Bugzilla – Bug 105598
Three drivers for the same device loaded at once
Last modified: 2005-08-24 19:55:20 UTC
I have a NetGear MA301 PCMCIA WLAN card in my desktop system that is connected to the main board via a PCI card. YaST2 discovers this card correctly, but the WLAN configuration did not succeed. After some investigation it turned out that three(!) different kernel drivers were loaded for this card at the same time: orinoco_plx, hostap_plx and prism2_plx. I wonder how that is even possible... After unloading them and just loading the orinoco_plx module, the WLAN worked...
Created attachment 46559 [details] hwinfo.txt
Created attachment 46569 [details] y2logs.tgz
The modules are loaded even before the network proposal. Christian, maybe hotplug does it? How can we find out?
Could it be ther kernel itself that triggers module loading according to the PCI ID? modules.pcimap maybe?
After a reboot, only the prism2_plx module was loaded, as it was configured in /etc/sysconfig/hardware/hwcfg-bus-pci-*
Created attachment 46628 [details] A tiny hotplug logging howto
Created attachment 46629 [details] The script for nice logging output
Hi Lenz, please remove the hwcfg file for your wlan device and follow the attached howto. The interesting pci events have usually a low sequence number (below 100). So 'show_event_log 1 100 pci > hotplug.pci.log' should show you what hotplug does for cpi devices.
Hmm, but this card is not really used as a hot-pluggable device - it should rather be treated as a built-in PCI network card. In fact, the PLX interface does not support hot-plugging the PCMICA card. it would probably kill the device. Does that matter in this case? Or do your instructions above apply nevertheless?
It does not matter if you can physically hotplug a device. We treat them all the same way: detect device --> hotplug event --> initialize device --> register interface (*) --> hotplug event --> setup interface (*) Of course there are exceptions, as always, but we tryto minimize them. (*) /dev/hda is also an interface, even if called device register interface can also mean reister new bus system, where we detect new devices
Joe, this is the same case with the *_plx modules. *** This bug has been marked as a duplicate of 106003 ***