fix(iso): harden boot network bring-up - v3.20

audit/internal/app/support_bundle.go

@@ -36,6 +36,8 @@ var supportBundleCommands = []struct {
 	{name: "system/dmesg-tail.txt", cmd: []string{"sh", "-c", "dmesg | tail -n 200"}},
 }
 
+const supportBundleGlob = "bee-support-*.tar.gz"
+
 func BuildSupportBundle(exportDir string) (string, error) {
 	exportDir = strings.TrimSpace(exportDir)
 	if exportDir == "" {
@@ -86,34 +88,64 @@ func BuildSupportBundle(exportDir string) (string, error) {
 	return archivePath, nil
 }
 
+func LatestSupportBundlePath() (string, error) {
+	return latestSupportBundlePath(os.TempDir())
+}
+
 func cleanupOldSupportBundles(dir string) error {
-	matches, err := filepath.Glob(filepath.Join(dir, "bee-support-*.tar.gz"))
+	matches, err := filepath.Glob(filepath.Join(dir, supportBundleGlob))
 	if err != nil {
 		return err
 	}
-	type entry struct {
-		path string
-		mod  time.Time
+	entries := supportBundleEntries(matches)
+	for path, mod := range entries {
+		if time.Since(mod) > 24*time.Hour {
+			_ = os.Remove(path)
+			delete(entries, path)
+		}
 	}
-	list := make([]entry, 0, len(matches))
+	ordered := orderSupportBundles(entries)
+	if len(ordered) > 3 {
+		for _, old := range ordered[3:] {
+			_ = os.Remove(old)
+		}
+	}
+	return nil
+}
+
+func latestSupportBundlePath(dir string) (string, error) {
+	matches, err := filepath.Glob(filepath.Join(dir, supportBundleGlob))
+	if err != nil {
+		return "", err
+	}
+	ordered := orderSupportBundles(supportBundleEntries(matches))
+	if len(ordered) == 0 {
+		return "", os.ErrNotExist
+	}
+	return ordered[0], nil
+}
+
+func supportBundleEntries(matches []string) map[string]time.Time {
+	entries := make(map[string]time.Time, len(matches))
 	for _, match := range matches {
 		info, err := os.Stat(match)
 		if err != nil {
 			continue
 		}
-		if time.Since(info.ModTime()) > 24*time.Hour {
-			_ = os.Remove(match)
-			continue
-		}
-		list = append(list, entry{path: match, mod: info.ModTime()})
+		entries[match] = info.ModTime()
 	}
-	sort.Slice(list, func(i, j int) bool { return list[i].mod.After(list[j].mod) })
-	if len(list) > 3 {
-		for _, old := range list[3:] {
-			_ = os.Remove(old.path)
-		}
+	return entries
+}
+
+func orderSupportBundles(entries map[string]time.Time) []string {
+	ordered := make([]string, 0, len(entries))
+	for path := range entries {
+		ordered = append(ordered, path)
 	}
-	return nil
+	sort.Slice(ordered, func(i, j int) bool {
+		return entries[ordered[i]].After(entries[ordered[j]])
+	})
+	return ordered
 }
 
 func writeJournalDump(dst string) error {

audit/internal/webui/api.go

@@ -2,7 +2,6 @@ package webui
 
 import (
 	"bufio"
-	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -87,15 +86,16 @@ func streamJob(w http.ResponseWriter, r *http.Request, j *jobState) {
 	}
 }
 
-// runCmdJob runs an exec.Cmd as a background job, streaming stdout+stderr lines.
-func runCmdJob(j *jobState, cmd *exec.Cmd) {
+// streamCmdJob runs an exec.Cmd and streams stdout+stderr lines into j.
+func streamCmdJob(j *jobState, cmd *exec.Cmd) error {
 	pr, pw := io.Pipe()
 	cmd.Stdout = pw
 	cmd.Stderr = pw
 
 	if err := cmd.Start(); err != nil {
-		j.finish(err.Error())
-		return
+		_ = pw.Close()
+		_ = pr.Close()
+		return err
 	}
 	// Lower the CPU scheduling priority of stress/audit subprocesses to nice+10
 	// so the X server and kernel interrupt handling remain responsive under load
@@ -104,8 +104,10 @@ func runCmdJob(j *jobState, cmd *exec.Cmd) {
 		_ = syscall.Setpriority(syscall.PRIO_PROCESS, cmd.Process.Pid, 10)
 	}
 
+	scanDone := make(chan error, 1)
 	go func() {
 		scanner := bufio.NewScanner(pr)
+		scanner.Buffer(make([]byte, 0, 64*1024), 1024*1024)
 		for scanner.Scan() {
 			// Split on \r to handle progress-bar style output (e.g. \r overwrites)
 			// and strip ANSI escape codes so logs are readable in the browser.
@@ -117,15 +119,21 @@ func runCmdJob(j *jobState, cmd *exec.Cmd) {
 				}
 			}
 		}
+		if err := scanner.Err(); err != nil && !errors.Is(err, io.ErrClosedPipe) {
+			scanDone <- err
+			return
+		}
+		scanDone <- nil
 	}()
 
 	err := cmd.Wait()
 	_ = pw.Close()
+	scanErr := <-scanDone
+	_ = pr.Close()
 	if err != nil {
-		j.finish(err.Error())
-	} else {
-		j.finish("")
+		return err
 	}
+	return scanErr
 }
 
 // ── Audit ─────────────────────────────────────────────────────────────────────
@@ -417,15 +425,23 @@ func (h *handler) handleAPIExportList(w http.ResponseWriter, r *http.Request) {
 }
 
 func (h *handler) handleAPIExportBundle(w http.ResponseWriter, r *http.Request) {
-	archive, err := app.BuildSupportBundle(h.opts.ExportDir)
-	if err != nil {
-		writeError(w, http.StatusInternalServerError, err.Error())
+	if globalQueue.hasActiveTarget("support-bundle") {
+		writeError(w, http.StatusConflict, "support bundle task is already pending or running")
 		return
 	}
+	t := &Task{
+		ID:        newJobID("support-bundle"),
+		Name:      "Support Bundle",
+		Target:    "support-bundle",
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+	}
+	globalQueue.enqueue(t)
 	writeJSON(w, map[string]string{
-		"status": "ok",
-		"path":   archive,
-		"url":    "/export/support.tar.gz",
+		"status":  "queued",
+		"task_id": t.ID,
+		"job_id":  t.ID,
+		"url":     "/export/support.tar.gz",
 	})
 }
 
@@ -513,10 +529,7 @@ func (h *handler) handleAPIInstallToRAM(w http.ResponseWriter, r *http.Request)
 		writeError(w, http.StatusServiceUnavailable, "app not configured")
 		return
 	}
-	h.installMu.Lock()
-	installRunning := h.installJob != nil && !h.installJob.isDone()
-	h.installMu.Unlock()
-	if installRunning {
+	if globalQueue.hasActiveTarget("install") {
 		writeError(w, http.StatusConflict, "install to disk is already running")
 		return
 	}
@@ -631,35 +644,23 @@ func (h *handler) handleAPIInstallRun(w http.ResponseWriter, r *http.Request) {
 		writeError(w, http.StatusConflict, "install to RAM task is already pending or running")
 		return
 	}
-
-	h.installMu.Lock()
-	if h.installJob != nil && !h.installJob.isDone() {
-		h.installMu.Unlock()
-		writeError(w, http.StatusConflict, "install already running")
+	if globalQueue.hasActiveTarget("install") {
+		writeError(w, http.StatusConflict, "install task is already pending or running")
 		return
 	}
-	j := &jobState{}
-	h.installJob = j
-	h.installMu.Unlock()
-
-	logFile := platform.InstallLogPath(req.Device)
-	go runCmdJob(j, exec.CommandContext(context.Background(), "bee-install", req.Device, logFile))
-
-	w.WriteHeader(http.StatusNoContent)
-}
-
-func (h *handler) handleAPIInstallStream(w http.ResponseWriter, r *http.Request) {
-	h.installMu.Lock()
-	j := h.installJob
-	h.installMu.Unlock()
-	if j == nil {
-		if !sseStart(w) {
-			return
-		}
-		sseWrite(w, "done", "")
-		return
+	t := &Task{
+		ID:        newJobID("install"),
+		Name:      "Install to Disk",
+		Target:    "install",
+		Priority:  20,
+		Status:    TaskPending,
+		CreatedAt: time.Now(),
+		params: taskParams{
+			Device: req.Device,
+		},
 	}
-	streamJob(w, r, j)
+	globalQueue.enqueue(t)
+	writeJSON(w, map[string]string{"task_id": t.ID, "job_id": t.ID})
 }
 
 // ── Metrics SSE ───────────────────────────────────────────────────────────────

audit/internal/webui/api_test.go

@@ -1,6 +1,7 @@
 package webui
 
 import (
+	"encoding/json"
 	"net/http/httptest"
 	"strings"
 	"testing"
@@ -62,3 +63,40 @@ func TestHandleAPISATRunDecodesBodyWithoutContentLength(t *testing.T) {
 		t.Fatalf("burn profile=%q want smoke", got)
 	}
 }
+
+func TestHandleAPIExportBundleQueuesTask(t *testing.T) {
+	globalQueue.mu.Lock()
+	originalTasks := globalQueue.tasks
+	globalQueue.tasks = nil
+	globalQueue.mu.Unlock()
+	t.Cleanup(func() {
+		globalQueue.mu.Lock()
+		globalQueue.tasks = originalTasks
+		globalQueue.mu.Unlock()
+	})
+
+	h := &handler{opts: HandlerOptions{ExportDir: t.TempDir()}}
+	req := httptest.NewRequest("POST", "/api/export/bundle", nil)
+	rec := httptest.NewRecorder()
+
+	h.handleAPIExportBundle(rec, req)
+
+	if rec.Code != 200 {
+		t.Fatalf("status=%d body=%s", rec.Code, rec.Body.String())
+	}
+	var body map[string]string
+	if err := json.Unmarshal(rec.Body.Bytes(), &body); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if body["task_id"] == "" {
+		t.Fatalf("missing task_id in response: %v", body)
+	}
+	globalQueue.mu.Lock()
+	defer globalQueue.mu.Unlock()
+	if len(globalQueue.tasks) != 1 {
+		t.Fatalf("tasks=%d want 1", len(globalQueue.tasks))
+	}
+	if got := globalQueue.tasks[0].Target; got != "support-bundle" {
+		t.Fatalf("target=%q want support-bundle", got)
+	}
+}

audit/internal/webui/metricsdb.go

@@ -4,6 +4,8 @@ import (
 	"database/sql"
 	"encoding/csv"
 	"io"
+	"os"
+	"path/filepath"
 	"strconv"
 	"time"
 
@@ -20,6 +22,9 @@ type MetricsDB struct {
 
 // openMetricsDB opens (or creates) the metrics database at the given path.
 func openMetricsDB(path string) (*MetricsDB, error) {
+	if err := os.MkdirAll(filepath.Dir(path), 0755); err != nil {
+		return nil, err
+	}
 	db, err := sql.Open("sqlite", path+"?_journal=WAL&_busy_timeout=5000")
 	if err != nil {
 		return nil, err
@@ -132,7 +137,7 @@ func (m *MetricsDB) loadSamples(query string, args ...any) ([]platform.LiveMetri
 	defer rows.Close()
 
 	type sysRow struct {
-		ts          int64
+		ts            int64
 		cpu, mem, pwr float64
 	}
 	var sysRows []sysRow
@@ -156,7 +161,10 @@ func (m *MetricsDB) loadSamples(query string, args ...any) ([]platform.LiveMetri
 	maxTS := sysRows[len(sysRows)-1].ts
 
 	// Load GPU rows in range
-	type gpuKey struct{ ts int64; idx int }
+	type gpuKey struct {
+		ts  int64
+		idx int
+	}
 	gpuData := map[gpuKey]platform.GPUMetricRow{}
 	gRows, err := m.db.Query(
 		`SELECT ts,gpu_index,temp_c,usage_pct,mem_usage_pct,power_w FROM gpu_metrics WHERE ts>=? AND ts<=? ORDER BY ts,gpu_index`,
@@ -174,7 +182,10 @@ func (m *MetricsDB) loadSamples(query string, args ...any) ([]platform.LiveMetri
 	}
 
 	// Load fan rows in range
-	type fanKey struct{ ts int64; name string }
+	type fanKey struct {
+		ts   int64
+		name string
+	}
 	fanData := map[fanKey]float64{}
 	fRows, err := m.db.Query(
 		`SELECT ts,name,rpm FROM fan_metrics WHERE ts>=? AND ts<=?`, minTS, maxTS,
@@ -192,7 +203,10 @@ func (m *MetricsDB) loadSamples(query string, args ...any) ([]platform.LiveMetri
 	}
 
 	// Load temp rows in range
-	type tempKey struct{ ts int64; name string }
+	type tempKey struct {
+		ts   int64
+		name string
+	}
 	tempData := map[tempKey]platform.TempReading{}
 	tRows, err := m.db.Query(
 		`SELECT ts,name,grp,celsius FROM temp_metrics WHERE ts>=? AND ts<=?`, minTS, maxTS,

audit/internal/webui/pages.go

@@ -926,7 +926,7 @@ func renderExport(exportDir string) string {
 	return `<div class="grid2">
 <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">Creates a tar.gz archive of all audit files, SAT results, and logs.</p>
-<a class="btn btn-primary" href="/export/support.tar.gz">&#8595; Download Support Bundle</a>
+` + renderSupportBundleInline() + `
 </div></div>
 <div class="card"><div class="card-head">Export Files</div><div class="card-body">
 <table><tr><th>File</th></tr>` + rows.String() + `</table>
@@ -1024,6 +1024,77 @@ func listExportFiles(exportDir string) ([]string, error) {
 	return entries, nil
 }
 
+func renderSupportBundleInline() string {
+	return `<button id="support-bundle-btn" class="btn btn-primary" onclick="supportBundleBuild()">Build Support Bundle</button>
+<a id="support-bundle-download" class="btn btn-secondary" href="/export/support.tar.gz" style="display:none">&#8595; Download Support Bundle</a>
+<div id="support-bundle-status" style="margin-top:12px;font-size:13px;color:var(--muted)">No support bundle built in this session.</div>
+<div id="support-bundle-log" class="terminal" style="display:none;margin-top:12px;max-height:260px"></div>
+<script>
+(function(){
+var _supportBundleES = null;
+window.supportBundleBuild = function() {
+  var btn = document.getElementById('support-bundle-btn');
+  var status = document.getElementById('support-bundle-status');
+  var log = document.getElementById('support-bundle-log');
+  var download = document.getElementById('support-bundle-download');
+  if (_supportBundleES) {
+    _supportBundleES.close();
+    _supportBundleES = null;
+  }
+  btn.disabled = true;
+  btn.textContent = 'Building...';
+  status.textContent = 'Queueing support bundle task...';
+  status.style.color = 'var(--muted)';
+  log.style.display = '';
+  log.textContent = '';
+  download.style.display = 'none';
+
+  fetch('/api/export/bundle', {method:'POST'}).then(function(r){
+    return r.json().then(function(j){
+      if (!r.ok) throw new Error(j.error || r.statusText);
+      return j;
+    });
+  }).then(function(data){
+    if (!data.task_id) throw new Error('missing task id');
+    status.textContent = 'Building support bundle...';
+    _supportBundleES = new EventSource('/api/tasks/' + data.task_id + '/stream');
+    _supportBundleES.onmessage = function(e) {
+      log.textContent += e.data + '\n';
+      log.scrollTop = log.scrollHeight;
+    };
+    _supportBundleES.addEventListener('done', function(e) {
+      _supportBundleES.close();
+      _supportBundleES = null;
+      btn.disabled = false;
+      btn.textContent = 'Build Support Bundle';
+      if (e.data) {
+        status.textContent = 'Error: ' + e.data;
+        status.style.color = 'var(--crit-fg)';
+        return;
+      }
+      status.textContent = 'Support bundle ready.';
+      status.style.color = 'var(--ok-fg)';
+      download.style.display = '';
+    });
+    _supportBundleES.onerror = function() {
+      if (_supportBundleES) _supportBundleES.close();
+      _supportBundleES = null;
+      btn.disabled = false;
+      btn.textContent = 'Build Support Bundle';
+      status.textContent = 'Support bundle stream disconnected.';
+      status.style.color = 'var(--crit-fg)';
+    };
+  }).catch(function(e){
+    btn.disabled = false;
+    btn.textContent = 'Build Support Bundle';
+    status.textContent = 'Error: ' + e;
+    status.style.color = 'var(--crit-fg)';
+  });
+};
+})();
+</script>`
+}
+
 // ── Display Resolution ────────────────────────────────────────────────────────
 
 func renderDisplayInline() string {
@@ -1113,7 +1184,7 @@ function installToRAM() {
 
 <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>
-<a class="btn btn-primary" href="/export/support.tar.gz">&#8595; Download Support Bundle</a>
+` + renderSupportBundleInline() + `
 </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>
@@ -1292,21 +1363,23 @@ function installStart() {
     headers: {'Content-Type': 'application/json'},
     body: JSON.stringify({device: _installSelected.device})
   }).then(function(r){
-    if (r.status === 204) {
-      installStreamLog();
-    } else {
-      return r.json().then(function(j){ throw new Error(j.error || r.statusText); });
-    }
+    return r.json().then(function(j){
+      if (!r.ok) throw new Error(j.error || r.statusText);
+      return j;
+    });
+  }).then(function(j){
+    if (!j.task_id) throw new Error('missing task id');
+    installStreamLog(j.task_id);
   }).catch(function(e){
     status.textContent = 'Error: ' + e;
     status.style.color = 'var(--crit-fg)';
   });
 }
 
-function installStreamLog() {
+function installStreamLog(taskId) {
   var term = document.getElementById('install-terminal');
   var status = document.getElementById('install-status');
-  var es = new EventSource('/api/install/stream');
+  var es = new EventSource('/api/tasks/' + taskId + '/stream');
   es.onmessage = function(e) {
     term.textContent += e.data + '\n';
     term.scrollTop = term.scrollHeight;

audit/internal/webui/server.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"html"
+	"log/slog"
 	"mime"
 	"net/http"
 	"os"
@@ -143,9 +144,6 @@ type handler struct {
 	latest   *platform.LiveMetricSample
 	// metrics persistence (nil if DB unavailable)
 	metricsDB *MetricsDB
-	// install job (at most one at a time)
-	installJob *jobState
-	installMu  sync.Mutex
 	// pending network change (rollback on timeout)
 	pendingNet   *pendingNetChange
 	pendingNetMu sync.Mutex
@@ -180,7 +178,11 @@ func NewHandler(opts HandlerOptions) http.Handler {
 			if len(samples) > 0 {
 				h.setLatestMetric(samples[len(samples)-1])
 			}
+		} else {
+			slog.Warn("metrics history unavailable", "path", metricsDBPath, "err", err)
 		}
+	} else {
+		slog.Warn("metrics db disabled", "path", metricsDBPath, "err", err)
 	}
 	h.startMetricsCollector()
 
@@ -266,7 +268,6 @@ func NewHandler(opts HandlerOptions) http.Handler {
 	// Install
 	mux.HandleFunc("GET /api/install/disks", h.handleAPIInstallDisks)
 	mux.HandleFunc("POST /api/install/run", h.handleAPIInstallRun)
-	mux.HandleFunc("GET /api/install/stream", h.handleAPIInstallStream)
 
 	// Metrics — SSE stream of live sensor data + server-side SVG charts + CSV export
 	mux.HandleFunc("GET /api/metrics/stream", h.handleAPIMetricsStream)
@@ -366,9 +367,13 @@ func (h *handler) handleRuntimeHealthJSON(w http.ResponseWriter, r *http.Request
 }
 
 func (h *handler) handleSupportBundleDownload(w http.ResponseWriter, r *http.Request) {
-	archive, err := app.BuildSupportBundle(h.opts.ExportDir)
+	archive, err := app.LatestSupportBundlePath()
 	if err != nil {
-		http.Error(w, fmt.Sprintf("build support bundle: %v", err), http.StatusInternalServerError)
+		if errors.Is(err, os.ErrNotExist) {
+			http.Error(w, "support bundle not built yet", http.StatusNotFound)
+			return
+		}
+		http.Error(w, fmt.Sprintf("locate support bundle: %v", err), http.StatusInternalServerError)
 		return
 	}
 	w.Header().Set("Cache-Control", "no-store")

audit/internal/webui/server_test.go

@@ -259,6 +259,17 @@ func TestSupportBundleEndpointReturnsArchive(t *testing.T) {
 	if err := os.WriteFile(filepath.Join(exportDir, "bee-audit.log"), []byte("audit log"), 0644); err != nil {
 		t.Fatal(err)
 	}
+	archive, err := os.CreateTemp(os.TempDir(), "bee-support-server-test-*.tar.gz")
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { _ = os.Remove(archive.Name()) })
+	if _, err := archive.WriteString("support-bundle"); err != nil {
+		t.Fatal(err)
+	}
+	if err := archive.Close(); err != nil {
+		t.Fatal(err)
+	}
 
 	handler := NewHandler(HandlerOptions{ExportDir: exportDir})
 	rec := httptest.NewRecorder()

audit/internal/webui/tasks.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net/http"
 	"os"
+	"os/exec"
 	"path/filepath"
 	"sort"
 	"strings"
@@ -40,6 +41,7 @@ var taskNames = map[string]string{
 	"sat-stress":      "SAT Stress (stressapptest)",
 	"platform-stress": "Platform Thermal Cycling",
 	"audit":           "Audit",
+	"support-bundle":  "Support Bundle",
 	"install":         "Install to Disk",
 	"install-to-ram":  "Install to RAM",
 }
@@ -213,6 +215,10 @@ var (
 	runSATStressPackCtx = func(a *app.App, ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
 		return a.RunSATStressPackCtx(ctx, baseDir, durationSec, logFunc)
 	}
+	buildSupportBundle = app.BuildSupportBundle
+	installCommand     = func(ctx context.Context, device string, logPath string) *exec.Cmd {
+		return exec.CommandContext(ctx, "bee-install", device, logPath)
+	}
 )
 
 // enqueue adds a task to the queue and notifies the worker.
@@ -410,9 +416,9 @@ func setCPUGovernor(governor string) {
 
 // runTask executes the work for a task, writing output to j.
 func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
-	if q.opts == nil || q.opts.App == nil {
-		j.append("ERROR: app not configured")
-		j.finish("app not configured")
+	if q.opts == nil {
+		j.append("ERROR: handler options not configured")
+		j.finish("handler options not configured")
 		return
 	}
 	a := q.opts.App
@@ -429,6 +435,10 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 
 	switch t.Target {
 	case "nvidia":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		diagLevel := t.params.DiagLevel
 		if t.params.BurnProfile != "" && diagLevel <= 0 {
 			diagLevel = resolveBurnPreset(t.params.BurnProfile).NvidiaDiag
@@ -446,6 +456,10 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 			archive, err = a.RunNvidiaAcceptancePack("", j.append)
 		}
 	case "nvidia-stress":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		dur := t.params.Duration
 		if t.params.BurnProfile != "" && dur <= 0 {
 			dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
@@ -457,10 +471,22 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 			ExcludeGPUIndices: t.params.ExcludeGPUIndices,
 		}, j.append)
 	case "memory":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		archive, err = runMemoryAcceptancePackCtx(a, ctx, "", j.append)
 	case "storage":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		archive, err = runStorageAcceptancePackCtx(a, ctx, "", j.append)
 	case "cpu":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		dur := t.params.Duration
 		if t.params.BurnProfile != "" && dur <= 0 {
 			dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
@@ -471,33 +497,65 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 		j.append(fmt.Sprintf("CPU stress duration: %ds", dur))
 		archive, err = runCPUAcceptancePackCtx(a, ctx, "", dur, j.append)
 	case "amd":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		archive, err = runAMDAcceptancePackCtx(a, ctx, "", j.append)
 	case "amd-mem":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		archive, err = runAMDMemIntegrityPackCtx(a, ctx, "", j.append)
 	case "amd-bandwidth":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		archive, err = runAMDMemBandwidthPackCtx(a, ctx, "", j.append)
 	case "amd-stress":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		dur := t.params.Duration
 		if t.params.BurnProfile != "" && dur <= 0 {
 			dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
 		}
 		archive, err = runAMDStressPackCtx(a, ctx, "", dur, j.append)
 	case "memory-stress":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		dur := t.params.Duration
 		if t.params.BurnProfile != "" && dur <= 0 {
 			dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
 		}
 		archive, err = runMemoryStressPackCtx(a, ctx, "", dur, j.append)
 	case "sat-stress":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		dur := t.params.Duration
 		if t.params.BurnProfile != "" && dur <= 0 {
 			dur = resolveBurnPreset(t.params.BurnProfile).DurationSec
 		}
 		archive, err = runSATStressPackCtx(a, ctx, "", dur, j.append)
 	case "platform-stress":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		opts := resolvePlatformStressPreset(t.params.BurnProfile)
 		archive, err = a.RunPlatformStress(ctx, "", opts, j.append)
 	case "audit":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		result, e := a.RunAuditNow(q.opts.RuntimeMode)
 		if e != nil {
 			err = e
@@ -506,7 +564,22 @@ func (q *taskQueue) runTask(t *Task, j *jobState, ctx context.Context) {
 				j.append(line)
 			}
 		}
+	case "support-bundle":
+		j.append("Building support bundle...")
+		archive, err = buildSupportBundle(q.opts.ExportDir)
+	case "install":
+		if strings.TrimSpace(t.params.Device) == "" {
+			err = fmt.Errorf("device is required")
+			break
+		}
+		installLogPath := platform.InstallLogPath(t.params.Device)
+		j.append("Install log: " + installLogPath)
+		err = streamCmdJob(j, installCommand(ctx, t.params.Device, installLogPath))
 	case "install-to-ram":
+		if a == nil {
+			err = fmt.Errorf("app not configured")
+			break
+		}
 		err = a.RunInstallToRAM(ctx, j.append)
 	default:
 		j.append("ERROR: unknown target: " + t.Target)

audit/internal/webui/tasks_test.go

@@ -3,7 +3,9 @@ package webui
 import (
 	"context"
 	"os"
+	"os/exec"
 	"path/filepath"
+	"strings"
 	"testing"
 	"time"
 
@@ -113,8 +115,6 @@ func TestTaskDisplayNameUsesNvidiaStressLoader(t *testing.T) {
 }
 
 func TestRunTaskHonorsCancel(t *testing.T) {
-	t.Parallel()
-
 	blocked := make(chan struct{})
 	released := make(chan struct{})
 	aRun := func(_ any, ctx context.Context, _ string, _ int, _ func(string)) (string, error) {
@@ -173,8 +173,6 @@ func TestRunTaskHonorsCancel(t *testing.T) {
 }
 
 func TestRunTaskUsesBurnProfileDurationForCPU(t *testing.T) {
-	t.Parallel()
-
 	var gotDuration int
 	q := &taskQueue{
 		opts: &HandlerOptions{App: &app.App{}},
@@ -202,3 +200,82 @@ func TestRunTaskUsesBurnProfileDurationForCPU(t *testing.T) {
 		t.Fatalf("duration=%d want %d", gotDuration, 5*60)
 	}
 }
+
+func TestRunTaskBuildsSupportBundleWithoutApp(t *testing.T) {
+	dir := t.TempDir()
+	q := &taskQueue{
+		opts: &HandlerOptions{ExportDir: dir},
+	}
+	tk := &Task{
+		ID:        "support-bundle-1",
+		Name:      "Support Bundle",
+		Target:    "support-bundle",
+		Status:    TaskRunning,
+		CreatedAt: time.Now(),
+	}
+	j := &jobState{}
+
+	var gotExportDir string
+	orig := buildSupportBundle
+	buildSupportBundle = func(exportDir string) (string, error) {
+		gotExportDir = exportDir
+		return filepath.Join(exportDir, "bundle.tar.gz"), nil
+	}
+	defer func() { buildSupportBundle = orig }()
+
+	q.runTask(tk, j, context.Background())
+
+	if gotExportDir != dir {
+		t.Fatalf("exportDir=%q want %q", gotExportDir, dir)
+	}
+	if j.err != "" {
+		t.Fatalf("unexpected error: %q", j.err)
+	}
+	if !strings.Contains(strings.Join(j.lines, "\n"), "Archive: "+filepath.Join(dir, "bundle.tar.gz")) {
+		t.Fatalf("lines=%v", j.lines)
+	}
+}
+
+func TestRunTaskInstallUsesSharedCommandStreaming(t *testing.T) {
+	q := &taskQueue{
+		opts: &HandlerOptions{},
+	}
+	tk := &Task{
+		ID:        "install-1",
+		Name:      "Install to Disk",
+		Target:    "install",
+		Status:    TaskRunning,
+		CreatedAt: time.Now(),
+		params:    taskParams{Device: "/dev/sda"},
+	}
+	j := &jobState{}
+
+	var gotDevice string
+	var gotLogPath string
+	orig := installCommand
+	installCommand = func(ctx context.Context, device string, logPath string) *exec.Cmd {
+		gotDevice = device
+		gotLogPath = logPath
+		return exec.CommandContext(ctx, "sh", "-c", "printf 'line1\nline2\n'")
+	}
+	defer func() { installCommand = orig }()
+
+	q.runTask(tk, j, context.Background())
+
+	if gotDevice != "/dev/sda" {
+		t.Fatalf("device=%q want /dev/sda", gotDevice)
+	}
+	if gotLogPath == "" {
+		t.Fatal("expected install log path")
+	}
+	logs := strings.Join(j.lines, "\n")
+	if !strings.Contains(logs, "Install log: ") {
+		t.Fatalf("missing install log line: %v", j.lines)
+	}
+	if !strings.Contains(logs, "line1") || !strings.Contains(logs, "line2") {
+		t.Fatalf("missing streamed output: %v", j.lines)
+	}
+	if j.err != "" {
+		t.Fatalf("unexpected error: %q", j.err)
+	}
+}

bible-local/decisions/2026-04-01-memtest-build-strategy.md

@@ -23,6 +23,63 @@ Current evidence from the archived `easy-bee-nvidia-v3.14-amd64` logs dated 2026
 
 So the assumption "live-build built-in memtest integration is enough on this stack" is currently false for this project until proven otherwise by a real built ISO.
 
+Additional evidence from the archived `easy-bee-nvidia-v3.17-dirty-amd64` logs dated 2026-04-01:
+
+- the build now completes successfully because memtest is non-blocking by default
+- `lb binary_memtest` still runs and installs `memtest86+`
+- the project-owned hook `config/hooks/normal/9100-memtest.hook.binary` does execute
+- but it executes too early for its current target paths:
+  - `binary/boot/grub/grub.cfg` is still missing at hook time
+  - `binary/isolinux/live.cfg` is still missing at hook time
+  - memtest binaries are also still absent in `binary/boot/`
+- later in the build, live-build does create intermediate bootloader configs with memtest lines in the workdir
+- but the final ISO still lacks memtest binaries and still lacks memtest lines in extracted ISO `boot/grub/grub.cfg` and `isolinux/live.cfg`
+
+So the assumption "the current normal binary hook path is late enough to patch final memtest artifacts" is also false.
+
+## Known Failed Attempts
+
+These approaches were already tried and should not be repeated blindly:
+
+1. Built-in live-build memtest only.
+Reason it failed:
+- `lb binary_memtest` runs, but the final ISO still misses memtest binaries and menu entries.
+
+2. Fixing only the memtest file names for Debian Bookworm.
+Reason it failed:
+- correct file names alone do not make the files appear in the final ISO.
+
+3. Copying memtest from `chroot/boot/` into `binary/boot/` via a binary hook.
+Reason it failed:
+- in this stack `chroot/boot/` is often empty for memtest payloads at the relevant time.
+
+4. Fallback extraction from cached `memtest86+` `.deb`.
+Reason it failed:
+- this was explored already and was not enough to stabilize the final ISO path end-to-end.
+
+5. Restoring explicit memtest menu entries in source bootloader templates only.
+Reason it failed:
+- memtest lines in source templates or intermediate workdir configs do not guarantee the final ISO contains them.
+
+6. Patching `binary/boot/grub/grub.cfg` and `binary/isolinux/live.cfg` from the current `config/hooks/normal/9100-memtest.hook.binary`.
+Reason it failed:
+- the hook runs before those files exist, so the hook cannot patch them there.
+
+## What This Means
+
+When revisiting memtest later, start from the constraints above rather than retrying the same patterns:
+
+- do not assume the built-in memtest stage is sufficient
+- do not assume `chroot/boot/` will contain memtest payloads
+- do not assume source bootloader templates are the last writer of final ISO configs
+- do not assume the current normal binary hook timing is late enough for final patching
+
+Any future memtest fix must explicitly identify:
+
+- where the memtest binaries are reliably available at build time
+- which exact build stage writes the final bootloader configs that land in the ISO
+- and a post-build proof from a real ISO, not only from intermediate workdir files
+
 ## Decision
 
 For `bee`, memtest must be treated as an explicit ISO artifact with explicit post-build validation.
@@ -44,12 +101,17 @@ Project rules from now on:
 Current implementation direction:
 
 - keep the live-build memtest stage enabled if it helps package acquisition
-- but enforce memtest explicitly in a project-owned binary hook
-- patch the generated `binary/boot/grub/grub.cfg` and `binary/isolinux/live.cfg` directly in the binary stage if memtest entries are missing
+- do not rely on the current early `binary_hooks` timing for final patching
+- prefer a post-`lb build` recovery step in `build.sh` that:
+  - patches the fully materialized `LB_DIR/binary` tree
+  - injects memtest binaries there
+  - ensures final bootloader entries there
+  - reruns late binary stages (`binary_checksums`, `binary_iso`, `binary_zsync`) after the patch
 
 ## Consequences
 
 - Future memtest changes must begin by reading this ADR and the commits listed above.
+- Future memtest changes must also begin by reading the failed-attempt list above.
 - We should stop re-introducing "prefer built-in live-build memtest" as a default assumption without new evidence.
 - Memtest validation in `build.sh` is not optional; it is the acceptance gate that prevents another silent regression.
 - If we change memtest strategy again, we must update this ADR with the exact build evidence that justified the change.

bible-local/docs/iso-build-rules.md

@@ -39,3 +39,17 @@ Rules:
   a real ISO.
 - If you reference memtest files manually, verify the exact package file list
   first for the target Debian release.
+
+Known bad loops for this repository:
+
+- Do not retry built-in-only memtest without new evidence. We already proved
+  that `lb binary_memtest` can run while the final ISO still has no memtest.
+- Do not assume fixing memtest file names is enough. Correct names did not fix
+  the final artifact path.
+- Do not assume `chroot/boot/` contains memtest payloads at the time hooks run.
+- Do not assume source `grub.cfg` / `live.cfg.in` are the final writers of ISO
+  bootloader configs.
+- Do not assume the current `config/hooks/normal/9100-memtest.hook.binary`
+  timing is late enough to patch final `binary/boot/grub/grub.cfg` or
+  `binary/isolinux/live.cfg`; logs from 2026-04-01 showed those files were not
+  present yet when the hook executed.

iso/builder/build.sh

@@ -272,6 +272,59 @@ memtest_fail() {
     return 0
 }
 
+iso_memtest_present() {
+    iso_path="$1"
+
+    [ -f "$iso_path" ] || return 1
+
+    if command -v bsdtar >/dev/null 2>&1; then
+        :
+    elif command -v xorriso >/dev/null 2>&1; then
+        :
+    else
+        return 1
+    fi
+
+    iso_list_files "$iso_path" | grep -q '^boot/memtest86+x64\.bin$' || return 1
+    iso_list_files "$iso_path" | grep -q '^boot/memtest86+x64\.efi$' || return 1
+
+    grub_cfg="$(mktemp)"
+    isolinux_cfg="$(mktemp)"
+
+    iso_extract_file "$iso_path" boot/grub/grub.cfg > "$grub_cfg" 2>/dev/null || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+    iso_extract_file "$iso_path" isolinux/live.cfg > "$isolinux_cfg" 2>/dev/null || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+
+    grep -q 'Memory Test (memtest86+)' "$grub_cfg" || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+    grep -q '/boot/memtest86+x64\.efi' "$grub_cfg" || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+    grep -q '/boot/memtest86+x64\.bin' "$grub_cfg" || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+    grep -q 'Memory Test (memtest86+)' "$isolinux_cfg" || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+    grep -q '/boot/memtest86+x64\.bin' "$isolinux_cfg" || {
+        rm -f "$grub_cfg" "$isolinux_cfg"
+        return 1
+    }
+
+    rm -f "$grub_cfg" "$isolinux_cfg"
+    return 0
+}
+
 validate_iso_memtest() {
     iso_path="$1"
     echo "=== validating memtest in ISO ==="
@@ -335,6 +388,127 @@ validate_iso_memtest() {
     echo "=== memtest validation OK ==="
 }
 
+append_memtest_grub_entry() {
+    grub_cfg="$1"
+    [ -f "$grub_cfg" ] || return 1
+    grep -q 'Memory Test (memtest86+)' "$grub_cfg" && return 0
+    grep -q '### BEE MEMTEST ###' "$grub_cfg" && return 0
+
+    cat >> "$grub_cfg" <<'EOF'
+
+### BEE MEMTEST ###
+if [ "${grub_platform}" = "efi" ]; then
+    menuentry "Memory Test (memtest86+)" {
+        chainloader /boot/memtest86+x64.efi
+    }
+else
+    menuentry "Memory Test (memtest86+)" {
+        linux16 /boot/memtest86+x64.bin
+    }
+fi
+### /BEE MEMTEST ###
+EOF
+}
+
+append_memtest_isolinux_entry() {
+    isolinux_cfg="$1"
+    [ -f "$isolinux_cfg" ] || return 1
+    grep -q 'Memory Test (memtest86+)' "$isolinux_cfg" && return 0
+    grep -q '### BEE MEMTEST ###' "$isolinux_cfg" && return 0
+
+    cat >> "$isolinux_cfg" <<'EOF'
+
+# ### BEE MEMTEST ###
+label memtest
+    menu label ^Memory Test (memtest86+)
+    linux /boot/memtest86+x64.bin
+# ### /BEE MEMTEST ###
+EOF
+}
+
+copy_memtest_from_deb() {
+    deb="$1"
+    dst_boot="$2"
+    tmpdir="$(mktemp -d)"
+
+    dpkg-deb -x "$deb" "$tmpdir"
+    for f in memtest86+x64.bin memtest86+x64.efi; do
+        if [ -f "$tmpdir/boot/$f" ]; then
+            cp "$tmpdir/boot/$f" "$dst_boot/$f"
+        fi
+    done
+    rm -rf "$tmpdir"
+}
+
+recover_iso_memtest() {
+    lb_dir="$1"
+    iso_path="$2"
+    binary_boot="$lb_dir/binary/boot"
+    grub_cfg="$lb_dir/binary/boot/grub/grub.cfg"
+    isolinux_cfg="$lb_dir/binary/isolinux/live.cfg"
+
+    echo "=== attempting memtest recovery in binary tree ==="
+
+    mkdir -p "$binary_boot"
+
+    for root in \
+        "$lb_dir/chroot/boot" \
+        "/boot"; do
+        for f in memtest86+x64.bin memtest86+x64.efi; do
+            if [ ! -f "$binary_boot/$f" ] && [ -f "$root/$f" ]; then
+                cp "$root/$f" "$binary_boot/$f"
+                echo "memtest recovery: copied $f from $root"
+            fi
+        done
+    done
+
+    if [ ! -f "$binary_boot/memtest86+x64.bin" ] || [ ! -f "$binary_boot/memtest86+x64.efi" ]; then
+        for dir in \
+            "$lb_dir/cache/packages.binary" \
+            "$lb_dir/cache/packages.chroot" \
+            "$lb_dir/chroot/var/cache/apt/archives" \
+            "${BEE_CACHE_DIR:-${DIST_DIR}/cache}/lb-packages" \
+            "/var/cache/apt/archives"; do
+            [ -d "$dir" ] || continue
+            deb="$(find "$dir" -maxdepth 1 -type f -name 'memtest86+*.deb' 2>/dev/null | head -1)"
+            [ -n "$deb" ] || continue
+            echo "memtest recovery: extracting payload from $deb"
+            copy_memtest_from_deb "$deb" "$binary_boot"
+            break
+        done
+    fi
+
+    if [ ! -f "$binary_boot/memtest86+x64.bin" ] || [ ! -f "$binary_boot/memtest86+x64.efi" ]; then
+        tmpdl="$(mktemp -d)"
+        if (
+            cd "$tmpdl" && apt-get download memtest86+ >/dev/null 2>&1
+        ); then
+            deb="$(find "$tmpdl" -maxdepth 1 -type f -name 'memtest86+*.deb' 2>/dev/null | head -1)"
+            if [ -n "$deb" ]; then
+                echo "memtest recovery: downloaded $deb"
+                copy_memtest_from_deb "$deb" "$binary_boot"
+            fi
+        fi
+        rm -rf "$tmpdl"
+    fi
+
+    if [ -f "$grub_cfg" ]; then
+        append_memtest_grub_entry "$grub_cfg" && echo "memtest recovery: ensured GRUB entry"
+    else
+        echo "memtest recovery: WARNING: missing $grub_cfg"
+    fi
+
+    if [ -f "$isolinux_cfg" ]; then
+        append_memtest_isolinux_entry "$isolinux_cfg" && echo "memtest recovery: ensured isolinux entry"
+    else
+        echo "memtest recovery: WARNING: missing $isolinux_cfg"
+    fi
+
+    run_optional_step_sh "rebuild live-build checksums after memtest recovery" "91-lb-checksums" "lb binary_checksums 2>&1"
+    run_optional_step_sh "rebuild ISO after memtest recovery" "92-lb-binary-iso" "rm -f '$iso_path' && lb binary_iso 2>&1"
+    run_optional_step_sh "rebuild zsync after memtest recovery" "93-lb-zsync" "lb binary_zsync 2>&1"
+}
+
 AUDIT_VERSION_EFFECTIVE="$(resolve_audit_version)"
 ISO_VERSION_EFFECTIVE="$(resolve_iso_version)"
 ISO_BASENAME="easy-bee-${BEE_GPU_VENDOR}-v${ISO_VERSION_EFFECTIVE}-amd64"
@@ -451,6 +625,32 @@ run_step_sh() {
     run_step "${step_name}" "${step_slug}" sh -c "${step_script}"
 }
 
+run_optional_step_sh() {
+    step_name="$1"
+    step_slug="$2"
+    step_script="$3"
+
+    if [ "${BEE_REQUIRE_MEMTEST:-0}" = "1" ]; then
+        run_step_sh "${step_name}" "${step_slug}" "${step_script}"
+        return 0
+    fi
+
+    step_log="${LOG_DIR}/${step_slug}.log"
+    echo ""
+    echo "=== optional step: ${step_name} ==="
+    echo "=== optional step log: ${step_log} ==="
+    set +e
+    sh -c "${step_script}" > "${step_log}" 2>&1
+    step_status=$?
+    set -e
+    cat "${step_log}"
+    if [ "${step_status}" -ne 0 ]; then
+        echo "WARNING: optional step failed: ${step_name} (see ${step_log})" >&2
+    else
+        echo "=== optional step OK: ${step_name} ==="
+    fi
+}
+
 start_build_log
 
 # Auto-detect kernel ABI: refresh apt index, then query current linux-image-amd64 dependency.
@@ -857,6 +1057,10 @@ fi
 ISO_RAW="${LB_DIR}/live-image-amd64.hybrid.iso"
 if [ -f "$ISO_RAW" ]; then
     dump_memtest_debug "post-build" "${LB_DIR}" "$ISO_RAW"
+    if ! iso_memtest_present "$ISO_RAW"; then
+        recover_iso_memtest "${LB_DIR}" "$ISO_RAW"
+        dump_memtest_debug "post-recovery" "${LB_DIR}" "$ISO_RAW"
+    fi
     validate_iso_memtest "$ISO_RAW"
     cp "$ISO_RAW" "$ISO_OUT"
     echo ""

iso/overlay/etc/systemd/system/bee-audit.service

@@ -1,9 +1,14 @@
 [Unit]
-Description=Bee: on-demand hardware audit (not started automatically)
+Description=Bee: hardware audit
+After=bee-preflight.service bee-network.service bee-nvidia.service
+Before=bee-web.service
 
 [Service]
 Type=oneshot
 RemainAfterExit=yes
-ExecStart=/bin/sh -c 'curl -sf -X POST http://localhost/api/audit/run >/dev/null'
+ExecStart=/usr/local/bin/bee-log-run /appdata/bee/export/bee-audit.log /usr/local/bin/bee audit --runtime auto --output file:/appdata/bee/export/bee-audit.json
 StandardOutput=journal
 StandardError=journal
+
+[Install]
+WantedBy=multi-user.target

iso/overlay/etc/systemd/system/bee-web.service

@@ -1,5 +1,6 @@
 [Unit]
 Description=Bee: hardware audit web viewer
+After=bee-audit.service
 
 [Service]
 Type=simple

iso/overlay/usr/local/bin/bee-network.sh

@@ -6,25 +6,66 @@ LOG_PREFIX="bee-network"
 
 log() { echo "[$LOG_PREFIX] $*"; }
 
-# find physical interfaces: exclude lo and virtual (docker/virbr/veth/tun/tap)
-interfaces=$(ip -o link show \
-    | awk -F': ' '{print $2}' \
-    | grep -v '^lo$' \
-    | grep -vE '^(docker|virbr|veth|tun|tap|br-|bond|dummy)' \
-    | sort)
+list_interfaces() {
+    ip -o link show \
+        | awk -F': ' '{print $2}' \
+        | grep -v '^lo$' \
+        | grep -vE '^(docker|virbr|veth|tun|tap|br-|bond|dummy)' \
+        | sort
+}
 
-if [ -z "$interfaces" ]; then
+# Give udev a short chance to expose late NICs before the first scan.
+if command -v udevadm >/dev/null 2>&1; then
+    udevadm settle --timeout=5 >/dev/null 2>&1 || log "WARN: udevadm settle timed out"
+fi
+
+started_ifaces=""
+started_count=0
+scan_pass=1
+
+# Some server NICs appear a bit later after module/firmware init. Do a small
+# bounded rescan window without turning network bring-up into a boot blocker.
+while [ "$scan_pass" -le 3 ]; do
+    interfaces=$(list_interfaces)
+
+    if [ -n "$interfaces" ]; then
+        for iface in $interfaces; do
+            case " $started_ifaces " in
+                *" $iface "*) continue ;;
+            esac
+
+            log "bringing up $iface"
+            if ! ip link set "$iface" up; then
+                log "WARN: could not bring up $iface"
+                continue
+            fi
+
+            carrier=$(cat "/sys/class/net/$iface/carrier" 2>/dev/null || true)
+            if [ "$carrier" = "1" ]; then
+                log "carrier detected on $iface"
+            else
+                log "carrier not detected yet on $iface"
+            fi
+
+            # DHCP in background — non-blocking, keep dhclient verbose output in the service log.
+            dhclient -4 -v -nw "$iface" &
+            log "DHCP started for $iface (pid $!)"
+
+            started_ifaces="$started_ifaces $iface"
+            started_count=$((started_count + 1))
+        done
+    fi
+
+    if [ "$scan_pass" -ge 3 ]; then
+        break
+    fi
+    scan_pass=$((scan_pass + 1))
+    sleep 2
+done
+
+if [ "$started_count" -eq 0 ]; then
     log "no physical interfaces found"
     exit 0
 fi
 
-for iface in $interfaces; do
-    log "bringing up $iface"
-    ip link set "$iface" up || { log "WARN: could not bring up $iface"; continue; }
-
-    # DHCP in background — non-blocking, keep dhclient verbose output in the service log.
-    dhclient -4 -v -nw "$iface" &
-    log "DHCP started for $iface (pid $!)"
-done
-
-log "done"
+log "done (interfaces started: $started_count)"