regarding trill implementation

Hi,
    I have got the point where packet data is truncated. The function
nf_bridge_maybe_copy_header( ) in  br_dev_queue_push_xmit( ) actually
truncates it.
Can you please tell me the purpose of nf_bridge_maybe_copy_headers( ).

 As far as i have understood, it is used for vlan or pppoe header
encapsulation. I don't understand the purpose of netfilters as vlan
and pppoe data is copied from skb->nf_bridge->data to skb->data but do
so in case of when netfilters are configured.

Should i put my encapsulation headers also in skb->nf_bridge->data?
And modify the function nf_bridge_encap_header_len(const struct
sk_buff *skb) by adding a "case" for trill header. So that header will
be put in skb->data in  function   skb_copy_to_linear_data_offset(skb,
-header_size, skb->nf_bridge->data, header_size);

Temporarily, i have called a function create_trill_header( ) in
br_dev_queue_push_xmit( ).

Actual is:

  int br_dev_queue_push_xmit(struct sk_buff *skb)
  {
          /* ip_fragment doesn't copy the MAC header */
          if (nf_bridge_maybe_copy_header(skb) ||
              (packet_length(skb) > skb->dev->mtu && !skb_is_gso(skb))) {
                  kfree_skb(skb);
          } else {
                  skb_push(skb, ETH_HLEN);
                  dev_queue_xmit(skb);
          }

          return 0;
  }


Now after addition of trill hooks:

int br_dev_queue_push_xmit(struct sk_buff *skb)
{
	/* ip_fragment doesn't copy the MAC header */
	if (nf_bridge_maybe_copy_header(skb) ||
	    (packet_length(skb) > skb->dev->mtu && !skb_is_gso(skb))) {
		kfree_skb(skb);
	} else {
	    printk ("Outside 1");	
			if(eth_hdr(skb)->h_proto == htons(0x0800) && skb->protocol != htons(0x22f3))
			{
				printk("Entery 1");
		
    				printk("Entery 2");
     				skb = create_trill_header(skb,false,0);
					
			}

		skb_push(skb, ETH_HLEN);
		trill_testing(skb);
		dev_queue_xmit(skb);
	}

	return 0;
}

It is working somehow fine. But i have got the problem with checks
before calling encapsulating function as shown above. I don't get upon
things i have to put in checks as given above.

Any idea and help......

How do i distinguish between packets which come from other bridges and
from it own connected hosts?

Should i extend the data structure of bridge "net_bridge_fdb_entry" by
adding trill fields as in solaris trill for generating trill data
base?

On Fri, 9 Mar 2012 23:13:42 +0500
Syed Muhammad Mohsin Kazmi [off-list ref] wrote:

Hi,
    I have got the point where packet data is truncated. The function
nf_bridge_maybe_copy_header( ) in  br_dev_queue_push_xmit( ) actually
truncates it.
Can you please tell me the purpose of nf_bridge_maybe_copy_headers( ).

Packets that go through a transparent bridge need to have special processing.
The original header is copied to an area above the header, so that netfilter
rules see the original header, not the header of destination.

    

               +----------+
               | Original |
               | Ethernet |
               | Header   |
               +----------+<-------+ skb->data
               | Bridged  |
               | Ethernet |
               | Header   |
               +----------+
               | IP and   |
               | rest     |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               |          |
               +----------+


When allocating skb's, dev_alloc_skb and other routines setup the
required padding.

 As far as i have understood, it is used for vlan or pppoe header
encapsulation. I don't understand the purpose of netfilters as vlan
and pppoe data is copied from skb->nf_bridge->data to skb->data but do
so in case of when netfilters are configured.

Should i put my encapsulation headers also in skb->nf_bridge->data?
And modify the function nf_bridge_encap_header_len(const struct
sk_buff *skb) by adding a "case" for trill header. So that header will
be put in skb->data in  function   skb_copy_to_linear_data_offset(skb,
-header_size, skb->nf_bridge->data, header_size);

Temporarily, i have called a function create_trill_header( ) in
br_dev_queue_push_xmit( ).

Actual is:

  int br_dev_queue_push_xmit(struct sk_buff *skb)
  {
          /* ip_fragment doesn't copy the MAC header */
          if (nf_bridge_maybe_copy_header(skb) ||
              (packet_length(skb) > skb->dev->mtu && !skb_is_gso(skb))) {
                  kfree_skb(skb);
          } else {
                  skb_push(skb, ETH_HLEN);
                  dev_queue_xmit(skb);
          }

          return 0;
  }


Now after addition of trill hooks:

Trill should be a tunnel (look at gretap) that can be added to bridge
transparently.