPack(_ context.Context, baseDir string, _ func(string)) (string, error) {
+	return "", nil
+}
+
 func (f fakeSAT) RunCPUAcceptancePack(_ context.Context, baseDir string, durationSec int, _ func(string)) (string, error) {
 	if f.runCPUFn != nil {
 		return f.runCPUFn(baseDir, durationSec)

audit/internal/app/blackbox.go

@@ -365,7 +365,6 @@ func (w *blackboxWorker) currentFlushPeriod() time.Duration {
 
 func (w *blackboxWorker) finishCycle(duration time.Duration, err error) {
 	w.mu.Lock()
-	defer w.mu.Unlock()
 	w.lastDuration = duration
 	if err != nil {
 		w.status = "degraded"
@@ -383,6 +382,10 @@ func (w *blackboxWorker) finishCycle(duration time.Duration, err error) {
 		}
 		w.flushPeriod = adjustFlushPeriod(w.flushPeriod, duration, true, w.fastCycles)
 	}
+	w.mu.Unlock()
+	// persistState must be called without w.mu held: it acquires rt.mu then
+	// each worker.mu inside persistStateLocked, so holding w.mu here would
+	// cause a deadlock (w.mu → rt.mu → w.mu).
 	w.runtime.persistState()
 }
 

audit/internal/collector/collector.go

@@ -49,7 +49,8 @@ func Run(_ runtimeenv.Mode) schema.HardwareIngestRequest {
 	snap.VROCLicense = collectVROCLicense(snap.PCIeDevices)
 	snap.PowerSupplies = collectPSUs(derefString(snap.Board.Manufacturer))
 	snap.PowerSupplies = enrichPSUsWithTelemetry(snap.PowerSupplies, sensorDoc)
-	snap.Sensors = buildSensorsFromDoc(sensorDoc)
+	snap.Sensors = mergeIPMISensors(buildSensorsFromDoc(sensorDoc), collectIPMISensors())
+	snap.EventLogs = append(collectIPMISEL(), collectDmesgErrors()...)
 	finalizeSnapshot(&snap, collectedAt)
 
 	// remaining collectors added in steps 1.8 – 1.10

audit/internal/collector/dmesg_events.go

@@ -0,0 +1,129 @@
+package collector
+
+import (
+	"bee/audit/internal/schema"
+	"log/slog"
+	"os/exec"
+	"regexp"
+	"strings"
+	"time"
+)
+
+// dmesg -T output: [Thu Jun 18 14:23:45 2026] message
+// dmesg without -T: [  123.456789] message
+var dmesgTimestampRE = regexp.MustCompile(`^\[([^\]]+)\]\s*(.*)$`)
+
+// Keywords that indicate an error or hardware problem worth capturing.
+var dmesgErrorPatterns = []*regexp.Regexp{
+	regexp.MustCompile(`(?i)\berr(or)?\b`),
+	regexp.MustCompile(`(?i)\bfail(ed|ure)?\b`),
+	regexp.MustCompile(`(?i)\bfault\b`),
+	regexp.MustCompile(`(?i)\bwarn(ing)?\b`),
+	regexp.MustCompile(`(?i)\bAER\b`),
+	regexp.MustCompile(`(?i)\bXid\b`),
+	regexp.MustCompile(`(?i)\bNVRM\b`),
+	regexp.MustCompile(`(?i)\bpanic\b`),
+	regexp.MustCompile(`(?i)\bcorrected\b`),
+	regexp.MustCompile(`(?i)\buncorrect`),
+	regexp.MustCompile(`(?i)\bECC\b`),
+	regexp.MustCompile(`(?i)\btimeout\b`),
+	regexp.MustCompile(`(?i)\breset\b`),
+	regexp.MustCompile(`(?i)\bdead\b`),
+	regexp.MustCompile(`(?i)\bhang\b`),
+	regexp.MustCompile(`(?i)\bstall\b`),
+	regexp.MustCompile(`(?i)\bdisabled\b`),
+}
+
+// collectDmesgErrors runs `dmesg -T` (or `dmesg` without -T on failure) and
+// returns only lines that match known error/warning patterns.
+func collectDmesgErrors() []schema.HardwareEventLog {
+	out, err := exec.Command("dmesg", "-T").Output()
+	if err != nil || len(out) == 0 {
+		// Fallback: dmesg without human-readable timestamps
+		out, err = exec.Command("dmesg").Output()
+		if err != nil || len(out) == 0 {
+			return nil
+		}
+	}
+	entries := parseDmesgErrors(string(out))
+	if len(entries) == 0 {
+		return nil
+	}
+	slog.Info("dmesg: collected error entries", "count", len(entries))
+	return entries
+}
+
+func parseDmesgErrors(output string) []schema.HardwareEventLog {
+	var entries []schema.HardwareEventLog
+	collectedAt := time.Now().UTC().Format(time.RFC3339)
+
+	for _, line := range strings.Split(output, "\n") {
+		line = strings.TrimSpace(line)
+		if line == "" {
+			continue
+		}
+
+		var timestamp, message string
+		if m := dmesgTimestampRE.FindStringSubmatch(line); m != nil {
+			timestamp = strings.TrimSpace(m[1])
+			message = strings.TrimSpace(m[2])
+		} else {
+			message = line
+		}
+
+		if message == "" {
+			continue
+		}
+		if !matchesAny(message, dmesgErrorPatterns) {
+			continue
+		}
+
+		severity := dmesgSeverity(message)
+		source := "dmesg"
+
+		var eventTime *string
+		if timestamp != "" {
+			t := timestamp
+			eventTime = &t
+		} else {
+			eventTime = &collectedAt
+		}
+
+		entries = append(entries, schema.HardwareEventLog{
+			Source:    source,
+			EventTime: eventTime,
+			Severity:  &severity,
+			Message:   message,
+		})
+	}
+	return entries
+}
+
+func matchesAny(s string, patterns []*regexp.Regexp) bool {
+	for _, p := range patterns {
+		if p.MatchString(s) {
+			return true
+		}
+	}
+	return false
+}
+
+func dmesgSeverity(msg string) string {
+	lower := strings.ToLower(msg)
+	switch {
+	case strings.Contains(lower, "panic") ||
+		strings.Contains(lower, "aer") ||
+		strings.Contains(lower, "uncorrect") ||
+		strings.Contains(lower, "xid") ||
+		strings.Contains(lower, "nvrm"):
+		return statusCritical
+	case strings.Contains(lower, "error") ||
+		strings.Contains(lower, "fault") ||
+		strings.Contains(lower, "fail") ||
+		strings.Contains(lower, "dead") ||
+		strings.Contains(lower, "hang"):
+		return statusCritical
+	default:
+		return statusWarning
+	}
+}

audit/internal/collector/ipmi_sel.go

@@ -0,0 +1,90 @@
+package collector
+
+import (
+	"bee/audit/internal/schema"
+	"fmt"
+	"log/slog"
+	"os/exec"
+	"strings"
+)
+
+// collectIPMISEL runs `ipmitool sel list` and returns parsed event log entries.
+// Returns nil if ipmitool is unavailable or the SEL is empty.
+func collectIPMISEL() []schema.HardwareEventLog {
+	out, err := exec.Command("ipmitool", "sel", "list").Output()
+	if err != nil || len(out) == 0 {
+		return nil
+	}
+	entries := parseIPMISELOutput(string(out))
+	if len(entries) == 0 {
+		return nil
+	}
+	slog.Info("ipmi sel: collected", "entries", len(entries))
+	return entries
+}
+
+// parseIPMISELOutput parses `ipmitool sel list` output.
+// Line format: ID | date | time | sensor | event description | direction
+// Example:    1 | 06/18/2026 | 14:23:45 | Temperature #0x30 | Upper Critical going high | Asserted
+func parseIPMISELOutput(output string) []schema.HardwareEventLog {
+	var entries []schema.HardwareEventLog
+	for _, line := range strings.Split(output, "\n") {
+		line = strings.TrimSpace(line)
+		if line == "" {
+			continue
+		}
+		parts := strings.SplitN(line, "|", 6)
+		if len(parts) < 5 {
+			continue
+		}
+		id := strings.TrimSpace(parts[0])
+		date := strings.TrimSpace(parts[1])
+		timeStr := strings.TrimSpace(parts[2])
+		sensor := strings.TrimSpace(parts[3])
+		event := strings.TrimSpace(parts[4])
+		direction := ""
+		if len(parts) == 6 {
+			direction = strings.TrimSpace(parts[5])
+		}
+
+		var eventTime *string
+		if date != "" && timeStr != "" {
+			t := fmt.Sprintf("%s %s", date, timeStr)
+			eventTime = &t
+		}
+
+		message := event
+		if direction != "" && strings.EqualFold(direction, "Deasserted") {
+			message = event + " (Deasserted)"
+		}
+
+		severity := ipmiSELSeverity(event)
+		isActive := !strings.EqualFold(direction, "Deasserted")
+
+		entry := schema.HardwareEventLog{
+			Source:    "ipmi-sel",
+			EventTime: eventTime,
+			Severity:  &severity,
+			MessageID: &id,
+			Message:   message,
+			IsActive:  &isActive,
+		}
+		if sensor != "" {
+			entry.ComponentRef = &sensor
+		}
+		entries = append(entries, entry)
+	}
+	return entries
+}
+
+func ipmiSELSeverity(event string) string {
+	lower := strings.ToLower(event)
+	switch {
+	case strings.Contains(lower, "critical") || strings.Contains(lower, "non-recoverable"):
+		return statusCritical
+	case strings.Contains(lower, "non-critical") || strings.Contains(lower, "warning") || strings.Contains(lower, "degraded"):
+		return statusWarning
+	default:
+		return "info"
+	}
+}

audit/internal/collector/ipmi_sensors.go

@@ -0,0 +1,216 @@
+package collector
+
+import (
+	"bee/audit/internal/schema"
+	"log/slog"
+	"os/exec"
+	"strconv"
+	"strings"
+)
+
+// collectIPMISensors runs `ipmitool sensor` and returns parsed sensor readings.
+// Returns nil if ipmitool is unavailable or produces no output.
+func collectIPMISensors() *schema.HardwareSensors {
+	out, err := exec.Command("ipmitool", "sensor").Output()
+	if err != nil || len(out) == 0 {
+		return nil
+	}
+	result := parseIPMISensorOutput(string(out))
+	if result == nil {
+		return nil
+	}
+	slog.Info("ipmi sensors: collected",
+		"fans", len(result.Fans),
+		"temperatures", len(result.Temperatures),
+		"power", len(result.Power),
+		"other", len(result.Other),
+	)
+	return result
+}
+
+// parseIPMISensorOutput parses `ipmitool sensor` text output.
+// Each line: name | value | unit | status | lnr | lcr | lnc | unc | ucr | unr
+func parseIPMISensorOutput(output string) *schema.HardwareSensors {
+	result := &schema.HardwareSensors{}
+	seen := map[string]struct{}{}
+
+	for _, line := range strings.Split(output, "\n") {
+		line = strings.TrimSpace(line)
+		if line == "" {
+			continue
+		}
+		parts := strings.Split(line, "|")
+		if len(parts) < 4 {
+			continue
+		}
+		name := strings.TrimSpace(parts[0])
+		rawVal := strings.TrimSpace(parts[1])
+		unit := strings.TrimSpace(parts[2])
+		status := strings.TrimSpace(parts[3])
+
+		if name == "" || rawVal == "na" || rawVal == "N/A" || rawVal == "" {
+			continue
+		}
+
+		value, err := strconv.ParseFloat(rawVal, 64)
+		if err != nil {
+			continue
+		}
+
+		statusStr := normalizeIPMISensorStatus(status)
+
+		switch {
+		case strings.EqualFold(unit, "RPM"):
+			if duplicateSensor(seen, "fan", name) {
+				continue
+			}
+			rpm := int(value)
+			item := schema.HardwareFanSensor{Name: name, RPM: &rpm}
+			if statusStr != "" {
+				item.Status = &statusStr
+			}
+			result.Fans = append(result.Fans, item)
+
+		case strings.EqualFold(unit, "degrees C") || strings.EqualFold(unit, "C"):
+			if duplicateSensor(seen, "temp", name) {
+				continue
+			}
+			item := schema.HardwareTemperatureSensor{Name: name, Celsius: &value}
+			if len(parts) >= 9 {
+				if unc := parseIPMIThreshold(parts[7]); unc != nil {
+					item.ThresholdWarningCelsius = unc
+				}
+				if ucr := parseIPMIThreshold(parts[8]); ucr != nil {
+					item.ThresholdCriticalCelsius = ucr
+				}
+			}
+			if statusStr != "" {
+				item.Status = &statusStr
+			} else {
+				item.Status = deriveTemperatureStatus(item.Celsius, item.ThresholdWarningCelsius, item.ThresholdCriticalCelsius)
+			}
+			result.Temperatures = append(result.Temperatures, item)
+
+		case strings.EqualFold(unit, "Volts") || strings.EqualFold(unit, "V"):
+			if duplicateSensor(seen, "power", name) {
+				continue
+			}
+			item := schema.HardwarePowerSensor{Name: name, VoltageV: &value}
+			if statusStr != "" {
+				item.Status = &statusStr
+			}
+			result.Power = append(result.Power, item)
+
+		case strings.EqualFold(unit, "Watts") || strings.EqualFold(unit, "W"):
+			if duplicateSensor(seen, "power", name) {
+				continue
+			}
+			item := schema.HardwarePowerSensor{Name: name, PowerW: &value}
+			if statusStr != "" {
+				item.Status = &statusStr
+			}
+			result.Power = append(result.Power, item)
+
+		case strings.EqualFold(unit, "Amps") || strings.EqualFold(unit, "A"):
+			if duplicateSensor(seen, "power", name) {
+				continue
+			}
+			item := schema.HardwarePowerSensor{Name: name, CurrentA: &value}
+			if statusStr != "" {
+				item.Status = &statusStr
+			}
+			result.Power = append(result.Power, item)
+
+		default:
+			if duplicateSensor(seen, "other", name) {
+				continue
+			}
+			item := schema.HardwareOtherSensor{Name: name, Value: &value}
+			if unit != "" {
+				item.Unit = &unit
+			}
+			if statusStr != "" {
+				item.Status = &statusStr
+			}
+			result.Other = append(result.Other, item)
+		}
+	}
+
+	if len(result.Fans) == 0 && len(result.Temperatures) == 0 && len(result.Power) == 0 && len(result.Other) == 0 {
+		return nil
+	}
+	return result
+}
+
+func parseIPMIThreshold(raw string) *float64 {
+	s := strings.TrimSpace(raw)
+	if s == "" || s == "na" || s == "N/A" {
+		return nil
+	}
+	v, err := strconv.ParseFloat(s, 64)
+	if err != nil {
+		return nil
+	}
+	return &v
+}
+
+func normalizeIPMISensorStatus(s string) string {
+	switch strings.ToLower(s) {
+	case "ok":
+		return statusOK
+	case "cr", "ucr", "lcr":
+		return statusCritical
+	case "nc", "unc", "lnc", "nr", "unr", "lnr":
+		return statusWarning
+	case "ns", "na":
+		return ""
+	default:
+		return ""
+	}
+}
+
+// mergeIPMISensors appends IPMI sensor entries into existing, skipping names already present.
+func mergeIPMISensors(existing, ipmi *schema.HardwareSensors) *schema.HardwareSensors {
+	if ipmi == nil {
+		return existing
+	}
+	if existing == nil {
+		return ipmi
+	}
+
+	existingNames := map[string]struct{}{}
+	for _, s := range existing.Fans {
+		existingNames["fan\x00"+s.Name] = struct{}{}
+	}
+	for _, s := range existing.Temperatures {
+		existingNames["temp\x00"+s.Name] = struct{}{}
+	}
+	for _, s := range existing.Power {
+		existingNames["power\x00"+s.Name] = struct{}{}
+	}
+	for _, s := range existing.Other {
+		existingNames["other\x00"+s.Name] = struct{}{}
+	}
+
+	for _, s := range ipmi.Fans {
+		if _, ok := existingNames["fan\x00"+s.Name]; !ok {
+			existing.Fans = append(existing.Fans, s)
+		}
+	}
+	for _, s := range ipmi.Temperatures {
+		if _, ok := existingNames["temp\x00"+s.Name]; !ok {
+			existing.Temperatures = append(existing.Temperatures, s)
+		}
+	}
+	for _, s := range ipmi.Power {
+		if _, ok := existingNames["power\x00"+s.Name]; !ok {
+			existing.Power = append(existing.Power, s)
+		}
+	}
+	for _, s := range ipmi.Other {
+		if _, ok := existingNames["other\x00"+s.Name]; !ok {
+			existing.Other = append(existing.Other, s)
+		}
+	}
+	return existing
+}

audit/internal/collector/raid.go

@@ -766,7 +766,7 @@ func parseMDAdmPlatformLicense(raw string) *string {
 
 func queryDeviceSerial(devPath string) string {
 	if out, err := exec.Command("nvme", "id-ctrl", devPath, "-o", "json").Output(); err == nil {
-		var ctrl nvmeIDCtrl
+		var ctrl NVMeIDCtrl
 		if json.Unmarshal(out, &ctrl) == nil {
 			if v := cleanDMIValue(strings.TrimSpace(ctrl.SerialNumber)); v != "" {
 				return v

audit/internal/collector/storage.go

@@ -84,16 +84,19 @@ func collectStorage() []schema.HardwareStorage {
 	return result
 }
 
-// jsonInt64 accepts both a bare JSON number and a JSON-quoted number string.
-// lsblk -J emits LOG-SEC / PHY-SEC as integers on util-linux ≥ 2.37 (Debian 12)
-// but older versions emit them as strings. This type handles both.
-type jsonInt64 int64
+// JSONInt64 accepts a bare JSON number (512), a JSON-quoted number string
+// ("512" — lsblk -J on util-linux < 2.37, and nvme-cli for large 64-bit
+// counters that would lose precision as JS numbers), or a {"lo":n,"hi":n}
+// object (128-bit NVMe counters on some nvme-cli versions; hi is ignored as
+// no real counter exceeds 64 bits). Shared by lsblk and nvme-cli JSON output
+// across the collector and the human-readable disk report.
+type JSONInt64 int64
 
-func (j *jsonInt64) UnmarshalJSON(data []byte) error {
+func (j *JSONInt64) UnmarshalJSON(data []byte) error {
 	// bare number: 512
 	var n int64
 	if err := json.Unmarshal(data, &n); err == nil {
-		*j = jsonInt64(n)
+		*j = JSONInt64(n)
 		return nil
 	}
 	// quoted string: "512"
@@ -101,24 +104,32 @@ func (j *jsonInt64) UnmarshalJSON(data []byte) error {
 	if err := json.Unmarshal(data, &s); err == nil {
 		n, err := strconv.ParseInt(strings.TrimSpace(s), 10, 64)
 		if err == nil {
-			*j = jsonInt64(n)
+			*j = JSONInt64(n)
 		}
 		return nil
 	}
+	// {"lo":n,"hi":n} 128-bit counter object
+	var obj struct {
+		Lo int64 `json:"lo"`
+	}
+	if err := json.Unmarshal(data, &obj); err == nil {
+		*j = JSONInt64(obj.Lo)
+		return nil
+	}
 	return nil // null or unexpected type — leave zero
 }
 
 // lsblkDevice is a minimal lsblk JSON record.
 type lsblkDevice struct {
-	Name   string   `json:"name"`
-	Type   string   `json:"type"`
-	Size   string   `json:"size"`
-	Serial string   `json:"serial"`
-	Model  string   `json:"model"`
-	Tran   string   `json:"tran"`
-	Hctl   string   `json:"hctl"`
-	LogSec jsonInt64 `json:"log-sec"`
-	PhySec jsonInt64 `json:"phy-sec"`
+	Name   string    `json:"name"`
+	Type   string    `json:"type"`
+	Size   string    `json:"size"`
+	Serial string    `json:"serial"`
+	Model  string    `json:"model"`
+	Tran   string    `json:"tran"`
+	Hctl   string    `json:"hctl"`
+	LogSec JSONInt64 `json:"log-sec"`
+	PhySec JSONInt64 `json:"phy-sec"`
 }
 
 type lsblkRoot struct {
@@ -423,32 +434,36 @@ func enrichWithSmartctl(dev lsblkDevice) schema.HardwareStorage {
 	return s
 }
 
-// nvmeSmartLog is the subset of `nvme smart-log -o json` output we care about.
-// nvme-cli emits most counters as JSON strings (e.g. "power_on_hours":"49"),
-// so all numeric fields use jsonInt64 which accepts both bare numbers and
-// quoted strings. Field names match nvme-cli JSON output, not NVMe spec prose.
-type nvmeSmartLog struct {
-	CriticalWarning  jsonInt64 `json:"critical_warning"`
-	PercentageUsed   jsonInt64 `json:"percent_used"`
-	AvailableSpare   jsonInt64 `json:"avail_spare"`
-	SpareThreshold   jsonInt64 `json:"spare_thresh"`
-	Temperature      jsonInt64 `json:"temperature"`
-	PowerOnHours     jsonInt64 `json:"power_on_hours"`
-	PowerCycles      jsonInt64 `json:"power_cycles"`
-	UnsafeShutdowns  jsonInt64 `json:"unsafe_shutdowns"`
-	DataUnitsRead    jsonInt64 `json:"data_units_read"`
-	DataUnitsWritten jsonInt64 `json:"data_units_written"`
-	ControllerBusy   jsonInt64 `json:"controller_busy_time"`
-	MediaErrors      jsonInt64 `json:"media_errors"`
-	NumErrLogEntries jsonInt64 `json:"num_err_log_entries"`
+// NVMeSmartLog is the subset of `nvme smart-log -o json` output shared by the
+// structured collector and the human-readable disk report. nvme-cli emits
+// most counters as JSON strings (e.g. "power_on_hours":"49") or, on some
+// versions, as {"lo":n,"hi":n} objects — all numeric fields use JSONInt64,
+// which accepts bare numbers, quoted strings, and lo/hi objects. Field names
+// match nvme-cli JSON output, not NVMe spec prose.
+type NVMeSmartLog struct {
+	CriticalWarning  JSONInt64 `json:"critical_warning"`
+	PercentageUsed   JSONInt64 `json:"percent_used"`
+	AvailableSpare   JSONInt64 `json:"avail_spare"`
+	SpareThreshold   JSONInt64 `json:"spare_thresh"`
+	Temperature      JSONInt64 `json:"temperature"`
+	PowerOnHours     JSONInt64 `json:"power_on_hours"`
+	PowerCycles      JSONInt64 `json:"power_cycles"`
+	UnsafeShutdowns  JSONInt64 `json:"unsafe_shutdowns"`
+	DataUnitsRead    JSONInt64 `json:"data_units_read"`
+	DataUnitsWritten JSONInt64 `json:"data_units_written"`
+	ControllerBusy   JSONInt64 `json:"controller_busy_time"`
+	MediaErrors      JSONInt64 `json:"media_errors"`
+	NumErrLogEntries JSONInt64 `json:"num_err_log_entries"`
 }
 
-// nvmeIDCtrl is the subset of `nvme id-ctrl -o json` output.
-type nvmeIDCtrl struct {
-	ModelNumber   string `json:"mn"`
-	SerialNumber  string `json:"sn"`
-	FirmwareRev   string `json:"fr"`
-	TotalCapacity int64  `json:"tnvmcap"`
+// NVMeIDCtrl is the subset of `nvme id-ctrl -o json` output shared by the
+// structured collector and the human-readable disk report.
+type NVMeIDCtrl struct {
+	ModelNumber   string    `json:"mn"`
+	SerialNumber  string    `json:"sn"`
+	FirmwareRev   string    `json:"fr"`
+	TotalCapacity JSONInt64 `json:"tnvmcap"`
+	NVMCapacity   JSONInt64 `json:"nvmcap"`
 }
 
 func enrichWithNVMe(dev lsblkDevice) schema.HardwareStorage {
@@ -481,7 +496,7 @@ func enrichWithNVMe(dev lsblkDevice) schema.HardwareStorage {
 
 	// id-ctrl: model, serial, firmware, capacity
 	if out, err := exec.Command("nvme", "id-ctrl", devPath, "-o", "json").Output(); err == nil {
-		var ctrl nvmeIDCtrl
+		var ctrl NVMeIDCtrl
 		if json.Unmarshal(out, &ctrl) == nil {
 			if v := cleanDMIValue(strings.TrimSpace(ctrl.ModelNumber)); v != "" {
 				s.Model = &v
@@ -502,7 +517,7 @@ func enrichWithNVMe(dev lsblkDevice) schema.HardwareStorage {
 
 	// smart-log: wear telemetry
 	if out, err := exec.Command("nvme", "smart-log", devPath, "-o", "json").Output(); err == nil {
-		var log nvmeSmartLog
+		var log NVMeSmartLog
 		if json.Unmarshal(out, &log) == nil {
 			if log.PowerOnHours > 0 {
 				v := int64(log.PowerOnHours)

audit/internal/collector/storage_discovery_test.go

@@ -56,7 +56,7 @@ func TestJsonInt64UnmarshalBothFormats(t *testing.T) {
 		{`null`, 0},
 	}
 	for _, tc := range cases {
-		var v jsonInt64
+		var v JSONInt64
 		if err := v.UnmarshalJSON([]byte(tc.json)); err != nil {
 			t.Fatalf("UnmarshalJSON(%s): unexpected error %v", tc.json, err)
 		}

audit/internal/collector/storage_health_test.go

@@ -9,7 +9,7 @@ import (
 
 // TestNVMeSmartLogUnmarshal verifies that nvme-cli JSON output (where most
 // counters are quoted strings and field names differ from NVMe spec prose)
-// is correctly parsed into nvmeSmartLog.
+// is correctly parsed into NVMeSmartLog.
 func TestNVMeSmartLogUnmarshal(t *testing.T) {
 	t.Parallel()
 
@@ -30,7 +30,7 @@ func TestNVMeSmartLogUnmarshal(t *testing.T) {
 		"media_errors": "0",
 		"num_err_log_entries": "0"
 	}`
-	var log nvmeSmartLog
+	var log NVMeSmartLog
 	if err := json.Unmarshal([]byte(raw), &log); err != nil {
 		t.Fatalf("json.Unmarshal failed: %v", err)
 	}

audit/internal/platform/benchmark.go

@@ -4008,14 +4008,23 @@ func renderPowerBenchReport(result NvidiaPowerBenchResult) string {
 			idleW = result.ServerPower.IdleW
 		}
 
-		// Build header: Run | GPU 0 | GPU 1 | ... | Server wall W | Per GPU wall W | Platform eff.
+		// Build header: Run | GPU 0 | GPU 1 | ... | GPU total W | Server itself W | Server wall W | Per GPU wall W | Platform eff.
 		headers := []string{"Run"}
 		for _, idx := range allGPUIndices {
 			headers = append(headers, fmt.Sprintf("GPU %d W", idx))
 		}
-		headers = append(headers, "Server wall W", "Per GPU wall W", "Platform eff.")
+		headers = append(headers, "GPU total W", "Server itself W", "Server wall W", "Per GPU wall W", "Platform eff.")
 
 		var rampRows [][]string
+		if idleW > 0 {
+			idleRow := []string{"0 (idle)"}
+			for range allGPUIndices {
+				idleRow = append(idleRow, "—")
+			}
+			// No load: GPU total is negligible, all draw is the server's own baseline.
+			idleRow = append(idleRow, "—", fmt.Sprintf("%.0f", idleW), fmt.Sprintf("%.0f", idleW), "—", "—")
+			rampRows = append(rampRows, idleRow)
+		}
 		for _, step := range result.RampSteps {
 			row := []string{fmt.Sprintf("%d", step.StepIndex)}
 			for _, idx := range allGPUIndices {
@@ -4036,6 +4045,16 @@ func renderPowerBenchReport(result NvidiaPowerBenchResult) string {
 				}
 				row = append(row, gpuPwr)
 			}
+			// GPU total W = sum of observed GPU power (nvidia-smi)
+			gpuTotal := "—"
+			if step.TotalObservedPowerW > 0 {
+				gpuTotal = fmt.Sprintf("%.0f", step.TotalObservedPowerW)
+			}
+			// Server itself W = server wall power minus GPU total (non-GPU baseline draw)
+			serverItself := "—"
+			if step.ServerLoadedW > 0 && step.TotalObservedPowerW > 0 {
+				serverItself = fmt.Sprintf("%.0f", step.ServerLoadedW-step.TotalObservedPowerW)
+			}
 			// Server wall W
 			serverWall := "—"
 			if step.ServerLoadedW > 0 {
@@ -4055,7 +4074,7 @@ func renderPowerBenchReport(result NvidiaPowerBenchResult) string {
 				}
 				platEff = fmt.Sprintf("%.2f", eff)
 			}
-			row = append(row, serverWall, perGPUWall, platEff)
+			row = append(row, gpuTotal, serverItself, serverWall, perGPUWall, platEff)
 			rampRows = append(rampRows, row)
 		}
 		b.WriteString(fmtMDTable(headers, rampRows))
@@ -4617,6 +4636,8 @@ func (s *System) RunNvidiaPowerBench(ctx context.Context, baseDir string, opts N
 			ramp.AvgFanRPM = singleRun.AvgFanRPM
 			ramp.AvgFanDutyCyclePct = singleRun.AvgFanDutyCyclePct
 		}
+		firstSummary := firstCalib.Summary
+		ramp.PerGPUTelemetry = map[int]*BenchmarkTelemetrySummary{firstIdx: &firstSummary}
 		if !firstCalib.Completed {
 			ramp.Status = "FAILED"
 			ramp.Notes = append(ramp.Notes, fmt.Sprintf("GPU %d did not complete single-card %s", firstIdx, benchmarkPowerEngineLabel(benchmarkPowerEngine())))

audit/internal/platform/confidential_compute.go

@@ -0,0 +1,248 @@
+package platform
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+)
+
+// ConfidentialComputingStatus summarizes whether this server can run NVIDIA
+// Confidential Computing: CPU-side TEE support (Intel TDX / AMD SEV-SNP) and
+// GPU firmware CC capability, as reported by `nvidia-smi conf-compute -q`.
+type ConfidentialComputingStatus struct {
+	CollectedAt time.Time `json:"collected_at"`
+
+	// GPU-reported fields, parsed from `nvidia-smi conf-compute -q`.
+	NvidiaSMIAvailable bool   `json:"nvidia_smi_available"`
+	CCState            string `json:"cc_state,omitempty"`             // ON / OFF
+	MultiGPUMode       string `json:"multi_gpu_mode,omitempty"`       // Protected PCIe / ...
+	CPUCCCapability    string `json:"cpu_cc_capability,omitempty"`    // e.g. "INTEL TDX", "AMD SEV-SNP", "NONE"
+	GPUCCCapability    string `json:"gpu_cc_capability,omitempty"`    // e.g. "CC Capable", "Not Capable"
+	CCGPUsReadyState   string `json:"cc_gpus_ready_state,omitempty"`  // Ready / Not Ready
+
+	// Host-side evidence that the CPU's TEE is actually active in the running
+	// kernel (BIOS + kernel cmdline + firmware), independent of what the GPU
+	// driver reports. Used as a fallback when the NVIDIA driver isn't loaded.
+	HostAMDSEVSupported   bool `json:"host_amd_sev_supported"`
+	HostAMDSEVESSupported bool `json:"host_amd_sev_es_supported"`
+	HostAMDSEVSNPActive   bool `json:"host_amd_sev_snp_active"`
+	HostIntelTDXActive    bool `json:"host_intel_tdx_active"`
+
+	// GPUCanRunCC is true when the GPU firmware reports CC-capable.
+	GPUCanRunCC bool `json:"gpu_can_run_cc"`
+	// CPUCanRunCC is true when either the GPU driver or the host kernel
+	// reports an active/available CPU TEE (SEV-SNP or TDX).
+	CPUCanRunCC bool `json:"cpu_can_run_cc"`
+	// Ready is true when both the CPU and the GPU support Confidential
+	// Computing, regardless of whether CC mode is currently enabled.
+	Ready bool `json:"ready"`
+
+	Notes []string `json:"notes,omitempty"`
+}
+
+// RunConfidentialComputingCheckPack runs a read-only check of whether this
+// server can run NVIDIA Confidential Computing: it queries the GPU driver
+// (`nvidia-smi conf-compute -q`) and inspects host kernel/dmesg evidence of
+// AMD SEV-SNP / Intel TDX support. It changes nothing on the system.
+func (s *System) RunConfidentialComputingCheckPack(ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
+	if ctx == nil {
+		ctx = context.Background()
+	}
+	if baseDir == "" {
+		baseDir = "/var/log/bee-sat"
+	}
+	ts := time.Now().UTC().Format("20060102-150405")
+	runDir := filepath.Join(baseDir, "confidential-computing-"+ts)
+	if err := os.MkdirAll(runDir, 0755); err != nil {
+		return "", err
+	}
+	verboseLog := filepath.Join(runDir, "verbose.log")
+
+	status := ConfidentialComputingStatus{CollectedAt: time.Now().UTC()}
+
+	// GPU firmware / driver state.
+	ccOut, ccErr := runSATCommandCtx(ctx, verboseLog, "nvidia-smi-conf-compute-q", []string{"nvidia-smi", "conf-compute", "-q"}, nil, logFunc)
+	_ = os.WriteFile(filepath.Join(runDir, "01-nvidia-smi-conf-compute-q.log"), ccOut, 0644)
+	if ccErr == nil {
+		status.NvidiaSMIAvailable = true
+		fields := parseConfComputeFields(ccOut)
+		status.CCState = fields["CC State"]
+		status.MultiGPUMode = fields["Multi-GPU Mode"]
+		status.CPUCCCapability = fields["CPU CC Capabilities"]
+		status.GPUCCCapability = fields["GPU CC Capabilities"]
+		status.CCGPUsReadyState = fields["CC GPUs Ready State"]
+	} else {
+		status.Notes = append(status.Notes, "nvidia-smi conf-compute -q unavailable (no NVIDIA driver, or GPU not present): "+firstLine(string(ccOut)))
+	}
+
+	// Host kernel evidence, independent of the GPU driver.
+	dmesgOut, _ := runSATCommandCtx(ctx, verboseLog, "dmesg", []string{"dmesg"}, nil, nil)
+	ccDmesgLines := filterConfComputeDmesgLines(dmesgOut)
+	_ = os.WriteFile(filepath.Join(runDir, "02-dmesg-cc-relevant.log"), []byte(strings.Join(ccDmesgLines, "\n")+"\n"), 0644)
+
+	lowerDmesg := strings.ToLower(strings.Join(ccDmesgLines, "\n"))
+	status.HostAMDSEVSNPActive = strings.Contains(lowerDmesg, "sev-snp enabled")
+	status.HostIntelTDXActive = strings.Contains(lowerDmesg, "tdx module") && strings.Contains(lowerDmesg, "module initialized") ||
+		strings.Contains(lowerDmesg, "virt/tdx: module initialized")
+
+	for i, path := range []string{
+		"/sys/module/kvm_amd/parameters/sev",
+		"/sys/module/kvm_amd/parameters/sev_es",
+		"/sys/module/kvm_amd/parameters/sev_snp",
+	} {
+		name := fmt.Sprintf("sysfs-%s", filepath.Base(path))
+		out, err := runSATCommandCtx(ctx, verboseLog, name, []string{"cat", path}, nil, nil)
+		_ = os.WriteFile(filepath.Join(runDir, fmt.Sprintf("03-%02d-%s.log", i+1, name)), out, 0644)
+		if err != nil {
+			continue
+		}
+		val := strings.TrimSpace(string(out))
+		switch filepath.Base(path) {
+		case "sev":
+			status.HostAMDSEVSupported = strings.EqualFold(val, "Y")
+		case "sev_es":
+			status.HostAMDSEVESSupported = strings.EqualFold(val, "Y")
+		case "sev_snp":
+			if strings.EqualFold(val, "Y") {
+				status.HostAMDSEVSNPActive = true
+			}
+		}
+	}
+
+	status.GPUCanRunCC = strings.EqualFold(strings.TrimSpace(status.GPUCCCapability), "CC Capable")
+	cpuCapReported := strings.TrimSpace(status.CPUCCCapability)
+	status.CPUCanRunCC = status.HostAMDSEVSNPActive || status.HostIntelTDXActive ||
+		(cpuCapReported != "" && !strings.EqualFold(cpuCapReported, "NONE"))
+	status.Ready = status.CPUCanRunCC && status.GPUCanRunCC
+
+	if !status.NvidiaSMIAvailable {
+		status.Notes = append(status.Notes, "GPU CC capability unknown — install the NVIDIA driver to query it with `nvidia-smi conf-compute -q`.")
+	}
+
+	summary := renderConfidentialComputingSummary(status)
+	if err := os.WriteFile(filepath.Join(runDir, "summary.txt"), []byte(summary), 0644); err != nil {
+		return "", err
+	}
+	report := renderConfidentialComputingReport(status)
+	if err := os.WriteFile(filepath.Join(runDir, "confidential-computing-report.txt"), []byte(report), 0644); err != nil {
+		return "", err
+	}
+
+	return runDir, nil
+}
+
+// parseConfComputeFields parses the indented "Key : Value" block emitted by
+// `nvidia-smi conf-compute -q`, e.g.:
+//
+//	CC State                : OFF
+//	Multi-GPU Mode          : Protected PCIe
+//	CPU CC Capabilities     : INTEL TDX
+//	GPU CC Capabilities     : CC Capable
+//	CC GPUs Ready State     : Not Ready
+func parseConfComputeFields(out []byte) map[string]string {
+	fields := map[string]string{}
+	for _, line := range strings.Split(string(out), "\n") {
+		idx := strings.Index(line, ":")
+		if idx < 0 {
+			continue
+		}
+		key := strings.TrimSpace(line[:idx])
+		val := strings.TrimSpace(line[idx+1:])
+		if key == "" || val == "" {
+			continue
+		}
+		fields[key] = val
+	}
+	return fields
+}
+
+// filterConfComputeDmesgLines returns the dmesg lines relevant to CPU
+// Confidential Computing support (AMD SEV/SEV-ES/SEV-SNP, Intel TDX).
+func filterConfComputeDmesgLines(dmesgOut []byte) []string {
+	var lines []string
+	scanner := bytes.Split(dmesgOut, []byte("\n"))
+	for _, raw := range scanner {
+		lower := strings.ToLower(string(raw))
+		if strings.Contains(lower, "sev") || strings.Contains(lower, "tdx") {
+			lines = append(lines, string(raw))
+		}
+	}
+	return lines
+}
+
+func renderConfidentialComputingSummary(status ConfidentialComputingStatus) string {
+	var b strings.Builder
+	fmt.Fprintf(&b, "run_at_utc=%s\n", status.CollectedAt.Format(time.RFC3339))
+	fmt.Fprintf(&b, "nvidia_smi_available=%t\n", status.NvidiaSMIAvailable)
+	fmt.Fprintf(&b, "cc_state=%s\n", status.CCState)
+	fmt.Fprintf(&b, "multi_gpu_mode=%s\n", status.MultiGPUMode)
+	fmt.Fprintf(&b, "cpu_cc_capability=%s\n", status.CPUCCCapability)
+	fmt.Fprintf(&b, "gpu_cc_capability=%s\n", status.GPUCCCapability)
+	fmt.Fprintf(&b, "cc_gpus_ready_state=%s\n", status.CCGPUsReadyState)
+	fmt.Fprintf(&b, "host_amd_sev_supported=%t\n", status.HostAMDSEVSupported)
+	fmt.Fprintf(&b, "host_amd_sev_es_supported=%t\n", status.HostAMDSEVESSupported)
+	fmt.Fprintf(&b, "host_amd_sev_snp_active=%t\n", status.HostAMDSEVSNPActive)
+	fmt.Fprintf(&b, "host_intel_tdx_active=%t\n", status.HostIntelTDXActive)
+	fmt.Fprintf(&b, "cpu_can_run_cc=%t\n", status.CPUCanRunCC)
+	fmt.Fprintf(&b, "gpu_can_run_cc=%t\n", status.GPUCanRunCC)
+	fmt.Fprintf(&b, "ready=%t\n", status.Ready)
+	if status.Ready {
+		fmt.Fprintln(&b, "overall_status=OK")
+	} else {
+		fmt.Fprintln(&b, "overall_status=NOT_READY")
+	}
+	return b.String()
+}
+
+func renderConfidentialComputingReport(status ConfidentialComputingStatus) string {
+	var b strings.Builder
+	line := strings.Repeat("=", 80)
+	b.WriteString(line + "\n")
+	b.WriteString("Confidential Computing Readiness\n")
+	b.WriteString(line + "\n\n")
+
+	verdict := "NOT READY"
+	if status.Ready {
+		verdict = "READY"
+	}
+	fmt.Fprintf(&b, "Verdict: %s\n\n", verdict)
+
+	b.WriteString("-- CPU ----------------------------------------------------------------------\n")
+	fmt.Fprintf(&b, "  Reported by GPU driver : %s\n", nonEmptyOr(status.CPUCCCapability, "unknown"))
+	fmt.Fprintf(&b, "  AMD SEV supported      : %t\n", status.HostAMDSEVSupported)
+	fmt.Fprintf(&b, "  AMD SEV-ES supported    : %t\n", status.HostAMDSEVESSupported)
+	fmt.Fprintf(&b, "  AMD SEV-SNP active      : %t\n", status.HostAMDSEVSNPActive)
+	fmt.Fprintf(&b, "  Intel TDX active        : %t\n", status.HostIntelTDXActive)
+	fmt.Fprintf(&b, "  Can run CC              : %t\n\n", status.CPUCanRunCC)
+
+	b.WriteString("-- GPU ----------------------------------------------------------------------\n")
+	fmt.Fprintf(&b, "  nvidia-smi available    : %t\n", status.NvidiaSMIAvailable)
+	fmt.Fprintf(&b, "  GPU CC Capabilities     : %s\n", nonEmptyOr(status.GPUCCCapability, "unknown"))
+	fmt.Fprintf(&b, "  CC State (current)      : %s\n", nonEmptyOr(status.CCState, "unknown"))
+	fmt.Fprintf(&b, "  Multi-GPU Mode          : %s\n", nonEmptyOr(status.MultiGPUMode, "unknown"))
+	fmt.Fprintf(&b, "  CC GPUs Ready State     : %s\n", nonEmptyOr(status.CCGPUsReadyState, "unknown"))
+	fmt.Fprintf(&b, "  Can run CC              : %t\n\n", status.GPUCanRunCC)
+
+	if len(status.Notes) > 0 {
+		b.WriteString("-- Notes ----------------------------------------------------------------------\n")
+		for _, n := range status.Notes {
+			fmt.Fprintf(&b, "  - %s\n", n)
+		}
+		b.WriteString("\n")
+	}
+
+	fmt.Fprintf(&b, "Collected : %s\n", status.CollectedAt.Format("2006-01-02 15:04:05 UTC"))
+	b.WriteString(line + "\n")
+	return b.String()
+}
+
+func nonEmptyOr(v, fallback string) string {
+	if strings.TrimSpace(v) == "" {
+		return fallback
+	}
+	return v
+}

audit/internal/platform/sat.go

@@ -182,9 +182,16 @@ func (s *System) DetectGPUVendor() string {
 		return "amd"
 	}
 	if raw, err := exec.Command("lspci", "-nn").Output(); err == nil {
-		text := strings.ToLower(string(raw))
-		if strings.Contains(text, "advanced micro devices") || strings.Contains(text, "amd/ati") {
-			return "amd"
+		// Only match AMD GPU device classes [0300]=VGA, [0302]=3D controller, [0380]=Display.
+		// AMD CPUs also appear in lspci as "Advanced Micro Devices" (Root Complex, IOMMU, etc.)
+		// so matching vendor alone causes false positives on AMD CPU servers without GPUs.
+		for _, line := range strings.Split(strings.ToLower(string(raw)), "\n") {
+			if !strings.Contains(line, "advanced micro devices") && !strings.Contains(line, "amd/ati") {
+				continue
+			}
+			if strings.Contains(line, "[0300]") || strings.Contains(line, "[0302]") || strings.Contains(line, "[0380]") {
+				return "amd"
+			}
 		}
 	}
 	return ""
@@ -723,12 +730,14 @@ func (s *System) RunStorageAcceptancePack(ctx context.Context, baseDir string, e
 		}
 		prefix := fmt.Sprintf("%02d-%s", index+1, filepath.Base(devPath))
 		commands := storageSATCommands(devPath, extended)
+		deviceOutputs := make(map[string][]byte, len(commands))
 		for cmdIndex, job := range commands {
 			if ctx.Err() != nil {
 				break
 			}
 			name := fmt.Sprintf("%s-%02d-%s.log", prefix, cmdIndex+1, job.name)
 			out, err := runSATCommandCtx(ctx, verboseLog, job.name, job.cmd, nil, logFunc)
+			deviceOutputs[job.name] = out
 			if writeErr := os.WriteFile(filepath.Join(runDir, name), out, 0644); writeErr != nil {
 				return "", writeErr
 			}
@@ -737,7 +746,28 @@ func (s *System) RunStorageAcceptancePack(ctx context.Context, baseDir string, e
 			key := filepath.Base(devPath) + "_" + strings.ReplaceAll(job.name, "-", "_")
 			fmt.Fprintf(&summary, "%s_rc=%d\n", key, rc)
 			fmt.Fprintf(&summary, "%s_status=%s\n", key, status)
+
+			// smartctl -t short only launches the self-test on the drive firmware and
+			// returns immediately ("Testing has begun"); unlike `nvme device-self-test
+			// --wait`, smartctl has no blocking mode, so we must poll the drive
+			// ourselves until the self-test actually finishes.
+			if job.name == "smartctl-self-test-short" && err == nil {
+				statusName := "smartctl-self-test-status"
+				statusOut := waitForSmartctlSelfTest(ctx, verboseLog, devPath, logFunc)
+				deviceOutputs[statusName] = statusOut
+				statusFile := fmt.Sprintf("%s-%02d-%s.log", prefix, cmdIndex+2, statusName)
+				if writeErr := os.WriteFile(filepath.Join(runDir, statusFile), statusOut, 0644); writeErr != nil {
+					return "", writeErr
+				}
+				sStatus, sRC := classifySATResult(statusName, statusOut, nil)
+				stats.Add(sStatus)
+				sKey := filepath.Base(devPath) + "_" + strings.ReplaceAll(statusName, "-", "_")
+				fmt.Fprintf(&summary, "%s_rc=%d\n", sKey, sRC)
+				fmt.Fprintf(&summary, "%s_status=%s\n", sKey, sStatus)
+			}
 		}
+		reportText := GenerateDiskReportText(index+1, devPath, deviceOutputs, time.Now().UTC())
+		_ = os.WriteFile(filepath.Join(runDir, "disk-"+prefix+"-report.txt"), []byte(reportText), 0644)
 	}
 
 	writeSATStats(&summary, stats)
@@ -1170,6 +1200,42 @@ func runSATCommandCtx(ctx context.Context, verboseLog, name string, cmd []string
 	return out, err
 }
 
+// smartctlSelfTestPollInterval/Timeout bound how long we poll the drive after
+// launching `smartctl -t short`, which SMART/ATA specs put at ~2 minutes.
+const (
+	smartctlSelfTestPollInterval = 5 * time.Second
+	smartctlSelfTestTimeout      = 4 * time.Minute
+)
+
+// waitForSmartctlSelfTest polls `smartctl -a` until the short self-test
+// started on devPath finishes (or the timeout/context elapses) and returns
+// the final output, which reflects the actual test result rather than the
+// "Testing has begun" launch acknowledgement.
+func waitForSmartctlSelfTest(ctx context.Context, verboseLog, devPath string, logFunc func(string)) []byte {
+	deadline := time.Now().Add(smartctlSelfTestTimeout)
+	var last []byte
+	for {
+		out, _ := runSATCommandCtx(ctx, verboseLog, "smartctl-self-test-status", []string{"smartctl", "-a", devPath}, nil, nil)
+		last = out
+		if ctx.Err() != nil {
+			return last
+		}
+		lower := bytes.ToLower(out)
+		if !bytes.Contains(lower, []byte("self-test routine in progress")) &&
+			!bytes.Contains(lower, []byte("% of test remaining")) {
+			return last
+		}
+		if time.Now().After(deadline) {
+			return last
+		}
+		select {
+		case <-ctx.Done():
+			return last
+		case <-time.After(smartctlSelfTestPollInterval):
+		}
+	}
+}
+
 func listStorageDevices() ([]string, error) {
 	out, err := satExecCommand("lsblk", "-dn", "-o", "NAME,TYPE,TRAN").Output()
 	if err != nil {
@@ -1178,26 +1244,27 @@ func listStorageDevices() ([]string, error) {
 	return parseStorageDevices(string(out)), nil
 }
 
+// storageSATCommands returns the commands to run for a single storage device.
+// extended=false (Check): read-only SMART/NVMe data collection, no self-test.
+// extended=true (Load): data collection + short self-test.
 func storageSATCommands(devPath string, extended bool) []satJob {
 	if strings.Contains(filepath.Base(devPath), "nvme") {
-		selfTestLevel := "1"
-		if extended {
-			selfTestLevel = "2"
-		}
-		return []satJob{
+		jobs := []satJob{
 			{name: "nvme-id-ctrl", cmd: []string{"nvme", "id-ctrl", devPath, "-o", "json"}},
 			{name: "nvme-smart-log", cmd: []string{"nvme", "smart-log", devPath, "-o", "json"}},
-			{name: "nvme-device-self-test", cmd: []string{"nvme", "device-self-test", devPath, "-s", selfTestLevel, "--wait"}},
 		}
+		if extended {
+			jobs = append(jobs, satJob{name: "nvme-device-self-test", cmd: []string{"nvme", "device-self-test", devPath, "-s", "1", "--wait"}})
+		}
+		return jobs
+	}
+	jobs := []satJob{
+		{name: "smartctl-health", cmd: []string{"smartctl", "-H", "-A", "-i", devPath}},
 	}
-	smartTestType := "short"
 	if extended {
-		smartTestType = "long"
-	}
-	return []satJob{
-		{name: "smartctl-health", cmd: []string{"smartctl", "-H", "-A", devPath}},
-		{name: "smartctl-self-test-short", cmd: []string{"smartctl", "-t", smartTestType, devPath}},
+		jobs = append(jobs, satJob{name: "smartctl-self-test-short", cmd: []string{"smartctl", "-t", "short", devPath}})
 	}
+	return jobs
 }
 
 func (s *satStats) Add(status string) {

audit/internal/platform/sat_test.go

@@ -14,14 +14,42 @@ import (
 func TestStorageSATCommands(t *testing.T) {
 	t.Parallel()
 
-	nvme := storageSATCommands("/dev/nvme0n1", false)
-	if len(nvme) != 3 || nvme[2].cmd[0] != "nvme" {
-		t.Fatalf("unexpected nvme commands: %#v", nvme)
+	// Check mode (extended=false): read-only collection, no self-test.
+	nvmeCheck := storageSATCommands("/dev/nvme0n1", false)
+	if len(nvmeCheck) != 2 {
+		t.Fatalf("check nvme: want 2 commands, got %d: %#v", len(nvmeCheck), nvmeCheck)
+	}
+	if nvmeCheck[0].name != "nvme-id-ctrl" || nvmeCheck[1].name != "nvme-smart-log" {
+		t.Fatalf("check nvme: unexpected command names: %#v", nvmeCheck)
 	}
 
-	sata := storageSATCommands("/dev/sda", false)
-	if len(sata) != 2 || sata[0].cmd[0] != "smartctl" {
-		t.Fatalf("unexpected sata commands: %#v", sata)
+	sataCheck := storageSATCommands("/dev/sda", false)
+	if len(sataCheck) != 1 || sataCheck[0].cmd[0] != "smartctl" {
+		t.Fatalf("check sata: want 1 smartctl command, got %#v", sataCheck)
+	}
+
+	// Load mode (extended=true): collection + short self-test.
+	nvmeLoad := storageSATCommands("/dev/nvme0n1", true)
+	if len(nvmeLoad) != 3 || nvmeLoad[2].name != "nvme-device-self-test" {
+		t.Fatalf("load nvme: want 3 commands with self-test last, got %#v", nvmeLoad)
+	}
+	if got := nvmeLoad[2].cmd[len(nvmeLoad[2].cmd)-3]; got != "-s" {
+		t.Fatalf("load nvme: want -s flag, got %q", got)
+	}
+	if got := nvmeLoad[2].cmd[len(nvmeLoad[2].cmd)-2]; got != "1" {
+		t.Fatalf("load nvme: want self-test level 1, got %q", got)
+	}
+
+	sataLoad := storageSATCommands("/dev/sda", true)
+	if len(sataLoad) != 2 || sataLoad[1].name != "smartctl-self-test-short" {
+		t.Fatalf("load sata: want 2 commands with short self-test last, got %#v", sataLoad)
+	}
+	// cmd is: smartctl -t short /dev/sda
+	if got := sataLoad[1].cmd[1]; got != "-t" {
+		t.Fatalf("load sata: want -t flag at index 1, got %q", got)
+	}
+	if got := sataLoad[1].cmd[2]; got != "short" {
+		t.Fatalf("load sata: want short at index 2, got %q", got)
 	}
 }
 

audit/internal/platform/storage_report.go

@@ -0,0 +1,548 @@
+package platform
+
+import (
+	"bee/audit/internal/collector"
+	"encoding/json"
+	"fmt"
+	"math"
+	"path/filepath"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// GenerateDiskReportText builds a human-readable text report for one storage
+// device from the raw command outputs collected during storage SAT.
+//
+// outputs keys match satJob.name: "nvme-id-ctrl", "nvme-smart-log",
+// "smartctl-health", "smartctl-self-test-short".
+func GenerateDiskReportText(index int, devPath string, outputs map[string][]byte, ts time.Time) string {
+	var b strings.Builder
+	devName := filepath.Base(devPath)
+	line := strings.Repeat("=", 80)
+	b.WriteString(line + "\n")
+	fmt.Fprintf(&b, "Disk %-3d  %s\n", index, devPath)
+	b.WriteString(line + "\n")
+
+	isNVMe := strings.Contains(devName, "nvme")
+	if isNVMe {
+		writeNVMeReport(&b, outputs)
+	} else {
+		writeSATAReport(&b, outputs)
+	}
+
+	b.WriteString("\n")
+	fmt.Fprintf(&b, "Collected : %s\n", ts.UTC().Format("2006-01-02 15:04:05 UTC"))
+	b.WriteString(line + "\n")
+	return b.String()
+}
+
+// ── NVMe ─────────────────────────────────────────────────────────────────────
+
+func writeNVMeReport(b *strings.Builder, outputs map[string][]byte) {
+	// id-ctrl
+	var ctrl collector.NVMeIDCtrl
+	if data := outputs["nvme-id-ctrl"]; len(data) > 0 {
+		_ = json.Unmarshal(data, &ctrl)
+	}
+
+	model := strings.TrimSpace(ctrl.ModelNumber)
+	serial := strings.TrimSpace(ctrl.SerialNumber)
+	firmware := strings.TrimSpace(ctrl.FirmwareRev)
+
+	capacityGB := ""
+	if ctrl.TotalCapacity > 0 {
+		capacityGB = formatCapacityGB(uint64(ctrl.TotalCapacity))
+	} else if ctrl.NVMCapacity > 0 {
+		capacityGB = formatCapacityGB(uint64(ctrl.NVMCapacity))
+	}
+
+	writeField(b, "Model", model)
+	writeField(b, "Serial", serial)
+	writeField(b, "Firmware", firmware)
+	if capacityGB != "" {
+		writeField(b, "Capacity", capacityGB)
+	}
+
+	// smart-log
+	data := outputs["nvme-smart-log"]
+	if len(data) == 0 {
+		b.WriteString("\n(no SMART data)\n")
+		return
+	}
+	var sl collector.NVMeSmartLog
+	if err := json.Unmarshal(data, &sl); err != nil {
+		fmt.Fprintf(b, "\n(SMART parse error: %v)\n", err)
+		return
+	}
+
+	tempC := int(sl.Temperature) - 273
+	if tempC < 0 {
+		tempC = 0
+	}
+
+	critWarnStr := "OK"
+	if sl.CriticalWarning != 0 {
+		critWarnStr = fmt.Sprintf("0x%02X", sl.CriticalWarning)
+	}
+
+	poh := uint64(sl.PowerOnHours)
+	pc := uint64(sl.PowerCycles)
+	us := uint64(sl.UnsafeShutdowns)
+	me := uint64(sl.MediaErrors)
+	nel := uint64(sl.NumErrLogEntries)
+
+	// data_units are in 1000 × 512-byte sectors = 512,000 bytes each
+	readBytes := uint64(sl.DataUnitsRead) * 512000
+	writtenBytes := uint64(sl.DataUnitsWritten) * 512000
+
+	writeSectionHeader(b, "Health")
+	writeField(b, "Temperature", fmt.Sprintf("%d °C", tempC))
+	writeField(b, "Critical Warning", critWarnStr)
+	writeField(b, "Percentage Used", fmt.Sprintf("%d %%", sl.PercentageUsed))
+	writeField(b, "Available Spare", fmt.Sprintf("%d %%  (threshold: %d %%)", sl.AvailableSpare, sl.SpareThreshold))
+
+	writeSectionHeader(b, "Usage")
+	writeField(b, "Power On Hours", fmt.Sprintf("%s h", formatUint(poh)))
+	writeField(b, "Power Cycles", formatUint(pc))
+	writeField(b, "Unsafe Shutdowns", formatUint(us))
+	writeField(b, "Data Written", formatBytesHuman(float64(writtenBytes)))
+	writeField(b, "Data Read", formatBytesHuman(float64(readBytes)))
+
+	writeSectionHeader(b, "Errors")
+	writeField(b, "Media Errors", formatUint(me))
+	writeField(b, "Error Log Entries", formatUint(nel))
+
+	capacityBytes := uint64(ctrl.TotalCapacity)
+	if capacityBytes == 0 {
+		capacityBytes = uint64(ctrl.NVMCapacity)
+	}
+	ri := resourceInfo{
+		powerOnHours:  poh,
+		powerCycles:   pc,
+		writtenBytes:  writtenBytes,
+		readBytes:     readBytes,
+		capacityBytes: capacityBytes,
+	}
+	writeResourceSection(b, ri)
+
+	if selfTest := outputs["nvme-device-self-test"]; len(selfTest) > 0 {
+		writeSectionHeader(b, "Self-Test")
+		result := parseSelfTestResult(string(selfTest))
+		writeField(b, "Result", result)
+	}
+
+	writeConclusionSection(b, ri)
+}
+
+// ── SATA / SAS (smartctl) ────────────────────────────────────────────────────
+
+var (
+	smartHealthRE   = regexp.MustCompile(`(?i)SMART overall-health self-assessment test result:\s*(\S+)`)
+	smartAttrLineRE = regexp.MustCompile(
+		`^\s*(\d{1,3})\s+(\S+)\s+0x[0-9a-fA-F]+\s+(\d{1,3})\s+(\d{1,3})\s+(\d{1,3})\s+\S+\s+\S+\s+\S+\s+(.+?)\s*$`,
+	)
+	smartModelRE    = regexp.MustCompile(`(?im)^Device Model:\s*(.+)$`)
+	smartSerialRE   = regexp.MustCompile(`(?im)^Serial Number:\s*(.+)$`)
+	smartFirmwareRE = regexp.MustCompile(`(?im)^Firmware Version:\s*(.+)$`)
+	smartCapacityRE = regexp.MustCompile(`(?im)^User Capacity:\s*(.+)$`)
+)
+
+type smartAttr struct {
+	ID        int
+	Name      string
+	Value     int
+	Worst     int
+	Threshold int
+	Raw       string
+}
+
+func writeSATAReport(b *strings.Builder, outputs map[string][]byte) {
+	data := outputs["smartctl-health"]
+	if len(data) == 0 {
+		b.WriteString("\n(no SMART data)\n")
+		return
+	}
+	text := string(data)
+
+	// Identity
+	if m := smartModelRE.FindStringSubmatch(text); m != nil {
+		writeField(b, "Model", strings.TrimSpace(m[1]))
+	}
+	if m := smartSerialRE.FindStringSubmatch(text); m != nil {
+		writeField(b, "Serial", strings.TrimSpace(m[1]))
+	}
+	if m := smartFirmwareRE.FindStringSubmatch(text); m != nil {
+		writeField(b, "Firmware", strings.TrimSpace(m[1]))
+	}
+	var capacityBytes uint64
+	if m := smartCapacityRE.FindStringSubmatch(text); m != nil {
+		cap := strings.TrimSpace(m[1])
+		capacityBytes = parseLeadingUint(cap)
+		// trim everything after "[" if present (e.g. "500,107,862,016 bytes [500 GB]")
+		if idx := strings.Index(cap, "["); idx > 0 {
+			cap = strings.TrimSpace(cap[idx+1:])
+			cap = strings.TrimSuffix(cap, "]")
+		}
+		writeField(b, "Capacity", cap)
+	}
+
+	writeSectionHeader(b, "Health")
+	health := "unknown"
+	if m := smartHealthRE.FindStringSubmatch(text); m != nil {
+		health = strings.TrimSpace(m[1])
+	}
+	writeField(b, "SMART Overall Health", health)
+
+	attrs := parseSMARTAttrs(text)
+	if len(attrs) > 0 {
+		writeSectionHeader(b, "SMART Attributes")
+		fmt.Fprintf(b, "  %-4s  %-32s  %5s  %5s  %5s  %s\n", "ID", "Attribute", "Value", "Worst", "Thresh", "Raw")
+		b.WriteString("  " + strings.Repeat("-", 72) + "\n")
+		for _, a := range attrs {
+			fmt.Fprintf(b, "  %-4d  %-32s  %5d  %5d  %5d  %s\n",
+				a.ID, a.Name, a.Value, a.Worst, a.Threshold, a.Raw)
+		}
+	}
+
+	var poh, pc, writtenLBAs, readLBAs uint64
+	var readValue int
+	hasReadValue := false
+	for _, a := range attrs {
+		switch a.ID {
+		case 9: // Power_On_Hours
+			poh = parseLeadingUint(a.Raw)
+		case 12: // Power_Cycle_Count
+			pc = parseLeadingUint(a.Raw)
+		case 241: // Total_LBAs_Written
+			writtenLBAs = parseLeadingUint(a.Raw)
+		case 242: // Total_LBAs_Read
+			readLBAs = parseLeadingUint(a.Raw)
+			readValue = a.Value
+			hasReadValue = true
+		}
+	}
+	const sataSectorBytes = 512
+	ri := resourceInfo{
+		powerOnHours:   poh,
+		powerCycles:    pc,
+		writtenBytes:   writtenLBAs * sataSectorBytes,
+		readBytes:      readLBAs * sataSectorBytes,
+		capacityBytes:  capacityBytes,
+		readPercent:    100 - readValue,
+		hasReadPercent: hasReadValue,
+	}
+	writeResourceSection(b, ri)
+
+	selfTest := outputs["smartctl-self-test-status"]
+	if len(selfTest) == 0 {
+		selfTest = outputs["smartctl-self-test-short"]
+	}
+	if len(selfTest) > 0 {
+		writeSectionHeader(b, "Self-Test")
+		result := parseSelfTestResult(string(selfTest))
+		writeField(b, "Result", result)
+	}
+
+	writeConclusionSection(b, ri)
+}
+
+func parseSMARTAttrs(text string) []smartAttr {
+	var attrs []smartAttr
+	inTable := false
+	for _, line := range strings.Split(text, "\n") {
+		if strings.Contains(line, "ATTRIBUTE_NAME") {
+			inTable = true
+			continue
+		}
+		if !inTable {
+			continue
+		}
+		m := smartAttrLineRE.FindStringSubmatch(line)
+		if m == nil {
+			if strings.TrimSpace(line) == "" {
+				inTable = false
+			}
+			continue
+		}
+		id, _ := strconv.Atoi(m[1])
+		val, _ := strconv.Atoi(m[3])
+		worst, _ := strconv.Atoi(m[4])
+		thresh, _ := strconv.Atoi(m[5])
+		attrs = append(attrs, smartAttr{
+			ID:        id,
+			Name:      m[2],
+			Value:     val,
+			Worst:     worst,
+			Threshold: thresh,
+			Raw:       strings.TrimSpace(m[6]),
+		})
+	}
+	return attrs
+}
+
+// parseSelfTestResult extracts a one-line summary from nvme device-self-test,
+// smartctl -a (post-completion status), or smartctl -t short (launch ack) output.
+func parseSelfTestResult(text string) string {
+	text = strings.TrimSpace(text)
+	if text == "" {
+		return "no output"
+	}
+	lines := strings.Split(text, "\n")
+	// smartctl -a: "Self-test execution status:      (   0)\n\tThe previous
+	// self-test routine completed\n\twithout error ..." — the description
+	// wraps onto following indented, colon-free continuation lines.
+	for i, line := range lines {
+		if strings.Contains(strings.ToLower(line), "self-test execution status") {
+			parts := []string{strings.TrimSpace(line)}
+			for j := i + 1; j < len(lines) && j < i+4; j++ {
+				cont := strings.TrimSpace(lines[j])
+				if cont == "" || strings.Contains(cont, ":") {
+					break
+				}
+				parts = append(parts, cont)
+			}
+			return strings.Join(parts, " ")
+		}
+	}
+	// nvme device-self-test: look for "Short Device Self-Test Status : 0x0" or similar
+	for _, line := range lines {
+		l := strings.ToLower(line)
+		if strings.Contains(l, "self-test status") || strings.Contains(l, "self test status") {
+			return strings.TrimSpace(line)
+		}
+	}
+	// smartctl -t short: "Testing has begun" or "Short BGST started"
+	for _, line := range lines {
+		l := strings.ToLower(line)
+		if strings.Contains(l, "testing has begun") || strings.Contains(l, "started") || strings.Contains(l, "complete") {
+			return strings.TrimSpace(line)
+		}
+	}
+	// fallback: last non-empty line
+	for i := len(lines) - 1; i >= 0; i-- {
+		if s := strings.TrimSpace(lines[i]); s != "" {
+			return s
+		}
+	}
+	return "done"
+}
+
+// ── Resource (pseudographic usage bars) ────────────────────────────────────────
+
+// designLifeYears/dwpd model the drive's rated endurance: 1 drive-write-per-day
+// for 5 years, the baseline enterprise endurance spec used when the vendor's
+// own TBW/DWPD rating isn't available from SMART/NVMe data.
+const (
+	designLifeYears = 5
+	dwpd            = 1.0
+)
+
+type resourceInfo struct {
+	powerOnHours   uint64
+	powerCycles    uint64
+	writtenBytes   uint64
+	readBytes      uint64
+	capacityBytes  uint64
+	readPercent    int  // only meaningful when hasReadPercent
+	hasReadPercent bool // true when the source SMART attribute exposes a normalized read-wear value
+}
+
+func writeResourceSection(b *strings.Builder, r resourceInfo) {
+	writeSectionHeader(b, "Resource")
+
+	const maxLifeHours = designLifeYears * 365 * 24
+	upFrac := float64(r.powerOnHours) / float64(maxLifeHours)
+	fmt.Fprintf(b, "  %-9s %s  %s / %s (%s)\n",
+		"Uptime", progressBar(upFrac, 24), formatHoursHuman(r.powerOnHours), formatHoursHuman(maxLifeHours), formatPercent(upFrac*100))
+
+	if r.capacityBytes > 0 {
+		maxWritten := float64(r.capacityBytes) * dwpd * designLifeYears * 365
+		wFrac := float64(r.writtenBytes) / maxWritten
+		fmt.Fprintf(b, "  %-9s %s  %s / %s (%s, %g DWPD×%dy)\n",
+			"Written", progressBar(wFrac, 24), formatBytesHuman(float64(r.writtenBytes)), formatBytesHuman(maxWritten), formatPercent(wFrac*100), dwpd, designLifeYears)
+	} else {
+		fmt.Fprintf(b, "  %-9s %s\n", "Written", formatBytesHuman(float64(r.writtenBytes)))
+	}
+
+	if r.hasReadPercent {
+		fmt.Fprintf(b, "  %-9s %s  %s (%d%%)\n",
+			"Read", progressBar(float64(r.readPercent)/100, 24), formatBytesHuman(float64(r.readBytes)), r.readPercent)
+	} else {
+		fmt.Fprintf(b, "  %-9s %s\n", "Read", formatBytesHuman(float64(r.readBytes)))
+	}
+}
+
+// ── Conclusion (new-vs-used verdict) ────────────────────────────────────────
+
+// Thresholds for treating a drive as "new": less than one full drive-write
+// (110% of capacity, headroom for provisioning/overprovisioning rounding),
+// less than a bit over two full drive-reads (210% of capacity), under a
+// week of power-on time, and under 30 power cycles. Any one violation is
+// enough to call the drive used — these are deliberately loose bounds, not
+// a wear/endurance judgment (see -- Resource -- for that).
+const (
+	newDiskMaxWrittenFrac = 1.10
+	newDiskMaxReadFrac    = 2.10
+	newDiskMaxUptimeHours = 7 * 24
+	newDiskMaxPowerCycles = 30
+)
+
+func writeConclusionSection(b *strings.Builder, r resourceInfo) {
+	writeSectionHeader(b, "Conclusion")
+
+	var reasons, notes []string
+	isNew := true
+
+	if r.capacityBytes > 0 {
+		writtenFrac := float64(r.writtenBytes) / float64(r.capacityBytes)
+		readFrac := float64(r.readBytes) / float64(r.capacityBytes)
+		if writtenFrac >= newDiskMaxWrittenFrac {
+			isNew = false
+			reasons = append(reasons, fmt.Sprintf(
+				"data written %s (%s of capacity)",
+				formatBytesHuman(float64(r.writtenBytes)), formatPercent(writtenFrac*100)))
+		}
+		if readFrac >= newDiskMaxReadFrac {
+			isNew = false
+			reasons = append(reasons, fmt.Sprintf(
+				"data read %s (%s of capacity)",
+				formatBytesHuman(float64(r.readBytes)), formatPercent(readFrac*100)))
+		}
+	} else {
+		notes = append(notes, "capacity unknown — write/read criteria not evaluated")
+	}
+
+	if r.powerOnHours >= newDiskMaxUptimeHours {
+		isNew = false
+		reasons = append(reasons, fmt.Sprintf("uptime %s", formatHoursHuman(r.powerOnHours)))
+	}
+
+	if r.powerCycles >= newDiskMaxPowerCycles {
+		isNew = false
+		reasons = append(reasons, fmt.Sprintf("power cycles %s", formatUint(r.powerCycles)))
+	}
+
+	if isNew {
+		writeField(b, "Disk Condition", "NEW")
+	} else {
+		writeField(b, "Disk Condition", "USED")
+		b.WriteString("  Reason:\n")
+		for _, reason := range reasons {
+			fmt.Fprintf(b, "    - %s\n", reason)
+		}
+	}
+	for _, note := range notes {
+		fmt.Fprintf(b, "  Note: %s\n", note)
+	}
+}
+
+// progressBar renders a fixed-width pseudographic bar, e.g. "[######------]".
+func progressBar(frac float64, width int) string {
+	if math.IsNaN(frac) || frac < 0 {
+		frac = 0
+	}
+	if frac > 1 {
+		frac = 1
+	}
+	filled := int(math.Round(frac * float64(width)))
+	return "[" + strings.Repeat("#", filled) + strings.Repeat("-", width-filled) + "]"
+}
+
+// formatBytesHuman renders a decimal (SI) human-readable byte size, e.g. "1.23 TB".
+func formatBytesHuman(n float64) string {
+	units := []string{"B", "KB", "MB", "GB", "TB", "PB"}
+	i := 0
+	for n >= 1000 && i < len(units)-1 {
+		n /= 1000
+		i++
+	}
+	if i == 0 {
+		return fmt.Sprintf("%.0f %s", n, units[i])
+	}
+	return fmt.Sprintf("%.2f %s", n, units[i])
+}
+
+// formatHoursHuman renders an hour count as a human-scaled duration (hours,
+// days, or years) so uptimes don't show as raw four/five-digit hour counts.
+func formatHoursHuman(hours uint64) string {
+	if hours < 48 {
+		return fmt.Sprintf("%d h", hours)
+	}
+	days := float64(hours) / 24
+	if days < 365 {
+		return fmt.Sprintf("%.0f d", days)
+	}
+	years := days / 365
+	if years == math.Trunc(years) {
+		return fmt.Sprintf("%.0f y", years)
+	}
+	return fmt.Sprintf("%.1f y", years)
+}
+
+// formatPercent renders a percentage with extra precision below 1% (e.g.
+// "0.03%"), where a rounded "0%" would hide any usage at all.
+func formatPercent(pct float64) string {
+	if pct > 0 && pct < 1 {
+		return fmt.Sprintf("%.2f%%", pct)
+	}
+	return fmt.Sprintf("%.0f%%", pct)
+}
+
+// parseLeadingUint parses the leading run of digits/commas in s (e.g. from a
+// SMART raw value or "500,107,862,016 bytes") into a uint64, ignoring the rest.
+func parseLeadingUint(s string) uint64 {
+	s = strings.TrimSpace(s)
+	end := 0
+	for end < len(s) && (s[end] >= '0' && s[end] <= '9' || s[end] == ',') {
+		end++
+	}
+	digits := strings.ReplaceAll(s[:end], ",", "")
+	n, _ := strconv.ParseUint(digits, 10, 64)
+	return n
+}
+
+// ── Formatting helpers ────────────────────────────────────────────────────────
+
+func writeSectionHeader(b *strings.Builder, title string) {
+	b.WriteString("\n")
+	header := "-- " + title + " "
+	header += strings.Repeat("-", max(0, 76-len(header)))
+	b.WriteString(header + "\n")
+}
+
+func writeField(b *strings.Builder, label, value string) {
+	fmt.Fprintf(b, "  %-20s : %s\n", label, value)
+}
+
+func formatCapacityGB(bytes uint64) string {
+	gb := float64(bytes) / 1e9
+	if gb >= 1000 {
+		return fmt.Sprintf("%.2g TB", gb/1000)
+	}
+	return fmt.Sprintf("%.0f GB", math.Round(gb))
+}
+
+func formatUint(n uint64) string {
+	if n == 0 {
+		return "0"
+	}
+	s := strconv.FormatUint(n, 10)
+	// insert thousand separators
+	var out []byte
+	for i, c := range s {
+		if i > 0 && (len(s)-i)%3 == 0 {
+			out = append(out, ',')
+		}
+		out = append(out, byte(c))
+	}
+	return string(out)
+}
+
+func max(a, b int) int {
+	if a > b {
+		return a
+	}
+	return b
+}

audit/internal/platform/storage_report_test.go

@@ -0,0 +1,151 @@
+package platform
+
+import (
+	"strings"
+	"testing"
+	"time"
+)
+
+var testNVMeIdCtrl = []byte(`{
+  "mn": "SAMSUNG MZ1L2960HCJR-00A07 ",
+  "sn": "S665NN0X415495",
+  "fr": "GDC7602Q",
+  "tnvmcap": 960197124096
+}`)
+
+var testNVMeSmartLog = []byte(`{
+  "critical_warning": 0,
+  "temperature": 311,
+  "avail_spare": 100,
+  "spare_thresh": 10,
+  "percent_used": 0,
+  "data_units_read": 1023456,
+  "data_units_written": 738281,
+  "power_cycles": 32,
+  "power_on_hours": 1234,
+  "unsafe_shutdowns": 3,
+  "media_errors": 0,
+  "num_err_log_entries": 0
+}`)
+
+// lo/hi variant emitted by some nvme-cli versions
+var testNVMeSmartLogLoHi = []byte(`{
+  "critical_warning": 0,
+  "temperature": {"lo": 311, "hi": 0},
+  "avail_spare": 100,
+  "spare_thresh": 10,
+  "percent_used": 0,
+  "data_units_read": {"lo": 1023456, "hi": 0},
+  "data_units_written": {"lo": 738281, "hi": 0},
+  "power_cycles": {"lo": 32, "hi": 0},
+  "power_on_hours": {"lo": 1234, "hi": 0},
+  "unsafe_shutdowns": {"lo": 3, "hi": 0},
+  "media_errors": {"lo": 0, "hi": 0},
+  "num_err_log_entries": {"lo": 0, "hi": 0}
+}`)
+
+var testSmartCtlHealth = []byte(`
+smartctl 7.3 2022-02-28 r5338 [x86_64-linux-5.15.0] (local build)
+Copyright (C) 2002-22, Bruce Allen, Christian Franke, www.smartmontools.org
+
+=== START OF INFORMATION SECTION ===
+Device Model:     SAMSUNG MZ1L2960HCJR-00A07
+Serial Number:    S665NN0X415495
+Firmware Version: GDC7602Q
+User Capacity:    960,197,124,096 bytes [960 GB]
+
+=== START OF READ SMART DATA SECTION ===
+SMART overall-health self-assessment test result: PASSED
+
+SMART Attributes Data Structure revision number: 1
+Vendor Specific SMART Attributes with Thresholds:
+ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
+  5 Reallocated_Sector_Ct   0x0032   100   100   000    Old_age   Always       -       0
+  9 Power_On_Hours           0x0032   100   100   000    Old_age   Always       -       1234
+ 12 Power_Cycle_Count        0x0032   100   100   000    Old_age   Always       -       45
+177 Wear_Leveling_Count      0x0013   097   097   000    Pre-fail  Always       -       30
+190 Airflow_Temperature_Cel  0x0032   063   045   000    Old_age   Always       -       37
+`)
+
+func TestGenerateDiskReportNVMe(t *testing.T) {
+	t.Parallel()
+	outputs := map[string][]byte{
+		"nvme-id-ctrl":   testNVMeIdCtrl,
+		"nvme-smart-log": testNVMeSmartLog,
+	}
+	report := GenerateDiskReportText(1, "/dev/nvme0n1", outputs, time.Unix(0, 0).UTC())
+
+	assertContains(t, report, "Disk 1", "/dev/nvme0n1")
+	assertContains(t, report, "SAMSUNG MZ1L2960HCJR-00A07")
+	assertContains(t, report, "S665NN0X415495")
+	assertContains(t, report, "GDC7602Q")
+	assertContains(t, report, "38 °C")     // 311 K - 273
+	assertContains(t, report, "1,234 h")   // power_on_hours with separator
+	assertContains(t, report, "32")        // power_cycles
+	assertContains(t, report, "3")         // unsafe_shutdowns
+	assertContains(t, report, "378.00 GB") // data_units_written * 512000, human-scaled
+}
+
+// TestGenerateDiskReportNVMeDataUnitsScaleToTB verifies that heavy write/read
+// counters render in the "-- Usage --" section as TB/PB, not raw GB, matching
+// the "-- Resource --" section which already used formatBytesHuman.
+func TestGenerateDiskReportNVMeDataUnitsScaleToTB(t *testing.T) {
+	t.Parallel()
+	heavy := []byte(`{
+  "critical_warning": 0,
+  "temperature": 307,
+  "avail_spare": 100,
+  "spare_thresh": 10,
+  "percent_used": 0,
+  "data_units_read": "252420478",
+  "data_units_written": "103834055",
+  "power_cycles": "45",
+  "power_on_hours": "45",
+  "unsafe_shutdowns": "35",
+  "media_errors": "0",
+  "num_err_log_entries": "0"
+}`)
+	outputs := map[string][]byte{
+		"nvme-id-ctrl":   testNVMeIdCtrl,
+		"nvme-smart-log": heavy,
+	}
+	report := GenerateDiskReportText(1, "/dev/nvme0n1", outputs, time.Unix(0, 0).UTC())
+
+	assertContains(t, report, "Data Written         : 53.16 TB")
+	assertContains(t, report, "Data Read            : 129.24 TB")
+}
+
+func TestGenerateDiskReportNVMeLoHi(t *testing.T) {
+	t.Parallel()
+	outputs := map[string][]byte{
+		"nvme-id-ctrl":   testNVMeIdCtrl,
+		"nvme-smart-log": testNVMeSmartLogLoHi,
+	}
+	report := GenerateDiskReportText(1, "/dev/nvme0n1", outputs, time.Unix(0, 0).UTC())
+	assertContains(t, report, "38 °C")
+	assertContains(t, report, "1,234 h")
+}
+
+func TestGenerateDiskReportSATA(t *testing.T) {
+	t.Parallel()
+	outputs := map[string][]byte{
+		"smartctl-health": testSmartCtlHealth,
+	}
+	report := GenerateDiskReportText(2, "/dev/sda", outputs, time.Unix(0, 0).UTC())
+
+	assertContains(t, report, "Disk 2", "/dev/sda")
+	assertContains(t, report, "SAMSUNG MZ1L2960HCJR-00A07")
+	assertContains(t, report, "S665NN0X415495")
+	assertContains(t, report, "PASSED")
+	assertContains(t, report, "Reallocated_Sector_Ct")
+	assertContains(t, report, "Power_On_Hours")
+}
+
+func assertContains(t *testing.T, text string, needles ...string) {
+	t.Helper()
+	for _, needle := range needles {
+		if !strings.Contains(text, needle) {
+			t.Errorf("report missing %q\nreport:\n%s", needle, text)
+		}
+	}
+}

audit/internal/platform/techdump.go

@@ -25,6 +25,9 @@ var techDumpFixedCommands = []struct {
 	{Name: "sensors", Args: []string{"-j"}, File: "sensors.json"},
 	{Name: "ipmitool", Args: []string{"fru", "print"}, File: "ipmitool-fru.txt"},
 	{Name: "ipmitool", Args: []string{"sdr"}, File: "ipmitool-sdr.txt"},
+	{Name: "ipmitool", Args: []string{"sensor"}, File: "ipmitool-sensor.txt"},
+	{Name: "ipmitool", Args: []string{"sel", "list"}, File: "ipmitool-sel.txt"},
+	{Name: "ipmitool", Args: []string{"sel", "time", "get"}, File: "ipmitool-sel-time.txt"},
 	{Name: "nvme", Args: []string{"list", "-o", "json"}, File: "nvme-list.json"},
 }
 
@@ -35,6 +38,7 @@ var techDumpNvidiaCommands = []struct {
 }{
 	{Name: "nvidia-smi", Args: []string{"-q"}, File: "nvidia-smi-q.txt"},
 	{Name: "nvidia-smi", Args: []string{"--query-gpu=index,pci.bus_id,serial,vbios_version,temperature.gpu,power.draw,ecc.errors.uncorrected.aggregate.total,ecc.errors.corrected.aggregate.total,clocks_throttle_reasons.hw_slowdown", "--format=csv,noheader,nounits"}, File: "nvidia-smi-query.csv"},
+	{Name: "nvidia-smi", Args: []string{"conf-compute", "-q"}, File: "nvidia-smi-conf-compute-q.txt"},
 }
 
 type lsblkDumpRoot struct {

audit/internal/webui/api.go

@@ -135,7 +135,7 @@ func defaultTaskPriority(target string, params taskParams) int {
 		return taskPriorityBurn
 	case "nvidia", "nvidia-targeted-stress", "nvidia-targeted-power", "nvidia-pulse",
 		"nvidia-interconnect", "nvidia-bandwidth", "memory", "storage", "cpu",
-		"amd", "amd-mem", "amd-bandwidth":
+		"amd", "amd-mem", "amd-bandwidth", "confidential-computing":
 		if params.StressMode {
 			return taskPriorityValidateStress
 		}
@@ -1292,6 +1292,22 @@ func (h *handler) handleAPIInstallToRAM(w http.ResponseWriter, r *http.Request)
 	_ = json.NewEncoder(w).Encode(map[string]string{"task_id": t.ID})
 }
 
+func (h *handler) handleAPISystemReboot(w http.ResponseWriter, r *http.Request) {
+	if err := exec.Command("systemctl", "reboot").Start(); err != nil {
+		writeError(w, http.StatusInternalServerError, "reboot failed: "+err.Error())
+		return
+	}
+	writeJSON(w, map[string]string{"status": "rebooting"})
+}
+
+func (h *handler) handleAPISystemShutdown(w http.ResponseWriter, r *http.Request) {
+	if err := exec.Command("systemctl", "poweroff").Start(); err != nil {
+		writeError(w, http.StatusInternalServerError, "shutdown failed: "+err.Error())
+		return
+	}
+	writeJSON(w, map[string]string{"status": "shutting down"})
+}
+
 // ── Tools ─────────────────────────────────────────────────────────────────────
 
 var standardTools = []string{

audit/internal/webui/huawei_elabel.go

@@ -0,0 +1,280 @@
+package webui
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"os/exec"
+	"strconv"
+	"strings"
+	"time"
+)
+
+type huaweiField struct {
+	Name     string `json:"name"`
+	Key      string `json:"key"`
+	Value    string `json:"value"`
+	ReadOnly bool   `json:"read_only,omitempty"`
+}
+
+type huaweiChange struct {
+	Key   string `json:"key"`
+	Value string `json:"value"`
+}
+
+type huaweiFieldDef struct {
+	Name    string
+	Key     string
+	FruID   byte
+	TypeID  byte
+	FieldID byte
+	Special string // "chassis-type" | "guid"
+}
+
+var huaweiElabelDefs = []huaweiFieldDef{
+	{"Device Name",            "DeviceName",            0x00, 0x06, 0x01, ""},
+	{"Device Serial Number",   "DeviceSerialNumber",    0x00, 0x06, 0x03, ""},
+	{"Product Name",           "ProductName",           0x00, 0x03, 0x01, ""},
+	{"Product Serial Number",  "ProductSerialNumber",   0x00, 0x03, 0x04, ""},
+	{"Product Asset Tag",      "ProductAssetTag",       0x00, 0x03, 0x05, ""},
+	{"Product Manufacturer",   "ProductManufacturer",   0x00, 0x03, 0x00, ""},
+	{"Mainboard Manufacturer", "MainboardManufacturer", 0x00, 0x02, 0x01, ""},
+	{"Board Product Name",     "BoardProductName",      0x00, 0x02, 0x02, ""},
+	{"Chassis Part Number",    "ChassisPartnumber",     0x00, 0x01, 0x01, ""},
+	{"Chassis Type",           "ChassisType",           0x00, 0x01, 0x00, "chassis-type"},
+	{"IO Chassis Serial",      "IOChassisSerialNumber", 0x01, 0x03, 0x04, ""},
+	{"IO Chassis Asset Tag",   "IOChassisAssetTag",     0x01, 0x03, 0x05, ""},
+	{"GUID",                   "GUID",                  0x00, 0x00, 0x00, "guid"},
+}
+
+// huaweiGetRaw reads a string elabel field via OEM IPMI raw command.
+// Protocol: ipmitool raw 0x30 0x90 0x05 <fru_id> <type_id> <field_id> 0x00 0x30
+// Response: <length_byte> <ascii_byte1> ... (null-terminated)
+func huaweiGetRaw(ctx context.Context, def huaweiFieldDef) (string, error) {
+	if def.Special == "guid" {
+		return huaweiGetGUID(ctx)
+	}
+	args := []string{
+		"0x30", "0x90", "0x05",
+		fmt.Sprintf("0x%02x", def.FruID),
+		fmt.Sprintf("0x%02x", def.TypeID),
+		fmt.Sprintf("0x%02x", def.FieldID),
+		"0x00", "0x30",
+	}
+	out, err := exec.CommandContext(ctx, "ipmitool", append([]string{"raw"}, args...)...).CombinedOutput()
+	if err != nil {
+		return "", err
+	}
+	return huaweiParseStringResponse(strings.TrimSpace(string(out)), def.Special), nil
+}
+
+// huaweiParseStringResponse decodes the OEM IPMI response bytes to a string.
+// Format: <length_byte> <byte1> <byte2> ...
+func huaweiParseStringResponse(hexOut, special string) string {
+	parts := strings.Fields(hexOut)
+	if len(parts) < 2 {
+		return ""
+	}
+	if special == "chassis-type" {
+		// Response: <length=1> <type_byte>
+		if len(parts) >= 2 {
+			n, err := strconv.ParseUint(parts[1], 16, 8)
+			if err == nil {
+				return fmt.Sprintf("0x%02x", n)
+			}
+		}
+		return ""
+	}
+	var sb strings.Builder
+	for _, p := range parts[1:] {
+		b, err := strconv.ParseUint(p, 16, 8)
+		if err != nil || b == 0 {
+			break
+		}
+		sb.WriteByte(byte(b))
+	}
+	return strings.TrimRight(sb.String(), "\x00")
+}
+
+// huaweiGetGUID reads the system GUID via standard IPMI Get System GUID (0x06 0x08).
+func huaweiGetGUID(ctx context.Context) (string, error) {
+	out, err := exec.CommandContext(ctx, "ipmitool", "raw", "0x06", "0x08").CombinedOutput()
+	if err != nil {
+		return "", err
+	}
+	parts := strings.Fields(strings.TrimSpace(string(out)))
+	if len(parts) != 16 {
+		return "", nil
+	}
+	// Format as UUID: 4-2-2-2-6 byte groups
+	// iBMC returns bytes in reversed order; re-reverse to get canonical UUID.
+	var bytes [16]string
+	for i, p := range parts {
+		bytes[15-i] = p
+	}
+	return fmt.Sprintf("%s%s%s%s-%s%s-%s%s-%s%s-%s%s%s%s%s%s",
+		bytes[0], bytes[1], bytes[2], bytes[3],
+		bytes[4], bytes[5],
+		bytes[6], bytes[7],
+		bytes[8], bytes[9],
+		bytes[10], bytes[11], bytes[12], bytes[13], bytes[14], bytes[15],
+	), nil
+}
+
+// huaweiChunks splits a value into 19-byte chunks for the OEM IPMI SET protocol.
+// Key byte: bit7=1 means more chunks follow; bits 0-6 = offset into string.
+func huaweiChunks(value string) [][]string {
+	if len(value) == 0 {
+		return [][]string{{"0x00", "0x01", "0x00"}}
+	}
+	const maxLen = 63
+	if len(value) > maxLen {
+		value = value[:maxLen]
+	}
+	const chunkSize = 19
+	var chunks [][]string
+	for offset := 0; offset < len(value); {
+		end := offset + chunkSize
+		if end > len(value) {
+			end = len(value)
+		}
+		isLast := end >= len(value)
+		key := byte(offset)
+		if !isLast {
+			key |= 0x80
+		}
+		args := []string{
+			fmt.Sprintf("0x%02x", key),
+			fmt.Sprintf("0x%02x", end-offset),
+		}
+		for _, b := range []byte(value[offset:end]) {
+			args = append(args, fmt.Sprintf("0x%02x", b))
+		}
+		chunks = append(chunks, args)
+		offset = end
+	}
+	return chunks
+}
+
+func (h *handler) handleAPIHuaweiElabelRead(w http.ResponseWriter, r *http.Request) {
+	ctx, cancel := context.WithTimeout(r.Context(), 60*time.Second)
+	defer cancel()
+
+	var fields []huaweiField
+	for _, def := range huaweiElabelDefs {
+		val, err := huaweiGetRaw(ctx, def)
+		if err != nil {
+			// First field failure likely means no Huawei BMC — abort with error.
+			if len(fields) == 0 {
+				msg := strings.TrimSpace(err.Error())
+				writeError(w, http.StatusInternalServerError, "huawei elabel not available: "+msg)
+				return
+			}
+			val = ""
+		}
+		fields = append(fields, huaweiField{
+			Name:     def.Name,
+			Key:      def.Key,
+			Value:    val,
+			ReadOnly: def.Special == "guid" || def.Special == "chassis-type",
+		})
+	}
+	writeJSON(w, fields)
+}
+
+func (h *handler) handleAPIHuaweiElabelWrite(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		Changes []huaweiChange `json:"changes"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		writeError(w, http.StatusBadRequest, "invalid JSON")
+		return
+	}
+	if len(req.Changes) == 0 {
+		writeError(w, http.StatusUnprocessableEntity, "no changes provided")
+		return
+	}
+
+	defByKey := make(map[string]huaweiFieldDef, len(huaweiElabelDefs))
+	for _, d := range huaweiElabelDefs {
+		defByKey[d.Key] = d
+	}
+
+	for _, c := range req.Changes {
+		def, ok := defByKey[c.Key]
+		if !ok {
+			writeError(w, http.StatusUnprocessableEntity, "unknown field key: "+c.Key)
+			return
+		}
+		if def.Special == "guid" || def.Special == "chassis-type" {
+			writeError(w, http.StatusUnprocessableEntity, "field is read-only: "+c.Key)
+			return
+		}
+		if len(c.Value) > 63 {
+			writeError(w, http.StatusUnprocessableEntity, "value too long (max 63 chars): "+c.Key)
+			return
+		}
+		for _, ch := range c.Value {
+			if ch < 0x20 || ch > 0x7E {
+				writeError(w, http.StatusUnprocessableEntity, "non-printable character in value for: "+c.Key)
+				return
+			}
+		}
+	}
+
+	t := &Task{
+		ID:        newJobID("huawei-elabel-write"),
+		Name:      fmt.Sprintf("Huawei Elabel Write (%d field(s))", len(req.Changes)),
+		Target:    "huawei-elabel-write",
+		Priority:  defaultTaskPriority("huawei-elabel-write", taskParams{}),
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params:    taskParams{HuaweiElabelChanges: req.Changes},
+	}
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID})
+}
+
+func runHuaweiElabelWriteTask(ctx context.Context, j *jobState, p taskParams) error {
+	defByKey := make(map[string]huaweiFieldDef, len(huaweiElabelDefs))
+	for _, d := range huaweiElabelDefs {
+		defByKey[d.Key] = d
+	}
+
+	// Enable device name effective flag before writing.
+	enableCmd := exec.CommandContext(ctx, "ipmitool", "raw", "0x30", "0x90", "0x21", "0x04", "0x01")
+	if out, err := enableCmd.CombinedOutput(); err != nil {
+		j.append("Warning: enable flag: " + strings.TrimSpace(string(out)))
+	}
+
+	for _, c := range p.HuaweiElabelChanges {
+		def := defByKey[c.Key]
+		setPrefix := []string{
+			"0x30", "0x90", "0x04",
+			fmt.Sprintf("0x%02x", def.FruID),
+			fmt.Sprintf("0x%02x", def.TypeID),
+			fmt.Sprintf("0x%02x", def.FieldID),
+		}
+
+		chunks := huaweiChunks(c.Value)
+		j.append(fmt.Sprintf("Setting %s = %q (%d chunk(s))", c.Key, c.Value, len(chunks)))
+
+		for _, chunk := range chunks {
+			args := append([]string{"raw"}, setPrefix...)
+			args = append(args, chunk...)
+			cmd := exec.CommandContext(ctx, "ipmitool", args...)
+			if err := streamCmdJob(j, cmd); err != nil {
+				return fmt.Errorf("set %s: %w", c.Key, err)
+			}
+		}
+
+		// Commit after each field.
+		commitCmd := exec.CommandContext(ctx, "ipmitool", "raw", "0x30", "0x90", "0x06", "0x00", "0xAA")
+		if out, err := commitCmd.CombinedOutput(); err != nil {
+			return fmt.Errorf("commit after %s: %w (output: %s)", c.Key, err, strings.TrimSpace(string(out)))
+		}
+		j.append("Committed " + c.Key)
+	}
+	return nil
+}

audit/internal/webui/ipmi_fru.go

@@ -0,0 +1,217 @@
+package webui
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strings"
+	"time"
+	"unicode"
+)
+
+type fruField struct {
+	Name     string `json:"name"`
+	Value    string `json:"value"`
+	Editable bool   `json:"editable"`
+	Area     string `json:"area,omitempty"`
+	Index    int    `json:"index,omitempty"`
+}
+
+type fruChange struct {
+	Area  string `json:"area"`
+	Index int    `json:"index"`
+	Name  string `json:"name"`
+	Value string `json:"value"`
+}
+
+// fruEditableFields maps display name → area + index for ipmitool fru edit.
+var fruEditableFields = map[string]struct {
+	Area  string
+	Index int
+}{
+	// Chassis — vendor doc names and ipmitool abbreviated names
+	"Chassis Part Number":   {"c", 0},
+	"Chassis Serial Number": {"c", 1},
+	"Chassis Serial":        {"c", 1},
+	// Board — vendor doc names and ipmitool abbreviated names
+	"Board Manufacturer":  {"b", 0},
+	"Board Mfg":           {"b", 0},
+	"Board Product Name":  {"b", 1},
+	"Board Product":       {"b", 1},
+	"Board Serial Number": {"b", 2},
+	"Board Serial":        {"b", 2},
+	"Board Part Number":   {"b", 3},
+	// Product — vendor doc names and ipmitool abbreviated names
+	"Product Manufacturer":  {"p", 0},
+	"Product Name":          {"p", 1},
+	"Product Part Number":   {"p", 2},
+	"Product Version":       {"p", 3},
+	"Product Serial Number": {"p", 4},
+	"Product Serial":        {"p", 4},
+	"Product Asset Tag":     {"p", 5},
+}
+
+// fruExtraBaseIndex gives the starting ipmitool field index for each area's
+// repeated "<Area> Extra" custom fields, per the vendor FRU field doc (Chassis
+// extra fields start at 2, Board at 5, Product at 7). ipmitool fru print
+// emits one identically-named line per custom field, so parseFRUOutput
+// counts occurrences to recover the real index for each one.
+var fruExtraBaseIndex = map[string]struct {
+	Area string
+	Base int
+}{
+	"Chassis Extra": {"c", 2},
+	"Board Extra":   {"b", 5},
+	"Product Extra": {"p", 7},
+}
+
+func parseFRUOutput(output string) []fruField {
+	var fields []fruField
+	extraSeen := map[string]int{}
+	for _, line := range strings.Split(output, "\n") {
+		// Lines look like: " Field Name            : value"
+		trimmed := strings.TrimLeft(line, " \t")
+		if trimmed == "" {
+			continue
+		}
+		colon := strings.Index(trimmed, " : ")
+		valueOffset := 3
+		if colon < 0 {
+			// try ": " with no leading space before colon
+			colon = strings.Index(trimmed, ": ")
+			valueOffset = 2
+			if colon < 0 {
+				continue
+			}
+		}
+		name := strings.TrimSpace(trimmed[:colon])
+		value := strings.TrimSpace(trimmed[colon+valueOffset:])
+		if name == "" {
+			continue
+		}
+		editable, area, idx := fruFieldMeta(name, extraSeen)
+		fields = append(fields, fruField{Name: name, Value: value, Editable: editable, Area: area, Index: idx})
+	}
+	return fields
+}
+
+func fruFieldMeta(name string, extraSeen map[string]int) (editable bool, area string, index int) {
+	if e, ok := fruExtraBaseIndex[name]; ok {
+		idx := e.Base + extraSeen[name]
+		extraSeen[name]++
+		return true, e.Area, idx
+	}
+	if e, ok := fruEditableFields[name]; ok {
+		return true, e.Area, e.Index
+	}
+	// All fields are shown as editable; server will reject unknown fields.
+	return true, "", 0
+}
+
+func (h *handler) handleAPIIPMIFRURead(w http.ResponseWriter, r *http.Request) {
+	ctx, cancel := context.WithTimeout(r.Context(), 30*time.Second)
+	defer cancel()
+
+	out, err := exec.CommandContext(ctx, "ipmitool", "fru", "print", "0").CombinedOutput()
+	if err != nil {
+		msg := strings.TrimSpace(string(out))
+		if msg == "" {
+			msg = err.Error()
+		}
+		writeError(w, http.StatusInternalServerError, "ipmitool fru print: "+msg)
+		return
+	}
+
+	fields := parseFRUOutput(string(out))
+	writeJSON(w, fields)
+}
+
+func (h *handler) handleAPIIPMIFRUWrite(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		Changes []fruChange `json:"changes"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		writeError(w, http.StatusBadRequest, "invalid JSON")
+		return
+	}
+	if len(req.Changes) == 0 {
+		writeError(w, http.StatusUnprocessableEntity, "no changes provided")
+		return
+	}
+	validAreas := map[string]bool{"c": true, "b": true, "p": true}
+	for i, c := range req.Changes {
+		if c.Area == "" {
+			e, ok := fruEditableFields[c.Name]
+			if !ok {
+				writeError(w, http.StatusUnprocessableEntity, "field not writable via ipmitool: "+c.Name)
+				return
+			}
+			req.Changes[i].Area = e.Area
+			req.Changes[i].Index = e.Index
+			c = req.Changes[i]
+		}
+		if !validAreas[c.Area] {
+			writeError(w, http.StatusUnprocessableEntity, "invalid area: "+c.Area)
+			return
+		}
+		if c.Index < 0 || c.Index > 9 {
+			writeError(w, http.StatusUnprocessableEntity, fmt.Sprintf("invalid index %d", c.Index))
+			return
+		}
+		if len(c.Value) > 64 {
+			writeError(w, http.StatusUnprocessableEntity, "value too long (max 64 chars)")
+			return
+		}
+		for _, ch := range c.Value {
+			if ch > unicode.MaxASCII || (ch < 0x20 && ch != 0) {
+				writeError(w, http.StatusUnprocessableEntity, "value contains non-printable characters")
+				return
+			}
+		}
+	}
+
+	t := &Task{
+		ID:        newJobID("ipmi-fru-write"),
+		Name:      fmt.Sprintf("IPMI FRU Write (%d field(s))", len(req.Changes)),
+		Target:    "ipmi-fru-write",
+		Priority:  defaultTaskPriority("ipmi-fru-write", taskParams{}),
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params:    taskParams{FRUChanges: req.Changes},
+	}
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID})
+}
+
+func runIPMIFRUWriteTask(ctx context.Context, j *jobState, exportDir string, p taskParams) error {
+	// Backup current FRU state
+	backupDir := filepath.Join(exportDir, "fru-backups")
+	if err := os.MkdirAll(backupDir, 0755); err != nil {
+		return fmt.Errorf("mkdir fru-backups: %w", err)
+	}
+	stamp := time.Now().Format("20060102150405")
+	backupPath := filepath.Join(backupDir, "fru-"+stamp+".txt")
+
+	backupOut, err := exec.CommandContext(ctx, "ipmitool", "fru", "print", "0").CombinedOutput()
+	if err != nil {
+		return fmt.Errorf("backup fru print: %w", err)
+	}
+	if err := os.WriteFile(backupPath, backupOut, 0644); err != nil {
+		return fmt.Errorf("write backup: %w", err)
+	}
+	j.append("Backup saved to " + backupPath)
+
+	// Apply changes
+	for _, c := range p.FRUChanges {
+		j.append(fmt.Sprintf("Setting %s (%s %d) = %q", c.Name, c.Area, c.Index, c.Value))
+		cmd := exec.CommandContext(ctx, "ipmitool", "fru", "edit", "0", "field", c.Area, fmt.Sprintf("%d", c.Index), c.Value)
+		if err := streamCmdJob(j, cmd); err != nil {
+			return fmt.Errorf("fru edit %s %d: %w", c.Area, c.Index, err)
+		}
+	}
+	return nil
+}

audit/internal/webui/ipmi_fru_test.go

@@ -0,0 +1,59 @@
+package webui
+
+import "testing"
+
+func TestParseFRUOutputExtraFields(t *testing.T) {
+	// Realistic ipmitool fru print output: repeated "<Area> Extra" lines
+	// (one per custom field) must resolve to sequential indices per the
+	// vendor FRU doc (Chassis Extra starts at 2, Board Extra at 5, Product
+	// Extra at 7), not all collapse onto the same index.
+	out := `
+ Product Manufacturer : Inspur
+ Product Name         : NF5280M6
+ Product Part Number  : PN123
+ Product Version      : 1.0
+ Product Serial       : SN123
+ Product Asset Tag    : ASSET01
+ Product Extra        : custom-p1
+ Board Mfg             : Inspur
+ Board Product         : BoardX
+ Board Serial          : BSN1
+ Board Part Number     : BPN1
+ Board Extra           : custom-b1
+ Board Extra           : custom-b2
+ Board Extra           : custom-b3
+ Chassis Part Number  : CPN1
+ Chassis Serial        : CSN1
+ Chassis Extra         : front-half
+ Chassis Extra         : back-half
+`
+	fields := parseFRUOutput(out)
+
+	byName := map[string][]fruField{}
+	for _, f := range fields {
+		byName[f.Name] = append(byName[f.Name], f)
+	}
+
+	assertMeta := func(name string, occurrence int, wantArea string, wantIndex int) {
+		t.Helper()
+		list := byName[name]
+		if occurrence >= len(list) {
+			t.Fatalf("expected occurrence %d of %q, got %d entries", occurrence, name, len(list))
+		}
+		f := list[occurrence]
+		if f.Area != wantArea || f.Index != wantIndex {
+			t.Errorf("%s[%d] = area:%q index:%d, want area:%q index:%d", name, occurrence, f.Area, f.Index, wantArea, wantIndex)
+		}
+		if !f.Editable {
+			t.Errorf("%s[%d] expected editable", name, occurrence)
+		}
+	}
+
+	assertMeta("Product Asset Tag", 0, "p", 5)
+	assertMeta("Product Extra", 0, "p", 7)
+	assertMeta("Board Extra", 0, "b", 5)
+	assertMeta("Board Extra", 1, "b", 6)
+	assertMeta("Board Extra", 2, "b", 7)
+	assertMeta("Chassis Extra", 0, "c", 2)
+	assertMeta("Chassis Extra", 1, "c", 3)
+}

audit/internal/webui/layout.go

@@ -68,10 +68,9 @@ tbody tr:hover td{background:rgba(0,0,0,.03)}
 .chip-warn{background:var(--warn-bg);color:var(--warn-fg);border:1px solid #c9ba9b}
 .chip-fail{background:var(--crit-bg);color:var(--crit-fg);border:1px solid var(--crit-border)}
 .chip-unknown{background:var(--surface-2);color:var(--muted);border:1px solid var(--border)}
-/* Tasks nav badge */
-.tasks-nav-btn{display:flex;justify-content:space-between;align-items:center;padding:10px 16px;color:rgba(255,255,255,.55);font-size:12px;text-decoration:none;border-top:1px solid rgba(255,255,255,.12);margin-top:auto;transition:color .15s}
-.tasks-nav-btn:hover{color:#fff}
-.tasks-nav-count{background:var(--accent);color:#fff;border-radius:10px;padding:1px 7px;font-size:11px;font-weight:700;display:none}
+/* Nav separator and tasks count badge */
+.nav-sep{height:1px;background:rgba(255,255,255,.12);margin:6px 0}
+.tasks-nav-count{background:var(--accent);color:#fff;border-radius:10px;padding:1px 7px;font-size:11px;font-weight:700;display:none;margin-left:auto}
 .tasks-nav-count.active{display:inline}
 /* Output terminal */
 .terminal{background:#1b1c1d;border:1px solid rgba(0,0,0,.2);border-radius:4px;padding:14px;font-family:monospace;font-size:12px;color:#b5cea8;max-height:400px;overflow-y:auto;white-space:pre-wrap;word-break:break-all;user-select:text;-webkit-user-select:text}
@@ -98,15 +97,21 @@ tbody tr:hover td{background:rgba(0,0,0,.03)}
 }
 
 func layoutNav(active string, buildLabel string) string {
-	items := []struct{ id, label, href string }{
-		{"dashboard", "Dashboard", "/"},
-		{"audit", "1. Audit", "/audit"},
-		{"check", "2. Check", "/check"},
-		{"load", "3. Load", "/load"},
-		{"speed", "4. Speed", "/speed"},
-		{"endurance", "5. Endurance", "/endurance"},
-		{"tools", "6. Tools", "/tools"},
-		{"settings", "7. Settings", "/settings"},
+	type navItem struct {
+		id, label, href string
+		sep              bool
+	}
+	items := []navItem{
+		{id: "dashboard", label: "Dashboard", href: "/"},
+		{id: "audit", label: "1. Audit", href: "/audit"},
+		{id: "check", label: "2. Check", href: "/check"},
+		{id: "load", label: "3. Load", href: "/load"},
+		{id: "burn", label: "4. Burn", href: "/burn"},
+		{id: "benchmark", label: "5. Benchmark", href: "/benchmark"},
+		{sep: true},
+		{id: "tasks", label: "Tasks", href: "/tasks"},
+		{id: "tools", label: "Tools", href: "/tools"},
+		{id: "settings", label: "Settings", href: "/settings"},
 	}
 	var b strings.Builder
 	b.WriteString(`<aside class="sidebar">`)
@@ -126,19 +131,23 @@ func layoutNav(active string, buildLabel string) string {
 	}
 	b.WriteString(`<nav class="nav">`)
 	for _, item := range items {
+		if item.sep {
+			b.WriteString(`<div class="nav-sep"></div>`)
+			continue
+		}
 		cls := "nav-item"
 		if item.id == active {
 			cls += " active"
 		}
-		b.WriteString(fmt.Sprintf(`<a class="%s" href="%s">%s</a>`, cls, item.href, item.label))
+		if item.id == "tasks" {
+			b.WriteString(fmt.Sprintf(`<a class="%s" href="%s" id="tasks-nav-item">%s<span class="tasks-nav-count" id="tasks-nav-count"></span></a>`, cls, item.href, item.label))
+		} else {
+			b.WriteString(fmt.Sprintf(`<a class="%s" href="%s">%s</a>`, cls, item.href, item.label))
+		}
 	}
 	b.WriteString(`</nav>`)
-	b.WriteString(`<a href="/tasks" class="tasks-nav-btn" id="tasks-nav-btn">`)
-	b.WriteString(`<span>Tasks</span>`)
-	b.WriteString(`<span class="tasks-nav-count" id="tasks-nav-count"></span>`)
-	b.WriteString(`</a>`)
 	b.WriteString(`<script>`)
-	b.WriteString(`(function(){function u(){fetch('/api/tasks',{cache:'no-store'}).then(function(r){return r.json();}).then(function(d){var n=Array.isArray(d)?d.filter(function(t){return t.status==='pending'||t.status==='running';}).length:0;var c=document.getElementById('tasks-nav-count');var b=document.getElementById('tasks-nav-btn');if(c){c.textContent=n>0?String(n):'';c.className='tasks-nav-count'+(n>0?' active':'');}if(b){b.style.color=n>0?'#f6c90e':'';}}).catch(function(){});}u();setInterval(u,5000);})();`)
+	b.WriteString(`(function(){function u(){fetch('/api/tasks',{cache:'no-store'}).then(function(r){return r.json();}).then(function(d){var n=Array.isArray(d)?d.filter(function(t){return t.status==='pending'||t.status==='running';}).length:0;var c=document.getElementById('tasks-nav-count');var el=document.getElementById('tasks-nav-item');if(c){c.textContent=n>0?String(n):'';c.className='tasks-nav-count'+(n>0?' active':'');}if(el){el.style.color=n>0?'#f6c90e':'';}}).catch(function(){});}u();setInterval(u,5000);})();`)
 	b.WriteString(`</script>`)
 	b.WriteString(`</aside>`)
 	return b.String()

audit/internal/webui/page_benchmark.go

@@ -612,19 +612,6 @@ func renderPowerBenchmarkResultsCard(exportDir string) string {
 	return b.String()
 }
 
-// renderSpeed renders the Speed page (step 4): performance benchmarks.
-// Uses the same benchmark infrastructure; defaults to Standard profile (throughput/bandwidth).
-// For long-duration stability/overnight runs, see Endurance (step 5).
-func renderSpeed(opts HandlerOptions) string {
-	base := renderBenchmark(opts)
-	return `<div class="alert alert-info" style="margin-bottom:16px"><strong>Speed:</strong> Measures GPU compute throughput and memory bandwidth. For overnight stability testing, go to <a href="/endurance">5. Endurance</a>.</div>` + base
-}
-
-// renderEndurance renders the Endurance page (step 5): long-duration reliability tests.
-// Focuses on Stability and Overnight profiles for multi-hour burn validation.
-// For short load tests, see Load (step 3). For throughput measurement, see Speed (step 4).
-func renderEndurance(opts HandlerOptions) string {
-	base := renderBenchmark(opts)
-	return `<div class="alert alert-warn" style="margin-bottom:16px"><strong>Endurance:</strong> Long-duration reliability tests — Stability (several hours) and Overnight (8+ h) profiles. These profiles run hardware at sustained load; results show whether the server holds its performance envelope over time.</div>
-<div class="alert alert-info" style="margin-bottom:16px">Use the <strong>Stability</strong> or <strong>Overnight</strong> profile in the setup card below. The Standard profile is available too but is better suited for the <a href="/speed">4. Speed</a> page.</div>` + base
-}
+// renderSpeed and renderEndurance are legacy wrappers; canonical page is 5. Benchmark at /benchmark.
+func renderSpeed(opts HandlerOptions) string    { return renderBenchmark(opts) }
+func renderEndurance(opts HandlerOptions) string { return renderBenchmark(opts) }

audit/internal/webui/page_burn.go

@@ -1,13 +1,8 @@
 package webui
 
-// renderLoad renders the Load page (step 3): sustained stress tests.
-// For non-destructive status checks, see Check (step 2).
-// For DCGM targeted diagnostics (targeted_stress, targeted_power, pulse), see Check → Validate mode.
-func renderLoad() string { return renderBurn() }
-
 func renderBurn() string {
 	return `<div class="alert alert-warn" style="margin-bottom:16px"><strong>&#9888; Warning:</strong> Stress tests on this page run hardware at high load. Repeated or prolonged use may reduce hardware lifespan. Use only when necessary.</div>
-<div class="alert alert-info" style="margin-bottom:16px"><strong>Scope:</strong> Load runs sustained GPU compute and CPU/memory stress recipes. DCGM diagnostics (<code>targeted_stress</code>, <code>targeted_power</code>, <code>pulse_test</code>) and NCCL/NVBandwidth are on the <a href="/check">2. Check</a> page. For overnight endurance runs, see <a href="/endurance">5. Endurance</a>.</div>
+<div class="alert alert-info" style="margin-bottom:16px"><strong>Scope:</strong> Burn runs sustained GPU compute and CPU/memory stress recipes. DCGM targeted diagnostics (<code>targeted_stress</code>, <code>targeted_power</code>, <code>pulse_test</code>) and NCCL/NVBandwidth are on the <a href="/load">3. Load</a> page. For performance benchmarks, see <a href="/benchmark">5. Benchmark</a>.</div>
 <p style="color:var(--muted);font-size:13px;margin-bottom:16px">Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
 
 <div class="card" style="margin-bottom:16px">

audit/internal/webui/page_export_tools.go

@@ -402,95 +402,226 @@ loadNvidiaSelfHeal();
 }
 
 func renderTools() string {
-	return `<div class="card" style="margin-bottom:16px">
-  <div class="card-head">System Install</div>
-  <div class="card-body">
-    <div style="margin-bottom:20px">
-    <div style="font-weight:600;margin-bottom:8px">Install to RAM</div>
-    <p id="boot-source-text" style="color:var(--muted);font-size:13px;margin-bottom:8px">Detecting boot source...</p>
-    <p id="ram-status-text" style="color:var(--muted);font-size:13px;margin-bottom:8px">Checking...</p>
-    <button id="ram-install-btn" class="btn btn-primary" onclick="installToRAM()" style="display:none">&#9654; Copy to RAM</button>
-    </div>
-    <div style="border-top:1px solid var(--line);padding-top:20px">
-    <div style="font-weight:600;margin-bottom:8px">Install to Disk</div>` +
-		renderInstallInline() + `
-    </div>
-  </div>
-</div>
-<script>
-fetch('/api/system/ram-status').then(r=>r.json()).then(d=>{
-  const boot = document.getElementById('boot-source-text');
-  const txt = document.getElementById('ram-status-text');
-  const btn = document.getElementById('ram-install-btn');
-  let source = d.device || d.source || 'unknown source';
-  let kind = d.kind || 'unknown';
-  let label = source;
-  if (kind === 'ram') label = 'RAM';
-  else if (kind === 'usb') label = 'USB (' + source + ')';
-  else if (kind === 'cdrom') label = 'CD-ROM (' + source + ')';
-  else if (kind === 'disk') label = 'disk (' + source + ')';
-  else label = source;
-  boot.textContent = 'Current boot source: ' + label + '.';
-  txt.textContent = d.blocked_reason || d.message || 'Checking...';
-  if (d.status === 'ok' || d.in_ram) {
-    txt.style.color = 'var(--ok, green)';
-  } else if (d.status === 'failed') {
-    txt.style.color = 'var(--err, #b91c1c)';
-  } else {
-    txt.style.color = 'var(--muted)';
-  }
-  if (d.can_start_task) {
-    btn.style.display = '';
-    btn.disabled = false;
-  } else {
-    btn.style.display = 'none';
-  }
-});
-function installToRAM() {
-  document.getElementById('ram-install-btn').disabled = true;
-  fetch('/api/system/install-to-ram', {method:'POST'}).then(r=>r.json()).then(d=>{
-    window.location.href = '/tasks#' + d.task_id;
-  });
-}
-</script>
+	return renderNVMeFormatCard() + `
 
-<div class="card"><div class="card-head">Support Bundle</div><div class="card-body">
-<p style="font-size:13px;color:var(--muted);margin-bottom:12px">Downloads a tar.gz archive of all audit files, SAT results, and logs.</p>
-` + renderSupportBundleInline() + `
-<div style="border-top:1px solid var(--border);margin-top:16px;padding-top:16px">
-  <div style="font-weight:600;margin-bottom:8px">USB Black-Box</div>
-  ` + renderUSBExportInline() + `
-</div>
+` + renderFRUEditorCard() + `
+
+` + renderRAIDMgmtCard()
+}
+
+func renderFRUEditorCard() string {
+	return `<div class="card"><div class="card-head card-head-actions">FRU / Elabel<div class="card-head-buttons"><button class="btn btn-sm btn-secondary" onclick="fruAllRead()">Read All</button></div></div><div class="card-body">
+<p style="font-size:13px;color:var(--muted);margin-bottom:12px">Reads and edits hardware identity fields from all available sources. Each field shows its source method.</p>
+<div id="fru-all-status" style="font-size:13px;color:var(--muted);margin-bottom:8px"></div>
+<div id="fru-src-status" style="display:none;margin-bottom:10px"></div>
+<div id="fru-all-table"></div>
 </div></div>
-
-<div class="card"><div class="card-head">Tool Check <button class="btn btn-sm btn-secondary" onclick="checkTools()" style="margin-left:auto">&#8635; Check</button></div>
-<div class="card-body"><div id="tools-table"><p style="color:var(--muted);font-size:13px">Checking...</p></div></div></div>
-
-<div class="card"><div class="card-head">NVIDIA Self Heal</div><div class="card-body">` +
-		renderNvidiaSelfHealInline() + `</div></div>
-
-<div class="card"><div class="card-head">Network</div><div class="card-body">` +
-		renderNetworkInline() + `</div></div>
-
-<div class="card"><div class="card-head">Services</div><div class="card-body">` +
-		renderServicesInline() + `</div></div>
-
-` + renderNVMeFormatCard() + `
-
-` + renderSAADMICard() + `
-
+<style>
+.fru-chip{display:inline-block;font-size:10px;font-weight:600;letter-spacing:.02em;padding:1px 6px;border-radius:3px;vertical-align:middle;white-space:nowrap;margin-right:8px;flex-shrink:0}
+.fru-chip-ipmi{background:#e8e8e8;color:#555}
+.fru-chip-huawei{background:#fff0e6;color:#b83}
+.fru-chip-saa{background:#e6f0ff;color:#557}
+.fru-inp-wrap{display:flex;align-items:center;gap:0}
+</style>
 <script>
-function checkTools() {
-  document.getElementById('tools-table').innerHTML = '<p style="color:var(--muted);font-size:13px">Checking...</p>';
-  fetch('/api/tools/check').then(r=>r.json()).then(tools => {
-    const rows = tools.map(t =>
-      '<tr><td>'+t.Name+'</td><td><span class="badge '+(t.OK ? 'badge-ok' : 'badge-err')+'">'+(t.OK ? '&#10003; '+t.Path : '&#10007; missing')+'</span></td></tr>'
-    ).join('');
-    document.getElementById('tools-table').innerHTML =
-      '<table><tr><th>Tool</th><th>Status</th></tr>'+rows+'</table>';
+(function(){
+var _actBtn='width:22px;height:22px;padding:0;font-size:13px;line-height:1;border:1px solid var(--line);border-radius:3px;background:var(--surface);cursor:pointer;vertical-align:middle;';
+var _inp='width:100%;padding:3px 6px;border:1.5px solid #888;border-radius:3px;font-size:13px;font-family:monospace;background:var(--surface);color:var(--ink);';
+
+var SOURCES = [
+  {
+    id: 'ipmi-fru',
+    label: 'IPMI FRU',
+    chipClass: 'fru-chip-ipmi',
+    url: '/api/tools/ipmi-fru',
+    writeUrl: '/api/tools/ipmi-fru/write',
+    rowAttrs: function(f) {
+      return 'data-source="ipmi-fru" data-area="'+esc(f.area||'')+'" data-index="'+(f.index||0)+'" data-name="'+esc(f.name)+'"';
+    },
+    writeBody: function(inp) {
+      return JSON.stringify({changes:[{area:inp.dataset.area,index:parseInt(inp.dataset.index,10),name:inp.dataset.name,value:inp.value}]});
+    },
+    fieldName: function(f) { return f.name; },
+    fieldValue: function(f) { return f.value||''; },
+    readOnly: function(f) { return false; },
+  },
+  {
+    id: 'huawei',
+    label: 'Huawei iBMC',
+    chipClass: 'fru-chip-huawei',
+    url: '/api/tools/huawei-elabel',
+    writeUrl: '/api/tools/huawei-elabel/write',
+    rowAttrs: function(f) {
+      return 'data-source="huawei" data-key="'+esc(f.key)+'"';
+    },
+    writeBody: function(inp) {
+      return JSON.stringify({changes:[{key:inp.dataset.key,value:inp.value}]});
+    },
+    fieldName: function(f) { return f.name; },
+    fieldValue: function(f) { return f.value||''; },
+    readOnly: function(f) { return !!f.read_only; },
+  },
+  {
+    id: 'saa-dmi',
+    label: 'SAA DMI',
+    chipClass: 'fru-chip-saa',
+    url: '/api/tools/saa-dmi',
+    writeUrl: '/api/tools/saa-dmi/write',
+    rowAttrs: function(f) {
+      return 'data-source="saa-dmi" data-shn="'+esc(f.shn)+'"';
+    },
+    writeBody: function(inp) {
+      return JSON.stringify({changes:[{shn:inp.dataset.shn,value:inp.value}]});
+    },
+    fieldName: function(f) { return f.name; },
+    fieldValue: function(f) { return f.value||''; },
+    readOnly: function(f) { return false; },
+  },
+];
+
+function esc(s){return String(s==null?'':s).replace(/&/g,'&amp;').replace(/</g,'&lt;').replace(/>/g,'&gt;').replace(/"/g,'&quot;');}
+
+function renderSrcStatus(perSource) {
+  var bar = document.getElementById('fru-src-status');
+  if (!perSource.length) { bar.style.display = 'none'; bar.innerHTML = ''; return; }
+  var html = '';
+  perSource.forEach(function(p) {
+    var state, color;
+    if (p.ok) {
+      state = p.count + ' field(s) available';
+      color = 'var(--ok-fg,green)';
+    } else if (/not activated|product key|SFT-DCMS|SFT-OOB/i.test(p.reason)) {
+      state = 'requires Supermicro license (SFT-OOB-LIC / SFT-DCMS-SINGLE) — activate on BMC';
+      color = 'var(--crit-fg,#9f3a38)';
+    } else {
+      state = p.reason || 'unavailable';
+      color = 'var(--muted)';
+    }
+    html += '<div style="display:flex;align-items:center;gap:8px;font-size:12px;margin:3px 0">'
+      + '<span class="fru-chip '+p.src.chipClass+'">'+p.src.label+'</span>'
+      + '<span style="color:'+color+'">'+esc(state)+'</span>'
+      + '</div>';
   });
+  bar.innerHTML = html;
+  bar.style.display = '';
 }
-checkTools();
+
+window.fruAllRead = function() {
+  var status = document.getElementById('fru-all-status');
+  var table  = document.getElementById('fru-all-table');
+  status.textContent = 'Reading…'; status.style.color = 'var(--muted)';
+  table.innerHTML = '';
+
+  var fetches = SOURCES.map(function(src) {
+    return fetch(src.url, {cache:'no-store'})
+      .then(function(r){ return r.json().then(function(d){ if(!r.ok) throw new Error(d.error||r.statusText); return d; }); });
+  });
+
+  Promise.allSettled(fetches).then(function(results) {
+    var rows = '';
+    var totalFields = 0;
+    var perSource = [];
+
+    results.forEach(function(res, i) {
+      var src = SOURCES[i];
+      if (res.status === 'rejected' || !Array.isArray(res.value) || res.value.length === 0) {
+        var reason = '';
+        if (res.status === 'rejected' && res.reason) reason = res.reason.message;
+        else reason = 'no editable fields returned';
+        perSource.push({src:src, ok:false, count:0, reason:reason});
+        return;
+      }
+      perSource.push({src:src, ok:true, count:res.value.length, reason:''});
+      res.value.forEach(function(f) {
+        var val = esc(src.fieldValue(f));
+        var ro  = src.readOnly(f);
+        var attrs = ro ? '' : (' '+src.rowAttrs(f));
+        rows += '<tr>'
+          + '<td style="white-space:nowrap;padding-right:4px;vertical-align:middle">'
+          +   '<span class="fru-chip '+src.chipClass+'">'+src.label+'</span>'
+          + '</td>'
+          + '<td style="color:var(--muted);white-space:nowrap;padding-right:16px;vertical-align:middle;font-size:13px">'+esc(src.fieldName(f))+'</td>'
+          + '<td style="vertical-align:middle">'
+          +   (ro
+              ? '<span style="font-family:monospace;font-size:13px;color:var(--muted)">'+val+'</span>'
+              : '<input class="fru-uni-inp" style="'+_inp+'" value="'+val+'" data-original="'+val+'"'+attrs+' oninput="fruUniChanged(this)">')
+          + '</td>'
+          + '<td class="fru-uni-act" style="display:none;white-space:nowrap;padding-left:6px;vertical-align:middle">'
+          +   '<button style="'+_actBtn+'color:var(--ok-fg,green);margin-right:3px" title="Save" onclick="fruUniSave(this)">✓</button>'
+          +   '<button style="'+_actBtn+'color:var(--crit-fg,#9f3a38)" title="Cancel" onclick="fruUniCancel(this)">✗</button>'
+          +   '<span class="fru-uni-msg" style="font-size:11px;margin-left:5px;color:var(--muted)"></span>'
+          + '</td>'
+          + '</tr>';
+        totalFields++;
+      });
+    });
+
+    renderSrcStatus(perSource);
+
+    if (totalFields === 0) {
+      status.textContent = 'No editable fields available — see per-source status below.';
+      status.style.color = 'var(--crit-fg,#9f3a38)';
+      table.innerHTML = '';
+      return;
+    }
+
+    table.innerHTML = '<table style="width:100%;border-collapse:collapse">'+rows+'</table>';
+    status.textContent = totalFields + ' field(s) loaded';
+    status.style.color = 'var(--muted)';
+  });
+};
+
+window.fruUniChanged = function(inp) {
+  var row = inp.closest('tr');
+  row.querySelector('.fru-uni-act').style.display = inp.value !== inp.dataset.original ? '' : 'none';
+  row.querySelector('.fru-uni-msg').textContent = '';
+};
+
+window.fruUniCancel = function(btn) {
+  var row = btn.closest('tr');
+  var inp = row.querySelector('.fru-uni-inp');
+  inp.value = inp.dataset.original;
+  row.querySelector('.fru-uni-act').style.display = 'none';
+  row.querySelector('.fru-uni-msg').textContent = '';
+};
+
+window.fruUniSave = function(btn) {
+  var row = btn.closest('tr');
+  var inp = row.querySelector('.fru-uni-inp');
+  var msg = row.querySelector('.fru-uni-msg');
+  var cancelBtn = row.querySelectorAll('.fru-uni-act button')[1];
+  var src = SOURCES.find(function(s){ return s.id === inp.dataset.source; });
+  if (!src) { msg.textContent = 'Unknown source'; msg.style.color='var(--crit-fg)'; return; }
+
+  btn.disabled = true; cancelBtn.disabled = true;
+  msg.textContent = '…'; msg.style.color = 'var(--muted)';
+
+  fetch(src.writeUrl, {method:'POST', headers:{'Content-Type':'application/json'}, body:src.writeBody(inp)})
+    .then(function(r){ return r.json().then(function(d){ if(!r.ok) throw new Error(d.error||r.statusText); return d; }); })
+    .then(function(d) {
+      var poll = setInterval(function() {
+        fetch('/api/tasks',{cache:'no-store'}).then(function(r){return r.json();}).then(function(tasks){
+          var t = Array.isArray(tasks) ? tasks.find(function(x){return x.id===d.task_id;}) : null;
+          if (!t) return;
+          if (t.status==='done') {
+            clearInterval(poll);
+            inp.dataset.original = inp.value;
+            row.querySelector('.fru-uni-act').style.display = 'none';
+            msg.textContent = ''; msg.style.color = '';
+          } else if (t.status==='failed'||t.status==='cancelled') {
+            clearInterval(poll);
+            msg.textContent = t.error||t.status; msg.style.color = 'var(--crit-fg)';
+            btn.disabled = false; cancelBtn.disabled = false;
+          }
+        });
+      }, 1500);
+    })
+    .catch(function(e) {
+      msg.textContent = 'Error: '+e.message; msg.style.color = 'var(--crit-fg)';
+      btn.disabled = false; cancelBtn.disabled = false;
+    });
+};
+})();
 </script>`
 }
 

audit/internal/webui/page_settings.go

@@ -7,34 +7,76 @@ func renderSettings(opts HandlerOptions) string {
 	if version == "" {
 		version = "dev"
 	}
-	return `<div class="grid2">
-
-<div class="card">
-  <div class="card-head">Blackbox Logging</div>
+	return `<div class="card" style="margin-bottom:16px">
+  <div class="card-head">System Install</div>
   <div class="card-body">
-    <p style="font-size:13px;color:var(--muted);margin-bottom:14px">Continuous hardware monitoring that writes a rolling log of sensor readings to the export directory. Useful for capturing thermal or power anomalies during long runs.</p>
-    <div style="display:flex;gap:8px;align-items:center">
-      <button class="btn btn-primary btn-sm" onclick="blackboxToggle('enable')">Enable</button>
-      <button class="btn btn-secondary btn-sm" onclick="blackboxToggle('disable')">Disable</button>
-      <span id="blackbox-status" style="font-size:12px;color:var(--muted)">Loading...</span>
+    <div style="margin-bottom:20px">
+    <div style="font-weight:600;margin-bottom:8px">Install to RAM</div>
+    <p id="boot-source-text" style="color:var(--muted);font-size:13px;margin-bottom:8px">Detecting boot source...</p>
+    <p id="ram-status-text" style="color:var(--muted);font-size:13px;margin-bottom:8px">Checking...</p>
+    <button id="ram-install-btn" class="btn btn-primary" onclick="installToRAM()" style="display:none">&#9654; Copy to RAM</button>
+    </div>
+    <div style="border-top:1px solid var(--line);padding-top:20px">
+    <div style="font-weight:600;margin-bottom:8px">Install to Disk</div>` +
+		renderInstallInline() + `
     </div>
   </div>
 </div>
+<script>
+fetch('/api/system/ram-status').then(r=>r.json()).then(d=>{
+  const boot = document.getElementById('boot-source-text');
+  const txt = document.getElementById('ram-status-text');
+  const btn = document.getElementById('ram-install-btn');
+  let kind = d.kind || 'unknown';
+  let source = d.device || d.source || 'unknown source';
+  let label = kind==='ram'?'RAM':kind==='usb'?'USB ('+source+')':kind==='cdrom'?'CD-ROM ('+source+')':kind==='disk'?'disk ('+source+')':source;
+  boot.textContent = 'Current boot source: ' + label + '.';
+  txt.textContent = d.blocked_reason || d.message || 'Checking...';
+  txt.style.color = (d.status==='ok'||d.in_ram)?'var(--ok,green)':d.status==='failed'?'var(--err,#b91c1c)':'var(--muted)';
+  if (d.can_start_task) { btn.style.display=''; btn.disabled=false; } else { btn.style.display='none'; }
+});
+function installToRAM() {
+  document.getElementById('ram-install-btn').disabled = true;
+  fetch('/api/system/install-to-ram', {method:'POST'}).then(r=>r.json()).then(d=>{
+    window.location.href = '/tasks#' + d.task_id;
+  });
+}
+</script>
+
+<div class="card"><div class="card-head">Support Bundle</div><div class="card-body">
+<p style="font-size:13px;color:var(--muted);margin-bottom:12px">Downloads a tar.gz archive of all audit files, SAT results, and logs.</p>
+` + renderSupportBundleInline() + `
+<div style="border-top:1px solid var(--border);margin-top:16px;padding-top:16px">
+  <div style="font-weight:600;margin-bottom:8px">USB Black-Box</div>
+  ` + renderUSBExportInline() + `
+</div>
+</div></div>
+
+<div class="card"><div class="card-head">Tool Check <button class="btn btn-sm btn-secondary" onclick="checkTools()" style="margin-left:auto">&#8635; Check</button></div>
+<div class="card-body"><div id="tools-table"><p style="color:var(--muted);font-size:13px">Checking...</p></div></div></div>
+<script>
+function checkTools() {
+  document.getElementById('tools-table').innerHTML = '<p style="color:var(--muted);font-size:13px">Checking...</p>';
+  fetch('/api/tools/check').then(r=>r.json()).then(tools => {
+    const rows = tools.map(t =>
+      '<tr><td>'+t.Name+'</td><td><span class="badge '+(t.OK?'badge-ok':'badge-err')+'">'+(t.OK?'&#10003; '+t.Path:'&#10007; missing')+'</span></td></tr>'
+    ).join('');
+    document.getElementById('tools-table').innerHTML = '<table><tr><th>Tool</th><th>Status</th></tr>'+rows+'</table>';
+  });
+}
+checkTools();
+</script>
+
+<div class="card"><div class="card-head">NVIDIA Self Heal</div><div class="card-body">` +
+		renderNvidiaSelfHealInline() + `</div></div>
+
+<div class="card"><div class="card-head">Network</div><div class="card-body">` +
+		renderNetworkInline() + `</div></div>
+
+<div class="card"><div class="card-head">Services</div><div class="card-body">` +
+		renderServicesInline() + `</div></div>
 
 <div class="card">
-  <div class="card-head">NVIDIA Recovery</div>
-  <div class="card-body">
-    <p style="font-size:13px;color:var(--muted);margin-bottom:14px">Reset NVIDIA GPU driver state. Use when <code>nvidia-smi</code> reports errors or GPUs appear stuck after a failed test.</p>
-    <div style="display:flex;gap:8px;align-items:center">
-      <button class="btn btn-danger btn-sm" onclick="nvidiaReset()">Reset NVIDIA Driver</button>
-      <span id="nvidia-reset-status" style="font-size:12px;color:var(--muted)"></span>
-    </div>
-  </div>
-</div>
-
-</div>
-
-<div class="card" style="margin-top:0">
   <div class="card-head">Build Info</div>
   <div class="card-body">
     <table style="width:auto">
@@ -46,32 +88,28 @@ func renderSettings(opts HandlerOptions) string {
   </div>
 </div>
 
+<div class="card">
+  <div class="card-head">Power</div>
+  <div class="card-body">
+    <div style="display:flex;gap:8px;align-items:center">
+      <button class="btn btn-secondary btn-sm" onclick="systemPower('reboot')">Reboot</button>
+      <button class="btn btn-secondary btn-sm" onclick="systemPower('shutdown')">Shutdown</button>
+      <span id="power-status" style="font-size:12px;color:var(--muted)"></span>
+    </div>
+  </div>
+</div>
+
 <script>
-(function() {
-  fetch('/api/blackbox/status', {cache:'no-store'}).then(r => r.json()).then(d => {
-    var el = document.getElementById('blackbox-status');
-    if (el) el.textContent = d.enabled ? 'Enabled' : 'Disabled';
-  }).catch(() => {
-    var el = document.getElementById('blackbox-status');
-    if (el) el.textContent = 'Status unavailable';
-  });
-})();
-function blackboxToggle(action) {
-  var el = document.getElementById('blackbox-status');
-  if (el) el.textContent = 'Updating...';
-  fetch('/api/blackbox/' + action, {method:'POST', cache:'no-store'})
-    .then(r => r.json())
-    .then(d => { if (el) el.textContent = d.enabled ? 'Enabled' : 'Disabled'; })
-    .catch(err => { if (el) el.textContent = 'Error: ' + err.message; });
+function systemPower(action) {
+  var label = action === 'reboot' ? 'reboot' : 'shut down';
+  if (!confirm('Are you sure you want to ' + label + ' the server?')) return;
+  var el = document.getElementById('power-status');
+  if (el) el.textContent = action === 'reboot' ? 'Rebooting...' : 'Shutting down...';
+  fetch('/api/system/' + action, {method: 'POST'})
+    .then(function(r) { return r.json(); })
+    .catch(function(e) { if (el) el.textContent = 'Error: ' + e.message; });
 }
-function nvidiaReset() {
-  var el = document.getElementById('nvidia-reset-status');
-  if (!confirm('Reset NVIDIA driver? This will interrupt any running GPU tasks.')) return;
-  if (el) el.textContent = 'Resetting...';
-  fetch('/api/gpu/nvidia-reset', {method:'POST', cache:'no-store'})
-    .then(r => r.json())
-    .then(d => { if (el) el.textContent = d.error ? ('Error: ' + d.error) : 'Done — driver reset.'; })
-    .catch(err => { if (el) el.textContent = 'Error: ' + err.message; });
-}
-</script>`
+</script>
+
+`
 }

audit/internal/webui/page_validate.go

@@ -68,6 +68,14 @@ func validateTotalStressSec(n int) int {
 }
 
 func renderValidate(opts HandlerOptions) string {
+	return renderValidateMode(opts, false)
+}
+
+func renderValidateStress(opts HandlerOptions) string {
+	return renderValidateMode(opts, true)
+}
+
+func renderValidateMode(opts HandlerOptions, stressDefault bool) string {
 	inv := loadValidateInventory(opts)
 	n := inv.NvidiaGPUCount
 	validateTotalStr := validateFmtDur(validateTotalValidateSec(n))
@@ -76,26 +84,49 @@ func renderValidate(opts HandlerOptions) string {
 	if n > 0 {
 		gpuNote = fmt.Sprintf(" (%d GPU)", n)
 	}
-	return `<div class="alert alert-info" style="margin-bottom:16px"><strong>Non-destructive:</strong> Validate tests collect diagnostics only. They do not write to disks, do not run sustained load, and do not increment hardware wear counters.</div>
-<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
+	estStr := validateTotalStr
+	if stressDefault {
+		estStr = stressTotalStr
+	}
+	alert := `<div class="alert alert-info" style="margin-bottom:16px"><strong>Non-destructive:</strong> Validate tests collect diagnostics only. They do not write to disks, do not run sustained load, and do not increment hardware wear counters.</div>`
+	if stressDefault {
+		alert = `<div class="alert alert-warn" style="margin-bottom:16px"><strong>&#9888; Stress mode:</strong> Runs extended load tests — CPU stress-ng, memory passes, DCGM targeted diagnostics. Higher wear than Validate.</div>`
+	}
 
-	<div class="card" style="margin-bottom:16px">
-	  <div class="card-head">Validate Profile</div>
-	  <div class="card-body validate-profile-body">
-	    <div class="validate-profile-col">
-	      <div class="form-row" style="margin:12px 0 0"><label>Mode</label></div>
-	      <label class="cb-row"><input type="radio" name="sat-mode" id="sat-mode-validate" value="validate" checked onchange="satModeChanged()"><span>Validate — quick non-destructive check</span></label>
-	      <label class="cb-row"><input type="radio" name="sat-mode" id="sat-mode-stress" value="stress" onchange="satModeChanged()"><span>Stress — thorough load test (` + stressTotalStr + gpuNote + `)</span></label>
-	    </div>
-	    <div class="validate-profile-col validate-profile-action">
-	      <p style="color:var(--muted);font-size:12px;margin:0 0 10px">Runs validate modules sequentially. Validate: ` + validateTotalStr + gpuNote + `; Stress: ` + stressTotalStr + gpuNote + `. Estimates are based on real log data and scale with GPU count.</p>
-	      <button type="button" class="btn btn-primary" onclick="runAllSAT()">Validate one by one</button>
-	      <div style="margin-top:12px">
-	        <span id="sat-all-status" style="font-size:12px;color:var(--muted)"></span>
-	      </div>
-	    </div>
-	  </div>
-	</div>
+	stressOnlyCards := ""
+	if stressDefault {
+		stressOnlyCards = renderSATCard("nvidia-targeted-stress", "NVIDIA GPU Targeted Stress", "runNvidiaValidateSet('nvidia-targeted-stress')", "", renderValidateCardBody(
+			inv.NVIDIA,
+			`Runs a controlled NVIDIA DCGM load to check stability under moderate stress.`,
+			`<code>dcgmi diag targeted_stress</code>`,
+			validateFmtDur(platform.SATEstimatedNvidiaTargetedStressSec)+` (all GPUs simultaneously).`,
+		)) +
+			renderSATCard("nvidia-targeted-power", "NVIDIA Targeted Power", "runNvidiaValidateSet('nvidia-targeted-power')", "", renderValidateCardBody(
+				inv.NVIDIA,
+				`Checks that the GPU can sustain its declared power delivery envelope. Pass/fail determined by DCGM.`,
+				`<code>dcgmi diag targeted_power</code>`,
+				validateFmtDur(platform.SATEstimatedNvidiaTargetedPowerSec)+` (all GPUs simultaneously).`,
+			)) +
+			renderSATCard("nvidia-pulse", "NVIDIA PSU Pulse Test", "runNvidiaFabricValidate('nvidia-pulse')", "", renderValidateCardBody(
+				inv.NVIDIA,
+				`Tests power supply transient response by pulsing all GPUs simultaneously between idle and full load. Synchronous pulses across all GPUs create worst-case PSU load spikes — running per-GPU would miss PSU-level failures.`,
+				`<code>dcgmi diag pulse_test</code>`,
+				validateFmtDur(platform.SATEstimatedNvidiaPulseTestSec)+` (all GPUs simultaneously; measured on 8-GPU system).`,
+			))
+	}
+
+	satStressModeJS := "function satStressMode() { return false; }"
+	if stressDefault {
+		satStressModeJS = "function satStressMode() { return true; }"
+	}
+
+	return alert + `
+<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
+<div style="display:flex;align-items:center;gap:12px;margin-bottom:16px">
+  <button type="button" class="btn btn-primary" onclick="runAllSAT()">Run All</button>
+  <span id="sat-all-status" style="font-size:12px;color:var(--muted)"></span>
+  <span style="font-size:12px;color:var(--muted)">est. ` + estStr + gpuNote + `</span>
+</div>
 
 <div class="grid3">
 ` + renderSATCard("cpu", "CPU", "runSAT('cpu')", "", renderValidateCardBody(
@@ -112,9 +143,9 @@ func renderValidate(opts HandlerOptions) string {
 		)) +
 		renderSATCard("storage", "Storage", "runSAT('storage')", "", renderValidateCardBody(
 			inv.Storage,
-			`Scans all storage devices and runs the matching health or self-test path for each device type.`,
-			`<code>lsblk</code>; NVMe: <code>nvme</code>; SATA/SAS: <code>smartctl</code>`,
-			`Seconds in Validate (NVMe: instant device query; SATA/SAS: short self-test). Up to ~1 h per device in Stress (extended self-test, device-dependent).`,
+			`Collects SMART data and runs a short self-test on each storage device.`,
+			`<code>lsblk</code>; NVMe: <code>nvme id-ctrl</code>, <code>nvme smart-log</code>, <code>nvme device-self-test -s 1</code>; SATA/SAS: <code>smartctl -H -A</code>, <code>smartctl -t short</code>`,
+			`~2 min per device (NVMe short self-test; SATA/SAS short self-test — duration device-dependent).`,
 		)) +
 		`</div>
 <div style="height:1px;background:var(--border);margin:16px 0"></div>
@@ -122,7 +153,7 @@ func renderValidate(opts HandlerOptions) string {
   <div class="card-head">NVIDIA GPU Selection</div>
   <div class="card-body">
     <p style="font-size:12px;color:var(--muted);margin:0 0 8px">` + inv.NVIDIA + `</p>
-    <p style="font-size:12px;color:var(--muted);margin:0 0 10px">All NVIDIA validate tasks use only the GPUs selected here. The same selection is used by Validate one by one.</p>
+    <p style="font-size:12px;color:var(--muted);margin:0 0 10px">All NVIDIA validate tasks use only the GPUs selected here. The same selection is used by Run All.</p>
     <div style="display:flex;gap:8px;flex-wrap:wrap;margin-bottom:8px">
       <button class="btn btn-sm btn-secondary" type="button" onclick="satSelectAllGPUs()">Select All</button>
       <button class="btn btn-sm btn-secondary" type="button" onclick="satSelectNoGPUs()">Clear</button>
@@ -143,46 +174,19 @@ func renderValidate(opts HandlerOptions) string {
 			validateFmtDur(platform.SATEstimatedNvidiaGPUValidateSec),
 			validateFmtDur(platform.SATEstimatedNvidiaGPUStressSec)),
 	)) +
-		`<div id="sat-card-nvidia-targeted-stress">` +
-		renderSATCard("nvidia-targeted-stress", "NVIDIA GPU Targeted Stress", "runNvidiaValidateSet('nvidia-targeted-stress')", "", renderValidateCardBody(
-			inv.NVIDIA,
-			`Runs a controlled NVIDIA DCGM load to check stability under moderate stress.`,
-			`<code>dcgmi diag targeted_stress</code>`,
-		"Skipped in Validate. Stress: " + validateFmtDur(platform.SATEstimatedNvidiaTargetedStressSec) + ` (all GPUs simultaneously).<p id="sat-ts-mode-hint" style="color:var(--warn-fg);font-size:12px;margin:8px 0 0">Only runs in Stress mode. Switch mode above to enable in Run All.</p>`,
-		)) +
-		`</div>` +
-		`<div id="sat-card-nvidia-targeted-power">` +
-		renderSATCard("nvidia-targeted-power", "NVIDIA Targeted Power", "runNvidiaValidateSet('nvidia-targeted-power')", "", renderValidateCardBody(
-			inv.NVIDIA,
-			`Checks that the GPU can sustain its declared power delivery envelope. Pass/fail determined by DCGM.`,
-			`<code>dcgmi diag targeted_power</code>`,
-		"Skipped in Validate. Stress: " + validateFmtDur(platform.SATEstimatedNvidiaTargetedPowerSec) + ` (all GPUs simultaneously).<p id="sat-tp-mode-hint" style="color:var(--warn-fg);font-size:12px;margin:8px 0 0">Only runs in Stress mode. Switch mode above to enable in Run All.</p>`,
-		)) +
-		`</div>` +
-		`<div id="sat-card-nvidia-pulse">` +
-		renderSATCard("nvidia-pulse", "NVIDIA PSU Pulse Test", "runNvidiaFabricValidate('nvidia-pulse')", "", renderValidateCardBody(
-			inv.NVIDIA,
-			`Tests power supply transient response by pulsing all GPUs simultaneously between idle and full load. Synchronous pulses across all GPUs create worst-case PSU load spikes — running per-GPU would miss PSU-level failures.`,
-			`<code>dcgmi diag pulse_test</code>`,
-			`Skipped in Validate. Stress: `+validateFmtDur(platform.SATEstimatedNvidiaPulseTestSec)+` (all GPUs simultaneously; measured on 8-GPU system).`+`<p id="sat-pt-mode-hint" style="color:var(--warn-fg);font-size:12px;margin:8px 0 0">Only runs in Stress mode. Switch mode above to enable in Run All.</p>`,
-		)) +
-		`</div>` +
-		`<div id="sat-card-nvidia-interconnect">` +
+		stressOnlyCards +
 		renderSATCard("nvidia-interconnect", "NVIDIA Interconnect (NCCL)", "runNvidiaFabricValidate('nvidia-interconnect')", "", renderValidateCardBody(
 			inv.NVIDIA,
 			`Verifies NVLink/NVSwitch fabric bandwidth using NCCL all_reduce_perf across all selected GPUs. Pass/fail based on achieved bandwidth vs. theoretical.`,
 			`<code>all_reduce_perf</code> (NCCL tests)`,
-			`Validate and Stress: `+validateFmtDur(platform.SATEstimatedNvidiaInterconnectSec)+` (all GPUs simultaneously, requires ≥2).`,
+			validateFmtDur(platform.SATEstimatedNvidiaInterconnectSec)+` (all GPUs simultaneously, requires ≥2).`,
 		)) +
-		`</div>` +
-		`<div id="sat-card-nvidia-bandwidth">` +
 		renderSATCard("nvidia-bandwidth", "NVIDIA Bandwidth (NVBandwidth)", "runNvidiaFabricValidate('nvidia-bandwidth')", "", renderValidateCardBody(
 			inv.NVIDIA,
 			`Validates GPU memory copy and peer-to-peer bandwidth paths using NVBandwidth.`,
 			`<code>nvbandwidth</code>`,
-			`Validate and Stress: `+validateFmtDur(platform.SATEstimatedNvidiaBandwidthSec)+` (all GPUs simultaneously; nvbandwidth runs all built-in tests without a time limit — duration set by the tool).`,
+			validateFmtDur(platform.SATEstimatedNvidiaBandwidthSec)+` (all GPUs simultaneously; nvbandwidth runs all built-in tests without a time limit — duration set by the tool).`,
 		)) +
-		`</div>` +
 		`</div>
 <div class="grid3" style="margin-top:16px">
 ` + renderSATCard("amd", "AMD GPU", "runAMDValidateSet()", "", renderValidateCardBody(
@@ -197,36 +201,15 @@ func renderValidate(opts HandlerOptions) string {
   <div class="card-body"><div id="sat-terminal" class="terminal"></div></div>
 </div>
 <style>
-.validate-profile-body { display:grid; grid-template-columns:1fr 1fr 1fr; gap:24px; align-items:stretch; }
-.validate-profile-col { min-width:0; display:flex; flex-direction:column; }
-.validate-profile-action { display:flex; flex-direction:column; align-items:center; justify-content:center; }
 .validate-card-body { padding:0; }
 .validate-card-section { padding:12px 16px 0; }
 .validate-card-section:last-child { padding-bottom:16px; }
 .sat-gpu-row { display:flex; align-items:flex-start; gap:8px; padding:6px 0; cursor:pointer; font-size:13px; }
 .sat-gpu-row input[type=checkbox] { width:16px; height:16px; margin-top:2px; flex-shrink:0; }
-@media(max-width:900px){ .validate-profile-body { grid-template-columns:1fr; } }
 </style>
 <script>
 let satES = null;
-function satStressMode() {
-  return document.querySelector('input[name="sat-mode"]:checked')?.value === 'stress';
-}
-function satModeChanged() {
-  const stress = satStressMode();
-  [
-    {card: 'sat-card-nvidia-targeted-stress', hint: 'sat-ts-mode-hint'},
-    {card: 'sat-card-nvidia-targeted-power',  hint: 'sat-tp-mode-hint'},
-    {card: 'sat-card-nvidia-pulse',           hint: 'sat-pt-mode-hint'},
-  ].forEach(function(item) {
-    const card = document.getElementById(item.card);
-    if (card) {
-      card.style.opacity = stress ? '1' : '0.5';
-      const hint = document.getElementById(item.hint);
-      if (hint) hint.style.display = stress ? 'none' : '';
-    }
-  });
-}
+` + satStressModeJS + `
 function satLabels() {
   return {nvidia:'Validate GPU', 'nvidia-targeted-stress':'NVIDIA Targeted Stress (dcgmi diag targeted_stress)', 'nvidia-targeted-power':'NVIDIA Targeted Power (dcgmi diag targeted_power)', 'nvidia-pulse':'NVIDIA PSU Pulse Test (dcgmi diag pulse_test)', 'nvidia-interconnect':'NVIDIA Interconnect (NCCL all_reduce_perf)', 'nvidia-bandwidth':'NVIDIA Bandwidth (NVBandwidth)', memory:'Validate Memory', storage:'Validate Storage', cpu:'Validate CPU', amd:'Validate AMD GPU', 'amd-mem':'AMD GPU MEM Integrity', 'amd-bandwidth':'AMD GPU MEM Bandwidth'};
 }
@@ -667,7 +650,7 @@ func renderCheck(opts HandlerOptions) string {
 	if n > 0 {
 		gpuNote = fmt.Sprintf(" (%d GPU)", n)
 	}
-	return `<div class="alert alert-info" style="margin-bottom:16px"><strong>Non-destructive:</strong> Check tests collect diagnostics only — no writes to disks, no sustained load, no hardware wear counters incremented. For stress testing, go to <a href="/load">3. Load</a>.</div>
+	return `<div class="alert alert-info" style="margin-bottom:16px"><strong>Non-destructive:</strong> Check tests collect diagnostics only — no writes to disks, no sustained load, no hardware wear counters incremented. For stress testing, go to <a href="/burn">4. Burn</a>.</div>
 <div style="display:flex;align-items:center;gap:12px;margin-bottom:16px">
   <button type="button" class="btn btn-primary" onclick="runAllCheckSAT()">Run All Checks</button>
   <span id="sat-all-status" style="font-size:12px;color:var(--muted)"></span>
@@ -689,9 +672,15 @@ func renderCheck(opts HandlerOptions) string {
 		)) +
 		renderSATCard("storage", "Storage", "runSAT('storage')", "", renderValidateCardBody(
 			inv.Storage,
-			`Scans all storage devices and runs the matching health or self-test path for each.`,
-			`<code>lsblk</code>; NVMe: <code>nvme</code>; SATA/SAS: <code>smartctl</code>`,
-			`Seconds (NVMe: instant device query; SATA/SAS: short self-test).`,
+			`Collects SMART health and attributes for each storage device. No self-test is triggered — read-only query only.`,
+			`<code>lsblk</code>; NVMe: <code>nvme id-ctrl</code>, <code>nvme smart-log</code>; SATA/SAS: <code>smartctl -H -A</code>`,
+			`Seconds — instantaneous device query, no wear counters incremented.`,
+		)) +
+		renderSATCard("confidential-computing", "Confidential Computing", "runSAT('confidential-computing')", "", renderValidateCardBody(
+			inv.NVIDIA,
+			`Checks whether this server can run NVIDIA Confidential Computing: CPU TEE support (Intel TDX / AMD SEV-SNP) and GPU firmware CC capability. Read-only — changes nothing.`,
+			`<code>nvidia-smi conf-compute -q</code>, <code>dmesg</code>, <code>/sys/module/kvm_amd/parameters/*</code>`,
+			`Seconds — read-only query only.`,
 		)) +
 		`</div>
 <div style="height:1px;background:var(--border);margin:16px 0"></div>
@@ -754,7 +743,7 @@ func renderCheck(opts HandlerOptions) string {
 <script>
 let satES = null;
 function satLabels() {
-  return {nvidia:'Check GPU (DCGM L2)', 'nvidia-interconnect':'NVIDIA Interconnect (NCCL)', 'nvidia-bandwidth':'NVIDIA Bandwidth (NVBandwidth)', memory:'Check Memory', storage:'Check Storage', cpu:'Check CPU', amd:'Check AMD GPU', 'amd-mem':'AMD GPU MEM Integrity', 'amd-bandwidth':'AMD GPU MEM Bandwidth'};
+  return {nvidia:'Check GPU (DCGM L2)', 'nvidia-interconnect':'NVIDIA Interconnect (NCCL)', 'nvidia-bandwidth':'NVIDIA Bandwidth (NVBandwidth)', memory:'Check Memory', storage:'Check Storage', cpu:'Check CPU', amd:'Check AMD GPU', 'amd-mem':'AMD GPU MEM Integrity', 'amd-bandwidth':'AMD GPU MEM Bandwidth', 'confidential-computing':'Check Confidential Computing'};
 }
 let satNvidiaGPUsPromise = null;
 function loadSatNvidiaGPUs() {
@@ -890,7 +879,7 @@ function runAllCheckSAT() {
   status.textContent = 'Enqueuing...';
   const nvidiaIndices = satSelectedGPUIndices();
   const nvidiaAllTargets = ['nvidia', 'nvidia-interconnect', 'nvidia-bandwidth'];
-  const baseTargets = ['cpu', 'memory', 'storage'];
+  const baseTargets = ['cpu', 'memory', 'storage', 'confidential-computing'];
   const amdTargets = selectedAMDValidateTargets();
   const expanded = [];
   baseTargets.forEach(t => expanded.push({target: t}));

audit/internal/webui/pages.go

@@ -33,36 +33,36 @@ func renderPage(page string, opts HandlerOptions) string {
 	case "load":
 		pageID = "load"
 		title = "3. Load"
-		body = renderLoad()
-	case "speed":
-		pageID = "speed"
-		title = "4. Speed"
-		body = renderSpeed(opts)
-	case "endurance":
-		pageID = "endurance"
-		title = "5. Endurance"
-		body = renderEndurance(opts)
+		body = renderValidateStress(opts)
+	case "burn":
+		pageID = "burn"
+		title = "4. Burn"
+		body = renderBurn()
+	case "benchmark":
+		pageID = "benchmark"
+		title = "5. Benchmark"
+		body = renderBenchmark(opts)
 	case "tools":
 		pageID = "tools"
-		title = "6. Tools"
+		title = "Tools"
 		body = renderTools()
 	case "settings":
 		pageID = "settings"
-		title = "7. Settings"
+		title = "Settings"
 		body = renderSettings(opts)
 	// Legacy routes (redirected at HTTP level in handlePage; these are fallbacks)
 	case "validate", "tests":
-		pageID = "check"
-		title = "2. Check"
-		body = renderCheck(opts)
-	case "burn", "burn-in":
 		pageID = "load"
 		title = "3. Load"
-		body = renderLoad()
-	case "benchmark":
-		pageID = "speed"
-		title = "4. Speed"
-		body = renderSpeed(opts)
+		body = renderValidate(opts)
+	case "burn-in":
+		pageID = "burn"
+		title = "4. Burn"
+		body = renderBurn()
+	case "speed", "endurance":
+		pageID = "benchmark"
+		title = "5. Benchmark"
+		body = renderBenchmark(opts)
 	case "tasks":
 		pageID = "tasks"
 		title = "Tasks"

audit/internal/webui/raid_mgmt.go

@@ -0,0 +1,857 @@
+package webui
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"os"
+	"os/exec"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// --- Response types ---
+
+type raidDriveInfo struct {
+	Slot   string  `json:"slot,omitempty"`
+	Device string  `json:"device,omitempty"`
+	Model  string  `json:"model,omitempty"`
+	SizeGB float64 `json:"size_gb,omitempty"`
+	Serial string  `json:"serial,omitempty"`
+	State  string  `json:"state,omitempty"`
+}
+
+type raidArrayInfo struct {
+	Name     string   `json:"name"`
+	Level    string   `json:"level,omitempty"`
+	Members  []string `json:"members"`
+	Degraded bool     `json:"degraded"`
+}
+
+type raidControllerInfo struct {
+	ID            string          `json:"id"`
+	Type          string          `json:"type"`
+	Index         int             `json:"index"`
+	Model         string          `json:"model"`
+	ForeignDrives []raidDriveInfo `json:"foreign_drives"`
+	FreeDrives    []raidDriveInfo `json:"free_drives"`
+	AllDrives     []raidDriveInfo `json:"all_drives"`
+	Arrays        []raidArrayInfo `json:"arrays,omitempty"`
+}
+
+type raidStatusResp struct {
+	Controllers []raidControllerInfo `json:"controllers"`
+}
+
+// --- LSI/storcli detection ---
+
+func detectLSIControllers() []raidControllerInfo {
+	ctrlOut, err := exec.Command("storcli64", "/call", "show", "J").Output()
+	if err != nil {
+		return nil
+	}
+
+	var ctrlDoc struct {
+		Controllers []struct {
+			ResponseData struct {
+				Basics struct {
+					Controller int    `json:"Controller"`
+					Model      string `json:"Model"`
+				} `json:"Basics"`
+			} `json:"Response Data"`
+		} `json:"Controllers"`
+	}
+	if err := json.Unmarshal(ctrlOut, &ctrlDoc); err != nil || len(ctrlDoc.Controllers) == 0 {
+		return nil
+	}
+
+	driveOut, _ := exec.Command("storcli64", "/call/eall/sall", "show", "all", "J").Output()
+
+	var driveDoc struct {
+		Controllers []struct {
+			ResponseData struct {
+				DriveInformation []struct {
+					EIDSlt string `json:"EID:Slt"`
+					State  string `json:"State"`
+					Size   string `json:"Size"`
+					Intf   string `json:"Intf"`
+					Med    string `json:"Med"`
+					Model  string `json:"Model"`
+					SN     string `json:"SN"`
+				} `json:"Drive Information"`
+			} `json:"Response Data"`
+		} `json:"Controllers"`
+	}
+	if len(driveOut) > 0 {
+		json.Unmarshal(driveOut, &driveDoc) //nolint:errcheck
+	}
+
+	var controllers []raidControllerInfo
+	for i, c := range ctrlDoc.Controllers {
+		ctrl := raidControllerInfo{
+			ID:            fmt.Sprintf("lsi-%d", c.ResponseData.Basics.Controller),
+			Type:          "lsi",
+			Index:         c.ResponseData.Basics.Controller,
+			Model:         c.ResponseData.Basics.Model,
+			ForeignDrives: []raidDriveInfo{},
+			FreeDrives:    []raidDriveInfo{},
+			AllDrives:     []raidDriveInfo{},
+		}
+		if ctrl.Model == "" {
+			ctrl.Model = fmt.Sprintf("LSI Controller %d", ctrl.Index)
+		}
+
+		if i < len(driveDoc.Controllers) {
+			for _, d := range driveDoc.Controllers[i].ResponseData.DriveInformation {
+				info := raidDriveInfo{
+					Slot:   strings.TrimSpace(d.EIDSlt),
+					Model:  strings.TrimSpace(d.Model),
+					State:  strings.TrimSpace(d.State),
+					SizeGB: raidParseHumanSizeGB(d.Size),
+					Serial: strings.TrimSpace(d.SN),
+				}
+				ctrl.AllDrives = append(ctrl.AllDrives, info)
+				switch strings.TrimSpace(d.State) {
+				case "Frgn":
+					ctrl.ForeignDrives = append(ctrl.ForeignDrives, info)
+				case "UGood", "JBOD":
+					ctrl.FreeDrives = append(ctrl.FreeDrives, info)
+				}
+			}
+		}
+
+		controllers = append(controllers, ctrl)
+	}
+	return controllers
+}
+
+// --- VROC/mdadm detection ---
+
+var raidMDStatDegradedRx = regexp.MustCompile(`\[[U_]+\]`)
+
+type mdStatEntry struct {
+	Name     string
+	Level    string
+	Members  []string
+	Degraded bool
+}
+
+func parseRAIDMDStat(raw string) []mdStatEntry {
+	var entries []mdStatEntry
+	var cur *mdStatEntry
+	for _, line := range strings.Split(raw, "\n") {
+		if strings.HasPrefix(line, "Personalities") || strings.HasPrefix(line, "unused devices") {
+			continue
+		}
+		if idx := strings.Index(line, " : "); idx > 0 {
+			name := strings.TrimSpace(line[:idx])
+			rest := line[idx+3:]
+			entry := mdStatEntry{Name: name}
+			for _, tok := range strings.Fields(rest) {
+				if strings.HasPrefix(tok, "raid") || strings.HasPrefix(tok, "linear") {
+					entry.Level = tok
+				}
+				if bk := strings.Index(tok, "["); bk > 0 && strings.HasSuffix(tok, "]") {
+					entry.Members = append(entry.Members, tok[:bk])
+				}
+			}
+			entries = append(entries, entry)
+			cur = &entries[len(entries)-1]
+			continue
+		}
+		if cur != nil {
+			if m := raidMDStatDegradedRx.FindString(line); m != "" && strings.Contains(m, "_") {
+				cur.Degraded = true
+			}
+		}
+	}
+	return entries
+}
+
+// raidVROCPortRx matches lines like "          Port2 : /dev/sda (SERIAL123)"
+// or "          Port3 : - no device attached -" from `mdadm --detail-platform`.
+var raidVROCPortRx = regexp.MustCompile(`^\s*Port\d+\s*:\s*(\S+)`)
+
+// parseVROCPorts returns the block device basenames (e.g. "sda") that are
+// physically wired to the VROC I/O controller's ports, per `mdadm
+// --detail-platform` output. Drives attached directly to the CPU (or to a
+// separate HBA) rather than through this controller's ports are excluded.
+func parseVROCPorts(raw string) map[string]bool {
+	ports := map[string]bool{}
+	for _, line := range strings.Split(raw, "\n") {
+		m := raidVROCPortRx.FindStringSubmatch(line)
+		if m == nil {
+			continue
+		}
+		dev := m[1]
+		if !strings.HasPrefix(dev, "/dev/") {
+			continue
+		}
+		ports[strings.TrimPrefix(dev, "/dev/")] = true
+	}
+	return ports
+}
+
+func detectVROCController() *raidControllerInfo {
+	out, err := exec.Command("mdadm", "--detail-platform").CombinedOutput()
+	if err != nil && len(out) == 0 {
+		return nil
+	}
+	hasVROC := false
+	for _, line := range strings.Split(string(out), "\n") {
+		lower := strings.ToLower(line)
+		if strings.Contains(lower, "license") || strings.Contains(lower, "intel") || strings.Contains(lower, "platform") {
+			hasVROC = true
+			break
+		}
+	}
+	if !hasVROC {
+		return nil
+	}
+
+	ctrl := &raidControllerInfo{
+		ID:            "vroc-0",
+		Type:          "vroc",
+		Model:         "Intel VROC",
+		ForeignDrives: []raidDriveInfo{},
+		FreeDrives:    []raidDriveInfo{},
+		AllDrives:     []raidDriveInfo{},
+	}
+
+	ports := parseVROCPorts(string(out))
+	// Some mdadm builds omit the "Port" lines from --detail-platform. When
+	// we can't determine which drives are actually wired to this
+	// controller, fall back to showing every disk not already in an array
+	// rather than hiding everything.
+	portsKnown := len(ports) > 0
+
+	inArray := map[string]bool{}
+	raw, err := os.ReadFile("/proc/mdstat")
+	if err == nil {
+		for _, arr := range parseRAIDMDStat(string(raw)) {
+			ctrl.Arrays = append(ctrl.Arrays, raidArrayInfo{
+				Name:     arr.Name,
+				Level:    arr.Level,
+				Members:  arr.Members,
+				Degraded: arr.Degraded,
+			})
+			for _, m := range arr.Members {
+				inArray[m] = true
+			}
+		}
+	}
+
+	lsblkOut, err := exec.Command("lsblk", "-J", "-d", "-o", "NAME,SIZE,TYPE,MODEL,SERIAL").Output()
+	if err == nil {
+		var lsblkDoc struct {
+			BlockDevices []struct {
+				Name   string `json:"name"`
+				Size   string `json:"size"`
+				Type   string `json:"type"`
+				Model  string `json:"model"`
+				Serial string `json:"serial"`
+			} `json:"blockdevices"`
+		}
+		if json.Unmarshal(lsblkOut, &lsblkDoc) == nil {
+			for _, d := range lsblkDoc.BlockDevices {
+				// Only consider disks wired to this controller's ports -
+				// drives attached directly to the CPU (or another
+				// controller) never show up as VROC ports and are skipped.
+				if d.Type != "disk" || (portsKnown && !ports[d.Name]) {
+					continue
+				}
+				info := raidDriveInfo{
+					Device: "/dev/" + d.Name,
+					Model:  strings.TrimSpace(d.Model),
+					Serial: strings.TrimSpace(d.Serial),
+					State:  "available",
+				}
+				if inArray[d.Name] {
+					info.State = "member"
+				}
+				ctrl.AllDrives = append(ctrl.AllDrives, info)
+				if info.State == "available" {
+					ctrl.FreeDrives = append(ctrl.FreeDrives, info)
+				}
+			}
+		}
+	}
+
+	return ctrl
+}
+
+// --- API handlers ---
+
+func (h *handler) handleAPIRAIDStatus(w http.ResponseWriter, r *http.Request) {
+	resp := raidStatusResp{Controllers: []raidControllerInfo{}}
+
+	if lsi := detectLSIControllers(); len(lsi) > 0 {
+		resp.Controllers = append(resp.Controllers, lsi...)
+	}
+	if vroc := detectVROCController(); vroc != nil {
+		resp.Controllers = append(resp.Controllers, *vroc)
+	}
+
+	writeJSON(w, resp)
+}
+
+func (h *handler) handleAPIRAIDForeignAction(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		ControllerID string `json:"controller_id"`
+		Action       string `json:"action"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		writeError(w, http.StatusBadRequest, "invalid JSON")
+		return
+	}
+	if req.Action != "import" && req.Action != "clear" {
+		writeError(w, http.StatusBadRequest, "action must be 'import' or 'clear'")
+		return
+	}
+	ctrlIdx, ok := parseLSIControllerIndex(req.ControllerID)
+	if !ok {
+		writeError(w, http.StatusBadRequest, "invalid controller_id")
+		return
+	}
+
+	target := "raid-foreign-clear"
+	name := fmt.Sprintf("RAID Foreign Clear (ctrl %d)", ctrlIdx)
+	if req.Action == "import" {
+		target = "raid-foreign-import"
+		name = fmt.Sprintf("RAID Foreign Import (ctrl %d)", ctrlIdx)
+	}
+
+	t := &Task{
+		ID:        newJobID(target),
+		Name:      name,
+		Target:    target,
+		Priority:  defaultTaskPriority(target, taskParams{}),
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params:    taskParams{RAIDController: ctrlIdx},
+	}
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID})
+}
+
+func (h *handler) handleAPIRAIDCreateMirror(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		ControllerID string   `json:"controller_id"`
+		Devices      []string `json:"devices"`
+		ArrayName    string   `json:"array_name"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		writeError(w, http.StatusBadRequest, "invalid JSON")
+		return
+	}
+	if len(req.Devices) < 2 {
+		writeError(w, http.StatusBadRequest, "at least 2 devices required")
+		return
+	}
+
+	var target, name string
+	var params taskParams
+
+	switch {
+	case strings.HasPrefix(req.ControllerID, "lsi-"):
+		ctrlIdx, ok := parseLSIControllerIndex(req.ControllerID)
+		if !ok {
+			writeError(w, http.StatusBadRequest, "invalid controller_id")
+			return
+		}
+		target = "raid-lsi-create-mirror"
+		name = fmt.Sprintf("Create RAID 1 Mirror (LSI ctrl %d)", ctrlIdx)
+		params = taskParams{RAIDController: ctrlIdx, RAIDDevices: req.Devices}
+
+	case req.ControllerID == "vroc-0":
+		arrayName := strings.TrimSpace(req.ArrayName)
+		if arrayName == "" {
+			arrayName = "bee-mirror0"
+		}
+		target = "raid-vroc-create-mirror"
+		name = fmt.Sprintf("Create VROC RAID 1 (%s)", arrayName)
+		params = taskParams{RAIDDevices: req.Devices, RAIDArrayName: arrayName}
+
+	default:
+		writeError(w, http.StatusBadRequest, "unknown controller_id")
+		return
+	}
+
+	t := &Task{
+		ID:        newJobID(target),
+		Name:      name,
+		Target:    target,
+		Priority:  defaultTaskPriority(target, taskParams{}),
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params:    params,
+	}
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID})
+}
+
+func (h *handler) handleAPIRAIDPrepareDrive(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		ControllerID string `json:"controller_id"`
+		Slot         string `json:"slot"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		writeError(w, http.StatusBadRequest, "invalid JSON")
+		return
+	}
+	ctrlIdx, ok := parseLSIControllerIndex(req.ControllerID)
+	if !ok {
+		writeError(w, http.StatusBadRequest, "invalid controller_id")
+		return
+	}
+	if _, _, ok := parseRAIDSlot(req.Slot); !ok {
+		writeError(w, http.StatusBadRequest, "invalid slot")
+		return
+	}
+
+	t := &Task{
+		ID:        newJobID("raid-lsi-prepare-drive"),
+		Name:      fmt.Sprintf("Prepare drive %s (LSI ctrl %d)", req.Slot, ctrlIdx),
+		Target:    "raid-lsi-prepare-drive",
+		Priority:  defaultTaskPriority("raid-lsi-prepare-drive", taskParams{}),
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params:    taskParams{RAIDController: ctrlIdx, RAIDSlot: req.Slot},
+	}
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID})
+}
+
+func parseLSIControllerIndex(id string) (int, bool) {
+	if !strings.HasPrefix(id, "lsi-") {
+		return 0, false
+	}
+	n, err := strconv.Atoi(strings.TrimPrefix(id, "lsi-"))
+	if err != nil || n < 0 {
+		return 0, false
+	}
+	return n, true
+}
+
+// --- Task runner functions ---
+
+func runRAIDForeignClearTask(ctx context.Context, j *jobState, ctrl int) error {
+	j.append(fmt.Sprintf("Clearing foreign configuration on controller %d...", ctrl))
+	cmd := exec.CommandContext(ctx, "storcli64", fmt.Sprintf("/c%d/fall", ctrl), "del", "noprompt")
+	return streamCmdJob(j, cmd)
+}
+
+func runRAIDForeignImportTask(ctx context.Context, j *jobState, ctrl int) error {
+	j.append(fmt.Sprintf("Importing foreign configuration on controller %d...", ctrl))
+	cmd := exec.CommandContext(ctx, "storcli64", fmt.Sprintf("/c%d/fall", ctrl), "import", "noprompt")
+	return streamCmdJob(j, cmd)
+}
+
+func runRAIDLSICreateMirrorTask(ctx context.Context, j *jobState, ctrl int, drives []string) error {
+	driveList := strings.Join(drives, ",")
+	j.append(fmt.Sprintf("Creating RAID 1 on controller %d with drives: %s", ctrl, driveList))
+	cmd := exec.CommandContext(ctx, "storcli64",
+		fmt.Sprintf("/c%d", ctrl),
+		"add", "vd", "type=raid1",
+		fmt.Sprintf("drives=%s", driveList),
+		"pdperarray=2",
+	)
+	return streamCmdJob(j, cmd)
+}
+
+// parseRAIDSlot splits a storcli "EID:Slt" identifier (e.g. "252:0") into
+// enclosure and slot numbers.
+func parseRAIDSlot(slot string) (eid int, slt int, ok bool) {
+	parts := strings.SplitN(strings.TrimSpace(slot), ":", 2)
+	if len(parts) != 2 {
+		return 0, 0, false
+	}
+	eid, err1 := strconv.Atoi(strings.TrimSpace(parts[0]))
+	slt, err2 := strconv.Atoi(strings.TrimSpace(parts[1]))
+	if err1 != nil || err2 != nil {
+		return 0, 0, false
+	}
+	return eid, slt, true
+}
+
+func runRAIDPrepareDriveTask(ctx context.Context, j *jobState, ctrl int, slot string) error {
+	eid, slt, ok := parseRAIDSlot(slot)
+	if !ok {
+		return fmt.Errorf("invalid slot %q", slot)
+	}
+	j.append(fmt.Sprintf("Preparing drive %s on controller %d (set good, force)...", slot, ctrl))
+	cmd := exec.CommandContext(ctx, "storcli64",
+		fmt.Sprintf("/c%d/e%d/s%d", ctrl, eid, slt),
+		"set", "good", "force",
+	)
+	return streamCmdJob(j, cmd)
+}
+
+func runRAIDVROCCreateMirrorTask(ctx context.Context, j *jobState, devices []string, arrayName string) error {
+	if arrayName == "" {
+		arrayName = "bee-mirror0"
+	}
+	devPath := "/dev/md/" + arrayName
+	args := []string{
+		"--create", devPath,
+		"--level=1",
+		fmt.Sprintf("--raid-devices=%d", len(devices)),
+		"--run",
+	}
+	args = append(args, devices...)
+	j.append(fmt.Sprintf("Creating VROC RAID 1 array %s with: %s", devPath, strings.Join(devices, " ")))
+	cmd := exec.CommandContext(ctx, "mdadm", args...)
+	return streamCmdJob(j, cmd)
+}
+
+// raidParseHumanSizeGB parses storcli size strings like "1.818 TB", "745.211 GB".
+func raidParseHumanSizeGB(s string) float64 {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return 0
+	}
+	upper := strings.ToUpper(s)
+	var mul float64
+	var numStr string
+	switch {
+	case strings.Contains(upper, " TB"):
+		mul = 1024
+		numStr = strings.TrimSpace(strings.SplitN(upper, " T", 2)[0])
+	case strings.Contains(upper, " GB"):
+		mul = 1
+		numStr = strings.TrimSpace(strings.SplitN(upper, " G", 2)[0])
+	case strings.Contains(upper, " MB"):
+		mul = 1.0 / 1024
+		numStr = strings.TrimSpace(strings.SplitN(upper, " M", 2)[0])
+	default:
+		return 0
+	}
+	v, err := strconv.ParseFloat(numStr, 64)
+	if err != nil {
+		return 0
+	}
+	return v * mul
+}
+
+// --- UI card ---
+
+func renderRAIDMgmtCard() string {
+	return `<div class="card"><div class="card-head card-head-actions">RAID Controller Management<div class="card-head-buttons"><button class="btn btn-sm btn-secondary" onclick="raidLoad()">&#8635; Refresh</button></div></div><div class="card-body">
+<div id="raid-status" style="font-size:13px;color:var(--muted);margin-bottom:8px">Loading...</div>
+<div id="raid-content"></div>
+<div id="raid-out-wrap" style="display:none;margin-top:14px">
+  <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:4px">
+    <span id="raid-out-label" style="font-size:12px;font-weight:600;color:var(--muted)">Output</span>
+    <span id="raid-out-status" style="font-size:12px"></span>
+  </div>
+  <div id="raid-terminal" class="terminal" style="max-height:260px;width:100%;box-sizing:border-box"></div>
+</div>
+</div></div>
+<script>
+(function(){
+function escHtml(s) {
+  return String(s||'').replace(/&/g,'&amp;').replace(/</g,'&lt;').replace(/>/g,'&gt;').replace(/"/g,'&quot;');
+}
+
+var _raidControllers = [];
+
+function raidLoad() {
+  var status = document.getElementById('raid-status');
+  var content = document.getElementById('raid-content');
+  status.textContent = 'Detecting RAID controllers...';
+  status.style.color = 'var(--muted)';
+  content.innerHTML = '';
+  fetch('/api/tools/raid/status', {cache:'no-store'})
+    .then(function(r) {
+      if (!r.ok) return r.json().then(function(e) { throw new Error(e.error || r.statusText); });
+      return r.json();
+    })
+    .then(function(data) {
+      _raidControllers = data.controllers || [];
+      if (_raidControllers.length === 0) {
+        status.textContent = 'No RAID controllers detected.';
+        return;
+      }
+      status.textContent = _raidControllers.length + ' controller(s) detected.';
+      content.innerHTML = _raidControllers.map(function(c, i) {
+        return raidRenderController(c, i);
+      }).join('<hr style="margin:16px 0;border:none;border-top:1px solid var(--border)">');
+    })
+    .catch(function(e) {
+      status.textContent = 'Error: ' + e.message;
+      status.style.color = 'var(--crit-fg)';
+    });
+}
+
+function raidRenderController(c, idx) {
+  var html = '';
+  var typeLabel = c.type === 'lsi' ? 'LSI / Broadcom' : 'Intel VROC';
+  html += '<div style="font-weight:600;font-size:13px;margin-bottom:10px">' + typeLabel + ' &mdash; ' + escHtml(c.model) + '</div>';
+
+  if (c.type === 'lsi') {
+    var foreign = c.foreign_drives || [];
+    if (foreign.length > 0) {
+      html += '<div style="background:var(--warn-bg,rgba(240,192,0,0.1));border:1px solid var(--warn-border,#c8a800);border-radius:4px;padding:10px 12px;margin-bottom:12px">';
+      html += '<div style="font-weight:600;font-size:13px;margin-bottom:6px">&#9888;&#xFE0E; Foreign Configuration Detected (' + foreign.length + ' drive(s))</div>';
+      html += '<table style="margin-bottom:10px"><tr><th>Slot</th><th>Model</th><th>Size</th><th>State</th></tr>';
+      foreign.forEach(function(d) {
+        html += '<tr>'
+          + '<td style="font-family:monospace">' + escHtml(d.slot) + '</td>'
+          + '<td>' + escHtml(d.model||'—') + '</td>'
+          + '<td>' + (d.size_gb > 0 ? Math.round(d.size_gb) + ' GB' : '—') + '</td>'
+          + '<td><span class="badge badge-warn">' + escHtml(d.state) + '</span></td>'
+          + '</tr>';
+      });
+      html += '</table>';
+      html += '<div style="display:flex;gap:8px;flex-wrap:wrap">';
+      html += '<button class="btn btn-sm btn-primary" onclick="raidForeignAction(\'' + escHtml(c.id) + '\',\'import\',this)">Import Foreign Config</button>';
+      html += '<button class="btn btn-sm btn-secondary" style="color:var(--crit-fg)" onclick="raidForeignAction(\'' + escHtml(c.id) + '\',\'clear\',this)">Clear Foreign Config</button>';
+      html += '</div></div>';
+    }
+
+    html += raidRenderAllDrives(c, idx);
+    html += raidRenderMirrorSection(c, idx, 'lsi');
+  }
+
+  if (c.type === 'vroc') {
+    var arrays = c.arrays || [];
+    if (arrays.length > 0) {
+      html += '<div style="font-size:12px;font-weight:600;color:var(--muted);margin-bottom:6px;text-transform:uppercase;letter-spacing:.04em">Active Arrays</div>';
+      html += '<table style="margin-bottom:14px"><tr><th>Name</th><th>Level</th><th>Members</th><th>Status</th></tr>';
+      arrays.forEach(function(a) {
+        var badge = a.degraded
+          ? '<span class="badge badge-err">Degraded</span>'
+          : '<span class="badge badge-ok">OK</span>';
+        html += '<tr>'
+          + '<td style="font-family:monospace">' + escHtml(a.name) + '</td>'
+          + '<td>' + escHtml(a.level||'—') + '</td>'
+          + '<td style="font-family:monospace;font-size:12px">' + (a.members||[]).map(escHtml).join(', ') + '</td>'
+          + '<td>' + badge + '</td>'
+          + '</tr>';
+      });
+      html += '</table>';
+    }
+
+    html += raidRenderAllDrives(c, idx);
+    html += raidRenderMirrorSection(c, idx, 'vroc');
+  }
+
+  return html;
+}
+
+var RAID_READY_STATES = {'UGood': true, 'JBOD': true, 'available': true};
+var RAID_NO_PREPARE_STATES = {'UGood': true, 'JBOD': true, 'Frgn': true, 'Onln': true, 'Msng': true};
+
+function raidRenderAllDrives(c, idx) {
+  var drives = c.all_drives || [];
+  var isLSI = c.type === 'lsi';
+  if (drives.length === 0) {
+    return '<p style="font-size:13px;color:var(--muted);margin-bottom:12px">No drives detected on this controller.</p>';
+  }
+  var html = '<div style="font-size:12px;font-weight:600;color:var(--muted);margin-bottom:6px;text-transform:uppercase;letter-spacing:.04em">All Drives on This Controller</div>';
+  html += '<table style="margin-bottom:14px"><tr><th>' + (isLSI ? 'Slot' : 'Device') + '</th><th>Model</th><th>Size</th><th>State</th>' + (isLSI ? '<th></th>' : '') + '</tr>';
+  drives.forEach(function(d) {
+    var ready = !!RAID_READY_STATES[d.state];
+    var badgeClass = ready ? 'badge-ok' : 'badge-warn';
+    var actionCell = '';
+    if (isLSI && !RAID_NO_PREPARE_STATES[d.state]) {
+      actionCell = '<td><button class="btn btn-sm btn-secondary" onclick="raidPrepareDrive(\'' + escHtml(c.id) + '\',\'' + escHtml(d.slot) + '\',this)">Prepare</button></td>';
+    } else if (isLSI) {
+      actionCell = '<td></td>';
+    }
+    html += '<tr>'
+      + '<td style="font-family:monospace">' + escHtml(isLSI ? d.slot : d.device) + '</td>'
+      + '<td>' + escHtml(d.model||'—') + (d.serial ? ' [' + escHtml(d.serial) + ']' : '') + '</td>'
+      + '<td>' + (d.size_gb > 0 ? Math.round(d.size_gb) + ' GB' : '—') + '</td>'
+      + '<td><span class="badge ' + badgeClass + '">' + escHtml(d.state||'—') + '</span></td>'
+      + actionCell
+      + '</tr>';
+  });
+  html += '</table>';
+  return html;
+}
+
+function raidPrepareDrive(ctrlID, slot, btn) {
+  if (!confirm('Prepare drive ' + slot + ' on ' + ctrlID + ' for array creation?\n\nThis forces the drive into Unconfigured Good state. If it currently belongs to a virtual drive or holds data, that data will become inaccessible.')) {
+    return;
+  }
+  var original = btn ? btn.textContent : '';
+  if (btn) { btn.disabled = true; btn.textContent = 'Preparing...'; }
+  raidShowOutput('Prepare drive ' + slot, '', '');
+  fetch('/api/tools/raid/prepare-drive', {
+    method: 'POST',
+    headers: {'Content-Type': 'application/json'},
+    body: JSON.stringify({controller_id: ctrlID, slot: slot})
+  })
+    .then(function(r) { return r.json(); })
+    .then(function(d) {
+      if (d.error) throw new Error(d.error);
+      raidStreamTask(d.task_id, 'Prepare drive ' + slot, function() {
+        if (btn) { btn.disabled = false; btn.textContent = original; }
+        raidLoad();
+      });
+    })
+    .catch(function(e) {
+      raidShowOutput('Error', 'failed', e.message);
+      if (btn) { btn.disabled = false; btn.textContent = original; }
+    });
+}
+
+function raidRenderMirrorSection(c, idx, kind) {
+  var free = c.free_drives || [];
+  var html = '<div style="font-size:12px;font-weight:600;color:var(--muted);margin-bottom:6px;text-transform:uppercase;letter-spacing:.04em">Create RAID 1 Mirror</div>';
+
+  if (free.length < 2) {
+    html += '<p style="font-size:13px;color:var(--muted)">No unconfigured drives available (need at least 2).</p>';
+    return html;
+  }
+
+  html += '<p style="font-size:13px;color:var(--muted);margin-bottom:8px">Select exactly 2 drives:</p>';
+  html += '<div>';
+  free.forEach(function(d) {
+    var val = kind === 'lsi' ? d.slot : d.device;
+    var label = kind === 'lsi'
+      ? escHtml(d.slot) + (d.model ? ' &mdash; ' + escHtml(d.model) : '') + (d.size_gb > 0 ? ' (' + Math.round(d.size_gb) + ' GB)' : '')
+      : escHtml(d.device) + (d.model ? ' &mdash; ' + escHtml(d.model) : '') + (d.serial ? ' [' + escHtml(d.serial) + ']' : '');
+    html += '<label style="display:block;margin-bottom:4px;font-size:13px;cursor:pointer">'
+      + '<input type="checkbox" class="raid-mirror-check-' + idx + '" value="' + escHtml(val) + '"> '
+      + label + '</label>';
+  });
+  html += '</div>';
+
+  if (kind === 'vroc') {
+    html += '<div style="margin-top:10px;display:flex;align-items:center;gap:8px;flex-wrap:wrap">'
+      + '<label style="font-size:13px">Array name:&nbsp;<input type="text" id="vroc-arrayname-' + idx + '" value="bee-mirror0" style="font-family:monospace;padding:2px 6px;width:140px"></label>';
+  } else {
+    html += '<div style="margin-top:10px;display:flex;gap:8px">';
+  }
+
+  html += '<button class="btn btn-sm btn-primary raid-mirror-btn-' + idx + '" onclick="raidCreateMirror(\'' + escHtml(c.id) + '\',' + idx + ',\'' + kind + '\',this)">Create Mirror</button>';
+  html += '</div>';
+
+  return html;
+}
+
+function raidForeignAction(ctrlID, action, btn) {
+  if (action === 'clear' && !confirm('Clear foreign configuration on ' + ctrlID + '?\n\nThis will DELETE the foreign RAID metadata. Data on those drives may become inaccessible.')) {
+    return;
+  }
+  var original = btn ? btn.textContent : '';
+  if (btn) { btn.disabled = true; btn.textContent = action === 'import' ? 'Importing...' : 'Clearing...'; }
+  raidShowOutput('RAID foreign ' + action, '', '');
+  fetch('/api/tools/raid/foreign', {
+    method: 'POST',
+    headers: {'Content-Type': 'application/json'},
+    body: JSON.stringify({controller_id: ctrlID, action: action})
+  })
+    .then(function(r) { return r.json(); })
+    .then(function(d) {
+      if (d.error) throw new Error(d.error);
+      var actionLabel = action === 'import' ? 'Import foreign config' : 'Clear foreign config';
+      raidStreamTask(d.task_id, actionLabel, function() {
+        if (btn) { btn.disabled = false; btn.textContent = original; }
+        raidLoad();
+      });
+    })
+    .catch(function(e) {
+      raidShowOutput('Error', 'failed', e.message);
+      if (btn) { btn.disabled = false; btn.textContent = original; }
+    });
+}
+
+function raidCreateMirror(ctrlID, idx, kind, btn) {
+  var checks = document.querySelectorAll('.raid-mirror-check-' + idx + ':checked');
+  if (checks.length !== 2) {
+    alert('Select exactly 2 drives.');
+    return;
+  }
+  var devices = Array.from(checks).map(function(c) { return c.value; });
+  var arrayName = '';
+  if (kind === 'vroc') {
+    var nameEl = document.getElementById('vroc-arrayname-' + idx);
+    arrayName = nameEl ? nameEl.value.trim() : 'bee-mirror0';
+    if (!arrayName) arrayName = 'bee-mirror0';
+  }
+  var original = btn ? btn.textContent : '';
+  if (btn) { btn.disabled = true; btn.textContent = 'Creating...'; }
+  raidShowOutput('Create RAID 1', '', '');
+  fetch('/api/tools/raid/create-mirror', {
+    method: 'POST',
+    headers: {'Content-Type': 'application/json'},
+    body: JSON.stringify({controller_id: ctrlID, devices: devices, array_name: arrayName})
+  })
+    .then(function(r) { return r.json(); })
+    .then(function(d) {
+      if (d.error) throw new Error(d.error);
+      raidStreamTask(d.task_id, 'Create RAID 1 mirror', function() {
+        if (btn) { btn.disabled = false; btn.textContent = original; }
+        raidLoad();
+      });
+    })
+    .catch(function(e) {
+      raidShowOutput('Error', 'failed', e.message);
+      if (btn) { btn.disabled = false; btn.textContent = original; }
+    });
+}
+
+function raidShowOutput(label, status, text) {
+  var wrap = document.getElementById('raid-out-wrap');
+  var labelEl = document.getElementById('raid-out-label');
+  var statusEl = document.getElementById('raid-out-status');
+  var term = document.getElementById('raid-terminal');
+  wrap.style.display = 'block';
+  labelEl.textContent = label;
+  if (status === 'ok') {
+    statusEl.textContent = '✓ done';
+    statusEl.style.color = 'var(--ok-fg)';
+  } else if (status === 'failed') {
+    statusEl.textContent = '✗ failed';
+    statusEl.style.color = 'var(--crit-fg)';
+  } else {
+    statusEl.textContent = status;
+    statusEl.style.color = 'var(--muted)';
+  }
+  if (text !== undefined) {
+    term.textContent = text;
+    term.scrollTop = term.scrollHeight;
+  }
+}
+
+function raidStreamTask(taskID, taskName, onDone) {
+  var term = document.getElementById('raid-terminal');
+  term.textContent = '';
+  raidShowOutput(taskName || 'Running…', 'running…', undefined);
+  var es = new EventSource('/api/tasks/' + taskID + '/stream');
+  es.onmessage = function(e) {
+    term.textContent += e.data + '\n';
+    term.scrollTop = term.scrollHeight;
+  };
+  es.addEventListener('done', function(e) {
+    es.close();
+    if (!e.data) {
+      raidShowOutput(taskName, 'ok', undefined);
+    } else {
+      raidShowOutput(taskName, 'failed', undefined);
+      term.textContent += '\nFailed: ' + e.data;
+      term.scrollTop = term.scrollHeight;
+    }
+    if (onDone) onDone();
+  });
+  es.onerror = function() {
+    es.close();
+    raidShowOutput(taskName, 'failed', undefined);
+    if (onDone) onDone();
+  };
+}
+
+window.raidLoad = raidLoad;
+window.raidForeignAction = raidForeignAction;
+window.raidCreateMirror = raidCreateMirror;
+window.raidPrepareDrive = raidPrepareDrive;
+raidLoad();
+})();
+</script>`
+}

audit/internal/webui/saa_dmi.go

@@ -28,10 +28,12 @@ var (
 	shnRE        = regexp.MustCompile(`^[A-Za-z0-9_]{1,16}$`)
 	dmiSectionRE = regexp.MustCompile(`^\[(.+?)\]$`)
 	// Item Name   {SHN}   = value   // comment
-	dmiItemRE    = regexp.MustCompile(`^(.+?)\s+\{([A-Za-z0-9]{1,16})\}\s*=\s*(.*)$`)
+	// SHN may contain parentheses, e.g. {PS(4)LC} for power supply fields
+	dmiItemRE    = regexp.MustCompile(`^(.+?)\s+\{([A-Za-z0-9_()\-]{1,24})\}\s*=\s*(.*)$`)
 	dmiVersionRE = regexp.MustCompile(`(?i)^version\s*=`)
 )
 
+
 // parseDMIFile parses the DMI.txt produced by "saa GetDmiInfo".
 // Real format (from SAA User Guide 4.8.1):
 //
@@ -90,7 +92,9 @@ func (h *handler) handleAPISAADMIRead(w http.ResponseWriter, r *http.Request) {
 	defer os.RemoveAll(tmpDir)
 
 	dmiFile := filepath.Join(tmpDir, "DMI.txt")
-	out, err := exec.CommandContext(ctx, "saa", "-c", "GetDmiInfo", "--file", dmiFile, "--overwrite").CombinedOutput()
+	cmd := exec.CommandContext(ctx, "saa", "-c", "GetDmiInfo", "--file", dmiFile, "--overwrite")
+	cmd.Dir = "/usr/local/bin"
+	out, err := cmd.CombinedOutput()
 	if err != nil {
 		msg := strings.TrimSpace(string(out))
 		if msg == "" {
@@ -167,7 +171,9 @@ func runSAADMIWriteTask(ctx context.Context, j *jobState, exportDir string, p ta
 	dmiFile := filepath.Join(tmpDir, "DMI.txt")
 
 	j.append("Reading current DMI configuration...")
-	if err := streamCmdJob(j, exec.CommandContext(ctx, "saa", "-c", "GetDmiInfo", "--file", dmiFile, "--overwrite")); err != nil {
+	getCmd := exec.CommandContext(ctx, "saa", "-c", "GetDmiInfo", "--file", dmiFile, "--overwrite")
+	getCmd.Dir = "/usr/local/bin"
+	if err := streamCmdJob(j, getCmd); err != nil {
 		return fmt.Errorf("GetDmiInfo: %w", err)
 	}
 
@@ -189,13 +195,16 @@ func runSAADMIWriteTask(ctx context.Context, j *jobState, exportDir string, p ta
 	for _, c := range p.SAADmiChanges {
 		j.append("Setting " + c.Shn + " = " + c.Value)
 		cmd := exec.CommandContext(ctx, "saa", "-c", "EditDmiInfo", "--file", dmiFile, "--shn", c.Shn, "--value", c.Value)
+		cmd.Dir = "/usr/local/bin"
 		if err := streamCmdJob(j, cmd); err != nil {
 			return fmt.Errorf("EditDmiInfo %s: %w", c.Shn, err)
 		}
 	}
 
 	j.append("Applying changes to hardware...")
-	if err := streamCmdJob(j, exec.CommandContext(ctx, "saa", "-c", "ChangeDmiInfo", "--file", dmiFile)); err != nil {
+	changeCmd := exec.CommandContext(ctx, "saa", "-c", "ChangeDmiInfo", "--file", dmiFile)
+	changeCmd.Dir = "/usr/local/bin"
+	if err := streamCmdJob(j, changeCmd); err != nil {
 		return fmt.Errorf("ChangeDmiInfo: %w", err)
 	}
 
@@ -203,99 +212,3 @@ func runSAADMIWriteTask(ctx context.Context, j *jobState, exportDir string, p ta
 	return nil
 }
 
-func renderSAADMICard() string {
-	return `<div class="card"><div class="card-head">SAA &#8212; DMI <button class="btn btn-sm btn-secondary" onclick="saaDMIRead()" style="margin-left:auto">Read</button></div><div class="card-body">
-<p style="font-size:13px;color:var(--muted);margin-bottom:12px">Reads and edits DMI fields via SAA (In-Band). Requires <code>saa</code> on PATH.</p>
-<div id="saa-dmi-status" style="font-size:13px;color:var(--muted);margin-bottom:8px"></div>
-<div id="saa-dmi-table"></div>
-<div id="saa-dmi-save-row" style="display:none;margin-top:12px">
-  <button class="btn btn-primary" id="saa-dmi-save-btn" onclick="saaDMISave()">Save</button>
-  <span id="saa-dmi-save-msg" style="font-size:13px;color:var(--muted);margin-left:10px"></span>
-</div>
-<script>
-function saaDMIEsc(s) {
-  return String(s==null?'':s).replace(/[&<>"']/g,function(c){return{'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[c];});
-}
-function saaDMIUpdateSaveBtn() {
-  var inputs = document.querySelectorAll('#saa-dmi-table input[data-original]');
-  var dirty = [];
-  inputs.forEach(function(inp){if(inp.value!==inp.dataset.original)dirty.push(inp);});
-  var row = document.getElementById('saa-dmi-save-row');
-  var btn = document.getElementById('saa-dmi-save-btn');
-  if(dirty.length>0){row.style.display='';btn.textContent='Save ('+dirty.length+' changed)';}
-  else{row.style.display='none';}
-}
-function saaDMIRead() {
-  var status = document.getElementById('saa-dmi-status');
-  var table = document.getElementById('saa-dmi-table');
-  var saveRow = document.getElementById('saa-dmi-save-row');
-  status.textContent = 'Reading...';
-  status.style.color = 'var(--muted)';
-  table.innerHTML = '';
-  saveRow.style.display = 'none';
-  fetch('/api/tools/saa-dmi').then(function(r){return r.json().then(function(d){if(!r.ok)throw new Error(d.error||('HTTP '+r.status));return d;});}).then(function(fields){
-    status.textContent = fields.length+' field(s) loaded.';
-    var rows = fields.map(function(f){
-      return '<tr>'
-        +'<td style="font-size:13px;white-space:nowrap;padding-right:8px">'+saaDMIEsc(f.name)+'</td>'
-        +'<td style="font-family:monospace;font-size:13px;white-space:nowrap;padding-right:8px">'+saaDMIEsc(f.shn)+'</td>'
-        +'<td><input type="text" value="'+saaDMIEsc(f.value)+'" data-shn="'+saaDMIEsc(f.shn)+'" data-original="'+saaDMIEsc(f.value)+'" oninput="saaDMIMarkDirty(this)" style="width:100%;font-family:monospace;font-size:13px;border:1px solid var(--line);padding:3px 6px;border-radius:3px"></td>'
-        +'<td id="saa-dmi-dirty-'+saaDMIEsc(f.shn)+'" style="font-size:12px;color:var(--warn,#b45309);width:50px;padding-left:6px"></td>'
-        +'</tr>';
-    }).join('');
-    table.innerHTML = '<table style="width:100%;border-collapse:collapse"><tr><th style="text-align:left;font-size:13px;padding-bottom:6px">Field</th><th style="text-align:left;font-size:13px;padding-bottom:6px">Shn</th><th style="text-align:left;font-size:13px;padding-bottom:6px">Value</th><th></th></tr>'+rows+'</table>';
-  }).catch(function(e){
-    status.textContent = 'Error: '+e.message;
-    status.style.color = 'var(--crit-fg,#9f3a38)';
-  });
-}
-function saaDMIMarkDirty(inp) {
-  var shn = inp.dataset.shn;
-  var cell = document.getElementById('saa-dmi-dirty-'+shn);
-  if(cell)cell.textContent = inp.value!==inp.dataset.original?'changed':'';
-  saaDMIUpdateSaveBtn();
-}
-function saaDMIWaitTask(taskID) {
-  var msg = document.getElementById('saa-dmi-save-msg');
-  msg.textContent = 'Task '+taskID+' queued...';
-  msg.style.color = 'var(--muted)';
-  var timer = setInterval(function(){
-    fetch('/api/tasks').then(function(r){return r.json();}).then(function(tasks){
-      var task = (tasks||[]).find(function(t){return t.id===taskID;});
-      if(!task)return;
-      if(task.status==='done'||task.status==='failed'||task.status==='cancelled'){
-        clearInterval(timer);
-        msg.textContent = task.status==='done'?'Saved. Reboot to apply.':'Failed: '+(task.error||task.status);
-        msg.style.color = task.status==='done'?'var(--ok,green)':'var(--crit-fg,#9f3a38)';
-        document.getElementById('saa-dmi-save-btn').disabled = false;
-      }
-    }).catch(function(){});
-  }, 1500);
-}
-function saaDMISave() {
-  var inputs = document.querySelectorAll('#saa-dmi-table input[data-original]');
-  var changes = [];
-  inputs.forEach(function(inp){if(inp.value!==inp.dataset.original)changes.push({shn:inp.dataset.shn,value:inp.value});});
-  if(!changes.length)return;
-  var names = changes.map(function(c){return c.shn;}).join(', ');
-  if(!window.confirm('Apply DMI changes for: '+names+'?\n\nThe server will need to be rebooted for changes to take effect.'))return;
-  var btn = document.getElementById('saa-dmi-save-btn');
-  var msg = document.getElementById('saa-dmi-save-msg');
-  btn.disabled = true;
-  msg.textContent = 'Submitting...';
-  msg.style.color = 'var(--muted)';
-  fetch('/api/tools/saa-dmi/write',{
-    method:'POST',
-    headers:{'Content-Type':'application/json'},
-    body:JSON.stringify({changes:changes})
-  }).then(function(r){return r.json().then(function(d){if(!r.ok)throw new Error(d.error||('HTTP '+r.status));return d;});}).then(function(d){
-    saaDMIWaitTask(d.task_id);
-  }).catch(function(e){
-    msg.textContent = 'Error: '+e.message;
-    msg.style.color = 'var(--crit-fg,#9f3a38)';
-    btn.disabled = false;
-  });
-}
-</script>
-</div></div>`
-}

audit/internal/webui/server.go

@@ -264,6 +264,7 @@ func NewHandler(opts HandlerOptions) http.Handler {
 	mux.HandleFunc("POST /api/sat/nvidia-stress/run", h.handleAPISATRun("nvidia-stress"))
 	mux.HandleFunc("POST /api/sat/memory/run", h.handleAPISATRun("memory"))
 	mux.HandleFunc("POST /api/sat/storage/run", h.handleAPISATRun("storage"))
+	mux.HandleFunc("POST /api/sat/confidential-computing/run", h.handleAPISATRun("confidential-computing"))
 	mux.HandleFunc("POST /api/sat/cpu/run", h.handleAPISATRun("cpu"))
 	mux.HandleFunc("POST /api/sat/amd/run", h.handleAPISATRun("amd"))
 	mux.HandleFunc("POST /api/sat/amd-mem/run", h.handleAPISATRun("amd-mem"))
@@ -316,6 +317,14 @@ func NewHandler(opts HandlerOptions) http.Handler {
 	mux.HandleFunc("POST /api/tools/nvme-format/run", h.handleAPINVMeFormatRun)
 	mux.HandleFunc("GET /api/tools/saa-dmi", h.handleAPISAADMIRead)
 	mux.HandleFunc("POST /api/tools/saa-dmi/write", h.handleAPISAADMIWrite)
+	mux.HandleFunc("GET /api/tools/ipmi-fru", h.handleAPIIPMIFRURead)
+	mux.HandleFunc("POST /api/tools/ipmi-fru/write", h.handleAPIIPMIFRUWrite)
+	mux.HandleFunc("GET /api/tools/huawei-elabel", h.handleAPIHuaweiElabelRead)
+	mux.HandleFunc("POST /api/tools/huawei-elabel/write", h.handleAPIHuaweiElabelWrite)
+	mux.HandleFunc("GET /api/tools/raid/status", h.handleAPIRAIDStatus)
+	mux.HandleFunc("POST /api/tools/raid/foreign", h.handleAPIRAIDForeignAction)
+	mux.HandleFunc("POST /api/tools/raid/create-mirror", h.handleAPIRAIDCreateMirror)
+	mux.HandleFunc("POST /api/tools/raid/prepare-drive", h.handleAPIRAIDPrepareDrive)
 
 	// GPU presence / tools
 	mux.HandleFunc("GET /api/gpu/presence", h.handleAPIGPUPresence)
@@ -327,6 +336,8 @@ func NewHandler(opts HandlerOptions) http.Handler {
 	// System
 	mux.HandleFunc("GET /api/system/ram-status", h.handleAPIRAMStatus)
 	mux.HandleFunc("POST /api/system/install-to-ram", h.handleAPIInstallToRAM)
+	mux.HandleFunc("POST /api/system/reboot", h.handleAPISystemReboot)
+	mux.HandleFunc("POST /api/system/shutdown", h.handleAPISystemShutdown)
 
 	// Preflight
 	mux.HandleFunc("GET /api/preflight", h.handleAPIPreflight)
@@ -1424,13 +1435,13 @@ func (h *handler) handlePage(w http.ResponseWriter, r *http.Request) {
 	// Redirect legacy routes to new named pages
 	switch page {
 	case "validate", "tests":
-		http.Redirect(w, r, "/check", http.StatusMovedPermanently)
-		return
-	case "burn", "burn-in":
 		http.Redirect(w, r, "/load", http.StatusMovedPermanently)
 		return
-	case "benchmark":
-		http.Redirect(w, r, "/speed", http.StatusMovedPermanently)
+	case "burn-in":
+		http.Redirect(w, r, "/burn", http.StatusMovedPermanently)
+		return
+	case "speed", "endurance":
+		http.Redirect(w, r, "/benchmark", http.StatusMovedPermanently)
 		return
 	}
 	body := renderPage(page, h.opts)

audit/internal/webui/server_test.go

@@ -666,54 +666,64 @@ func TestTasksPageRendersOpenLinksAndPaginationControls(t *testing.T) {
 
 func TestToolsPageRendersNvidiaSelfHealSection(t *testing.T) {
 	handler := NewHandler(HandlerOptions{})
-	rec := httptest.NewRecorder()
-	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/tools", nil))
-	if rec.Code != http.StatusOK {
-		t.Fatalf("status=%d", rec.Code)
+
+	// /tools: only NVMe Block Format and Supermicro DMI remain
+	recTools := httptest.NewRecorder()
+	handler.ServeHTTP(recTools, httptest.NewRequest(http.MethodGet, "/tools", nil))
+	if recTools.Code != http.StatusOK {
+		t.Fatalf("tools status=%d", recTools.Code)
 	}
-	body := rec.Body.String()
-	if !strings.Contains(body, `NVIDIA Self Heal`) {
-		t.Fatalf("tools page missing nvidia self heal section: %s", body)
+	toolsBody := recTools.Body.String()
+	if !strings.Contains(toolsBody, `NVMe Block Format`) {
+		t.Fatalf("tools page missing nvme block format section: %s", toolsBody)
 	}
-	if !strings.Contains(body, `Restart GPU Drivers`) {
-		t.Fatalf("tools page missing restart gpu drivers button: %s", body)
+	if !strings.Contains(toolsBody, `/api/tools/nvme-formats`) || !strings.Contains(toolsBody, `/api/tools/nvme-format/run`) {
+		t.Fatalf("tools page missing nvme format api usage: %s", toolsBody)
 	}
-	if !strings.Contains(body, `nvidiaRestartDrivers()`) {
-		t.Fatalf("tools page missing nvidiaRestartDrivers action: %s", body)
+
+	// /settings: system install, support bundle, tool check, nvidia self heal, network, services
+	recSettings := httptest.NewRecorder()
+	handler.ServeHTTP(recSettings, httptest.NewRequest(http.MethodGet, "/settings", nil))
+	if recSettings.Code != http.StatusOK {
+		t.Fatalf("settings status=%d", recSettings.Code)
 	}
-	if !strings.Contains(body, `/api/gpu/nvidia-status`) {
-		t.Fatalf("tools page missing nvidia status api usage: %s", body)
+	settingsBody := recSettings.Body.String()
+	if !strings.Contains(settingsBody, `NVIDIA Self Heal`) {
+		t.Fatalf("settings page missing nvidia self heal section: %s", settingsBody)
 	}
-	if !strings.Contains(body, `nvidiaResetGPU(`) {
-		t.Fatalf("tools page missing nvidiaResetGPU action: %s", body)
+	if !strings.Contains(settingsBody, `Restart GPU Drivers`) {
+		t.Fatalf("settings page missing restart gpu drivers button: %s", settingsBody)
 	}
-	if !strings.Contains(body, `id="boot-source-text"`) {
-		t.Fatalf("tools page missing boot source field: %s", body)
+	if !strings.Contains(settingsBody, `nvidiaRestartDrivers()`) {
+		t.Fatalf("settings page missing nvidiaRestartDrivers action: %s", settingsBody)
 	}
-	if !strings.Contains(body, `USB Black-Box`) {
-		t.Fatalf("tools page missing usb black-box section: %s", body)
+	if !strings.Contains(settingsBody, `/api/gpu/nvidia-status`) {
+		t.Fatalf("settings page missing nvidia status api usage: %s", settingsBody)
 	}
-	if !strings.Contains(body, `/api/blackbox/status`) {
-		t.Fatalf("tools page missing black-box status api usage: %s", body)
+	if !strings.Contains(settingsBody, `nvidiaResetGPU(`) {
+		t.Fatalf("settings page missing nvidiaResetGPU action: %s", settingsBody)
 	}
-	if !strings.Contains(body, `NVMe Block Format`) {
-		t.Fatalf("tools page missing nvme block format section: %s", body)
+	if !strings.Contains(settingsBody, `id="boot-source-text"`) {
+		t.Fatalf("settings page missing boot source field: %s", settingsBody)
 	}
-	if !strings.Contains(body, `/api/tools/nvme-formats`) || !strings.Contains(body, `/api/tools/nvme-format/run`) {
-		t.Fatalf("tools page missing nvme format api usage: %s", body)
+	if !strings.Contains(settingsBody, `USB Black-Box`) {
+		t.Fatalf("settings page missing usb black-box section: %s", settingsBody)
+	}
+	if !strings.Contains(settingsBody, `/api/blackbox/status`) {
+		t.Fatalf("settings page missing black-box status api usage: %s", settingsBody)
 	}
 }
 
 func TestBenchmarkPageRendersGPUSelectionControls(t *testing.T) {
 	handler := NewHandler(HandlerOptions{})
 	rec := httptest.NewRecorder()
-	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/speed", nil))
+	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/benchmark", nil))
 	if rec.Code != http.StatusOK {
 		t.Fatalf("status=%d", rec.Code)
 	}
 	body := rec.Body.String()
 	for _, needle := range []string{
-		`href="/speed"`,
+		`href="/benchmark"`,
 		`id="benchmark-gpu-list"`,
 		`/api/gpu/nvidia`,
 		`/api/bee-bench/nvidia/perf/run`,
@@ -769,7 +779,7 @@ func TestBenchmarkPageRendersSavedResultsTable(t *testing.T) {
 
 	handler := NewHandler(HandlerOptions{ExportDir: exportDir})
 	rec := httptest.NewRecorder()
-	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/speed", nil))
+	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/benchmark", nil))
 	if rec.Code != http.StatusOK {
 		t.Fatalf("status=%d", rec.Code)
 	}
@@ -834,10 +844,10 @@ func TestCheckPageRendersNvidiaFabricCards(t *testing.T) {
 	}
 }
 
-func TestLoadPageRendersGoalBasedNVIDIACards(t *testing.T) {
+func TestBurnPageRendersGoalBasedNVIDIACards(t *testing.T) {
 	handler := NewHandler(HandlerOptions{})
 	rec := httptest.NewRecorder()
-	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/load", nil))
+	handler.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/burn", nil))
 	if rec.Code != http.StatusOK {
 		t.Fatalf("status=%d", rec.Code)
 	}
@@ -1217,7 +1227,8 @@ func TestDashboardRendersRuntimeHealthTable(t *testing.T) {
   ],
   "services":[
     {"name":"bee-web","status":"active"},
-    {"name":"bee-nvidia","status":"inactive"}
+    {"name":"bee-audit","status":"inactive"},
+    {"name":"bee-nvidia","status":"failed"}
   ]
 }`
 	if err := os.WriteFile(filepath.Join(exportDir, "runtime-health.json"), []byte(health), 0644); err != nil {
@@ -1271,7 +1282,7 @@ func TestDashboardRendersRuntimeHealthTable(t *testing.T) {
 		`Bee Services`,
 		`CUDA runtime is not ready for GPU SAT.`,
 		`Missing: nvidia-smi`,
-		`bee-nvidia=inactive`,
+		`bee-nvidia=failed`,
 		// Hardware Summary card — component health badges
 		`Hardware Summary`,
 		`>CPU<`,

audit/internal/webui/task_report.go

@@ -232,6 +232,9 @@ func renderTaskReportFragment(report taskReport, charts map[string]string, logTe
 	if powerCard := renderTaskPowerResultsCard(report.Target, logText); powerCard != "" {
 		b.WriteString(powerCard)
 	}
+	if report.Target == "storage" {
+		b.WriteString(renderStorageDiskReportCards(logText))
+	}
 
 	if len(report.Charts) > 0 {
 		for _, chart := range report.Charts {
@@ -369,3 +372,60 @@ func formatTaskDuration(sec int) string {
 	}
 	return fmt.Sprintf("%dh %02dm %02ds", sec/3600, (sec%3600)/60, sec%60)
 }
+
+// renderStorageDiskReportCards reads disk-*-report.txt files from the storage
+// SAT run directory and renders one card per disk.
+func renderStorageDiskReportCards(logText string) string {
+	runDir := taskStorageRunDirFromLog(logText)
+	if runDir == "" {
+		return ""
+	}
+	entries, err := os.ReadDir(runDir)
+	if err != nil {
+		return ""
+	}
+
+	var cards []string
+	for _, entry := range entries {
+		name := entry.Name()
+		if !strings.HasPrefix(name, "disk-") || !strings.HasSuffix(name, "-report.txt") {
+			continue
+		}
+		data, err := os.ReadFile(filepath.Join(runDir, name))
+		if err != nil || len(data) == 0 {
+			continue
+		}
+		// Extract disk label from filename: "disk-01-nvme0n1-report.txt" → "Disk 01 — nvme0n1"
+		stem := strings.TrimPrefix(strings.TrimSuffix(name, "-report.txt"), "disk-")
+		// stem is like "01-nvme0n1"
+		parts := strings.SplitN(stem, "-", 2)
+		title := "Disk " + stem
+		if len(parts) == 2 {
+			title = "Disk " + parts[0] + " — " + parts[1]
+		}
+		card := `<div class="card">` +
+			`<div class="card-head">` + html.EscapeString(title) + `</div>` +
+			`<div class="card-body" style="padding:0">` +
+			`<pre style="margin:0;padding:16px;font-size:12px;line-height:1.6;overflow-x:auto;white-space:pre">` +
+			html.EscapeString(string(data)) +
+			`</pre></div></div>`
+		cards = append(cards, card)
+	}
+	return strings.Join(cards, "\n")
+}
+
+// taskStorageRunDirFromLog finds the storage SAT run directory path logged as
+// "Archive: /path/to/storage-YYYYMMDD-HHMMSS".
+func taskStorageRunDirFromLog(logText string) string {
+	for _, line := range strings.Split(logText, "\n") {
+		line = strings.TrimSpace(line)
+		if !strings.HasPrefix(line, "Archive:") {
+			continue
+		}
+		path := strings.TrimSpace(strings.TrimPrefix(line, "Archive:"))
+		if strings.Contains(filepath.Base(path), "storage-") && !strings.HasSuffix(path, ".tar.gz") {
+			return path
+		}
+	}
+	return ""
+}

audit/internal/webui/task_runner.go

@@ -272,6 +272,12 @@ func executeTaskWithOptions(opts *HandlerOptions, t *Task, j *jobState, ctx cont
 			break
 		}
 		archive, err = runStorageAcceptancePackCtx(a, ctx, "", t.params.StressMode, j.append)
+	case "confidential-computing":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
+		archive, err = runConfidentialComputingCheckPackCtx(a, ctx, "", j.append)
 	case "cpu":
 		if a == nil {
 			err = fmt.Errorf("app not configured")
@@ -388,6 +394,40 @@ func executeTaskWithOptions(opts *HandlerOptions, t *Task, j *jobState, ctx cont
 			break
 		}
 		err = runSAADMIWriteTask(ctx, j, opts.ExportDir, t.params)
+	case "ipmi-fru-write":
+		if len(t.params.FRUChanges) == 0 {
+			err = fmt.Errorf("no changes provided")
+			break
+		}
+		err = runIPMIFRUWriteTask(ctx, j, opts.ExportDir, t.params)
+	case "huawei-elabel-write":
+		if len(t.params.HuaweiElabelChanges) == 0 {
+			err = fmt.Errorf("no changes provided")
+			break
+		}
+		err = runHuaweiElabelWriteTask(ctx, j, t.params)
+	case "raid-foreign-clear":
+		err = runRAIDForeignClearTask(ctx, j, t.params.RAIDController)
+	case "raid-foreign-import":
+		err = runRAIDForeignImportTask(ctx, j, t.params.RAIDController)
+	case "raid-lsi-create-mirror":
+		if len(t.params.RAIDDevices) < 2 {
+			err = fmt.Errorf("at least 2 drives required")
+			break
+		}
+		err = runRAIDLSICreateMirrorTask(ctx, j, t.params.RAIDController, t.params.RAIDDevices)
+	case "raid-lsi-prepare-drive":
+		if strings.TrimSpace(t.params.RAIDSlot) == "" {
+			err = fmt.Errorf("no drive slot provided")
+			break
+		}
+		err = runRAIDPrepareDriveTask(ctx, j, t.params.RAIDController, t.params.RAIDSlot)
+	case "raid-vroc-create-mirror":
+		if len(t.params.RAIDDevices) < 2 {
+			err = fmt.Errorf("at least 2 devices required")
+			break
+		}
+		err = runRAIDVROCCreateMirrorTask(ctx, j, t.params.RAIDDevices, t.params.RAIDArrayName)
 	default:
 		j.append("ERROR: unknown target: " + t.Target)
 		j.finish("unknown target")

audit/internal/webui/tasks.go

@@ -45,6 +45,7 @@ var taskNames = map[string]string{
 	"nvidia-stress":          "NVIDIA GPU Stress",
 	"memory":                 "Memory SAT",
 	"storage":                "Storage SAT",
+	"confidential-computing": "Confidential Computing Check",
 	"cpu":                    "CPU SAT",
 	"amd":                    "AMD GPU SAT",
 	"amd-mem":                "AMD GPU MEM Integrity",
@@ -140,7 +141,13 @@ type taskParams struct {
 	Device             string      `json:"device,omitempty"` // for install
 	LBAF               int         `json:"lbaf,omitempty"`
 	PlatformComponents []string    `json:"platform_components,omitempty"`
-	SAADmiChanges      []saaChange `json:"saa_dmi_changes,omitempty"`
+	SAADmiChanges         []saaChange      `json:"saa_dmi_changes,omitempty"`
+	FRUChanges            []fruChange      `json:"fru_changes,omitempty"`
+	HuaweiElabelChanges   []huaweiChange   `json:"huawei_elabel_changes,omitempty"`
+	RAIDController     int          `json:"raid_controller,omitempty"`
+	RAIDDevices        []string     `json:"raid_devices,omitempty"`
+	RAIDArrayName      string       `json:"raid_array_name,omitempty"`
+	RAIDSlot           string       `json:"raid_slot,omitempty"`
 }
 
 type persistedTask struct {
@@ -306,6 +313,9 @@ var (
 	runStorageAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, extended bool, logFunc func(string)) (string, error) {
 		return a.RunStorageAcceptancePackCtx(ctx, baseDir, extended, logFunc)
 	}
+	runConfidentialComputingCheckPackCtx = func(a *app.App, ctx context.Context, baseDir string, logFunc func(string)) (string, error) {
+		return a.RunConfidentialComputingCheckPackCtx(ctx, baseDir, logFunc)
+	}
 	runCPUAcceptancePackCtx = func(a *app.App, ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
 		return a.RunCPUAcceptancePackCtx(ctx, baseDir, durationSec, logFunc)
 	}
@@ -1019,6 +1029,12 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 			break
 		}
 		archive, err = runStorageAcceptancePackCtx(a, ctx, "", t.params.StressMode, j.append)
+	case "confidential-computing":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
+		archive, err = runConfidentialComputingCheckPackCtx(a, ctx, "", j.append)
 	case "cpu":
 		if a == nil {
 			err = fmt.Errorf("app not configured")

bible-local/README.md

@@ -13,6 +13,7 @@ Generic engineering rules live in `bible/rules/patterns/`.
 | `docs/hardware-ingest-contract.md` | Current Reanimator hardware ingest JSON contract |
 | `docs/validate-vs-burn.md` | Validate and Validate -> Stress hardware test policy |
 | `decisions/` | Architectural decision log, including read-only submodule policy |
+| `proposals/` | RFCs and contract change proposals for Reanimator Core |
 
 ## Validate Test Matrix
 

bible-local/backlog.md

@@ -1,5 +1,103 @@
 # Backlog
 
+## Сбор SFP-модулей
+
+**Статус:** не реализовано.
+
+### Источник данных
+
+`ethtool -m <iface>` / `ethtool --module-info <iface>` — читает EEPROM SFP/SFP+/QSFP28/QSFP-DD по стандарту MSA (SFF-8472 / SFF-8636).
+
+Доступные поля из EEPROM:
+- Идентификатор модуля: `Identifier` (SFP, SFP+, QSFP28, …)
+- Тип коннектора: `Connector`
+- Вендор: `Vendor name`, `Vendor OUI`, `Vendor PN`, `Vendor SN`, `Vendor rev`
+- Оптика: `Wavelength`, `Transceiver type` (10GBase-SR, LR, DAC, …)
+- Телеметрия DOM (если модуль поддерживает): `Laser tx bias current`, `Transmit avg optical power`, `Receive avg optical power`, `Module temperature`, `Module voltage`
+- Статус: `Rx power high alarm`, `Tx power low warning`, …
+
+Для QSFP28 данные повторяются на 4 канала (lane 0–3).
+
+Инструмент требует root. На bee ISO — доступен (`ethtool` входит в образ).
+
+### Scope для bee
+
+1. Собирать список сетевых интерфейсов через `ip -j link show` (только `ether`, без `lo`/VLAN/bond).
+2. Для каждого интерфейса пробовать `ethtool -m <iface>`. Если модуль отсутствует или не поддерживает EEPROM read — тихо пропускать.
+3. Связывать интерфейс с PCIe-устройством через `ethtool -i <iface>` → поле `bus-info` (BDF) → сопоставление с `pcie_devices[].slot`.
+
+### Gap в контракте
+
+Текущий контракт v2.10 имеет в `pcie_devices[]` скалярные поля:
+- `sfp_temperature_c`, `sfp_tx_power_dbm`, `sfp_rx_power_dbm`, `sfp_voltage_v`, `sfp_bias_ma`
+
+Этого **недостаточно**:
+- Одна NIC-карта может иметь несколько портов — нужен массив, а не скаляр.
+- Нет полей идентификации модуля (vendor, part_number, serial_number, wavelength, connector).
+- Нет разбивки по каналам для QSFP28.
+
+### Предлагаемое расширение контракта
+
+Добавить в `pcie_devices[]` массив `sfp_modules[]`:
+
+```json
+"pcie_devices": [
+  {
+    "slot": "0000:3b:00.0",
+    "device_class": "EthernetController",
+    "model": "ConnectX-6 Dx",
+    "sfp_modules": [
+      {
+        "port":            0,
+        "identifier":      "QSFP28",
+        "connector":       "LC",
+        "vendor":          "Mellanox",
+        "part_number":     "MFA1A00-C003",
+        "serial_number":   "MT2124VS09999",
+        "revision":        "A",
+        "wavelength_nm":   850,
+        "transceiver_type": "100GBase-SR4",
+        "temperature_c":   36.4,
+        "voltage_v":       3.29,
+        "tx_power_dbm":   -1.8,
+        "rx_power_dbm":   -2.1,
+        "bias_ma":         7.2
+      }
+    ]
+  }
+]
+```
+
+Поля `sfp_modules[]`:
+
+| Поле | Тип | Описание |
+|---|---|---|
+| `port` | int | Номер порта на NIC (0-based) |
+| `identifier` | string | `SFP`, `SFP+`, `QSFP28`, `QSFP-DD`, … |
+| `connector` | string | `LC`, `MPO`, `DAC`, … |
+| `vendor` | string | Производитель модуля |
+| `part_number` | string | Партномер |
+| `serial_number` | string | Серийный номер |
+| `revision` | string | Ревизия |
+| `wavelength_nm` | int | Длина волны, нм |
+| `transceiver_type` | string | `10GBase-SR`, `100GBase-SR4`, `DAC`, … |
+| `temperature_c` | float | Температура модуля, °C |
+| `voltage_v` | float | Напряжение, В |
+| `tx_power_dbm` | float | TX оптическая мощность, dBm |
+| `rx_power_dbm` | float | RX оптическая мощность, dBm |
+| `bias_ma` | float | Bias current, мА |
+
+Старые скалярные поля `sfp_temperature_c` / `sfp_tx_power_dbm` / `sfp_rx_power_dbm` / `sfp_voltage_v` / `sfp_bias_ma` на уровне `pcie_devices[]` — **вывести из контракта** (deprecated), заменить на `sfp_modules[]`.
+
+### Порядок реализации
+
+1. Согласовать расширение контракта с Reanimator Core (bump до v2.11).
+2. Добавить `ethtool` parser в `audit/internal/collector/` — новый файл `sfp.go`.
+3. Дополнить schema в `audit/internal/schema/` типом `SFPModule`.
+4. Добавить `sfp_modules` в `PCIeDevice` в schema.
+5. Заполнять в NIC-коллекторе: связь интерфейс → BDF → `pcie_devices[].sfp_modules`.
+6. Показывать в TUI и web UI в разделе PCIe/NIC.
+
 ## BMC версия через IPMI
 
 **Статус:** реализовано.

bible-local/docs/hardware-ingest-contract.md

@@ -1,7 +1,7 @@
 ---
 title: Hardware Ingest JSON Contract
-version: "2.10"
-updated: "2026-04-29"
+version: "2.11"
+updated: "2026-06-19"
 maintainer: Reanimator Core
 audience: external-integrators, ai-agents
 language: ru
@@ -9,7 +9,7 @@ language: ru
 
 # Интеграция с Reanimator: контракт JSON-импорта аппаратного обеспечения
 
-Версия: **2.10** · Дата: **2026-04-29**
+Версия: **2.11** · Дата: **2026-06-19**
 
 Документ описывает формат JSON для передачи данных об аппаратном обеспечении серверов в систему **Reanimator** (управление жизненным циклом аппаратного обеспечения).
 Предназначен для разработчиков смежных систем (Redfish-коллекторов, агентов мониторинга, CMDB-экспортёров) и может быть включён в документацию интегрируемых проектов.
@@ -22,6 +22,7 @@ language: ru
 
 | Версия | Дата | Изменения |
 |--------|------|-----------|
+| 2.11 | 2026-06-19 | В `pcie_devices[]` добавлен необязательный массив `sfp_modules[]` с идентификацией и DOM telemetry SFP/QSFP-модулей. Скалярные поля `sfp_temperature_c` / `sfp_tx_power_dbm` / `sfp_rx_power_dbm` / `sfp_voltage_v` / `sfp_bias_ma` помечены как deprecated (принимаются, но `sfp_modules[]` имеет приоритет) |
 | 2.10 | 2026-04-29 | Для `hardware.storage[]` добавлены необязательные числовые поля `logical_block_size_bytes`, `physical_block_size_bytes`, `metadata_bytes_per_block` для нормализованного описания формата блока накопителя |
 | 2.9 | 2026-03-19 | Добавлена необязательная секция `hardware.platform_config` — произвольный объект с настройками платформы (BIOS/Redfish); хранится как latest-snapshot per machine |
 | 2.8 | 2026-03-15 | Поле `location` удалено из всех `sensors.*`; сенсоры передаются только по `name` и измеренным значениям |
@@ -422,11 +423,12 @@ GET /ingest/hardware/jobs/{job_id}
 | `battery_temperature_c` | float | нет | Температура батареи / supercap, °C |
 | `battery_voltage_v` | float | нет | Напряжение батареи / supercap, В |
 | `battery_replace_required` | bool | нет | Требуется замена батареи / supercap |
-| `sfp_temperature_c` | float | нет | Температура SFP/optic, °C |
-| `sfp_tx_power_dbm` | float | нет | TX optical power, dBm |
-| `sfp_rx_power_dbm` | float | нет | RX optical power, dBm |
-| `sfp_voltage_v` | float | нет | Напряжение SFP, В |
-| `sfp_bias_ma` | float | нет | Bias current SFP, мА |
+| `sfp_temperature_c` | float | нет | Температура SFP/optic, °C *(deprecated since 2.11)* |
+| `sfp_tx_power_dbm` | float | нет | TX optical power, dBm *(deprecated since 2.11)* |
+| `sfp_rx_power_dbm` | float | нет | RX optical power, dBm *(deprecated since 2.11)* |
+| `sfp_voltage_v` | float | нет | Напряжение SFP, В *(deprecated since 2.11)* |
+| `sfp_bias_ma` | float | нет | Bias current SFP, мА *(deprecated since 2.11)* |
+| `sfp_modules` | array | нет | Установленные SFP/QSFP-модули по портам (см. sfp_modules[]) |
 | `bdf` | string | нет | Deprecated alias для `slot`; при наличии ingest нормализует его в `slot` |
 | `device_class` | string | нет | Класс устройства (см. список ниже) |
 | `manufacturer` | string | нет | Производитель |
@@ -444,10 +446,43 @@ GET /ingest/hardware/jobs/{job_id}
 `numa_node` передавайте для NIC / InfiniBand / RAID / GPU, когда источник знает CPU/NUMA affinity. Поле сохраняется в snapshot-атрибутах PCIe-компонента и дублируется в telemetry для topology use cases.
 Поля `temperature_c` и `power_w` используйте для device-level telemetry GPU / accelerator / smart PCIe devices. Они не влияют на идентификацию компонента.
 
+**Deprecated поля sfp_\*:** Скалярные поля `sfp_temperature_c`, `sfp_tx_power_dbm`, `sfp_rx_power_dbm`, `sfp_voltage_v`, `sfp_bias_ma` продолжают приниматься, но помечены как deprecated since 2.11. Если в payload одновременно присутствуют `sfp_modules[]` и deprecated sfp_-скаляры — приоритет у `sfp_modules[]`, скаляры игнорируются. Deprecated поля будут удалены в версии 3.0.
+
 **Генерация serial_number при отсутствии или `"N/A"`:** `{board_serial}-PCIE-{slot}`, где `slot` для PCIe равен BDF.
 
 `slot` — единственный канонический адрес компонента. Для PCIe в `slot` передавайте BDF. Поле `bdf` сохраняется только как переходный alias на входе и не должно использоваться как отдельная координата рядом со `slot`.
 
+#### pcie_devices[].sfp_modules[]
+
+Необязательный массив установленных SFP/QSFP-модулей для данного PCIe-устройства. Один элемент — один порт. Используйте для многопортовых NIC (ConnectX-6 Dx, Intel X710, Mellanox HDR и др.).
+
+| Поле | Тип | Обязательно | Описание |
+|------|-----|-------------|----------|
+| `port` | int | **да** | Номер порта на NIC (0-based). Ключ дедупликации внутри устройства |
+| `identifier` | string | нет | Тип модуля: `SFP`, `SFP+`, `SFP28`, `QSFP+`, `QSFP28`, `QSFP-DD`, `DAC` |
+| `connector` | string | нет | Тип разъёма: `LC`, `MPO`, `RJ45`, `DAC`, `AOC`, `No separable connector` |
+| `vendor` | string | нет | Производитель модуля из EEPROM |
+| `part_number` | string | нет | Партномер из EEPROM |
+| `serial_number` | string | нет | Серийный номер из EEPROM |
+| `revision` | string | нет | Ревизия из EEPROM |
+| `wavelength_nm` | int | нет | Длина волны, нм (0 для DAC/медных кабелей) |
+| `transceiver_type` | string | нет | `10GBase-SR`, `10GBase-LR`, `25GBase-SR`, `100GBase-SR4`, `DAC`, … |
+| `temperature_c` | float | нет | Температура модуля, °C (DOM telemetry) |
+| `voltage_v` | float | нет | Напряжение питания, В (DOM telemetry) |
+| `tx_power_dbm` | float | нет | TX оптическая мощность, dBm (DOM telemetry) |
+| `rx_power_dbm` | float | нет | RX оптическая мощность, dBm (DOM telemetry) |
+| `bias_ma` | float | нет | Bias current, мА (DOM telemetry) |
+
+**Ключ дедупликации:** `(pcie_devices[].slot, sfp_modules[].port)`.
+
+**Правила ingest:**
+- При каждом импорте — полная замена `sfp_modules[]` для данного `pcie_devices[].slot` (upsert всего массива целиком).
+- Если `sfp_modules` отсутствует или `null` — существующие данные SFP не трогать.
+- Если `sfp_modules: []` (пустой массив) — трактовать как «модули не обнаружены», очистить сохранённые данные.
+- Дубли по `port` внутри одного `pcie_devices[]` — невалидны, endpoint возвращает `400` с описанием поля.
+- Модули без `serial_number` допустимы (многие DAC-кабели не имеют SN); сохраняются по ключу `(slot, port)`.
+- Изменение `serial_number` или `part_number` модуля на порту создаёт событие `COMPONENT_CHANGED` для PCIe-устройства с описанием «SFP module replaced on port N».
+
 **Значения `device_class`:**
 
 | Значение | Назначение |
@@ -472,16 +507,47 @@ GET /ingest/hardware/jobs/{job_id}
     "numa_node": 0,
     "temperature_c": 48.5,
     "power_w": 18.2,
-    "sfp_temperature_c": 36.2,
-    "sfp_tx_power_dbm": -1.8,
-    "sfp_rx_power_dbm": -2.1,
     "device_class": "EthernetController",
-    "manufacturer": "Intel",
-    "model": "X710 10GbE",
-    "serial_number": "K65472-003",
-    "firmware": "9.20 0x8000d4ae",
+    "manufacturer": "Mellanox",
+    "model": "ConnectX-6 Dx",
+    "serial_number": "MT2012X12345",
+    "firmware": "22.35.2010",
     "mac_addresses": ["3c:fd:fe:aa:bb:cc", "3c:fd:fe:aa:bb:cd"],
-    "status": "OK"
+    "status": "OK",
+    "sfp_modules": [
+      {
+        "port": 0,
+        "identifier": "QSFP28",
+        "connector": "LC",
+        "vendor": "Mellanox",
+        "part_number": "MFA1A00-C003",
+        "serial_number": "MT2124VS09999",
+        "revision": "A",
+        "wavelength_nm": 850,
+        "transceiver_type": "100GBase-SR4",
+        "temperature_c": 36.4,
+        "voltage_v": 3.29,
+        "tx_power_dbm": -1.8,
+        "rx_power_dbm": -2.1,
+        "bias_ma": 7.2
+      },
+      {
+        "port": 1,
+        "identifier": "QSFP28",
+        "connector": "LC",
+        "vendor": "Mellanox",
+        "part_number": "MFA1A00-C003",
+        "serial_number": "MT2124VS09998",
+        "revision": "A",
+        "wavelength_nm": 850,
+        "transceiver_type": "100GBase-SR4",
+        "temperature_c": 35.9,
+        "voltage_v": 3.28,
+        "tx_power_dbm": -1.9,
+        "rx_power_dbm": -2.3,
+        "bias_ma": 7.1
+      }
+    ]
   }
 ]
 ```
@@ -793,7 +859,24 @@ PSU без `serial_number` игнорируется.
         "model": "X710 10GbE",
         "serial_number": "K65472-003",
         "mac_addresses": ["3c:fd:fe:aa:bb:cc", "3c:fd:fe:aa:bb:cd"],
-        "status": "OK"
+        "status": "OK",
+        "sfp_modules": [
+          {
+            "port": 0,
+            "identifier": "SFP+",
+            "connector": "LC",
+            "vendor": "Intel",
+            "part_number": "FTLX8574D3BCV-IT",
+            "serial_number": "FNS123456789",
+            "wavelength_nm": 850,
+            "transceiver_type": "10GBase-SR",
+            "temperature_c": 34.1,
+            "voltage_v": 3.30,
+            "tx_power_dbm": -2.5,
+            "rx_power_dbm": -3.0,
+            "bias_ma": 6.8
+          }
+        ]
       }
     ],
     "power_supplies": [

bible-local/proposals/2026-06-19-sfp-modules-contract-rfc.md

@@ -0,0 +1,201 @@
+# RFC: SFP Module Identity and Telemetry in Hardware Ingest Contract
+
+**Date:** 2026-06-19
+**Status:** proposal
+**Target:** Reanimator Core — ingest schema, validation, and storage
+**Requested by:** bee (hardware audit LiveCD)
+**Contract version affected:** 2.10 → 2.11
+
+---
+
+## Задача для агента
+
+Внести изменения в контракт ingest и реализовать их в Reanimator Core:
+
+1. Добавить массив `sfp_modules[]` внутри `pcie_devices[]`.
+2. Обновить JSON-схему и валидатор ingest endpoint.
+3. Реализовать хранение/upsert данных SFP-модулей.
+4. Пометить скалярные поля `sfp_*` на уровне `pcie_devices[]` как deprecated.
+5. Обновить документ контракта (`hardware-ingest-contract.md`) до версии 2.11.
+
+Ниже — всё необходимое для принятия решений без дополнительных вопросов.
+
+---
+
+## Контекст: откуда берутся данные
+
+**Источник в bee:** `ethtool -m <iface>` (читает EEPROM SFP/SFP+/QSFP28/QSFP-DD по стандарту MSA SFF-8472 / SFF-8636).
+
+**Связь с PCIe:** `ethtool -i <iface>` возвращает `bus-info` = BDF (`0000:3b:00.0`), который совпадает с `pcie_devices[].slot`. Так bee связывает SFP-данные конкретного интерфейса с PCIe-устройством.
+
+**Один NIC — несколько модулей:** карта ConnectX-6 Dx (2 порта), Intel X710 (4 порта), Mellanox HDR (2 порта). Каждый порт — отдельный `ethtool -m`, отдельный SFP-модуль. Одного скаляра на устройство недостаточно.
+
+**QSFP28/QSFP-DD:** 4-канальные модули возвращают telemetry отдельно по каждому каналу (lane). В предложенной схеме lane-уровень не включён в первую версию — только агрегированные значения модуля в целом. Расширение до lane-уровня — отдельный RFC если понадобится.
+
+---
+
+## Проблема с текущим контрактом v2.10
+
+В `pcie_devices[]` есть пять скалярных полей:
+
+```
+sfp_temperature_c   float
+sfp_tx_power_dbm    float
+sfp_rx_power_dbm    float
+sfp_voltage_v       float
+sfp_bias_ma         float
+```
+
+Ограничения:
+- **Нет идентификации модуля** — vendor, part_number, serial_number, wavelength отсутствуют; модуль нельзя инвентаризировать как самостоятельный компонент.
+- **Только один набор значений на устройство** — невозможно описать 4-портовый NIC.
+- **Нет типа модуля** — SFP, QSFP28, DAC-кабель не различаются.
+- **Нет connector/transceiver_type** — невозможно понять, оптика это или медь.
+
+---
+
+## Предлагаемое изменение схемы
+
+### Новая структура `sfp_modules[]`
+
+Добавляется как необязательное поле внутри каждого объекта `pcie_devices[]`.
+
+```json
+"pcie_devices": [
+  {
+    "slot": "0000:3b:00.0",
+    "device_class": "EthernetController",
+    "model": "ConnectX-6 Dx",
+    "manufacturer": "Mellanox",
+    "serial_number": "MT2012X12345",
+    "status": "OK",
+    "sfp_modules": [
+      {
+        "port": 0,
+        "identifier": "QSFP28",
+        "connector": "LC",
+        "vendor": "Mellanox",
+        "part_number": "MFA1A00-C003",
+        "serial_number": "MT2124VS09999",
+        "revision": "A",
+        "wavelength_nm": 850,
+        "transceiver_type": "100GBase-SR4",
+        "temperature_c": 36.4,
+        "voltage_v": 3.29,
+        "tx_power_dbm": -1.8,
+        "rx_power_dbm": -2.1,
+        "bias_ma": 7.2
+      },
+      {
+        "port": 1,
+        "identifier": "QSFP28",
+        "connector": "LC",
+        "vendor": "Mellanox",
+        "part_number": "MFA1A00-C003",
+        "serial_number": "MT2124VS09998",
+        "revision": "A",
+        "wavelength_nm": 850,
+        "transceiver_type": "100GBase-SR4",
+        "temperature_c": 35.9,
+        "voltage_v": 3.28,
+        "tx_power_dbm": -1.9,
+        "rx_power_dbm": -2.3,
+        "bias_ma": 7.1
+      }
+    ]
+  }
+]
+```
+
+### Поля `sfp_modules[]`
+
+| Поле | Тип | Обязательно | Описание |
+|---|---|---|---|
+| `port` | int | **да** | Номер порта на NIC (0-based). Ключ дедупликации внутри устройства. |
+| `identifier` | string | нет | Тип модуля: `SFP`, `SFP+`, `SFP28`, `QSFP+`, `QSFP28`, `QSFP-DD`, `DAC` |
+| `connector` | string | нет | Тип разъёма: `LC`, `MPO`, `RJ45`, `DAC`, `AOC`, `No separable connector` |
+| `vendor` | string | нет | Производитель модуля из EEPROM |
+| `part_number` | string | нет | Партномер из EEPROM |
+| `serial_number` | string | нет | Серийный номер из EEPROM |
+| `revision` | string | нет | Ревизия из EEPROM |
+| `wavelength_nm` | int | нет | Длина волны, нм (0 для DAC/медных кабелей) |
+| `transceiver_type` | string | нет | `10GBase-SR`, `10GBase-LR`, `25GBase-SR`, `100GBase-SR4`, `DAC`, … |
+| `temperature_c` | float | нет | Температура модуля, °C (DOM telemetry) |
+| `voltage_v` | float | нет | Напряжение питания, В (DOM telemetry) |
+| `tx_power_dbm` | float | нет | TX оптическая мощность, dBm (DOM telemetry) |
+| `rx_power_dbm` | float | нет | RX оптическая мощность, dBm (DOM telemetry) |
+| `bias_ma` | float | нет | Bias current, мА (DOM telemetry) |
+
+**Ключ дедупликации:** `(pcie_device.slot, sfp_modules[].port)`.
+
+**Модули без серийного номера** — допустимы; многие DAC-кабели не имеют SN. Не игнорировать, сохранять по ключу `(slot, port)`.
+
+---
+
+## Deprecated поля
+
+Следующие поля на уровне `pcie_devices[]` помечаются как **deprecated** начиная с v2.11:
+
+```
+sfp_temperature_c
+sfp_tx_power_dbm
+sfp_rx_power_dbm
+sfp_voltage_v
+sfp_bias_ma
+```
+
+**Поведение при получении deprecated полей:**
+- Продолжать принимать и сохранять (не ломать существующих интеграторов).
+- Если одновременно присутствуют `sfp_modules[]` и deprecated скаляры — приоритет у `sfp_modules[]`; скаляры игнорируются.
+- В документации пометить как `deprecated since 2.11, will be removed in 3.0`.
+
+**Не удалять** deprecated поля из валидации в этом PR — только пометить в документации и changelog.
+
+---
+
+## Правила ingest для `sfp_modules[]`
+
+- `sfp_modules[]` хранится как snapshot-атрибут PCIe-компонента (аналогично `mac_addresses`).
+- При каждом импорте — полная замена `sfp_modules[]` для данного `pcie_devices[].slot` (upsert всего массива целиком, не merge по портам).
+- Если `sfp_modules` отсутствует или `null` — существующие данные SFP не трогать (не затирать).
+- Если `sfp_modules: []` (пустой массив) — трактовать как «модули не обнаружены», очистить сохранённые данные.
+- Изменение `serial_number` или `part_number` модуля на порту — создавать событие `COMPONENT_CHANGED` для PCIe-устройства с описанием «SFP module replaced on port N».
+
+---
+
+## Изменения в документе контракта
+
+Файл: `bible-local/docs/hardware-ingest-contract.md`
+
+1. Заголовок версии: `2.10` → `2.11`, дата → `2026-06-19`.
+2. Добавить в changelog:
+   ```
+   | 2.11 | 2026-06-19 | В `pcie_devices[]` добавлен необязательный массив `sfp_modules[]`
+   с идентификацией и DOM telemetry SFP/QSFP-модулей. Скалярные поля
+   sfp_temperature_c / sfp_tx_power_dbm / sfp_rx_power_dbm / sfp_voltage_v /
+   sfp_bias_ma помечены как deprecated (принимаются, но sfp_modules[] имеет приоритет). |
+   ```
+3. В секции `pcie_devices` добавить строку в таблицу полей:
+   ```
+   | `sfp_modules` | array | нет | Установленные SFP/QSFP-модули по портам (см. sfp_modules[]) |
+   ```
+4. Добавить подсекцию `#### pcie_devices[].sfp_modules[]` с таблицей полей и примером JSON (из раздела выше).
+5. Пометить deprecated поля в таблице: добавить суффикс `*(deprecated since 2.11)*`.
+6. Обновить полный пример JSON — добавить `sfp_modules` к NIC-записи в `pcie_devices`.
+
+---
+
+## Что не нужно делать в этом PR
+
+- Не добавлять lane-level данные QSFP (tx_power_dbm_lane_0 и т.п.) — отдельный RFC.
+- Не удалять deprecated поля — только пометить.
+- Не создавать отдельную top-level секцию `network_ports` — данные остаются вложенными в `pcie_devices`.
+- Не менять логику идентификации PCIe-компонента — `serial_number` SFP-модуля не является ключом для самостоятельного компонента.
+
+---
+
+## Валидация
+
+Единственное обязательное поле в `sfp_modules[]` — `port` (int, >= 0).  
+Все остальные поля опциональны.  
+Дубли по `port` внутри одного `pcie_devices[]` — невалидны, возвращать `400` с описанием поля.

iso/builder/build.sh

@@ -1419,6 +1419,13 @@ rm -rf \
 if [ "$BEE_GPU_VENDOR" != "nvidia" ]; then
     rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-nvidia-load"
     rm -f "${OVERLAY_STAGE_DIR}/etc/systemd/system/bee-nvidia.service"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-gpu-burn"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-john-gpu-stress"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-nccl-gpu-stress"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-nvidia-recover"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-dcgmproftester-staggered"
+    rm -f "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-check-nvswitch"
+    rm -rf "${OVERLAY_STAGE_DIR}/etc/systemd/system/nvidia-fabricmanager.service.d"
 fi
 
 # --- inject authorized_keys for SSH access ---
@@ -1473,7 +1480,7 @@ cp "${BUILDER_DIR}/smoketest.sh" "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-smokete
 chmod +x "${OVERLAY_STAGE_DIR}/usr/local/bin/bee-smoketest"
 
 # --- vendor utilities (optional pre-fetched binaries) ---
-for tool in storcli64 sas2ircu sas3ircu arcconf ssacli; do
+for tool in storcli64 sas2ircu sas3ircu arcconf ssacli saa; do
     if [ -f "${VENDOR_DIR}/${tool}" ]; then
         cp "${VENDOR_DIR}/${tool}" "${OVERLAY_STAGE_DIR}/usr/local/bin/${tool}"
         chmod +x "${OVERLAY_STAGE_DIR}/usr/local/bin/${tool}" || true
@@ -1483,6 +1490,17 @@ for tool in storcli64 sas2ircu sas3ircu arcconf ssacli; do
     fi
 done
 
+# saa companion directories — saa searches for these relative to CWD (/usr/local/bin)
+for saa_subdir in acpica_bin ExternalData tool stunnel GO_SNMP; do
+    if [ -d "${VENDOR_DIR}/${saa_subdir}" ]; then
+        cp -r "${VENDOR_DIR}/${saa_subdir}" "${OVERLAY_STAGE_DIR}/usr/local/bin/"
+        find "${OVERLAY_STAGE_DIR}/usr/local/bin/${saa_subdir}" -type f -exec chmod +x {} \; 2>/dev/null || true
+        echo "vendor saa: ${saa_subdir}/ (included)"
+    else
+        echo "vendor saa: ${saa_subdir}/ (not found, skipped)"
+    fi
+done
+
 # --- NVIDIA kernel modules and userspace libs ---
 if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
     run_step "build NVIDIA ${NVIDIA_DRIVER_VERSION} modules" "40-nvidia-module" \

iso/overlay/usr/local/bin/bee-selfheal

@@ -67,7 +67,8 @@ if ! mkdir "${LOCK_DIR}" 2>/dev/null; then
 fi
 trap 'rmdir "${LOCK_DIR}" >/dev/null 2>&1 || true' EXIT
 
-if have_nvidia_gpu && [ ! -e /dev/nvidia0 ]; then
+GPU_VENDOR=$(cat /etc/bee-gpu-vendor 2>/dev/null || echo "")
+if [ "$GPU_VENDOR" = "nvidia" ] && have_nvidia_gpu && [ ! -e /dev/nvidia0 ]; then
     log_event "NVIDIA GPU detected but /dev/nvidia0 is missing"
     restart_service bee-nvidia.service || true
 fi

iso/vendor/ExternalData/supportAutoDST

@@ -0,0 +1,37 @@
+(UTC-10:00) Aleutian Islands
+(UTC-09:00) Alaska
+(UTC-08:00) Baja California
+(UTC-08:00) Pacific Time (US & Canada)
+(UTC-07:00) Mountain Time (US & Canada)
+(UTC-06:00) Central Time (US & Canada)
+(UTC-06:00) Easter Island
+(UTC-05:00) Eastern Time (US & Canada)
+(UTC-05:00) Haiti
+(UTC-05:00) Havana
+(UTC-05:00) Indiana (East)
+(UTC-05:00) Turks and Caicos
+(UTC-04:00) Asuncion
+(UTC-04:00) Atlantic Time (Canada)
+(UTC-04:00) Santiago
+(UTC-03:30) Newfoundland
+(UTC-03:00) Saint Pierre and Miquelon
+(UTC-01:00) Azores
+(UTC+00:00) Dublin, Edinburgh, Lisbon, London
+(UTC+01:00) Casablanca
+(UTC+01:00) Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna
+(UTC+01:00) Belgrade, Bratislava, Budapest, Ljubljana, Prague
+(UTC+01:00) Brussels, Copenhagen, Madrid, Paris
+(UTC+01:00) Sarajevo, Skopje, Warsaw, Zagreb
+(UTC+02:00) Athens, Bucharest
+(UTC+02:00) Beirut
+(UTC+02:00) Chisinau
+(UTC+02:00) Gaza, Hebron
+(UTC+02:00) Helsinki, Kyiv, Riga, Sofia, Tallinn, Vilnius
+(UTC+02:00) Jerusalem
+(UTC+09:30) Adelaide
+(UTC+10:00) Canberra, Melbourne, Sydney
+(UTC+10:00) Hobart
+(UTC+10:30) Lord Howe Island
+(UTC+11:00) Norfolk Island
+(UTC+12:00) Auckland, Wellington
+(UTC+12:45) Chatham Islands

iso/vendor/ExternalData/timezone.txt

@@ -0,0 +1,139 @@
+(UTC-12:00) International Date Line West
+(UTC-11:00) Coordinated Universal Time-11
+(UTC-10:00) Aleutian Islands
+(UTC-10:00) Hawaii
+(UTC-09:30) Marquesas Islands
+(UTC-09:00) Alaska
+(UTC-09:00) Coordinated Universal Time-09
+(UTC-08:00) Baja California
+(UTC-08:00) Coordinated Universal Time-08
+(UTC-08:00) Pacific Time (US & Canada)
+(UTC-07:00) Arizona
+(UTC-07:00) Chihuahua, La Paz, Mazatlan
+(UTC-07:00) Mountain Time (US & Canada)
+(UTC-07:00) Yukon
+(UTC-06:00) Central America
+(UTC-06:00) Central Time (US & Canada)
+(UTC-06:00) Easter Island
+(UTC-06:00) Guadalajara, Mexico City, Monterrey
+(UTC-06:00) Saskatchewan
+(UTC-05:00) Bogota, Lima, Quito, Rio Branco
+(UTC-05:00) Chetumal
+(UTC-05:00) Eastern Time (US & Canada)
+(UTC-05:00) Haiti
+(UTC-05:00) Havana
+(UTC-05:00) Indiana (East)
+(UTC-05:00) Turks and Caicos
+(UTC-04:00) Atlantic Time (Canada)
+(UTC-04:00) Caracas
+(UTC-04:00) Cuiaba
+(UTC-04:00) Georgetown, La Paz, Manaus, San Juan
+(UTC-04:00) Santiago
+(UTC-03:30) Newfoundland
+(UTC-03:00) Asuncion
+(UTC-03:00) Araguaina
+(UTC-03:00) Brasilia
+(UTC-03:00) Cayenne, Fortaleza
+(UTC-03:00) City of Buenos Aires
+(UTC-03:00) Greenland
+(UTC-03:00) Montevideo
+(UTC-03:00) Punta Arenas
+(UTC-03:00) Saint Pierre and Miquelon
+(UTC-03:00) Salvador
+(UTC-02:00) Coordinated Universal Time-02
+(UTC-01:00) Azores
+(UTC-01:00) Cabo Verde Is.
+(UTC+00:00) Coordinated Universal Time
+(UTC+00:00) Dublin, Edinburgh, Lisbon, London
+(UTC+00:00) Monrovia, Reykjavik
+(UTC+00:00) Sao Tome
+(UTC+01:00) Casablanca
+(UTC+01:00) Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna
+(UTC+01:00) Belgrade, Bratislava, Budapest, Ljubljana, Prague
+(UTC+01:00) Brussels, Copenhagen, Madrid, Paris
+(UTC+01:00) Sarajevo, Skopje, Warsaw, Zagreb
+(UTC+01:00) West Central Africa
+(UTC+02:00) Amman
+(UTC+02:00) Athens, Bucharest
+(UTC+02:00) Beirut
+(UTC+02:00) Cairo
+(UTC+02:00) Chisinau
+(UTC+02:00) Damascus
+(UTC+02:00) Gaza, Hebron
+(UTC+02:00) Harare, Pretoria
+(UTC+02:00) Helsinki, Kyiv, Riga, Sofia, Tallinn, Vilnius
+(UTC+02:00) Jerusalem
+(UTC+02:00) Juba
+(UTC+02:00) Kaliningrad
+(UTC+02:00) Khartoum
+(UTC+02:00) Tripoli
+(UTC+02:00) Windhoek
+(UTC+03:00) Baghdad
+(UTC+03:00) Istanbul
+(UTC+03:00) Kuwait, Riyadh
+(UTC+03:00) Minsk
+(UTC+03:00) Moscow, St. Petersburg
+(UTC+03:00) Nairobi
+(UTC+03:00) Volgograd
+(UTC+03:30) Tehran
+(UTC+04:00) Abu Dhabi, Muscat
+(UTC+04:00) Astrakhan, Ulyanovsk
+(UTC+04:00) Baku
+(UTC+04:00) Izhevsk, Samara
+(UTC+04:00) Port Louis
+(UTC+04:00) Saratov
+(UTC+04:00) Tbilisi
+(UTC+04:00) Yerevan
+(UTC+04:30) Kabul
+(UTC+05:00) Ashgabat, Tashkent
+(UTC+05:00) Astana
+(UTC+05:00) Ekaterinburg
+(UTC+05:00) Islamabad, Karachi
+(UTC+05:00) Qyzylorda
+(UTC+05:30) Chennai, Kolkata, Mumbai, New Delhi
+(UTC+05:30) Sri Jayawardenepura
+(UTC+05:45) Kathmandu
+(UTC+06:00) Dhaka
+(UTC+06:00) Omsk
+(UTC+06:30) Yangon (Rangoon)
+(UTC+07:00) Bangkok, Hanoi, Jakarta
+(UTC+07:00) Barnaul, Gorno-Altaysk
+(UTC+07:00) Hovd
+(UTC+07:00) Krasnoyarsk
+(UTC+07:00) Novosibirsk
+(UTC+07:00) Tomsk
+(UTC+08:00) Beijing, Chongqing, Hong Kong, Urumqi
+(UTC+08:00) Irkutsk
+(UTC+08:00) Kuala Lumpur, Singapore
+(UTC+08:00) Perth
+(UTC+08:00) Taipei
+(UTC+08:00) Ulaanbaatar
+(UTC+08:45) Eucla
+(UTC+09:00) Chita
+(UTC+09:00) Osaka, Sapporo, Tokyo
+(UTC+09:00) Pyongyang
+(UTC+09:00) Seoul
+(UTC+09:00) Yakutsk
+(UTC+09:30) Adelaide
+(UTC+09:30) Darwin
+(UTC+10:00) Brisbane
+(UTC+10:00) Canberra, Melbourne, Sydney
+(UTC+10:00) Guam, Port Moresby
+(UTC+10:00) Hobart
+(UTC+10:00) Vladivostok
+(UTC+10:30) Lord Howe Island
+(UTC+11:00) Bougainville Island
+(UTC+11:00) Chokurdakh
+(UTC+11:00) Magadan
+(UTC+11:00) Norfolk Island
+(UTC+11:00) Sakhalin
+(UTC+11:00) Solomon Is., New Caledonia
+(UTC+12:00) Anadyr, Petropavlovsk-Kamchatsky
+(UTC+12:00) Auckland, Wellington
+(UTC+12:00) Coordinated Universal Time+12
+(UTC+12:00) Fiji
+(UTC+12:45) Chatham Islands
+(UTC+13:00) Coordinated Universal Time+13
+(UTC+13:00) Nuku'alofa
+(UTC+13:00) Samoa
+(UTC+14:00) Kiritimati Island