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

Re: autoloading module for pcmcia network card



Hello,
Thomas Friedrichsmeier [Mittwoch, 2. April 2003 14:53]:
>| After I finally broke off the adaptor-cable of my old network-card, I now
>| have a new noname 8139-based one.
Assuming it's a cardbus card...
>| Problem is, that the cardmgr can find and activate the card, but does not
>| load the required modules or start the interface. The card works just
>| fine, if I modprobe the 8139too-module and configure the interface by
>| hand, but I'd rather have this happen automatically, like it worked with
>| my old card.
(8139too is said to be somewhat suspicious, though.)
>| Apr  2 14:42:50 flyer kernel: Linux Kernel Card Services 3.1.22
... you can try if the pcmcia-cs package does a better job.
>| Apr  2 14:42:50 flyer kernel:   options:  [pci] [cardbus] [pm]
>| Apr  2 14:42:51 flyer kernel: PCI: Guessed IRQ 9 for device 00:13.0
>| Apr  2 14:42:51 flyer kernel: PCI: Sharing IRQ 9 with 00:13.1
>| Apr  2 14:42:51 flyer kernel: PCI: Guessed IRQ 9 for device 00:13.1
>| Apr  2 14:42:51 flyer kernel: PCI: Sharing IRQ 9 with 00:13.0
>| Apr  2 14:42:51 flyer kernel: Yenta IRQ list 0c98, PCI irq9
>| Apr  2 14:42:51 flyer kernel: Socket status: 30000827
>| Apr  2 14:42:51 flyer kernel: Yenta IRQ list 0c98, PCI irq9
>| Apr  2 14:42:51 flyer kernel: Socket status: 30000007
>| Apr  2 14:42:52 flyer kernel: cs: cb_alloc(bus 2): vendor 0x10ec, device
>| 0x8139
Now here something is. But not identified from the config list.
>| Apr  2 14:42:52 flyer kernel: PCI: Enabling device 02:00.0 (0000 -> 0003)
Seems not to be cardbus device! Is that all about ? Next, there should be 
cardmgr: entries like socket : ID string in the syslog; then cardmgr 
executing modprobe.
>| Apr  2 14:42:52 flyer kernel: cs: IO port probe 0x0c00-0x0cff: clean.
>| Apr  2 14:42:52 flyer kernel: cs: IO port probe 0x0800-0x08ff: clean.
>| Apr  2 14:42:52 flyer kernel: cs: IO port probe 0x0100-0x04ff: excluding
>| 0x408-0x40f 0x480-0x48f 0x4d0-0x4d7
>| Apr  2 14:42:52 flyer kernel: cs: IO port probe 0x0a00-0x0aff: clean.
>| I tried adding the following lines to /etc/pcmcia/config:
>| device "8139too"
>|   class "network" module "net/8139too"
I think here's a module missing ... As i understand it (and we deal about a 
cardbus card, pcmcia version 2.x), there should be also a pcmcia-module 
assigned to the class. They usually end with _cs for 2.4 kernel; e.g. 
 "pcnet_cs"  for the "net/8390".
If it's not a cardbus card, you might consider the "hotplug" package instead 
of pcmcia cs. It's said to be better with pcmcia 32 bit cards ( though 
pcmcia-sc supports that also) and also handles PCI and USB.
Better to run only either of them. 
>| card "NoName 10/100"
>|   manfid 0x10ec, 0x8139
>|   bind "8139too"
Seems to be alright. Did you get the name string from 'cardctl ident' ?
You know, it has to match _exactly_ ....
I would first try to find out if cardmgr (cs) would accept the config file at 
all. Try to decrease the amount of information, e.g. exclude the name string 
or even the 'manfid' ... down to a pure "function network_adapter " or even 
"anonymous" card. Until *something* is recognized.
See the pcmcia manpage + PCMCIA HOWTO chapter 3.

After all this set up correctly, and my card still wasn't identified, i 
compiled the 'standalone' package, and it worked out of the box.

This implied to disable kernel pcmcia-modules (CONFIG_PCMCIA=n),
in /etc/defaults/ choose i82365 as socket, and configure (with make) and 
recompile cardmgr and the other tools. It's all described in the Readme.
(It's a little bit freaky if you've got kernel-module-version activated and 
less experience with make-kpkg...i can send you my notes if you like.
I had to solve some dependency tasks, also.)
It's but a chance, only if anything else is done.

-- 
                                                                              
                          micha.



Reply to: