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

Re: ethernet routing problem?



On Mon, 3 Mar 1997, Steve Izma wrote:

> I've been setting up Debian 1.2.4 (25 Jan.: Cheapbytes distribution)
> on two new Pentium 150 systems and I can't get network routing over
> ethernet to work.
>
>  Installation of netbase and netstd seemed to go well using dselect,
> except for an unsurprising temporary problem in finding the right i/o
> port for the ethernet card. I'm using the D-link DE220P, which the ne
> driver easily finds. Ifconfig gives this report, which I believe shows
> correct configuration of the driver to the card:
>
> eth0      Link encap:10Mbps Ethernet  HWaddr 00:80:C8:2D:7D:8F
>           inet addr:192.54.242.228  Bcast:192.54.242.255  Mask:255.255.255.0
>           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
>           RX packets:0 errors:0 dropped:0 overruns:0
>           TX packets:4117 errors:0 dropped:0 overruns:0
>           Interrupt:11 Base address:0x220 
> 
> Pinging the localhost works fine, but pinging anything else on the
> network (even the immediately adjacent device connected via coax)
> produces no result. I'm sure it's not a wiring problem because the
> rest of the machines on the network are not affected and I get the
> same negative results using ping on both new machines, which are wired
> into the network in different locations in the office.

does the card have more than one connector? i.e. BNC and/or RJ-45 and/or
AUI connectors?

If so, make sure that the card is correctly configured to use the right
connector for your network.

4117 TX packets and 0 RX packets makes me a bit suspicious that, e.g., the
card is configured to use the RJ-45 UTP connector when you are running on a
coax network - or vice-versa.

check the output of 'dmesg'  (or 'tail /var/log/messages') - cabling
problems like this usually show error messages in the logs.

> Trying ftp produces the error: "no route to host".
> So I assume this is some sort of routing problem. 

no, it's not routing. you get that same 'no route to host' message when
your computer can't figure out a host's ethernet address using arp. to
send a packet to a host on the local ethernet, your machine will send
out an 'arp who-has' request and wait for an 'arp is-at' reply.  This builds
up the kernel's arp table so it knows which hosts (ethernet cards, actually)
correspond to which IP addresses.
 
here's an example of what happens - excerpt of output from tcpdump:

# tcpdump -l | grep arp
    09:47:52.279492 arp who-has taz.net.au tell proxy.taz.net.au
    09:47:52.279492 arp who-has siva.taz.net.au tell proxy.taz.net.au
    09:47:52.279492 arp reply siva.taz.net.au is-at 0:0:c0:a:44:a5
    09:47:52.279492 arp reply taz.net.au is-at 0:0:c0:5:c3:14
    .
    .
    09:48:01.741658 arp who-has kali.taz.net.au tell siva.taz.net.au
    09:48:01.741658 arp reply kali.taz.net.au is-at 0:0:c0:bc:2f:41

and this is part of what the arp table looks like on siva:

# arp -a
Address                 HWtype  HWaddress           Flags Mask Iface
taz.net.au              ether   00:00:C0:05:C3:14   C     *     eth0
kali.taz.net.au         ether   00:00:C0:BC:2F:41   C     *     eth0
proxy.taz.net.au        ether   00:00:C0:9F:96:42   C     *     eth0



if your machine is sending out the arp who-has request and doesn't get
the arp reply then it will be unable to send a packet to the target host.

my guess is that it is unable to get the arp 'is-at' reply because it is
using the wrong connector.  you say your network is coax...make sure that
the card is using the BNC connector and NOT an AUI or RJ-45 connector. 

Depending on the type of card you have, you may have to configure this by
setting jumpers on the card.  Some cards can be set up with ms-dos based
configuration programs.  Some cards even have linux config tools.  Some
newer cards can auto-detect whether the network is coax or rj-45. 



> The problem is identical on both machines with the new distribution.
> I've compared everything I can think of to another machine we have
> running Debian 1.1 (dating from last June). All the files in init.d
> appear to be the same (/etc/init.d/net*). The machine with 1.1 works
> fine.

it looks like a hardware fault or hardware configuration problem to me,
not software.

how many other machines are on the network?  do you get any response if you
do a broadcast ping (i.e. ping the broadcast address 192.54.242.255)?

craig


Reply to: