I just booted into single user mode, and this happened: (I'm running
GENERIC on a soekris with LFS on /)

# mount -a -w /dev/wd0a /
# mount
/dev/wd0a on / type lfs (local)
mfs:12 on /dev type mfs (synchronous, local)
# uvm_fault(0xc088fca0, 0xaa55a000, 0, 2) -> 0xe
kernel: supervisor trap page fault, code=3D0
Stopped in pid 9.1 (ioflush) at netbsd:getnewbuf+0xe1:  movl    %eax,0x5c(%=
edx)
db> bt
getnewbuf(0,0,0,0,c0cf6000) at netbsd:getnewbuf+0xe1
getblk(ca38cf18,9d924,0,800,0) at netbsd:getblk+0x176
lfs_writeinode(c0cf6000,caddcf68,ca38ded8,1,200) at netbsd:lfs_writeinode+0=
xf7
lfs_segwrite(c0cf7000,5,c08134a4,c0c21d00,206) at netbsd:lfs_segwrite+0x24c
lfs_sync(c0cf7000,3,ca380f6c,ca38bad4,ffffffff) at netbsd:lfs_sync+0x4f
sync_fsync(cad03f14,0,cad03f3c,c0404cac,c0653e20) at netbsd:sync_fsync+0x8e
VOP_FSYNC(cadd3694,ca380f6c,8,0,0) at netbsd:VOP_FSYNC+0x4c
sched_sync(ca38bad4,966000,96f000,0,c0100321) at netbsd:sched_sync+0x1b3
db> sync
syncing disks... uvm_fault(0xc088fca0, 0, 0, 1) -> 0xe
kernel: supervisor trap page fault, code=3D0
Stopped in pid 10.1 (aiodoned) at       netbsd:lfs_cluster_aiodone+0x19c:  =
    m
ovl     0x24(%edi),%eax