I did some fooling around with USB_DEBUG and related defines today.

After watching the output roll by on the iMac screen for a while, I
began to suspect that the complicated internal hubbage of the iMac
keyboard might have something to do with it.

So I dug up a cheap Win98 USB keyboard from another project and gave
that a try. Lo and behold, the iMac doesn't want to panic any more.

Interestingly, the keyboard works correctly while it is asking for the
root device. However once it goes to multi-user the keyboard is totally
dead -- the "shift" workaround has no effect, and I see no messages from
the USB drivers at all.

So we appear to have one problem which is iMac/Apple _keyboard_ specific
and another which is not.

1.4.2 works fine (modulo a few quirks during boot) with both keyboards.

I know that -current started to go downhill in mid-may, as of the "glue"
cleanup changes. After all the fooling around I've done, I think the glue
changes weren't the cause, but they somehow managed to expose the failure.

Now that I can prevent the panic, I think I'll see if I can isolate the
commit that brings in the dead-keyboard problem for me.

Todd Whitesel
toddpw @ best.com