Fatal trap 12: page fault while in kernel mode   [pre-v2-fix stress, boot @ ~05:47]
cpuid = 3; lapic id = 3
[...]
current process		= 633            (an /root/nvspam re-glob/re-open)
kernel: type 12 trap, code=0

CPU3 stopping CPUs: 0x00000037
 stopped
Stopped at      nvme_open+0x13:  testb   $0x2,0x44c(%rdx)
db>

# Context: 6x detach/attach cycles with the v1-fixed kernel. v1 closed the
# ioctl path; this crash is the SAME teardown-lifetime defect surfacing
# through nvme_open(): a spammer's open() dispatched on a devfs node whose
# si_drv1 still pointed at the kfree'd softns, so `sc = nsc->sc` yielded a
# dangling softc pointer and the NVME_SC_UNLOADING check faulted reading
# sc->flags (offset 0x44c). Motivated fix v2: NULL cdev->si_drv1 at
# nvme_disk_detach() + NULL guards at every dispatch entry.
# (Reconstructed from the serial console tail captured at crash time; each
# QEMU boot truncates boot-nvme.log, so the raw block was not retained.)