[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: AltQ




On 26-Jan-00 Chris Cappuccio wrote:
> I ported ALTQ to OpenBSD several months ago, based mostly on the code used to
> run ALTQ on NetBSD.  I submitted this back to the author, through his work it
> evolved along with several other parts of ALTQ to become version 2.1.

Yes Chris, I remember seeing it on the list someday (on Oct 12th 1999), but
it is not a real port (in the port tree) yet, as it needs kernel rebuild. (No
offense)
And indeed, in AltQ 2.1, they support several cards based on stock 2.6 .

> I asked Kenjiro Cho, author of ALTQ, why not do its work at ether_input() and
> ether_output(), and he said:
> 
> "Packet scheduling should be done from a driver interrupts (tx complete
> interrupts) and it is if_start (xx_start in the driver) in the BSD systems."

Kenjiro told us about the interrupts, and also about a if_done hook (wich is
obviously found nowhere, except for ip_rsvp_vif_done in netinet mroute.[ch] )


Then, I would suggest envisaging the integration of AltQ into the base source
(despite Theo's reluctancy to change the network drivers' API), at least the
kernel parts.
For the userland parts, we'll have to see (on the long run) if we can get rid
of an altqd daemon, and try to merge the controls with ipfilter (which, btw, is
not IP centric only :)


        Thierry