ipv4: Fix oops with FIB_TRIE

It seems we can fix this by disabling preemption while we re-balance the
trie. This is with the CONFIG_CLASSIC_RCU. It's been stress-tested at high
loads continuesly taking a full BGP table up/down via iproute -batch.

Note. fib_trie is not updated for CONFIG_PREEMPT_RCU

Reported-by: Andrei Popa
Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index ec0ae49..33c7c85 100644 (file)
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -986,9 +986,12 @@ fib_find_node(struct trie *t, u32 key)
 static struct node *trie_rebalance(struct trie *t, struct tnode *tn)
 {
        int wasfull;
-       t_key cindex, key = tn->key;
+       t_key cindex, key;
        struct tnode *tp;
 
+       preempt_disable();
+       key = tn->key;
+
        while (tn != NULL && (tp = node_parent((struct node *)tn)) != NULL) {
                cindex = tkey_extract_bits(key, tp->pos, tp->bits);
                wasfull = tnode_full(tp, tnode_get_child(tp, cindex));
@@ -1007,6 +1010,7 @@ static struct node *trie_rebalance(struct trie *t, struct tnode *tn)
        if (IS_TNODE(tn))
                tn = (struct tnode *)resize(t, (struct tnode *)tn);
 
+       preempt_enable();
        return (struct node *)tn;
 }