Copy mac_len in skb_clone() as well

ANK says: "It is rarely used, that's wy it was not noticed. But in the places, where it is used, it should be disaster."

---

 net/core/skbuff.c |    1 +
 1 file changed, 1 insertion(+)

@@ -463,6 +463,7 @@ #endif
 	memcpy(n->cb, skb->cb, sizeof(skb->cb));
 	C(len);
 	C(data_len);
+	C(mac_len);
 	C(csum);
 	C(local_df);
 	n->cloned = 1;

From: Alexey Dobriyan <[email protected]>
Date: Wed, 14 Mar 2007 16:07:11 +0300

Applied.

What bug triggered that helped you discover this? Or is it merely from a code audit?

Thanks.

Ohhh, it is a fairy-tale to tell the truth :)
We had some unexplainable problems with java application in OpenVZ kernel. It didn't work sometimes, but worked fine (!) with CONFIG_SLAB_DEBUG. Alexey blamed java :), but ...
Then we found that poising one of the bits in slab cache was curing it. After that we found that the problem is related to fclone cache. And then we found that not all the fields are initialized during cloning. The bug was related to our own skb->field we introduced,
but we analyzed the code and found this as well.

Thanks,
Kirill

Hello!

I asked the same question. :-)

openvz added some another fields to skbuff and when it was found that they are lost while clone, he tried to figure out how all this works and looked for another examples of this kind.

As I understand, the problem can be seen only in xfrmX_tunnel_input. If uninitialized mac_len obtained from slab is more than current head room it could corrupt memory.

Also, it looks like the fix is incomplete. copy_skb_header() also does not copy this field. But it will be initialized to 0 by alloc_skb in this case and xfrmX_tunnel_input() just will not copy mac header.

Alexey

From: Kirill Korotaev <[email protected]>
Date: Thu, 15 Mar 2007 13:33:12 +0300

Thanks for the detailed information.