DF-2685/DF-2686 fix-validation kernel build log
=================================================

Validated kernel:  DragonFly 6.5-DEVELOPMENT #1: Sun Aug 30 19:03:41 UTC 2026
                   root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC  (x86_64)
                   /boot/kernel/kernel 119340672 bytes, Aug 30 19:10
Stock kernel was:  DragonFly 6.5-DEVELOPMENT #0: Thu Jul  2 06:02:54 UTC 2026

Exact commands run inside the guest (as root) for the validated kernel:

  cd /usr/src
  patch -p1 < /tmp/fix.diff                    # the two-hunk fix (see fix.diff in both packs)
  nohup sh -c 'make -j4 nativekernel > /tmp/kernel_build.log 2>&1; echo BUILD_RC=$? >> /tmp/kernel_build.log' &
  # ... build observed in progress via tail(1) over ~20 minutes ...
  # tail of /tmp/kernel_build.log at completion:
  #   --------------------------------------------------------------
  #   >>> Kernel build for X86_64_GENERIC completed on Sun Aug 30 19:07:49 UTC 2026
  #   --------------------------------------------------------------
  #   BUILD_RC=0
  make installkernel
  #   --------------------------------------------------------------
  #   >>> Kernel install for X86_64_GENERIC completed on Sun Aug 30 19:11:32 UTC 2026
  #   --------------------------------------------------------------
  reboot
  # after reboot:
  uname -v
  #   DragonFly 6.5-DEVELOPMENT #1: Sun Aug 30 19:03:41 UTC 2026  root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC

NOTE: the full untrimmed compiler output lived in the guest's tmpfs /tmp and
was lost across the validation reboot; the completion banners, BUILD_RC=0,
the installed-kernel timestamps and the running #1 kernel identity above are
the durable record. Patch presence in the built tree (verified post-boot):

  $ grep -n "VM_FAULT_USER_WIRE) != 0" /usr/src/sys/vm/vm_fault.c
  386:		 ((fs->fault_flags & VM_FAULT_USER_WIRE) != 0)))
  $ grep -c "pip_wakeup(fs->first_ba->object)" /usr/src/sys/vm/vm_fault.c
  13            (stock tree: 12 + the fixed line 2242)

Fix validation outcomes (full logs in the respective packs):
  DF-2685  stock:    bad=16253937 of 16777216 words zeroed by mlock (rc=2)
           patched:  bad=0, "OK: data intact after mlock" (rc=0)
  DF-2686  stock:    "vm_fault: pager read error" + refcount_wait objtrm1 long-wait
                     warnings repeating + victim stuck D1E objtrm1 forever
           patched:  pager error still surfaces (SIGSEGV), process exits fully,
                     zero long-wait warnings