wifi gone (atheros) [SOLVED]

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

I just had to do a kernel upgrade on a first-generation macbook that had been working fine for a couple of years to sort out a udev poblem. After the upgrade everything was fine except that the wifi had vanished. I decided if it was broken it was about time I moved to the ath5k module anyway, so I unmerged madwifi-ng and rebuilt the kernel with ath5k. The module loads but isn't associated with the device, I can't work out why but I bet it's something basic I'm missing. I believe ath5k should work fine with these early macbooks.

Here's lsmod:

BradN · Posted: Sun Aug 08, 2010 12:07 am Post subject:

CONFIG_ATH_COMMON=m
# CONFIG_ATH_DEBUG is not set
CONFIG_ATH5K=m
# CONFIG_ATH5K_DEBUG is not set

Try turning on these debug settings, install kernel modules, reboot, and then rmmod and modprobe ath5k to see if it shows any messages in dmesg.

From the supported device list on debian wiki:
PCI: 168C:0023 Atheros Communications Inc. AR5008 Wireless Network Adapter
PCI: 168C:0024 Atheros Communications Inc. AR5008 Wireless Network Adapter

So assuming your device identifies as 0023 or 0024 it should be picked up by the driver. Make sure the old driver isn't present just in case it's messing something up.

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

Turned on both debug settings and it made no change to the dmesg output at all.

Can't find any trace of older modules being loaded. /etc/udev/rules.d/70-persistent-net-rules did mention the old module. I removed the file and let it autogenerate a new one; the new one ONLY has a line for eth0, nothing else. Is that a symptom or a cause?

Have been round the houses with this most of today and am completely stumped.

Graham

BradN · Posted: Sun Aug 08, 2010 4:17 pm Post subject:

It's a symptom, udev will only make a record of it if it finds it... if the driver loads...

What's the output of lspci -n?

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

lspci -n:

BradN · Posted: Sun Aug 08, 2010 7:08 pm Post subject:

02:00.0 0280: 168c:0024 (rev 01) <- that device ID looks correct, I have no idea why the driver isn't picking it up.

I guess now it's down to maybe trying older/newer kernels and seeing if there's a bug in your version.

cach0rr0 · Posted: Sun Aug 08, 2010 8:23 pm Post subject:

might try out madwifi again, but on a new kernel; zen-sources includes the madwifi drivers, and is what ive been using in lieu of ath9k.
that would be my suggestion at least. tells you if it's ath5k specific, or something a bit deeper down in the wireless networking stack on that kernel revision
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

I was using 2.6.23-r6 when the udev update lost me my hard drive. I switched to 2.6.34-r1 which was the one I put on pastebin, and found a configuration that gave me my hard drive back but no wifi. I also previously tried 2.6.32-r7 with ath5k with the same results. Now I've also tried 2.6.30-r8 with madwifi (I used the version in portage) with the same results - the ath_pci driver module is loaded apparently ok but lspci -k shows no driver associated with the Atheros network adapter. So it's not likely to be the ath5k driver that's the problem, if madwifi behaves the same.

So I think there are two possibilities: either my hardware died, coincidentally at the exact moment I changed kernel, or there's something I'm doing wrong somewhere. I still think me doing something wrong with configuration somewhere seems more likely.

Here's my current dmesg:

BradN · Posted: Sun Aug 08, 2010 10:31 pm Post subject:

If this wireless is on a pci card, you could try removing and reinserting it - it's possible that if an address line or data line isn't connected correctly, the card could show up in lspci but not respond correctly when the driver probes for it. It's kinda a long shot, because you'll often see crashing or data corruption more often, and the PCI ids are showing correctly so at least most connections are good... If a data line is screwed up you can often see one of the ID numbers it returns off by a bit even!

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

I was being dumb, it's not a hardware problem: I left an OSX partition on it, though I never use it, and the wireless in that is still working fine. So it's a configuration problem. Or udev, which is still giving me problems (just realised my cdrom is no longer recognised as /dev/hda).

BradN · Posted: Tue Aug 10, 2010 8:28 pm Post subject:

You should consider building your wireless driver into the kernel instead of as a module - that will rule out problems with udev as the driver will have already been set up by the time udev is loaded.

marinheiro · Tux's lil' helper Joined: 16 Jan 2004 Posts: 140

I copied the kernel config from http://www.odi.ch/prog/macbookpro/config-2.6.34.txt, and (with some minor changes for the graphics and hard drive) it worked. It uses ath9k as an inbuilt module, but there are a lot of other differences from my old .configure - which was a mess- so I'm not exactly sure what made the difference (at this point I'm a bit out of patience to investigate more). I'm going to mark this as solved, anyway.

For anyone else with macbook problems - http://www.odi.ch/prog/macbookpro is a good resource.

Thanks to both of you, and especially to Odi for his site.

Graham

BradN · Posted: Tue Aug 10, 2010 10:18 pm Post subject:

Oh wow, we were trying to use the wrong driver the whole time. You'd think 5xxx chip model number would mean 5k series, but not in wireless land I guess...