powerpc: Re-enable preemption before cpu_die()

start_secondary() is called shortly after _start and also via

cpu_idle()->cpu_die()->pseries_mach_cpu_die()

start_secondary() expects a preempt_count() of 0. pseries_mach_cpu_die() is
called via the cpu_idle() routine with preemption disabled, resulting in the
following repeating message during rapid cpu offline/online tests
with CONFIG_PREEMPT=y:

BUG: scheduling while atomic: swapper/0/0x00000002
Modules linked in: autofs4 binfmt_misc dm_mirror dm_region_hash dm_log [last unloaded: scsi_wait_scan]
Call Trace:
[c00000010e7079c0] [c0000000000133ec] .show_stack+0xd8/0x218 (unreliable)
[c00000010e707aa0] [c0000000006a47f0] .dump_stack+0x28/0x3c
[c00000010e707b20] [c00000000006e7a4] .__schedule_bug+0x7c/0x9c
[c00000010e707bb0] [c000000000699d9c] .schedule+0x104/0x800
[c00000010e707cd0] [c000000000015b24] .cpu_idle+0x1c4/0x1d8
[c00000010e707d70] [c0000000006aa1b4] .start_secondary+0x398/0x3d4
[c00000010e707e30] [c000000000008278] .start_secondary_resume+0x10/0x14

Move the cpu_die() call inside the existing preemption enabled block of
cpu_idle(). This is safe as the idle task is affined to a single CPU so the
debug_smp_processor_id() tests (from cpu_should_die()) won't trigger as we are
in a "migration disabled" region.

Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
Acked-by: Will Schmidt <will_schmidt@vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Nathan Fontenot <nfont@austin.ibm.com>
Cc: Robert Jennings <rcj@linux.vnet.ibm.com>
Cc: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

arch/powerpc/kernel/idle.c

@@ -94,9 +94,9 @@ void cpu_idle(void)
 		HMT_medium();
 		ppc64_runlatch_on();
 		tick_nohz_restart_sched_tick();
+		preempt_enable_no_resched();
 		if (cpu_should_die())
 			cpu_die();
-		preempt_enable_no_resched();
 		schedule();
 		preempt_disable();
 	}