[PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Timo Teräs]

When GRE tunnel is in NBMA mode, this patch allows an application to use
a PF_PACKET socket to:
- send a packet to specific NBMA address with sendto()
- use recvfrom() to receive packet and check which NBMA address it came from

Signed-off-by: Timo Teras <timo.teras@iki.fi>

---

This is useful for implementing NHRP (and other low level protocols) over
NBMA GRE tunnels.

Posted about a week ago, but got no reply. Could someone comment on this?
Or commit it? Thanks.

net/ipv4/ip_gre.c |   12 +++++++++++-
1 files changed, 11 insertions(+), 1 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index f151900..b1e3816 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -1033,7 +1033,6 @@ static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
	return 0;
}

-#ifdef CONFIG_NET_IPGRE_BROADCAST
/* Nice toy. Unfortunately, useless in real life :-)
  It allows to construct virtual multiprotocol broadcast "LAN"
  over the Internet, provided multicast routing is tuned.
@@ -1092,10 +1091,19 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
	return -t->hlen;
}

+static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
+{
+	struct iphdr *iph = (struct iphdr *) skb_mac_header(skb);
+	memcpy(haddr, &iph->saddr, 4);
+	return 4;
+}
+
static const struct header_ops ipgre_header_ops = {
	.create	= ipgre_header,
+	.parse  = ipgre_header_parse,
};

+#ifdef CONFIG_NET_IPGRE_BROADCAST
static int ipgre_open(struct net_device *dev)
{
	struct ip_tunnel *t = netdev_priv(dev);
@@ -1197,6 +1205,8 @@ static int ipgre_tunnel_init(struct net_device *dev)
			dev->stop = ipgre_close;
		}
#endif
+	} else {
+		dev->header_ops = &ipgre_header_ops;
	}

	if (!tdev && tunnel->parms.link)

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Patrick McHardy]

Timo Teräs wrote:
> When GRE tunnel is in NBMA mode, this patch allows an application to use
> a PF_PACKET socket to:
> - send a packet to specific NBMA address with sendto()
> - use recvfrom() to receive packet and check which NBMA address it came 
> from
> 
> Signed-off-by: Timo Teras <timo.teras@iki.fi>
> 
> ---
> 
> This is useful for implementing NHRP (and other low level protocols) over
> NBMA GRE tunnels.
> 
> Posted about a week ago, but got no reply. Could someone comment on this?
> Or commit it? Thanks.


Your mailer mangled tabs, it won't apply like this. Try sending to
yourself for testing, then repost when it works properly.

[PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Timo Teräs]

When GRE tunnel is in NBMA mode, this patch allows an application to use
a PF_PACKET socket to:
- send a packet to specific NBMA address with sendto()
- use recvfrom() to receive packet and check which NBMA address it came from

This is required to implement properly NHRP over GRE tunnel.

Signed-off-by: Timo Teras <timo.teras@iki.fi>

---
> Your mailer mangled tabs, it won't apply like this. Try sending to
> yourself for testing, then repost when it works properly. 

Sorry. Tabs were ok, but the initial white space on context lines was somehow
removed. Apparently a bug in my editor where I cut/pasted from.

 net/ipv4/ip_gre.c |   12 +++++++++++-
 1 files changed, 11 insertions(+), 1 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index f151900..b1e3816 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -1033,7 +1033,6 @@ static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 	return 0;
 }
 
-#ifdef CONFIG_NET_IPGRE_BROADCAST
 /* Nice toy. Unfortunately, useless in real life :-)
    It allows to construct virtual multiprotocol broadcast "LAN"
    over the Internet, provided multicast routing is tuned.
@@ -1092,10 +1091,19 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
 	return -t->hlen;
 }
 
+static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
+{
+	struct iphdr *iph = (struct iphdr*) skb_mac_header(skb);
+	memcpy(haddr, &iph->saddr, 4);
+	return 4;
+}
+
 static const struct header_ops ipgre_header_ops = {
 	.create	= ipgre_header,
+	.parse	= ipgre_header_parse,
 };
 
+#ifdef CONFIG_NET_IPGRE_BROADCAST
 static int ipgre_open(struct net_device *dev)
 {
 	struct ip_tunnel *t = netdev_priv(dev);
@@ -1197,6 +1205,8 @@ static int ipgre_tunnel_init(struct net_device *dev)
 			dev->stop = ipgre_close;
 		}
 #endif
+	} else {
+		dev->header_ops = &ipgre_header_ops;
 	}
 
 	if (!tdev && tunnel->parms.link)

[PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Timo Teräs]

When GRE tunnel is in NBMA mode, this patch allows an application to use
a PF_PACKET socket to:
- send a packet to specific NBMA address with sendto()
- use recvfrom() to receive packet and check which NBMA address it came from

This is required to implement properly NHRP over GRE tunnel.

Signed-off-by: Timo Teras <timo.teras@iki.fi>

---
Patrick McHardy wrote:
> Your mailer mangled tabs, it won't apply like this. Try sending to
> yourself for testing, then repost when it works properly.

Tabs were ok, I think. But somehow it seems that cut'n'paste from my
default editor loses the space from context lines. It should be good now.

 net/ipv4/ip_gre.c |   12 +++++++++++-
 1 files changed, 11 insertions(+), 1 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index f151900..b1e3816 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -1033,7 +1033,6 @@ static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 	return 0;
 }
 
-#ifdef CONFIG_NET_IPGRE_BROADCAST
 /* Nice toy. Unfortunately, useless in real life :-)
    It allows to construct virtual multiprotocol broadcast "LAN"
    over the Internet, provided multicast routing is tuned.
@@ -1092,10 +1091,19 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
 	return -t->hlen;
 }
 
+static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
+{
+	struct iphdr *iph = (struct iphdr*) skb_mac_header(skb);
+	memcpy(haddr, &iph->saddr, 4);
+	return 4;
+}
+
 static const struct header_ops ipgre_header_ops = {
 	.create	= ipgre_header,
+	.parse	= ipgre_header_parse,
 };
 
+#ifdef CONFIG_NET_IPGRE_BROADCAST
 static int ipgre_open(struct net_device *dev)
 {
 	struct ip_tunnel *t = netdev_priv(dev);
@@ -1197,6 +1205,8 @@ static int ipgre_tunnel_init(struct net_device *dev)
 			dev->stop = ipgre_close;
 		}
 #endif
+	} else {
+		dev->header_ops = &ipgre_header_ops;
 	}
 
 	if (!tdev && tunnel->parms.link)

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Patrick McHardy]

Timo Teräs wrote:
> When GRE tunnel is in NBMA mode, this patch allows an application to use
> a PF_PACKET socket to:
> - send a packet to specific NBMA address with sendto()
> - use recvfrom() to receive packet and check which NBMA address it 
> came from
>
> This is required to implement properly NHRP over GRE tunnel.
>
> Signed-off-by: Timo Teras <timo.teras@iki.fi>
>
> ---
> Patrick McHardy wrote:
>> Your mailer mangled tabs, it won't apply like this. Try sending to
>> yourself for testing, then repost when it works properly.
>
> Tabs were ok, I think. But somehow it seems that cut'n'paste from my
> default editor loses the space from context lines. It should be good now.

Might have been a problem on my side, I just updated my mailer
and it still displays tabs broken, but the actual patch seems
fine. Sorry for the noise ...

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Alexey Kuznetsov]

Hello!

> When GRE tunnel is in NBMA mode, this patch allows an application to use
> a PF_PACKET socket to:
> - send a packet to specific NBMA address with sendto()
> - use recvfrom() to receive packet and check which NBMA address it came from
> 
> This is required to implement properly NHRP over GRE tunnel.

Ack. This is good idea.

Frankly, I was sure ip_gre worked in this way all these years.
I do not remember any reasons why it was crippled.

The only dubious case is when next hop is set using routing tables.
But code in ipgre_tunnel_xmit() is ready to accept this situation,
it checks for zero destination address and fixes it when it is able to.

Alexey

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[David Miller]

From: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Date: Tue, 23 Oct 2007 23:03:24 +0400

> Hello!
> 
> > When GRE tunnel is in NBMA mode, this patch allows an application to use
> > a PF_PACKET socket to:
> > - send a packet to specific NBMA address with sendto()
> > - use recvfrom() to receive packet and check which NBMA address it came from
> > 
> > This is required to implement properly NHRP over GRE tunnel.
> 
> Ack. This is good idea.

Patch applied, thanks everyone!

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Alexey Kuznetsov]

Hello!

Me wrote:

> Ack. This is good idea.
> 
> Frankly, I was sure ip_gre worked in this way all these years.
> I do not remember any reasons why it was crippled.
> 
> The only dubious case is when next hop is set using routing tables.
> But code in ipgre_tunnel_xmit() is ready to accept this situation,
> it checks for zero destination address and fixes it when it is able to.

Nevertheless, it does not work.

The reason is that NOARP arp entries on device with initialized
hard_header are initialized not to all zeros, but to dev->dev_addr.
So that, netxthop from routing tables is ignored and all gre packets
are lost in loopback. Not good.

The problem can be ignored. I am even not sure that someone uses
this feature. Actually, it was not recommended in documentation.

Alternatively, arp.c can be changed to generate 0 addresses instead
of dev->dev_addr. Normally it is equally good, but I am not sure
about possible side effects.

Another thoughts?

Alexey

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Timo Teräs]

Hi,

Alexey Kuznetsov wrote:
>> The only dubious case is when next hop is set using routing tables.
>> But code in ipgre_tunnel_xmit() is ready to accept this situation,
>> it checks for zero destination address and fixes it when it is able to.
> 
> Nevertheless, it does not work.
> 
> Another thoughts?

Could you explain this a little more?

I was able to set a nbma gre tunnel, add routes to it and it worked
perfectly ok.

Link-level next hop worked:
ip route add <route> via <link-level-address> dev <tunnel-dev> onlink

Also normal route with static neighbor worked:
ip neigh add <router-tunnel-ip> lladdr <link-level-address> nud permanent dev <tunnel-dev>
ip route add <route> via <router-tunnel-ip>

In both cases packets were routed as expected to hosts accessible using
the added route. And sendto() sent a packet correctly with destination
link-level-address set. Did I miss/misunderstand something?

Cheers,
  Timo

Re: [PATCH RESEND] ip_gre: sendto/recvfrom NBMA address

[Alexey Kuznetsov]

Hello!

> I was able to set a nbma gre tunnel, add routes to it and it worked
> perfectly ok.
> 
> Link-level next hop worked:
> ip route add <route> via <link-level-address> dev <tunnel-dev> onlink

This can work if you use gre0. By plain luck it has all-zero dev_addr.

It will break on nbma devices set with:

ip tunnel add XXX mode gre local a.b.c.d [key whatever] ...

Alexey

[RFC][PATCH][IPV4] ip_gre: use skb->{mac,network}_header consistently

[Timo Teräs]

From: Timo Teras <timo.teras@iki.fi>

Make both send and receive paths use consistently skb->*_header:
  mac_header = outer IPv4 header
  network_header = encapsulated packet (e.g. inner IPv4 header)

Signed-off-by: Timo Teras <timo.teras@iki.fi>
---
Currently in send path:
  network_header = outer IPv4 header
  transport_header = encapsulated packet
And receive path:
  network_header = encapsulated packet
  mac_header = GRE header

mac_header used to be the outer IPv4 header on receive path, but it was
broken/changed when skb_reset_mac_header() was introduced.

This also makes ipgre_header_parse() work correctly as it assumes that
mac_header points to the outer IPv4 header always. Though, it returns
zeros for outgoing packets that are tcpdumped from gre tunnel without
explicit local address, since the routing is done in ipgre_tunnel_xmit().

 net/ipv4/ip_gre.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index 02b02a8..eba1ade 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -613,7 +613,7 @@ static int ipgre_rcv(struct sk_buff *skb)
 				offset += 4;
 		}
 
-		skb_reset_mac_header(skb);
+		skb->mac_header = skb->network_header;
 		__pskb_pull(skb, offset);
 		skb_reset_network_header(skb);
 		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
@@ -826,9 +826,8 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 		old_iph = ip_hdr(skb);
 	}
 
-	skb->transport_header = skb->network_header;
 	skb_push(skb, gre_hlen);
-	skb_reset_network_header(skb);
+	skb_reset_mac_header(skb);
 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
 			      IPSKB_REROUTED);
@@ -839,7 +838,7 @@ static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 	 *	Push down and install the IPIP header.
 	 */
 
-	iph 			=	ip_hdr(skb);
+	iph 			=	(struct iphdr *) skb_mac_header(skb);
 	iph->version		=	4;
 	iph->ihl		=	sizeof(struct iphdr) >> 2;
 	iph->frag_off		=	df;
@@ -1070,6 +1069,7 @@ static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
 	struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen);
 	__be16 *p = (__be16*)(iph+1);
 
+	skb_reset_mac_header(skb);
 	memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
 	p[0]		= t->parms.o_flags;
 	p[1]		= htons(type);
-- 
1.5.2.5

[PATCH][IPV4] ip_gre: set mac_header correctly in receive path

[Timo Teräs]

From: Timo Teras <timo.teras@iki.fi>

mac_header update in ipgre_recv() was incorrectly changed to
skb_reset_mac_header() when it was introduced.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
---
This replaces my earlier patch titled "ip_gre: use skb->{mac,
network}_header consistently". Apparently I hadn't done my homework how
to use *_header correctly. And I should have done a bit more testing to
figure out the previous patch does not work.

But the main problem was the receive path in the first place, and this
patch fixes it.

The bug was introduced in commit 459a98ed881802dee55897441bc7f77af614368e.
There might be other similar incorrect replaces.


 net/ipv4/ip_gre.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index 02b02a8..4b93f32 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -613,7 +613,7 @@ static int ipgre_rcv(struct sk_buff *skb)
 				offset += 4;
 		}
 
-		skb_reset_mac_header(skb);
+		skb->mac_header = skb->network_header;
 		__pskb_pull(skb, offset);
 		skb_reset_network_header(skb);
 		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
-- 
1.5.2.5

Re: [PATCH][IPV4] ip_gre: set mac_header correctly in receive path

[David Miller]

From: Timo_Teräs <timo.teras@iki.fi>
Date: Wed, 19 Dec 2007 20:10:41 +0200

> From: Timo Teras <timo.teras@iki.fi>
> 
> mac_header update in ipgre_recv() was incorrectly changed to
> skb_reset_mac_header() when it was introduced.
> 
> Signed-off-by: Timo Teras <timo.teras@iki.fi>

Patch applied, thanks.

> ---
> This replaces my earlier patch titled "ip_gre: use skb->{mac,
> network}_header consistently". Apparently I hadn't done my homework how
> to use *_header correctly. And I should have done a bit more testing to
> figure out the previous patch does not work.
> 
> But the main problem was the receive path in the first place, and this
> patch fixes it.
> 
> The bug was introduced in commit 459a98ed881802dee55897441bc7f77af614368e.
> There might be other similar incorrect replaces.

That commit has two other identical bad conversions, I'll
fix them up.