Please add madwifi back into intrepid's eeepc, eeepc-lean for 701s and 900s

[bedfojo]

As requested at: http://forum.eeeuser.com/viewtopic.php?id=46649&p=6 in post #138.

It seems that for eeepc 701s and 900s, the FLOSS driver ath5k for the wireless chips in gives much worse performance than the madwifi driver. See: https://lists.ath5k.org/pipermail/ath5k … 01269.html

Apparently, ath5k developers know about this problem but it is low on their list see: http://www.mail-archive.com/ath5k-devel … 01168.html

It is reported from http://forum.eeeuser.com/viewtopic.php?id=46649&p=6 at post #135 that "the ath5k driver does not handle tx power at all. There is no dynamic adjustments nor can you manually adjust the power. The result for the eeepc (at least my 900) is that the wifi hardware is set to max power causing clipping and poor performance as well as driving up power consumption in the eeepc. An intrepid kernel with the original wifi module would be a big plus for the eeepc wireless performance as well as lower its power consumption, at least until the madwifi crew get a patch for the ath5k driver that works with our chips."

[Adam McDaniel (adamrmcd)]

Assigning myself.

Note that the ath5k driver should still remain the default as this is what Intrepid itself is aiming for.
I can restore madwifi's ath_pci module frombranch "madwifi-hal-0.10.5.6" back onto the intrepid eeepc/eeepc-lean kernels.

(Note Madwifi is still the default on the hardy eeepc kernel.)

To enable (when it's ready)
1. Add "blacklist ath5k" to /etc/modprobe.d/eeepc
2. Append "ath_pci" to /etc/modules
3. Reboot

(I imagine eventually there will be an easier config option applied to eeepc-config, when this fix is released)

[Fabio Albieri (chareos)]

I'm desperately waiting for ath_pci to happen too ;)

[Fabio Albieri (chareos)]

I'm sorry to bother you, but this issue is well standing still. How's going ?

[Adam McDaniel (adamrmcd)]

Unfortunately I encountered some nasty regression issues when trying to integrate ath_pci back into the kernel. I believe that there are some problems with the hal-0.10.5.6 branch code running on 2.6.27;

I haven't had a chance to debug them yet, however, on a semi-related note, there are some upstream updates to ath5k coming which promise to improve on some of the performance issues.

Keep an eye on this ticket, or the array.org news, if there's movement on either front i'll post updates on both fronts :)

[Arne Bockholdt (u-launchpad-net-bockholdt-com)]

It seems that linux-backports-modules-intrepid-generic package from intrepid comes with an updated version of ath5k module which works much better than the version from eeepc kernel on my 701.

Maybe you should recompile this package with your headers and .config and make a linux-backports-modules-intrepid-eeepc to work around the compatibility problems with latest madwifi and 2.6.27 kernels.

[Adam McDaniel (adamrmcd)]

That is exactly what the plan is :)

At least this is one aspect where I actually own the hardware to test on rather then have to guess how it'll react.

[Fabio Albieri (chareos)]

Hi there,

Still no news ?
Thanks

[Emilio (emillan)]

Fabio (and others): You can install madwifi yourself.

Follow the instructions here to install the driver: http://wiki.eeeuser.com/ubuntu

And then use Adam's instructions above to blacklist the ath5k driver and enable the ath_pci/madwifi driver.

Hope this helps!
Emilio

[armo (r-mo)]

Not sure how related this is but my ath5k performance is really bad often not giving me a connection at all. I see lots of these messages (differing frequencies) in kern.log even when I'm right next to the access point.

[code]
ath5k phy0: noise floor calibration timeout (2462MHz)
ath5k phy0: ath5k_chan_set: unable to reset channel (2462 Mhz)
wlan0: failed to set freq to 2462 MHz for scan
[/code]

[Weedy (weedy2887)]

do I still need to compile madwifi or have you made any progress?

[Paula (paula-ke)]

Weedy wrote:
> do I still need to compile madwifi or have you made any progress?
>
>
It still performs poorly by passable. Can't get any better than 24M in
speed and still no control over xmit power (which is the real problem)
Not sure what kind of progress the ath people have made.
--Paula

[Weedy (weedy2887)]

I ment is there a ppa or something for it, has eeepc-kernel been fixed or do i need to start compiling.

[Weedy (weedy2887)]

has this been fixed in later kernels yet?

[Gianmarco De Francisci Morales (gianmarco-dfm)]

It seems this problem is not being handled.
Should I compile madwifi by hand (I'm using 2.6.27.8-eeepc-lean) ?