fix(iso): make memtest non-blocking by default

This commit is contained in:
Mikhail Chusavitin
2026-04-01 08:33:36 +03:00
parent 942f11937f
commit 3472afea32
5 changed files with 363 additions and 38 deletions

View File

@@ -38,6 +38,7 @@ export BEE_GPU_VENDOR
. "${BUILDER_DIR}/VERSIONS"
export PATH="$PATH:/usr/local/go/bin"
: "${BEE_REQUIRE_MEMTEST:=0}"
# Allow git to read the bind-mounted repo (different UID inside container).
git config --global safe.directory "${REPO_ROOT}"
@@ -177,6 +178,16 @@ dump_memtest_debug() {
fi
done
echo "-- source binary hooks --"
for hook in \
"${BUILDER_DIR}/config/hooks/normal/9100-memtest.hook.binary"; do
if [ -f "$hook" ]; then
echo " hook: $hook"
else
echo " (missing $hook)"
fi
done
if [ -n "$lb_dir" ] && [ -d "$lb_dir" ]; then
echo "-- live-build workdir package lists --"
for pkg in \
@@ -203,6 +214,20 @@ dump_memtest_debug() {
echo " (missing $lb_dir/binary/boot)"
fi
echo "-- live-build binary grub cfg --"
if [ -f "$lb_dir/binary/boot/grub/grub.cfg" ]; then
grep -n 'Memory Test\|memtest' "$lb_dir/binary/boot/grub/grub.cfg" || echo " (no memtest lines)"
else
echo " (missing $lb_dir/binary/boot/grub/grub.cfg)"
fi
echo "-- live-build binary isolinux cfg --"
if [ -f "$lb_dir/binary/isolinux/live.cfg" ]; then
grep -n 'Memory Test\|memtest' "$lb_dir/binary/isolinux/live.cfg" || echo " (no memtest lines)"
else
echo " (missing $lb_dir/binary/isolinux/live.cfg)"
fi
echo "-- live-build package cache --"
if [ -d "$lb_dir/cache/packages.chroot" ]; then
find "$lb_dir/cache/packages.chroot" -maxdepth 1 -name 'memtest86+*.deb' -print | sed 's/^/ /' || true
@@ -235,45 +260,75 @@ dump_memtest_debug() {
memtest_fail() {
msg="$1"
iso_path="${2:-}"
echo "ERROR: ${msg}" >&2
level="WARNING"
if [ "${BEE_REQUIRE_MEMTEST:-0}" = "1" ]; then
level="ERROR"
fi
echo "${level}: ${msg}" >&2
dump_memtest_debug "failure" "${LB_DIR:-}" "$iso_path" >&2
exit 1
if [ "${BEE_REQUIRE_MEMTEST:-0}" = "1" ]; then
exit 1
fi
return 0
}
validate_iso_memtest() {
iso_path="$1"
echo "=== validating memtest in ISO ==="
[ -f "$iso_path" ] || memtest_fail "ISO not found for validation: $iso_path" "$iso_path"
require_iso_reader "$iso_path"
[ -f "$iso_path" ] || {
memtest_fail "ISO not found for validation: $iso_path" "$iso_path"
return 0
}
require_iso_reader "$iso_path" || return 0
iso_list_files "$iso_path" | grep -q '^boot/memtest86+x64\.bin$' || {
memtest_fail "memtest BIOS binary missing in ISO: boot/memtest86+x64.bin" "$iso_path"
return 0
}
iso_list_files "$iso_path" | grep -q '^boot/memtest86+x64\.efi$' || {
memtest_fail "memtest EFI binary missing in ISO: boot/memtest86+x64.efi" "$iso_path"
return 0
}
grub_cfg="$(mktemp)"
isolinux_cfg="$(mktemp)"
iso_extract_file "$iso_path" boot/grub/grub.cfg > "$grub_cfg" || memtest_fail "failed to extract boot/grub/grub.cfg from ISO" "$iso_path"
iso_extract_file "$iso_path" isolinux/live.cfg > "$isolinux_cfg" || memtest_fail "failed to extract isolinux/live.cfg from ISO" "$iso_path"
iso_extract_file "$iso_path" boot/grub/grub.cfg > "$grub_cfg" || {
memtest_fail "failed to extract boot/grub/grub.cfg from ISO" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
iso_extract_file "$iso_path" isolinux/live.cfg > "$isolinux_cfg" || {
memtest_fail "failed to extract isolinux/live.cfg from ISO" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
grep -q 'Memory Test (memtest86+)' "$grub_cfg" || {
memtest_fail "GRUB menu entry for memtest is missing" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
grep -q '/boot/memtest86+x64\.efi' "$grub_cfg" || {
memtest_fail "GRUB memtest EFI path is missing" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
grep -q '/boot/memtest86+x64\.bin' "$grub_cfg" || {
memtest_fail "GRUB memtest BIOS path is missing" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
grep -q 'Memory Test (memtest86+)' "$isolinux_cfg" || {
memtest_fail "isolinux menu entry for memtest is missing" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
grep -q '/boot/memtest86+x64\.bin' "$isolinux_cfg" || {
memtest_fail "isolinux memtest path is missing" "$iso_path"
rm -f "$grub_cfg" "$isolinux_cfg"
return 0
}
rm -f "$grub_cfg" "$isolinux_cfg"
@@ -292,6 +347,10 @@ cleanup_build_log() {
status="${1:-$?}"
trap - EXIT INT TERM HUP
if [ "${STEP_LOG_ACTIVE:-0}" = "1" ]; then
cleanup_step_log "${status}" || true
fi
if [ "${BUILD_LOG_ACTIVE:-0}" = "1" ]; then
BUILD_LOG_ACTIVE=0
exec 1>&3 2>&4
@@ -335,6 +394,63 @@ start_build_log() {
echo "=== build log archive: ${LOG_ARCHIVE} ==="
}
cleanup_step_log() {
status="${1:-$?}"
if [ "${STEP_LOG_ACTIVE:-0}" = "1" ]; then
STEP_LOG_ACTIVE=0
exec 1>&5 2>&6
exec 5>&- 6>&-
if [ -n "${STEP_TEE_PID:-}" ]; then
wait "${STEP_TEE_PID}" 2>/dev/null || true
fi
rm -f "${STEP_LOG_PIPE}"
fi
return "${status}"
}
run_step() {
step_name="$1"
step_slug="$2"
shift 2
step_log="${LOG_DIR}/${step_slug}.log"
echo ""
echo "=== step: ${step_name} ==="
echo "=== step log: ${step_log} ==="
STEP_LOG_PIPE="$(mktemp -u "${TMPDIR:-/tmp}/bee-step-log.XXXXXX")"
mkfifo "${STEP_LOG_PIPE}"
exec 5>&1 6>&2
tee "${step_log}" < "${STEP_LOG_PIPE}" >&5 &
STEP_TEE_PID=$!
exec > "${STEP_LOG_PIPE}" 2>&1
STEP_LOG_ACTIVE=1
set +e
"$@"
step_status=$?
set -e
cleanup_step_log "${step_status}"
if [ "${step_status}" -ne 0 ]; then
echo "ERROR: step failed: ${step_name} (see ${step_log})" >&2
exit "${step_status}"
fi
echo "=== step OK: ${step_name} ==="
}
run_step_sh() {
step_name="$1"
step_slug="$2"
step_script="$3"
run_step "${step_name}" "${step_slug}" sh -c "${step_script}"
}
start_build_log
# Auto-detect kernel ABI: refresh apt index, then query current linux-image-amd64 dependency.
@@ -370,8 +486,8 @@ echo "Debian: ${DEBIAN_VERSION}, Kernel ABI: ${DEBIAN_KERNEL_ABI}, Go: ${GO_VERS
echo "Audit version: ${AUDIT_VERSION_EFFECTIVE}, ISO version: ${ISO_VERSION_EFFECTIVE}"
echo ""
echo "=== syncing git submodules ==="
git -C "${REPO_ROOT}" submodule update --init --recursive
run_step "sync git submodules" "05-git-submodules" \
git -C "${REPO_ROOT}" submodule update --init --recursive
# --- compile bee binary (static, Linux amd64) ---
# Shared between variants — built once, reused on second pass.
@@ -383,13 +499,13 @@ if [ -f "$BEE_BIN" ]; then
fi
if [ "$NEED_BUILD" = "1" ]; then
echo "=== building bee binary ==="
cd "${REPO_ROOT}/audit"
GOOS=linux GOARCH=amd64 CGO_ENABLED=0 \
go build \
-ldflags "-s -w -X main.Version=${AUDIT_VERSION_EFFECTIVE}" \
-o "$BEE_BIN" \
./cmd/bee
run_step_sh "build bee binary" "10-build-bee" \
"cd '${REPO_ROOT}/audit' && \
env GOOS=linux GOARCH=amd64 CGO_ENABLED=0 \
go build \
-ldflags '-s -w -X main.Version=${AUDIT_VERSION_EFFECTIVE}' \
-o '${BEE_BIN}' \
./cmd/bee"
echo "binary: $BEE_BIN"
if command -v stat >/dev/null 2>&1; then
BEE_SIZE_BYTES="$(stat -c '%s' "$BEE_BIN" 2>/dev/null || stat -f '%z' "$BEE_BIN")"
@@ -408,9 +524,8 @@ fi
# --- NVIDIA-only build steps ---
GPU_BURN_WORKER_BIN="${DIST_DIR}/bee-gpu-burn-worker-linux-amd64"
if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
echo ""
echo "=== downloading cuBLAS/cuBLASLt/cudart ${NCCL_CUDA_VERSION} userspace ==="
sh "${BUILDER_DIR}/build-cublas.sh" \
run_step "download cuBLAS/cuBLASLt/cudart ${NCCL_CUDA_VERSION} userspace" "20-cublas" \
sh "${BUILDER_DIR}/build-cublas.sh" \
"${CUBLAS_VERSION}" \
"${CUDA_USERSPACE_VERSION}" \
"${NCCL_CUDA_VERSION}" \
@@ -424,8 +539,8 @@ if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
fi
if [ "$GPU_STRESS_NEED_BUILD" = "1" ]; then
echo "=== building bee-gpu-burn worker ==="
gcc -O2 -s -Wall -Wextra \
run_step "build bee-gpu-burn worker" "21-gpu-burn-worker" \
gcc -O2 -s -Wall -Wextra \
-I"${CUBLAS_CACHE}/include" \
-o "$GPU_BURN_WORKER_BIN" \
"${BUILDER_DIR}/bee-gpu-stress.c" \
@@ -546,9 +661,8 @@ done
# --- NVIDIA kernel modules and userspace libs ---
if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
echo ""
echo "=== building NVIDIA ${NVIDIA_DRIVER_VERSION} modules ==="
sh "${BUILDER_DIR}/build-nvidia-module.sh" "${NVIDIA_DRIVER_VERSION}" "${DIST_DIR}" "${DEBIAN_KERNEL_ABI}"
run_step "build NVIDIA ${NVIDIA_DRIVER_VERSION} modules" "40-nvidia-module" \
sh "${BUILDER_DIR}/build-nvidia-module.sh" "${NVIDIA_DRIVER_VERSION}" "${DIST_DIR}" "${DEBIAN_KERNEL_ABI}"
KVER="${DEBIAN_KERNEL_ABI}-amd64"
NVIDIA_CACHE="${DIST_DIR}/nvidia-${NVIDIA_DRIVER_VERSION}-${KVER}"
@@ -576,9 +690,8 @@ if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
fi
# --- build / download NCCL ---
echo ""
echo "=== downloading NCCL ${NCCL_VERSION}+cuda${NCCL_CUDA_VERSION} ==="
sh "${BUILDER_DIR}/build-nccl.sh" "${NCCL_VERSION}" "${NCCL_CUDA_VERSION}" "${DIST_DIR}" "${NCCL_SHA256:-}"
run_step "download NCCL ${NCCL_VERSION}+cuda${NCCL_CUDA_VERSION}" "50-nccl" \
sh "${BUILDER_DIR}/build-nccl.sh" "${NCCL_VERSION}" "${NCCL_CUDA_VERSION}" "${DIST_DIR}" "${NCCL_SHA256:-}"
NCCL_CACHE="${DIST_DIR}/nccl-${NCCL_VERSION}+cuda${NCCL_CUDA_VERSION}"
@@ -591,9 +704,8 @@ if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
echo "=== cuBLAS: $(ls "${CUBLAS_CACHE}/lib/" | wc -l) files injected into /usr/lib/ ==="
# --- build nccl-tests ---
echo ""
echo "=== building nccl-tests ${NCCL_TESTS_VERSION} ==="
sh "${BUILDER_DIR}/build-nccl-tests.sh" \
run_step "build nccl-tests ${NCCL_TESTS_VERSION}" "60-nccl-tests" \
sh "${BUILDER_DIR}/build-nccl-tests.sh" \
"${NCCL_TESTS_VERSION}" \
"${NCCL_VERSION}" \
"${NCCL_CUDA_VERSION}" \
@@ -607,9 +719,8 @@ if [ "$BEE_GPU_VENDOR" = "nvidia" ]; then
cp "${NCCL_TESTS_CACHE}/lib/"* "${OVERLAY_STAGE_DIR}/usr/lib/" 2>/dev/null || true
echo "=== all_reduce_perf injected ==="
echo ""
echo "=== building john jumbo ${JOHN_JUMBO_COMMIT} ==="
sh "${BUILDER_DIR}/build-john.sh" "${JOHN_JUMBO_COMMIT}" "${DIST_DIR}"
run_step "build john jumbo ${JOHN_JUMBO_COMMIT}" "70-john" \
sh "${BUILDER_DIR}/build-john.sh" "${JOHN_JUMBO_COMMIT}" "${DIST_DIR}"
JOHN_CACHE="${DIST_DIR}/john-${JOHN_JUMBO_COMMIT}"
mkdir -p "${OVERLAY_STAGE_DIR}/usr/local/lib/bee/john"
rsync -a --delete "${JOHN_CACHE}/run/" "${OVERLAY_STAGE_DIR}/usr/local/lib/bee/john/run/"
@@ -730,10 +841,10 @@ BEE_GPU_VENDOR_UPPER="$(echo "${BEE_GPU_VENDOR}" | tr 'a-z' 'A-Z')"
export BEE_GPU_VENDOR_UPPER
cd "${LB_DIR}"
lb clean 2>&1 | tail -3
lb config 2>&1 | tail -5
run_step_sh "live-build clean" "80-lb-clean" "lb clean 2>&1 | tail -3"
run_step_sh "live-build config" "81-lb-config" "lb config 2>&1 | tail -5"
dump_memtest_debug "pre-build" "${LB_DIR}"
lb build 2>&1
run_step_sh "live-build build" "90-lb-build" "lb build 2>&1"
# --- persist deb package cache back to shared location ---
# This allows the second variant to reuse all downloaded packages.

View File

@@ -0,0 +1,139 @@
#!/bin/sh
# Ensure memtest is present in the final ISO even if live-build's built-in
# memtest stage does not copy the binaries or expose menu entries.
set -e
: "${BEE_REQUIRE_MEMTEST:=0}"
MEMTEST_FILES="memtest86+x64.bin memtest86+x64.efi"
BINARY_BOOT_DIR="binary/boot"
GRUB_CFG="binary/boot/grub/grub.cfg"
ISOLINUX_CFG="binary/isolinux/live.cfg"
log() {
echo "memtest hook: $*"
}
fail_or_warn() {
msg="$1"
if [ "${BEE_REQUIRE_MEMTEST}" = "1" ]; then
log "ERROR: ${msg}"
exit 1
fi
log "WARNING: ${msg}"
return 0
}
copy_memtest_file() {
src="$1"
base="$(basename "$src")"
dst="${BINARY_BOOT_DIR}/${base}"
[ -f "$src" ] || return 1
mkdir -p "${BINARY_BOOT_DIR}"
cp "$src" "$dst"
log "copied ${base} from ${src}"
}
extract_memtest_from_deb() {
deb="$1"
tmpdir="$(mktemp -d)"
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/${f}"
fi
done
rm -rf "$tmpdir"
}
ensure_memtest_binaries() {
missing=0
for f in ${MEMTEST_FILES}; do
[ -f "${BINARY_BOOT_DIR}/${f}" ] || missing=1
done
[ "$missing" -eq 1 ] || return 0
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
done
missing=0
for f in ${MEMTEST_FILES}; do
[ -f "${BINARY_BOOT_DIR}/${f}" ] || missing=1
done
[ "$missing" -eq 1 ] || return 0
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)"
[ -n "$deb" ] || continue
extract_memtest_from_deb "$deb"
break
done
missing=0
for f in ${MEMTEST_FILES}; do
if [ ! -f "${BINARY_BOOT_DIR}/${f}" ]; then
fail_or_warn "missing ${BINARY_BOOT_DIR}/${f}"
missing=1
fi
done
[ "$missing" -eq 0 ] || return 0
}
ensure_grub_entry() {
[ -f "$GRUB_CFG" ] || {
fail_or_warn "missing ${GRUB_CFG}"
return 0
}
grep -q '### BEE MEMTEST ###' "$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
log "appended memtest entry to ${GRUB_CFG}"
}
ensure_isolinux_entry() {
[ -f "$ISOLINUX_CFG" ] || {
fail_or_warn "missing ${ISOLINUX_CFG}"
return 0
}
grep -q '### BEE MEMTEST ###' "$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
log "appended memtest entry to ${ISOLINUX_CFG}"
}
log "ensuring memtest binaries and menu entries in binary image"
ensure_memtest_binaries
ensure_grub_entry
ensure_isolinux_entry
log "memtest assets ready"