Date:        Thu, 20 Sep 2001 11:17:05 +0200
    From:        Manuel Bouyer <bouyer@antioche.lip6.fr>
    Message-ID:  <20010920111705.C8195@antioche.lip6.fr>

  | No, I don't.

OK, I must have misinterpreted what you said earlier (what was
actually happening).

  | raidframe should overwrite the root device with raid0a.
  | It works this way on i386 and alpha, at last.
  | Greg explicitely added code to raidframe to do this, if I read it properly
  | the magic happens in rf_netbsdkintf.c:rf_buildroothack()

Yes, I see - that sets booted_device ... unfortunately the sparc port doesn't
use booted_device in the call to setroot() the way that most of the other
ports do, instead it uses the data struct it used to parse the boot path
in the call to setroot() - see arch/sparc/sparc/autoconf.c:cpu_rootconf()

The RAIDframe code setting booted_device on the sparc isn't going to
achieve anything useful (why booted_device is being set at all on the
sparc I'm not sure, it doesn't seem to be used for anything I could see).

kre