Skip to contentSkip to conversation

Net: Don't send ICMP_FRAG_NEEDED for GSO packets

PATCH Applied Rusty Russell 3 participants 3 messages Mar 28, 2008

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Net: Don't send ICMP_FRAG_NEEDED for GSO packets

Commit 9af3912ec9e30509b76cb376abb65a4d8af27df3 added a new check to send ICMP fragmentation needed for large packets.

Unlike the check in ip_finish_output(), it doesn't check for GSO. I triggered this with the GSO TUN patches, so not sure if it needs to be in 2.6.25 or not (do we otherwise forward GSO packets?).

net/ipv4/ip_forward.c +1 -1
diff -r 69b8317ff51c net/ipv4/ip_forward.c
@@ -85,7 +85,7 @@ int ip_forward(struct sk_buff *skb)
 	if (opt->is_strictroute && rt->rt_dst != rt->rt_gateway)
 		goto sr_failed;
 
-	if (unlikely(skb->len > dst_mtu(&rt->u.dst) &&
+	if (unlikely(skb->len > dst_mtu(&rt->u.dst) && !skb_is_gso(skb) &&
 		     (ip_hdr(skb)->frag_off & htons(IP_DF))) && !skb->local_df) {
 		IP_INC_STATS(IPSTATS_MIB_FRAGFAILS);
 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
#1 lore raw
Quoted 9 lines · Rusty Russell

Commit 9af3912ec9e30509b76cb376abb65a4d8af27df3 added a new check to send ICMP fragmentation needed for large packets.

Unlike the check in ip_finish_output(), it doesn't check for GSO. I triggered this with the GSO TUN patches, so not sure if it needs to be in 2.6.25 or not (do we otherwise forward GSO packets?).

Since this is a regression for Xen, I would say yes.

Thanks,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
maintainer Applied
#2 lore raw

From: Herbert Xu <[email protected]>
Date: Fri, 28 Mar 2008 16:40:43 +0800

Quoted 12 lines · Rusty Russell
Quoted 9 lines

Commit 9af3912ec9e30509b76cb376abb65a4d8af27df3 added a new check to send ICMP fragmentation needed for large packets.

Unlike the check in ip_finish_output(), it doesn't check for GSO. I triggered this with the GSO TUN patches, so not sure if it needs to be in 2.6.25 or not (do we otherwise forward GSO packets?).

Since this is a regression for Xen, I would say yes.

Applied.

But Rusty, when referencing commits, always provide a text string of the commit's header line because in different trees and backports the commit ID may change.

I fixed it up this time...

Thanks.