ICMP echo conntrack timeout
Subsystems (2)
PATCH 1/1 ICMP echo conntrack timeout
Hello, netfilter developers!
I am trying to delploy a conntrackd-based HA router and (if possible) a flow statistics collector. I have discovered that nf_conntrack treats each ICMP echo request/reply pair as a separate connection (in net/ipv4/netfilter/nf_conntrack_proto_icmp.c:icmp_packet() function). This has several problems:
- excessive conntrackd traffic when the ping is running over the router (one new "connection" per echo request/reply pair).
- should there be a duplicated ICMP echo reply (such as when when pinging a cluster IP address), only the first echo reply is seen as ESTABLISHED, the rest is INVALID.
- no "per-flow" statistics available, as there is no notion of the "flow" at all.
I think it would be better to keep the default timeout ofnf_ct_icmp_timeout even after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.
Thanks,
-Yenya
From: Jan "Yenya" Kasprzak <[email protected]>
Date: Mon, 1 Jun 2009 18:53:20 +0200
Subject: [PATCH] [nf_conntrack] Keep the ICMP ct entries longer
Current conntrack code kills the ICMP conntrack entry as soon as the first reply is received. This is incorrect, as we then see only the first ICMP echo reply out of several possible duplicates as ESTABLISHED, while the rest will be INVALID. Also this unnecessarily increases the conntrackd traffic on H-A firewalls.
Make all the ICMP conntrack entries (including the replied ones) last for the default of nf_conntrack_icmp{,v6}_timeout seconds.
--- include/net/netfilter/ipv4/nf_conntrack_icmp.h | 11 ----------- include/net/netfilter/ipv6/nf_conntrack_icmpv6.h | 7 ------- include/net/netfilter/nf_conntrack.h | 3 --- net/ipv4/netfilter/nf_conntrack_proto_icmp.c | 18 ++++++------------ net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c | 18 ++++++------------ 5 files changed, 12 insertions(+), 45 deletions(-) delete mode 100644 include/net/netfilter/ipv4/nf_conntrack_icmp.h
include/net/netfilter/ipv4/nf_conntrack_icmp.h deleted +0 -11
@@ -1,11 +0,0 @@ -#ifndef _NF_CONNTRACK_ICMP_H -#define _NF_CONNTRACK_ICMP_H -/* ICMP tracking. */ -#include <asm/atomic.h> - -struct ip_ct_icmp -{ -/* Optimization: when number in == number out, forget immediately. */ -atomic_t count; -}; -#endif /* _NF_CONNTRACK_ICMP_H */
include/net/netfilter/ipv6/nf_conntrack_icmpv6.h +0 -7
@@ -9,7 +9,6 @@#ifndef _NF_CONNTRACK_ICMPV6_H#define _NF_CONNTRACK_ICMPV6_H -#include <asm/atomic.h> #ifndef ICMPV6_NI_QUERY #define ICMPV6_NI_QUERY 139
@@ -18,10 +17,4 @@#define ICMPV6_NI_REPLY 140#endif -struct nf_ct_icmpv6 -{ -/* Optimization: when number in == number out, forget immediately. */ -atomic_t count; -}; -#endif /* _NF_CONNTRACK_ICMPV6_H */
include/net/netfilter/nf_conntrack.h +0 -3
@@ -23,7 +23,6 @@#include <linux/netfilter/nf_conntrack_dccp.h> #include <linux/netfilter/nf_conntrack_sctp.h> #include <linux/netfilter/nf_conntrack_proto_gre.h> -#include <net/netfilter/ipv4/nf_conntrack_icmp.h> #include <net/netfilter/ipv6/nf_conntrack_icmpv6.h> #include <net/netfilter/nf_conntrack_tuple.h>
@@ -34,8 +33,6 @@ union nf_conntrack_proto {struct nf_ct_dccp dccp;struct ip_ct_sctp sctp;struct ip_ct_tcp tcp; -struct ip_ct_icmp icmp; -struct nf_ct_icmpv6 icmpv6;struct nf_ct_gre gre; };
net/ipv4/netfilter/nf_conntrack_proto_icmp.c +6 -12
@@ -82,18 +82,13 @@ static int icmp_packet(struct nf_conn *ct, u_int8_t pf, unsigned int hooknum) { -/* Try to delete connection immediately after all replies: -won't actually vanish as we still have skb, and del_timer -means this will only run once even if count hits zero twice -(theoretically possible with SMP) */ -if (CTINFO2DIR (ctinfo) ==IP_CT_DIR_REPLY ) { -if (atomic_dec_and_test (&ct->proto.icmp.count)) -nf_ct_kill_acct (ct, ctinfo, skb); - }else { -atomic_inc (&ct->proto.icmp.count); +/* Do not immediately delete the connection after the first +successful reply to avoid excessive conntrackd traffic +and also to handle correctly ICMP echo reply duplicates. */ +if (CTINFO2DIR (ctinfo) !=IP_CT_DIR_REPLY )nf_conntrack_event_cache (IPCT_PROTOINFO_VOLATILE , ct); -nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmp_timeout); - } + +nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmp_timeout);return NF_ACCEPT ; }
@@ -117,7 +112,6 @@ static bool icmp_new(struct nf_conn *ct, const struct sk_buff *skb,nf_ct_dump_tuple_ip (&ct->tuplehash[0 ].tuple);return false; } -atomic_set (&ct->proto.icmp.count,0 );return true; }
net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c +6 -12
@@ -95,18 +95,13 @@ static int icmpv6_packet(struct nf_conn *ct, u_int8_t pf, unsigned int hooknum) { -/* Try to delete connection immediately after all replies: -won't actually vanish as we still have skb, and del_timer -means this will only run once even if count hits zero twice -(theoretically possible with SMP) */ -if (CTINFO2DIR (ctinfo) ==IP_CT_DIR_REPLY ) { -if (atomic_dec_and_test (&ct->proto.icmp.count)) -nf_ct_kill_acct (ct, ctinfo, skb); - }else { -atomic_inc (&ct->proto.icmp.count); +/* Do not immediately delete the connection after the first +successful reply to avoid excessive conntrackd traffic +and also to handle correctly ICMP echo reply duplicates. */ +if (CTINFO2DIR (ctinfo) !=IP_CT_DIR_REPLY )nf_conntrack_event_cache (IPCT_PROTOINFO_VOLATILE , ct); -nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmpv6_timeout); - } + +nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmpv6_timeout);return NF_ACCEPT ; }
@@ -132,7 +127,6 @@ static bool icmpv6_new(struct nf_conn *ct, const struct sk_buff *skb, type + 128);return false; } -atomic_set (&ct->proto.icmp.count,0 );return true; }
Signature · 11 lines
1.6.0.6
--
| Jan "Yenya" Kasprzak <kas at {fi.muni.cz - work | yenya.net - private}> |
| GPG: ID 1024/D3498839 Fingerprint 0D99A7FB206605D7 8B35FCDE05B18A5E |
| http://www.fi.muni.cz/~kas/ Journal: http://www.fi.muni.cz/~kas/blog/ |Quoted 2 lines
Quoted 2 lines
If we wanted to trade simplicity and kewl design for usability I think <<
we all know the URL of the Apple Store. --jmorris42 @LWN <<
Quoted 15 lines · Jan Kasprzak
Hello, netfilter developers!
I am trying to delploy a conntrackd-based HA router and (if possible) a flow statistics collector. I have discovered that
nf_conntracktreats each ICMP echo request/reply pair as a separate connection (in net/ipv4/netfilter/nf_conntrack_proto_icmp.c:icmp_packet() function). This has several problems:- excessive conntrackd traffic when the ping is running over the router (one new "connection" per echo request/reply pair).
- should there be a duplicated ICMP echo reply (such as when when pinging a cluster IP address), only the first echo reply is seen as ESTABLISHED, the rest is INVALID.
Quoted 6 lines
- no "per-flow" statistics available, as there is no notion of the "flow" at all.
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.
I think this patch makes sense, it also improves behaviour in cases with asymetric MTUs where a fragmentation required is sent after the reply passed through conntrack and conntrack can't associate it to the already gone connection anymore.
Unless someone has concerns I'm missing, I'll apply this.
Pablo Neira Ayuso
maintainerHi Jan,
Quoted 10 lines · Jan Kasprzak
Hello, netfilter developers!
I am trying to delploy a conntrackd-based HA router and (if possible) a flow statistics collector. I have discovered that
nf_conntracktreats each ICMP echo request/reply pair as a separate connection (in net/ipv4/netfilter/nf_conntrack_proto_icmp.c:icmp_packet() function). This has several problems:- excessive conntrackd traffic when the ping is running over the router (one new "connection" per echo request/reply pair).
Indeed, one event per new ICMP echo packet is too much. This can be also a way to exhaust resources of the conntrack subsystem.
Quoted 5 lines
- should there be a duplicated ICMP echo reply (such as when when pinging a cluster IP address), only the first echo reply is seen as ESTABLISHED, the rest is INVALID.
- no "per-flow" statistics available, as there is no notion of the "flow" at all.
Indeed, no packet accounting at all.
Quoted 4 lines
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.
The only problem that I see is that you patch relaxes the current checking that we're doing. I mean, for every packet in one direction we only accept one ICMP reply packet. With your patch, we can accept more than one packet in the reply direction.
I have a similar patch in my internal tree. It allows only one reply packet per original and many packet in the original direction as you want. I'll post it asap so we can discuss on it.
Signature · 1 line
"Los honestos son inadaptados sociales" -- Les Luthiers
Pablo Neira Ayuso wrote:
: The only problem that I see is that you patch relaxes the current : checking that we're doing. I mean, for every packet in one direction we : only accept one ICMP reply packet. With your patch, we can accept more : than one packet in the reply direction.
I think it is perfectly legal, as we _want_ to se the duplicates,
if there are any.
-Yenya
Signature · 3 lines
| Jan "Yenya" Kasprzak <kas at {fi.muni.cz - work | yenya.net - private}> |
| GPG: ID 1024/D3498839 Fingerprint 0D99A7FB206605D7 8B35FCDE05B18A5E |
| http://www.fi.muni.cz/~kas/ Journal: http://www.fi.muni.cz/~kas/blog/ |Quoted 2 lines
Quoted 2 lines
If we wanted to trade simplicity and kewl design for usability I think <<
we all know the URL of the Apple Store. --jmorris42 @LWN <<
Quoted 9 lines · Pablo Neira Ayuso
Quoted 4 lines
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.The only problem that I see is that you patch relaxes the current checking that we're doing. I mean, for every packet in one direction we only accept one ICMP reply packet. With your patch, we can accept more than one packet in the reply direction.
Thats the intention, isn't it? :) I don't see a problem with this, conntrack is supposed to accept valid responses and I don't think its unreasonable to consider duplicate echo-replies as valid.
Pablo Neira Ayuso
maintainerQuoted 6 lines · Patrick McHardy
Quoted 5 lines · Pablo Neira Ayuso
Quoted 4 lines
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.
Quoted 8 lines
Quoted 4 lines
The only problem that I see is that you patch relaxes the current checking that we're doing. I mean, for every packet in one direction we only accept one ICMP reply packet. With your patch, we can accept more than one packet in the reply direction.
Thats the intention, isn't it? :) I don't see a problem with this, conntrack is supposed to accept valid responses and I don't think its unreasonable to consider duplicate echo-replies as valid.
I only wanted to point with this patch we're doing more relaxed ICMP tracking, but I'm fine with this.
BTW, with this patch, we can add state synchronization in conntrackd for ICMP (some bits are still missing to support this). This is something that I don't particularly find very useful, but some people have requested this.
Signature · 1 line
"Los honestos son inadaptados sociales" -- Les Luthiers
Quoted 14 lines · Pablo Neira Ayuso
Quoted 5 lines · Patrick McHardy
Quoted 4 lines · Pablo Neira Ayuso
Quoted 4 lines
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.Quoted 7 lines
Quoted 4 lines
The only problem that I see is that you patch relaxes the current checking that we're doing. I mean, for every packet in one direction we only accept one ICMP reply packet. With your patch, we can accept more than one packet in the reply direction.
Thats the intention, isn't it? :) I don't see a problem with this, conntrack is supposed to accept valid responses and I don't think its unreasonable to consider duplicate echo-replies as valid.
Quoted 7 lines
I only wanted to point with this patch we're doing more relaxed ICMP tracking, but I'm fine with this.
BTW, with this patch, we can add state synchronization in conntrackd for ICMP (some bits are still missing to support this). This is something that I don't particularly find very useful, but some people have requested this.
I guess this really helps for a "ping-demonstration" where you pull the plug and the ping keeps running :)
Pablo Neira Ayuso
maintainerQuoted 18 lines · Patrick McHardy
Quoted 17 lines · Pablo Neira Ayuso
Quoted 5 lines · Patrick McHardy
Quoted 4 lines · Pablo Neira Ayuso
Quoted 4 lines
I think it would be better to keep the default timeout of
nf_ct_icmp_timeouteven after the echo reply is received. Feel free to correct me why early deleting of ICMP conntrack entries is needed, or consider applying the following patch.Quoted 7 lines
Quoted 4 lines
The only problem that I see is that you patch relaxes the current checking that we're doing. I mean, for every packet in one direction we only accept one ICMP reply packet. With your patch, we can accept more than one packet in the reply direction.
Thats the intention, isn't it? :) I don't see a problem with this, conntrack is supposed to accept valid responses and I don't think its unreasonable to consider duplicate echo-replies as valid.
I only wanted to point with this patch we're doing more relaxed ICMP tracking, but I'm fine with this.
Quoted 7 lines
Quoted 4 lines
BTW, with this patch, we can add state synchronization in conntrackd for ICMP (some bits are still missing to support this). This is something that I don't particularly find very useful, but some people have requested this.
I guess this really helps for a "ping-demonstration" where you pull the plug and the ping keeps running :)
Indeed :). For some strange reason, this seems to be one of the very first tests that people do to make sure that their HA firewall works.
Signature · 1 line
"Los honestos son inadaptados sociales" -- Les Luthiers
Quoted 3 lines · Jan Kasprzak
From: Jan "Yenya" Kasprzak <[email protected]>
Date: Mon, 1 Jun 2009 18:53:20 +0200
Subject: [PATCH] [nf_conntrack] Keep the ICMP ct entries longer
Your patch doesn't apply to the latest tree. Could you rediff againstgit://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git
please?
Thanks.
Patrick McHardy wrote:
: Jan Kasprzak wrote:
: >From: Jan "Yenya" Kasprzak <[email protected]>
: >Date: Mon, 1 Jun 2009 18:53:20 +0200
: >Subject: [PATCH] [nf_conntrack] Keep the ICMP ct entries longer :
: Your patch doesn't apply to the latest tree. Could you rediff against : git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git
: please?
Yes, patch attached.
-Yenya
Signature · 3 lines
| Jan "Yenya" Kasprzak <kas at {fi.muni.cz - work | yenya.net - private}> |
| GPG: ID 1024/D3498839 Fingerprint 0D99A7FB206605D7 8B35FCDE05B18A5E |
| http://www.fi.muni.cz/~kas/ Journal: http://www.fi.muni.cz/~kas/blog/ |Quoted 3 lines
Quoted 2 lines
If we wanted to trade simplicity and kewl design for usability I think <<
we all know the URL of the Apple Store. --jmorris42 @LWN <<From
3413b225e0c0950ba6e76f7be863eb90e95a78e3Mon Sep 17 00:00:00 2001
From: Jan "Yenya" Kasprzak <[email protected]>
Date: Mon, 1 Jun 2009 18:53:20 +0200
Subject: [PATCH] [nf_conntrack] Keep the ICMP ct entries longer
Current conntrack code kills the ICMP conntrack entry as soon as the first reply is received. This is incorrect, as we then see only the first ICMP echo reply out of several possible duplicates as ESTABLISHED, while the rest will be INVALID. Also this unnecessarily increases the conntrackd traffic on H-A firewalls.
Make all the ICMP conntrack entries (including the replied ones) last for the default of nf_conntrack_icmp{,v6}_timeout seconds.
--- include/net/netfilter/ipv4/nf_conntrack_icmp.h | 11 ----------- include/net/netfilter/ipv6/nf_conntrack_icmpv6.h | 7 ------- include/net/netfilter/nf_conntrack.h | 3 --- net/ipv4/netfilter/nf_conntrack_proto_icmp.c | 16 ++++------------ net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c | 16 ++++------------ 5 files changed, 8 insertions(+), 45 deletions(-) delete mode 100644 include/net/netfilter/ipv4/nf_conntrack_icmp.h
include/net/netfilter/ipv4/nf_conntrack_icmp.h deleted +0 -11
@@ -1,11 +0,0 @@ -#ifndef _NF_CONNTRACK_ICMP_H -#define _NF_CONNTRACK_ICMP_H -/* ICMP tracking. */ -#include <asm/atomic.h> - -struct ip_ct_icmp -{ -/* Optimization: when number in == number out, forget immediately. */ -atomic_t count; -}; -#endif /* _NF_CONNTRACK_ICMP_H */
include/net/netfilter/ipv6/nf_conntrack_icmpv6.h +0 -7
@@ -9,7 +9,6 @@#ifndef _NF_CONNTRACK_ICMPV6_H#define _NF_CONNTRACK_ICMPV6_H -#include <asm/atomic.h> #ifndef ICMPV6_NI_QUERY #define ICMPV6_NI_QUERY 139
@@ -18,10 +17,4 @@#define ICMPV6_NI_REPLY 140#endif -struct nf_ct_icmpv6 -{ -/* Optimization: when number in == number out, forget immediately. */ -atomic_t count; -}; -#endif /* _NF_CONNTRACK_ICMPV6_H */
include/net/netfilter/nf_conntrack.h +0 -3
@@ -23,7 +23,6 @@#include <linux/netfilter/nf_conntrack_dccp.h> #include <linux/netfilter/nf_conntrack_sctp.h> #include <linux/netfilter/nf_conntrack_proto_gre.h> -#include <net/netfilter/ipv4/nf_conntrack_icmp.h> #include <net/netfilter/ipv6/nf_conntrack_icmpv6.h> #include <net/netfilter/nf_conntrack_tuple.h>
@@ -34,8 +33,6 @@ union nf_conntrack_proto {struct nf_ct_dccp dccp;struct ip_ct_sctp sctp;struct ip_ct_tcp tcp; -struct ip_ct_icmp icmp; -struct nf_ct_icmpv6 icmpv6;struct nf_ct_gre gre; };
net/ipv4/netfilter/nf_conntrack_proto_icmp.c +4 -12
@@ -82,17 +82,10 @@ static int icmp_packet(struct nf_conn *ct, u_int8_t pf, unsigned int hooknum) { -/* Try to delete connection immediately after all replies: -won't actually vanish as we still have skb, and del_timer -means this will only run once even if count hits zero twice -(theoretically possible with SMP) */ -if (CTINFO2DIR (ctinfo) ==IP_CT_DIR_REPLY ) { -if (atomic_dec_and_test (&ct->proto.icmp.count)) -nf_ct_kill_acct (ct, ctinfo, skb); - }else { -atomic_inc (&ct->proto.icmp.count); -nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmp_timeout); - } +/* Do not immediately delete the connection after the first +successful reply to avoid excessive conntrackd traffic +and also to handle correctly ICMP echo reply duplicates. */ +nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmp_timeout);return NF_ACCEPT ; }
@@ -116,7 +109,6 @@ static bool icmp_new(struct nf_conn *ct, const struct sk_buff *skb,nf_ct_dump_tuple_ip (&ct->tuplehash[0 ].tuple);return false; } -atomic_set (&ct->proto.icmp.count,0 );return true; }
net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c +4 -12
@@ -95,17 +95,10 @@ static int icmpv6_packet(struct nf_conn *ct, u_int8_t pf, unsigned int hooknum) { -/* Try to delete connection immediately after all replies: -won't actually vanish as we still have skb, and del_timer -means this will only run once even if count hits zero twice -(theoretically possible with SMP) */ -if (CTINFO2DIR (ctinfo) ==IP_CT_DIR_REPLY ) { -if (atomic_dec_and_test (&ct->proto.icmp.count)) -nf_ct_kill_acct (ct, ctinfo, skb); - }else { -atomic_inc (&ct->proto.icmp.count); -nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmpv6_timeout); - } +/* Do not immediately delete the connection after the first +successful reply to avoid excessive conntrackd traffic +and also to handle correctly ICMP echo reply duplicates. */ +nf_ct_refresh_acct (ct, ctinfo, skb, nf_ct_icmpv6_timeout);return NF_ACCEPT ; }
@@ -131,7 +124,6 @@ static bool icmpv6_new(struct nf_conn *ct, const struct sk_buff *skb, type + 128);return false; } -atomic_set (&ct->proto.icmp.count,0 );return true; }
Signature · 1 line
1.6.0.6
Patrick McHardy
AppliedQuoted 10 lines · Jan Kasprzak
Patrick McHardy wrote:
: Jan Kasprzak wrote:
: >From: Jan "Yenya" Kasprzak <[email protected]>
: >Date: Mon, 1 Jun 2009 18:53:20 +0200
: >Subject: [PATCH] [nf_conntrack] Keep the ICMP ct entries longer : : Your patch doesn't apply to the latest tree. Could you rediff against :git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git
: please?Yes, patch attached.
Applied, thanks Jan.