Atheros AR8132 PCI-E FE Controller conflict?
 

Acer laptop 5517-5997, (a recent bargain basement model) It is a few weeks old and have just got it back with a new motherboard. (the video chipset fried after 2 days.)

I've read the past thread on this chipset, and tried the solution offered to no avail. What I'm experiencing is this:

The eth0 device is intermittently dissapearing. Sometimes I boot the machine and lspci shows the FE controller. Sometimes not. When it does, the device (from ifconfig) is numbered with the reserved boot address (165.something... I think), and any attempt to configure the interface manually, causes the device to dissapear.

I've used modprobe to remove and install the default module, the test module recommended in the previous thread, and the most recent module available from the vendors site. All for soot. Now I can't get the device to show up again.

This machine also has an Atheros WiFi device installed (AR5B95), I have blown away this module presuming a possible memory address conflict. (do they still have those?) No luck.

My questions are as follows: How dependable is lspci? Do motherboards still use jumpers for memory address selection, and do they do it with PCI-E? (Haven't touched hardware in a _while_ can you tell?) I could just see someone at Acer sticking a new board in and not bothering to set a jumper.

Or is there some alternative to lspci which will give me a look at what is going on in the bus?

Anyway, any help our troubleshooting advice would be appreciated.
Thanks in advance!

Much of this can be caused by fast ethernet switching being set in the BIOS. If the nic is used, the bios disables the wifi. Eliminate this first.
lspci is pretty good, but the bios rules over the kernel. You also have /proc/bus /sys/devices and /sys/bus to grok. Nothing has jumpers anymore except CPUs and IDE disks as a rule. There are occasions where different physical connections are required, but it's a minimum. It's mainly automagic these days.

SUGGEST: Power off, enter setup, disable fast ethernet switching, and come back at them then. You should be able to set up eth0 and wlan0. Try to do it in the init scripts, and don't leave it to network manager unless you live in X. If you can't find fast ethernet switching, presume it exists. Leave the nic unused if you want wifi, and turn wifi off when load up the nic.

Only network related options I have in BIOS is to turn off PXE boot, which i've done as a matter of housekeeping. I reinstalled Slack 13 and the device showed up in lspci again.

I'm presuming the 0x.00.0 preceding the device name in lspci is a memory address or interrupt number or something? I've got the WiFi on 02.00.0 and the NIC on 08.00.0

Looked at these proc files. Not much to glean without parsing the binary output of some of them, and that is a bit above my ability at this point.

Any other thoughts?

Thanks!

The agp, pcie and pci buses are inclined to be merged in the lspci o/p. lspci shows me

0:00:00 (Northbridge)
then northbridge devices (of which memory is one)
then (00:12:0)southbridge devices
These are legs that the cpu can directly toggle

Once you get to 02:nn:n and above, they seem to be driver or software buses i.e. the device is invisible without the driver loaded. Typically one device per major number - These will be pc card bus, nic, wifi, soundcard, etc. The usb bus doesn't show in lspci.

Gave up.
 

I looked around, and it looks like the driver for this chipset is being worked on by some MAC guys or something, but nothing imminent. In the mean time I would consider ACER products with this Atheros NIC to be simply not Linux compatable.

So I bought an

AGILER USB TO RJ45 LAN ADAPTER

off amazon for $4, and the kernel sees it and it configures fine. It is a little klunky and I'm torqued at loosing a USB port because ACER can't figure out how to do Ethernet properly. I mean seriously, the technology is twenty years old, and the machine was and AMD based $325 slab. How is it possible that they've added a _new_ 100BaseT chip? And given the state of Ethernet, how is it possible that there even _IS_ a new 100BaseT chip?

On the good news front, I found the WiFi works a treat. So no beef there.

If anyone ever sorts out a driver for this chip please reply to this thread!