[comp.unix.aix] IP routing subnet mask bug?

todd@janus.UUCP (Todd Booth) (10/20/89)

We have several logical networks on one physical network and are
having problems with IBM's PS/2 AIX V 1.1.  All networks have been
set up as class B (16 bits network and 16 bits host).  Of the 16 bits
host address, the AIX machines are using the first 10 as their subnet
address and the last 6 as their local host address.  

One problem is even though we set the subnet mask to 6 bits during
ifconfig, the IP routing logic seems to ignore this information.  We
can set up the route properly to other subnets via the local
interface, but applications can't find the route "unreachable".  Our
workaround to this is to route all "unreachable" IP packets via the
local interface (and hope) thats right.

The second problem is that machines on our subnet, for example,
130.224.4.1 see packets destined for other subnets, for example,
130.224.0.63.  I realize that in this case we have all 0's for
the subnet.  (Is there a way to configure AIX to ignore the *old*
style broadcast methods?)

--todd booth / Quotron

robin@itcatl.UUCP (Robin Cutshaw) (10/23/89)

We have several different types of machines on our backbone.  The only
PS/2 running AIX 1.1 seems to send a subnet mask of 0.0.0.255 when in
fact we use 255.0 0.0.  This causes at least one of our systems to
take itself off the net (for all intents and purposes).

Anyone else seen this?

robin
-- 
----------
Robin Cutshaw		{gatech,rebel}!itcatl!robin
Disc Access, Atlanta, Ga.  (404)261-1264   (formerly ITC)

todd@janus.UUCP (Todd Booth) (10/24/89)

In article <394@itcatl.UUCP>, robin@itcatl.UUCP (Robin Cutshaw) writes:
> 
> The only
> PS/2 running AIX 1.1 seems to send a subnet mask of 0.0.0.255

How did you set up AIX to perform the subnet mask server function?

> This causes at least one of our systems to
> take itself off the net (for all intents and purposes).

Assuming your system isn't diskless, just give the mask at ifconfig time.
ifconfig net0 netmask 255.255.255.0

--todd booth / Quotron

robin@itcatl.UUCP (Robin Cutshaw) (10/25/89)

In article <468@janus.UUCP>, todd@janus.UUCP (Todd Booth) writes:
> How did you set up AIX to perform the subnet mask server function?
> 
> > This causes at least one of our systems to
> > take itself off the net (for all intents and purposes).
> 
> Assuming your system isn't diskless, just give the mask at ifconfig time.
> ifconfig net0 netmask 255.255.255.0
> 

Guess I should have included more detail.  I am using ifconfig with
netmask 255.0.0.0 and broadcast MYNET.255.255.255.  ifconfig shows
the netmask to be ff000000 but it actually arp's 000000ff...

robin

-- 
----------
Robin Cutshaw		{gatech,rebel}!itcatl!robin
Disc Access, Atlanta, Ga.  (404)261-1264   (formerly ITC)

dyer@spdcc.COM (Steve Dyer) (10/25/89)

In article <397@itcatl.UUCP> robin@itcatl.UUCP (Robin Cutshaw) writes:
>Guess I should have included more detail.  I am using ifconfig with
>netmask 255.0.0.0 and broadcast MYNET.255.255.255.  ifconfig shows
>the netmask to be ff000000 but it actually arp's 000000ff...

Uh, I don't know what you mean by "arp"ing 000000ff, but...  I'd ask:

What's your (Class A) net number for MYNET?  Strictly speaking, most systems
derived from the 4.x network code don't require you to specify a netmask if
your network convention is the same as the "natural" netmask for your network
number.

Urgh--that's awkwardly phrased.
Class A networks (001-127.h.h.h) have a "natural" netmask of 255.0.0.0.
Class B networks (128-191.n.h.h) have a "natural" netmask of 255.255.0.0, and
class C networks (192-223.n.n.h) have a "natural" netmask of 255.255.255.0.

I'm running with a class A net address (net 18) with a netmask of 255.255.0.0
(16 bits of host) with AIX PS/2, and it works flawlessly.  I suspect you're
neglecting to mention some bit of information here.

-- 
Steve Dyer
dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer
dyer@arktouros.mit.edu, dyer@hstbme.mit.edu

lcc@seashell.seas.ucla.edu (Locus Computing Corporation) (11/11/89)

In article <328@ursa-major.SPDCC.COM> dyer@ursa-major.spdcc.COM (Steve Dyer) writes:
>In article <397@itcatl.UUCP> robin@itcatl.UUCP (Robin Cutshaw) writes:
>>Guess I should have included more detail.  I am using ifconfig with
>>netmask 255.0.0.0 and broadcast MYNET.255.255.255.  ifconfig shows
>>the netmask to be ff000000 but it actually arp's 000000ff...
>
	Well, I tried to respond to this ages ago, but it looks like it
never made it out.  First off the system doesn't arp ff000000 or 000000ff, but
it DOES respond to ICMP mask request/responses.  As it turned out there
were TWO bugs w.r.t. ICMP mask responses.  1)  The subnet mask was NOT being
used, consequently the requester only saw the "normal" mask (i.e. no subnet
mask in use) which is potentially wrong.  2)  The mask was not sent out
in the canonical IP byte ordering (leastwise not on little-endian machines).
	Both of these problems HAVE been fixed in version 1.2 (at least that
is what I THINK IBM is calling it this week).  A bug fix should be available
for your first generation system.  Check with your SE.

							Brian Horn
							PMTS AIX development
							Locus Computing Corp.