# DF-3534 fix validation (kernel #1 + DF-3531 layer-2 module + ngt_shutdown
# NG_NODE_UNREF fix; vmstat -m | grep -iE netgraph):
=== FINAL FIX STATE: kernel#1 + module(layer2+3534) ===
DragonFly dfbsd 6.5-DEVELOPMENT #1: Mon Sep 14 09:24:11
--- baseline ---
      netgraph_node       2     256       0    390M        2
--- nodd 500 ---
done nodd 500
           netgraph       7    1008       0    390M      507
      netgraph_node       2     256       0    390M      502      <-- FLAT
--- stall-close 2000 ---
done: 2000 rounds (mode=stall/leak)
           netgraph       9   1.27K       0    390M    2.45K
      netgraph_node       2     256       0    390M    4.40K      <-- FLAT
--- stall-close 2000 (2nd) ---
done: 2000 rounds (mode=stall/leak)
           netgraph       9   1.27K       0    390M    4.40K      <-- FLAT
      netgraph_node       2     256       0    390M    8.30K      <-- FLAT
(hooks/paths/socks in-use 0 throughout; ~9K node allocations cycled
 through 4500 lifecycles with in-use never moving off 2)
#
# NOTE: a later drain-mode stress run in the same session panicked with
# 'memory chunk already free' — attributed to the DF-3531 layer-1/2
# close-vs-fire race (in-flight timer item double-free), NOT to this
# node-ref fix (nodd mode — no timers involved — clean and flat).