introduce the pickup flag to take over connections

PATCH In review Pablo Neira Ayuso 2 participants 4 messages Jul 7, 2006 – Jul 17, 2006

Subsystems (2)

Networking netfilter

lore

Share & tools

lore

Pablo Neira Ayuso

author maintainer

2006-07-07 02:16 UTC #0 lore raw

PATCH 10/10 introduce the pickup flag to take over connections

This patch introduces a new flag called IPS_PICKUP that forces the protocol handler to pick up the required information in order to ensure that the connection will reach a successful state. Currently, the only client is the TCP protocol helper.

More arguments conntrackd or whatever failover solution need this.

@Jozsef: You know better TCP sequence tracking in-deep details, can you see any problem with this?

Signed-off-by Pablo Neira Ayuso <[email protected]>

Signature · 2 lines

The dawn of the fourth age of Linux firewalling is coming; a time of
great struggle and heroic deeds -- J.Kadlecsik got inspired by J.Morris

[CONNTRACK] Introduce the pickup flag to take over connections

Signed-off-by Pablo Neira Ayuso <[email protected]>

net/ipv4/netfilter/ip_conntrack_proto_tcp.c +35 ~~-20~~

Index: net-2.6/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
===================================================================
@@ -425,10 +425,10 @@ static unsigned int get_conntrack_index(
    we doesn't have to deal with fragments. 
 */
 
-static inline __u32 segment_seq_plus_len(__u32 seq,
-					 size_t len,
-					 struct iphdr *iph,
-					 struct tcphdr *tcph)
+static inline __u32 segment_seq_plus_len(const __u32 seq,
+					 const size_t len,
+					 const struct iphdr *iph,
+					 const struct tcphdr *tcph)
 {
 	return (seq + len - (iph->ihl + tcph->doff)*4
 		+ (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));

@@ -890,6 +890,28 @@ static int tcp_error(struct sk_buff *skb
 	return NF_ACCEPT;
 }
 
+static void tcp_pickup_connection(struct ip_conntrack *conntrack,
+				  const struct sk_buff *skb,
+				  const struct iphdr *iph,
+				  const struct tcphdr *th)
+{
+	conntrack->proto.tcp.seen[0].td_end =
+		segment_seq_plus_len(ntohl(th->seq), skb->len, iph, th);
+	conntrack->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
+	if (conntrack->proto.tcp.seen[0].td_maxwin == 0)
+		conntrack->proto.tcp.seen[0].td_maxwin = 1;
+	conntrack->proto.tcp.seen[0].td_maxend =
+		conntrack->proto.tcp.seen[0].td_end + 
+		conntrack->proto.tcp.seen[0].td_maxwin;
+		conntrack->proto.tcp.seen[0].td_scale = 0;
+
+	/* We assume SACK. Should we assume window scaling too? */
+	conntrack->proto.tcp.seen[0].flags =
+	conntrack->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM;
+	conntrack->proto.tcp.seen[0].loose = 
+	conntrack->proto.tcp.seen[1].loose = ip_ct_tcp_loose;
+}
+
 /* Returns verdict for packet, or -1 for invalid. */
 static int tcp_packet(struct ip_conntrack *conntrack,
 		      const struct sk_buff *skb,

@@ -907,6 +929,14 @@ static int tcp_packet(struct ip_conntrac
 	BUG_ON(th == NULL);
 	
 	write_lock_bh(&tcp_lock);
+
+	/*
+	 * This conntrack was added via ctnetlink or ct_sync and needs to
+	 * take over sequence tracking in order to work properly.
+	 */
+	if (test_and_clear_bit(IPS_PICKUP, &conntrack->status))
+		tcp_pickup_connection(conntrack, skb, iph, th);
+
 	old_state = conntrack->proto.tcp.state;
 	dir = CTINFO2DIR(ctinfo);
 	index = get_conntrack_index(th);

@@ -1116,22 +1146,7 @@ static int tcp_new(struct ip_conntrack *
 		 * its history is lost for us.
 		 * Let's try to use the data from the packet.
 		 */
-		conntrack->proto.tcp.seen[0].td_end =
-			segment_seq_plus_len(ntohl(th->seq), skb->len,
-					     iph, th);
-		conntrack->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
-		if (conntrack->proto.tcp.seen[0].td_maxwin == 0)
-			conntrack->proto.tcp.seen[0].td_maxwin = 1;
-		conntrack->proto.tcp.seen[0].td_maxend =
-			conntrack->proto.tcp.seen[0].td_end + 
-			conntrack->proto.tcp.seen[0].td_maxwin;
-		conntrack->proto.tcp.seen[0].td_scale = 0;
-
-		/* We assume SACK. Should we assume window scaling too? */
-		conntrack->proto.tcp.seen[0].flags =
-		conntrack->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM;
-		conntrack->proto.tcp.seen[0].loose = 
-		conntrack->proto.tcp.seen[1].loose = ip_ct_tcp_loose;
+		tcp_pickup_connection(conntrack, skb, iph, th);
 	}
     
 	conntrack->proto.tcp.seen[1].td_end = 0;

include/linux/netfilter/nf_conntrack_common.h +4 -0

Index: net-2.6/include/linux/netfilter/nf_conntrack_common.h
===================================================================
@@ -73,6 +73,10 @@ enum ip_conntrack_status {
 	/* Connection has fixed timeout. */
 	IPS_FIXED_TIMEOUT_BIT = 10,
 	IPS_FIXED_TIMEOUT = (1 << IPS_FIXED_TIMEOUT_BIT),
+
+	/* Pick up connection information if required */
+	IPS_PICKUP_BIT = 11,
+	IPS_PICKUP = (1 << IPS_PICKUP_BIT),
 };
 
 /* Connection tracking event bits */

net/netfilter/nf_conntrack_proto_tcp.c +37 ~~-20~~

93 lines not on this page · raw message

Jozsef Kadlecsik

2006-07-10 13:26 UTC #1 lore raw

Hi Pablo,

Quoted 9 lines · Pablo Neira Ayuso

This patch introduces a new flag called IPS_PICKUP that forces the protocol handler to pick up the required information in order to ensure that the connection will reach a successful state. Currently, the only client is the TCP protocol helper.
More arguments conntrackd or whatever failover solution need this.
@Jozsef: You know better TCP sequence tracking in-deep details, can you see any problem with this?

Nothing against it as I see. But I'd regard it as a preliminary step toward creating conntrack entries by conntrackd/failover: the missing bits are the flags (SACK and WSCALE) and the window scale factor. We assume SACK is on (which thus don't hurt), but disabled window scaling can really bite if it's actually on.

Best regards,
Jozsef
-
E-mail : [email protected], [email protected] PGP key : http://www.kfki.hu/~kadlec/pgp_public_key.txt
Address : KFKI Research Institute for Particle and Nuclear Physics

          H-1525 Budapest 114, POB. 49, Hungary

Pablo Neira Ayuso

author maintainer

2006-07-13 20:22 UTC #2 lore raw

Quoted 12 lines · Jozsef Kadlecsik

Hi Pablo,
Quoted 7 lines · Pablo Neira Ayuso
This patch introduces a new flag called IPS_PICKUP that forces the protocol handler to pick up the required information in order to ensure that the connection will reach a successful state. Currently, the only client is the TCP protocol helper.
More arguments conntrackd or whatever failover solution need this.

Quoted 8 lines

Quoted 2 lines
@Jozsef: You know better TCP sequence tracking in-deep details, can you see any problem with this?
Nothing against it as I see. But I'd regard it as a preliminary step toward creating conntrack entries by conntrackd/failover: the missing bits are the flags (SACK and WSCALE) and the window scale factor. We assume SACK is on (which thus don't hurt), but disabled window scaling can really bite if it's actually on.

Interesting, since the scale factor is only advertised in the SYN+ACK packet, we need a patch to introduce a new ctnetlink attribute that contains the scale factor. I think that should be enough.

Thanks for the clue.

Signature · 1 line

The dawn of the fourth age of Linux firewalling is coming; a time of great struggle and heroic deeds -- J.Kadlecsik got inspired by J.Morris

Jozsef Kadlecsik

2006-07-17 08:17 UTC #3 lore raw

Quoted 22 lines · Pablo Neira Ayuso

Quoted 12 lines · Jozsef Kadlecsik
Hi Pablo,
Quoted 7 lines · Pablo Neira Ayuso
This patch introduces a new flag called IPS_PICKUP that forces the protocol handler to pick up the required information in order to ensure that the connection will reach a successful state. Currently, the only client is the TCP protocol helper.
More arguments conntrackd or whatever failover solution need this.
Quoted 8 lines
Quoted 2 lines
@Jozsef: You know better TCP sequence tracking in-deep details, can you see any problem with this?
Nothing against it as I see. But I'd regard it as a preliminary step toward creating conntrack entries by conntrackd/failover: the missing bits are the flags (SACK and WSCALE) and the window scale factor. We assume SACK is on (which thus don't hurt), but disabled window scaling can really bite if it's actually on.

Quoted 3 lines

Interesting, since the scale factor is only advertised in the SYN+ACK packet, we need a patch to introduce a new ctnetlink attribute that contains the scale factor. I think that should be enough.

Yes, that'd be sufficient (of course it can be advertised in both directions). Window scaling is a must when someone wants to utilize a high badwidth connection as much as possible in one TCP stream and conntrack can go totally wrong if it uses wrong scale factors.

Best regards,
Jozsef
-
E-mail : [email protected], [email protected] PGP key : http://www.kfki.hu/~kadlec/pgp_public_key.txt
Address : KFKI Research Institute for Particle and Nuclear Physics

          H-1525 Budapest 114, POB. 49, Hungary