[PATCH] IPV4 IPComp : threshold comparison

[PATCH] IPV4 IPComp : threshold comparison

[Mitsuru KANDA / 神田 充]

Hello,

I'm not sure how exactly we should check threshold.
In IPsec tunnel mode, the original comparison code 
seems to be including IP header length.

Could you check it?

Regards,
-mk


===== ipcomp.c 1.6 vs edited =====
--- 1.6/net/ipv4/ipcomp.c	Tue May 13 05:58:03 2003
+++ edited/ipcomp.c	Sun May 18 01:42:22 2003
@@ -168,6 +168,7 @@
 		struct iphdr	iph;
 		char 		buf[60];
 	} tmp_iph;
+	int hdr_len = 0;
 
 	if (skb->ip_summed == CHECKSUM_HW && skb_checksum_help(skb) == NULL) {
 		err = -EINVAL;
@@ -180,7 +181,11 @@
 		goto error;
 
 	/* Don't bother compressing */
-	if (skb->len < ipcd->threshold) {
+	if (!x->props.mode) {
+		iph = skb->nh.iph;
+		hdr_len = iph->ihl * 4;
+	}
+	if ((skb->len - hdr_len) < ipcd->threshold) {
 		if (x->props.mode) {
 			ipcomp_tunnel_encap(x, skb);
 			iph = skb->nh.iph;

Re: [PATCH] IPV4 IPComp : threshold comparison

[James Morris]

On Sun, 18 May 2003, Mitsuru KANDA / [ISO-2022-JP] ^[$B?@ED^[(B ^[$B=<^[(B wrote:

> I'm not sure how exactly we should check threshold.

We should check the length of the IP payload, which the current code is 
only currently doing for tunnel mode.

> In IPsec tunnel mode, the original comparison code 
> seems to be including IP header length.

It is counting the inner IP header, which is correct, as it is part of the 
final IP payload.  The transport mode calculation is incorrect.

> Could you check it?

Your patch fixes the calculation for transport mode, and looks correct.


- James
-- 
James Morris
<jmorris@intercode.com.au>

Re: [PATCH] IPV4 IPComp : threshold comparison

[David S. Miller]

   From: Mitsuru KANDA / 神田 充 <mk@linux-ipv6.org>
   Date: Sun, 18 May 2003 02:00:01 +0900

   ===== ipcomp.c 1.6 vs edited =====
   --- 1.6/net/ipv4/ipcomp.c	Tue May 13 05:58:03 2003
   +++ edited/ipcomp.c	Sun May 18 01:42:22 2003

I am applying this, but please you USAGI guys start to
generate properly rooted patched for me ok?

I want something of the form:

   --- a/net/ipv4/ipcomp.c	Tue May 13 05:58:03 2003
   +++ b/net/ipv4/ipcomp.c	Sun May 18 01:42:22 2003

Almost everyone sends me patches like this, USAGI are
the one exception.  If everyone sends me patches this
way, then all of my patch applying scripts know they can
simply use "patch -p1" when sitting at top of kernel tree.

Probably all of the USAGI patches look this way because of
CVS or something like that.

Thank you.