Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 14 Aug 1996 10:16:22 -0700 (PDT)
From:      Jon Inouye <jinouye@cse.ogi.edu>
To:        Nate Williams <nate@mt.sri.com>
Cc:        mobile@freebsd.org
Subject:   Re: 2.1.5 vs. 2.2-SNAP
Message-ID:  <Pine.HPP.3.95.960814095732.15198A-100000@indurain.cse.ogi.edu>
In-Reply-To: <199608141650.KAA27565@rocky.mt.sri.com>

next in thread | previous in thread | raw e-mail | index | archive | help

On Wed, 14 Aug 1996, Nate Williams wrote:

> They should be almost *exactly* the same.
> 

Okay, I'll spend some time investigating why the PC card
ioctl PIOCSDRV fails for my modem card in 2.1.5 but not 2.2.

> > If so, I may try to migrate some
> > 2.1 kernel modifications to 2.2-SNAP rather than to 2.1.5.
> 
> What kernel modifications?  The code in -current (therefore any of the
> recent SNAPS) is mostly functional already, although some of the Nomad
> 'features' don't yet exist in the -current code.  And, given the lack of
> interest and help I've received it's not obvious when/if the code will
> ever make it into -current.  (Yes, I'm a bit frustrated!)
> 

I have some kernel modifications to support physical media independence,
something folks at Intel call "media switching".  It adjusts the network
stack to survive changes in network interfaces --- enabled through
PC card technology.  Some of this can be accomplished using a modified
arp (to filter out invalid ARP entries) and route flush but doing so
at user level has some synchronization problems --- the kernel is a
better place for atomic operations.  I also reset TCP retransmission
timers to avoid the backoff when no interface is available.
Resetting retransmission timers on correspondent hosts is on the
todo list; I plan on using Ramon Caceres' triple ack hack.  Other
dependencies on ifnet pointers include multicast groups and tunnels,
the BPF, and routed/gated.  Migrating tunnels is impossible
with most implementations of mrouted (fixed endpoints) and breaks
the concept of efficient multicasting.  Migrating mcast groups and BPF
filters to different interfaces may not be a good idea.  Opinions?

Then there's Jim Binkley's Mobile IP modifications, which allow survival
across IP address changes (assuming you have connectivity to your home
agent).  Otherwise, long-term TCP/UDP connections like telnet, X11,
and a certain distributed real-time MPEG player break when the new
interface has a different IP address.

-JI




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.HPP.3.95.960814095732.15198A-100000>