Chris G. Demetriou writes:
 > Robert Elz <kre@munnari.OZ.AU> writes:
 > 
 > > or the old SRMs do some init of the nvram that the newer ones
 > > don't do, or don't do properly (that's the kind of bug I can
 > > believe would not be found easily by the SRM developers, as they
 > > quite possibly never tested on a system which hadn't moved up
 > > through the SRM versions in sequence).
 > 
 > I'd bet on this.

Whatever the reasons, I can now add two more datapoints; both boards
now correctly boot without amnesia after doing 5.4 -> 4.9 -> 5.5

They also seem to shut down to SRM properly.

I did have one problem, however: at least the 4.9 fwupdate would only
work when run with a vga console; on serial it failed to get input.  

--
Dan.