802.1q + device removal causing hang

Hi,

In 2.4.20-pre6, 2.4.20-pre4, 2.4.19, and 2.4.18 at least, I have been having problems with vlan shutdown when removing tg3.o. rmmod hangs on rtnl_sem, call trace:

	rtnetlink_event
	unregister_netdev
	vlan_device_event
	notifier_call_chain
	unregister_netdevice
	unregister_netdev
	[tg3 module]
	[tg3 module]
	pci_unregister_driver
	...

The hang is happening at rmmod time. All other operation appears to be fine. If I manually make sure all vlans are removed first, everything works, so it's probably something to do with the NETDEV_UNREGISTER case in vlan_device_event() in vlan.c.

Steps to reproduce:

modprobe tg3
ifconfig eth0 up
vconfig add eth0 10
ifconfig eth0.10 1.2.3.4
ifconfig eth0.10 down
ifconfig eth0 down
rmmod tg3

Note that the IP address stays assigned even with eth0.10 set down. Everything works if the IP assignment is not made. Tested on 2.4.20-pre6.

Does anybody see what would be causing this (perhaps unregister_netdev reentering)?

(Also, why is the "ifconfig eth0 up" necessary for vconfig? It allows "ifconfig eth0 down" _after_ the vlan is created, and vlans can still be removed (but not created) at that point.)

Simon-

[ Stormix Technologies Inc. ][ NetNation Communications Inc. ]
[ [email protected] ][ [email protected] ]
[ Opinions expressed are not necessarily those of my employers. ]

Try this:

@@ -626,7 +626,7 @@
 			ret = unregister_vlan_dev(dev,
 						  VLAN_DEV_INFO(vlandev)->vlan_id);
 
-			unregister_netdev(vlandev);
+			unregister_netdevice(vlandev);
 
 			/* Group was destroyed? */
 			if (ret == 1)

Hello David, list,

Ebtables is a project similar to iptables, but working on the bridge netfilter hooks. It allows for a basic transparent firewall, making a brouter and doing MAC source address and destination address manipulation. The firewall part has currently modules for basic IP filtering, 802.1q filtering, ARP filtering, logging and a mark match/target.
Ebtables has been under development for over 1.5 year and has more than 100 users, I think.

The patch is 3662 lines long, so I won't list it in this mail. It is available at:
http://users.pandora.be/bart.de.schuymer/ebtables/v2.0/ebtables-v2.0_vs_2.5.34.diff
or, gzipped:
http://users.pandora.be/bart.de.schuymer/ebtables/v2.0/ebtables-v2.0_vs_2.5.34.diff.gz

It is vs 2.5.34, I can make a patch vs 2.4.x when the time is right. Comments/questions are appreciated.

For more information, see
http://users.pandora.be/bart.de.schuymer/ebtables/
There is an ebtables hacking howto, some basic examples and some real life examples from users. And ofcourse the userspace program.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Thu, 12 Sep 2002 08:36:52 +0200

ARP filtering

People should use ARP tables for arp filtering, that is why I wrote it. ARP filtering should not need to be bridge specific.

Next, has Lennert Buytenhek, the bridging maintainer, approved of your changes to the bridging layer APIs?

Hello David, Lennert, list,

Well, a bridge can also just _bridge_ ARP packets between two sides of the bridge. The ARP module can filter out those packets. These packets will not pass through the ARP code of the Linux kernel. Ofcourse, the ebtables ARP module can be easily adjusted for arptables, I will do this later if nobody beats me to it... For the same reason, basic ebtables IP filtering is not redundant.

OK. This is to Lennert:
Could you please have a look at the ebtables patch located at

http://users.pandora.be/bart.de.schuymer/ebtables/v2.0/ebtables-v2.0_vs_2.5.34.diff

and approve the changes made to the bridging layer API? They are necessary to make a brouter and to deal with bogus NETFILTER_DEBUG warnings if the option is compiled in the kernel. Any questions will be gladly answered... Note that the brouting facility has been working for atleast three months and it has already been used in real-life situations, there's an example usage on the ebtables homepage. Dealing with NETFILTER_DEBUG warnings consists of setting nf_debug to zero when the netfilter hooks change from bridge hooks to some other stack's hooks and vice versa. See the patch.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Fri, 13 Sep 2002 05:20:41 +0200

Well, a bridge can also just _bridge_ ARP packets between two sides of the bridge. The ARP module can filter out those packets. These packets will not pass through the ARP code of the Linux kernel. Ofcourse, the ebtables ARP module can be easily adjusted for arptables

No, I think I understand the difference and why you're problem space does not intersect what arptables handles.

It may not be nice that we can't immediately just reuse ipv4/netfilter handlers for bridging, but I'm not going to require that you make that work before I'll accept your patch.

Once you work things out with Lennert and he approves the changes, I'll apply your patch.

If you mean using iptables on bridged packets, this is possible with the br-nf patch. It is not trivial however, 2 new fields to the sk_buff need to be added, a small change in the IP fragment code and a small change in ip_tables.c, a change to netfilter.h and netfilter.c. The br-nf patch has been under development for over 1.5 years, most work done by Lennert with me helping now and then...
So, if you would accept br-nf, that would be great. I think Lennert and me would need some time to agree on a few things before submitting anything, however...

Cool, looking forward to his comments.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Fri, 13 Sep 2002 08:12:27 +0200

It is not trivial however, 2 new fields to the sk_buff need to be added, a small change in the IP fragment code and a small change in ip_tables.c, a change to netfilter.h and netfilter.c.

I've seen these changes, they are very buggy. The IPv4 copies added are just ugly and are buggy too, they potentially copy past the end of the packet buffer.

So, if you would accept br-nf, that would be great.

You need to remove the IPv4 bits, that copy of the MAC has to happen at a different layer, it does not belong in IPv4. At best, everyone shouldn't eat that header copy.

Franks a lot,
David S. Miller
[email protected]

You mean this part? This is the only copy added to generic code that I can find.

@@ -883,6 +885,7 @@
                iph->tot_len = htons(len + hlen);
 
                ip_send_check(iph);
+               memcpy(skb2->data - 16, skb->data - 16, 16);
 
                err = output(skb2);
                if (err)

If this code is buggy, isn't the following bit from ip_output.c buggy too? (around line 170)

        if (hh) {
                read_lock_bh(&hh->hh_lock);

                read_unlock_bh(&hh->hh_lock);
                skb_push(skb, hh->hh_len);
                return hh->hh_output(skb);
        } else if (dst->neighbour)

What if I make the memcpy conditional on "if (skb->physindev != NULL)"?

cheers,
Lennert

From: Lennert Buytenhek <[email protected]>
Date: Fri, 13 Sep 2002 14:45:18 +0200

What if I make the memcpy conditional on "if (skb->physindev != NULL)"?

First explain to me why the copy is needed for.

memcpy(skb2->data - 16, skb->data - 16, 16);

This is for purely bridged packets.
IP connection tracking first gathers all fragments, before the bridged IP packets are sent out the packet is fragmented. However, since this fragmenting actually happens while the IP packet "is in the bridge code", no existing code makes sure the fragments' Ethernet headers are correctly filled. Now, AFAIK an Ethernet header has length 14 bytes, so don't ask me why the magic number 16 is used (Lennert got IP fragmenting fixed). A nice comment in front of that copy is certainly needed...

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Sat, 14 Sep 2002 09:05:40 +0200

   memcpy(skb2->data - 16, skb->data - 16, 16);

This is for purely bridged packets.

Why is it being added, therefore, to ip_queue_xmit() which is only ever invoked by TCP output processing?

If the patch adds the call somewhere else, please correct me, but I specifically remember it being added to ip_queue_xmit() which is why I barfed when seeing it :-)

I've never seen this in the patch. It sure isn't in it now.

To be more precise:
net/ipv4/netfilter/ip_conntrack_standalone.c:ip_refrag() is (or can be) attached to the NF_IP_POST_ROUTING hook. This function calls:
net/ipv4/ip_output.c:ip_fragment()
In this function the copy of the Ethernet frame is added for each fragment (by the br-nf patch).
The bridge-netfilter patch lets IP packets/frames passing the NF_BR_POST_ROUTING hook go through the NF_IP_POST_ROUTING hook, so the ip_fragment() code is executed while the IP packet/frame is really in the bridge code. After this, the fragments get queued:
net/bridge/br_forward.c:br_dev_queue_push_xmit() calls dev_queue_xmit()

Lennert's previous mail says in which cases and why this header copy has to be explicitly done.

The following document might be useful to know what we are doing:
http://users.pandora.be/bart.de.schuymer/ebtables/br_fw_ia/br_fw_ia.html

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Mon, 16 Sep 2002 23:41:17 +0200

net/ipv4/ip_output.c:ip_fragment()
In this function the copy of the Ethernet frame is added for each fragment (by the br-nf patch).

'output' callback arg to ip_fragment() must generate correct hardware headers when necessary. This hack usage of it via netfilter, in this weird bridging case, is violating this requirement.

Normally ip_finish_output2() is going to make this.

If it can't do the job properly, pass instead a routine that can do what netfilter needs.

Lennert says:

So if we postpone FORWARD and POST_ROUTING until after br_dev_xmit, we effectively reverse refragmentation and neighbor resolution. But refragmentation messes up the hardware header.

The 16byte hardware header copy fixes this by copying to each fragment the hardware header that was tacked onto or was already present on the bigger packet. It's ugly, I admit. There's currently no better way though.

I don't understand why you can't add on the hardware header some other way.

If ip_finish_output doesn't put the right hardware header on there, you have to use as 'okfn' (what netfilter sends down as 'output' to ip_fragment) some routine which will do it correctly.

Aha. In our case, the output function is net/bridge/br_forward.c:__dev_queue_push_xmit(). This is because __br_forward_finish() (same file) uses this as okfn. Remember the IP hooks are "faked" on the bridge hooks, so functions attached to NF_IP_POST_ROUTING are called when the IP packet/frame passes the NF_BR_POST_ROUTING hook. They are not called earlier. All of this assuming that the destination device according to the routing table is a (logical) bridge device. If not, the packets go through the IP code and netfilter hooks normally.

So, what if we were to add the following code to the start of __dev_queue_push_xmit():

	if (skb->protocol == __constant_htons(ETH_P_IP)) {
		struct dst_entry *dst = skb->dst;
		if (hh) {
			read_lock_bh(&hh->hh_lock);
  			memcpy(skb->data - 16, hh->hh_data, 16);
			read_unlock_bh(&hh->hh_lock);
		}
	}

hh being NULL for an unfragmented IP packet and else non-NULL? Do realize that I (I can't speak for Lennert ofcourse) am not very familiar to the workings of the IP code.

Then we can remove the memcpy from ip_fragment(). Does that make sense?

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Tue, 17 Sep 2002 21:10:06 +0200

Then we can remove the memcpy from ip_fragment(). Does that make sense?

That sounds like the kind of solution I'd like used.

This is just to elaborate upon what Bart said earlier.

In the "L2 switched frame" case, we have a bit of a nasty problem with IP fragmentation. And in the "L3 'switched' frame" case (brouted frame), we have an ordering problem with IP fragmentation and neighbor resolution.

This is what the call stack looks like when we have a purely bridged frame (that needs to be netfiltered):

	net_rx_action
	  -> br_handle_frame
	    -> PF_BRIDGE/PRE_ROUTING
	      -> br_handle_frame_finish
		-> br_forward
		  -> PF_BRIDGE/FORWARD
		    -> __br_forward_finish
		      -> PF_BRIDGE/POST_ROUTING
			-> dev_queue_xmit

This case is easy to see. With ip_conntrack enabled, packets are reassembled in PRE_ROUTING and refragmented in POST_ROUTING. This refragmenting messes up the hardware header, so the fragments will leave the box with incorrect HW headers.

The broute case is a bit harder to see. If L3 (routed) packets are destined for a bridge device, we don't know what subdevice (slave port) they will go to until the bridge layer's br_dev_xmit has its way. However, we would like to be able to use the real outgoing interface (physoutdev) in FORWARD and POST_ROUTING.

To be able to do this, we postpone calling IPv4/FORWARD and IPv4/POST_ROUTING until after PF_BRIDGE/POST_ROUTING has happened, because at that point we know physoutdev so we can feed it to said IPv4 hooks.

But. Packet refragmentation normally happens in IPv4/POST_ROUTING. We don't want to do it there though, because that would cause the eventual call to IPv4/FORWARD and IPv4/POST_ROUTING to see all fragments instead of one packet (which goes against the idea of conntrack).

So if we postpone FORWARD and POST_ROUTING until after br_dev_xmit, we effectively reverse refragmentation and neighbor resolution. But refragmentation messes up the hardware header.

The 16byte hardware header copy fixes this by copying to each fragment the hardware header that was tacked onto or was already present on the bigger packet. It's ugly, I admit. There's currently no better way though.

(And Bart, I chose 16 because 16-byte aligned 16-byte copies should be cheaper than 2-byte aligned 14-byte copies, and there should be at least 16 bytes before skb->data at this point anyway. That is, if I understood the code correctly.)

cheers,
Lennert

Hello David,

The following link points to the ebtables patch approved by Lennert:
http://users.pandora.be/bart.de.schuymer/ebtables/v2.0/ebtables-v2.0_vs_2.5.35-try3.diff

Changes:

cleanup brouter interface in the bridge code + brouter bugfix.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Mon, 16 Sep 2002 08:50:16 +0200

The following link points to the ebtables patch approved by Lennert:
http://users.pandora.be/bart.de.schuymer/ebtables/v2.0/ebtables-v2.0_vs_2.5.35-try3.diff

I'm applying this, but please in the future do not put these "*** deal with blah blah" seperators in one big patch file.

Give me one bug patch I can just do a clean "patch -p1 <your_patch" with, thanks.

Sure it's just noise and patch perhaps knows how to skip over it, but lots of other patch reading tools might not be able to. (for example diffstat gives a lot of different paths as the patch base directory because you've done the patch this way).

Hello David, list,

David, you asked for a patch for mapping the IPv4 hooks onto the bridge hooks, this is the patch.

Comments and suggestions for better solutions are welcome.

I don't know how to (easily) move the copy of the Ethernet header out of ip_output.c::ip_fragment.c. The patch adds 3 members to the skbuff...

There is a little text file explaining the source code more in-depth here:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre1-against-2.5.39-comments.txt
A high-level explanation of what we're doing is here:
http://users.pandora.be/bart.de.schuymer/ebtables/br_fw_ia/br_fw_ia.html
The patch is also available at:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre1-against-2.5.42.diff

Note that Lennert has not yet responded to my private RFC to him, but Halloween is coming, so we should at least try to get this into 2.5...

cheers,
Bart

Here's the patch:

@@ -117,17 +117,23 @@
 /* This is gross, but inline doesn't cut it for avoiding the function
    call in fast path: gcc doesn't inline (needs value tracking?). --RR */
 #ifdef CONFIG_NETFILTER_DEBUG
-#define NF_HOOK nf_hook_slow
+#define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
+ nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN)
+#define NF_HOOK_THRESH nf_hook_slow
 #else
 #define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
 (list_empty(&nf_hooks[(pf)][(hook)])					\
  ? (okfn)(skb)								\
- : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn)))
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN))
+#define NF_HOOK_THRESH(pf, hook, skb, indev, outdev, okfn, thresh)	\
+(list_empty(&nf_hooks[(pf)][(hook)])					\
+ ? (okfn)(skb)								\
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), (thresh)))
 #endif
 
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev, struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *));
+		 int (*okfn)(struct sk_buff *), int thresh);
 
 /* Call setsockopt() */
 int nf_setsockopt(struct sock *sk, int pf, int optval, char *opt, 

@@ -52,8 +52,10 @@
 enum nf_ip_hook_priorities {
 	NF_IP_PRI_FIRST = INT_MIN,
 	NF_IP_PRI_CONNTRACK = -200,
+	NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD = -175,
 	NF_IP_PRI_MANGLE = -150,
 	NF_IP_PRI_NAT_DST = -100,
+	NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT = -50,
 	NF_IP_PRI_FILTER = 0,
 	NF_IP_PRI_NAT_SRC = 100,
 	NF_IP_PRI_LAST = INT_MAX,

--- linux-2.5.42/include/linux/netfilter_bridge.h Sat Oct 12 06:22:09 2002
+++ linux-2.5.42-brnf/include/linux/netfilter_bridge.h Sun Oct 13 11:45:19
2002
@@ -22,14 +22,27 @@
#define NF_BR_BROUTING 5
#define NF_BR_NUMHOOKS 6

+/* Masks for skb->brnfmask */
+#define BRNF_PKT_TYPE 0x01
+#define BRNF_BRIDGED_DNAT 0x02
+#define BRNF_COPY_HEADER 0x04
+#define BRNF_DONT_TAKE_PARENT 0x08
+
enum nf_br_hook_priorities {
NF_BR_PRI_FIRST = INT_MIN,
- NF_BR_PRI_FILTER_BRIDGED = -200,
- NF_BR_PRI_FILTER_OTHER = 200,
NF_BR_PRI_NAT_DST_BRIDGED = -300,
+ NF_BR_PRI_FILTER_BRIDGED = -200,
+ NF_BR_PRI_BRNF = 0,
NF_BR_PRI_NAT_DST_OTHER = 100,
+ NF_BR_PRI_FILTER_OTHER = 200,
NF_BR_PRI_NAT_SRC = 300,
NF_BR_PRI_LAST = INT_MAX,
};

+/* Used in br_netfilter.c */
+struct bridge_skb_cb {
+ union {
+ __u32 ipv4;
+ } daddr;
+};
#endif

@@ -140,6 +140,8 @@
  *	@sk: Socket we are owned by
  *	@stamp: Time we arrived
  *	@dev: Device we arrived on/are leaving by
+ *	@physindev: Physical device we arrived on - see br_netfilter.c
+ *	@physoutdev: Phsical device we will leave by - see br_netfilter.c
  *	@h: Transport layer header
  *	@nh: Network layer header
  *	@mac: Link layer header

@@ -166,6 +168,7 @@
  *	@nfcache: Cache info
  *	@nfct: Associated connection, if any
  *	@nf_debug: Netfilter debugging
+ *	@brnfmask: Info about a bridged frame - see br_netfilter.c
  *	@tc_index: Traffic control index
  */
 

@@ -178,6 +181,8 @@
 	struct sock		*sk;
 	struct timeval		stamp;
 	struct net_device	*dev;
+	struct net_device	*physindev;
+	struct net_device	*physoutdev;
 
 	union {
 		struct tcphdr	*th;

@@ -236,6 +241,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
         unsigned int		nf_debug;
 #endif
+	unsigned int		brnfmask;
 #endif /* CONFIG_NETFILTER */
 #if defined(CONFIG_HIPPI)
 	union {

@@ -45,6 +45,8 @@
 {
 	printk(KERN_INFO "NET4: Ethernet Bridge 008 for NET4.0\n");
 
+	if (br_netfilter_init())
+		return 1;
 	br_handle_frame_hook = br_handle_frame;
 	br_ioctl_hook = br_ioctl_deviceless_stub;
 #if defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE)

@@ -63,6 +65,7 @@
 
 static void __exit br_deinit(void)
 {
+	br_netfilter_fini();
 	unregister_netdevice_notifier(&br_device_notifier);
 	br_call_ioctl_atomic(__br_clear_ioctl_hook);
 

@@ -30,7 +30,7 @@
 	return 1;
 }
 
-static int __dev_queue_push_xmit(struct sk_buff *skb)
+int br_dev_queue_push_xmit(struct sk_buff *skb)
 {
 	skb_push(skb, ETH_HLEN);
 	dev_queue_xmit(skb);

@@ -38,10 +38,10 @@
 	return 0;
 }
 
-static int __br_forward_finish(struct sk_buff *skb)
+int br_forward_finish(struct sk_buff *skb)
 {
 	NF_HOOK(PF_BRIDGE, NF_BR_POST_ROUTING, skb, NULL, skb->dev,
-			__dev_queue_push_xmit);
+			br_dev_queue_push_xmit);
 
 	return 0;
 }

@@ -53,7 +53,7 @@
 	skb->nf_debug = 0;
 #endif
 	NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 static void __br_forward(struct net_bridge_port *to, struct sk_buff *skb)

@@ -64,7 +64,7 @@
 	skb->dev = to->dev;
 
 	NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, indev, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 /* called under bridge lock */

@@ -49,7 +49,7 @@
 			br_pass_frame_up_finish);
 }
 
-static int br_handle_frame_finish(struct sk_buff *skb)
+int br_handle_frame_finish(struct sk_buff *skb)
 {
 	struct net_bridge *br;
 	unsigned char *dest;

@@ -144,8 +144,10 @@
 /* br_forward.c */
 extern void br_deliver(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_dev_queue_push_xmit(struct sk_buff *skb);
 extern void br_forward(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_forward_finish(struct sk_buff *skb);
 extern void br_flood_deliver(struct net_bridge *br,
 		      struct sk_buff *skb,
 		      int clone);

@@ -166,6 +168,7 @@
 			   int *ifindices);
 
 /* br_input.c */
+extern int br_handle_frame_finish(struct sk_buff *skb);
 extern int br_handle_frame(struct sk_buff *skb);
 
 /* br_ioctl.c */

@@ -176,6 +179,10 @@
 	     unsigned long arg1,
 	     unsigned long arg2);
 extern int br_ioctl_deviceless_stub(unsigned long arg);
+
+/* br_netfilter.c */
+extern int br_netfilter_init(void);
+extern void br_netfilter_fini(void);
 
 /* br_stp.c */
 extern int br_is_root_bridge(struct net_bridge *br);

@@ -9,6 +9,11 @@
 bridge-objs	:= br.o br_device.o br_fdb.o br_forward.o br_if.o br_input.o \
 			br_ioctl.o br_notify.o br_stp.o br_stp_bpdu.o \
 			br_stp_if.o br_stp_timer.o
+
+ifeq ($(CONFIG_NETFILTER),y)
+bridge-objs	+= br_netfilter.o
+endif
+
 obj-$(CONFIG_BRIDGE_NF_EBTABLES) += netfilter/
 
 include $(TOPDIR)/Rules.make

@@ -342,10 +342,15 @@
 			       const struct net_device *indev,
 			       const struct net_device *outdev,
 			       struct list_head **i,
-			       int (*okfn)(struct sk_buff *))
+			       int (*okfn)(struct sk_buff *),
+			       int hook_thresh)
 {
 	for (*i = (*i)->next; *i != head; *i = (*i)->next) {
 		struct nf_hook_ops *elem = (struct nf_hook_ops *)*i;
+
+		if (hook_thresh > elem->priority)
+			continue;
+
 		switch (elem->hook(hook, skb, indev, outdev, okfn)) {
 		case NF_QUEUE:
 			return NF_QUEUE;

@@ -413,6 +418,8 @@
 {
 	int status;
 	struct nf_info *info;
+	struct net_device *physindev;
+	struct net_device *physoutdev;
 
 	if (!queue_handler[pf].outfn) {
 		kfree_skb(skb);

@@ -435,11 +442,16 @@
 	if (indev) dev_hold(indev);
 	if (outdev) dev_hold(outdev);
 
+	if ((physindev = skb->physindev)) dev_hold(physindev);
+	if ((physoutdev = skb->physoutdev)) dev_hold(physoutdev);
+
 	status = queue_handler[pf].outfn(skb, info, queue_handler[pf].data);
 	if (status < 0) {
 		/* James M doesn't say fuck enough. */
 		if (indev) dev_put(indev);
 		if (outdev) dev_put(outdev);
+		if (physindev) dev_put(physindev);
+		if (physoutdev) dev_put(physoutdev);
 		kfree(info);
 		kfree_skb(skb);
 		return;

@@ -449,7 +461,8 @@
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev,
 		 struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *))
+		 int (*okfn)(struct sk_buff *),
+		 int hook_thresh)
 {
 	struct list_head *elem;
 	unsigned int verdict;

@@ -481,7 +494,7 @@
 
 	elem = &nf_hooks[pf][hook];
 	verdict = nf_iterate(&nf_hooks[pf][hook], &skb, hook, indev,
-			     outdev, &elem, okfn);
+			     outdev, &elem, okfn, hook_thresh);
 	if (verdict == NF_QUEUE) {
 		NFDEBUG("nf_hook: Verdict = QUEUE.\n");
 		nf_queue(skb, elem, pf, hook, indev, outdev, okfn);

@@ -530,7 +543,7 @@
 		verdict = nf_iterate(&nf_hooks[info->pf][info->hook],
 				     &skb, info->hook, 
 				     info->indev, info->outdev, &elem,
-				     info->okfn);
+				     info->okfn, INT_MIN);
 	}
 
 	switch (verdict) {

@@ -234,6 +234,8 @@
 	skb->sk		  = NULL;
 	skb->stamp.tv_sec = 0;	/* No idea about time */
 	skb->dev	  = NULL;
+	skb->physindev	  = NULL;
+	skb->physoutdev	  = NULL;
 	skb->dst	  = NULL;
 	memset(skb->cb, 0, sizeof(skb->cb));
 	skb->pkt_type	  = PACKET_HOST;	/* Default type */

@@ -248,6 +250,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	skb->nf_debug	  = 0;
 #endif
+	skb->brnfmask	  = 0;
 #endif
 #ifdef CONFIG_NET_SCHED
 	skb->tc_index	  = 0;

@@ -363,6 +366,8 @@
 	n->sk = NULL;
 	C(stamp);
 	C(dev);
+	C(physindev);
+	C(physoutdev);
 	C(h);
 	C(nh);
 	C(mac);

@@ -392,6 +397,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	C(nf_debug);
 #endif
+	C(brnfmask);
 #endif /*CONFIG_NETFILTER*/
 #if defined(CONFIG_HIPPI)
 	C(private);

@@ -418,6 +424,8 @@
 	new->list	= NULL;
 	new->sk		= NULL;
 	new->dev	= old->dev;
+	new->physindev	= old->physindev;
+	new->physoutdev	= old->physoutdev;
 	new->priority	= old->priority;
 	new->protocol	= old->protocol;
 	new->dst	= dst_clone(old->dst);

@@ -438,6 +446,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	new->nf_debug	= old->nf_debug;
 #endif
+	new->brnfmask	= old->brnfmask;
 #endif
 #ifdef CONFIG_NET_SCHED
 	new->tc_index	= old->tc_index;

@@ -75,6 +75,7 @@
 #include <net/inetpeer.h>
 #include <linux/igmp.h>
 #include <linux/netfilter_ipv4.h>
+#include <linux/netfilter_bridge.h>
 #include <linux/mroute.h>
 #include <linux/netlink.h>
 

@@ -908,6 +909,18 @@
 		iph->tot_len = htons(len + hlen);
 
 		ip_send_check(iph);
+
+		/*
+		 *	Fragments with a bridge device destination need
+		 *	to get the Ethernet header copied here, as
+		 *	br_dev_queue_push_xmit() can't do this.
+		 *	See net/bridge/br_netfilter.c
+		 */
+
+#ifdef CONFIG_NETFILTER
+		if (skb->brnfmask & BRNF_COPY_HEADER)
+			memcpy(skb2->data - 16, skb->data - 16, 16);
+#endif
 
 		err = output(skb2);
 		if (err)

@@ -121,12 +121,14 @@
 static inline int
 ip_packet_match(const struct iphdr *ip,
 		const char *indev,
+		const char *physindev,
 		const char *outdev,
+		const char *physoutdev,
 		const struct ipt_ip *ipinfo,
 		int isfrag)
 {
 	size_t i;
-	unsigned long ret;
+	unsigned long ret, ret2;
 
 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
 

@@ -156,7 +158,13 @@
 			& ((const unsigned long *)ipinfo->iniface_mask)[i];
 	}
 
-	if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
+	for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret2 |= (((const unsigned long *)physindev)[i]
+			^ ((const unsigned long *)ipinfo->iniface)[i])
+			& ((const unsigned long *)ipinfo->iniface_mask)[i];
+	}
+
+	if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
 			indev, ipinfo->iniface,
 			ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");

@@ -169,7 +177,13 @@
 			& ((const unsigned long *)ipinfo->outiface_mask)[i];
 	}
 
-	if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
+	for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret2 |= (((const unsigned long *)physoutdev)[i]
+			^ ((const unsigned long *)ipinfo->outiface)[i])
+			& ((const unsigned long *)ipinfo->outiface_mask)[i];
+	}
+
+	if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
 			outdev, ipinfo->outiface,
 			ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");

@@ -268,6 +282,7 @@
 	/* Initializing verdict to NF_DROP keeps gcc happy. */
 	unsigned int verdict = NF_DROP;
 	const char *indev, *outdev;
+	const char *physindev, *physoutdev;
 	void *table_base;
 	struct ipt_entry *e, *back;
 

@@ -277,6 +292,9 @@
 	datalen = (*pskb)->len - ip->ihl * 4;
 	indev = in ? in->name : nulldevname;
 	outdev = out ? out->name : nulldevname;
+	physindev = (*pskb)->physindev ? (*pskb)->physindev->name : nulldevname;
+	physoutdev = (*pskb)->physoutdev ? (*pskb)->physoutdev->name : nulldevname;
+
 	/* We handle fragments by dealing with the first fragment as
 	 * if it was a normal packet.  All other fragments are treated
 	 * normally, except that they will NEVER match rules that ask

@@ -311,7 +329,8 @@
 		IP_NF_ASSERT(e);
 		IP_NF_ASSERT(back);
 		(*pskb)->nfcache |= e->nfcache;
-		if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
+		if (ip_packet_match(ip, indev, physindev, outdev, physoutdev,
+		    &e->ip, offset)) {
 			struct ipt_entry_target *t;
 
 			if (IPT_MATCH_ITERATE(e, do_match,

@@ -285,10 +285,13 @@
 	level_string[1] = '0' + (loginfo->level % 8);
 	spin_lock_bh(&log_lock);
 	printk(level_string);
-	printk("%sIN=%s OUT=%s ",
-	       loginfo->prefix,
-	       in ? in->name : "",
-	       out ? out->name : "");
+	printk("%sIN=%s ", loginfo->prefix, in ? in->name : "");
+	if ((*pskb)->physindev && in != (*pskb)->physindev)
+		printk("PHYSIN=%s ", (*pskb)->physindev->name);
+	printk("OUT=%s ", out ? out->name : "");
+	if ((*pskb)->physoutdev && out != (*pskb)->physoutdev)
+		printk("PHYSOUT=%s ", (*pskb)->physoutdev->name);
+
 	if (in && !out) {
 		/* MAC logging for input chain only. */
 		printk("MAC=");

**pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct iphdr *iph;
+ __u32 len;
+ struct sk_buff *skb;
+
+ if ((*pskb)->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ if ((skb = skb_share_check(*pskb, GFP_ATOMIC)) == NULL)
+ goto out;
+
+ if (!pskb_may_pull(skb, sizeof(struct iphdr)))
+ goto inhdr_error;
+
+ iph = skb->nh.iph;
+ if (iph->ihl < 5 || iph->version != 4)
+ goto inhdr_error;
+
+ if (!pskb_may_pull(skb, 4*iph->ihl))
+ goto inhdr_error;
+
+ iph = skb->nh.iph;
+ if (ip_fast_csum((__u8 *)iph, iph->ihl) != 0)
+ goto inhdr_error;
+
+ len = ntohs(iph->tot_len);
+ if (skb->len < len || len < 4*iph->ihl)
+ goto inhdr_error;
+
+ if (skb->len > len) {
+ __pskb_trim(skb, len);
+ if (skb->ip_summed == CHECKSUM_HW)
+ skb->ip_summed = CHECKSUM_NONE;
+ }
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_PRE_ROUTING);
+#endif
+
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ skb->brnfmask |= BRNF_PKT_TYPE;
+ }
+
+ skb->physindev = skb->dev;
+ skb->dev = bridge_parent(skb->dev);
+ store_orig_dstaddr(skb);
+
+ NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, skb->dev, NULL,
+ br_nf_pre_routing_finish);
+
+ return NF_STOLEN;
+
+inhdr_error:
+// IP_INC_STATS_BH(IpInHdrErrors);
+out:
+ return NF_DROP;
+}
+
+
+/* PF_BRIDGE/LOCAL_IN ************************************************/ +/* The packet is locally destined, which requires a real
+ * dst_entry, so detach the fake one. On the way up, the
+ * packet would pass through PRE_ROUTING again (which already
+ * took place when the packet entered the bridge), but we
+ * register an IPv4 PRE_ROUTING 'sabotage' hook that will
+ * prevent this from happening.
+ */
+static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ if (skb->dst == (struct dst_entry *)&__fake_rtable) {
+ dst_release(skb->dst);
+ skb->dst = NULL;
+ }
+
+ return NF_ACCEPT;
+}
+
+
+/* PF_BRIDGE/FORWARD *************************************************/ +static int br_nf_forward_finish(struct sk_buff *skb)
+{
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_BR_FORWARD);
+#endif
+
+ if (skb->brnfmask & BRNF_PKT_TYPE) {
+ skb->pkt_type = PACKET_OTHERHOST;
+ skb->brnfmask ^= BRNF_PKT_TYPE;
+ }
+
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_FORWARD, skb, skb->physindev,
+ skb->dev, br_forward_finish, 1);
+
+ return 0;
+}
+
+/* This is the 'purely bridged' case. We pass the packet to
+ * netfilter with indev and outdev set to the bridge device,
+ * but we are still able to filter on the 'real' indev/outdev
+ * because another bit of the bridge-nf patch overloads the
+ * '-i' and '-o' iptables interface checks to take
+ * skb->phys{in,out}dev into account as well (so both the real
+ * device and the bridge device will match).
+ */
+static unsigned int br_nf_forward(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_BR_FORWARD);
+#endif
+
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ skb->brnfmask |= BRNF_PKT_TYPE;
+ }
+
+ skb->physoutdev = skb->dev;
+
+ NF_HOOK(PF_INET, NF_IP_FORWARD, skb, bridge_parent(skb->physindev),
+ bridge_parent(skb->dev), br_nf_forward_finish);
+
+ return NF_STOLEN;
+}
+
+
+/* PF_BRIDGE/LOCAL_OUT ***********************************************/ +static int br_nf_local_out_finish(struct sk_buff *skb)
+{
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug &= ~(1 << NF_BR_LOCAL_OUT);
+#endif
+
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
+ br_forward_finish, NF_BR_PRI_FIRST + 1);
+
+ return 0;
+}
+
+
+/* This function sees both locally originated IP packets and forwarded
+ * IP packets (in both cases the destination device is a bridge
+ * device). It also sees bridged-and-DNAT'ed packets.
+ * For the sake of interface transparency (i.e. properly
+ * overloading the '-o' option), we steal packets destined to
+ * a bridge device away from the PF_INET/FORWARD and PF_INET/OUTPUT hook
+ * functions, and give them back later, when we have determined the real
+ * output device. This is done in here.
+ *
+ * If (skb->brnfmask & BRNF_BRIDGED_DNAT) then the packet is bridged
+ * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
+ * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
+ * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
+ * will be executed.
+ * Otherwise, if skb->physindev is NULL, the bridge-nf code never touched
+ * this packet before, and so the packet was locally originated. We fake
+ * the PF_INET/LOCAL_OUT hook.
+ * Finally, if skb->physindev isn't NULL, then the packet was IP routed,
+ * so we fake the PF_INET/FORWARD hook. ipv4_sabotage_out() makes sure
+ * even routed packets that didn't arrive on a bridge interface have their
+ * skb->physindev set.
+ */
+
+static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*_okfn)(struct sk_buff *))
+{
+ int (*okfn)(struct sk_buff *skb);
+ struct net_device *realindev;
+ struct sk_buff *skb = *pskb;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ /* Sometimes we get packets with NULL ->dst here (for example,
+ * running a dhcp client daemon triggers this).
+ */
+ if (skb->dst == NULL)
+ return NF_ACCEPT;
+
+ skb->physoutdev = skb->dev;
+
+ realindev = skb->physindev;
+
+ /* Bridged, take PF_BRIDGE/FORWARD.
+ * (see big note in front of br_nf_pre_routing_finish)
+ */
+ if (skb->brnfmask & BRNF_BRIDGED_DNAT) {
+ okfn = br_forward_finish;
+
+ if (skb->brnfmask & BRNF_PKT_TYPE) {
+ skb->pkt_type = PACKET_OTHERHOST;
+ skb->brnfmask ^= BRNF_PKT_TYPE;
+ }
+
+ NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, realindev,
+ skb->dev, okfn);
+ } else {
+ okfn = br_nf_local_out_finish;
+ /* IP forwarded traffic has a physindev, locally
+ * generated traffic hasn't.
+ */
+ if (realindev != NULL) {
+ if (((skb->brnfmask & BRNF_DONT_TAKE_PARENT) == 0) &&
+ has_bridge_parent(realindev))
+ realindev = bridge_parent(realindev);
+
+ NF_HOOK_THRESH(PF_INET, NF_IP_FORWARD, skb, realindev,
+ bridge_parent(skb->dev), okfn,
+ NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD + 1);
+ } else {
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_LOCAL_OUT);
+#endif
+
+ NF_HOOK_THRESH(PF_INET, NF_IP_LOCAL_OUT, skb, realindev,
+ bridge_parent(skb->dev), okfn,
+ NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT + 1);
+ }
+ }
+
+ return NF_STOLEN;
+}
+
+
+/* PF_BRIDGE/POST_ROUTING ********************************************/ +static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+
+ /* Be very paranoid. */
+ if (skb->mac.raw < skb->head || skb->mac.raw + ETH_HLEN > skb->data) {
+ printk(KERN_CRIT "br_netfilter: Argh!! br_nf_post_routing: "
+ "bad mac.raw pointer.");
+ if (skb->dev != NULL) {
+ printk("[%s]", skb->dev->name);
+ if (has_bridge_parent(skb->dev))
+ printk("[%s]", bridge_parent(skb->dev)->name);
+ }
+ printk("\n");
+ return NF_ACCEPT;
+ }
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ /* Sometimes we get packets with NULL ->dst here (for example,
+ * running a dhcp client daemon triggers this).
+ */
+ if (skb->dst == NULL)
+ return NF_ACCEPT;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_POST_ROUTING);
+#endif
+
+ /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
+ * about the value of skb->pkt_type.
+ */
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ skb->brnfmask |= BRNF_PKT_TYPE;
+ }
+
+ /* Fragmented packets need a good Ethernet header, tell this to
+ * ip_output.c::ip_fragment().
+ */
+ skb->brnfmask |= BRNF_COPY_HEADER;
+
+ NF_HOOK(PF_INET, NF_IP_POST_ROUTING, skb, NULL,
+ bridge_parent(skb->dev), br_dev_queue_push_xmit);
+
+ return NF_STOLEN;
+}
+
+
+/* IPv4/SABOTAGE *****************************************************/ +
+/* Don't hand locally destined packets to PF_INET/PRE_ROUTING
+ * for the second time.
+ */
+static unsigned int ipv4_sabotage_in(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ if (in->hard_start_xmit == br_dev_xmit &&
+ okfn != br_nf_pre_routing_finish) {
+ okfn(*pskb);
+ return NF_STOLEN;
+ }
+
+ return NF_ACCEPT;
+}
+
+/* Postpone execution of PF_INET/FORWARD, PF_INET/LOCAL_OUT
+ * and PF_INET/POST_ROUTING until we have done the forwarding
+ * decision in the bridge code and have determined skb->physoutdev.
+ */
+static unsigned int ipv4_sabotage_out(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ if (out->hard_start_xmit == br_dev_xmit &&

+	    okfn != br_nf_forward_finish &&
+	    okfn != br_nf_local_out_finish &&
+	    okfn != br_dev_queue_push_xmit) {

+ struct sk_buff *skb = *pskb;
+
+ /* This frame will arrive on PF_BRIDGE/LOCAL_OUT and we
+ * will need the indev then. For a brouter, the real indev
+ * can be a bridge port, so we make sure br_nf_local_out()
+ * doesn't use the bridge parent of the indev by using
+ * the BRNF_DONT_TAKE_PARENT mask.
+ */
+ if (hook == NF_IP_FORWARD && skb->physindev == NULL) {
+ skb->brnfmask &= BRNF_DONT_TAKE_PARENT;
+ skb->physindev = (struct net_device *)in;
+ }
+ okfn(skb);
+ return NF_STOLEN;
+ }
+
+ return NF_ACCEPT;
+}
+
+/* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
+ * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
+ * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
+ * ip_refrag() can return NF_STOLEN.
+ */
+static struct nf_hook_ops br_nf_ops[] = {
+ { { NULL, NULL }, br_nf_pre_routing, PF_BRIDGE, NF_BR_PRE_ROUTING,
NF_BR_PRI_BRNF },
+ { { NULL, NULL }, br_nf_local_in, PF_BRIDGE, NF_BR_LOCAL_IN, NF_BR_PRI_BRNF
},
+ { { NULL, NULL }, br_nf_forward, PF_BRIDGE, NF_BR_FORWARD, NF_BR_PRI_BRNF },
+ { { NULL, NULL }, br_nf_local_out, PF_BRIDGE, NF_BR_LOCAL_OUT,
NF_BR_PRI_FIRST },
+ { { NULL, NULL }, br_nf_post_routing, PF_BRIDGE, NF_BR_POST_ROUTING,
NF_BR_PRI_LAST },
+ { { NULL, NULL }, ipv4_sabotage_in, PF_INET, NF_IP_PRE_ROUTING,
NF_IP_PRI_FIRST },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_FORWARD,
NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_LOCAL_OUT,
NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_POST_ROUTING,
NF_IP_PRI_FIRST }
+};
+
+#define NUMHOOKS (sizeof(br_nf_ops)/sizeof(br_nf_ops[0]))
+
+int br_netfilter_init(void)
+{
+ int i;
+
+ for (i = 0; i < NUMHOOKS; i++) {
+ int ret;
+
+ if ((ret = nf_register_hook(&br_nf_ops[i])) >= 0)
+ continue;
+
+ while (i--)
+ nf_unregister_hook(&br_nf_ops[i]);
+
+ return ret;
+ }
+
+ printk(KERN_NOTICE "Bridge firewalling registered\n");
+
+ return 0;
+}
+
+void br_netfilter_fini(void)
+{
+ int i;
+
+ for (i = NUMHOOKS - 1; i >= 0; i--)
+ nf_unregister_hook(&br_nf_ops[i]);
+}

These changes cannot go in:

1) There is no reason the 'okfn' you use cannot be the function doing the MAC header copy.

This is how this is supposed to work.

I explained in that long thread a few weeks ago how this copy may not be placed in the generic IP code. This is final, you must find a way to make this copy without touching ipv4/*.c

2) The netfilter changes need to be approved by the netfilter team.

I suspect, like myself, they will barf at the phys{in,out}dev additions to sk_buff. We already have enough junk sitting in sk_buff making it larger than it needs to be.

Perhaps you can hang this off the nf_conntrack pointer and specify a destructor.

3) The bridging layer changes need to be approved by Lennert. But I'd suggest working out #1 and #2 first.

Thanks.

Lennert appears to have dropped off the net.

http://www.PowerDNS.com          Versatile DNS Software & Services
http://www.tk                              the dot in .tk
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO

On Monday 14 October 2002 20:01, David S. Miller wrote:
Hello,

These are probably stupid questions to you, but here it goes.

I've checked the skb->dst->hh field and it (or skb->dst itself) was NULL for purely bridged packets. So we'd have to fill this in ourselves. Can the bridge code go fill in a skb->dst and skb->dst->hh? Is this considered clean?

I added a third member as well... It's needed too, in my opinion. There could ofcourse be added a pointer to a struct containing these three values (and a copied Ethernet header). Then we go from 3 to 1 extra member... Anyway, it's not like Lennert and me like adding new members, but we need to save those things somewhere...

So if I change

struct nf_conntrack {
	atomic_t use;
	void (*destroy)(struct nf_conntrack *);
};

into this:

struct nf_conntrack {
	atomic_t use;
	void (*destroy)(struct nf_conntrack *);
	struct brnf_data *brnf;
};

I can keep the copy of the Ethernet header in the struct brnf_data too (then I don't have to touch skbuff->dst).
The skbuff->nfct field can already be in use by an IP connection tracker (or something), so I can't use my own destroy function.
So I'd have to go do something in net/ipv4/netfilter/ip_conntrack_core.c::destroy_conntrack() and I don't know that stuff.
I sure don't like this solution more than the current situation.

Anyway, mapping the IPv4 hooks onto the bridge hooks is in my opinion by definition a hack. But a very useful hack. So if you want this in the kernel you'll have to be forgiving. Or present a nice solution, because I and probably Lennert really don't see a nice(r) solution.

So, the best solution I can think of is adding a skbuff->brnf pointer to a struct brnf_data. This will get rid of the copy in ip_output.c. Is that enough? This will uglify the ip_tables.c patch however.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Mon, 14 Oct 2002 20:58:53 +0200

Can the bridge code go fill in a skb->dst and skb->dst->hh? Is this considered clean?

If it is a properly formed 'dst' entry, it will get cleaned up at SKB free and there will be no problems.

   So if I change
   struct nf_conntrack {

You shouldn't be touching nf_conntrack, you should perhaps instead do something like:

struct nf_ct_info {
	union {
		struct nf_conntrack *master;
		struct nf_bridge_info *brinfo;
	} u;
};

But again, you need to get these sorts of extensions and core changes approved by the netfilter team.

I'm the wrong person to ask about how they would prefer this stuff be done.

So if you want this in the kernel you'll have to be forgiving. Or present a nice solution, because I and probably Lennert really don't see a nice(r) solution.

It is my job to show you why a piece of code isn't going to go in. It is not my job to help you dream up a better solution.

Because, frankly I don't care about bridge netfiltering.

I do care about keeping the code as clean as possible so I don't run into road blocks when trying to rework input/output processing just because I let some bogon hack into the tree I must continue to support.

You do care about bridge netfiltering, so you are going to be the one to find the clean solution that doesn't touch net/ipv4/*.c :-)

This is life in the kernel hacking community :-)

So, the best solution I can think of is adding a skbuff->brnf pointer to a struct brnf_data. This will get rid of the copy in ip_output.c. Is that enough? This will uglify the ip_tables.c patch however.

That could work too, I think you'll need to specify a seperate destructor in that case, and all this stuff ifdef'd on whether bridge netfiltering is enabled or not.

Again, talk to the netfilter folks. They may even have ideas for you that you haven't dreamt of yet.

Franks a lot,
David S. Miller
[email protected]

You were the one who asked for that patch.

Ack.

I care about Linux. I absolutely don't need a bridging firewall for anything. I just happen to know something about it.

This brings me to another question: I've been told it is the general concensus that this bridge firewall should be compiled in the kernel if CONFIG_NETFILTER=y. Or should it be a user option? It is predicted that using a user option will give alot of questions about the bridge firewall not working.

Will do.

cheers,
Bart

From: Bart De Schuymer <[email protected]>
Date: Mon, 14 Oct 2002 21:29:56 +0200

You were the one who asked for that patch.

I asked for the patch to be cleaned up to eliminate the net/ipv4/*.c hacking. :)

This brings me to another question: I've been told it is the general concensus that this bridge firewall should be compiled in the kernel if CONFIG_NETFILTER=y.

I don't have any strong opinion here.

Hello,

This is a follow-up from the previous RFC for the bridge-nf patch. The new patch adds one member to the skbuff, a pointer to a struct nf_bridge_info. There is still a need to change ip_output.c, but the change is the analogue as is done for the skbuff->nfct pointer field. So, for me this is a clean solution. The copy of the Ethernet header is no longer done in ip_fragment().

The patch is available at:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre2-against-2.5.44.diff
An incremental diff, for 2.5.42, against the previous patch is here:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre2.001-against-2.5.42.diff

David, are you happy with this solution?
Other comments?

Here's the patch.

@@ -117,17 +117,23 @@
 /* This is gross, but inline doesn't cut it for avoiding the function
    call in fast path: gcc doesn't inline (needs value tracking?). --RR */
 #ifdef CONFIG_NETFILTER_DEBUG
-#define NF_HOOK nf_hook_slow
+#define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
+ nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN)
+#define NF_HOOK_THRESH nf_hook_slow
 #else
 #define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
 (list_empty(&nf_hooks[(pf)][(hook)])					\
  ? (okfn)(skb)								\
- : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn)))
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN))
+#define NF_HOOK_THRESH(pf, hook, skb, indev, outdev, okfn, thresh)	\
+(list_empty(&nf_hooks[(pf)][(hook)])					\
+ ? (okfn)(skb)								\
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), (thresh)))
 #endif
 
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev, struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *));
+		 int (*okfn)(struct sk_buff *), int thresh);
 
 /* Call setsockopt() */
 int nf_setsockopt(struct sock *sk, int pf, int optval, char *opt, 

@@ -52,8 +52,10 @@
 enum nf_ip_hook_priorities {
 	NF_IP_PRI_FIRST = INT_MIN,
 	NF_IP_PRI_CONNTRACK = -200,
+	NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD = -175,
 	NF_IP_PRI_MANGLE = -150,
 	NF_IP_PRI_NAT_DST = -100,
+	NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT = -50,
 	NF_IP_PRI_FILTER = 0,
 	NF_IP_PRI_NAT_SRC = 100,
 	NF_IP_PRI_LAST = INT_MAX,

--- linux-2.5.44/include/linux/netfilter_bridge.h Sat Oct 19 06:01:57 2002
+++ linux-2.5.44-brnf/include/linux/netfilter_bridge.h Sun Oct 20 21:57:52
2002
@@ -6,6 +6,7 @@

#include <linux/config.h>
#include <linux/netfilter.h>
+#include <asm/atomic.h>

/* Bridge Hooks */
/* After promisc drops, checksum checks. */
@@ -22,14 +23,39 @@
#define NF_BR_BROUTING 5
#define NF_BR_NUMHOOKS 6

+#define BRNF_PKT_TYPE 0x01
+#define BRNF_BRIDGED_DNAT 0x02
+#define BRNF_DONT_TAKE_PARENT 0x04
+
enum nf_br_hook_priorities {
NF_BR_PRI_FIRST = INT_MIN,
- NF_BR_PRI_FILTER_BRIDGED = -200,
- NF_BR_PRI_FILTER_OTHER = 200,
NF_BR_PRI_NAT_DST_BRIDGED = -300,
+ NF_BR_PRI_FILTER_BRIDGED = -200,
+ NF_BR_PRI_BRNF = 0,
NF_BR_PRI_NAT_DST_OTHER = 100,
+ NF_BR_PRI_FILTER_OTHER = 200,
NF_BR_PRI_NAT_SRC = 300,
NF_BR_PRI_LAST = INT_MAX,
+};
+
+static inline
+struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb) +{
+ struct nf_bridge_info **nf_bridge = &(skb->nf_bridge);
+
+ if ((*nf_bridge = kmalloc(sizeof(**nf_bridge), GFP_ATOMIC)) != NULL) {
+ atomic_set(&(*nf_bridge)->use, 1);
+ (*nf_bridge)->mask = 0;
+ (*nf_bridge)->physindev = (*nf_bridge)->physoutdev = NULL;
+ }
+
+ return *nf_bridge;
+}
+
+struct bridge_skb_cb {
+ union {
+ __u32 ipv4;
+ } daddr;
};

#endif

@@ -96,6 +96,14 @@
 struct nf_ct_info {
 	struct nf_conntrack *master;
 };
+
+struct nf_bridge_info {
+	atomic_t use;
+	struct net_device *physindev;
+	struct net_device *physoutdev;
+	unsigned int mask;
+	unsigned long hh[16 / sizeof(unsigned long)];
+};
 #endif
 
 struct sk_buff_head {

@@ -166,6 +174,7 @@
  *	@nfcache: Cache info
  *	@nfct: Associated connection, if any
  *	@nf_debug: Netfilter debugging
+ *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  *	@tc_index: Traffic control index
  */
 

@@ -236,6 +245,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
         unsigned int		nf_debug;
 #endif
+	struct nf_bridge_info	*nf_bridge;
 #endif /* CONFIG_NETFILTER */
 #if defined(CONFIG_HIPPI)
 	union {

@@ -1145,6 +1155,17 @@
 {
 	if (nfct)
 		atomic_inc(&nfct->master->use);
+}
+
+static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
+{
+	if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
+		kfree(nf_bridge);
+}
+static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
+{
+	if (nf_bridge)
+		atomic_inc(&nf_bridge->use);
 }
 #endif
 

@@ -45,6 +45,10 @@
 {
 	printk(KERN_INFO "NET4: Ethernet Bridge 008 for NET4.0\n");
 
+#ifdef CONFIG_NETFILTER
+	if (br_netfilter_init())
+		return 1;
+#endif
 	br_handle_frame_hook = br_handle_frame;
 	br_ioctl_hook = br_ioctl_deviceless_stub;
 #if defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE)

@@ -63,6 +67,9 @@
 
 static void __exit br_deinit(void)
 {
+#ifdef CONFIG_NETFILTER
+	br_netfilter_fini();
+#endif
 	unregister_netdevice_notifier(&br_device_notifier);
 	br_call_ioctl_atomic(__br_clear_ioctl_hook);
 

@@ -30,18 +30,23 @@
 	return 1;
 }
 
-static int __dev_queue_push_xmit(struct sk_buff *skb)
+int br_dev_queue_push_xmit(struct sk_buff *skb)
 {
+#ifdef CONFIG_NETFILTER
+	if (skb->nf_bridge)
+		memcpy(skb->data - 16, skb->nf_bridge->hh, 16);
+#endif
 	skb_push(skb, ETH_HLEN);
+
 	dev_queue_xmit(skb);
 
 	return 0;
 }
 
-static int __br_forward_finish(struct sk_buff *skb)
+int br_forward_finish(struct sk_buff *skb)
 {
 	NF_HOOK(PF_BRIDGE, NF_BR_POST_ROUTING, skb, NULL, skb->dev,
-			__dev_queue_push_xmit);
+			br_dev_queue_push_xmit);
 
 	return 0;
 }

@@ -53,7 +58,7 @@
 	skb->nf_debug = 0;
 #endif
 	NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 static void __br_forward(struct net_bridge_port *to, struct sk_buff *skb)

@@ -64,7 +69,7 @@
 	skb->dev = to->dev;
 
 	NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, indev, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 /* called under bridge lock */

@@ -49,7 +49,7 @@
 			br_pass_frame_up_finish);
 }
 
-static int br_handle_frame_finish(struct sk_buff *skb)
+int br_handle_frame_finish(struct sk_buff *skb)
 {
 	struct net_bridge *br;
 	unsigned char *dest;

@@ -144,8 +144,10 @@
 /* br_forward.c */
 extern void br_deliver(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_dev_queue_push_xmit(struct sk_buff *skb);
 extern void br_forward(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_forward_finish(struct sk_buff *skb);
 extern void br_flood_deliver(struct net_bridge *br,
 		      struct sk_buff *skb,
 		      int clone);

@@ -166,6 +168,7 @@
 			   int *ifindices);
 
 /* br_input.c */
+extern int br_handle_frame_finish(struct sk_buff *skb);
 extern int br_handle_frame(struct sk_buff *skb);
 
 /* br_ioctl.c */

@@ -176,6 +179,10 @@
 	     unsigned long arg1,
 	     unsigned long arg2);
 extern int br_ioctl_deviceless_stub(unsigned long arg);
+
+/* br_netfilter.c */
+extern int br_netfilter_init(void);
+extern void br_netfilter_fini(void);
 
 /* br_stp.c */
 extern int br_is_root_bridge(struct net_bridge *br);

@@ -9,6 +9,11 @@
 bridge-objs	:= br.o br_device.o br_fdb.o br_forward.o br_if.o br_input.o \
 			br_ioctl.o br_notify.o br_stp.o br_stp_bpdu.o \
 			br_stp_if.o br_stp_timer.o
+
+ifeq ($(CONFIG_NETFILTER),y)
+bridge-objs	+= br_netfilter.o
+endif
+
 obj-$(CONFIG_BRIDGE_NF_EBTABLES) += netfilter/
 
 include $(TOPDIR)/Rules.make

@@ -342,10 +342,15 @@
 			       const struct net_device *indev,
 			       const struct net_device *outdev,
 			       struct list_head **i,
-			       int (*okfn)(struct sk_buff *))
+			       int (*okfn)(struct sk_buff *),
+			       int hook_thresh)
 {
 	for (*i = (*i)->next; *i != head; *i = (*i)->next) {
 		struct nf_hook_ops *elem = (struct nf_hook_ops *)*i;
+
+		if (hook_thresh > elem->priority)
+			continue;
+
 		switch (elem->hook(hook, skb, indev, outdev, okfn)) {
 		case NF_QUEUE:
 			return NF_QUEUE;

@@ -413,6 +418,8 @@
 {
 	int status;
 	struct nf_info *info;
+	struct net_device *physindev = NULL;
+	struct net_device *physoutdev = NULL;
 
 	if (!queue_handler[pf].outfn) {
 		kfree_skb(skb);

@@ -435,11 +442,20 @@
 	if (indev) dev_hold(indev);
 	if (outdev) dev_hold(outdev);
 
+	if (skb->nf_bridge) {
+		physindev = skb->nf_bridge->physindev;
+		if (physindev) dev_hold(physindev);
+		physoutdev = skb->nf_bridge->physoutdev;
+		if (physoutdev) dev_hold(physoutdev);
+	}
+
 	status = queue_handler[pf].outfn(skb, info, queue_handler[pf].data);
 	if (status < 0) {
 		/* James M doesn't say fuck enough. */
 		if (indev) dev_put(indev);
 		if (outdev) dev_put(outdev);
+		if (physindev) dev_put(physindev);
+		if (physoutdev) dev_put(physoutdev);
 		kfree(info);
 		kfree_skb(skb);
 		return;

@@ -449,7 +465,8 @@
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev,
 		 struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *))
+		 int (*okfn)(struct sk_buff *),
+		 int hook_thresh)
 {
 	struct list_head *elem;
 	unsigned int verdict;

@@ -481,7 +498,7 @@
 
 	elem = &nf_hooks[pf][hook];
 	verdict = nf_iterate(&nf_hooks[pf][hook], &skb, hook, indev,
-			     outdev, &elem, okfn);
+			     outdev, &elem, okfn, hook_thresh);
 	if (verdict == NF_QUEUE) {
 		NFDEBUG("nf_hook: Verdict = QUEUE.\n");
 		nf_queue(skb, elem, pf, hook, indev, outdev, okfn);

@@ -530,7 +547,7 @@
 		verdict = nf_iterate(&nf_hooks[info->pf][info->hook],
 				     &skb, info->hook, 
 				     info->indev, info->outdev, &elem,
-				     info->okfn);
+				     info->okfn, INT_MIN);
 	}
 
 	switch (verdict) {

@@ -248,6 +248,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	skb->nf_debug	  = 0;
 #endif
+	skb->nf_bridge	  = NULL;
 #endif
 #ifdef CONFIG_NET_SCHED
 	skb->tc_index	  = 0;

@@ -327,6 +328,7 @@
 	}
 #ifdef CONFIG_NETFILTER
 	nf_conntrack_put(skb->nfct);
+	nf_bridge_put(skb->nf_bridge);
 #endif
 	skb_headerinit(skb, NULL, 0);  /* clean state */
 	kfree_skbmem(skb);

@@ -392,6 +394,7 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	C(nf_debug);
 #endif
+	C(nf_bridge);
 #endif /*CONFIG_NETFILTER*/
 #if defined(CONFIG_HIPPI)
 	C(private);

@@ -404,6 +407,7 @@
 	skb->cloned = 1;
 #ifdef CONFIG_NETFILTER
 	nf_conntrack_get(skb->nfct);
+	nf_bridge_get(skb->nf_bridge);
 #endif
 	return n;
 }

@@ -438,6 +442,8 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	new->nf_debug	= old->nf_debug;
 #endif
+	new->nf_bridge	= old->nf_bridge;
+	nf_bridge_get(new->nf_bridge);
 #endif
 #ifdef CONFIG_NET_SCHED
 	new->tc_index	= old->tc_index;

@@ -396,6 +396,8 @@
 	/* Connection association is same as pre-frag packet */
 	to->nfct = from->nfct;
 	nf_conntrack_get(to->nfct);
+	to->nf_bridge = from->nf_bridge;
+	nf_bridge_get(to->nf_bridge);
 #ifdef CONFIG_NETFILTER_DEBUG
 	to->nf_debug = from->nf_debug;
 #endif

@@ -121,12 +121,14 @@
 static inline int
 ip_packet_match(const struct iphdr *ip,
 		const char *indev,
+		const char *physindev,
 		const char *outdev,
+		const char *physoutdev,
 		const struct ipt_ip *ipinfo,
 		int isfrag)
 {
 	size_t i;
-	unsigned long ret;
+	unsigned long ret, ret2;
 
 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
 

@@ -156,7 +158,13 @@
 			& ((const unsigned long *)ipinfo->iniface_mask)[i];
 	}
 
-	if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
+	for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret2 |= (((const unsigned long *)physindev)[i]
+			^ ((const unsigned long *)ipinfo->iniface)[i])
+			& ((const unsigned long *)ipinfo->iniface_mask)[i];
+	}
+
+	if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
 			indev, ipinfo->iniface,
 			ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");

@@ -169,7 +177,13 @@
 			& ((const unsigned long *)ipinfo->outiface_mask)[i];
 	}
 
-	if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
+	for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret2 |= (((const unsigned long *)physoutdev)[i]
+			^ ((const unsigned long *)ipinfo->outiface)[i])
+			& ((const unsigned long *)ipinfo->outiface_mask)[i];
+	}
+
+	if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
 			outdev, ipinfo->outiface,
 			ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");

@@ -268,6 +282,7 @@
 	/* Initializing verdict to NF_DROP keeps gcc happy. */
 	unsigned int verdict = NF_DROP;
 	const char *indev, *outdev;
+	const char *physindev, *physoutdev;
 	void *table_base;
 	struct ipt_entry *e, *back;
 

@@ -277,6 +292,16 @@
 	datalen = (*pskb)->len - ip->ihl * 4;
 	indev = in ? in->name : nulldevname;
 	outdev = out ? out->name : nulldevname;
+	if ((*pskb)->nf_bridge) {
+		physindev = (*pskb)->nf_bridge->physindev ?
+			(*pskb)->nf_bridge->physindev->name : nulldevname;
+		physoutdev = (*pskb)->nf_bridge->physoutdev ?
+			(*pskb)->nf_bridge->physoutdev->name : nulldevname;
+	} else {
+		physindev = nulldevname;
+		physoutdev = nulldevname;
+	}
+
 	/* We handle fragments by dealing with the first fragment as
 	 * if it was a normal packet.  All other fragments are treated
 	 * normally, except that they will NEVER match rules that ask

@@ -311,7 +336,8 @@
 		IP_NF_ASSERT(e);
 		IP_NF_ASSERT(back);
 		(*pskb)->nfcache |= e->nfcache;
-		if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
+		if (ip_packet_match(ip, indev, physindev, outdev, physoutdev,
+		    &e->ip, offset)) {
 			struct ipt_entry_target *t;
 
 			if (IPT_MATCH_ITERATE(e, do_match,

@@ -285,10 +285,18 @@
 	level_string[1] = '0' + (loginfo->level % 8);
 	spin_lock_bh(&log_lock);
 	printk(level_string);
-	printk("%sIN=%s OUT=%s ",
-	       loginfo->prefix,
-	       in ? in->name : "",
-	       out ? out->name : "");
+	printk("%sIN=%s ", loginfo->prefix, in ? in->name : "");
+	if ((*pskb)->nf_bridge) {
+		struct net_device *physindev = (*pskb)->nf_bridge->physindev;
+		struct net_device *physoutdev = (*pskb)->nf_bridge->physoutdev;
+
+		if (physindev && in != physindev)
+			printk("PHYSIN=%s ", physindev->name);
+		printk("OUT=%s ", out ? out->name : "");
+		if (physoutdev && out != physoutdev)
+			printk("PHYSOUT=%s ", physoutdev->name);
+	}
+
 	if (in && !out) {
 		/* MAC logging for input chain only. */
 		printk("MAC=");

@@ -0,0 +1,616 @@
+/*
+ *	Handle firewalling
+ *	Linux ethernet bridge
+ *
+ *	Authors:
+ *	Lennert Buytenhek               <[email protected]>
+ *	Bart De Schuymer		<[email protected]>
+ *
+ *	$Id: bridge-nf-0.0.10-dev-pre2-against-2.5.42.diff,v 1.1 2002/10/19 

10:46:51 bdschuym Exp $
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ *
+ * Lennert dedicates this file to Kerstin Wurdinger.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/ip.h>
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+#include <linux/if_ether.h>
+#include <linux/netfilter_bridge.h>
+#include <linux/netfilter_ipv4.h>
+#include <linux/in_route.h>
+#include <net/ip.h>
+#include <asm/uaccess.h>
+#include <asm/checksum.h>
+#include "br_private.h"
+
+
+#define skb_origaddr(skb) (((struct bridge_skb_cb *) \
+ (skb->cb))->daddr.ipv4)
+#define store_orig_dstaddr(skb) (skb_origaddr(skb) = (skb)->nh.iph->daddr)
+#define dnat_took_place(skb) (skb_origaddr(skb) != (skb)->nh.iph->daddr)
+#define clear_cb(skb) (memset(&skb_origaddr(skb), 0, \
+ sizeof(struct bridge_skb_cb)))
+
+#define has_bridge_parent(device) ((device)->br_port != NULL)
+#define bridge_parent(device) (&((device)->br_port->br->dev))
+
+/* We need these fake structures to make netfilter happy --
+ * lots of places assume that skb->dst != NULL, which isn't
+ * all that unreasonable.
+ *
+ * Currently, we fill in the PMTU entry because netfilter
+ * refragmentation needs it, and the rt_flags entry because
+ * ipt_REJECT needs it. Future netfilter modules might
+ * require us to fill additional fields.
+ */
+static struct net_device __fake_net_device = {
+ hard_header_len: ETH_HLEN
+};
+
+static struct rtable __fake_rtable = {
+ u: {
+ dst: {
+ __refcnt: ATOMIC_INIT(1),
+ dev: &__fake_net_device,
+ pmtu: 1500
+ }
+ },
+
+ rt_flags: 0
+};
+
+
+/* PF_BRIDGE/PRE_ROUTING *********************************************/ +static void __br_dnat_complain(void)
+{
+ static unsigned long last_complaint = 0;
+
+ if (jiffies - last_complaint >= 5 * HZ) {
+ printk(KERN_WARNING "Performing cross-bridge DNAT requires IP "
+ "forwarding to be enabled\n");
+ last_complaint = jiffies;
+ }
+}
+
+
+/* This requires some explaining. If DNAT has taken place,
+ * we will need to fix up the destination Ethernet address,
+ * and this is a tricky process.
+ *
+ * There are two cases to consider:
+ * 1. The packet was DNAT'ed to a device in the same bridge
+ * port group as it was received on. We can still bridge
+ * the packet.
+ * 2. The packet was DNAT'ed to a different device, either
+ * a non-bridged device or another bridge port group.
+ * The packet will need to be routed.
+ *
+ * The correct way of distinguishing between these two cases is to
+ * call ip_route_input() and to look at skb->dst->dev, which is
+ * changed to the destination device if ip_route_input() succeeds.
+ *
+ * Let us first consider the case that ip_route_input() succeeds:
+ *
+ * If skb->dst->dev equals the logical bridge device the packet
+ * came in on, we can consider this bridging. We then call
+ * skb->dst->output() which will make the packet enter br_nf_local_out()
+ * not much later. In that function it is assured that the iptables
+ * FORWARD chain is traversed for the packet.
+ *
+ * Otherwise, the packet is considered to be routed and we just
+ * change the destination MAC address so that the packet will
+ * later be passed up to the IP stack to be routed.
+ *
+ * Let us now consider the case that ip_route_input() fails:
+ *
+ * After a "echo '0' > /proc/sys/net/ipv4/ip_forward" ip_route_input()
+ * will fail, while ip_route_output() will return success. The source
+ * address for ip_route_output() is set to zero, so ip_route_output()
+ * thinks we're handling a locally generated packet and won't care
+ * if IP forwarding is allowed. We send a warning message to the users's
+ * log telling her to put IP forwarding on.
+ *
+ * ip_route_input() will also fail if there is no route available.
+ * In that case we just drop the packet.
+ *
+ * --Lennert, 20020411
+ * --Bart, 20020416 (updated)
+ * --Bart, 20021007 (updated)
+ */
+
+static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb) +{
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug |= (1 << NF_BR_PRE_ROUTING) | (1 << NF_BR_FORWARD);
+#endif
+
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ skb->nf_bridge->mask |= BRNF_PKT_TYPE;
+ }
+
+ skb->dev = bridge_parent(skb->dev);
+ skb->dst->output(skb);
+ return 0;
+}
+
+static int br_nf_pre_routing_finish(struct sk_buff *skb)
+{
+ struct net_device *dev = skb->dev;
+ struct iphdr *iph = skb->nh.iph;
+ struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_BR_PRE_ROUTING);
+#endif
+
+ if (nf_bridge->mask & BRNF_PKT_TYPE) {
+ skb->pkt_type = PACKET_OTHERHOST;
+ nf_bridge->mask ^= BRNF_PKT_TYPE;
+ }
+
+ if (dnat_took_place(skb)) {
+ if (ip_route_input(skb, iph->daddr, iph->saddr, iph->tos,
+ dev)) {
+ struct rtable *rt;
+
+ if (!ip_route_output(&rt, iph->daddr, 0, iph->tos, 0)) {
+ /* Bridged-and-DNAT'ed traffic doesn't
+ * require ip_forwarding.
+ */
+ if (((struct dst_entry *)rt)->dev == dev) {
+ skb->dst = (struct dst_entry *)rt;
+ goto bridged_dnat;
+ }
+ __br_dnat_complain();
+ dst_release((struct dst_entry *)rt);
+ }
+ kfree_skb(skb);
+ return 0;
+ } else {
+ if (skb->dst->dev == dev) {
+bridged_dnat:
+ /* Tell br_nf_local_out this is a
+ * bridged frame
+ */
+ nf_bridge->mask |= BRNF_BRIDGED_DNAT;
+ skb->dev = nf_bridge->physindev;
+ clear_cb(skb);
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING,

+					       skb, skb->dev, NULL,
+					       br_nf_pre_routing_finish_bridge,
+					       1);

+ return 0;
+ }
+ memcpy(skb->mac.ethernet->h_dest, dev->dev_addr,
+ ETH_ALEN);
+ }
+ } else {
+ skb->dst = (struct dst_entry *)&__fake_rtable;
+ dst_hold(skb->dst);
+ }
+
+ clear_cb(skb);
+ skb->dev = nf_bridge->physindev;
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
+ br_handle_frame_finish, 1);
+
+ return 0;
+}
+
+/* Replicate the checks that IPv4 does on packet reception.
+ * Set skb->dev to the bridge device (i.e. parent of the
+ * receiving device) to make netfilter happy, the REDIRECT
+ * target in particular. Save the original destination IP
+ * address to be able to detect DNAT afterwards.
+ */
+static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct iphdr *iph;
+ __u32 len;
+ struct sk_buff *skb;
+ struct nf_bridge_info *nf_bridge;
+
+ if ((*pskb)->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ if ((skb = skb_share_check(*pskb, GFP_ATOMIC)) == NULL)
+ goto out;
+
+ if (!pskb_may_pull(skb, sizeof(struct iphdr)))
+ goto inhdr_error;
+
+ iph = skb->nh.iph;
+ if (iph->ihl < 5 || iph->version != 4)
+ goto inhdr_error;
+
+ if (!pskb_may_pull(skb, 4*iph->ihl))
+ goto inhdr_error;
+
+ iph = skb->nh.iph;
+ if (ip_fast_csum((__u8 *)iph, iph->ihl) != 0)
+ goto inhdr_error;
+
+ len = ntohs(iph->tot_len);
+ if (skb->len < len || len < 4*iph->ihl)
+ goto inhdr_error;
+
+ if (skb->len > len) {
+ __pskb_trim(skb, len);
+ if (skb->ip_summed == CHECKSUM_HW)
+ skb->ip_summed = CHECKSUM_NONE;
+ }
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_PRE_ROUTING);
+#endif
+ if ((nf_bridge = nf_bridge_alloc(skb)) == NULL)
+ return NF_DROP;
+
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ nf_bridge->mask |= BRNF_PKT_TYPE;
+ }
+
+ nf_bridge->physindev = skb->dev;
+ skb->dev = bridge_parent(skb->dev);
+ store_orig_dstaddr(skb);
+
+ NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, skb->dev, NULL,
+ br_nf_pre_routing_finish);
+
+ return NF_STOLEN;
+
+inhdr_error:
+// IP_INC_STATS_BH(IpInHdrErrors);
+out:
+ return NF_DROP;
+}
+
+
+/* PF_BRIDGE/LOCAL_IN ************************************************/ +/* The packet is locally destined, which requires a real
+ * dst_entry, so detach the fake one. On the way up, the
+ * packet would pass through PRE_ROUTING again (which already
+ * took place when the packet entered the bridge), but we
+ * register an IPv4 PRE_ROUTING 'sabotage' hook that will
+ * prevent this from happening.
+ */
+static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ if (skb->dst == (struct dst_entry *)&__fake_rtable) {
+ dst_release(skb->dst);
+ skb->dst = NULL;
+ }
+
+ return NF_ACCEPT;
+}
+
+
+/* PF_BRIDGE/FORWARD *************************************************/ +static int br_nf_forward_finish(struct sk_buff *skb)
+{
+ struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_BR_FORWARD);
+#endif
+
+ if (nf_bridge->mask & BRNF_PKT_TYPE) {
+ skb->pkt_type = PACKET_OTHERHOST;
+ nf_bridge->mask ^= BRNF_PKT_TYPE;
+ }
+
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_FORWARD, skb, nf_bridge->physindev,
+ skb->dev, br_forward_finish, 1);
+
+ return 0;
+}
+
+/* This is the 'purely bridged' case. We pass the packet to
+ * netfilter with indev and outdev set to the bridge device,
+ * but we are still able to filter on the 'real' indev/outdev
+ * because another bit of the bridge-nf patch overloads the
+ * '-i' and '-o' iptables interface checks to take
+ * skb->phys{in,out}dev into account as well (so both the real
+ * device and the bridge device will match).
+ */
+static unsigned int br_nf_forward(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+ struct nf_bridge_info *nf_bridge;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_BR_FORWARD);
+#endif
+
+ nf_bridge = skb->nf_bridge;
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ nf_bridge->mask |= BRNF_PKT_TYPE;
+ }
+
+ nf_bridge->physoutdev = skb->dev;
+
+ NF_HOOK(PF_INET, NF_IP_FORWARD, skb, bridge_parent(nf_bridge->physindev),
+ bridge_parent(skb->dev), br_nf_forward_finish);
+
+ return NF_STOLEN;
+}
+
+
+/* PF_BRIDGE/LOCAL_OUT ***********************************************/ +static int br_nf_local_out_finish(struct sk_buff *skb)
+{
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug &= ~(1 << NF_BR_LOCAL_OUT);
+#endif
+
+ NF_HOOK_THRESH(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
+ br_forward_finish, NF_BR_PRI_FIRST + 1);
+
+ return 0;
+}
+
+
+/* This function sees both locally originated IP packets and forwarded
+ * IP packets (in both cases the destination device is a bridge
+ * device). It also sees bridged-and-DNAT'ed packets.
+ * For the sake of interface transparency (i.e. properly
+ * overloading the '-o' option), we steal packets destined to
+ * a bridge device away from the PF_INET/FORWARD and PF_INET/OUTPUT hook
+ * functions, and give them back later, when we have determined the real
+ * output device. This is done in here.
+ *
+ * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
+ * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
+ * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
+ * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
+ * will be executed.
+ * Otherwise, if nf_bridge->physindev is NULL, the bridge-nf code never touched
+ * this packet before, and so the packet was locally originated. We fake
+ * the PF_INET/LOCAL_OUT hook.
+ * Finally, if nf_bridge->physindev isn't NULL, then the packet was IP routed,
+ * so we fake the PF_INET/FORWARD hook. ipv4_sabotage_out() makes sure
+ * even routed packets that didn't arrive on a bridge interface have their
+ * nf_bridge->physindev set.
+ */
+
+static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*_okfn)(struct sk_buff *))
+{
+ int (*okfn)(struct sk_buff *skb);
+ struct net_device *realindev;
+ struct sk_buff *skb = *pskb;
+ struct nf_bridge_info *nf_bridge;
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ /* Sometimes we get packets with NULL ->dst here (for example,
+ * running a dhcp client daemon triggers this).
+ */
+ if (skb->dst == NULL)
+ return NF_ACCEPT;
+
+ nf_bridge = skb->nf_bridge;
+ nf_bridge->physoutdev = skb->dev;
+
+ realindev = nf_bridge->physindev;
+
+ /* Bridged, take PF_BRIDGE/FORWARD.
+ * (see big note in front of br_nf_pre_routing_finish)
+ */
+ if (nf_bridge->mask & BRNF_BRIDGED_DNAT) {
+ okfn = br_forward_finish;
+
+ if (nf_bridge->mask & BRNF_PKT_TYPE) {
+ skb->pkt_type = PACKET_OTHERHOST;
+ nf_bridge->mask ^= BRNF_PKT_TYPE;
+ }
+
+ NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, realindev,
+ skb->dev, okfn);
+ } else {
+ okfn = br_nf_local_out_finish;
+ /* IP forwarded traffic has a physindev, locally
+ * generated traffic hasn't.
+ */
+ if (realindev != NULL) {
+ if (((nf_bridge->mask & BRNF_DONT_TAKE_PARENT) == 0) &&
+ has_bridge_parent(realindev))
+ realindev = bridge_parent(realindev);
+
+ NF_HOOK_THRESH(PF_INET, NF_IP_FORWARD, skb, realindev,
+ bridge_parent(skb->dev), okfn,
+ NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD + 1);
+ } else {
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_LOCAL_OUT);
+#endif
+
+ NF_HOOK_THRESH(PF_INET, NF_IP_LOCAL_OUT, skb, realindev,
+ bridge_parent(skb->dev), okfn,
+ NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT + 1);
+ }
+ }
+
+ return NF_STOLEN;
+}
+
+
+/* PF_BRIDGE/POST_ROUTING ********************************************/ +static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ struct sk_buff *skb = *pskb;
+ struct nf_bridge_info *nf_bridge = (*pskb)->nf_bridge;
+
+ /* Be very paranoid. */
+ if (skb->mac.raw < skb->head || skb->mac.raw + ETH_HLEN > skb->data) {
+ printk(KERN_CRIT "br_netfilter: Argh!! br_nf_post_routing: "
+ "bad mac.raw pointer.");
+ if (skb->dev != NULL) {
+ printk("[%s]", skb->dev->name);
+ if (has_bridge_parent(skb->dev))
+ printk("[%s]", bridge_parent(skb->dev)->name);
+ }
+ printk("\n");
+ return NF_ACCEPT;
+ }
+
+ if (skb->protocol != __constant_htons(ETH_P_IP))
+ return NF_ACCEPT;
+
+ /* Sometimes we get packets with NULL ->dst here (for example,
+ * running a dhcp client daemon triggers this).
+ */
+ if (skb->dst == NULL)
+ return NF_ACCEPT;
+
+#ifdef CONFIG_NETFILTER_DEBUG
+ skb->nf_debug ^= (1 << NF_IP_POST_ROUTING);
+#endif
+
+ /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
+ * about the value of skb->pkt_type.
+ */
+ if (skb->pkt_type == PACKET_OTHERHOST) {
+ skb->pkt_type = PACKET_HOST;
+ nf_bridge->mask |= BRNF_PKT_TYPE;
+ }
+
+ memcpy(nf_bridge->hh, skb->data - 16, 16);
+
+ NF_HOOK(PF_INET, NF_IP_POST_ROUTING, skb, NULL,
+ bridge_parent(skb->dev), br_dev_queue_push_xmit);
+
+ return NF_STOLEN;
+}
+
+
+/* IPv4/SABOTAGE *****************************************************/ +
+/* Don't hand locally destined packets to PF_INET/PRE_ROUTING
+ * for the second time.
+ */
+static unsigned int ipv4_sabotage_in(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ if (in->hard_start_xmit == br_dev_xmit &&
+ okfn != br_nf_pre_routing_finish) {
+ okfn(*pskb);
+ return NF_STOLEN;
+ }
+
+ return NF_ACCEPT;
+}
+
+/* Postpone execution of PF_INET/FORWARD, PF_INET/LOCAL_OUT
+ * and PF_INET/POST_ROUTING until we have done the forwarding
+ * decision in the bridge code and have determined skb->physoutdev.
+ */
+static unsigned int ipv4_sabotage_out(unsigned int hook, struct sk_buff **pskb,
+ const struct net_device *in, const struct net_device *out,
+ int (*okfn)(struct sk_buff *))
+{
+ if (out->hard_start_xmit == br_dev_xmit &&

+	    okfn != br_nf_forward_finish &&
+	    okfn != br_nf_local_out_finish &&
+	    okfn != br_dev_queue_push_xmit) {

+ struct sk_buff *skb = *pskb;
+ struct nf_bridge_info *nf_bridge;
+
+ if (!skb->nf_bridge && !nf_bridge_alloc(skb))
+ return NF_DROP;
+
+ nf_bridge = skb->nf_bridge;
+
+ /* This frame will arrive on PF_BRIDGE/LOCAL_OUT and we
+ * will need the indev then. For a brouter, the real indev
+ * can be a bridge port, so we make sure br_nf_local_out()
+ * doesn't use the bridge parent of the indev by using
+ * the BRNF_DONT_TAKE_PARENT mask.
+ */
+ if (hook == NF_IP_FORWARD && nf_bridge->physindev == NULL) {
+ nf_bridge->mask &= BRNF_DONT_TAKE_PARENT;
+ nf_bridge->physindev = (struct net_device *)in;
+ }
+ okfn(skb);
+ return NF_STOLEN;
+ }
+
+ return NF_ACCEPT;
+}
+
+/* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
+ * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
+ * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
+ * ip_refrag() can return NF_STOLEN.
+ */
+static struct nf_hook_ops br_nf_ops[] = {
+ { { NULL, NULL }, br_nf_pre_routing, PF_BRIDGE, NF_BR_PRE_ROUTING,
NF_BR_PRI_BRNF },
+ { { NULL, NULL }, br_nf_local_in, PF_BRIDGE, NF_BR_LOCAL_IN, NF_BR_PRI_BRNF
},
+ { { NULL, NULL }, br_nf_forward, PF_BRIDGE, NF_BR_FORWARD, NF_BR_PRI_BRNF },
+ { { NULL, NULL }, br_nf_local_out, PF_BRIDGE, NF_BR_LOCAL_OUT,
NF_BR_PRI_FIRST },
+ { { NULL, NULL }, br_nf_post_routing, PF_BRIDGE, NF_BR_POST_ROUTING,
NF_BR_PRI_LAST },
+ { { NULL, NULL }, ipv4_sabotage_in, PF_INET, NF_IP_PRE_ROUTING,
NF_IP_PRI_FIRST },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_FORWARD,
NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_LOCAL_OUT,
NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT },
+ { { NULL, NULL }, ipv4_sabotage_out, PF_INET, NF_IP_POST_ROUTING,
NF_IP_PRI_FIRST }
+};
+
+#define NUMHOOKS (sizeof(br_nf_ops)/sizeof(br_nf_ops[0]))
+
+int br_netfilter_init(void)
+{
+ int i;
+
+ for (i = 0; i < NUMHOOKS; i++) {
+ int ret;
+
+ if ((ret = nf_register_hook(&br_nf_ops[i])) >= 0)
+ continue;
+
+ while (i--)
+ nf_unregister_hook(&br_nf_ops[i]);
+
+ return ret;
+ }
+
+ printk(KERN_NOTICE "Bridge firewalling registered\n");
+
+ return 0;
+}
+
+void br_netfilter_fini(void)
+{
+ int i;
+
+ for (i = NUMHOOKS - 1; i >= 0; i--)
+ nf_unregister_hook(&br_nf_ops[i]);
+}

From: Bart De Schuymer <[email protected]>
Date: Mon, 21 Oct 2002 00:20:37 +0200

This is a follow-up from the previous RFC for the bridge-nf patch. The new patch adds one member to the skbuff, a pointer to a struct nf_bridge_info. There is still a need to change ip_output.c, but the change is the analogue as is done for the skbuff->nfct pointer field. So, for me this is a clean solution. The copy of the Ethernet header is no longer done in ip_fragment().

This definitely looks a lot better.

I still want the netfilter team to 'ACK' the core/ipv4 netfilter changes before I apply this. :-)

Is [email protected] dead, or? Anyway, if you think the patch just totally sucks, don't worry, I can take it. Just say so.

cheers,
Bart

Hello David, Harald, others,

The following patch deals with the problems you still had with the earlier one. Changes:
1. add #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) everywhere
2. don't touch ip_tables.c
3. no ipt_physdev.c file yet. I'll try to make it this weekend.

As this ipt_physdev.c module is not essential I propose to already apply this patch. Harald, should I make this module for patch-o-magic or can I post it directly to David?

The patch is also here:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre3-against-2.5.44.diff
An incremental patch vs the previous one is here:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre3.001-against-2.5.44.diff

cheers,
Bart

@@ -117,17 +117,23 @@
 /* This is gross, but inline doesn't cut it for avoiding the function
    call in fast path: gcc doesn't inline (needs value tracking?). --RR */
 #ifdef CONFIG_NETFILTER_DEBUG
-#define NF_HOOK nf_hook_slow
+#define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
+ nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN)
+#define NF_HOOK_THRESH nf_hook_slow
 #else
 #define NF_HOOK(pf, hook, skb, indev, outdev, okfn)			\
 (list_empty(&nf_hooks[(pf)][(hook)])					\
  ? (okfn)(skb)								\
- : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn)))
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), INT_MIN))
+#define NF_HOOK_THRESH(pf, hook, skb, indev, outdev, okfn, thresh)	\
+(list_empty(&nf_hooks[(pf)][(hook)])					\
+ ? (okfn)(skb)								\
+ : nf_hook_slow((pf), (hook), (skb), (indev), (outdev), (okfn), (thresh)))
 #endif
 
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev, struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *));
+		 int (*okfn)(struct sk_buff *), int thresh);
 
 /* Call setsockopt() */
 int nf_setsockopt(struct sock *sk, int pf, int optval, char *opt, 

@@ -52,8 +52,10 @@
 enum nf_ip_hook_priorities {
 	NF_IP_PRI_FIRST = INT_MIN,
 	NF_IP_PRI_CONNTRACK = -200,
+	NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD = -175,
 	NF_IP_PRI_MANGLE = -150,
 	NF_IP_PRI_NAT_DST = -100,
+	NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT = -50,
 	NF_IP_PRI_FILTER = 0,
 	NF_IP_PRI_NAT_SRC = 100,
 	NF_IP_PRI_LAST = INT_MAX,

@@ -6,6 +6,7 @@
 
 #include <linux/config.h>
 #include <linux/netfilter.h>
+#include <asm/atomic.h>
 
 /* Bridge Hooks */
 /* After promisc drops, checksum checks. */

@@ -22,14 +23,39 @@
 #define NF_BR_BROUTING		5
 #define NF_BR_NUMHOOKS		6
 
+#define BRNF_PKT_TYPE			0x01
+#define BRNF_BRIDGED_DNAT		0x02
+#define BRNF_DONT_TAKE_PARENT		0x04
+
 enum nf_br_hook_priorities {
 	NF_BR_PRI_FIRST = INT_MIN,
-	NF_BR_PRI_FILTER_BRIDGED = -200,
-	NF_BR_PRI_FILTER_OTHER = 200,
 	NF_BR_PRI_NAT_DST_BRIDGED = -300,
+	NF_BR_PRI_FILTER_BRIDGED = -200,
+	NF_BR_PRI_BRNF = 0,
 	NF_BR_PRI_NAT_DST_OTHER = 100,
+	NF_BR_PRI_FILTER_OTHER = 200,
 	NF_BR_PRI_NAT_SRC = 300,
 	NF_BR_PRI_LAST = INT_MAX,
+};
+
+static inline
+struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
+{
+	struct nf_bridge_info **nf_bridge = &(skb->nf_bridge);
+
+	if ((*nf_bridge = kmalloc(sizeof(**nf_bridge), GFP_ATOMIC)) != NULL) {
+		atomic_set(&(*nf_bridge)->use, 1);
+		(*nf_bridge)->mask = 0;
+		(*nf_bridge)->physindev = (*nf_bridge)->physoutdev = NULL;
+	}
+
+	return *nf_bridge;
+}
+
+struct bridge_skb_cb {
+	union {
+		__u32 ipv4;
+	} daddr;
 };
 
 #endif

@@ -96,6 +96,17 @@
 struct nf_ct_info {
 	struct nf_conntrack *master;
 };
+
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+struct nf_bridge_info {
+	atomic_t use;
+	struct net_device *physindev;
+	struct net_device *physoutdev;
+	unsigned int mask;
+	unsigned long hh[16 / sizeof(unsigned long)];
+};
+#endif
+
 #endif
 
 struct sk_buff_head {

@@ -166,6 +177,7 @@
  *	@nfcache: Cache info
  *	@nfct: Associated connection, if any
  *	@nf_debug: Netfilter debugging
+ *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  *	@tc_index: Traffic control index
  */
 

@@ -236,6 +248,9 @@
 #ifdef CONFIG_NETFILTER_DEBUG
         unsigned int		nf_debug;
 #endif
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	struct nf_bridge_info	*nf_bridge;
+#endif
 #endif /* CONFIG_NETFILTER */
 #if defined(CONFIG_HIPPI)
 	union {

@@ -1146,6 +1161,20 @@
 	if (nfct)
 		atomic_inc(&nfct->master->use);
 }
+
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
+{
+	if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
+		kfree(nf_bridge);
+}
+static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
+{
+	if (nf_bridge)
+		atomic_inc(&nf_bridge->use);
+}
+#endif
+
 #endif
 
 #endif	/* __KERNEL__ */

@@ -45,6 +45,10 @@
 {
 	printk(KERN_INFO "NET4: Ethernet Bridge 008 for NET4.0\n");
 
+#ifdef CONFIG_NETFILTER
+	if (br_netfilter_init())
+		return 1;
+#endif
 	br_handle_frame_hook = br_handle_frame;
 	br_ioctl_hook = br_ioctl_deviceless_stub;
 #if defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE)

@@ -63,6 +67,9 @@
 
 static void __exit br_deinit(void)
 {
+#ifdef CONFIG_NETFILTER
+	br_netfilter_fini();
+#endif
 	unregister_netdevice_notifier(&br_device_notifier);
 	br_call_ioctl_atomic(__br_clear_ioctl_hook);
 

@@ -30,18 +30,23 @@
 	return 1;
 }
 
-static int __dev_queue_push_xmit(struct sk_buff *skb)
+int br_dev_queue_push_xmit(struct sk_buff *skb)
 {
+#ifdef CONFIG_NETFILTER
+	if (skb->nf_bridge)
+		memcpy(skb->data - 16, skb->nf_bridge->hh, 16);
+#endif
 	skb_push(skb, ETH_HLEN);
+
 	dev_queue_xmit(skb);
 
 	return 0;
 }
 
-static int __br_forward_finish(struct sk_buff *skb)
+int br_forward_finish(struct sk_buff *skb)
 {
 	NF_HOOK(PF_BRIDGE, NF_BR_POST_ROUTING, skb, NULL, skb->dev,
-			__dev_queue_push_xmit);
+			br_dev_queue_push_xmit);
 
 	return 0;
 }

@@ -53,7 +58,7 @@
 	skb->nf_debug = 0;
 #endif
 	NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 static void __br_forward(struct net_bridge_port *to, struct sk_buff *skb)

@@ -64,7 +69,7 @@
 	skb->dev = to->dev;
 
 	NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, indev, skb->dev,
-			__br_forward_finish);
+			br_forward_finish);
 }
 
 /* called under bridge lock */

@@ -49,7 +49,7 @@
 			br_pass_frame_up_finish);
 }
 
-static int br_handle_frame_finish(struct sk_buff *skb)
+int br_handle_frame_finish(struct sk_buff *skb)
 {
 	struct net_bridge *br;
 	unsigned char *dest;

@@ -144,8 +144,10 @@
 /* br_forward.c */
 extern void br_deliver(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_dev_queue_push_xmit(struct sk_buff *skb);
 extern void br_forward(struct net_bridge_port *to,
 		struct sk_buff *skb);
+extern int br_forward_finish(struct sk_buff *skb);
 extern void br_flood_deliver(struct net_bridge *br,
 		      struct sk_buff *skb,
 		      int clone);

@@ -166,6 +168,7 @@
 			   int *ifindices);
 
 /* br_input.c */
+extern int br_handle_frame_finish(struct sk_buff *skb);
 extern int br_handle_frame(struct sk_buff *skb);
 
 /* br_ioctl.c */

@@ -176,6 +179,10 @@
 	     unsigned long arg1,
 	     unsigned long arg2);
 extern int br_ioctl_deviceless_stub(unsigned long arg);
+
+/* br_netfilter.c */
+extern int br_netfilter_init(void);
+extern void br_netfilter_fini(void);
 
 /* br_stp.c */
 extern int br_is_root_bridge(struct net_bridge *br);

@@ -9,6 +9,11 @@
 bridge-objs	:= br.o br_device.o br_fdb.o br_forward.o br_if.o br_input.o \
 			br_ioctl.o br_notify.o br_stp.o br_stp_bpdu.o \
 			br_stp_if.o br_stp_timer.o
+
+ifeq ($(CONFIG_NETFILTER),y)
+bridge-objs	+= br_netfilter.o
+endif
+
 obj-$(CONFIG_BRIDGE_NF_EBTABLES) += netfilter/
 
 include $(TOPDIR)/Rules.make

@@ -342,10 +342,15 @@
 			       const struct net_device *indev,
 			       const struct net_device *outdev,
 			       struct list_head **i,
-			       int (*okfn)(struct sk_buff *))
+			       int (*okfn)(struct sk_buff *),
+			       int hook_thresh)
 {
 	for (*i = (*i)->next; *i != head; *i = (*i)->next) {
 		struct nf_hook_ops *elem = (struct nf_hook_ops *)*i;
+
+		if (hook_thresh > elem->priority)
+			continue;
+
 		switch (elem->hook(hook, skb, indev, outdev, okfn)) {
 		case NF_QUEUE:
 			return NF_QUEUE;

@@ -413,6 +418,10 @@
 {
 	int status;
 	struct nf_info *info;
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	struct net_device *physindev = NULL;
+	struct net_device *physoutdev = NULL;
+#endif
 
 	if (!queue_handler[pf].outfn) {
 		kfree_skb(skb);

@@ -435,11 +444,24 @@
 	if (indev) dev_hold(indev);
 	if (outdev) dev_hold(outdev);
 
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	if (skb->nf_bridge) {
+		physindev = skb->nf_bridge->physindev;
+		if (physindev) dev_hold(physindev);
+		physoutdev = skb->nf_bridge->physoutdev;
+		if (physoutdev) dev_hold(physoutdev);
+	}
+#endif
+
 	status = queue_handler[pf].outfn(skb, info, queue_handler[pf].data);
 	if (status < 0) {
 		/* James M doesn't say fuck enough. */
 		if (indev) dev_put(indev);
 		if (outdev) dev_put(outdev);
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+		if (physindev) dev_put(physindev);
+		if (physoutdev) dev_put(physoutdev);
+#endif
 		kfree(info);
 		kfree_skb(skb);
 		return;

@@ -449,7 +471,8 @@
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
 		 struct net_device *indev,
 		 struct net_device *outdev,
-		 int (*okfn)(struct sk_buff *))
+		 int (*okfn)(struct sk_buff *),
+		 int hook_thresh)
 {
 	struct list_head *elem;
 	unsigned int verdict;

@@ -481,7 +504,7 @@
 
 	elem = &nf_hooks[pf][hook];
 	verdict = nf_iterate(&nf_hooks[pf][hook], &skb, hook, indev,
-			     outdev, &elem, okfn);
+			     outdev, &elem, okfn, hook_thresh);
 	if (verdict == NF_QUEUE) {
 		NFDEBUG("nf_hook: Verdict = QUEUE.\n");
 		nf_queue(skb, elem, pf, hook, indev, outdev, okfn);

@@ -530,7 +553,7 @@
 		verdict = nf_iterate(&nf_hooks[info->pf][info->hook],
 				     &skb, info->hook, 
 				     info->indev, info->outdev, &elem,
-				     info->okfn);
+				     info->okfn, INT_MIN);
 	}
 
 	switch (verdict) {

@@ -248,6 +248,9 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	skb->nf_debug	  = 0;
 #endif
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	skb->nf_bridge	  = NULL;
+#endif
 #endif
 #ifdef CONFIG_NET_SCHED
 	skb->tc_index	  = 0;

@@ -327,6 +330,9 @@
 	}
 #ifdef CONFIG_NETFILTER
 	nf_conntrack_put(skb->nfct);
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	nf_bridge_put(skb->nf_bridge);
+#endif
 #endif
 	skb_headerinit(skb, NULL, 0);  /* clean state */
 	kfree_skbmem(skb);

@@ -392,6 +398,9 @@
 #ifdef CONFIG_NETFILTER_DEBUG
 	C(nf_debug);
 #endif
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	C(nf_bridge);
+#endif
 #endif /*CONFIG_NETFILTER*/
 #if defined(CONFIG_HIPPI)
 	C(private);

@@ -404,6 +413,9 @@
 	skb->cloned = 1;
 #ifdef CONFIG_NETFILTER
 	nf_conntrack_get(skb->nfct);
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	nf_bridge_get(skb->nf_bridge);
+#endif
 #endif
 	return n;
 }

@@ -437,6 +449,10 @@
 	nf_conntrack_get(new->nfct);
 #ifdef CONFIG_NETFILTER_DEBUG
 	new->nf_debug	= old->nf_debug;
+#endif
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	new->nf_bridge	= old->nf_bridge;
+	nf_bridge_get(new->nf_bridge);
 #endif
 #endif
 #ifdef CONFIG_NET_SCHED

@@ -396,6 +396,10 @@
 	/* Connection association is same as pre-frag packet */
 	to->nfct = from->nfct;
 	nf_conntrack_get(to->nfct);
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	to->nf_bridge = from->nf_bridge;
+	nf_bridge_get(to->nf_bridge);
+#endif
 #ifdef CONFIG_NETFILTER_DEBUG
 	to->nf_debug = from->nf_debug;
 #endif

@@ -289,6 +289,18 @@
 	       loginfo->prefix,
 	       in ? in->name : "",
 	       out ? out->name : "");
+#if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)
+	if ((*pskb)->nf_bridge) {
+		struct net_device *physindev = (*pskb)->nf_bridge->physindev;
+		struct net_device *physoutdev = (*pskb)->nf_bridge->physoutdev;
+
+		if (physindev && in != physindev)
+			printk("PHYSIN=%s ", physindev->name);
+		if (physoutdev && out != physoutdev)
+			printk("PHYSOUT=%s ", physoutdev->name);
+	}
+#endif
+
 	if (in && !out) {
 		/* MAC logging for input chain only. */
 		printk("MAC=");

I'm fine with this patch

Please send the patch to me for review. I will pass it on very quickly once I have received it.

Live long and prosper
- Harald Welte / [email protected]               http://www.gnumonks.org/
============================================================================
"If this were a dictatorship, it'd be a heck of a lot easier, just so long
 as I'm the dictator."  --  George W. Bush Dec 18, 2000

From: Bart De Schuymer <[email protected]>
Date: Fri, 25 Oct 2002 08:01:16 +0200

The following patch deals with the problems you still had with the earlier one. Changes:
1. add #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) everywhere
2. don't touch ip_tables.c
3. no ipt_physdev.c file yet. I'll try to make it this weekend.

I've applied this to my tree, thanks.

Hello netfilter team and others,

DaveM suggested I talk to you (netfilter team) about this.

What's bridge-netfilter: the mapping of the IPv4 onto the bridge hooks, to make a powerful bridging firewall.

The problem: in the current br-nf patch we add 3 fields to the skbuff, which is (as expected) not acceptable. So we need to find a way to solve this.

We cannot use the control buffer to save this data because tcp uses it while we still need the brnf data.

The solution I like best (and David seems to not mind) is adding one pointer to a struct nf_bridge_info in the skbuff. So, adding one new member.

Another suggestion by David is this:

struct nf_ct_info {
	union {
		struct nf_conntrack *master;
		struct nf_bridge_info *brinfo;
	} u;
};

But I don't think this will not work because master will be in use while we need brinfo.

So another solution could be this:

struct nf_ct_info {
		struct nf_conntrack *master;
		struct nf_bridge_info *brinfo;
};

But I don't know anything about the intricacies of adding this.

Do you have any other suggestions? Comments? Help?
The current patch was already posted on lkml so I won't repeat it.

Also, could you have a look at the current patch, to spot any other obstacles/things you don't like?
The patch is available at:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre1-against-2.5.42.diff
There is a little text file explaining the source code more in-depth here:
http://users.pandora.be/bart.de.schuymer/ebtables/br-nf/bridge-nf-0.0.10-dev-pre1-against-2.5.39-comments.txt
A high-level explanation of what we're doing is here:
http://users.pandora.be/bart.de.schuymer/ebtables/br_fw_ia/br_fw_ia.html

Another question:
I've been told it is the general concensus that this bridge firewall should be compiled in the kernel if CONFIG_NETFILTER=y. Or should it be a user option? It is predicted that using a user option will give alot of questions about the bridge firewall not working.
Do you have any strong opinion about this?

cheers,
Bart

Hi Bart.

Thanks, unfortunately there was this huge delay before your email was answered. Sorry for this inconvenience.

Yes, I am aware of this project. In fact, Lennert did a presentation about this on the netfilter developer workshop in Nov 2001

That would be acceptable.

no, it will not work. It clashes as soon as you want to use connection tracking on a bridging firewall.

This is, of course, possible.

I don't see any big problems. If master stays the first member in struct nf_ct_info, it should even work if there is a mis-use somewhere in the code, referencing directly to nf_ct_info instead of nf_ct_info->master.

It's great seeing that the bridging stuff finally gets included.

I have read the new (2.5.44) patch.

The only issue that comes to my mind is:

ip_packet_match is getting a way too long argument list. Can't you write the matching against physical in/out devices as iptables match extension? (like ipt_physdev.c?)

Mh. Since bridging firewall is cool, but not something everybody will use by default [and it adds code as well as enlarges the skb], I think it should be a compiletime kernel config option.

Live long and prosper
- Harald Welte / [email protected]               http://www.gnumonks.org/
============================================================================
"If this were a dictatorship, it'd be a heck of a lot easier, just so long
 as I'm the dictator."  --  George W. Bush Dec 18, 2000

From: Harald Welte <[email protected]>
Date: Thu, 24 Oct 2002 10:16:56 +0200

Mh. Since bridging firewall is cool, but not something everybody will use by default [and it adds code as well as enlarges the skb], I think it should be a compiletime kernel config option.

This was my initial reaction, but both of us misunderstand what is going on I think.

If you use bridging, using netfilter on the bridged traffic "is not possible" without these bridge-netfilter changes.

So he's saying, if we have bridging enable and netfilter, should bridge-netfilter be on, and right now I say yes.

Bart, correct me if I'm wrong.

Ok ;)

Yes, this does perfectly make sense. This is the way to go.

Live long and prosper
- Harald Welte / [email protected]               http://www.gnumonks.org/
============================================================================
"If this were a dictatorship, it'd be a heck of a lot easier, just so long
 as I'm the dictator."  --  George W. Bush Dec 18, 2000

From: Simon Kirby <[email protected]>
Date: Thu, 12 Sep 2002 16:49:22 -0700