On Tue, 10 Sep 1996, Andrew Gallatin wrote:

.. I *think* this means that the kernel is in a.out format, and the PROM
.. wants it in ecoff -- did you run elf2ecoff on the kernel before
.. booting? 
No, but now I have tried it, see below.
The sources to elf2ecoff.c (from about 26 June 1996) had to be tweaked 
quite a bit due to changes in sys/exec_ecoff.h. I recall to have seen
commit log messages telling why the ECOFF headers were changed, and others
have possibly already updated elf2ecoff and friends. Otherwise, I can
send my updated version to whoever is in charge for it.

.. 
.. As I remember it (I've changed jobs and haven't seen a DECstation in 2
.. months), "?IO: 3/tftp, a.out err (-2)" means that the kernel is in the
.. wrong format, and an error like "1033120?IO:3/tftp, a.out err (-5)"
.. means that the text segment is too large for the PROM to deal with.

Yes, now it answers
   808000
   ?IO: 3/tftp, a.out err (-5)

Running the size command on NetBSD and on Ultrix for netbsd.elf and netbsd
(which is the output from elf2ecoff) gives the following results:

On NetBSD:
size netbsd.elf 
text    data    bss     dec     hex     filename
722144  85848   61968   869960  d4648   netbsd.elf

On Ultrix:
sok size netbsd.elf 
text    data    bss     dec     hex
0       808000  61972   869972  d4654
sok size netbsd
text    data    bss     dec     hex
808000  0       61972   869972  d4654

Is there an explanation for the differences between "size netbsd.elf" on
the two operating systems ?

.. 
.. This latter type of error is the type you'll get on the 'problem'
.. machines.  As Jonathan mentioned, to get one of the problem machines
.. to boot, you'll need to build a stripped down kernel.  Using the
.. Ultrix size on the elf2ecoff'ed kernels, I've found that a kernel of
.. this size will not boot:
.. text    data    bss     dec     hex
.. 1033120 0       89236   1122356 112034
.. 
.. And one of this size will:
.. text    data    bss     dec     hex
.. 855024  0       69556   924580  e1ba4

This does not comfort me very much as my kernel is smaller than any of
yours and still it fails to load.

I will, however, try to minimize it further, but I will probably not be
able to do any further experiments until next week.

/ Erik Bertelsen.