Build/install session for the DF-5432 single-fix kernel (guest /usr/src,
with-src snapshot, kernel A lineage). Captured outputs (the full untrimmed
logs lived at /tmp/build.log and /tmp/install.log inside the guest and were
lost with the required post-validation `vm.sh reset with-src`; the decisive
lines below were captured from the live session):

  $ cd /usr/src && git apply /tmp/fix.diff
  Checking patch sys/platform/pc64/acpica/acpi_madt.c...
  Hunk #3 succeeded at 638 (offset -3 lines).
  APPLIED
  577:madt_ioapic_addr_in_ram(uint32_t addr)
  641:		if (madt_ioapic_addr_in_ram(ioapic_ent->Address)) {

  $ make -j6 nativekernel KERNCONF=X86_64_GENERIC   (nohup, /tmp/build.log)
  ...
  linking kernel.debug
  12379249  1538723  10350144  24268116  1724d54  kernel.debug
  --- kernel.stripped ---
  objcopy --strip-debug kernel.debug kernel.stripped
  ... (module builds continue) ...
  <=== libiconv
  ------------------------------------------------------------
  RC=0                                    <- no -Werror failures, no warnings

  $ make installkernel KERNCONF=X86_64_GENERIC
  >>> Kernel install for X86_64_GENERIC completed on Sun Oct  4 20:50:56 UTC 2026
  RC=0
  -r-xr-xr-x  1 root wheel 119342760 Oct  4 20:49 /boot/kernel/kernel

Patched-kernel test: see run.fixed.log + monitor_fixed.txt (same blob, same
hint; MADT ioapic enumerator rejected, mptable fallback used, attacker page
untouched).