[parisc-linux] CONFIG_SMP update
Grant Grundler
grundler@cup.hp.com
Fri, 26 Jan 2001 16:47:53 -0800
[ This was posted to the CVS mailing list - I thought it
might be of general interest. -ggg
]
CVSROOT: /home/cvs/parisc
Module name: linux
Changes by: grundler 01/01/26 17:33:29
Modified files:
arch/parisc/kernel: Makefile head.S irq.c pdc.c setup.c smp.c
arch/parisc64/kernel: Makefile head.S
arch/parisc64/lib: Makefile
include/asm-parisc: pdc.h
Log message:
Two sets of fixes here: Makefiles and round two of CONFIG_SMP
First the news :^) (it's 11 now Matt ;^)
hppa64 won't build on head:
o hppa64-linux-gcc is broken. Workaround is to define lock in spinlock_t
as unsigned long and change __ldcw() macro to use ldcd isns.
amodra has fixed gcc in the devel branch but some other problems
prevent him from making that mainline at the moment.
o parisc64 Makefile complete the merge for those files.
o problem is pbus_set_ranges() isn't defined.
Looks like we picked up the new drivers/pci stuff anyway. *sigh*.
I'll look at this first thing next week.
parisc* Makefiles weren't using proper rules to generate .o from .S files.
This is a first hack. Willy will follow up with "clean" versions.
Second round of CONFIG_SMP fixes. I now have both CPU's "online"
and in wide-mode. IPI isn't working. And removing the mdelay's
in smp.c exposes a race condition I haven't figured out yet.
(it's not pdc calls or console output - unless our spinlocks are
broken). If anyone wants to look at either problem, I suggest with
a 2-way j5k running 32-bit to remove some of the variables.