[PATCH] ppc: fix powersave with interrupts disabled

[Brian Waite <linwoes@gmail.com>]

On Thu, 13 Jan 2005 15:58:34 -0500, Brian Waite <linwoes@gmail.com> wrote:
> On Wed, 12 Jan 2005 01:41:19 +0000, Linux Kernel Mailing List
> <linux-kernel@vger.kernel.org> wrote:
> > ChangeSet 1.2369, 2005/01/11 17:41:19-08:00, tglx@linutronix.de
> >
> >         [PATCH] ppc: fix idle with interrupts disabled
> >
> >         The idle-thread-preemption-fix.patch in mm1/2 leads to a stalled box on PPC
> >         machines which do not provide a powersave function and therefor poll the
> >         idle loop with interrupts disabled.  The patch reenables interrupts.
> There is still a stall with PPC  boxes that have powersave enabled. I
> use a 74xx based board and unless I disable powersave
> (ppc_md.power_save=NULL), I get a stall at:
> NET: Registered protocol family 2
> 
It looks like the problem has to do with entering the powersave
routine with irqs disabled. Here is a patch that will only enter
powersave if irqs are enabled:

Entering powersave on PPC while irqs are disabled causes a hang. Only
enter powersave if irqs are disabled.

Signed-off-by: Brian Waite <waite@skycomputers.com>
===== arch/ppc/kernel/idle.c 1.22 vs edited =====
--- 1.22/arch/ppc/kernel/idle.c Tue Jan 11 19:42:36 2005
+++ edited/arch/ppc/kernel/idle.c       Thu Jan 13 17:22:25 2005
@@ -39,8 +39,9 @@
        powersave = ppc_md.power_save;

        if (!need_resched()) {
-               if (powersave != NULL)
-                       powersave();
+               if ((powersave != NULL) && !irqs_disabled())
+                           powersave();
+
                else {
 #ifdef CONFIG_SMP
                        set_thread_flag(TIF_POLLING_NRFLAG);