IP Datapath Refactoring [PSARC/2009/331 FastTrack timeout 06/09/2009]

[James Carlson]

Erik Nordmark writes:
> James Carlson wrote:
> > I'd suggest both route(1M) and route(7P).  The latter is what routing
> > protocol authors are supposed to be reading.
> 
> OK.
> Do you want to see draft man pages? (It would essenntially be my "two 
> things" text above.)

No need; the discussion has cleared this up for me.

> > Yep; that's how BSD works.
> > 
> > It has RTF_CLONING for the former route, to indicate that when you
> > match it, you need to create a cloned route, and RTF_CLONED to mark
> > the entries that were created by the cloning process.
> > 
> >   http://www.daemon-systems.org/man/route.8.html
> 
> And I think RTF_CLONED maps to 'W' if I don't misremember. Would it make 
> sense to use 'W' instead of 'C' for Solaris in this context?

The table that I know about is in here:

  http://cvsweb.netbsd.org/cgi-bin/cvsweb.cgi/~checkout~/src/usr.bin/netstat/netstat.1?rev=1.51&content-type=text/plain

... and it maps RTF_CLONING to 'C' and RTF_CLONED to 'c'.  I don't
know of a flag that maps to 'W', but I guess it's possible that was
done on some system.

Lining up with BSD would be a nice thing, if possible, but if we have
to be different, I wouldn't complain too much.

> Note that right now the refactor-gate implementation for local 
> connections doesn't have visibility to the port numbers when it does the 
> ECMP behavior. We'd to fix that to get better spreading.

I'm not sure I follow.  I thought that the selection mechanism for
local connections was described as round-robin.  How would port
numbers or any ECMP get involved with that?

(I agree that if you're doing ECMP, the more flow-identifying stuff
you can get in there, the better.  But that might be tending towards
design review ...)

> > What other code (besides the TCP/IP stack) sends M_MULTIDATA?  It was
> > a private interface, and I don't know of any other users.
> > 
> > If we're disabling multidata, then why not start the process of
> > removing this old stuff?  We could at least start the notification
> > process on the MDT contracts.
> 
> Not enough hours in a day, and I don't want to spend those few hours on 
> debating whether or not Cassini is the best Ethernet driver ever.
> (As far as I know Cassini is the only driver that uses multidata.)

At least to me, that's not quite the point.  We're _breaking_ that
previous feature by permanently disabling it.  I think that may well
be a good thing -- I can believe that we get acceptably good
performance without the complication of MDT -- but I don't think it's
useful to say that those things are still "supported" but simply never
work.

Would we say we "support" LSO but then never allow anyone to use it?

I think doing this cleanly means obsoleting those old interfaces
properly.

-- 
James Carlson, Solaris Networking              <james.d.carlson at sun.com>
Sun Microsystems / 35 Network Drive        71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677