feat(ttl): Run Script 4's CPU stress as the same systemd unit

The thermal / safety macro started mlucas-avx2 with '&', exactly as
Script B used to, so it carried the same failure: the load ended quietly
when the console went away - in the one test where a long, steady full
load is the entire point of the run.

It now uses the same mlucas-amm transient unit, and 'systemctl is-active
mlucas-amm' joins 'bgctl list' and 'jobs' in the monitor loop.

Script 4 still has no teardown section, so the matching stop stays
manual, and it is now two commands rather than one:

    blanton_tr518.sh stop
    systemctl stop mlucas-amm

Closing Tera Term no longer stops the CPU load, and 'kill $(jobs -p)'
never could reach a unit. Script A's V1.1.8 history, CLAUDE.md,
ARCHITECTURE.md and the release note all say so now; the version stays
V1.1.8.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014RetWKFZFG1ZHcitQAyhwM

ARCHITECTURE.md

@@ -373,7 +373,7 @@ PROFILE_CHANNELS=( "0:high:8" "1:low:8" ... )   # "channel:operation:change_perc
 | **Script A（Pre-test）** | `1_Blanton_Script_A.ttl` V1.1.8 | root 登入 → `date -s` 對時 → 清 job log → **收掉前一輪殘留**（`hw-test-session finish` → `bgctl reset --yes` → `bgctl stop --all` → `bgctl list` 確認清空）→ `hw-test-session start` → **設定風扇轉速** → **等資料面就緒**（`wait_init.ttl`）→ DUT 清單（boot/version/fwutil/syseeprom/ssdhealth/TPM/NVMe）→ BMC enroll + version/status → source bash 工具 → 清/讀 7 組 PCIe AER + `ras-mc-ctl --summary` → `lspci` → margin → PMON → dmesg → **10G/1G 管理網路 ping** → **NFC 基線**（start → 30s → stop → report）→ 100G uplink 設定與狀態 → **一輪流量基線** → `show uptime` |
 | **Script B（Stress）** | `2_Blanton_Script_B.ttl` | **V1.1.8 起 `mlucas-avx2 -cpu 0:15` 由 `systemd-run --unit=mlucas-amm`（`Type=exec`、`Restart=always`）啟動，不再是 shell 的 `&` 背景 job** + `bgctl run` 的 `memtester` / SSD / USB 壓力（v1.1.7 起 SSD/USB 的 `--runtime`/`--passes` 讀 `config.ttl` 的 `SSD_RUNTIME`/`SSD_COUNT`、`USB_RUNTIME`/`USB_COUNT`；BMC memtester 不帶次數 = 跑到被停） + BMC DDR（`bmc-manager run memtester`）→ BMC USB net test（探測 cdc_ncm 介面，`systemd-run` 掛 `BMC_USB_RUNTIME` 秒的 ping，V1.1.8 前寫死 4 小時）→ 啟動 mgmt ping 監控 → **啟動 NFC 輪詢** → `jobs` / `systemctl is-active mlucas-amm` / `bgctl list` → traffic（依 `SWB_UNIT`）→ `while 1` 每輪 PMON + `bgctl list`/`jobs`/`systemctl is-active mlucas-amm` + `pause 60`（**V1.1.4 起 margin 不在迴圈內**）|
 | **Script C（Post-test）** | `3_Blanton_Script_C.ttl` | `kill $(jobs -p)` + **`systemctl stop mlucas-amm`**（systemd unit 不是 shell job，`jobs -p` 抓不到）+ `bgctl stop --all` / `reset` + 停 BMC USB unit、mgmt ping 與 **NFC 輪詢** → PMON + dmesg → 7 組 AER + `ras-mc-ctl` → BMC USB `journalctl` **寫進 `log/bmc_usb_net.log`**（不再印在 console） → `ip -s link show eth0/eth1` → `cat` mlucas log（並 `cp` 進 job 目錄一起帶走）→ **NFC `report`** → traffic `stop` + `report` → **BER 掃描**（依 `SWB_UNIT`）→ NVMe 健康 → `show reboot-cause` → **margin 掃描**（`EN_Margin`，V1.1.4 由 Script B 移來）→ `uptime` → **job log 複製到 USB（帶時戳）** → `exit` + wait『Script done』→ `hw-test-session finish` |
-| **Script 4（Thermal / Safety）** | `4_Blanton_Script_thermal_safety.ttl` | **獨立腳本，不接 A→B→C。** Script A 式的環境建立（登入 / session / 風扇 / `wait_init` / DUT 清單 / BMC / PCIe 清除 / `pcie_bus` / PMON / dmesg）→ **滿載**（`mlucas-avx2 -cpu 0:15` 背景 + `blanton_tr518.sh start`）→ `while 1`：PMON + `TH6_SWB0/1_power_readback.sh` + `bgctl list` / `jobs`，`pause 60`。**沒有收尾段** —— 中斷後 DUT 仍在 `port cd lb=mac` / `l2 learn off` / `test mode nr=yes`，要手動 `blanton_tr518.sh stop` |
+| **Script 4（Thermal / Safety）** | `4_Blanton_Script_thermal_safety.ttl` | **獨立腳本，不接 A→B→C。** Script A 式的環境建立（登入 / session / 風扇 / `wait_init` / DUT 清單 / BMC / PCIe 清除 / `pcie_bus` / PMON / dmesg）→ **滿載**（`mlucas-avx2 -cpu 0:15`，V1.1.8 起與 Script B 同樣掛成 `mlucas-amm` systemd unit + `blanton_tr518.sh start`）→ `while 1`：PMON + `TH6_SWB0/1_power_readback.sh` + `bgctl list` / `jobs` / `systemctl is-active mlucas-amm`，`pause 60`。**沒有收尾段** —— 中斷後 DUT 仍在 `port cd lb=mac` / `l2 learn off` / `test mode nr=yes`，且 CPU 仍滿載，要手動 `blanton_tr518.sh stop` + `systemctl stop mlucas-amm` |
 | **Script 5（Reboot Cycle）** | `5_Blanton_Script_Reboot_Cycle.ttl` | **獨立腳本，不接 A→B→C。** `while 1`：登入 → `sudo -i` → 對時 → 風扇 → **Check Pt1~Pt13**（DUT/BMC info、mgmt ping、NFC、SSD、USB、BMC USB、port status + 15s traffic、BER、NVMe、platform data、dmesg i2c、PCIe error counters）→ `show uptime` → `reboot` → `wait "Booting in blind mode"`。`cnt_cycle` 每輪 +1 印在標頭。**沒有結束條件** —— 要停就關掉 Tera Term 的 macro。**尚未上機驗證** |
 | **SWB loopback 流量** | `blanton_traffic_linespeed <sub> [-u 0\|1\|all]` | `init`（VLAN 30..137，成對 `cdN`/`cdN+32`）→ `clear` → `start`（`tx 200 length=324 PatternRandom=yes`，v0.7.0 起）→ `stop` → `report`（per-pair TX/RX 交叉比對 PASS/FAIL）；另有 `ps`（link/speed）與 `run`（一條龍）。**V0.6.0 起含 100G uplink `ce0`**（VLAN 138）：它沒有 loopback 對象，靠 **ingress mirror 打回自己**維持循環，`report` 最後一列就是它，底下再接一份跨 unit summary |
 | **FPGA 暫存器存取** | `cb_fpga <fn> <addr> [data]`、`fpga_rescan` | pcimem 打 `resource0` + BAR 相對 offset，32-bit word；BDF 於 source 時自動偵測；`pmc/icb/swb0/swb1_fpga` 走 VSPI 窗口 |
@@ -545,9 +545,10 @@ pwr_data
     搬到 `mlucas-amm` unit（好處：掉線不會停、`jobs` 看不見的問題消失、`Restart=always` 會自動拉回），
     代價是 Script C 少了那行 `systemctl stop` 就會讓 CPU 滿載燒進收工量測與下一輪測試 ——
     而 `show interfaces status` 之類的檢查完全看不出來。目前 `mlucas-amm` 與
-    `qfx5252-bmc-usb-net-test` 兩個 unit 都已在 Script C 收掉；
+    `qfx5252-bmc-usb-net-test` 兩個 unit 都已在 Script C 收掉。
-    ⚠️ **`4_Blanton_Script_thermal_safety.ttl` 的 `mlucas-avx2` 仍是 `&` 背景 job**，
+    ⚠️ **`4_Blanton_Script_thermal_safety.ttl` 用同一個 `mlucas-amm` unit，但它沒有收尾段**，
-    兩支腳本目前用不同機制，改動時不要互相套用。
+    所以那行 `systemctl stop mlucas-amm` 必須由測試員手動下 —— 這是 Script 4 目前
+    唯一一個「不下就會延續到下一輪」的負載。
 18. **`config.ttl` 少一個變數會讓 macro 當場中止**，不是當成 0。V1.1.8 新增 `EN_timestamp`
     （Script A `logopen` 的第 5 個參數）與 `BMC_USB_RUNTIME`（Script B 的 ping unit）。
     只更新 `.ttl` 卻沿用舊的 `config.ttl`，Script A 會停在 `logopen` 那一行。
@@ -883,8 +884,9 @@ margin status、全部 VRM 軌的 Vin/Vout/Iout/Temp、CB + ICB 溫感讀值，
   LTC2980 PMBus）摸同一批軌，各自都假設自己是唯一的寫入者。目前只靠文件與 help 提醒，
   沒有任何執行期檢查。
 - **Script 4 沒有收尾段**：`while 1` 之後就是檔案結尾，中斷時 DUT 停在 `port cd lb=mac` /
-  `l2 learn off` / `test mode nr=yes`，而且 `mlucas-avx2` 還在背景跑。要手動
+  `l2 learn off` / `test mode nr=yes`，而且 CPU 壓力還在跑。要手動
-  `blanton_tr518.sh stop` + `kill`，否則下一輪任何測試都在被污染的狀態上開始
+  `blanton_tr518.sh stop` + `systemctl stop mlucas-amm`（V1.1.8 起它是 systemd unit，
+  關掉 Tera Term 或 `kill $(jobs -p)` 都收不掉），否則下一輪任何測試都在被污染的狀態上開始
   （Script A 的 `bgctl reset/stop --all` 收得掉 job，但收不掉 bcm 的 loopback 設定）。
   可考慮加一支對應的收尾腳本，或在檔尾用 `messagebox` 提示要下哪些指令。
 - **margin 在 soak 期間不再取樣**（V1.1.4）：只剩 Script A 的起始值與 Script C 的結束值。

CLAUDE.md

@@ -261,8 +261,8 @@ Tera Term 連 COM port (115200-8-N-1)
   唯一線索，**不會報警**。V1.1.8 起出貨預設改為 `RUNTIME=3600`（1 小時），2 天 soak 要填 `172800`
 - ⚠️ **`utils/kill_all_process.ttl` 只有 `kill $(jobs -p)`，抓不到 systemd unit**：把壓力工作改成
   `systemd-run` 之後，Script C 一定要補對應的 `systemctl stop <unit>`，否則它會活過 Script C 燒進
-  下一輪，**收工的溫度／功耗全部是在滿載下量的**。V1.1.8 已為 `mlucas-amm` 補上；
+  下一輪，**收工的溫度／功耗全部是在滿載下量的**。V1.1.8 已為 `mlucas-amm` 補上。
-  **Script 4 的 `mlucas-avx2` 仍是 `&` 背景 job**，兩支目前用不同機制，改動時別互相套用
+  Script 4 也改用同一個 unit，但它**沒有收尾段**，所以那行 `systemctl stop` 得由人手動下
 - ⚠️ **`config.ttl` 少一個變數會讓 macro 當場中止，不是當成 0**：`EN_timestamp`（V1.1.8 新增，
   餵 Script A `logopen` 的第 5 個參數）、`BMC_USB_RUNTIME` 都是必要項。只更新 `.ttl` 卻沿用舊的
   `config.ttl`，Script A 會停在 `logopen` 那一行
@@ -276,7 +276,8 @@ Tera Term 連 COM port (115200-8-N-1)
 - ⚠️ **Script 4（thermal/safety）沒有收尾**：`while 1` 之後就是檔尾。中斷後 DUT 停在
   `port cd lb=mac` / `l2 learn off` / `test mode nr=yes`，`mlucas-avx2` 也還在背景。
   下一個測試會在被污染的狀態上開始，而且 **`show interfaces status` 看不出來**。
-  收尾：`blanton_tr518.sh stop` + `kill $(jobs -p)`
+  收尾：`blanton_tr518.sh stop` + `kill $(jobs -p)` + **`systemctl stop mlucas-amm`**
+  （V1.1.8 起 CPU 壓力是 systemd unit，`kill $(jobs -p)` 收不掉，少這一行 CPU 會一直滿載）
 - ⚠️ **Script 4 與 Script B 不能共用同一台 DUT**：兩者都會接管 `port cd` 的 loopback
 - ⚠️ **`settings/*.conf` 的 `VNOM` 是實測中位數，不是規格值**（2026-08-30 起）。它同時是
   `margin_status` 偏差 % 的分母、也是 `margin_set` 推算 high/low/OV/UV 的基準，所以改它會

docs/release-notes/v1.1.8.md

@@ -3,7 +3,7 @@
 ## ✨ New features
 
 **`mlucas` now runs as a systemd unit, so it survives a lost session**
-* The AMD CPU load is started as `systemd-run --unit=mlucas-amm --property=Type=exec --property=Restart=always --property=RestartSec=2`, instead of being backgrounded with `&` in the login shell.
+* The AMD CPU load is started as `systemd-run --unit=mlucas-amm --property=Type=exec --property=Restart=always --property=RestartSec=2`, instead of being backgrounded with `&` in the login shell. Script B and the thermal/safety macro (Script 4) both use it.
 * As a shell job it belonged to that one terminal: close Tera Term, drop the console, reconnect on a different session, and the load was gone — while `jobs` from the new session showed nothing to tell you. A multi-day soak could spend most of its length with an idle CPU and produce thermal numbers that looked fine because nothing was loading the part.
 * `Restart=always` brings it back if it exits on its own, and the log is appended (`>>`) inside the unit while being truncated (`>`) once before the unit starts, so a restart adds to the run rather than erasing it.
 * Script B now prints `systemctl is-active mlucas-amm` beside `jobs`, both in the one-shot check and in every pass of the monitor loop. One word — `active` — tells you the CPU is still loaded.
@@ -53,7 +53,7 @@ ls ~/hammer/tools/amd/mlucas-avx2   # must already be there - see below
 * **The three stress durations are deliberately different**: SSD and USB at 1 hour, the BMC USB ping at 2 days. Set each to match the soak you are actually running.
 * **`--runtime` is the job's overall limit, not a per-pass one.** `qfx5252-stress-*` stops after `RUNTIME` seconds whatever `--passes` says. For a 24 hour soak set `RUNTIME` to `86400`.
 * **`mlucas-amm` has not been run on hardware yet.** The unit resolves `~` through bash's fallback to the passwd entry, because systemd does not set `HOME` for a system service without `User=`. It should land on `/root/hammer`, the same path the interactive shell uses — confirm it on the bench, or change both paths to an absolute `/root/hammer/...`.
-* **Script 4 still backgrounds `mlucas-avx2` with `&`.** The two macros use different mechanisms now; do not copy one's teardown into the other.
+* **Script 4 uses the same unit but has no teardown.** The thermal/safety macro now starts `mlucas-amm` the same way Script B does, and it still ends at `while 1` — so stopping it means `blanton_tr518.sh stop` **and** `systemctl stop mlucas-amm`. Closing Tera Term does not stop the CPU load any more.
 * **Script 5 (reboot cycle) is still unverified on hardware**, still has no exit condition, and `wait "Booting in blind mode"` still has no `timeout` guard.
 * **`blanton_multiphase_margin.sh save` is permanent**, and must not be mixed with `margin.sh` — same rails, two access paths, no interlock.
 * **`READ_VIN` and `READ_IIN` are still not in the ADPM12200 coefficient table.**

src/Script_ABC_Blanton/1_Blanton_Script_A.ttl

@@ -387,6 +387,17 @@
 ;                                                                         !! ScriptB and ScriptC change together. Whoever moves a
 ;                                                                            stress job to systemd owns the matching stop in
 ;                                                                            ScriptC
+;                       Script4                                         Use the same mlucas-amm systemd unit as ScriptB
+;                                                                         The thermal / safety macro started mlucas the same way
+;                                                                         ScriptB used to, with '&', so it had the same failure:
+;                                                                         the load quietly ended when the console went away, in the
+;                                                                         one test where a long steady full load is the whole point
+;                                                                         'systemctl is-active mlucas-amm' joins 'bgctl list' and
+;                                                                         'jobs' in the monitor loop
+;                                                                         !! Script4 still has no teardown section, so the matching
+;                                                                            stop is manual: blanton_tr518.sh stop AND systemctl
+;                                                                            stop mlucas-amm. Closing Tera Term no longer stops the
+;                                                                            CPU load, and 'kill $(jobs -p)' never reached a unit
 ;                       config.ttl                                      Add EN_timestamp and BMC_USB_RUNTIME, raise stress runtimes
 ;                                                                         EN_timestamp = 1 feeds ScriptA's logopen; BMC_USB_RUNTIME
 ;                                                                         = 172800 feeds ScriptB's ping unit

src/Script_ABC_Blanton/4_Blanton_Script_thermal_safety.ttl

@@ -129,11 +129,15 @@ include "utils/setup_pmon.ttl"
 include "utils/show_dmesg.ttl"
 
 ; ========== Full Load ==========
-;  AMD CPU STRESS TEST 
+; AMD CPU STRESS TEST 
+; Clear the log file first
+wait prompt_sonic_root
+sendln "> ~/hammer/tools/amd/mlucas_amm_log"
 wait prompt_sonic_root
 sendln 'chmod +x ~/hammer/tools/amd/mlucas-avx2'
 wait prompt_sonic_root
-sendln "~/hammer/tools/amd/mlucas-avx2 -s l -cpu 0:15 > ~/hammer/tools/amd/mlucas_amm_log 2>&1 &"
+;sendln "~/hammer/tools/amd/mlucas-avx2 -s l -cpu 0:15 > ~/hammer/tools/amd/mlucas_amm_log 2>&1 &"
+sendln "systemd-run --unit=mlucas-amm --property=Type=exec --property=Restart=always --property=RestartSec=2 /bin/bash -c '~/hammer/tools/amd/mlucas-avx2 -s l -cpu 0:15 >> ~/hammer/tools/amd/mlucas_amm_log 2>&1'"
 
 ;  TR518 TEST 
 wait prompt_sonic_root
@@ -154,5 +158,7 @@ while 1
     sendln "bgctl list"
     wait prompt_sonic_root
     sendln "jobs"
+    wait prompt_sonic_root
+    sendln "systemctl is-active mlucas-amm"
     pause 60
 endwhile