feat(proxmox): add e1000e offload fix script for Intel NIC hang bug

Script disables TSO/GRO/GSO/tx/rx offload on all e1000e NICs to prevent
the known "hardware unit hang" bug affecting Intel I217-LM, I219-LM, and
82571EB controllers under Proxmox. Applies live + installs systemd
service for persistence.

Affected hosts identified: tsys1 (already off), tsys3 (already off),
tsys4 (needs fix), tsys5 (nic1/nic2, needs fix), tsys9 (needs fix).
tsys6/7 use Broadcom (bnx2/tg3), not affected.

Script is ready but NOT YET DEPLOYED — pending review/approval via
Redmine [#416]. Initial attempt to deploy did not execute properly.

Based on archive/provisioning/Dell/Server/fixeth.sh.
Refs: https://forum.proxmox.com/threads/e1000-driver-hang.58284/

💘 Generated with Crush

Assisted-by: Crush:glm-5.2

WORKING.md

@@ -5,6 +5,4 @@ A commit is blocked while any task below remains unchecked.
 
 ## Current Tasks
 
-- [x] Install iperf3 on all hosts + production VMs (persistent)
+- [x] Create e1000e offload fix script (repo only, not deployed yet)
-- [x] Create reusable any-to-any perf test script
-- [x] Enable iperf3 systemd service on hosts for on-demand testing

proxmox/perf/scripts/fix-e1000e-offload.sh

@@ -0,0 +1,79 @@
+#!/bin/bash
+###############################################################################
+# fix-e1000e-offload.sh — Disable offload on e1000e NICs to prevent hangs
+#
+# The Intel e1000e driver has a known bug where TSO/GSO/GRO offload causes
+# "hardware unit hang" resets on certain Intel NICs (I217-LM, I219-LM, 82571EB).
+# This script disables offload features on all e1000e interfaces and installs
+# a systemd service to persist across reboots.
+#
+# References:
+#   https://forum.proxmox.com/threads/e1000-driver-hang.58284/
+#   https://serverfault.com/questions/616485
+#
+# Run on each Proxmox host:
+#   PROX_HOST=pfv-tsys4 bash tests/remote.sh prox-file proxmox/perf/scripts/fix-e1000e-offload.sh
+###############################################################################
+set -euo pipefail
+
+echo "=== e1000e Offload Fix on $(hostname) ==="
+
+# Find all e1000e physical NICs (skip bridges, bonds, virtual interfaces)
+AFFECTED_NICS=()
+for nic_path in /sys/class/net/*; do
+    nic=$(basename "$nic_path")
+    [ "$nic" = "lo" ] && continue
+    # Skip bridges, bonds, virtual interfaces
+    [ -d "${nic_path}/bridge" ] && continue
+    [ -d "${nic_path}/bonding" ] && continue
+    case "$nic" in
+        tap*|veth*|fwpr*|fwln*|vmbr*|datanet*|storagenet*|tailscale*) continue ;;
+    esac
+
+    driver=$(ethtool -i "$nic" 2>/dev/null | awk '/^driver:/{print $2}')
+    if [ "$driver" = "e1000e" ]; then
+        AFFECTED_NICS+=("$nic")
+        echo "  Found e1000e NIC: $nic"
+    fi
+done
+
+if [ "${#AFFECTED_NICS[@]}" -eq 0 ]; then
+    echo "  No e1000e NICs found. Nothing to do."
+    exit 0
+fi
+
+# Apply fix live
+echo ""
+echo "--- Disabling offload features ---"
+for nic in "${AFFECTED_NICS[@]}"; do
+    echo "  $nic:"
+    ethtool -K "$nic" tso off gro off gso off tx off rx off 2>&1 | sed 's/^/    /' || true
+    tso_state=$(ethtool -k "$nic" 2>/dev/null | awk '/tcp-segmentation-offload/{print $2}' | head -1)
+    echo "    tso=$tso_state"
+done
+
+# Install systemd service for persistence
+echo ""
+echo "--- Installing systemd service ---"
+{
+    echo "[Unit]"
+    echo "Description=Disable offload on e1000e NICs (prevent hardware unit hang)"
+    echo "After=network.target"
+    echo "Wants=network.target"
+    echo ""
+    echo "[Service]"
+    echo "Type=oneshot"
+    echo "RemainAfterExit=yes"
+    for nic in "${AFFECTED_NICS[@]}"; do
+        echo "ExecStart=/sbin/ethtool -K $nic tso off gro off gso off tx off rx off"
+    done
+    echo ""
+    echo "[Install]"
+    echo "WantedBy=multi-user.target"
+} > /etc/systemd/system/fix-e1000e-offload.service
+
+systemctl daemon-reload
+systemctl enable fix-e1000e-offload.service
+echo "  Service installed and enabled (fix-e1000e-offload.service)"
+echo ""
+echo "=== Done. Affected NICs: ${AFFECTED_NICS[*]} ==="