Skip to contentSkip to conversation

Make udp_encap_rcv use pskb_may_pull

PATCH In review Olaf Kirch 3 participants 4 messages Nov 23, 2006 – Nov 24, 2006

Subsystems (3)
lore
Share & tools
lore

Olaf Kirch

author
#0 lore raw

PATCH 1/1 Make udp_encap_rcv use pskb_may_pull

Make udp_encap_rcv use pskb_may_pull

IPsec with NAT-T breaks on some notebooks using the latest e1000 chipset, when header split is enabled. When receiving sufficiently large packets, the driver puts everything up to and including the UDP header into the header portion of the skb, and the rest goes into the paged part. udp_encap_rcv forgets to use pskb_may_pull, and fails to decapsulate it. Instead, it passes it up it to the IKE daemon.

 net/ipv4/udp.c |   19 ++++++++++++++-----
 1 files changed, 14 insertions(+), 5 deletions(-)
net/ipv4/udp.c +14 -5
Index: linux-2.6.19-rc6/net/ipv4/udp.c
===================================================================
@@ -928,24 +928,33 @@ static int udp_encap_rcv(struct sock * s
 	return 1; 
 #else
 	struct udp_sock *up = udp_sk(sk);
-  	struct udphdr *uh = skb->h.uh;
+  	struct udphdr *uh;
 	struct iphdr *iph;
 	int iphlen, len;
   
-	__u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
-	__be32 *udpdata32 = (__be32 *)udpdata;
+	__u8 *udpdata;
+	__be32 *udpdata32;
 	__u16 encap_type = up->encap_type;
 
 	/* if we're overly short, let UDP handle it */
-	if (udpdata > skb->tail)
+	len = skb->len - sizeof(struct udphdr);
+	if (len <= 0)
 		return 1;
 
 	/* if this is not encapsulated socket, then just return now */
 	if (!encap_type)
 		return 1;
 
-	len = skb->tail - udpdata;
+	/* If this is a paged skb, make sure we pull up
+	 * whatever data we need to look at. */
+	if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
+		return 1;
 
+	/* Now we can get the pointers */
+	uh = skb->h.uh;
+	udpdata = (__u8 *)uh + sizeof(struct udphdr);
+	udpdata32 = (__be32 *)udpdata;
+  
 	switch (encap_type) {
 	default:
 	case UDP_ENCAP_ESPINUDP:
maintainer
#1 lore raw

From: Olaf Kirch <[email protected]>
Date: Thu, 23 Nov 2006 01:01:44 +0100

Quoted 10 lines

Make udp_encap_rcv use pskb_may_pull

IPsec with NAT-T breaks on some notebooks using the latest e1000 chipset, when header split is enabled. When receiving sufficiently large packets, the driver puts everything up to and including the UDP header into the header portion of the skb, and the rest goes into the paged part. udp_encap_rcv forgets to use pskb_may_pull, and fails to decapsulate it. Instead, it passes it up it to the IKE daemon.

Quoted 2 lines

Excellent catch, applied, thanks Olaf.

#2 lore raw

Hi David,

Quoted 7 lines · David Miller

From: Olaf Kirch <[email protected]>
Date: Thu, 23 Nov 2006 01:01:44 +0100

Quoted 2 lines

Make udp_encap_rcv use pskb_may_pull

Excellent catch, applied, thanks Olaf.

Should this go to -stable, too? Or are these kernels not affected, yet?

Regards

Ingo Oeser

maintainer
#3 lore raw

From: Ingo Oeser <[email protected]>
Date: Fri, 24 Nov 2006 11:54:15 +0100

Quoted 7 lines

Hi David,

Quoted 4 lines · David Miller

From: Olaf Kirch <[email protected]>
Date: Thu, 23 Nov 2006 01:01:44 +0100

Quoted 1 line
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

Make udp_encap_rcv use pskb_may_pull

Excellent catch, applied, thanks Olaf.

Should this go to -stable, too? Or are these kernels not affected, yet?

I planned to push this to -stable over the weekend, but thanks for reminding me anyways.