Skip to contentSkip to conversation

ipq+AF8-set+AF8-verdict problem in bridge+-iptables

DISCUSSION Yong 2 participants 2 messages Jul 16, 2003 netfilter

lore
Share & tools
lore

Yong

author
#0 lore raw

Hello Harald,

Thank you for your emails!

I want to use the iptable_queue in bridge+-iptables environment. I can get the packet in userspace using -j QUEUE command. However, If I change the packet size. for example, I change the ping icmp packet size to 400, the packet Ethernet header is changed. In my test, the MAC address is changed to 0xffffffffff. I can capture this packet using sniffer tool. Since the MAC address is changed, the other computer cannot receive the ICMP packet.

It seems that the bridge iptables patch changed something in the function ipq_set_verdict().

Regarding the MTU issue, you are right. the ipq_set_verdict function does not perform the IP fragment. If I send a packet larger than MTU, it is missing. Is it by design?

Can I modify the ipq_set_verdict function to perform the IP fragment? Is there any patch for this IP fragment issue?

Thank you again for your help!

Yong

Quoted 24 lines

----- Original Message -----
From: "Harald Welte" <[email protected]>
To: "Yong Li" <[email protected]>
Cc: <[email protected]>
Sent: Monday, July 14, 2003 4:07 PM
Subject: Re: ipq_set_verdict problem in bridge+-iptables

Quoted 6 lines · Yong Li

Hello All,

I encountered a problem with the ipq_set_verdict function. I want to modify the packet content and size using the ipq_set_verdict function. However, I found I cannot change the IP packet size more than 400+- bytes. Is it a known issue?

what do you mean by 'I cannot' ? What happens? Is an error returned to the ipq_set_verdict() call? Is the packet silently discarded? Is the packet truncated?

Anyway, it should work. But if you exceed the outgoing interface's MTU, I could imagine that no fragmentation happens...

Quoted 2 lines

Thanks in advance!
Yong

#1 lore raw
Quoted 9 lines · Yong

I want to use the iptable_queue in bridge+-iptables environment. I can get
the packet in userspace using -j QUEUE command. However, If I change the packet size. for example, I change the ping icmp packet size to 400, the packet Ethernet header is changed. In my test, the MAC address is changed to 0xffffffffff. I can capture this packet using sniffer tool. Since the MAC address is changed, the other computer cannot receive the ICMP packet.

It seems that the bridge iptables patch changed something in the function
ipq_set_verdict().

Please test the patch below.

cheers,
Bart

--- linux-2.6.0-test1/net/ipv4/netfilter/ip_queue.c.old 2003-07-16 19:35:07.000000000 +-0200
+-+-+- linux-2.6.0-test1/net/ipv4/netfilter/ip_queue.c 2003-07-16 19:39:05.000000000 +-0200
@@ -353,6 +-353,11 @@ ipq_mangle_ipv4(ipq_verdict_msg_t *v, st
}
if (e->skb->sk)
skb_set_owner_w(newskb, e->skb->sk);

+-#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+-			/* bridged packets already have their Ethernet header */
+-			if (e->skb->nf_bridge)
+-				memcpy(newskb->data - 16, e->skb->data - 16, 16);
+-#endif

kfree_skb(e->skb);
e->skb = newskb;
}