Re: [Cake] [PATCH net-next v14 6/7] sch_cake: Add overhead compensation support to the rate shaper

["Toke Høiland-Jørgensen" <toke@toke.dk>]

Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> writes:

> On Tue, May 22, 2018 at 10:44:53AM +0200, Toke Høiland-Jørgensen wrote:
>> 
>> 
>> On 22 May 2018 01:45:13 CEST, Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> wrote:
>> >On Mon, May 21, 2018 at 10:35:58PM +0200, Toke Høiland-Jørgensen wrote:
>> >> +static u32 cake_overhead(struct cake_sched_data *q, const struct
>> >sk_buff *skb)
>> >> +{
>> >> +	const struct skb_shared_info *shinfo = skb_shinfo(skb);
>> >> +	unsigned int hdr_len, last_len = 0;
>> >> +	u32 off = skb_network_offset(skb);
>> >> +	u32 len = qdisc_pkt_len(skb);
>> >> +	u16 segs = 1;
>> >> +
>> >> +	q->avg_netoff = cake_ewma(q->avg_netoff, off << 16, 8);
>> >> +
>> >> +	if (!shinfo->gso_size)
>> >> +		return cake_calc_overhead(q, len, off);
>> >> +
>> >> +	/* borrowed from qdisc_pkt_len_init() */
>> >> +	hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
>> >> +
>> >> +	/* + transport layer */
>> >> +	if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 |
>> >> +						SKB_GSO_TCPV6))) {
>> >> +		const struct tcphdr *th;
>> >> +		struct tcphdr _tcphdr;
>> >> +
>> >> +		th = skb_header_pointer(skb, skb_transport_offset(skb),
>> >> +					sizeof(_tcphdr), &_tcphdr);
>> >> +		if (likely(th))
>> >> +			hdr_len += __tcp_hdrlen(th);
>> >> +	} else {
>> >
>> >I didn't see some code limiting GSO packets to just TCP or UDP. Is it
>> >safe to assume that this packet is an UDP one, and not SCTP or ESP,
>> >for example?
>> 
>> As the comment says, I nicked this from the qdisc init code.
>> So I assume it's safe? :)
>
> As long as it doesn't go further than this, it is. As in, it is just
> validating if it can contain an UDP header, and if so, account for its
> size, without actually reading the header.
>
> Considering everything !TCP as UDP work as an approximation, which is
> quite accurate. SCTP header is just 4 bytes bigger than UDP header and
> is equal to ESP header size.

Yup, that seems close enough for our purposes. Thanks for explaining.
Didn't actually know that GSO handles other protocols as well :)

-Toke