Jamie Scuglia wrote:

> The funny thing is that it's not just happening on one machine we have,
> but many.  I thought it would be too much of a co-incidence to think
> that all these machines have bad memory somewhere.  All worked fine
> under Ultrix and were subjected to much heavier load.

I doubt it's a hardware problem.

> Anyway, re-seating the memory cards didn't help.  On some systems,
> we immediately get a segmentation fault when running "dump".  After
> rebooting the machine, it behaves better, but then still crashes
> the machine later for certain largish filesystem dumps.  It's almost
> like the "dump" command is responsible for this.
> 
> We get many different errors like these few:
> 
> 	trap: address error (store) in kernel mode
> 	panic: utlbmod: invalid segmap
> 
> 	trap: TBL miss (load or instr-fetch) in kernel mode
> 
> 	CPU memory read timeout error at 0x027361bc
> 	panic: panic("Mem error interrupt");
> 
> We've used "tar" to tar up a partition that has one Gb of data with
> no problems.
> 
> Is anyone using "dump" to backup their machines?  Or what are people
> using?

The problem seems to be related to accessing the character device (which
dump does and tar does not).  The problem is reproducable, but I don't
believe there's a known solution at this time.  Also the problem is
filesystem related and not tape drive related - the box I'd like to
backup has no local tape drive.  The last time I tried a -current kernel
with dumping was around November 21 with a 1.4O kernel and the problem
still existed then.

As far as backups, at the moment I don't :-(

Simon.