On Mon, Nov 27, 2006 at 09:15:40PM +0100, Tony Sarendal wrote:
Typos...
Bad cut and pasting.
>
> I have a suggestion for a feature for pf, adaptable tbr.
>
> Instead of doing estimation of overhead of a network link like
> a dsl uplink the tbr can be modified to adapt instead.
>
> In my case I have run PPPoE over a DSL link that is AAL5 LLC/SNAP.
> In altq_subr.c I changed:
> tbr->tbr_token -= TBR_SCALE(m_pktlen(m));
>
> to:
> if (m_pktlen(m) <= 38)
if (m_pktlen(m) > 38)
> tbr->tbr_token -=
> TBR_SCALE((m_pktlen(m)+8+3+3+2+2+18+6+2+47)/48*53);
> else
> tbr->tbr_token -= TBR_SCALE(106);
>
>
> my traffic shaper now works well even in the extreme cases where
> DSL overhead is 165% (TCP ACK, 40 byte IP => 106 bytes ATM).
>
> What would be really nice would be to be able to configure this
> from pf.conf, something like
> queue vr0_blk tbradapt PPPOE-LLCSNAP-AAL5 ...
>
> It would be simple to add more adapters for the other media once
> this is in place.
>
> RFC1483 bridging over AAL5 LLC/SNAP:
> if (m_pktlen(m) <= 46)
if (m_pktlen(m) > 46)
> tbr->tbr_token -=
> TBR_SCALE((m_pktlen(m)+8+3+3+2+2+18+47)/48*53);
> else
> tbr->tbr_token -= TBR_SCALE(106);
>
> or PPP over AAL5 VCMUX:
> tbr->tbr_token -= TBR_SCALE((m_pktlen(m)+8+2+47)48*53);
>
> and so on.
>
> Comments ?
>
These changes alone doesn't change the counters, so pfctl -sq -vv
still shows IP.
/Tony
--
---
Tony Sarendal - [EMAIL PROTECTED]
IP/Unix
-= The scorpion replied,
"I couldn't help it, it's my nature" =-
