On 30.05.2013 18:55, Jesper Dangaard Brouer wrote:

On Thu, 30 May 2013 07:39:10 -0700
Eric Dumazet [off-list ref] wrote:

quoted

On Thu, 2013-05-30 at 09:51 +0200, Jesper Dangaard Brouer wrote:

quoted

On Wed, 29 May 2013 08:52:04 -0700
Eric Dumazet [off-list ref] wrote:

quoted

quoted

I am not sure it will solve the ATM logic (with the 5 bytes
overhead per 48 bytes cell)

Are you talking about, that for GSO frames we are not adding a encap
overhead to each "sub" skb.

This part is now done properly in qdisc_pkt_len_init() since linux-3.9

Thanks for the pointer, but qdisc_pkt_len_init() only adds the
EthMAC+IP+TCP header size for each GSO segment (stored in
qdisc_skb_cb(skb)->pkt_len).
It is still missing the AAL5 encapsulation overhead per GSO segment.

Besides I can see that __qdisc_calculate_pkt_len() "forgets" this
information and overwrites qdisc_skb_cb(skb)->pkt_len (iif a stab is
defined on the qdisc).

Maybe change qdisc_pkt_len_init() to take the additional overhead value as
input and change __qdisc_calculate_pkt_len() use

	qdisc_pkt_len_init(skb, stab->szopts.overhead);

instead of

	pkt_len = skb->len + stab->szopts.overhead;

-Jussi