hi,

I'm experimenting with a diskless setup and found that there's a
really easy way to crash the client (amd486DX4/100, 3c509, Quantum
TRB850A running netbsd-current)(*): just do a
   dd if=/dev/rwd0d of=/dev/null bs=64k
and the deed is done! :-/
ok, that box's not completely diskless: I use the disk of this thing
to get the netboot code from and for /var and swap. the kernel is
loaded from an ss20 (running netbsd-current, of course), which holds
the root and user filesystems for my client, too.

a backtrace on the client looks like this:
_trap()
--- trap (number 4) ---
_wdc_ata_bio_intr()
_wdcintr()
_Xintr14()
--- interrupt ---
_idle()
bpendtleep()
_physio()
_wdread()
_spec_read()
_nfsspec_read()
_vn_read()
_dofileread()
_sys_read
_syscall
--- syscall (number 3) ---

on the server I see the following line in /var/log/messages:

May  7 15:27:49 dino /netbsd: nfsd send error 64


what problem am I stumbling over here? Is this a known "feature"?
any further info that I should provide?

-ks

ps.: This is not a real problem for me, as joe luser isn't able to login
     manually on this boxes. He gets an xterm connected to our
     opac-server right away on booting the beast.

(*) recycling old hardware to build "xterms" and "thin clients"
    replacing non-functional and unmaintainable Windoze-PeCees ;-)