Bug 28768

We have router on FreeBSD 4.3-stable with 2 network cards (IP only). 
The networks are divided on two parts: one card (fxp0) has addreses from
1 to 126, the other card (rl0) has addreses from 129 to 254. The first
interface has address .2, the second card has .129.

fxp0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
	inet xxx.xxx.xxx.129 netmask 0xffffff80 broadcast xxx.xxx.xxx.127
	ether 00:a0:c9:6e:dc:80 
	media: autoselect (100baseTX <full-duplex>) status: active
	supported media: autoselect 100baseTX <full-duplex> 100baseTX 10baseT/UTP <full-duplex> 10baseT/UTP 100baseTX <hw-loopback>
rl0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
	inet xxx.xxx.xxx.2 netmask 0xffffff80 broadcast xxx.xxx.xxx.127
	ether 00:a0:d2:a4:19:f9 
	media: autoselect (100baseTX <full-duplex>) status: active
	supported media: autoselect 100baseTX <full-duplex> 100baseTX 10baseT/UTP <full-duplex> 10baseT/UTP 100baseTX <hw-loopback>

Routing tables
Internet:
Destination        Gateway            Flags     Refs     Use     Netif Expire
default            xxx.xxx.xxx.5      UGSc       43     7708      rl0
127.0.0.1          127.0.0.1          UH          2     9310      lo0
xxx.xxx.xxx/25     link#2             UC          1        0      rl0 =>
xxx.xxx.xxx.2/32   0:a0:d2:a4:19:f9   ULS2c       1        0      lo0
xxx.xxx.xxx.5/32   0:4:27:19:82:e0    ULS2c       1        0      rl0
xxx.xxx.xxx.127    ff:ff:ff:ff:ff:ff  UHLWb       0      348      rl0
xxx.xxx.xxx.128    ff:ff:ff:ff:ff:ff  UHLWb       0        1     fxp0 =>
xxx.xxx.xxx.128/25 link#1             UC          2        0     fxp0 =>
xxx.xxx.xxx.129/32 0:a0:c9:6e:dc:80   ULS2c       0        0      lo0
xxx.xxx.xxx.132/32 0:a0:d2:a4:1a:4b   ULS2c       1        4     fxp0
xxx.xxx.xxx.255    ff:ff:ff:ff:ff:ff  UHLWb       1     1039     fxp0


Using the computer (address .132) we don't meet any mistakes in
connecting with other addresses on network 1-126 (and at Internet). When
we connect to the  router (address .2, other network cards) we don't
meet any problems in connecting too. When we are trying to connect to
the router on address his interface (.129), we can not do this. Ping on
.129 work, but any TCP connect was fail.

When I used a network scaner I noticed that the control sum in the packet 
of the router wasn't correct (we always get mistake in the control sum if
the packet goes out from the address 129. The control sum of address .2 
is correct and computer connects).

I changed the network cards by places but the mistake appears again
(address .129 isn't connected).

Before upgrading of the system (4.3-release) we had no problems. 
In May after first cvs there were no nistakes. But after June cvs the
system isn't work.

Of course I made changes in system and kernel every time after cvs. Now
we have working system after cvs (with changes on current date) but
having old kernel and new modules.

The cvs protocols have a lot of changes in network part. But I can't
believe that somebody could brake something in such place like the
routing and nobody could notice that...

Fix: 

I don't know.
How-To-Repeat: I have only one router, so I don't know it.