If a user calls the *alloc_node() functions with an invalid node
its possible to crash in alloc_pages_nodemask because NODE_DATA()
returns a bad node, which propogates into the node zonelist in
prepare_alloc_pages. This avoids that by not trying to allocate
new slabs against offline nodes.

(example backtrace)

  __alloc_pages_nodemask+0x128/0xf48
  allocate_slab+0x94/0x528
  new_slab+0x68/0xc8
  ___slab_alloc+0x44c/0x520
  __slab_alloc+0x50/0x68
  kmem_cache_alloc_node_trace+0xe0/0x230

Signed-off-by: Jeremy Linton <jeremy.lin...@arm.com>
---
 mm/slub.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/mm/slub.c b/mm/slub.c
index 51258eff4178..e03719bac1e2 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -2519,6 +2519,8 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t 
gfpflags, int node,
                if (unlikely(!node_match(page, searchnode))) {
                        stat(s, ALLOC_NODE_MISMATCH);
                        deactivate_slab(s, page, c->freelist, c);
+                       if (!node_online(searchnode))
+                               node = NUMA_NO_NODE;
                        goto new_slab;
                }
        }
-- 
2.14.3

Reply via email to