aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorEric Dumazet <eric.dumazet@gmail.com>2010-06-09 12:21:07 -0400
committerDavid S. Miller <davem@davemloft.net>2010-06-11 01:47:08 -0400
commit592fcb9dfafaa02dd0edc207bf5d3a0ee7a1f8df (patch)
tree3b7bbd5313c7f564cf8087658c36d32a87ca79c4
parent96b52e61be1ad4d4f8de39b9deaf253da804ea3b (diff)
ip: ip_ra_control() rcu fix
commit 66018506e15b (ip: Router Alert RCU conversion) introduced RCU lookups to ip_call_ra_chain(). It missed proper deinit phase : When ip_ra_control() deletes an ip_ra_chain, it should make sure ip_call_ra_chain() users can not start to use socket during the rcu grace period. It should also delay the sock_put() after the grace period, or we risk a premature socket freeing and corruptions, as raw sockets are not rcu protected yet. This delay avoids using expensive atomic_inc_not_zero() in ip_call_ra_chain(). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
-rw-r--r--include/net/ip.h5
-rw-r--r--net/ipv4/ip_sockglue.c19
2 files changed, 19 insertions, 5 deletions
diff --git a/include/net/ip.h b/include/net/ip.h
index 9982c97f0bdc..d52f01180361 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -61,7 +61,10 @@ struct ipcm_cookie {
61struct ip_ra_chain { 61struct ip_ra_chain {
62 struct ip_ra_chain *next; 62 struct ip_ra_chain *next;
63 struct sock *sk; 63 struct sock *sk;
64 void (*destructor)(struct sock *); 64 union {
65 void (*destructor)(struct sock *);
66 struct sock *saved_sk;
67 };
65 struct rcu_head rcu; 68 struct rcu_head rcu;
66}; 69};
67 70
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 08b9519a24f4..47fff528ff39 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -241,9 +241,13 @@ int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
241struct ip_ra_chain *ip_ra_chain; 241struct ip_ra_chain *ip_ra_chain;
242static DEFINE_SPINLOCK(ip_ra_lock); 242static DEFINE_SPINLOCK(ip_ra_lock);
243 243
244static void ip_ra_free_rcu(struct rcu_head *head) 244
245static void ip_ra_destroy_rcu(struct rcu_head *head)
245{ 246{
246 kfree(container_of(head, struct ip_ra_chain, rcu)); 247 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
248
249 sock_put(ra->saved_sk);
250 kfree(ra);
247} 251}
248 252
249int ip_ra_control(struct sock *sk, unsigned char on, 253int ip_ra_control(struct sock *sk, unsigned char on,
@@ -264,13 +268,20 @@ int ip_ra_control(struct sock *sk, unsigned char on,
264 kfree(new_ra); 268 kfree(new_ra);
265 return -EADDRINUSE; 269 return -EADDRINUSE;
266 } 270 }
271 /* dont let ip_call_ra_chain() use sk again */
272 ra->sk = NULL;
267 rcu_assign_pointer(*rap, ra->next); 273 rcu_assign_pointer(*rap, ra->next);
268 spin_unlock_bh(&ip_ra_lock); 274 spin_unlock_bh(&ip_ra_lock);
269 275
270 if (ra->destructor) 276 if (ra->destructor)
271 ra->destructor(sk); 277 ra->destructor(sk);
272 sock_put(sk); 278 /*
273 call_rcu(&ra->rcu, ip_ra_free_rcu); 279 * Delay sock_put(sk) and kfree(ra) after one rcu grace
280 * period. This guarantee ip_call_ra_chain() dont need
281 * to mess with socket refcounts.
282 */
283 ra->saved_sk = sk;
284 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
274 return 0; 285 return 0;
275 } 286 }
276 } 287 }