Using update_drv with jumpstart server?

lundman@lundman.net · *nix forums beginner Joined: 18 Mar 2005 Posts: 18

Ok this one was amusing.

Long time back, we added a line to "postinst" to eject the cdrom, so
that the machine can do the first reboot unsupervised. (and not boot
the jumpstart CD again).

The new machines use sata.

So the eject -f /dev/dsk/c*2 is a bit over-zealous. This would eject
the SATA /a mount as well, so the "last bits not quite yet saved" to
HDD would be lost. Which included my line addition to driver_aliases.

lundman@lundman.net · *nix forums beginner Joined: 18 Mar 2005 Posts: 18

Got a new set of servers in, with have a slightly newer flavour of
Intel gig ethernet. I found on sunsolve how to enable these nics using:

# update_drv -a -i '"pci8086,108c"' e1000g
# ifconfig e1000g0 plumb

and that does work nicely.

Then I moved to fix the jumpstart server to apply this correctly, the
jumpstart x86.miniroot itself was actually easy. gunzip, lofiadm,
modify /etc/driver_aliases, and rebuild it.

The I modified "postinst" script with jumpstart server to append the
same additional line to
/a/etc/driver_aliases, but I find that this fails after reboot. It can
not plumb e1000g0 in network/physical.

If I run the update_drv command (after taking out my driver_aliases
entry, since otherwise it just points out I already have it) it does
work, so clearly update_drv does more than just append to this file.
(truss confirms this).

I have attempted to run "chroot /a update_drv" from postinst, but I was
unable to solve this. (modctl unable to set major number..)

I have attempted to create a /etc/rcS.d/S10nic hoping it would be
executed before svcs run network-physical, but it is not.

What would be the proper way to solve this? I mean, I could add S10nic
to update_drv and force reboot, then the NEXT time it'll be ok, but
that feels a little cheesey.

I could modify network-physical script itself.. but perhaps there is a
cleaner way?

Google