On Tue, Dec 07, 2004 at 01:57:03PM -0500, Tim Kelly wrote:
> At 10:42 AM -0800 12/7/04, Matt Thomas wrote:
> >For what it's worth, my g4 has been running fine with that change.
> 
> It breaks MP completely, and the sync mesh patch Michael has is essential
> to both Riccardo's disk drive operating at all and illuminating problems
> within the interrupt code, but the commit breaks this and causes the hang
> Riccardo describes. I had independently confirmed this during my testing of
> your patch, which is why I asked what locore_subr.S he had.

my 9500/180MP (running an MP kernel) behaves the same with or without
rev 1.26 of locore_subr.S:  works fine except for hanging when setting
the time (either by hand or with ntpdate) or rebooting (after unmounting
the file systems).  resettodr() is the common element.

so this change doesn't seem to have caused any problems for this dual 604e box.
(and it works fine on my dual g4 as well, same as for others.)

the change was definitely a move in the right direction, do_pending_int()
expects PSL_EE to be on, and the Idle code was wrong for not satisfying that.
it's useful that it has reliably exposed other problems, it should help us
figure out how to fix them.

-Chuck