Running: netbsd-current as of late 11/93, with the Soeren Schmidt
	wd driver "fixes" (hey, they allow me to backup my boot disk
	w/o hanging..).
	486DX/33, 2xIDE drives (boot/swap), 1xSCSI disk (extra space..)

I just had my system hang with the WD light on; the system reported:

wdc0: busy too long, resetting
wdc0: busy too long, resetting
wdc0: busy too long, resetting

(with a delay in between each one).

A reboot failed, with the boot blocks "hanging" (with the wd drive
light *off*) after two or three turns of the "spinner".

Booting from floppy discovered that portions of the ilist of the boot
disk had been toasted; the kernel was one of these.  Does the boot
code do any sanity checking on the disk addresses they read from the
inode?

All of the inodes which fsck complained about were in a small range,
roughly between ~50 and ~75-80 (I should have taken notes during the
fsck, but didn't); they could have been nailed by a single
errant/misdirected/whatever buffer cache write (there are 64 inodes
per 8K disk block).

					- Bill

------------------------------------------------------------------------------