Guest: DragonFly dfbsd 6.5-DEVELOPMENT x86_64 (QEMU/KVM, 6 vCPUs, single-tenant audit guest)
Kernels used:
  baseline: #0: Thu Jul  2 06:02:54 UTC 2026  root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC  (stock, bug present)
  fixed   : #1: Wed Oct 14 08:13:16 UTC 2026  root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC  (same tree + fix.diff, rebuilt via `make nativekernel`)
Compiler: cc (gcc 8.3 [DragonFly] Release/2019-02-22)
hw.ncpu=6  kern.ipc.nmbclusters=33296  hw.machine=x86_64
Guest /usr/src/sys/dev/netif/wb/if_wb.c MD5 = bec15392abf0facf8fe77b4aefb23938 = audit-tree sys/dev/netif/wb/if_wb.c (byte-identical, 1638 lines)
Fixed kernel: /usr/src/sys/dev/netif/wb/if_wb.c line 976 = `if (rxstat & WB_RXSTAT_OWN)` (fix.diff applied)
wb is built into X86_64_GENERIC (sys/config/X86_64_GENERIC:246 `device wb`):
  ffffffff808dae30 t wb_rxeof   (same address in BOTH kernels — the one-line fix did not move the symbol)
  ffffffff808da6d0 t wb_init
  ffffffff808d9850 t wb_stop
  (wb_newbuf exists only as IPA-SRA clone `wb_newbuf.isra.5` — compiler-mangled ABI, deliberately NOT called directly)
Hardware: NO W89C840F / Compex RL100-ATX NIC in the guest (pci list has none) — the NIC is simulated
at the driver's only observation point: descriptor status/buffer writes, exactly what the chip's DMA
engine produces. Rig details in df4266_wb_poc.c header comment and VERDICT.md.
Module build: cc -O2 -Wall -Werror ... -DKLD_MODULE (see kld_build.log, clean, zero warnings)