The PPP interface strips off the PPP header from the packet before passing it up with netif_rx(). On the xmit path, it has to add the PPP header itself because dev->header_ops is NULL.
This means that a PF_PACKET, SOCK_RAW socket can't get the link level header and has to use the sll_protocol field of the sockaddr_ll to know what type of packet is received with recvfrom(2). I consider this a design flaw since with most (all?) other interfaces you only need to know the sll_hatype to know what type of packet you get. I have patched the PPP code to include the PPP header. I tried with IP and IPv6 over PPP and it works as expected. This patch, however, changes the interface to the user space in an incompatible way. But IMHO we could include it, since * PF_PACKET is Linux specific and not portable anyway. Most apps use libpcap instead of opening PF_PACKET sockets themselves. * Using strace on tcpdump, it seems that libpcap on Linux uses PF_PACKET/SOCK_DGRAM for PPP interfaces and thus is not affected by my patch. * Other apps using PF_PACKET/SOCK_RAW can easily be changed to PF_PACKET/SOCK_DGRAM if they don't want to see the link level header. After all, this is what SOCK_DGRAM is for. Currently SOCK_RAW and SOCK_DGRAM are the same although the packet(7) man page states that with SOCK_RAW packets are passed without any changes. This makes having SOCK_RAW besides SOCK_DGRAM useless for PPP. So what is your opinion about this change? Signed-off-by: Urs Thuermann <[EMAIL PROTECTED]> --- drivers/net/ppp_generic.c | 41 ++++++++++++++++++++++++++++------------- 1 file changed, 28 insertions(+), 13 deletions(-) Index: net-2.6/drivers/net/ppp_generic.c =================================================================== --- net-2.6.orig/drivers/net/ppp_generic.c 2008-02-08 11:09:03.000000000 +0100 +++ net-2.6/drivers/net/ppp_generic.c 2008-02-08 13:27:29.000000000 +0100 @@ -873,12 +873,32 @@ /* * Network interface unit routines. */ + +/* Put the 2-byte PPP protocol number on the front of skb */ +static int ppp_header(struct sk_buff *skb, struct net_device *dev, + unsigned short type, + const void *daddr, const void *saddr, unsigned len) +{ + unsigned char *pp; + int npi, proto; + + npi = ethertype_to_npindex(ntohs(skb->protocol)); + if (npi < 0) + return -dev->hard_header_len; + + pp = skb_push(skb, 2); + proto = npindex_to_proto[npi]; + pp[0] = proto >> 8; + pp[1] = proto; + + return dev->hard_header_len; +} + static int ppp_start_xmit(struct sk_buff *skb, struct net_device *dev) { struct ppp *ppp = (struct ppp *) dev->priv; - int npi, proto; - unsigned char *pp; + int npi; npi = ethertype_to_npindex(ntohs(skb->protocol)); if (npi < 0) @@ -897,16 +917,6 @@ goto outf; } - /* Put the 2-byte PPP protocol number on the front, - making sure there is room for the address and control fields. */ - if (skb_cow_head(skb, PPP_HDRLEN)) - goto outf; - - pp = skb_push(skb, 2); - proto = npindex_to_proto[npi]; - pp[0] = proto >> 8; - pp[1] = proto; - netif_stop_queue(dev); skb_queue_tail(&ppp->file.xq, skb); ppp_xmit_process(ppp); @@ -969,9 +979,14 @@ return err; } +static const struct header_ops ppp_header_ops ____cacheline_aligned = { + .create = ppp_header, +}; + static void ppp_setup(struct net_device *dev) { dev->hard_header_len = PPP_HDRLEN; + dev->header_ops = &ppp_header_ops; dev->mtu = PPP_MTU; dev->addr_len = 0; dev->tx_queue_len = 3; @@ -1677,10 +1692,10 @@ kfree_skb(skb); } else { /* chop off protocol */ + skb_reset_mac_header(skb); skb_pull_rcsum(skb, 2); skb->dev = ppp->dev; skb->protocol = htons(npindex_to_ethertype[npi]); - skb_reset_mac_header(skb); netif_rx(skb); ppp->dev->last_rx = jiffies; } -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html