A Dangerous conf bug/feature in 4.?BSD

[Amos Shapir]

We are running a vax/785 under 4.3BSD (but this might happen also in 4.2).
We have 3 disks on hp controllers - 2 rm03, 1 rp07; the latter is just back
from the shop and wasnt mounted. The configuration is:
disk            hp0     at mba? drive 0
disk            hp1     at mba? drive ?
disk            hp2     at mba? drive ?
which usually comes out at boot time as:
hp0 at mba0 drive 0
hp1 at mba0 drive 1
hp2 at mba1 drive 2	(other devs & mba's deleted)

Today, 2 strange things happened: a) a maintennance man noticed 1 of the
rm03 had the 'fault' on - it was just a fuse, but this drive is usually
mounted on /tmp, but the system didnt complain! (not even when it was
put offline to be fixed).
b) trying to newfs hp2 (which we thought was the rp07), we got
write error starting at some block.
'df' confirmed the suspicion: /tmp was now mounted on the rp07! A look
at /usr/adm/messages from a few days (and reboots) back, proved what had
happened:
hp0 at mba0 drive 0
hp1 at mba1 drive 2
Trying to autoconfigure, the driver skipped the faulted rm03, and configured
the rp07 as hp1. When accessing hp2, it became configured to the 2nd rm03,
which was fixed by now.
Luckily, the rp07 didnt contain any usefull information (and what it had
was cleaned by the /tmp-cleaning utilities); on a system which keeps some
disks unmounted, this phenomenon may be dangerous.
-- 
Amos Shapir
National Semiconductor (Israel)
6 Maskit st.
Herzlia 46104, Israel
34.49'E 32.10'N