On Sun, 12 Nov 1995, Allen Briggs wrote:

> Hmmm...  I wonder if it's related to the crash that I very occasionally
> see.  I assumed that I was getting it because of flaky hardware.  When
> one of my partitions gets nearly full, I get one or two "Bad block XX"
> messages where XX is some impossible number.  Then a panic: freeing free
> frag.

I saw that panic once; it seems to be something corrupting just the 
metadata, from what I have seen.  In other words, the problems primarily 
affect a contiguous segment of inodes, and invariably have "PARTIALLY 
ALLOCATED INODE.  CLEAR?" when I do an fsck from my miniroot.

I usually notice the problem when I try to run a binary and get something 
to the effect of "ls: bad fd" from the shell.

> > One common factor that seems to be coming up is blocks from apparently 
> > random places get thrown into another part of the filesystem.
> 
> Someone else saw this on a slow (MO) drive.  I have not seen it on my
> system at all.

(blush) I haven't seen it either.  I was mistaken, and I'm very glad I 
used the word "seems"... it reflected my lack of certainty well.  I'm 
still looking for clues to what the problem might be.

> The driver probably has a wierd bug or two.  It's also possible that you
> have bad termination or that one of the devices has a bug (unlikely).

I've triple-checked the termination, now.  Would enabling SCSI debugging 
in the kernel show some definitive message if there were termination 
problems?

> How is your cable length?

The internal 8" or 12" inch ribbon, a 3' to the external CD-ROM, and an 
18" to the external Viper 150 tape (which has a standard terminator; 
should it be one of those funky Apple pass-through types?).

By the way, thanks for your efforts, Allen.  I realize that you're very 
busy, and I really appreciate your time.

--scott