RH9 ethernet assistance please..

Quick scenario: Pentium 200 MMX w/128M RAM and 2 2g SCSI drives, Was
running RH 7.2 for almost 2 years with no problems. Decided to
"upgrade" to RH9 after backing up the entire system to 8mm tape.

Upgrade failed (twice) so I nuked the drives and did a fresh
workstation install with a bit of customization. Worked fine, X came
right up I can log in and "play".

Went to set up the ethernet connection using an old (circa 1990)
WD8013 card - same card that worked fine under RH7.2 and intially had
problems getting the driver to load with IRQ15, IO 0x240 and Mem
0xC8000. Read somewhere the newer kernels like to see IRQ 10 and mem
0xD0000 with these old nonPNP ISA cards. Changed the card to IRQ 10,
IO 0x280 and Mem 0XD0000

modprobe seemed to like that combo and loaded all the appropriate
modules (8390.o and wd.o) with no complaints. However, when I check
the status of the card, I get this:

ifconfig eth0
eth0
Link encap:Ethernet HWaddr 00:00:C0:64:7C:2B
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:0 errors:0 dropped:153 overruns:0 frame:0
TX packets:0 errors:12 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
RX bytes:0 (0.0 b) TX bytes:3192 (3.1 Kb)
Interrupt:10 Base address:0x290 Memory:d0000-d4000

Note the IO address of 0x290. Wierd 'cuz the card is at 0x280 and
this brings up the 1st question. What IO should I use for this card?
RH7.2 ran just peachy with this very card at IRQ15, IO 0x240 and Mem
0xC8000. Runs fine if I toss in a flopy with an ODI stack and boot
off that The 153 dropped RX packets raises an eyebrow as well.

Since I'd set this thing up to use DHCP for addressing (Served by a
Winblows NT 4 server running the DHCP server) I figured I'd try
downing the interface and bringing it back down. (I also restarted
the Network service) Here's what ifup netted me:

Jul 9 09:51:22 localhost ifup:
Jul 9 09:51:22 localhost ifup: Determining IP information for eth0...
Jul 9 09:51:26 localhost dhclient: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 8
Jul 9 09:51:34 localhost dhclient: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 13
Jul 9 09:51:34 localhost kernel: NETDEV WATCHDOG: eth0: transmit
timed out
Jul 9 09:51:47 localhost dhclient: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 18
Jul 9 09:52:05 localhost dhclient: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 15
Jul 9 09:52:05 localhost kernel: NETDEV WATCHDOG: eth0: transmit
timed out
Jul 9 09:52:20 localhost dhclient: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 7
Jul 9 09:52:27 localhost dhclient: No DHCPOFFERS received.
Jul 9 09:52:27 localhost ifup: failed.
Jul 9 09:52:27 localhost network: Bringing up interface eth0: failed

I then tried using dhclient directly and "told' it where to find the
DHCP server: (192.168.1.10)

Jul 9 09:53:44 localhost dhclient:
Jul 9 09:53:46 localhost dhclient: Listening on
LPF/eth0/00:00:c0:64:7c:2b
Jul 9 09:53:46 localhost dhclient: Sending on
LPF/eth0/00:00:c0:64:7c:2b
Jul 9 09:53:46 localhost dhclient: Sending on Socket/fallback
Jul 9 09:53:47 localhost dhclient: DHCPDISCOVER on eth0 to
192.168.1.10 port 67 interval 5
Jul 9 09:53:47 localhost kernel: NETDEV WATCHDOG: eth0: transmit
timed out
Jul 9 09:53:52 localhost dhclient: DHCPDISCOVER on eth0 to
192.168.1.10 port 67 interval 7
Jul 9 09:53:59 localhost dhclient: DHCPDISCOVER on eth0 to
192.168.1.10 port 67 interval 18
Jul 9 09:53:59 localhost kernel: NETDEV WATCHDOG: eth0: transmit
timed out
Jul 9 09:54:17 localhost dhclient: DHCPDISCOVER on eth0 to
192.168.1.10 port 67 interval 19
Jul 9 09:54:36 localhost dhclient: DHCPDISCOVER on eth0 to
192.168.1.10 port 67 interval 12
Jul 9 09:54:36 localhost kernel: NETDEV WATCHDOG: eth0: transmit
timed out
Jul 9 09:54:48 localhost dhclient: No DHCPOFFERS received.
Jul 9 09:54:48 localhost dhclient: No working leases in persistent
database - sleeping.

In both instances, I can watch the DHCP console on the NT machine and
see the server pick up the broadcast and assign an IP. The DHCPOFFER
never seems to get back to the RH machine tho.

I created a reservation for this machine and again, I can see the DHCP
server assign the IP ("reservation in use" message) so it appears as
if part of the process is working.

I thought maybe iptables was stopping the inbound stuff so I opened it
up all the way by using the security config tool and telling it "No
firewall".

iptables -L

Chain INPUT (policy ACCEPT) target
prot opt source destination
Chain FORWARD (policy ACCEPT) target
prot opt source destination
Chain OUTPUT (policy ACCEPT) target
prot opt source destination


For giggles, I tried setting the ipstatically with ifconfig. The
address took ok, so I restarted the network service and tried to get
off the box. No joy. I can successfully ping localhost (127.0.0.1)
and the NIC itself (192.168.1.100) but can't ping anything outside the
box. I tried to ping the thing from one of my other machines with no
success either. I tried connecting this machine directly to the NT
box with a crossover cable and met with the same (lack of) results.
Booting off the ol' ODI floppy, I can ping the NT server with this
config so I know my cable is ok.

I'm fresh out of ideas and have exhausted combinations of IRQ, IO
address and base mem addresses with this card.. The *only* one that
will actually load the modules is IRQ 10, IO 0x280 and mem 0xD0000


Here are some config files for your perusal. maybe I'm just not
seeing something obvious:

/etc/dhclient-eth0.conf
(I've tried this one absolutely blank as well. No joy.)
send dhcp-client-identifier "1:00:00:c0:64:7c:2b"
#send dhcp-client-identifier "01:00:00:c0:64:7c:2b"
send host-name "fester"; # temporary RHL ifup addition


/etc/modules.conf
alias scsi_hostadapter aic7xxx
alias eth0 wd
options wd mem=0xd0000 irq=10 io=0x280

/etc/sysconfig/networking/devices/ifcfg.eth0

USERCTL=no
PEERDNS=yes
TYPE=Ethernet
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
DHCP_HOSTNAME=fester

/etc/sysconfig/networking/profiles/default/ifcfg.eth0

USERCTL=no
PEERDNS=yes
TYPE=Ethernet
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
DHCP_HOSTNAME=fester


\etc\sysconfig\network-scripts>type ifcfg-eth0

USERCTL=no
PEERDNS=yes
TYPE=Ethernet
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
DHCP_HOSTNAME=fester


I'm a bit confused as to which file does what. Why three copies of
seemingly the same file??

(I know just enough about Linux to be really dangerous - But I'm GONNA
learn this stuff if it kills me!! ehheh)

TIA!

On Tue, 13 Jul 2004 11:58:12 -0700, The OTHER Kevin in San Diego
<skiddz *AT* adelphia *DOT* net> wrote:


>
>I'm a bit confused as to which file does what. Why three copies of
>seemingly the same file??

Oops.. didn't see that there were links to the original file..
heehhe

Pretty please with Linux on top???

In article <(E-Mail Removed)>, The OTHER Kevin
in San Diego wrote:
>Pretty please with Linux on top???

Hmmm...

OK, I found the posting on Tue, 13 Jul 2004 11:58:12 -0700. Lets see...

>Quick scenario: Pentium 200 MMX w/128M RAM and 2 2g SCSI drives, Was
>running RH 7.2 for almost 2 years with no problems. Decided to
>"upgrade" to RH9 after backing up the entire system to 8mm tape.

Off topic, but what went wrong? Truthfully, I've never tried to upgrade a
system, preferring to back up anything needed, then wipe and install from
scratch. It avoids a lot of loose ends. But anyway...

>Went to set up the ethernet connection using an old (circa 1990)
>WD8013 card - same card that worked fine under RH7.2 and intially had
>problems getting the driver to load with IRQ15, IO 0x240 and Mem
>0xC8000.

Yeah, that's an oldy that should work with few problems. Did you remember
to resd the Ethernet-HOWTO, especially sections 2.7, 2.9.8, and 4.41.2?

206074 Aug 25 2003 Ethernet-HOWTO

>Read somewhere the newer kernels like to see IRQ 10 and mem 0xD0000 with
>these old nonPNP ISA cards. Changed the card to IRQ 10, IO 0x280 and
>Mem 0XD0000

I suspect you are referring to the Ethernet-HOWTO above, but things _should_
work if you tell the kernel (via boot parameters or module clues in
/etc/modules.comf) what is going on, AND if your BIOS isn't trying to play
with your head. I'm assuming that you didn't _otherwise_ change hardware
between RH7.2 amd now.

>modprobe seemed to like that combo and loaded all the appropriate
>modules (8390.o and wd.o) with no complaints. However, when I check
>the status of the card, I get this:
>
>ifconfig eth0
>eth0
>Link encap:Ethernet HWaddr 00:00:C0:64:7C:2B

[compton ~]$ etherwhois 00:00:C0
00-00-C0 (hex) WESTERN DIGITAL CORPORATION
0000C0 (base 16) WESTERN DIGITAL CORPORATION
8105 IRVINE CENTER DRIVE
IRVINE CA 92718
[compton ~]$

Well, obviously, you can see the card. But I'd suggest looking at the
boot messages (in /var/log/messages or /var/log/boot perhaps) and seeing
what the kernel saw at boot time. Is it using the right parameters there?

>eth0
>Link encap:Ethernet HWaddr 00:00:C0:64:7C:2B
>UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
>RX packets:0 errors:0 dropped:153 overruns:0 frame:0
>TX packets:0 errors:12 dropped:0 overruns:0 carrier:0
>collisions:0 txqueuelen:100
>RX bytes:0 (0.0 b) TX bytes:3192 (3.1 Kb)
>Interrupt:10 Base address:0x290 Memory:d0000-d4000

All bets are off, because the card has no IP address. I don't even know why
it's showing in the /sbin/ifconfig output in this case.

>Note the IO address of 0x290. Wierd 'cuz the card is at 0x280 and
>this brings up the 1st question.

Again - boot messages, and look at /etc/modules.conf _SOMETHING_ is OK,
because you are reading what I take to be a valid MAC address, and you
can only do that if you can talk to the card.

>What IO should I use for this card?

An old copy of the Ethernet-HOWTO (v2.66, 6 July 1998) suggested

wd.c: 0x300, 0x280, 0x380, 0x240

>RH7.2 ran just peachy with this very card at IRQ15, IO 0x240 and Mem
>0xC8000. Runs fine if I toss in a flopy with an ODI stack and boot
>off that The 153 dropped RX packets raises an eyebrow as well.

The 153 dropped are somewhat meaningless, because the card isn't
configured.

>Since I'd set this thing up to use DHCP for addressing (Served by a
>Winblows NT 4 server running the DHCP server) I figured I'd try
>downing the interface and bringing it back down.

The only reason I've ever seen to use DHCP (or the earlier bootp) is if
you have more computers than you have IP addresses, or if the host is
moving from one network to another. Otherwise, I hard set ALL of the
addresses on networks I'm responsible for. However...

>In both instances, I can watch the DHCP console on the NT machine and
>see the server pick up the broadcast and assign an IP. The DHCPOFFER
>never seems to get back to the RH machine tho.

That sorta suggests that the I/O addresses are correct, because otherwise
you wouldn't be able to stuff packets out onto the wire.

>I thought maybe iptables was stopping the inbound stuff so I opened it
>up all the way by using the security config tool and telling it "No
>firewall".

Good point

>For giggles, I tried setting the ipstatically with ifconfig. The
>address took ok, so I restarted the network service and tried to get
>off the box. No joy. I can successfully ping localhost (127.0.0.1)
>and the NIC itself (192.168.1.100) but can't ping anything outside the
>box.

Actually, you aren't pinging the NIC, because the NIC has no brains to
be able to do anything - it's just an interface for the kernel to use IF
IT WANTS TO. The last item is stressed, because the kernel knows it's
own IP addresses, and knows not to waste bandwidth on the wire when it's
trying to talk to itself - you are secretly using the loopback, which
would work even if the NIC was burnt to a cinder (as long as ifconfig
was able to configure what remains). Try looking at the 'lo' transmit
count (/sbin/ifconfig -a) before and after pinging the NIC address. Note
that the 'lo' interface packet count is increased, not 'eth0'.

>I'm fresh out of ideas and have exhausted combinations of IRQ, IO
>address and base mem addresses with this card.. The *only* one that
>will actually load the modules is IRQ 10, IO 0x280 and mem 0xD0000

If you can down the system, try yanking the card out, and then booting and
looking at what I/O and IRQ is in use now (cat /proc/irq and /proc/ioports).
Then, install the card, and recheck. Is there an overlap? Also, look at
the contents of /etc/modules.conf and those boot messages.

>(I know just enough about Linux to be really dangerous - But I'm GONNA
>learn this stuff if it kills me!! ehheh)

Don't forget that there is a ton of information available in the HOWTOs. I
try to keep current on them, and I have 468 HOWTOs (including mini-howtos),
comprising of 708574 lines / 60 == 11809.56 pages equivalent - just over
3.9 _million_ words.

Old Guy

On Sun, 18 Jul 2004 15:58:23 -0500, (E-Mail Removed)
(Moe Trin) wrote:


>Off topic, but what went wrong? Truthfully, I've never tried to upgrade a
>system, preferring to back up anything needed, then wipe and install from
>scratch. It avoids a lot of loose ends. But anyway...

Nothin' really. The 7.2 system was a bit "off" and it's performance
was getting sluggish. This was pretty much a "just for giggles"
exercise. It ended up being a "nuke and reinstall" instead of an
upgrade.


>Yeah, that's an oldy that should work with few problems. Did you remember
>to resd the Ethernet-HOWTO, especially sections 2.7, 2.9.8, and 4.41.2?

206074 Aug 25 2003 Ethernet-HOWTO

That was my 1st venture into document-dom.

>I suspect you are referring to the Ethernet-HOWTO above, but things _should_
>work if you tell the kernel (via boot parameters or module clues in
>/etc/modules.comf) what is going on, AND if your BIOS isn't trying to play
>with your head. I'm assuming that you didn't _otherwise_ change hardware
>between RH7.2 amd now.

Module clues. Hardware didn't change in the slightest with the 1st
three attempts. The only harware that's changed since is this old
card.


>[compton ~]$ etherwhois 00:00:C0
>00-00-C0 (hex) WESTERN DIGITAL CORPORATION
>0000C0 (base 16) WESTERN DIGITAL CORPORATION
> 8105 IRVINE CENTER DRIVE
> IRVINE CA 92718
>[compton ~]$
>
>Well, obviously, you can see the card. But I'd suggest looking at the
>boot messages (in /var/log/messages or /var/log/boot perhaps) and seeing
>what the kernel saw at boot time. Is it using the right parameters there?

It appears to be using the right stuff at boot time, but even
"removing" the card from the system, rebooting and doing it all
manually leaves me with the same seemingly "dead" card.


>Again - boot messages, and look at /etc/modules.conf _SOMETHING_ is OK,
>because you are reading what I take to be a valid MAC address, and you
>can only do that if you can talk to the card.

hence my growing frustration. RH 7.2 took all of 5 mins to get TCP/IP
running and another 10 mins on top of that to get the samba server up.

>>What IO should I use for this card?
>
>An old copy of the Ethernet-HOWTO (v2.66, 6 July 1998) suggested
>
> wd.c: 0x300, 0x280, 0x380, 0x240

Tried all those with identical results each time.

>The only reason I've ever seen to use DHCP (or the earlier bootp) is if
>you have more computers than you have IP addresses, or if the host is
>moving from one network to another. Otherwise, I hard set ALL of the
>addresses on networks I'm responsible for. However...

I typically set up reservations for each machine in the DHCP server
instead of setting IPs manually, but this one I don't particularly
care which address it gets.

>That sorta suggests that the I/O addresses are correct, because otherwise
>you wouldn't be able to stuff packets out onto the wire.

My thoughts exactly.

>Actually, you aren't pinging the NIC, because the NIC has no brains to
>be able to do anything - it's just an interface for the kernel to use IF
>IT WANTS TO. The last item is stressed, because the kernel knows it's
>own IP addresses, and knows not to waste bandwidth on the wire when it's
>trying to talk to itself - you are secretly using the loopback, which
>would work even if the NIC was burnt to a cinder (as long as ifconfig
>was able to configure what remains). Try looking at the 'lo' transmit
>count (/sbin/ifconfig -a) before and after pinging the NIC address. Note
>that the 'lo' interface packet count is increased, not 'eth0'.

Didn't check that, but I will as I work on this ^&^&* thing..

>If you can down the system, try yanking the card out, and then booting and
>looking at what I/O and IRQ is in use now (cat /proc/irq and /proc/ioports).
>Then, install the card, and recheck. Is there an overlap? Also, look at
>the contents of /etc/modules.conf and those boot messages.

I thought I'd done that, but will do so again.

>Don't forget that there is a ton of information available in the HOWTOs. I
>try to keep current on them, and I have 468 HOWTOs (including mini-howtos),
>comprising of 708574 lines / 60 == 11809.56 pages equivalent - just over
>3.9 _million_ words.

I've just about HOW-TO'd myself blind over the past week or so. I'm
running out of on-line resources as well.. *BUT* I'm not giving up.

On Sun, 18 Jul 2004 20:11:41 -0700, The OTHER Kevin in San Diego
<skiddz *AT* adelphia *DOT* net> wrote:

OK, I give up. 2 weeks of this BS and I still can't get this effing
RH9 box to effing work.. No IO, IRQ or MEM conflicts, the %^&^&* wd
module loads fine the ^%&*^* DHCP server sees the DHCP broadcast,
issues an IP and the friggin' machine won't accept the DHCPOFFER..

boot messages and other logs show the driver being loaded with the
correct parameters but it's still failing.

Static addressing doesn't work for %&*%^&* either. Can't get off the
%^&%&* box with a known good working card. For giggles, I bolted in
an old 3C509 card. Same exact issue as with this WD8013 card. I even
crimped up a new patch cord, tested it with my Wirescope 350 and
replaced the old one. Same deal.

WTF is the problem with RH9??? Is the networking completely screwed
or is it just me??

Grrrr,,,

On Thu, 22 Jul 2004 23:48:13 -0700, The OTHER Kevin in San Diego wrote:
> Static addressing doesn't work for %&*%^&* either...

I would start with static addressing, and work on up.
Get ping to work. That's pretty basic.
Sounds like you're not getting out on the cable.

I do remember having some problems with a WD8013 card on an older machine.
ISTR having to explicitly tell it the IO address (took a few tries because
I mixed up decimal and hex) and specifically what IRQ to use. Furthermore,
ISTR I had to use some Windoze program to change the software controlled
IRQ on the board, since there was no other way to get at it. YMMV

> WTF is the problem with RH9??? Is the networking completely screwed or
> is it just me??

Dunno about RH9. I don't run RH anymore. I run SuSE these days. Works
for me. If you're frustrated kick the walls. Don't kick us.

--
Juhan Leemet
Logicognosis, Inc.

On Fri, 23 Jul 2004 15:39:30 -0200, Juhan Leemet
<(E-Mail Removed)> wrote:

>On Thu, 22 Jul 2004 23:48:13 -0700, The OTHER Kevin in San Diego wrote:
>> Static addressing doesn't work for %&*%^&* either...
>
>I would start with static addressing, and work on up.
>Get ping to work. That's pretty basic.
>Sounds like you're not getting out on the cable.

It's getting out alright. I can see the DHCP server (Winblows) assign
an address. The DHCPOFFER never gets accepted by the card. Static
addressing doesn't work. Something is preventing packets from coming
back IN to the machine.. Thought it was IPTABLES, but I shut that
down and still have the problem.

>I do remember having some problems with a WD8013 card on an older machine.
>ISTR having to explicitly tell it the IO address (took a few tries because
>I mixed up decimal and hex) and specifically what IRQ to use. Furthermore,
>ISTR I had to use some Windoze program to change the software controlled
>IRQ on the board, since there was no other way to get at it. YMMV

I run a DOS based utility to tweak settings on the card. Worked fine
under RH 7.2 for nearly 2 years.. Works just fine if I boot off a
floppy with a DOS/ODI stack on it.

>Dunno about RH9. I don't run RH anymore. I run SuSE these days. Works
>for me. If you're frustrated kick the walls. Don't kick us.

Not kicking, venting.. It should not be this much of an issue to get
one friggin' machine up on the network.

On Fri, 23 Jul 2004 16:45:17 -0700, The OTHER Kevin in San Diego wrote:
> On Fri, 23 Jul 2004 15:39:30 -0200, Juhan Leemet
> <(E-Mail Removed)> wrote:
>
>>On Thu, 22 Jul 2004 23:48:13 -0700, The OTHER Kevin in San Diego wrote:
>>> Static addressing doesn't work for %&*%^&* either...
>>
>>I would start with static addressing, and work on up.
>>Get ping to work. That's pretty basic.
>>Sounds like you're not getting out on the cable.
>
> It's getting out alright. I can see the DHCP server (Winblows) assign
> an address. The DHCPOFFER never gets accepted by the card. Static
> addressing doesn't work. Something is preventing packets from coming
> back IN to the machine.. Thought it was IPTABLES, but I shut that
> down and still have the problem.

So, it's getting from Linux->Windoze, but the return is not "accepted"? Is
Windoze sending to the correct IP address, or someplace else?

Statis addressing should be the easiest of all to setup. If that doesn't
work there is something fundamentally wrong. So, you put both machines in
each other's hosts files and they cannot ping each other? That's bad.

Try some diagnostic tools, on Linux, which put the ethernet in
"promiscuous mode" and accept everything. See if the diagnostics see the
packets "coming back" and if their address headers are correct.

--
Juhan Leemet
Logicognosis, Inc.

On Sat, 24 Jul 2004 02:07:04 -0200, Juhan Leemet
<(E-Mail Removed)> wrote:


>So, it's getting from Linux->Windoze, but the return is not "accepted"? Is
>Windoze sending to the correct IP address, or someplace else?

As far as Mr. Sniffer knows, the DHCPOFFER is being sent right back to
the same MAC address as the card in the RH machine. Something at the
Linux machine is preventing that traffic from hitting the NIC.

>Statis addressing should be the easiest of all to setup. If that doesn't
>work there is something fundamentally wrong. So, you put both machines in
>each other's hosts files and they cannot ping each other? That's bad.

I can assign a static IP to the interface and even if the routes are
correct, traffic will not come back TO the interface. I nuked the
config on my Cisco switch (2924XL) to remove *all* access list entries
and provided rudimentary config and the problem still exists.

>Try some diagnostic tools, on Linux, which put the ethernet in
>"promiscuous mode" and accept everything. See if the diagnostics see the
>packets "coming back" and if their address headers are correct.

Nothing on the lInux machine seems to indicate the packets are
actually "coming back" The only thing I can think of at this point is
some type of interface problem. That card has BNC or AUI interfaces
and I've got an external AUI to RJ45 transceiver connected. The
card's jumpers say to use the AUI interface and as far as I can tell,
there's no way to tell the wd.o module to use the AUI interface. The
card itself has a pair of jumpers to determine AUI or BNC interface.

Again, it wokred great under RH 7.2 and works just fine with a DOS/ODI
stack. right now.