xircom cardbus IBM laptop FEDORA problem
 

I have a strange situation.
Laptop T20 750Mhz
Dual boot : XP and Fedora
Before Fedora I used Suse8. Worked fine.

Problem:
When Fedora is loaded I can ping and dig hosts.
When trying to browse Internet. Can open first page. When click on some links - no connection anymore.
After that cannot ping (host unreachable) and cannot dig anything.

Tested card again from XP - works .
What it may be?

The card is Xircom Curdbus. Fedora detected it.

P.S.
May be it's normal but I found weird that during boot
first it's trying to load network, the status is FAILED load network or delayed..
and after it loads PCMCIA this one is OK>

What to check?
May be just go back to SUSE? :)
I really would like to make it work on Fedora.
Thanks.
Michael.

you're right on the load sequence. i had to rename my pcmcia (now it's /etc/rc.d/rc5.d/S09pcmcia) load script so it would load sooner than the eth0 otherwise my adsl would hang.

perhaps you should check to see if pcmcia has a higher number then the network script.

thanks a lot.

unfortunately it does the same after fix.
Now pcmcia is loaded before eth0. there is no error message.
I load Fedora. Can ping can dig ibm.com.
When I go to imb.com from browser network hangs.
Cannot ping and dig anymore.
may be you can suggest something else?

I seem to have a similar problem. Using SuSE 9.0 on a Toshiba Tecra.
Right after boot up the network is operational, then hangs after a few seconds. Interestingly it worked for a few minutes right after installation of SuSE, during the automatic update. Connected to ftp server, got update information and downloaded the patches.
Now after every boot up it behaves the same and refuses to work.

Berfore I used SuSE 8.2 and the network worked just fine.

maybe I should also mention that my notebook is connected to an SMC router with dhcp enabled and it get a valid ip.
Running "tcpdump -i eth0" i get

22:59:13.790335 arp who-has 192.168.11.1 tell 192.168.11.101

every two seconds or so. 192.168.11.1 is the local rout ip, 192.168.11.101 is the one of the notebook.
I can't ping any other machine from my notebook, and when trying to ping 192.168.11.101 from another pc (ip 192.168.11.103) then tcpdump shows

22:59:26.838900 arp who-has 192.168.11.101 tel 192.168.11.103

due to the predictable and almost instant nature of your problem there may be an irq conflict. i'm not familiar with fedora's configuration tools but you may be able to snoop around to see if you are conflicting with your sound card or something. there may also be a buffer issue. i have a strange problem with my provider (yahoo bb japan) that intermittantly forces me to reboot once every week or every two days due to all networking going down. my network runs 24/7 through a 12mb adsl/voice over ip connection. the adsl modem hands me an ip via dchp.

i'm still looking for some tools to give me an error trace or something.

more info

/etc/modules.conf
...
alias eth0 tulip
alias eth1 tulip
options tulip irq=10,11

I tried to connect on LAN. Main switch is connected to firewall and 128k ISDN and through cable modem. The same story. When it hangs it shows
proper address in ifconfig. Just you cannot ping.
Looks like OS problem.
I found the same it worked on SUSE8.2 and not works on Fedora.
Now we can see that it not works on Suse9.
Tomorrow will check IRQs.
Any news please post.
Michael.

Hmmm ... There is a new driver for xircom and i wonder if this is related?

what driver is being assigned to your xircom in Suse9 and Fedora? have a lok at your /etc/modules.conf (or equiv) and tell me what is aliased.

i switched away from the xircom specific driver and went back to the old tulip and i increased the reliability of my connection.

try the tulip driver. this is what i'm using in my rh7.3 firewall. i have dual RBE-100 one is to the adsl the other to my 192.168.0.0 net.

my rh9 box has an internal intel 82559 compatible pci so i don't know how this looks in rh9.

Hi Jeff,
my modules.conf:

alias usb-controller usb-uhci
alias sound-slot-0 cs46xx
post-install sound-slot-0 /bin/aumix-minimal -f /etc/.aumixrc -L >/dev/null 2>&1 || :
pre-remove sound-slot-0 /bin/aumix-minimal -f /etc/.aumixrc -S >/dev/null 2>&1 || :
alias eth0 xircom_cb
all what I can do: send one message. After that no network :)

If you can suggest something please provide detailed instructions.
I never installed drivers on Linux.
So if yu think tulip will help. Tell me how. And what is rpm or source exact name.

In my case this card worked on Suse 8.2 not 9.

I have 3com pcmcia but it is with dungle.
And i prefer real port card ( my IBM(Xircom) card)


Thanks.
Michael.

i hear you on the dongle. i also prefer the realport rbe-100 that i have. my only complaint is that they don't like to start if they're hot. when restarting the system i sometimes have to cool them down first.

do a locate tulip and see if you can find tulip listed under your current kernel version.

Here's mine from my RH7.3 box:
...
/lib/modules/2.4.20-20.7/kernel/drivers/net/tulip
/lib/modules/2.4.20-20.7/kernel/drivers/net/tulip/tulip.o
/lib/modules/2.4.20-20.7/kernel/drivers/net/tulip_old
/lib/modules/2.4.20-20.7/kernel/drivers/net/tulip_old/tulip_old.o
...

if you have these try changing xircom_cb to tulip or even tulip_old and see if your situation improves.

also, are you getting any irq conflict messages while booting?
if so you can specify the irq with the option command

options tulip irq=10,11


i hope this helps

Hello again,
mla or jeff, do you know if ACPI is enabled in Fedora? Because in SuSE 9.0 it is by default but wasn't in 8.2. Now I disabled it by stating acpi=off in the boot parameters and my network connection works.
So it seems to be a problem with acpi, at least for me and SuSE 9.0. Since better acpi support was a reason to move from 8.2 to 9.0, I hope there is a way to get both going properly.

hi there,
I used chkconfig and put acpid to off 345 (it was on by default on Fedora).
Didn't help. :(
I did locate tulip.
Jeff, it does exist and looks like yours (two more lines).
I have no mood to restart laptop again just to e-mail output ( I have my network 20 sec. :))) )
Tell me please how to go to previous driver (exact commands)
Thanks guys.
Michael.

P.S.
After 20 sec of working I type service network status it not shows
eth0 at all. service network restar failed for etc0.

Jeff,
I analysed your previous posts and found that I have to change name of
alias in modules.conf in order to load different driver. Yes?
I tried to change my alias eth0 to few tulip's.
restarted comp. Load is normal (bringing up network OK etc.)
But the situation is exactly the same. It works 20 seconds.
After I restart network service "bringing up interface eth0 : FAILED
Mistery.

If you have another "things to do" you are welcome.
Michael.