Compare commits
8 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
1c5cb45698 | ||
|
|
090b92ca73 | ||
|
|
2dccbc010c | ||
| e84c69d360 | |||
| c80a39e7ac | |||
| a5e0261ff2 | |||
| ee422ede3c | |||
| d560b2fead |
@@ -2831,15 +2831,6 @@ func powerBenchDurationSec(profile string) int {
|
||||
}
|
||||
}
|
||||
|
||||
func occupiedSlots(indices []int, current int) []int {
|
||||
out := make([]int, 0, len(indices))
|
||||
for _, idx := range indices {
|
||||
if idx != current {
|
||||
out = append(out, idx)
|
||||
}
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func cloneBenchmarkGPUInfoMap(src map[int]benchmarkGPUInfo) map[int]benchmarkGPUInfo {
|
||||
out := make(map[int]benchmarkGPUInfo, len(src))
|
||||
@@ -2887,9 +2878,7 @@ func renderPowerBenchReport(result NvidiaPowerBenchResult) string {
|
||||
b.WriteString("\n")
|
||||
for _, gpu := range result.GPUs {
|
||||
fmt.Fprintf(&b, "### GPU %d — %s\n\n", gpu.Index, gpu.Name)
|
||||
if gpu.OccupiedSlotsNote != "" {
|
||||
fmt.Fprintf(&b, "- %s\n", gpu.OccupiedSlotsNote)
|
||||
}
|
||||
|
||||
for _, note := range gpu.Notes {
|
||||
fmt.Fprintf(&b, "- %s\n", note)
|
||||
}
|
||||
@@ -2955,10 +2944,24 @@ func (s *System) RunNvidiaPowerBench(ctx context.Context, baseDir string, opts N
|
||||
}
|
||||
durationSec := powerBenchDurationSec(opts.Profile)
|
||||
_ = durationSec
|
||||
calibByIndex, restoreActions := runBenchmarkPowerCalibration(ctx, verboseLog, runDir, selected, infoByIndex, logFunc)
|
||||
// Phase 1: calibrate each GPU individually (sequentially, one at a time) to
|
||||
// establish a true single-card power baseline unaffected by neighbour heat.
|
||||
calibByIndex := make(map[int]benchmarkPowerCalibrationResult, len(selected))
|
||||
var allRestoreActions []benchmarkRestoreAction
|
||||
for _, idx := range selected {
|
||||
singleDir := filepath.Join(runDir, fmt.Sprintf("single-%02d", idx))
|
||||
_ = os.MkdirAll(singleDir, 0755)
|
||||
singleInfo := cloneBenchmarkGPUInfoMap(infoByIndex)
|
||||
logFunc(fmt.Sprintf("power calibration: GPU %d single-card baseline", idx))
|
||||
c, restore := runBenchmarkPowerCalibration(ctx, verboseLog, singleDir, []int{idx}, singleInfo, logFunc)
|
||||
allRestoreActions = append(allRestoreActions, restore...)
|
||||
if r, ok := c[idx]; ok {
|
||||
calibByIndex[idx] = r
|
||||
}
|
||||
}
|
||||
defer func() {
|
||||
for i := len(restoreActions) - 1; i >= 0; i-- {
|
||||
restoreActions[i].fn()
|
||||
for i := len(allRestoreActions) - 1; i >= 0; i-- {
|
||||
allRestoreActions[i].fn()
|
||||
}
|
||||
}()
|
||||
gpus := make([]NvidiaPowerBenchGPU, 0, len(selected))
|
||||
@@ -2975,11 +2978,6 @@ func (s *System) RunNvidiaPowerBench(ctx context.Context, baseDir string, opts N
|
||||
result.OverallStatus = "PARTIAL"
|
||||
}
|
||||
}
|
||||
occupied := occupiedSlots(selected, idx)
|
||||
note := ""
|
||||
if len(occupied) > 0 {
|
||||
note = fmt.Sprintf("Slot recommendation was measured while slots %s were populated; airflow in a different chassis fill pattern may differ.", joinIndexList(occupied))
|
||||
}
|
||||
gpus = append(gpus, NvidiaPowerBenchGPU{
|
||||
Index: idx,
|
||||
Name: info.Name,
|
||||
@@ -2991,8 +2989,6 @@ func (s *System) RunNvidiaPowerBench(ctx context.Context, baseDir string, opts N
|
||||
CalibrationAttempts: calib.Attempts,
|
||||
Derated: calib.Derated,
|
||||
Status: status,
|
||||
OccupiedSlots: occupied,
|
||||
OccupiedSlotsNote: note,
|
||||
Notes: append([]string(nil), calib.Notes...),
|
||||
CoolingWarning: calib.CoolingWarning,
|
||||
})
|
||||
@@ -3032,14 +3028,26 @@ func (s *System) RunNvidiaPowerBench(ctx context.Context, baseDir string, opts N
|
||||
for _, gpu := range gpus {
|
||||
singleByIndex[gpu.Index] = gpu
|
||||
}
|
||||
|
||||
// Phase 2: ramp — add one GPU per step and calibrate the growing subset
|
||||
// simultaneously. Step 1 reuses single-card results; steps 2..N run fresh
|
||||
// targeted_power with derating if degradation is detected.
|
||||
for step := 1; step <= len(result.RecommendedSlotOrder); step++ {
|
||||
subset := append([]int(nil), result.RecommendedSlotOrder[:step]...)
|
||||
stepDir := filepath.Join(runDir, fmt.Sprintf("step-%02d", step))
|
||||
_ = os.MkdirAll(stepDir, 0755)
|
||||
stepInfo := cloneBenchmarkGPUInfoMap(infoByIndex)
|
||||
stepCalib, stepRestore := runBenchmarkPowerCalibration(ctx, verboseLog, stepDir, subset, stepInfo, logFunc)
|
||||
for i := len(stepRestore) - 1; i >= 0; i-- {
|
||||
stepRestore[i].fn()
|
||||
var stepCalib map[int]benchmarkPowerCalibrationResult
|
||||
if step == 1 {
|
||||
// Single-GPU step — already measured in phase 1; reuse directly.
|
||||
stepCalib = calibByIndex
|
||||
logFunc(fmt.Sprintf("power ramp: step 1/%d — reusing single-card calibration for GPU %d", len(result.RecommendedSlotOrder), subset[0]))
|
||||
} else {
|
||||
stepInfo := cloneBenchmarkGPUInfoMap(infoByIndex)
|
||||
var stepRestore []benchmarkRestoreAction
|
||||
stepCalib, stepRestore = runBenchmarkPowerCalibration(ctx, verboseLog, stepDir, subset, stepInfo, logFunc)
|
||||
for i := len(stepRestore) - 1; i >= 0; i-- {
|
||||
stepRestore[i].fn()
|
||||
}
|
||||
}
|
||||
ramp := NvidiaPowerBenchStep{
|
||||
StepIndex: step,
|
||||
|
||||
@@ -280,8 +280,6 @@ type NvidiaPowerBenchGPU struct {
|
||||
CalibrationAttempts int `json:"calibration_attempts,omitempty"`
|
||||
Derated bool `json:"derated,omitempty"`
|
||||
Status string `json:"status"`
|
||||
OccupiedSlots []int `json:"occupied_slots,omitempty"`
|
||||
OccupiedSlotsNote string `json:"occupied_slots_note,omitempty"`
|
||||
Notes []string `json:"notes,omitempty"`
|
||||
// CoolingWarning mirrors BenchmarkGPUResult.CoolingWarning for the power workflow.
|
||||
CoolingWarning string `json:"cooling_warning,omitempty"`
|
||||
|
||||
@@ -552,9 +552,13 @@ func (s *System) RunMemoryAcceptancePack(ctx context.Context, baseDir string, si
|
||||
if passes <= 0 {
|
||||
passes = 1
|
||||
}
|
||||
// Bound memtester with a hard wall-clock timeout: ~2.5 min per 100 MB per
|
||||
// pass, plus a fixed 2-minute buffer. Without this, a stuck memory
|
||||
// controller can cause memtester to spin forever on a single subtest.
|
||||
timeoutSec := sizeMB*passes*150/100 + 120
|
||||
return runAcceptancePackCtx(ctx, baseDir, "memory", []satJob{
|
||||
{name: "01-free-before.log", cmd: []string{"free", "-h"}},
|
||||
{name: "02-memtester.log", cmd: []string{"memtester", fmt.Sprintf("%dM", sizeMB), fmt.Sprintf("%d", passes)}},
|
||||
{name: "02-memtester.log", cmd: []string{"timeout", fmt.Sprintf("%d", timeoutSec), "memtester", fmt.Sprintf("%dM", sizeMB), fmt.Sprintf("%d", passes)}},
|
||||
{name: "03-free-after.log", cmd: []string{"free", "-h"}},
|
||||
}, logFunc)
|
||||
}
|
||||
|
||||
@@ -1529,6 +1529,11 @@ func (h *handler) handleAPINetworkRollback(w http.ResponseWriter, _ *http.Reques
|
||||
writeJSON(w, map[string]string{"status": "rolled back"})
|
||||
}
|
||||
|
||||
func (h *handler) handleAPIBenchmarkResults(w http.ResponseWriter, r *http.Request) {
|
||||
w.Header().Set("Content-Type", "text/html; charset=utf-8")
|
||||
fmt.Fprint(w, renderBenchmarkResultsCard(h.opts.ExportDir))
|
||||
}
|
||||
|
||||
func (h *handler) rollbackPendingNetworkChange() error {
|
||||
h.pendingNetMu.Lock()
|
||||
pnc := h.pendingNet
|
||||
|
||||
@@ -2002,7 +2002,7 @@ func renderBenchmark(opts HandlerOptions) string {
|
||||
</div>
|
||||
</div>
|
||||
|
||||
` + renderBenchmarkResultsCard(opts.ExportDir) + `
|
||||
`+`<div id="benchmark-results-section">`+renderBenchmarkResultsCard(opts.ExportDir)+`</div>`+`
|
||||
|
||||
<div id="benchmark-output" style="display:none;margin-top:16px" class="card">
|
||||
<div class="card-head">Benchmark Output <span id="benchmark-title"></span></div>
|
||||
@@ -2188,7 +2188,9 @@ function runNvidiaBenchmark(kind) {
|
||||
if (e.data) failures += 1;
|
||||
term.textContent += (e.data ? '\nERROR: ' + e.data : '\nCompleted.') + '\n';
|
||||
term.scrollTop = term.scrollHeight;
|
||||
const isLast = (idx + 1 >= taskIds.length);
|
||||
streamNext(idx + 1, failures);
|
||||
if (isLast) { benchmarkRefreshResults(); }
|
||||
});
|
||||
benchmarkES.onerror = function() {
|
||||
if (benchmarkES) {
|
||||
@@ -2208,18 +2210,30 @@ function runNvidiaBenchmark(kind) {
|
||||
}
|
||||
|
||||
benchmarkLoadGPUs();
|
||||
|
||||
function benchmarkRefreshResults() {
|
||||
fetch('/api/benchmark/results')
|
||||
.then(function(r) { return r.text(); })
|
||||
.then(function(html) {
|
||||
const el = document.getElementById('benchmark-results-section');
|
||||
if (el) el.innerHTML = html;
|
||||
})
|
||||
.catch(function() {});
|
||||
}
|
||||
</script>`
|
||||
}
|
||||
|
||||
func renderBenchmarkResultsCard(exportDir string) string {
|
||||
maxIdx, runs := loadBenchmarkHistory(exportDir)
|
||||
return renderBenchmarkResultsCardFromRuns(
|
||||
"Perf Results",
|
||||
perf := renderBenchmarkResultsCardFromRuns(
|
||||
"Performance Results",
|
||||
"Composite score by saved benchmark run and GPU.",
|
||||
"No saved benchmark runs yet.",
|
||||
"No saved performance benchmark runs yet.",
|
||||
maxIdx,
|
||||
runs,
|
||||
)
|
||||
power := renderPowerBenchmarkResultsCard(exportDir)
|
||||
return perf + "\n" + power
|
||||
}
|
||||
|
||||
func renderBenchmarkResultsCardFromRuns(title, description, emptyMessage string, maxGPUIndex int, runs []benchmarkHistoryRun) string {
|
||||
@@ -2299,6 +2313,126 @@ func loadBenchmarkHistoryFromPaths(paths []string) (int, []benchmarkHistoryRun)
|
||||
return maxGPUIndex, runs
|
||||
}
|
||||
|
||||
func renderPowerBenchmarkResultsCard(exportDir string) string {
|
||||
baseDir := app.DefaultBeeBenchPowerDir
|
||||
if strings.TrimSpace(exportDir) != "" {
|
||||
baseDir = filepath.Join(exportDir, "bee-bench", "power")
|
||||
}
|
||||
paths, err := filepath.Glob(filepath.Join(baseDir, "power-*", "result.json"))
|
||||
if err != nil || len(paths) == 0 {
|
||||
return `<div class="card" style="margin-top:16px"><div class="card-head">Power / Thermal Fit Results</div><div class="card-body"><p style="color:var(--muted);font-size:13px">No saved power benchmark runs yet.</p></div></div>`
|
||||
}
|
||||
sort.Strings(paths)
|
||||
|
||||
type powerRun struct {
|
||||
generatedAt time.Time
|
||||
displayTime string
|
||||
result platform.NvidiaPowerBenchResult
|
||||
}
|
||||
var runs []powerRun
|
||||
for _, path := range paths {
|
||||
raw, err := os.ReadFile(path)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
var r platform.NvidiaPowerBenchResult
|
||||
if err := json.Unmarshal(raw, &r); err != nil {
|
||||
continue
|
||||
}
|
||||
runs = append(runs, powerRun{
|
||||
generatedAt: r.GeneratedAt,
|
||||
displayTime: r.GeneratedAt.Local().Format("2006-01-02 15:04:05"),
|
||||
result: r,
|
||||
})
|
||||
}
|
||||
sort.Slice(runs, func(i, j int) bool {
|
||||
return runs[i].generatedAt.After(runs[j].generatedAt)
|
||||
})
|
||||
|
||||
// Show only the most recent run's GPU slot table, plus a run history summary.
|
||||
var b strings.Builder
|
||||
b.WriteString(`<div class="card" style="margin-top:16px"><div class="card-head">Power / Thermal Fit Results</div><div class="card-body">`)
|
||||
|
||||
latest := runs[0].result
|
||||
b.WriteString(`<p style="font-size:12px;color:var(--muted);margin-bottom:10px">Latest run: ` + html.EscapeString(runs[0].displayTime))
|
||||
if latest.Hostname != "" {
|
||||
b.WriteString(` — ` + html.EscapeString(latest.Hostname))
|
||||
}
|
||||
if latest.OverallStatus != "" {
|
||||
statusColor := "var(--ok)"
|
||||
if latest.OverallStatus != "OK" {
|
||||
statusColor = "var(--warn)"
|
||||
}
|
||||
b.WriteString(` — <span style="color:` + statusColor + `;font-weight:600">` + html.EscapeString(latest.OverallStatus) + `</span>`)
|
||||
}
|
||||
b.WriteString(`</p>`)
|
||||
|
||||
if len(latest.GPUs) > 0 {
|
||||
b.WriteString(`<div style="overflow-x:auto"><table><thead><tr>`)
|
||||
b.WriteString(`<th>GPU</th><th>Model</th><th>Nominal W</th><th>Achieved W</th><th>P95 Observed W</th><th>Status</th>`)
|
||||
b.WriteString(`</tr></thead><tbody>`)
|
||||
for _, gpu := range latest.GPUs {
|
||||
derated := gpu.Derated || (gpu.DefaultPowerLimitW > 0 && gpu.AppliedPowerLimitW < gpu.DefaultPowerLimitW-1)
|
||||
rowStyle := ""
|
||||
achievedStyle := ""
|
||||
if derated {
|
||||
rowStyle = ` style="background:rgba(255,180,0,0.08)"`
|
||||
achievedStyle = ` style="color:#e6a000;font-weight:600"`
|
||||
}
|
||||
statusLabel := gpu.Status
|
||||
if statusLabel == "" {
|
||||
statusLabel = "OK"
|
||||
}
|
||||
statusColor := "var(--ok)"
|
||||
if statusLabel != "OK" {
|
||||
statusColor = "var(--warn)"
|
||||
}
|
||||
nominalStr := "-"
|
||||
if gpu.DefaultPowerLimitW > 0 {
|
||||
nominalStr = fmt.Sprintf("%.0f", gpu.DefaultPowerLimitW)
|
||||
}
|
||||
achievedStr := "-"
|
||||
if gpu.AppliedPowerLimitW > 0 {
|
||||
achievedStr = fmt.Sprintf("%.0f", gpu.AppliedPowerLimitW)
|
||||
}
|
||||
p95Str := "-"
|
||||
if gpu.MaxObservedPowerW > 0 {
|
||||
p95Str = fmt.Sprintf("%.0f", gpu.MaxObservedPowerW)
|
||||
}
|
||||
b.WriteString(`<tr` + rowStyle + `>`)
|
||||
b.WriteString(`<td>` + strconv.Itoa(gpu.Index) + `</td>`)
|
||||
b.WriteString(`<td>` + html.EscapeString(gpu.Name) + `</td>`)
|
||||
b.WriteString(`<td>` + nominalStr + `</td>`)
|
||||
b.WriteString(`<td` + achievedStyle + `>` + achievedStr + `</td>`)
|
||||
b.WriteString(`<td>` + p95Str + `</td>`)
|
||||
b.WriteString(`<td style="color:` + statusColor + `;font-weight:600">` + html.EscapeString(statusLabel) + `</td>`)
|
||||
b.WriteString(`</tr>`)
|
||||
}
|
||||
b.WriteString(`</tbody></table></div>`)
|
||||
}
|
||||
|
||||
if len(runs) > 1 {
|
||||
b.WriteString(`<details style="margin-top:12px"><summary style="font-size:12px;color:var(--muted);cursor:pointer">` + strconv.Itoa(len(runs)) + ` runs total</summary>`)
|
||||
b.WriteString(`<div style="overflow-x:auto;margin-top:8px"><table><thead><tr><th>#</th><th>Time</th><th>GPUs</th><th>Status</th></tr></thead><tbody>`)
|
||||
for i, run := range runs {
|
||||
statusColor := "var(--ok)"
|
||||
if run.result.OverallStatus != "OK" {
|
||||
statusColor = "var(--warn)"
|
||||
}
|
||||
b.WriteString(`<tr>`)
|
||||
b.WriteString(`<td>#` + strconv.Itoa(i+1) + `</td>`)
|
||||
b.WriteString(`<td>` + html.EscapeString(run.displayTime) + `</td>`)
|
||||
b.WriteString(`<td>` + strconv.Itoa(len(run.result.GPUs)) + `</td>`)
|
||||
b.WriteString(`<td style="color:` + statusColor + `;font-weight:600">` + html.EscapeString(run.result.OverallStatus) + `</td>`)
|
||||
b.WriteString(`</tr>`)
|
||||
}
|
||||
b.WriteString(`</tbody></table></div></details>`)
|
||||
}
|
||||
|
||||
b.WriteString(`</div></div>`)
|
||||
return b.String()
|
||||
}
|
||||
|
||||
// ── Burn ──────────────────────────────────────────────────────────────────────
|
||||
|
||||
func renderBurn() string {
|
||||
|
||||
@@ -263,6 +263,7 @@ func NewHandler(opts HandlerOptions) http.Handler {
|
||||
mux.HandleFunc("POST /api/sat/abort", h.handleAPISATAbort)
|
||||
mux.HandleFunc("POST /api/bee-bench/nvidia/perf/run", h.handleAPIBenchmarkNvidiaRunKind("nvidia-bench-perf"))
|
||||
mux.HandleFunc("POST /api/bee-bench/nvidia/power/run", h.handleAPIBenchmarkNvidiaRunKind("nvidia-bench-power"))
|
||||
mux.HandleFunc("GET /api/benchmark/results", h.handleAPIBenchmarkResults)
|
||||
|
||||
// Tasks
|
||||
mux.HandleFunc("GET /api/tasks", h.handleAPITasksList)
|
||||
|
||||
@@ -21,3 +21,4 @@ HIPBLASLT_VERSION=0.10.0.60304-76~22.04
|
||||
COMGR_VERSION=2.8.0.60304-76~22.04
|
||||
GO_VERSION=1.24.0
|
||||
AUDIT_VERSION=1.0.0
|
||||
MEMTEST_VERSION=6.10-4
|
||||
|
||||
@@ -23,9 +23,9 @@ lb config noauto \
|
||||
--bootloaders "grub-efi,syslinux" \
|
||||
--debian-installer none \
|
||||
--archive-areas "main contrib non-free non-free-firmware" \
|
||||
--mirror-bootstrap "https://deb.debian.org/debian" \
|
||||
--mirror-chroot "https://deb.debian.org/debian" \
|
||||
--mirror-binary "https://deb.debian.org/debian" \
|
||||
--mirror-bootstrap "http://mirror.mephi.ru/debian/" \
|
||||
--mirror-chroot "http://mirror.mephi.ru/debian/" \
|
||||
--mirror-binary "http://mirror.mephi.ru/debian/" \
|
||||
--security true \
|
||||
--linux-flavours "amd64" \
|
||||
--linux-packages "${LB_LINUX_PACKAGES}" \
|
||||
|
||||
@@ -161,6 +161,7 @@ run_variant() {
|
||||
-e GOMODCACHE=/cache/go-mod \
|
||||
-e TMPDIR=/cache/tmp \
|
||||
-e BEE_CACHE_DIR=/cache/bee \
|
||||
-e BEE_REQUIRE_MEMTEST=1 \
|
||||
-w /work \
|
||||
"${IMAGE_REF}" \
|
||||
sh /work/iso/builder/build.sh --variant "${_v}" \
|
||||
@@ -175,6 +176,7 @@ run_variant() {
|
||||
-e GOMODCACHE=/cache/go-mod \
|
||||
-e TMPDIR=/cache/tmp \
|
||||
-e BEE_CACHE_DIR=/cache/bee \
|
||||
-e BEE_REQUIRE_MEMTEST=1 \
|
||||
-w /work \
|
||||
"${IMAGE_REF}" \
|
||||
sh /work/iso/builder/build.sh --variant "${_v}"
|
||||
|
||||
@@ -57,6 +57,7 @@ OVERLAY_STAGE_DIR="${DIST_DIR}/overlay-stage-${BUILD_VARIANT}"
|
||||
export BEE_GPU_VENDOR BEE_NVIDIA_MODULE_FLAVOR BUILD_VARIANT
|
||||
|
||||
. "${BUILDER_DIR}/VERSIONS"
|
||||
export MEMTEST_VERSION
|
||||
export PATH="$PATH:/usr/local/go/bin"
|
||||
: "${BEE_REQUIRE_MEMTEST:=0}"
|
||||
|
||||
@@ -775,6 +776,7 @@ run_optional_step_sh() {
|
||||
return 0
|
||||
fi
|
||||
|
||||
mkdir -p "${LOG_DIR}" 2>/dev/null || true
|
||||
step_log="${LOG_DIR}/${step_slug}.log"
|
||||
echo ""
|
||||
echo "=== optional step: ${step_name} ==="
|
||||
@@ -798,13 +800,14 @@ start_build_log
|
||||
# install them on the fly so NVIDIA modules and ISO kernel always match.
|
||||
if [ -z "${DEBIAN_KERNEL_ABI}" ] || [ "${DEBIAN_KERNEL_ABI}" = "auto" ]; then
|
||||
echo "=== refreshing apt index to detect current kernel ABI ==="
|
||||
apt-get update -qq
|
||||
apt-get update -qq || echo "WARNING: apt-get update failed, trying cached index"
|
||||
DEBIAN_KERNEL_ABI=$(apt-cache depends linux-image-amd64 2>/dev/null \
|
||||
| awk '/Depends:.*linux-image-[0-9]/{print $2}' \
|
||||
| grep -oE '[0-9]+\.[0-9]+\.[0-9]+-[0-9]+' \
|
||||
| head -1)
|
||||
if [ -z "${DEBIAN_KERNEL_ABI}" ]; then
|
||||
echo "ERROR: could not auto-detect kernel ABI from apt-cache" >&2
|
||||
echo "Hint: set DEBIAN_KERNEL_ABI=x.y.z-N in iso/builder/VERSIONS to skip auto-detection" >&2
|
||||
exit 1
|
||||
fi
|
||||
echo "=== kernel ABI: ${DEBIAN_KERNEL_ABI} ==="
|
||||
|
||||
@@ -5,6 +5,8 @@ set -e
|
||||
|
||||
: "${BEE_REQUIRE_MEMTEST:=0}"
|
||||
|
||||
# memtest86+ 6.x uses memtest86+.bin (no x64 suffix) for the BIOS binary,
|
||||
# while 5.x used memtest86+x64.bin. We normalise both to x64 names in the ISO.
|
||||
MEMTEST_FILES="memtest86+x64.bin memtest86+x64.efi"
|
||||
BINARY_BOOT_DIR="binary/boot"
|
||||
GRUB_CFG="binary/boot/grub/grub.cfg"
|
||||
@@ -26,13 +28,13 @@ fail_or_warn() {
|
||||
|
||||
copy_memtest_file() {
|
||||
src="$1"
|
||||
base="$(basename "$src")"
|
||||
dst="${BINARY_BOOT_DIR}/${base}"
|
||||
dst_name="${2:-$(basename "$src")}"
|
||||
dst="${BINARY_BOOT_DIR}/${dst_name}"
|
||||
|
||||
[ -f "$src" ] || return 1
|
||||
mkdir -p "${BINARY_BOOT_DIR}"
|
||||
cp "$src" "$dst"
|
||||
log "copied ${base} from ${src}"
|
||||
log "copied ${dst_name} from ${src}"
|
||||
}
|
||||
|
||||
extract_memtest_from_deb() {
|
||||
@@ -41,14 +43,44 @@ extract_memtest_from_deb() {
|
||||
|
||||
log "extracting memtest payload from ${deb}"
|
||||
dpkg-deb -x "$deb" "$tmpdir"
|
||||
for f in ${MEMTEST_FILES}; do
|
||||
if [ -f "${tmpdir}/boot/${f}" ]; then
|
||||
copy_memtest_file "${tmpdir}/boot/${f}"
|
||||
fi
|
||||
done
|
||||
|
||||
# EFI binary: both 5.x and 6.x use memtest86+x64.efi
|
||||
if [ -f "${tmpdir}/boot/memtest86+x64.efi" ]; then
|
||||
copy_memtest_file "${tmpdir}/boot/memtest86+x64.efi"
|
||||
fi
|
||||
|
||||
# BIOS binary: 5.x = memtest86+x64.bin, 6.x = memtest86+.bin
|
||||
if [ -f "${tmpdir}/boot/memtest86+x64.bin" ]; then
|
||||
copy_memtest_file "${tmpdir}/boot/memtest86+x64.bin"
|
||||
elif [ -f "${tmpdir}/boot/memtest86+.bin" ]; then
|
||||
copy_memtest_file "${tmpdir}/boot/memtest86+.bin" "memtest86+x64.bin"
|
||||
fi
|
||||
|
||||
rm -rf "$tmpdir"
|
||||
}
|
||||
|
||||
download_and_extract_memtest() {
|
||||
tmpdl="$(mktemp -d)"
|
||||
if [ -n "${MEMTEST_VERSION:-}" ]; then
|
||||
pkg_spec="memtest86+=${MEMTEST_VERSION}"
|
||||
else
|
||||
pkg_spec="memtest86+"
|
||||
fi
|
||||
log "downloading ${pkg_spec} from apt"
|
||||
if ! ( cd "$tmpdl" && apt-get download "$pkg_spec" 2>/dev/null ); then
|
||||
log "apt download failed, retrying after apt-get update"
|
||||
apt-get update -qq >/dev/null 2>&1 || true
|
||||
( cd "$tmpdl" && apt-get download "$pkg_spec" 2>/dev/null ) || true
|
||||
fi
|
||||
deb="$(find "$tmpdl" -maxdepth 1 -type f -name 'memtest86+*.deb' 2>/dev/null | head -1)"
|
||||
if [ -n "$deb" ]; then
|
||||
extract_memtest_from_deb "$deb"
|
||||
else
|
||||
log "apt download of memtest86+ failed"
|
||||
fi
|
||||
rm -rf "$tmpdl"
|
||||
}
|
||||
|
||||
ensure_memtest_binaries() {
|
||||
missing=0
|
||||
for f in ${MEMTEST_FILES}; do
|
||||
@@ -56,10 +88,15 @@ ensure_memtest_binaries() {
|
||||
done
|
||||
[ "$missing" -eq 1 ] || return 0
|
||||
|
||||
# 1. Try files already placed by lb binary_memtest or chroot
|
||||
for root in chroot/boot /boot; do
|
||||
for f in ${MEMTEST_FILES}; do
|
||||
[ -f "${BINARY_BOOT_DIR}/${f}" ] || copy_memtest_file "${root}/${f}" || true
|
||||
done
|
||||
# 6.x BIOS binary may lack x64 in name — copy with normalised name
|
||||
if [ ! -f "${BINARY_BOOT_DIR}/memtest86+x64.bin" ]; then
|
||||
copy_memtest_file "${root}/memtest86+.bin" "memtest86+x64.bin" || true
|
||||
fi
|
||||
done
|
||||
|
||||
missing=0
|
||||
@@ -68,6 +105,7 @@ ensure_memtest_binaries() {
|
||||
done
|
||||
[ "$missing" -eq 1 ] || return 0
|
||||
|
||||
# 2. Try apt package cache (may be empty if lb binary_memtest already purged)
|
||||
for root in cache chroot/var/cache/apt/archives /var/cache/apt/archives; do
|
||||
[ -d "$root" ] || continue
|
||||
deb="$(find "$root" -type f \( -name 'memtest86+_*.deb' -o -name 'memtest86+*.deb' \) 2>/dev/null | head -1)"
|
||||
@@ -76,6 +114,15 @@ ensure_memtest_binaries() {
|
||||
break
|
||||
done
|
||||
|
||||
missing=0
|
||||
for f in ${MEMTEST_FILES}; do
|
||||
[ -f "${BINARY_BOOT_DIR}/${f}" ] || missing=1
|
||||
done
|
||||
[ "$missing" -eq 1 ] || return 0
|
||||
|
||||
# 3. Fallback: download fresh from apt (lb binary_memtest purges the cache)
|
||||
download_and_extract_memtest
|
||||
|
||||
missing=0
|
||||
for f in ${MEMTEST_FILES}; do
|
||||
if [ ! -f "${BINARY_BOOT_DIR}/${f}" ]; then
|
||||
|
||||
Reference in New Issue
Block a user