Skip to contentSkip to conversation

Introduce flag facilities to take over TCP connections

PATCH In review Pablo Neira Ayuso 2 participants 5 messages Nov 28, 2006 – Nov 29, 2006

Subsystems (2)
lore
Share & tools
lore

Pablo Neira Ayuso

author maintainer
#0 lore raw

PATCH 1/3 Introduce flag facilities to take over TCP connections

Hi Patrick,

I reworked the previous patch based on some of your suggestions. Let me know what you think.

Description:

This patch introduces two new flags called IPS_PICKUP that forces the protocol handler to pick up the window of valid TCP packets and IPS_IN_WINDOW to by pass window checkings.

Moreover, four new attributes to inject the window scale factor and internal TCP flags handling. These new facilities provide the appropiate mechanisms to take over TCP connections in failover settings with TCP tracking enabled.

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 flag facilities to take over TCP connections

This patch introduces two new flags called IPS_PICKUP that forces the protocol handler to pick up the window of valid TCP packets and IPS_IN_WINDOW to by pass window checkings.

Moreover, four new attributes to inject the window scale factor and internal TCP flags handling. These new facilities provide the appropiate mechanisms to take over TCP connections in failover settings with TCP tracking enabled.

net/ipv4/netfilter/ip_conntrack_proto_tcp.c +74 -16
Index: linux-2.6.git/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
===================================================================
@@ -345,8 +345,25 @@ static int tcp_to_nfattr(struct sk_buff 
 	nest_parms = NFA_NEST(skb, CTA_PROTOINFO_TCP);
 	NFA_PUT(skb, CTA_PROTOINFO_TCP_STATE, sizeof(u_int8_t),
 		&ct->proto.tcp.state);
-	read_unlock_bh(&tcp_lock);
 
+	/* skip window scale and flags dump if hard tracking is by passed */
+	if (test_bit(IPS_IN_WINDOW, &ct->status))
+		goto out;
+
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].td_scale);
+
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].td_scale);
+
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].flags);
+	
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].flags);
+
+out:
+	read_unlock_bh(&tcp_lock);
 	NFA_NEST_END(skb, nest_parms);
 
 	return 0;
@@ -357,7 +374,11 @@ nfattr_failure:
 }
 
 static const size_t cta_min_tcp[CTA_PROTOINFO_TCP_MAX] = {
-	[CTA_PROTOINFO_TCP_STATE-1]	= sizeof(u_int8_t),
+	[CTA_PROTOINFO_TCP_STATE-1]	      = sizeof(u_int8_t),
+	[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL-1] = sizeof(u_int8_t),
+	[CTA_PROTOINFO_TCP_WSCALE_REPLY-1]    = sizeof(u_int8_t),
+	[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL-1]  = sizeof(struct nf_ct_tcp_flags),
+	[CTA_PROTOINFO_TCP_FLAGS_REPLY-1]     = sizeof(struct nf_ct_tcp_flags)
 };
 
 static int nfattr_to_tcp(struct nfattr *cda[], struct ip_conntrack *ct)
@@ -381,6 +402,30 @@ static int nfattr_to_tcp(struct nfattr *
 	write_lock_bh(&tcp_lock);
 	ct->proto.tcp.state = 
 		*(u_int8_t *)NFA_DATA(tb[CTA_PROTOINFO_TCP_STATE-1]);
+
+	if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL-1]) {
+		struct nf_ct_tcp_flags *attr =
+			NFA_DATA(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL-1]);
+		ct->proto.tcp.seen[0].flags &= ~attr->mask;
+		ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
+	}
+
+	if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY-1]) {
+		struct nf_ct_tcp_flags *attr =
+			NFA_DATA(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY-1]);
+		ct->proto.tcp.seen[1].flags &= ~attr->mask;
+		ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
+	}
+
+	if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL-1] &&
+	    tb[CTA_PROTOINFO_TCP_WSCALE_REPLY-1] &&
+	    ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE &&
+	    ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
+		ct->proto.tcp.seen[0].td_scale = *(u_int8_t *)
+			NFA_DATA(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL-1]);
+		ct->proto.tcp.seen[1].td_scale = *(u_int8_t *)
+			NFA_DATA(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY-1]);
+	}
 	write_unlock_bh(&tcp_lock);
 
 	return 0;
@@ -424,10 +469,10 @@ 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 +935,22 @@ static int tcp_error(struct sk_buff *skb
 	return NF_ACCEPT;
 }
 
+static inline void tcp_pickup_window(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;
+}
+
 /* Returns verdict for packet, or -1 for invalid. */
 static int tcp_packet(struct ip_conntrack *conntrack,
 		      const struct sk_buff *skb,
@@ -912,6 +973,12 @@ static int tcp_packet(struct ip_conntrac
 	index = get_conntrack_index(th);
 	new_state = tcp_conntracks[dir][index][old_state];
 
+	/* pick up or by-pass window tracking */
+	if (test_bit(IPS_IN_WINDOW, &conntrack->status))
+		goto in_window;
+	else if (test_and_clear_bit(IPS_PICKUP, &conntrack->status))
+		tcp_pickup_window(conntrack, skb, iph, th);
+
 	switch (new_state) {
 	case TCP_CONNTRACK_IGNORE:
 		/* Ignored packets:
@@ -1116,16 +1183,7 @@ 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;
+		tcp_pickup_window(conntrack, skb, iph, th);
 
 		/* We assume SACK. Should we assume window scaling too? */
 		conntrack->proto.tcp.seen[0].flags =
include/linux/netfilter/nf_conntrack_common.h +8 -0
Index: linux-2.6.git/include/linux/netfilter/nf_conntrack_common.h
===================================================================
@@ -73,6 +73,14 @@ 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. */
+	IPS_PICKUP_BIT = 11,
+	IPS_PICKUP = (1 << IPS_PICKUP_BIT),
+
+	/* Ignore in-window checkings. */
+	IPS_IN_WINDOW_BIT = 12,
+	IPS_IN_WINDOW = (1 << IPS_IN_WINDOW_BIT),
 };
 
 /* Connection tracking event bits */
net/netfilter/nf_conntrack_proto_tcp.c +74 -16

133 lines not on this page · raw message

include/linux/netfilter/nfnetlink_conntrack.h +4 -0
Index: linux-2.6.git/include/linux/netfilter/nfnetlink_conntrack.h
===================================================================
@@ -83,6 +83,10 @@ enum ctattr_protoinfo {
 enum ctattr_protoinfo_tcp {
 	CTA_PROTOINFO_TCP_UNSPEC,
 	CTA_PROTOINFO_TCP_STATE,
+	CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
+	CTA_PROTOINFO_TCP_WSCALE_REPLY,
+	CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
+	CTA_PROTOINFO_TCP_FLAGS_REPLY,
 	__CTA_PROTOINFO_TCP_MAX
 };
 #define CTA_PROTOINFO_TCP_MAX (__CTA_PROTOINFO_TCP_MAX - 1)
include/linux/netfilter/nf_conntrack_tcp.h +5 -0
Index: linux-2.6.git/include/linux/netfilter/nf_conntrack_tcp.h
===================================================================
@@ -27,6 +27,11 @@ enum tcp_conntrack {
 /* This sender sent FIN first */
 #define IP_CT_TCP_FLAG_CLOSE_INIT		0x03
 
+struct nf_ct_tcp_flags {
+	u_int8_t flags;
+	u_int8_t mask;
+};
+
 #ifdef __KERNEL__
 
 struct ip_ct_tcp_state {
#1 lore raw
Quoted 4 lines · Pablo Neira Ayuso

Hi Patrick,

I reworked the previous patch based on some of your suggestions. Let me know what you think.

We're almost there :)

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 349

Quoted 1 line
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +1 -0
+	/* skip window scale and flags dump if hard tracking is by passed */

minor nitpick: bypassed

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 363

Quoted 5 lines
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +5 -0
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].flags);
+	
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].flags);

The attributes should contain the same data type in both directions, the receive side expects a structure.

author maintainer
#2 lore raw
Quoted 11 lines · Patrick McHardy
Quoted 4 lines · Pablo Neira Ayuso

Hi Patrick,

I reworked the previous patch based on some of your suggestions. Let me know what you think.

We're almost there :)

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 349

Quoted 1 line
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +1 -0
+	/* skip window scale and flags dump if hard tracking is by passed */

minor nitpick: bypassed

OK ;)

Quoted 8 lines

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 363

Quoted 5 lines
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +5 -0
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].flags);
+	
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].flags);

The attributes should contain the same data type in both directions, the receive side expects a structure.

how could the mask field make sense for the dumping?

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
#3 lore raw
Quoted 12 lines · Pablo Neira Ayuso
Quoted 8 lines · Patrick McHardy

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 363

Quoted 5 lines
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +5 -0
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].flags);
+	
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].flags);

The attributes should contain the same data type in both directions, the receive side expects a structure.

Quoted 1 line

how could the mask field make sense for the dumping?

It doesn't really make sense, but an attribute encapsulates one type of information and should always use the same representation. If this is sent over the wire to another host for example it would need to distinguish whether it contains the kernel->user representation or the user->kernel representation, which IMO is bad design.

author maintainer
#4 lore raw
Quoted 11 lines · Patrick McHardy
Quoted 10 lines · Pablo Neira Ayuso
Quoted 7 lines · Patrick McHardy

net/ipv4/netfilter/ip_conntrack_proto_tcp.c static int tcp_to_nfattr(struct sk_buff · line 363

Quoted 5 lines
net/ipv4/netfilter/ip_conntrack_proto_tcp.c +5 -0
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[0].flags);
+	
+	NFA_PUT(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, sizeof(u_int8_t),
+		&ct->proto.tcp.seen[1].flags);

The attributes should contain the same data type in both directions, the receive side expects a structure.

Quoted 8 lines
Quoted 1 line

how could the mask field make sense for the dumping?

It doesn't really make sense, but an attribute encapsulates one type of information and should always use the same representation. If this is sent over the wire to another host for example it would need to distinguish whether it contains the kernel->user representation or the user->kernel representation, which IMO is bad design.

Totally right, I'll rework this minor issue and get back to you.

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