I seem to be able to hang up a NetBSD process completely, in such
manner that it can't be killed and NetBSD won't reboot without
going to kernel debugger and telling it to reboot.

I am attempting to make an ISO image from files accessible via
NFS. The server is running on AmigaOS (NFSv2) and I suspect that
the server program I am using is buggy, which would cause
the trouble for NetBSD. Still, I think NetBSD shouldn't turn into
a vegetable while halting.

Mkhybrid starts reading data from the Amiga, which I know
as the hard drive rattles. After a while it just stops. Top
shows that it apparently is doing something, as mkhybrids
processor usage percentage constantly changes. ps shows the
"D" flag for mkhybrid - waiting for disk. This leads me to
think my Amiga NFS server is at fault.

When I try to break mkhybrid with CTRL-C nothing happens.
Issuing kill -9 to it is no good. And if mkhybrid is left
to do what ever it is doing, halt and reboot hangs. I just
get "syncing disks... 1 done". Usually I get "syncing disks... 3 done".

I tried killing mkhybrid (and nfsio processes) from kernel
debugger after reboot hung. I could see their PIDs in
debuggers ps output, but trying to kill those PIDs resulted
in "no such proc" message.


BTW, the Amiga NFS server isn't totally braindead. It quite well
in normal use, but as mkhybrid should pull about 400Mb of files in total for
the ISO image in this case, I get this trouble.

Oh, and another similar problem as above also appeared with
the Amiga's NFS server. As AmigaOS can very well do with rogue
programs, it once decided to crash on me, while the NetBSD box
had an NFS drive mounted from the Amiga. I tried umounting
the connection, but NetBSD didn't let me do it. Just said that
it wasn't currently mounted. Again, on reboot NetBSD hung up
in the syncing disks phase.

-- 
Arto Huusko  --  WWW: http://maailma.yok.utu.fi/
                  ** Divecalc **
The Diving Software @ http://maailma.yok.utu.fi/Divecalc