[PATCH] powerpc/mm: Fix a AB->BA deadlock scenario with nohash MMU context lock
Benjamin Herrenschmidt
benh at kernel.crashing.org
Wed Jun 3 14:53:37 EST 2009
The MMU context_lock can be taken from switch_mm() while the
rq->lock is held. The rq->lock can also be taken from interrupts,
thus if we get interrupted in destroy_context() with the context
lock held and that interrupt tries to take the rq->lock, there's
a possible deadlock scenario with another CPU having the rq->lock
and calling switch_mm() which takes our context lock.
The fix is to always ensure interrupts are off when taking our
context lock. The switch_mm() path is already good so this fixes
the destroy_context() path.
While at it, turn the context lock into a new style spinlock.
Signed-off-by: Benjamin Herrenschmidt <benh at kernel.crashing.org>
---
arch/powerpc/mm/mmu_context_nohash.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
--- linux-work.orig/arch/powerpc/mm/mmu_context_nohash.c 2009-06-03 14:14:47.000000000 +1000
+++ linux-work/arch/powerpc/mm/mmu_context_nohash.c 2009-06-03 14:25:30.000000000 +1000
@@ -46,7 +46,7 @@ static unsigned int next_context, nr_fre
static unsigned long *context_map;
static unsigned long *stale_map[NR_CPUS];
static struct mm_struct **context_mm;
-static spinlock_t context_lock = SPIN_LOCK_UNLOCKED;
+static DEFINE_SPINLOCK(context_lock);
#define CTX_MAP_SIZE \
(sizeof(unsigned long) * (last_context / BITS_PER_LONG + 1))
@@ -272,6 +272,7 @@ int init_new_context(struct task_struct
*/
void destroy_context(struct mm_struct *mm)
{
+ unsigned long flags;
unsigned int id;
if (mm->context.id == MMU_NO_CONTEXT)
@@ -279,7 +280,7 @@ void destroy_context(struct mm_struct *m
WARN_ON(mm->context.active != 0);
- spin_lock(&context_lock);
+ spin_lock_irqsave(&context_lock, flags);
id = mm->context.id;
if (id != MMU_NO_CONTEXT) {
__clear_bit(id, context_map);
@@ -290,7 +291,7 @@ void destroy_context(struct mm_struct *m
#endif
nr_free_contexts++;
}
- spin_unlock(&context_lock);
+ spin_unlock_irqrestore(&context_lock, flags);
}
#ifdef CONFIG_SMP
More information about the Linuxppc-dev
mailing list