ms-tech 8139 pcmcia/cardbus problem

Hello.

I have a cheap ms-tech fast-ethernet pcmcia card.
It works unfortnuately under windoze.
I think it is supposed to work, since even the linux-driver on the floppy says:
<rtl8319.c>
...
/* Version 1.11 is originally written by Donald Becker.
   Version 1.11a is modified by Clifford Chiang, [email protected],
   with Link Down Power Saving   */
...
</rtl8139.c>
of course, this driver does not compile 
(rtl8139.c: In function `rtl8129_open':
rtl8139.c:714: error: structure has no member named `tbusy')


I tried rtl8139 newest version and 8139too, too.
<syslog>
Oct 11 11:29:25 civita kernel: 8139too Fast Ethernet driver 0.9.26
Oct 11 11:29:25 civita kernel: eth0: RealTek RTL8139 Fast Ethernet at 0x4000, 78
:06:17:25:1b:52, IRQ 13
Oct 11 11:29:25 civita kernel: eth0:  Identified 8139 chip type 'RTL-8139C'
</syslog>


My problem is again this:
<syslog>
Oct 11 11:29:52 civita kernel: NETDEV WATCHDOG: eth0: transmit timed out
Oct 11 11:29:52 civita kernel: eth0: Tx queue start entry 4  dirty entry 0.
Oct 11 11:29:52 civita kernel: eth0:  Tx descriptor 0 is 00002000. (queue head)
Oct 11 11:29:52 civita kernel: eth0:  Tx descriptor 1 is 00002000.
Oct 11 11:29:52 civita kernel: eth0:  Tx descriptor 2 is 00002000.
Oct 11 11:29:52 civita kernel: eth0:  Tx descriptor 3 is 00002000.
Oct 11 11:29:52 civita kernel: eth0: Setting full-duplex based on auto-negotiate
d partner ability 0000.
</syslog>

I get a link stat, I try to ping, I hear and see some ping-flashes, but no ping gets through, it says 100% packet loss.
But I could rarely really ping from outside, with a lot of loss.

I read it might be the interrupts sharing.
I have pcmcia from kernel 2.4.22 (debian), no external pcmcia-source installed
<syslog>
Oct 11 11:23:17 civita kernel: Linux Kernel Card Services 3.1.22
Oct 11 11:23:17 civita kernel:   options:  [pci] [cardbus] [pm]
Oct 11 11:23:18 civita kernel: PCI: Found IRQ 13 for device 00:0a.0
Oct 11 11:23:18 civita kernel: PCI: Sharing IRQ 13 with 00:0a.1
Oct 11 11:23:18 civita kernel: PCI: Found IRQ 13 for device 00:0a.1
Oct 11 11:23:18 civita kernel: PCI: Sharing IRQ 13 with 00:0a.0
Oct 11 11:23:18 civita kernel: Yenta IRQ list 0698, PCI irq13
Oct 11 11:23:18 civita kernel: Socket status: 30000020
Oct 11 11:23:18 civita kernel: Yenta IRQ list 0698, PCI irq13
Oct 11 11:23:18 civita kernel: Socket status: 30000006
Oct 11 11:23:18 civita kernel: cs: cb_alloc(bus 2): vendor 0x10ec, device 0x813
9
...
Oct 11 11:23:18 civita kernel: PCI: Enabling device 02:00.0 (0000 -> 0003)
Oct 11 11:23:19 civita cardmgr[792]: watching 2 sockets
... (or)
Oct 11 11:51:59 civita kernel: PCI: Enabling device 06:00.0 (0000 -> 0003)
Oct 11 11:51:59 civita kernel: PCI: Setting latency timer of device 06:00.0 to
64
... depending on the slot I plug the thingy in.
<syslog>

<lspci>
00:0a.0 CardBus bridge: Texas Instruments PCI1225 (rev 01)
00:0a.1 CardBus bridge: Texas Instruments PCI1225 (rev 01)
06:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10)
</lspci>

</proc/interrupts>
13:          0          XT-PIC  Texas Instruments PCI1225, Texas Instruments PCI1225 (#2)
<//proc/interrupts>


both rtl8139 and 8139too show the same behaviour.

Do I really have a cardbus system, because my pcmcia of the laptop is ~3 yrs. old?
Do I have an IRQ conflict somewhere?
can anybody help me?


A lot of thanks for any reply. I'm not on the list. So please email me, but I'll check the archives anyway.

cheers, tobias.