fix: surface CPU thermal throttling in status and SAT results

A CPU that had thermally throttled (sysfs thermal_throttle counter >
0) still reported status "OK" everywhere: dmidecode-derived CPU status
only distinguishes populated/enabled/disabled and never looked at the
throttle flag the collector already recorded next to it, and neither
SAT path meant to catch this actually could:

- The routine "cpu" SAT pack (RunCPUAcceptancePack) only checked
  lscpu/sensors/stress-ng exit codes — stress-ng exits 0 whether or
  not the CPU throttled while running it, so an 89°C/throttled CPU
  right after a "successful" run still showed cpu:all as OK in
  component-status.json.
- The more thorough platform-stress test already detected throttling
  and fan-spindown correctly, but wrote its verdict as "Overall: FAIL
  — ..." with no "=", which parseSATKV can't parse — so even a real
  detected throttle event never reached the component-status DB.

Fixes:
- cpu_telemetry.go: escalate a CPU's status to Warning (only-escalate,
  same severity ranking already used elsewhere) when Throttled is set.
- sat.go: add a before/after thermal-throttle-counter check job around
  the "cpu" pack's stress-ng run, so a throttle event during the run
  fails that job and (via the existing FAILED->Warning DB mapping)
  flips cpu:all to Warning.
- platform_stress.go: emit a machine-readable overall_status= line
  alongside the human-readable verdict so platform-stress results
  actually reach ApplySATResultToDB.

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

audit/internal/collector/cpu_telemetry.go

@@ -37,12 +37,36 @@ func enrichCPUsWithTelemetry(cpus []schema.HardwareCPU, doc sensorsDoc) []schema
 		}
 		if value, ok := throttleBySocket[socket]; ok {
 			cpus[i].Throttled = &value
+			if value {
+				escalateCPUThrottleStatus(&cpus[i])
+			}
 		}
 	}
 
 	return cpus
 }
 
+// escalateCPUThrottleStatus raises a CPU's status to at least Warning when it
+// has hit thermal throttling (cpuPackageThrottled: a cumulative since-boot
+// sysfs counter, not "is throttling right now"). dmidecode-derived status
+// (parseCPUStatus) only distinguishes populated/enabled/disabled and has no
+// way to know about a live thermal event, so without this a CPU that
+// throttled during e.g. a stress test still reports status "OK" — see
+// enrichCPUsWithTelemetry.
+func escalateCPUThrottleStatus(cpu *schema.HardwareCPU) {
+	current := ""
+	if cpu.Status != nil {
+		current = strings.TrimSpace(*cpu.Status)
+	}
+	if current != "" && current != statusUnknown && StatusSeverity(statusWarning) <= StatusSeverity(current) {
+		return
+	}
+	status := statusWarning
+	cpu.Status = &status
+	desc := "CPU hit thermal throttling (package/core throttle count > 0 since boot)"
+	cpu.ErrorDescription = &desc
+}
+
 func cpuTempsFromSensors(doc sensorsDoc, cpuCount int) map[int]float64 {
 	out := map[int]float64{}
 	if len(doc) == 0 {

audit/internal/collector/cpu_telemetry_test.go

@@ -49,6 +49,12 @@ func TestEnrichCPUsWithTelemetry(t *testing.T) {
 	if got[0].Throttled == nil || !*got[0].Throttled {
 		t.Fatalf("cpu0 throttled mismatch: %#v", got[0].Throttled)
 	}
+	if got[0].Status == nil || *got[0].Status != statusWarning {
+		t.Fatalf("cpu0 status not escalated to Warning on throttle: %#v", got[0].Status)
+	}
+	if got[0].ErrorDescription == nil || *got[0].ErrorDescription == "" {
+		t.Fatalf("cpu0 error description not set on throttle")
+	}
 	if got[1].TemperatureC == nil || *got[1].TemperatureC != 58.0 {
 		t.Fatalf("cpu1 temperature mismatch: %#v", got[1].TemperatureC)
 	}
@@ -58,6 +64,26 @@ func TestEnrichCPUsWithTelemetry(t *testing.T) {
 	if got[1].Throttled != nil && *got[1].Throttled {
 		t.Fatalf("cpu1 throttled mismatch: %#v", got[1].Throttled)
 	}
+	if got[1].Status == nil || *got[1].Status != statusOK {
+		t.Fatalf("cpu1 status should remain OK when not throttled: %#v", got[1].Status)
+	}
+}
+
+func TestEscalateCPUThrottleStatusDoesNotDowngradeCritical(t *testing.T) {
+	status := statusCritical
+	desc := "pre-existing critical finding"
+	cpu := schema.HardwareCPU{
+		HardwareComponentStatus: schema.HardwareComponentStatus{Status: &status, ErrorDescription: &desc},
+	}
+
+	escalateCPUThrottleStatus(&cpu)
+
+	if *cpu.Status != statusCritical {
+		t.Fatalf("status downgraded from Critical: %#v", cpu.Status)
+	}
+	if *cpu.ErrorDescription != desc {
+		t.Fatalf("error description overwritten: %#v", cpu.ErrorDescription)
+	}
 }
 
 func mustWriteFile(t *testing.T, path, content string) {

audit/internal/platform/platform_stress.go

@@ -331,12 +331,24 @@ func writePlatformSummary(opts PlatformStressOptions, analyses []cycleAnalysis)
 	}
 
 	fmt.Fprintf(&b, "%s\n", strings.Repeat("=", 48))
+	// overall_status is the machine-readable twin of the "Overall: ..." line
+	// below, in the key=value vocabulary ApplySATResultToDB/parseSATKV
+	// expects (OK/FAILED/PARTIAL/UNSUPPORTED). Without it, a throttle FAIL
+	// detected right here never reaches component-status.json: the DB writer
+	// only recognizes "overall_status=", and "Overall: FAIL — ..." has no
+	// "=" for parseSATKV to split on. FAILED is the only token that maps to
+	// DB status "Warning" (see satStatusToDBStatus), so both throttle and
+	// fan-spindown findings use it — there's no separate "WARN" token in the
+	// SAT summary vocabulary today.
 	if totalThrottle > 0 {
 		fmt.Fprintf(&b, "Overall: FAIL — throttle detected in %d/%d cycles\n", totalThrottle, len(analyses))
+		fmt.Fprintf(&b, "overall_status=FAILED\n")
 	} else if totalFanWarn > 0 {
 		fmt.Fprintf(&b, "Overall: WARN — fast fan spindown in %d/%d cycles (cooling recovery risk)\n", totalFanWarn, len(analyses))
+		fmt.Fprintf(&b, "overall_status=FAILED\n")
 	} else {
 		fmt.Fprintf(&b, "Overall: PASS\n")
+		fmt.Fprintf(&b, "overall_status=OK\n")
 	}
 	return b.String()
 }

audit/internal/platform/platform_stress_test.go

@@ -2,6 +2,7 @@ package platform
 
 import (
 	"runtime"
+	"strings"
 	"testing"
 )
 
@@ -32,3 +33,52 @@ func TestPlatformStressMemoryMBOverride(t *testing.T) {
 		t.Fatalf("platformStressMemoryMB=%d want 8192", got)
 	}
 }
+
+// TestWritePlatformSummaryOverallStatusKV verifies writePlatformSummary emits
+// a parseable "overall_status=" line matching its human-readable "Overall:"
+// verdict. ApplySATResultToDB/parseSATKV (internal/app/component_status_db.go)
+// only recognizes "key=value" lines — a summary with only "Overall: FAIL —
+// ..." (no "=") silently never reaches component-status.json, so a real
+// thermal-throttle FAIL detected by this test would never show up as cpu:all
+// going Warning.
+func TestWritePlatformSummaryOverallStatusKV(t *testing.T) {
+	opts := PlatformStressOptions{Cycles: []PlatformStressCycle{{LoadSec: 60, IdleSec: 30}}}
+
+	tests := []struct {
+		name     string
+		analyses []cycleAnalysis
+		want     string
+	}{
+		{
+			name:     "clean run reports OK",
+			analyses: []cycleAnalysis{{maxCPUTemp: 70, maxGPUTemp: 60}},
+			want:     "overall_status=OK",
+		},
+		{
+			name:     "throttle detected reports FAILED",
+			analyses: []cycleAnalysis{{maxCPUTemp: 95, throttled: true}},
+			want:     "overall_status=FAILED",
+		},
+		{
+			name:     "fast fan spindown reports FAILED",
+			analyses: []cycleAnalysis{{fanAtCutAvg: 8000, fanMin15s: 2000, fanDropPct: 75}},
+			want:     "overall_status=FAILED",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			summary := writePlatformSummary(opts, tt.analyses)
+			found := false
+			for _, line := range strings.Split(summary, "\n") {
+				if line == tt.want {
+					found = true
+					break
+				}
+			}
+			if !found {
+				t.Fatalf("summary missing %q:\n%s", tt.want, summary)
+			}
+		})
+	}
+}

audit/internal/platform/sat.go

@@ -793,16 +793,71 @@ func (s *System) RunSATStressPack(ctx context.Context, baseDir string, durationS
 	}, logFunc)
 }
 
+// cpuThermalThrottleSysDir is the sysfs root the throttle-check scripts glob
+// under. Overridden in tests so they can point at a fake directory tree
+// instead of the real /sys.
+var cpuThermalThrottleSysDir = "/sys/devices/system/cpu"
+
+// cpuThrottleSumScript is the shell fragment both before/after scripts use to
+// sum the kernel's cumulative-since-boot thermal throttle counters across
+// every CPU.
+func cpuThrottleSumScript() string {
+	return fmt.Sprintf(`
+sum=0
+for f in %[1]s/cpu*/thermal_throttle/core_throttle_count %[1]s/cpu*/thermal_throttle/package_throttle_count; do
+  [ -f "$f" ] || continue
+  v=$(cat "$f" 2>/dev/null)
+  case "$v" in ''|*[!0-9]*) continue ;; esac
+  sum=$((sum + v))
+done
+`, cpuThermalThrottleSysDir)
+}
+
+// cpuThrottleBeforeScript snapshots the throttle counter sum into a file in
+// {{run_dir}} so cpuThrottleCheckScript can later diff before/after despite
+// each satJob running as an independent process.
+func cpuThrottleBeforeScript() string {
+	return cpuThrottleSumScript() + `echo "$sum" | tee {{run_dir}}/.cpu-throttle-before` + "\n"
+}
+
+// cpuThrottleCheckScript compares the after-run throttle counter sum against
+// the snapshot cpuThrottleBeforeScript took, and fails (non-zero exit) if it
+// increased — i.e. the CPU actually hit thermal throttling during this
+// specific run, not just at some earlier point this boot. classifySATResult
+// maps a failed job here to SAT status FAILED, which ApplySATResultToDB
+// records as component status "Warning" for cpu:all — without this, the
+// "cpu" SAT pack only checks stress-ng's exit code, which is 0 whether or
+// not the CPU throttled while running it.
+func cpuThrottleCheckScript() string {
+	return `before=$(cat {{run_dir}}/.cpu-throttle-before 2>/dev/null)
+case "$before" in ''|*[!0-9]*) before=0 ;; esac
+` + cpuThrottleSumScript() + `after=$sum
+echo "throttle_count_before=$before"
+echo "throttle_count_after=$after"
+if [ "$after" -gt "$before" ]; then
+  echo "THROTTLE DETECTED: CPU package/core hit thermal throttling during this stress-ng run ($before -> $after)"
+  exit 1
+fi
+echo "no new thermal throttling detected during this run"
+`
+}
+
+func cpuSATJobs(durationSec int) []satJob {
+	return []satJob{
+		{name: "01-lscpu.log", cmd: []string{"lscpu"}},
+		{name: "02-sensors-before.log", cmd: []string{"sensors"}},
+		{name: "02-thermal-throttle-before.log", cmd: []string{"sh", "-c", cpuThrottleBeforeScript()}, informational: true},
+		{name: "03-stress-ng.log", cmd: []string{"stress-ng", "--cpu", "0", "--cpu-method", "all", "--timeout", fmt.Sprintf("%d", durationSec)}, syncBracket: true},
+		{name: "04-sensors-after.log", cmd: []string{"sensors"}},
+		{name: "05-thermal-throttle-check.log", cmd: []string{"sh", "-c", cpuThrottleCheckScript()}},
+	}
+}
+
 func (s *System) RunCPUAcceptancePack(ctx context.Context, baseDir string, durationSec int, logFunc func(string)) (string, error) {
 	if durationSec <= 0 {
 		durationSec = 60
 	}
-	return runAcceptancePackCtx(ctx, baseDir, "cpu", []satJob{
+	return runAcceptancePackCtx(ctx, baseDir, "cpu", cpuSATJobs(durationSec), logFunc)
-		{name: "01-lscpu.log", cmd: []string{"lscpu"}},
-		{name: "02-sensors-before.log", cmd: []string{"sensors"}},
-		{name: "03-stress-ng.log", cmd: []string{"stress-ng", "--cpu", "0", "--cpu-method", "all", "--timeout", fmt.Sprintf("%d", durationSec)}, syncBracket: true},
-		{name: "04-sensors-after.log", cmd: []string{"sensors"}},
-	}, logFunc)
 }
 
 func (s *System) RunStorageAcceptancePack(ctx context.Context, baseDir string, extended bool, logFunc func(string)) (string, error) {

audit/internal/platform/sat_test.go

@@ -684,3 +684,94 @@ func TestRunAcceptancePackCtxRetriesJob(t *testing.T) {
 		t.Fatalf("summary=%q want flaky_status=OK", got)
 	}
 }
+
+func TestCPUSATJobsIncludeThrottleCheck(t *testing.T) {
+	t.Parallel()
+
+	jobs := cpuSATJobs(60)
+
+	if len(jobs) != 6 {
+		t.Fatalf("jobs=%d want 6", len(jobs))
+	}
+	if jobs[2].name != "02-thermal-throttle-before.log" || !jobs[2].informational {
+		t.Fatalf("throttle-before job=%+v want informational before-snapshot", jobs[2])
+	}
+	if jobs[5].name != "05-thermal-throttle-check.log" || jobs[5].informational {
+		t.Fatalf("throttle-check job=%+v want non-informational so it can fail overall_status", jobs[5])
+	}
+}
+
+// TestCPUThrottleCheckDetectsIncreaseDuringRun runs the real before/check
+// scripts (the same ones RunCPUAcceptancePack shells out to) against a fake
+// sysfs tree, guarding both that a genuine thermal-throttle event during the
+// stress-ng run is caught, and that a host with no new throttling (or none
+// at all — e.g. AMD/ARM, which lack this sysfs interface) stays OK.
+func TestCPUThrottleCheckDetectsIncreaseDuringRun(t *testing.T) {
+	fakeSys := t.TempDir()
+	oldDir := cpuThermalThrottleSysDir
+	cpuThermalThrottleSysDir = fakeSys
+	t.Cleanup(func() { cpuThermalThrottleSysDir = oldDir })
+
+	writeCounter := func(cpu, file, value string) {
+		path := filepath.Join(fakeSys, cpu, "thermal_throttle", file)
+		if err := os.MkdirAll(filepath.Dir(path), 0755); err != nil {
+			t.Fatalf("mkdir: %v", err)
+		}
+		if err := os.WriteFile(path, []byte(value+"\n"), 0644); err != nil {
+			t.Fatalf("write: %v", err)
+		}
+	}
+
+	run := func(t *testing.T, dir string) (string, error) {
+		t.Helper()
+		runDir, err := runAcceptancePackCtx(context.Background(), dir, "cpu-throttle-test", []satJob{
+			{name: "02-thermal-throttle-before.log", cmd: []string{"sh", "-c", cpuThrottleBeforeScript()}, informational: true},
+			{name: "05-thermal-throttle-check.log", cmd: []string{"sh", "-c", cpuThrottleCheckScript()}},
+		}, nil)
+		if err != nil {
+			return "", err
+		}
+		summary, err := os.ReadFile(filepath.Join(runDir, "summary.txt"))
+		if err != nil {
+			t.Fatalf("reading summary.txt: %v", err)
+		}
+		return string(summary), nil
+	}
+
+	t.Run("no throttling stays OK", func(t *testing.T) {
+		writeCounter("cpu0", "core_throttle_count", "0")
+		writeCounter("cpu1", "core_throttle_count", "0")
+
+		summary, err := run(t, t.TempDir())
+		if err != nil {
+			t.Fatalf("runAcceptancePackCtx error: %v", err)
+		}
+		if !strings.Contains(summary, "overall_status=OK") {
+			t.Fatalf("summary=%q want overall_status=OK", summary)
+		}
+	})
+
+	t.Run("throttle count increasing during run fails overall_status", func(t *testing.T) {
+		writeCounter("cpu0", "core_throttle_count", "3")
+		writeCounter("cpu1", "core_throttle_count", "0")
+
+		dir := t.TempDir()
+		runDir, err := runAcceptancePackCtx(context.Background(), dir, "cpu-throttle-test", []satJob{
+			{name: "02-thermal-throttle-before.log", cmd: []string{"sh", "-c", cpuThrottleBeforeScript()}, informational: true},
+		}, nil)
+		if err != nil {
+			t.Fatalf("runAcceptancePackCtx (before) error: %v", err)
+		}
+
+		// Simulate the stress-ng run itself tripping the throttle.
+		writeCounter("cpu0", "core_throttle_count", "5")
+
+		out, err := exec.Command("sh", "-c", strings.ReplaceAll(cpuThrottleCheckScript(), "{{run_dir}}", runDir)).CombinedOutput()
+		if err == nil {
+			t.Fatalf("check script succeeded, want failure; output=%s", out)
+		}
+		if !strings.Contains(string(out), "THROTTLE DETECTED") {
+			t.Fatalf("output=%s want THROTTLE DETECTED", out)
+		}
+	})
+}