>         [ ... random TLB miss seen on -current kernel ... ]
>          
[...]
> >> trap: TLB miss (load or instr. fetch) in kernel mode
> >> status=0x8fc34, cause=0x30000008, epc=0x8014aed8, vaddr=0x200c5ba8
> >> pid=109 cmd=syslogd usp=0x7ffffc10 ksp=0xc21cbe90
[...]
> I think the symptom was fixed by the recent change of splsoftclock()
> definition made by Jason Thorpe.  Can someone comfirm whether the
> ramdom panic is really cured in -current 1.4J kernel?

I built a -current kernel (netbsd14J) from yesterdays sup (from
the german mirror sup.de.netbsd.org).
It is up now for some hours and this looks very promising.
I used it with heavy disk and processor load (making a new kernel
and simultaneously using elm on a 35 MB mailfile with only 24 MB memory
in the box) and so far no problems. Usually the TLB misses came
immediately on startup or within the first 5..10 minutes.

THANK YOU VERY MUCH!

After Michael Hitch changes which solved the

	asc_get_status: cmdreg 11, fifo cnt 7
	rz0: Illegal request

panics this is the first time for over a year that I can build my own
working kernel on -current!

GREAT WORK!

Andreas