[pkg-fso-maint] Kernel and wifi

[Luca Capello]

Hi there!

On Sat, 28 Feb 2009 20:46:13 +0100, gregor herrmann wrote:
> On Sat, 28 Feb 2009 16:42:43 +0100, Christian Adams wrote:
>> >> * I can start wicd and wicd-client.py later, without getting an oops;
>> >>   but (as before) I have no eth0 device?!
>> eth0 gets enabled via frameworkd (there is also another way via sysfs
>> i currently don't remember)

wlan should be present in /proc/net/dev as soon as you insert the
module.  I guess with "enabled" you meant "brought up".

> Ah, I didn't know that, thanks.

I do not remember how frameworkd works in this specific case, but I
guess it disables every service (GSM, GPS, WiFi, Bluetooth) when it
starts because this is the only way to deal with a consistent state at
frameworkd boot.

>> > - should not this be ath0 or (which I would prefer) wlan0?
>
> With 2.6.24 it's eth0, and I didn't find any hints for another name.

Read below :-)

>> > - can you check if udev has assigned a persistend ifname to the  
>> > device,
>> >   please?  Look at
>> >     /etc/udev/rules.d/70-persistent-net.rules
>
> Yes:
>
> # cat /etc/udev/rules.d/70-persistent-net.rules
> [..]
> SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="00:12:cf:8f:08:56", ATTR{type}=="1", KERNEL=="eth*", NAME="eth0"

I do not know if we need to blame the kernel, the ar6000 driver or udev
for not picking a "better" name.  I was under the impression that any
WiFi interface should have the wlan* name, but I stopped being surprised
by the various ath*, ra* and so on.

>> test a)
[...]
>> 	no probs found
>
> Same here, except that I have wicd installed but not started.
>
> But I could produce an oops later by combining manual "ifconfig eth0
> down", activating wicd-client, and turning wifi on/off with o-p-p.
> (Don't ask me how exactly ...)

I know, I am sometime a bit boring, but I think the command order would
be a good start to understand what cause the oops.

BTW, I have never asked it before, but does the oops happen with plain
(i.e. not encrypted) WiFi networks as well?

>> test b)
[...]
>> 	ar6000 was loaded automatically
>> 	no krnl-oops
>> 	eth0 appeared after 'enabling' via frameworkd
>> 	ifup eth0 worked
>> 
>> 	eth0 comes up even when manually rmmod ar6000 when enabling via  
>> frameworkd

The only explanation I have is that frameworkd loads the module for you,
but I do not have the latest frameworkd sources here ATM to check.

> I tried again now with wicd activated but with the usb cable
> unplugged. Result: kernel oops, no GSM, no network via usb, o-p-p
> doesn't respond, ...
>
> If I boot with the usb cable connected I can at least ssh into the
> freerunner and see when the oops happens.
[...]
> Purging and re-installing wicd (to get a clean config) and unsetting
> the wired interface in wicd + manually down/upping usb0 leads to a
> working system - if I then disable wifi in o-p-p I get the oops
> again.

So it seems to me that the problem relies in the ar6000 module only,
but I am not an expert, sorry.

I will prepare a new Debian kernel package based on a more recent Git
snapshot (not sure if it will already 2.6.29), maybe the FSO-MS5 kernel
is missing some patches.

Thx, bye,
Gismo / Luca
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 314 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/pkg-fso-maint/attachments/20090301/47567bda/attachment.pgp