rt2x00 driver

[Gábor Molnár]

Hi!
As far as i can see, this is a catch of 22 situation... i personally
patched the hostapd source to ignore the lack of ackowledge. Maybe i
will write a mail to the linux-wireless mailing list, it seems that
the problem has to be solved on a higher level, maybe in the mac80211
stack.

regards,
 M.G.

2008/12/10 Iwo Mergler <iwo at call-direct.com.au>:
> Gábor Molnár wrote:
>> Hi!
>>
>> I have a rt73-baseb usb dongle, and i want to use it in master mode,
>> so i downloaded and compiled the latest rt2x00 kernel source from git,
>> and compiled it succesfully. HostAPd seems to work correctly at first
>> sight, sends out beacons, and the client can associate as well. But
>> when the client is associated, the hostapd log says, that "wlan1: STA
>> 00:14:a5:29:e0:98 IEEE 802.11: did not acknowledge association
>> response", and after that, when the client tries to send packages, the
>> hostapd forces it to reassociate.
>>
>> The hostapd part of the story
>> There was a thread earlier in connection with this problem:
>> http://readlist.com/lists/shmoo.com/hostap/1/5106.html. The case is
>> that the driver doesnt ackowledge the association response.
>>
>> The rt2x00 part of the story
>> I found a thread in the rt2x00 mailing list, in connection with this:
>> http://sourceforge.net/mailarchive/message.php?msg_id=200808082059.11968.IvDoorn%40gmail.com
>> (the discussion is about 3 different patch, the first one is the
>> important part, but in the end, the patch wasn't accepted) and another
>> thread in their forums:
>> http://rt2x00.serialmonkey.com/phpBB/viewtopic.php?f=5&t=4790&p=30366
>> . They say, that "there is simply no way to tell if the frame was
>> successfully send or not". So, the driver cannot acknowledge this, and
>> they say, that handling this situation is the responsibility of
>> hostapd.
>>
>> Now everybody says that this is the responsibility of the other one.
>> Now you can see the situation clear. What do you think about this
>> issue?
>>
>
> Hi Gábor,
>
> As far as I can tell, everyone is right.
>
> From memory, rt2x00 cannot tell if a specific frame got ack'ed, as the
> hardware can and does accept multiple frames for transmission, but only
> has a global failure counter. There have been some ideas floating around
> to fix this, but all would have meant some false positives and/or false
> negatives. Thus, the rt2x00 drivers return 'unknown' rather than 'success'
> or 'failure'.
>
> I believe mac80211 then translates 'unknown' into 'failure' and hostapd
> acts accordingly. There probably is a possible solution involving both,
> mac80211 and hostapd to deal with 'unknown', but most other hardware
> doesn't have the problem, so the motivation seems to be lacking.
>
> Personally, I hacked the rt73usb driver to return 'success'. I'm aware
> that this can cause problems, but within the scope of my application,
> it works.
>
> Kind regards,
>
> Iwo
>
>