Skip to contentSkip to conversation

IPVS: ICMPv6 checksum calculation

PATCH Superseded Simon Horman 4 participants 9 messages Aug 17, 2010 – Aug 23, 2010

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Simon Horman

author maintainer
#0 lore raw

PATCH 1/1 IPVS: ICMPv6 checksum calculation

From: Xiaoyu Du <[email protected]>

Correct the ICMPv6 checksum calculation when IPv6 is is used

---
 net/netfilter/ipvs/ip_vs_core.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

I wonder if this is -stable or .35 material.

net/netfilter/ipvs/ip_vs_core.c +0 -0
index 4f8ddba..124a1ae 100644
@@ -638,8 +638,12 @@ void ip_vs_nat_icmp_v6(struct sk_buff *skb,

struct ip_vs_protocol *pp,

/* And finally the ICMP checksum */ icmph->icmp6_cksum = 0;
- /* TODO IPv6: is this correct for ICMPv6? */
ip_vs_checksum_complete(skb, icmp_offset);
+ icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
+ &iph->daddr,
+ skb->len - icmp_offset, IPPROTO_ICMPV6,
+ skb->csum);
+
skb->ip_summed = CHECKSUM_UNNECESSARY;

if (inout)
--

maintainer
#1 lore raw

Hello,

Quoted 18 lines · Simon Horman

From: Xiaoyu Du <[email protected]>

Correct the ICMPv6 checksum calculation when IPv6 is is used

---
 net/netfilter/ipvs/ip_vs_core.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

I wonder if this is -stable or .35 material.

net/netfilter/ipvs/ip_vs_core.c +0 -0
index 4f8ddba..124a1ae 100644

net/netfilter/ipvs/ip_vs_core.c void ip_vs_nat_icmp_v6(struct sk_buff *skb, · line 642

Quoted 5 lines
net/netfilter/ipvs/ip_vs_core.c +0 -1
@@ -638,8 +638,12 @@ void ip_vs_nat_icmp_v6(struct sk_buff *skb,
 struct ip_vs_protocol *pp,
 
 	/* And finally the ICMP checksum */
 	icmph->icmp6_cksum = 0;
-	/* TODO IPv6: is this correct for ICMPv6? */

May be ip_vs_checksum_complete() is not needed here?
Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

	icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
		&iph->daddr,
		skb->len - icmp_offset, IPPROTO_ICMPV6,
		0);
	skb->csum_start = skb_network_header(skb) - skb->head +
			  icmp_offset;
	skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum);
	skb->ip_summed = CHECKSUM_PARTIAL;
Quoted 9 lines
net/netfilter/ipvs/ip_vs_core.c +5 -0
 	ip_vs_checksum_complete(skb, icmp_offset);
+	icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
+			&iph->daddr,
+			skb->len - icmp_offset, IPPROTO_ICMPV6,
+			skb->csum);
+
 	skb->ip_summed = CHECKSUM_UNNECESSARY;
 

if (inout)

Regards

--
Julian Anastasov <[email protected]>

author maintainer
#2 lore raw

[ CCing netdev for comment ]

Quoted 59 lines · Julian Anastasov

Hello,

Quoted 18 lines · Simon Horman

From: Xiaoyu Du <[email protected]>

Correct the ICMPv6 checksum calculation when IPv6 is is used

---
 net/netfilter/ipvs/ip_vs_core.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

I wonder if this is -stable or .35 material.

net/netfilter/ipvs/ip_vs_core.c +0 -0
index 4f8ddba..124a1ae 100644

net/netfilter/ipvs/ip_vs_core.c void ip_vs_nat_icmp_v6(struct sk_buff *skb, · line 642

Quoted 5 lines
net/netfilter/ipvs/ip_vs_core.c +0 -1
@@ -638,8 +638,12 @@ void ip_vs_nat_icmp_v6(struct sk_buff *skb,
 struct ip_vs_protocol *pp,
 
 	/* And finally the ICMP checksum */
 	icmph->icmp6_cksum = 0;
-	/* TODO IPv6: is this correct for ICMPv6? */

May be ip_vs_checksum_complete() is not needed here?
Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

	icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
		&iph->daddr,
		skb->len - icmp_offset, IPPROTO_ICMPV6,
		0);
	skb->csum_start = skb_network_header(skb) - skb->head +
			  icmp_offset;
	skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum);
	skb->ip_summed = CHECKSUM_PARTIAL;
Quoted 9 lines
net/netfilter/ipvs/ip_vs_core.c +5 -0
 	ip_vs_checksum_complete(skb, icmp_offset);
+	icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
+			&iph->daddr,
+			skb->len - icmp_offset, IPPROTO_ICMPV6,
+			skb->csum);
+
 	skb->ip_summed = CHECKSUM_UNNECESSARY;
 

if (inout)

Regards

--
Julian Anastasov <[email protected]>

maintainer
#3 lore raw

From: Simon Horman <[email protected]>
Date: Wed, 18 Aug 2010 09:21:49 +0900

Quoted 3 lines

[ CCing netdev for comment ]

On Tue, Aug 17, 2010 at 12:25:56PM +0300, Julian Anastasov wrote:

...

Quoted 7 lines
Quoted 7 lines

May be ip_vs_checksum_complete() is not needed here?
Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

Quoted 9 lines
Quoted 9 lines
	icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr,
		&iph->daddr,
		skb->len - icmp_offset, IPPROTO_ICMPV6,
		0);
	skb->csum_start = skb_network_header(skb) - skb->head +
			  icmp_offset;
	skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum);
	skb->ip_summed = CHECKSUM_PARTIAL;

Yes, using CHECKSUM_PARTIAL unconditionally would work.

#4 lore raw

Thanks, I learned this. You gave me a big cake.

2010/8/18 David Miller <[email protected]>:

Quoted 26 lines

From: Simon Horman <[email protected]>
Date: Wed, 18 Aug 2010 09:21:49 +0900

Quoted 3 lines

[ CCing netdev for comment ]

On Tue, Aug 17, 2010 at 12:25:56PM +0300, Julian Anastasov wrote:

 ...

Quoted 7 lines
Quoted 7 lines

     May be ip_vs_checksum_complete() is not needed here? Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

Quoted 9 lines
Quoted 9 lines

     icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr, &iph->daddr,
             skb->len - icmp_offset, IPPROTO_ICMPV6, 0);
     skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset;
     skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); skb->ip_summed = CHECKSUM_PARTIAL;

Yes, using CHECKSUM_PARTIAL unconditionally would work.

author maintainer
#5 lore raw
Quoted 25 lines · xiaoyu Du

2010/8/18 David Miller <[email protected]>:

Quoted 24 lines

From: Simon Horman <[email protected]>
Date: Wed, 18 Aug 2010 09:21:49 +0900

Quoted 3 lines

[ CCing netdev for comment ]

On Tue, Aug 17, 2010 at 12:25:56PM +0300, Julian Anastasov wrote:

 ...

Quoted 16 lines
Quoted 16 lines

     May be ip_vs_checksum_complete() is not needed here? Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

     icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr, &iph->daddr,
             skb->len - icmp_offset, IPPROTO_ICMPV6, 0);
     skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset;
     skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); skb->ip_summed = CHECKSUM_PARTIAL;

Quoted 3 lines
Quoted 2 lines

Yes, using CHECKSUM_PARTIAL unconditionally would work.

Thanks, I learned this. You gave me a big cake.

[ please don't top-post ]

Xiaoyu, are you in a position to test the code proposed by Julian?

#6 lore raw

I just replied to thank for his idea then, as he gived me some ideas on how to manipulate
checksum. I tested the code today, but it doesn't seem to work. (I just wrote a netfilter module and hooked the code on local_out , then ping6 the address on another PC,
It failed.). I'll find out why.

2010/8/18 Simon Horman <[email protected]>:

Quoted 26 lines · xiaoyu Du
Quoted 25 lines

2010/8/18 David Miller <[email protected]>:

Quoted 24 lines

From: Simon Horman <[email protected]>
Date: Wed, 18 Aug 2010 09:21:49 +0900

Quoted 3 lines

[ CCing netdev for comment ]

On Tue, Aug 17, 2010 at 12:25:56PM +0300, Julian Anastasov wrote:

 ...

Quoted 16 lines
Quoted 16 lines

     May be ip_vs_checksum_complete() is not needed here? Also, skb->csum is not valid for every ip_summed value. May be we need to apply CHECKSUM_PARTIAL kind of setup for the IPv6. Such example is net/ipv6/udp.c:udp6_ufo_send_check(). Later dev_queue_xmit() and skb_checksum_help() should take care for the next steps. Something like this can be tested:

     icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr, &iph->daddr,
             skb->len - icmp_offset, IPPROTO_ICMPV6, 0);
     skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset;
     skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); skb->ip_summed = CHECKSUM_PARTIAL;

Quoted 8 lines
Quoted 3 lines
Quoted 2 lines

Yes, using CHECKSUM_PARTIAL unconditionally would work.

Thanks, I learned this. You gave me a big cake.

[ please don't top-post ]

Xiaoyu, are you in a position to test the code proposed by Julian?

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#7 lore raw

Hello,

Quoted 6 lines · xiaoyu Du

I just replied to thank for his idea then, as he gived me some ideas on how to manipulate
checksum. I tested the code today, but it doesn't seem to work. (I just wrote a netfilter module and hooked the code on local_out , then ping6 the address on another PC,
It failed.). I'll find out why.

May be because I copied csum_ipv6_magic() code from
your patch. Please, try with one '~' before csum_ipv6_magic.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

     icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr, &iph->daddr,
             skb->len - icmp_offset, IPPROTO_ICMPV6, 0);
     skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset;
     skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); skb->ip_summed = CHECKSUM_PARTIAL;

Regards

--
Julian Anastasov <[email protected]>
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#8 lore raw

Yes, I had added it and test it, it works.

2010/8/20 Julian Anastasov <[email protected]>:

Quoted 29 lines

       Hello,

Quoted 6 lines · xiaoyu Du

I just replied to thank for his idea then, as he gived me some ideas on how to manipulate
checksum. I tested the code today, but it doesn't seem to work. (I just wrote a netfilter module and hooked the code on local_out , then ping6 the address on another PC,
It failed.). I'll find out why.

       May be because I copied csum_ipv6_magic() code from your patch. Please, try with one '~' before csum_ipv6_magic.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

     icmph->icmp6_cksum = csum_ipv6_magic(&iph->saddr, &iph->daddr,
             skb->len - icmp_offset, IPPROTO_ICMPV6, 0);
     skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset;
     skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); skb->ip_summed = CHECKSUM_PARTIAL;

Regards

--
Julian Anastasov <[email protected]>