Thread (26 messages) flat view 26 messages, 1 author, 2016-07-23
STALE3667d

Revision v1 of 43 in this series.

Revisions (43)
  1. v1 [diff vs current]
  2. v1 [diff vs current]
  3. v1 [diff vs current]
  4. v1 [diff vs current]
  5. v1 [diff vs current]
  6. v1 [diff vs current]
  7. v1 [diff vs current]
  8. v1 [diff vs current]
  9. v1 [diff vs current]
  10. v1 [diff vs current]
  11. v1 [diff vs current]
  12. v1 current
  13. v1 [diff vs current]
  14. v1 [diff vs current]
  15. v1 [diff vs current]
  16. v1 [diff vs current]
  17. v1 [diff vs current]
  18. v1 [diff vs current]
  19. v1 [diff vs current]
  20. v1 [diff vs current]
  21. v1 [diff vs current]
  22. v1 [diff vs current]
  23. v1 [diff vs current]
  24. v1 [diff vs current]
  25. v1 [diff vs current]
  26. v1 [diff vs current]
  27. v1 [diff vs current]
  28. v1 [diff vs current]
  29. v1 [diff vs current]
  30. v1 [diff vs current]
  31. v1 [diff vs current]
  32. v1 [diff vs current]
  33. v1 [diff vs current]
  34. v1 [diff vs current]
  35. v1 [diff vs current]
  36. v1 [diff vs current]
  37. v1 [diff vs current]
  38. v1 [diff vs current]
  39. v1 [diff vs current]
  40. v1 [diff vs current]
  41. v1 [diff vs current]
  42. v1 [diff vs current]
  43. v1 [diff vs current]

[PATCH 03/25] netfilter: cttimeout: unlink timeout obj again when hash resize happen

From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: 2016-07-23 11:09:13
Also in: netfilter-devel
Subsystem: netfilter, networking [general], the rest · Maintainers: Pablo Neira Ayuso, Florian Westphal, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds

From: Liping Zhang <redacted>

Imagine such situation, nf_conntrack_htable_size now is 4096, we are doing
ctnl_untimeout, and iterate on 3000# bucket.

Meanwhile, another user try to reduce hash size to 2048, then all nf_conn
are removed to the new hashtable. When this hash resize operation finished,
we still try to itreate ct begin from 3000# bucket, find nothing to do and
just return.

We may miss unlinking some timeout objects. And later we will end up with
invalid references to timeout object that are already gone.

So when we find that hash resize happened, try to unlink timeout objects
from the 0# bucket again.

Signed-off-by: Liping Zhang <redacted>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 net/netfilter/nfnetlink_cttimeout.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)
diff --git a/net/netfilter/nfnetlink_cttimeout.c b/net/netfilter/nfnetlink_cttimeout.c
index 3c84f14..4cdcd96 100644
--- a/net/netfilter/nfnetlink_cttimeout.c
+++ b/net/netfilter/nfnetlink_cttimeout.c
@@ -303,16 +303,24 @@ static void ctnl_untimeout(struct net *net, struct ctnl_timeout *timeout)
 {
 	struct nf_conntrack_tuple_hash *h;
 	const struct hlist_nulls_node *nn;
+	unsigned int last_hsize;
+	spinlock_t *lock;
 	int i;
 
 	local_bh_disable();
-	for (i = 0; i < nf_conntrack_htable_size; i++) {
-		nf_conntrack_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
-		if (i < nf_conntrack_htable_size) {
-			hlist_nulls_for_each_entry(h, nn, &nf_conntrack_hash[i], hnnode)
-				untimeout(h, timeout);
+restart:
+	last_hsize = nf_conntrack_htable_size;
+	for (i = 0; i < last_hsize; i++) {
+		lock = &nf_conntrack_locks[i % CONNTRACK_LOCKS];
+		nf_conntrack_lock(lock);
+		if (last_hsize != nf_conntrack_htable_size) {
+			spin_unlock(lock);
+			goto restart;
 		}
-		spin_unlock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
+
+		hlist_nulls_for_each_entry(h, nn, &nf_conntrack_hash[i], hnnode)
+			untimeout(h, timeout);
+		spin_unlock(lock);
 	}
 	local_bh_enable();
 }
-- 
2.1.4
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help