af_packet.c -ERRMSGSIZE logic error for VLAN Support

[jmerkey]

I guess a RAW packet interface needs to really be just that.  The af_packet.c logic is broken
with VLAN tagging.  When I attempt to regenerate packets through ethernet with 
AF_PACKET calls via sockaddr_ll request structures, the calculation being performed in
af_packet of:

err = -ERRMSGSIZE
if (len > (dev->mtu + dev->hard_header_len))
    goto out_unlock;

...
..

return err;

This is busted with VLAN tags.  I am sending packets out of the max length of 1514 with the
additional 4 bytes of VLAN info attached.  I am received packets of this size from the 
e1000 adapters (1518) from one side of the system and when I attempt to resend them 
via af_packet.c through another e1000 adapter with the ame driver code, it bombs since the packet is larger than the mtu (1500) + dev->hard_header_len (14).  A RAW packet interface needs to really be raw, and this includes taking into account the size of a VLAN header.  I realize this is a driver error since the driver returns these fields from the net_device structure, but it doesn't make much sense for a driver to enforce these limits in raw mode.  

manipulation of VLAN tags should be allowed through a raw interface, and the logic in 
af_packet.c is what is creating the error.  If this an error in the logic or is there some other
method recommended for dealing with the VLAN tagging info.

I would assume for IP and other types of sockets, some sort of sideband is ok for tagging 
the packets, but a raw interface should allow raw packet construction and transmission.

Jeff




-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/