On Mon, Dec 01, 2014 at 09:14:40AM +0900, Yasuaki Ishimatsu wrote:

(2014/12/01 7:16), Paul Mackerras wrote:

quoted

The bounds check for nodeid in ____cache_alloc_node gives false
positives on machines where the node IDs are not contiguous, leading
to a panic at boot time.  For example, on a POWER8 machine the node
IDs are typically 0, 1, 16 and 17.  This means that num_online_nodes()
returns 4, so when ____cache_alloc_node is called with nodeid = 16 the
VM_BUG_ON triggers.

Do you have the call trace? If you have it, please add it in the description.

I can get it easily enough.

quoted

To fix this, we instead compare the nodeid with MAX_NUMNODES, and
additionally make sure it isn't negative (since nodeid is an int).
The check is there mainly to protect the array dereference in the
get_node() call in the next line, and the array being dereferenced is
of size MAX_NUMNODES.  If the nodeid is in range but invalid, the
BUG_ON in the next line will catch that.

Signed-off-by: Paul Mackerras <redacted>

Do you need to backport it into -stable kernels?

It does need to go to stable, yes, for 3.10 and later.

quoted

---

diff --git a/mm/slab.c b/mm/slab.c
index eb2b2ea..f34e053 100644
--- a/mm/slab.c
+++ b/mm/slab.c

@@ -3076,7 +3076,7 @@ static void *____cache_alloc_node(struct kmem_cache *cachep, gfp_t flags,
 	void *obj;
 	int x;

quoted

-	VM_BUG_ON(nodeid > num_online_nodes());
+	VM_BUG_ON(nodeid < 0 || nodeid >= MAX_NUMNODES);

How about use:
	VM_BUG_ON(!node_online(nodeid));

That would not be better, since node_online() doesn't bounds-check its
argument.

When allocating the memory, the node of the memory being allocated must be
online. But your code cannot check the condition.

The following two lines:

quoted

 	n = get_node(cachep, nodeid);
 	BUG_ON(!n);

effectively check that condition already, as I tried to explain in the
commit message.

Regards,
Paul.