Ken Nakata wrote:
> 
> Mon, 20 Jul 1998 20:31:58 -0400, Michael R. Zucca wrote:
> >
> > Is there some way this value can be calculated by default or is it strictly a
> > hand-tuneable item?
> 
> And, of course, in the perfect world, kernel should calculate this
> value at boot time and allocate memory dynamically, don't you think?
> 
> Ken

Ahh, just a belated "Me Too!".

I recently attempted upgrading to a current snapshot (980627, I think),
and after single-user panics for both SBC and NCR kernels, I then tried
a more recent kernel snapshot (GENERIC77) SBC and NCR. This also paniced
easily in single user (but did appear slightly more stable).

The SBC panic gave the following fingerprint:
# ls -l
uvm_fault
  type 8, code [mmu,,ssw]
trap type 8, code = 0x402074d, v = 0x759b78
kernel program counter = 0x36d0
kernel: MMU fault trap
db> trace
_Debugger
_panic
faultstkadj
_ncr5380_data_xfer
_ncr5380_machine
_ncr5380_sched
_ncr5380_scsi_cmd
_scsipi_execute_xs
...

The NCR panic looks remarkably similar:
# more z
pdma in: cnt 11550720 (0xb04000) (pending cnt 4096)
panic: something is wrong
db> trace
_Debugger
_panic
_do_ncr5380_drq_intr
_ncr5380_drq_intr
_rbv_intr
_lev2intr
_mi_switch
_tsleep
_biowait
_bread
_ffs_blkatoff
...

BTW, this is a normally working IIsi, 17Mb RAM, FPU, 1Gb Quantum Fireball.
I normally run with a custom SBC kernel with reconnects enabled. At the
moment, I'm not running anything :(

Could this be the abovementioned UVM tuning problem?

Cheers,
-- 
Paul Ripke
BHP Information Technology
Open VMS, AXP & UNIX (AIX, HP/UX, DG/UX, SCO, SGI, Digital, SunOS...) Sysadmin
Computer Centre,    Five Islands Rd,    Port Kembla,    NSW 2505,    AUSTRALIA
   ripke.paul.pr@bhp.com.au  weripp@itwol.bhp.com.au  stixpjr@ozemail.com.au
            Anyone wishing to lay claim to the opinions expressed
                      herein, do so at their own risk.