powerpc/nohash: Allocate stale_map[cpu] on CPU_UP_PREPARE not CPU_ONLINE
authorMichael Ellerman <michael@ellerman.id.au>
Mon, 4 Apr 2011 20:57:27 +0000 (20:57 +0000)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Wed, 20 Apr 2011 07:01:18 +0000 (17:01 +1000)
Currently we allocate the stale_map for a cpu when it comes online,
this leaves open a small window where a process can be scheduled
on the cpu before the stale_map is allocated. Instead allocate
the stale_map at CPU_UP_PREPARE time, that way it will be always
available before tasks start running.

It is possible the cpu fails to come up, in which case we should free
the stale_map, so add a CPU_UP_CANCELED case to do that.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/mm/mmu_context_nohash.c

index c0aab52da3a56b73f4758ab5c874c9bb1374c68a..4d8fa911c73dde5e7853b3e0fa85298569177dc5 100644 (file)
@@ -338,12 +338,14 @@ static int __cpuinit mmu_context_cpu_notify(struct notifier_block *self,
                return NOTIFY_OK;
 
        switch (action) {
-       case CPU_ONLINE:
-       case CPU_ONLINE_FROZEN:
+       case CPU_UP_PREPARE:
+       case CPU_UP_PREPARE_FROZEN:
                pr_devel("MMU: Allocating stale context map for CPU %d\n", cpu);
                stale_map[cpu] = kzalloc(CTX_MAP_SIZE, GFP_KERNEL);
                break;
 #ifdef CONFIG_HOTPLUG_CPU
+       case CPU_UP_CANCELED:
+       case CPU_UP_CANCELED_FROZEN:
        case CPU_DEAD:
        case CPU_DEAD_FROZEN:
                pr_devel("MMU: Freeing stale context map for CPU %d\n", cpu);