On Mon, 26 Oct 1998 17:05:54 -0800  Jonathan Stone wrote:

> >>
> >>     c) 5100s :)
> 
> >Anyone else with one of these?
> 
> if there is, they're keeping quite.
> 
> >Again, failure at the same point (after the "WARNING: clock gained 117
> >days -- CHECK AND RESET THE DATE!" message).  Snoop from a Sun shows the
> >last thing it did was send read request for 8 kBytes from /sbin/init and
> >a reply was sent.  This is running diskless in case you didn't guess -
> >if you think it will help I can probably find a disk to try and boot
> >off.
> 
> Drat.  double drat. Hm. what happpens if you try an NFS mount with a
> read/write size of 512 bytes or 1k?  (is that easy to specify?)

Just tried 1k blocks (change some magic numbers in nfs_vfsops.c to do
this - there seemed to be no easy way).

> Re disks: knowing if the SII is being initialized and the SCSI bus
> probed might be helpful, it would eliminate another cause of potential
> lossage.

Probes ok, see previous message.

> Hm. is this dying at the point where the kernel is opening /dev/consoe
> "for real" rather than using its internal kluded-up version?  that
> might be worth looking into.

Tried commenting out the "*cn_tab = dccons;" around line 245 in
dev/dc.c, still no better.

Time to give up today - I'll look further tomorrow.

Simon.

ps:  I needed the following to stop the "le0: device timeout" messages
during the bootp phase.  Want me to commit it?

Index: pmax/pmax/dec_5100.c
===================================================================
RCS file: /cvsroot/src/sys/arch/pmax/pmax/dec_5100.c,v
retrieving revision 1.2
diff -c -r1.2 dec_5100.c
*** dec_5100.c	1998/04/27 06:10:32	1.2
--- dec_5100.c	1998/10/27 02:03:29
***************
*** 236,241 ****
--- 236,243 ----
  #endif
  	}
  
+ 	icsr = *((volatile u_int *)MIPS_PHYS_TO_KSEG1(KN230_SYS_ICSR));
+ 
  	/* handle clock interrupts ASAP */
  	if (mask & MIPS_INT_MASK_2) {
  		temp = c->regc;	/* XXX clear interrupt bits */