On Sun, May 12, 2013 at 03:25:10PM +0800, Marek Lindner wrote:

...

On Tuesday, May 07, 2013 19:54:02 Antonio Quartulli wrote:

...

In the gateway code in case of VLAN tagged frame the ethhdr pointer is moved forward by 4 bytes so that the offset of h_proto in struct ethhdr matches the real h_vlan_encapsulated_proto address in the skb.

This trick is correct but the code is not easy to understand and may lead to bugs in case of re-use of ethhdr for other purposes.

Use a standalone protocol variable and assign it accordingly to the header type

How about adding the key word "refactoring" somewhere into the subject line or the commit message ?

oky

...

...

@@ -626,23 +626,29 @@ bool batadv_gw_is_dhcp_target(struct sk_buff *skb, unsigned int *header_len) struct iphdr *iphdr; struct ipv6hdr *ipv6hdr; struct udphdr *udphdr;

• uint16_t proto;

  /* check for ethernet header */ if (!pskb_may_pull(skb, *header_len + ETH_HLEN)) return false; ethhdr = (struct ethhdr *)skb->data;

• proto = ntohs(ethhdr->h_proto); *header_len += ETH_HLEN;

  /* check for initial vlan header */

• if (ntohs(ethhdr->h_proto) == ETH_P_8021Q) {

• if (proto == ETH_P_8021Q) {

• struct vlan_ethhdr *vhdr;
  

While we are refactoring I suggest to not ntohs() the protocol every time but the protocol constants. The compiler will replace the constants with the appropriate number during the compilation, thus saving us from a runtime operation every time we check a packet.

oh right. I did not think about it. I guess I have to use __constant_htons instead of the plain hntons.

cheers,

-- Antonio Quartulli

..each of us alone is worth nothing.. Ernesto "Che" Guevara