problem still exits in kernel 3.03
ath5k module is not mature as ath_pci/hal provided by madwifi on Arothes AR5001 chip.
I can confirm this. No AP mode in ath5k, but with ath_pci yes. On AR5001X+ chipset.
]]>Turning off WPA encryption at the router eliminated the dropped wireless connections but I don't consider this to be much of a solution.
I don't have the laptop in front of me now, but here are the configuration settings I know: running linux kernel 2.6.35.4-1, ath5k driver, for Atheros chipset AR5212, using wicd.
Blacklisting ath5k in rc.conf and substituting madwifi's ath_pci doesn't solve the problem. The only noticeable change I see is that signal strength is significantly lowered while using ath_pci.
I'm thinking that perhaps a new wpa_supplicant template added to wicd could help, with a line that would either increase (or eliminate) the amount of time that would elapse before dropping the connection, but I don't know what such a modification would look like. The above posts about disabling rfkill looked promising but my Twinhead has no wireless shutoff switch; would disabling rfkill do anything for me?
]]>My ath5k completely died with kernel 2.6.34 update. wlan0 interface doesn't show, module doesn't load automatically. Loading it manually with modprobe doesn't start the card.
D-link DWL-G650 (Atheors AR5212).
Had to revert to 2.6.33.x to get it working.
Bump on this?
]]>D-link DWL-G650 (Atheors AR5212).
Had to revert to 2.6.33.x to get it working.
]]>i realized that the need of rfkill command was due to my thinkpad on/off switch also.(which previously it wasn't usable so that's why i had not test it again.)
The problem is that after switching on and of my wireless 2-3 times it stops responding and wireless interface is not recognized anymore.
As far as the not connecting issue concerned, up to now i have not any access point over here to check.
]]>I tested again with 2.6.34, and after rfkill unblock all, it appears but it keeps remaining unusable.
Is there any possibility for this issue to be arch specific?
]]>There were actually three factors with my problem.
1. No matter the driver I used, my WRT54G didn't like WPA encryption with my Aspire One. It would disassociate everyone connected whenever there was high bandwidth usage on it. So I switched to the DD-WRT firmware
2. After switching to DD-WRT, it didn't like WPA encryption with *anyone*, and would drop any computer that had high bandwidth usage. Thus I disabled encryption, and just to be sure, downgraded to a more tested version of DD-WRT.
3. Ath5k seems to kind of suck for my Aspire One. I frequently got "atk5k: no further txbuf available" in my dmesg after disabling encryption on the router. Which meant it continued to drop.
Thus, the fix for me was to disable encryption on the router, and switch my driver to ndiswrapper. Now things are working peachy keen. I do get dropped once in a long while with no indication of such, but the stability is leaps and bounds ahead of what it was before. I can even download 2GB torrents at 1MB/s without dropping. The one caveat is that speed seems to fluctuate. But that's a very acceptable loss.
]]>I reinstalled Arch last night, to switch fro x86_64 to i686 (for various reasons). When I was configuring wireless from the command line it worked fine with ath5k. To make a long story short, I realized that newer versions of the kernel automatically detect the wireless on/off switch on my thinkpad. This was not the case with 2.6.30, with which it did not matter what position the switch was in. I tried turning off the switch and got the "SIOCSIFFLAGS: Unknown error 132" message that I mentioned in my first post in this thread.
So it's possible that my problem with ath5k in the newer versions of the kernel had to do with my wireless switch being off. Duh. I am deeply embarrassed. Although I really thought I checked the switch when the problem first arose for me in 2.6.31 (because I already knew I had to turn it on in Ubuntu, but it didn't matter in Arch). Of course, I can't rule out that the solution has something to do with reinstalling Arch or with the change to i686, since I don't fully recall whether I checked the switch in the past or not.
Curiously madwifi continues to works fine in 2.6.32 regardless of what position the switch is in.
[Edit: I may have spoke too soon. Now when I scan for access points in wicd, I'm getting the "no wireless networks found" message, even though I'm still connected to an access point. Ath5k, when it works, also remains extremely slow to scan and slow to obtain and IP address, compared to madwifi. It's weird that ath5k would work intermittently.
"iwlist wlan0 scan" gives me this message: "wlan0 Failed to read scan data : Resource temporarily unavailable"
Turning off and on the wireless switch solved the problem. Very weird. I guess that's invoking rfkill somehow?]
]]>So really, the point of me posting here is to say, "me too", and also to ask... did ndiswrapper fix the problem for anyone, or did anyone else have the same problem with it as me? I'm just asking so I can be secure in knowing that the wifi isn't fried on this thing.
I would gladly give Tera's dev kernel a try, and I have repeatedly... it won't boot for me, presumably due to a different hard drive controller. No one's offering a 2.6.33-rc6 kernel package so I can check to see if it works for me. Zen-next is not an option, as rc5 has a regression that breaks ath5k, and it was fixed in rc6.
EDIT: Spoke too soon; I found a kernel26-rc in AUR and will be trying it out now. Hopefully I won't have too many problems downloading the source. :x
]]>kernel26-2.6.31.6-1-i686.pkg.tar.gz
(link of post)
]]>