(usagi-users 02131) USAGI STABLE4 binary package fails DAD tests on TAHI

[MAEDA Naoaki <maeda.naoaki@xxxxxxxxxxxxxx>]

Hi,

I am a newbie of this list.

I have tested the binary package of USAGI STABLE Release4
for RedHat using TAHI. The results are, some of the items in
Stateless Address Autoconfiguration related to DAD are failed.

One of the reason is, it seems that Linux-box ignores
(faked) DAD NA from TN because of the side effect of
CONFIG_PCNET32_VMWARE=y setting as following:

--------- in net/ethernet/eth.c
unsigned short eth_type_trans(struct sk_buff *skb, struct net_device *dev)
{
...
...

#ifdef CONFIG_PCNET32_VMWARE
        /*      For VMware hack.
         *      VMware network interfaces picks up MULTICAST packet
         *      which is generated by themselves.
         */
        if(*eth->h_dest&1)  /* BROADCAST or MULTICAST Packet */
        {
                if(memcmp(dev->dev_addr,eth->h_source, ETH_ALEN)==0)
                        return 0;
        }
#endif
....
----------

Needless to say, DAD NA is a multicast packet and its source address
is the same as Linux-box (that is why DAD NA packet is sent), so it also
hits this condition then it is dropped by net_rx_action().

I am not sure the reason why CONFIG_PCNET32_VMWARE has been set in
the binary package, but if it is not an intention, it should be off 
in the next binary release.

Regards,

MAEDA Naoaki