Date: Tue, 07 Sep 1999 12:29:10 +0100 From: Brian Somers <brian@Awfulhak.org> To: "Rodney W. Grimes" <freebsd@gndrsh.dnsmgr.net> Cc: brian@FreeBSD.org (Brian Somers), cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/usr.sbin/ppp command.c ip.c ipcp.c ipcp.h ppp.8 Message-ID: <199909071129.MAA87364@keep.lan.Awfulhak.org> In-Reply-To: Your message of "Tue, 07 Sep 1999 01:21:44 PDT." <199909070821.BAA05060@gndrsh.dnsmgr.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> > brian 1999/09/07 00:51:12 PDT
> >
> > Modified files:
> > usr.sbin/ppp command.c ip.c ipcp.c ipcp.h ppp.8
> > Log:
> > Introduce a forth IP packet queue. Urgent packets with
> > ip_tos == IPTOS_LOWDELAY now get precidence over urgent
> > packets with ip_tos != IPTOS_LOWDELAY and non-urgent packets
> > with ip_tos == IPTOS_LOWDELAY.
>
> I'm getting a bit concerned here, what is the depth of these
> queues? If it is the default of 50, then we now eat a lot
> of mbuf space should the ppp interface stall (and, trust me,
> they do stall). I'm not really worried about the memory,
> thats cheap enough, I'm worried about the amount of time
> it will take for the higher priority queues to drain so that
> the low priority traffic goes out after a stall. (Note, these
> stalls are usually cause on ISDN TA's with Additional Call Offering
> or Additional Outbound Call features running in ISDN Bonding mode.)
>
> It may require me to retune some OSPF timing parameters...
> or mark the packets high priority (which, they really aren't)
Ppp will eat up to 30 packets from the tun device (if you're using
MP, add (2 * links) to that). These packets will sit in one of the
four IP queues.
The highest priority queue is reserved for link level traffic.
There's no such thing in terms of IP traffic, so I'm sure you'll now
spot the deliberate mistake :-I
The second highest is reserved for ``urgent'' packets that have the
ip_tos also set to LOWDELAY.
The third highest is reserved for both ``urgent'' traffic and traffic
with the ip_tos set to LOWDELAY.
The lowest is reserved for everything else.
Once the total packets in these four (soon to become 3) queues
reaches the above number, ppp stops reading stuff from the tun
device. If nothing can be sent within the ``choked'' time (default 2
minutes), ppp dumps the packets and starts reading again.
If any of the link descriptors become writable (as reported by
select()), ppp pushes a packet from the head of the highest priority
IP queue, resets the choked timer and read()s more from the tun
device.
There's no overhead (apart from the minor ``find which queue has
data'' search) in these extra queues. The extra one (added with this
commit) is there so that (for example) telnet trafic that's marked as
LOWDELAY will actually skip other telnet trafic.
I think the only point worth arguing over is the previous commit that
changed the magic number of packets from 20 to 30.
> > Enhance the ``set urgent'' syntax to allow for urgent UDP
> > packets as well as urgent TCP packets.
>
> And is there a knob to turn this all off and go back to 1 or
> 2 queues only?
No. I don't think we need one.
> --
> Rod Grimes - KD7CAX - (RWG25) rgrimes@gndrsh.dnsmgr.net
>
--
Brian <brian@Awfulhak.org> <brian@FreeBSD.org>
<http://www.Awfulhak.org>; <brian@OpenBSD.org>
Don't _EVER_ lose your sense of humour ! <brian@FreeBSD.org.uk>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199909071129.MAA87364>