Netfilter problem with new 2.4.22
Subsystems (2)
More new info about this problem.
When I get ipt_REJECT.c from 2.4.21 and replace ipt_REJECT.c in 2.4.22, problem has disappeared. So when I compare two files ipt_REJECT.c from different versions of kernel:
3a4
Quoted 1 line
* Added support for ICMP type-3-code-13 (Maciej Soltysiak). [RFC 1812]
35a37,76
Quoted 32 lines
static inline struct rtable *route_reverse(struct
sk_buff*skb, int local){ struct iphdr *iph = skb->nh.iph; struct dst_entry *odst; struct rt_key key = {}; struct rtable *rt; if (local) { key.dst = iph->saddr; key.src = iph->daddr; key.tos = RT_TOS(iph->tos); if (ip_route_output_key(&rt, &key) != 0) return NULL; } else { /* non-local src, find valid iif to satisfy * rp-filter when calling ip_route_input. */ key.dst = iph->daddr; if (ip_route_output_key(&rt, &key) != 0) return NULL; odst = skb->dst; if (ip_route_input(skb, iph->saddr, iph->daddr, RT_TOS(iph->tos), rt->u.dst.dev) != 0) { dst_release(&rt->u.dst); return NULL; } dst_release(&rt->u.dst); rt = (struct rtable *)skb->dst; skb->dst = odst; }
Quoted 8 lines
if (rt->u.dst.error) {
dst_release(&rt->u.dst);
rt = NULL;
}
return rt;
}66,69c107
< /* Routing: if not headed for us, route won't like source */
< if (ip_route_output(&rt, oldskb->nh.iph->saddr,
< local ? oldskb->nh.iph->daddr : 0,
< RT_TOS(oldskb->nh.iph->tos), 0) != 0)
---
Quoted 1 line
if ((rt = route_reverse(oldskb, local)) == NULL)
332a371,373
Quoted 3 lines
case IPT_ICMP_ADMIN_PROHIBITED:
send_unreach(*pskb, ICMP_PKT_FILTERED);
break;As I think problem in new new route_reverse function which called from tcp_reset() procedure
So any new ideas?
Quoted 22 lines
Subject:
Netfilter problem with new 2.4.22
From:
Diadon <[email protected]>
Date:
Tue, 16 Sep 2003 14:22:37 +0400
To:
[email protected]After installing 2.4.22
this chain doesn't work
$IPPROG -A OUTPUT -p tcp --dport 113 -j REJECT --reject-with tcp-resetOn 2.4.21 all works fine
In tcpdump on 2.4.21:
14:41:41.752557 somehost.auth > somehost1.32825: R 0:0(0) ack 217583467 win 0 (DF)In tcpdump on 2.4.22:
nothing.......any ideas?
Quoted 7 lines · Diadon
Quoted 3 lines
On 2.4.21 all works fine
In tcpdump on 2.4.21:
14:41:41.752557 somehost.auth > somehost1.32825: R 0:0(0) ack217583467 win 0 (DF)
Quoted 2 lines
In tcpdump on 2.4.22:
nothing.......
Hi Diadon,
the problem seems to be that a dst for local input doesn't carry pmtu information, the pmtu is set by rt_set_nexthop which is skipped for local input. The packet is dropped by send_reset because of this check:
/* "Never happens" */
if (nskb->len > nskb->dst->pmtu)
goto free_nskb;;)
I've attached two possible fixes for this. The first one restores behaviour from before the routing changes for LOCAL_OUT, the other one removes the check since obviously "Never happens" is not true anymore (and it is not an error). Another possibility would be something like "if (nskb->dst->pmtu && nskb->len > nskb->dst->pmtu) ..." Someone from the coreteam should comment which solution is prefered.
Regards,
Patrick
===== net/ipv4/netfilter/ipt_REJECT.c 1.13 vs edited =====
net/ipv4/netfilter/ipt_REJECT.c +7 -6
@@ -34,16 +34,17 @@attach (new_skb, nfct); } -static inline struct rtable *route_reverse (struct sk_buff *skb,int local) +static inline struct rtable *route_reverse (struct sk_buff *skb,int hook) {struct iphdr *iph = skb->nh.iph;struct dst_entry *odst;struct rt_key key = {};struct rtable *rt; -if (local) { +if (hook !=NF_IP_FORWARD ) { key.dst = iph->saddr; - key.src = iph->daddr; +if (hook ==NF_IP_LOCAL_IN ) + key.src = iph->daddr; key.tos =RT_TOS (iph->tos);if (ip_route_output_key (&rt, &key) !=0 )
@@ -75,7 +76,7 @@ }/* Send RST reply */ -static void send_reset (struct sk_buff *oldskb,int local) +static void send_reset (struct sk_buff *oldskb,int hook) {struct sk_buff *nskb;struct tcphdr *otcph, *tcph;
@@ -104,7 +105,7 @@ csum_partial((char *)otcph, otcplen, 0)) != 0)return ; -if ((rt =route_reverse (oldskb, local)) ==NULL ) +if ((rt =route_reverse (oldskb, hook)) ==NULL )return ; hh_len = (rt->u.dst.dev->hard_header_len +15 )&~15 ;
@@ -372,7 +373,7 @@send_unreach (*pskb,ICMP_PKT_FILTERED );break ;case IPT_TCP_RESET : -send_reset (*pskb, hooknum ==NF_IP_LOCAL_IN ); +send_reset (*pskb, hooknum);case IPT_ICMP_ECHOREPLY :/* Doesn't happen. */ break ;
===== net/ipv4/netfilter/ipt_REJECT.c 1.13 vs edited =====
net/ipv4/netfilter/ipt_REJECT.c +0 -4
@@ -186,10 +186,6 @@ nskb->nh.iph->check =ip_fast_csum ((unsigned char *)nskb->nh.iph, nskb->nh.iph->ihl); -/* "Never happens" */ -if (nskb->len > nskb->dst->pmtu) - goto free_nskb; -connection_attach (nskb, oldskb->nfct); NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
Quoted 4 lines · Patrick McHardy
Hi Diadon,
the problem seems to be that a dst for local input doesn't carry pmtu information, the pmtu is set byrt_set_nexthopwhich is skipped for local input.
Thanks for debugging the problem, Patrick.
Quoted 6 lines
I've attached two possible fixes for this. The first one restores behaviour from before the routing changes for
LOCAL_OUT, the other one removes the check since obviously "Never happens" is not true anymore (and it is not an error). Another possibility would be something like "if (nskb->dst->pmtu&&nskb->len>nskb->dst->pmtu) ..." Someone from the coreteam should comment which solution is prefered.
I'd actually prefer the last solution (together with a comment why nskb->dst-pmtu can be zero).
I've included the patch into 'submitted' (will send it to davem later)
Quoted 2 lines
Regards,
Patrick
Signature · 5 lines
- Harald Welte <[email protected]> http://www.netfilter.org/ ============================================================================ "Fragmentation is like classful addressing -- an interesting early architectural error that shows how much experimentation was going on while IP was being designed." -- Paul Vixie
That patch is not work, after patching the kernel problem is not disappeared!
That patch is not work, after patching the kernel problem is not disappeared!