Re: bonding: don't increase rx_dropped after processing LACPDUs

[Jiri Bohac <jbohac@suse.cz>]

On Wed, May 02, 2012 at 04:10:39PM -0700, Jay Vosburgh wrote:
> 	Won't this make it impossible to bind a PF_PACKET socket to
> sll_protocol == ETH_P_SLOW and see the LACPDUs, but only when bonding is
> running 802.3ad? 

yes it will...

> This because the ptype_all check in
> __netif_receive_skb happens before the rx_handler, but the ptype_base
> check (bound packet socket, for example) happens after.  Currently,
> libpcap looks to bind to ETH_P_ALL, so it won't be affected.

Does it make any sense for the packet socket not to bind to
ETH_P_ALL? With all the rx_handlers interfering with the packet
(e.g. modifying skb->dev, modifying the MAC address), binding to
ETH_P_SLOW will never reliably get you the the original frame
with reliable information about the incoming device. 

> If so, is that something we care about?

I think not.
 

-- 
Jiri Bohac <jbohac@suse.cz>
SUSE Labs, SUSE CZ