Re: problems with madwifi and Atheros AR5418
Sorry if I used the wrong notation, that was 25 Mbits/s ~ 3.2 Mbytes/s
which is only slightly faster that 11g so it's not that "wow", but an
improvement nonetheless.
What modules is module assistant trying to build? The headers
shouldn't be a problem. If you've compiled your own kernel, just make
sure to keep the source tree from the compile (that means don't "make
clean" it too). I did however notice a few things weren't yet bleeding
edge enough for the new kernel. A common problem is that the
asm/semaphore.h header is now at linux/semaphore.h. If you take a look
in your source files and switch that around, things might work. For
example, see this bug on tp_ctrl:
http://sourceforge.net/tracker/index.php?func=detail&aid=2041469&group_id=1212&atid=101212
Kevin
On Mon, Sep 8, 2008 at 9:49 AM, Manuel Pégourié-Gonnard <mpg@elzevir.fr> wrote:
> Kevin Mitchell a écrit :
>> The ath9k in 2.6.27-rc5 is working famously for me on my Thinkpad T60
>> AR5418. I'm even creeping past 11g speeds (25MB/s on a good day).
>
> Wow. Here I can't do tests since my only other local machine has slow
> wifi, but yes, ath9k form 2.6.27-rc5 works fine too (except that that
> card won't reconnect to the AP if it goes down then up unless I reload
> the driver... fortunately, my AP doens't go down that often).
>
>> I used to use madwifi and didn't experience the kind of problems you
>> speak of. I did however notice that somewhere along the way, there was
>> a regression that slowed down the connection by a non negligible
>> amount http://madwifi.org/ticket/2043. You might try downloading 3401
>> snapshot (I think you use the "-r 3401" as an argument to svn). I had
>> meant to pinpoint exactly when and where this regression occured, but
>> then ath9k came out and there was an exponential decay in my caring
>> coefficient.
>>
> Well, since ath9k is working well, I don't think I'll try to go back to
> madwifi. Now the only drawback is, module-assistant doesn't succed any
> more in building modules: I'm under the impression it tries to use the
> 2.6.26 headers...
>
> Manuel.
>
>
Reply to: