feat(scenario): add nvbandwidth-acs-ab, an ACS off/on A/B for cross-socket P2P

Replaces the old nvbandwidth-all-gpu-power-watch scenario (dropped: the
power/IPMI-sampler framing was a dead-end — the blackbox shows every other
load test, incl. NCCL all-reduce across all 6 GPUs, passes with power fully
connected; only nvbandwidth on cross-socket GPU P2P reboots the box).

nvbandwidth-acs-ab is a two-phase reproduction for the MSI CG480-S6053 case.
Phase 1 disables PCIe ACS redirect at runtime via setpci (saving each
bridge's original ACS control word to /run/bee-acs-orig) then runs
nvbandwidth across all 6 GPUs — expected to survive. Phase 2 restores the
saved ACS values and repeats — expected to reproduce the reboot. Safe phase
first so its full blackbox reaches the USB stick before the risky phase can
reset the box (sync-bracket before the final job flushes phase-1 evidence).

ACS mutation runs via `bash -c` since scenario cmd is exec'd directly with no
shell. Both shipped copies (repo scenarios/ and image overlay) identical;
verified both parse via ParseScenarioJSON (4 jobs).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

iso/overlay/usr/share/bee/scenarios/nvbandwidth-acs-ab.json

@@ -0,0 +1,29 @@
+{
+  "name": "nvbandwidth-acs-ab",
+  "description": "Two-phase A/B for the MSI CG480-S6053 cross-socket GPU P2P hard-reboot. Phase 1 disables PCIe ACS redirect at runtime via setpci (P2P goes device-direct instead of up to the root complex) then runs nvbandwidth across all 6 GPUs — the run expected to SURVIVE. Phase 2 restores ACS to the BIOS default and repeats — the run expected to REPRODUCE the reboot. Safe phase runs first so its full blackbox reaches the USB stick before the risky phase can reset the machine.",
+  "timeout_sec": 1800,
+  "jobs": [
+    {
+      "name": "acs-disable",
+      "type": "command",
+      "cmd": ["bash", "-c", ": > /run/bee-acs-orig; for bdf in $(lspci -D | awk '{print $1}'); do setpci -s $bdf ECAP_ACS.w >/dev/null 2>&1 || continue; v=$(setpci -s $bdf ECAP_ACS+0x6.w 2>/dev/null) || continue; echo $bdf $v >> /run/bee-acs-orig; setpci -s $bdf ECAP_ACS+0x6.w=0000 >/dev/null 2>&1 || true; done; echo ACS disabled on $(wc -l < /run/bee-acs-orig) bridges; echo -n 'bridges still ReqRedir+: '; lspci -vvv 2>/dev/null | grep -c 'ReqRedir+' || true"]
+    },
+    {
+      "name": "nvbandwidth-acs-off",
+      "type": "command",
+      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
+      "gpu_indices": [0, 1, 2, 3, 4, 5]
+    },
+    {
+      "name": "acs-restore",
+      "type": "command",
+      "cmd": ["bash", "-c", "test -s /run/bee-acs-orig || { echo no saved ACS state; exit 0; }; while read -r bdf v; do setpci -s $bdf ECAP_ACS+0x6.w=$v >/dev/null 2>&1 || true; done < /run/bee-acs-orig; echo ACS restored on $(wc -l < /run/bee-acs-orig) bridges; echo -n 'bridges ReqRedir+ now: '; lspci -vvv 2>/dev/null | grep -c 'ReqRedir+' || true"]
+    },
+    {
+      "name": "nvbandwidth-acs-on",
+      "type": "command",
+      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
+      "gpu_indices": [0, 1, 2, 3, 4, 5]
+    }
+  ]
+}

iso/overlay/usr/share/bee/scenarios/nvbandwidth-all-gpu-power-watch.json

@@ -1,25 +0,0 @@
-{
-  "name": "nvbandwidth-all-gpu-power-watch",
-  "description": "Full nvbandwidth across all GPUs at once (the crash never reproduces on a single socket alone), with IPMI sensors and GPU power/temp sampled every 2s to check for a power-delivery correlation.",
-  "timeout_sec": 1800,
-  "jobs": [
-    {
-      "name": "ipmi-sensors",
-      "type": "sampler",
-      "interval_sec": 2,
-      "cmd": ["ipmitool", "sensor"]
-    },
-    {
-      "name": "gpu-power",
-      "type": "sampler",
-      "interval_sec": 2,
-      "cmd": ["nvidia-smi", "--query-gpu=index,power.draw,temperature.gpu,clocks.sm", "--format=csv"]
-    },
-    {
-      "name": "nvbandwidth-all",
-      "type": "command",
-      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
-      "gpu_indices": [0, 1, 2, 3, 4, 5]
-    }
-  ]
-}

scenarios/nvbandwidth-acs-ab.json

@@ -0,0 +1,29 @@
+{
+  "name": "nvbandwidth-acs-ab",
+  "description": "Two-phase A/B for the MSI CG480-S6053 cross-socket GPU P2P hard-reboot. Phase 1 disables PCIe ACS redirect at runtime via setpci (P2P goes device-direct instead of up to the root complex) then runs nvbandwidth across all 6 GPUs — the run expected to SURVIVE. Phase 2 restores ACS to the BIOS default and repeats — the run expected to REPRODUCE the reboot. Safe phase runs first so its full blackbox reaches the USB stick before the risky phase can reset the machine.",
+  "timeout_sec": 1800,
+  "jobs": [
+    {
+      "name": "acs-disable",
+      "type": "command",
+      "cmd": ["bash", "-c", ": > /run/bee-acs-orig; for bdf in $(lspci -D | awk '{print $1}'); do setpci -s $bdf ECAP_ACS.w >/dev/null 2>&1 || continue; v=$(setpci -s $bdf ECAP_ACS+0x6.w 2>/dev/null) || continue; echo $bdf $v >> /run/bee-acs-orig; setpci -s $bdf ECAP_ACS+0x6.w=0000 >/dev/null 2>&1 || true; done; echo ACS disabled on $(wc -l < /run/bee-acs-orig) bridges; echo -n 'bridges still ReqRedir+: '; lspci -vvv 2>/dev/null | grep -c 'ReqRedir+' || true"]
+    },
+    {
+      "name": "nvbandwidth-acs-off",
+      "type": "command",
+      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
+      "gpu_indices": [0, 1, 2, 3, 4, 5]
+    },
+    {
+      "name": "acs-restore",
+      "type": "command",
+      "cmd": ["bash", "-c", "test -s /run/bee-acs-orig || { echo no saved ACS state; exit 0; }; while read -r bdf v; do setpci -s $bdf ECAP_ACS+0x6.w=$v >/dev/null 2>&1 || true; done < /run/bee-acs-orig; echo ACS restored on $(wc -l < /run/bee-acs-orig) bridges; echo -n 'bridges ReqRedir+ now: '; lspci -vvv 2>/dev/null | grep -c 'ReqRedir+' || true"]
+    },
+    {
+      "name": "nvbandwidth-acs-on",
+      "type": "command",
+      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
+      "gpu_indices": [0, 1, 2, 3, 4, 5]
+    }
+  ]
+}

scenarios/nvbandwidth-all-gpu-power-watch.json

@@ -1,25 +0,0 @@
-{
-  "name": "nvbandwidth-all-gpu-power-watch",
-  "description": "Full nvbandwidth across all GPUs at once (the crash never reproduces on a single socket alone), with IPMI sensors and GPU power/temp sampled every 2s to check for a power-delivery correlation.",
-  "timeout_sec": 1800,
-  "jobs": [
-    {
-      "name": "ipmi-sensors",
-      "type": "sampler",
-      "interval_sec": 2,
-      "cmd": ["ipmitool", "sensor"]
-    },
-    {
-      "name": "gpu-power",
-      "type": "sampler",
-      "interval_sec": 2,
-      "cmd": ["nvidia-smi", "--query-gpu=index,power.draw,temperature.gpu,clocks.sm", "--format=csv"]
-    },
-    {
-      "name": "nvbandwidth-all",
-      "type": "command",
-      "cmd": ["dcgmi", "diag", "-r", "nvbandwidth", "-i", "{{gpus}}"],
-      "gpu_indices": [0, 1, 2, 3, 4, 5]
-    }
-  ]
-}