DF-5833 / run.final.log
=== DF-5833 ckpt2 rig, FINAL run (2026-10-13 15:0x, guest DragonFly 6.5-DEVELOPMENT #0 X86_64_GENERIC) ===
probe: note_off=17816 note_sz=160940 prstatus=201 vn=4 off_end=183696
tune: K=16 n_pad=0 residue=4064 (stage=ckpt2)
ATTEMPT 1: WORKERS-DRAINING done=0 children=3586 sleep=200
ATTEMPT 1 RESULT: img=285392896 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=3639 sleep=200
ATTEMPT 1: NT_PRSTATUS=3822 expected=3803 (race += 19)
ATTEMPT 2: WORKERS-DRAINING done=3 children=3318 sleep=200
ATTEMPT 2 RESULT: img=285175808 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=3339 sleep=200
ATTEMPT 2: NT_PRSTATUS=3549 expected=3535 (race += 14)
ATTEMPT 3: WORKERS-DRAINING done=1 children=3364 sleep=200
ATTEMPT 3 RESULT: img=285212672 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=3453 sleep=200
ATTEMPT 3: NT_PRSTATUS=3597 expected=3581 (race += 16)
ATTEMPT 4: WORKERS-DRAINING done=0 children=53 sleep=200
ATTEMPT 4 RESULT: img=282562560 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=96 sleep=200
ATTEMPT 4: NT_PRSTATUS=286 expected=270 (race += 16)
ATTEMPT 5: WORKERS-DRAINING done=1 children=3316 sleep=200
ATTEMPT 5 RESULT: img=285175808 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=3613 sleep=200
ATTEMPT 5: NT_PRSTATUS=3550 expected=3533 (race += 17)
ATTEMPT 6: WORKERS-DRAINING done=0 children=3630 sleep=200
ATTEMPT 6 RESULT: img=285425664 victim_alive=no
CKPT-RESULT rv=0 errno=0 children=3776 sleep=200
ATTEMPT 6: NT_PRSTATUS=3865 expected=3847 (race += 18)
CONTROLLER-DONE
--- victim.log (attempt 6)
VICTIM-SETUP sleep=200 workers=16 pad=0 filemaps=0
WORKERS-DRAINING done=0 children=3630 sleep=200
CKPT-RESULT rv=0 errno=0 children=3776 sleep=200
--- eater.log (excerpt)
eater: stalled (2074593 us) - at floor
EATER-READY free=18944 eaten=1160MB chunks=1160
DRIP 4194304 free=16653
--- parsed images
attempt 1: e_phnum=345 note_off=19384 note_sz=3057740 NT_PRSTATUS=3822
attempt 4: e_phnum=345 note_off=19384 note_sz=228940 NT_PRSTATUS=286
attempt 6 (parse2, complete-header analysis):
phnum=347 note_off=19496 note_sz=2670540 seg1_off=2695168
vn_count=4 siginfo csi_ckptpisz=4320 (correct) filehdr nfiles=1 (the ckpt fd)
WRITE_end_raw=2695056 round_page=2695168 seg1_off=2695168 CONSISTENT
-> the WRITE pass consumed the reservation to its final 112 bytes:
the DRYRUN->WRITE LWP delta was 1..6 notes (800 bytes each),
absorbed by the round_page slack.
=== earlier decisive runs ===
[ckpt2 v2, attempt 4] CKPT-RESULT rv=-1 errno=22 (EINVAL) children=98 (call-time snapshot 93):
the WRITE pass required more note space than the DRYRUN reserved ->
target_reserve wall hit mid-walk -> putallnotes error SWALLOWED by
__elfN(puthdr) -> dump aborted with EINVAL, no image. This is the
reservation-mismatch failure mode firing on the guest.
[ckpt2 v6, attempt 4] children=1448 done=10 -> NT=1671 expected=1665 (race += 6)
(measured with corrected arithmetic: 6 LWPs inserted between the
pre-syscall snapshot and the WRITE pass, dump still succeeded)