Luis R. Rodriguez a écrit :
> On Thu, Nov 26, 2009 at 01:47:14AM -0800, Gábor Stefanik wrote:
>> 2009/11/25 Benoit PAPILLAULT <benoit.papilla...@free.fr>:
>>> Benoit PAPILLAULT a écrit :
>>>> TX rate is dropping a lot (like 20 MBits down to 200 kbits/s). Using
>>>> another card in monitor mode, I still see packets sent at 54 Mbits rate
>>> I found out what is wrong with the TX throughput. In fact, when a packet is
>>> transmitted at 54 Mbits, the destination send back a 802.11 ACK which is
>>> ignored!!! As such, the packet is transmitted at 54 Mbits once again, then
>>> twice at 48 Mbits, then twice at 36 Mbits ... and so on down to 1 Mbits. Of
>>> course, this divide the throughput by 18 (twice for each 9 rates), leading
>>> to 200 - 300 kbits.
>>> When implementing proper TX status reporting, needed by minstrel rate
>>> control, I also found out that some 802.11 ACK where not received by the
>>> driver, but received by the card, leading to inconsistencies in the rate
>>> control code.
>>> Are there some registers that are responsible for ACK behavior?
>> ACK handling and retrying is one of the known weak points of the
>> zd1211(b) chipset - apparently enabling CR_SNIFFER causes the card to
>> ignore ACKs received. Also, the retry plan for unACKed outgoing
>> packets (retry count & rates) needs to be uploaded to the device on
>> firmware init, and can't be controlled per-packet.
>> Luis: Any chance of an open firmware release for the ZD1211, similarly
>> to what was done for AR9170/ZD1221? That would probably allow for
>> fixing these problems.
> I had asked a while back and there is no immediate motivation. I should
> note for ar9170 I did all the review myself during my own time, and can't say
> I have enough interest to do so for zd1211 as I can use my time to focus
> on *newer* items. I'm not even sure about the compiler that was used and
> still have yet to see more developer interest on ar9170.fw before making
> arguments in favor for any other release.
> BTW use my mcg...@gmail.com address if you want a quicker reponse, I tend
> to use my atheros address only for internal stuff.
I understand your concern. So far the current state of the driver is
usable to me. Bad performance in monitor mode is not a problem since I
only use it for debugging.
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
Zd1211-devs mailing list - http://zd1211.wiki.sourceforge.net/