7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Kimmo Huoso]

Im having vast amount of  Invalid misc, Tx excessive retries. my 
hardware is 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB 
Adapter(rev.A) [Atheros AR9001U-(2)NG]. Kernel is currently kernel26 
2.6.37.5-1 (arch linux x86_64), using carl9170-fw-1.9.2, wpa_supplicant  
0.7.3-3 (arch linux) and driver is now from compat-wireless-2011-03-31.

There seems to be  no difference with latest bleeding edge release of 
linux-wireless either which version of carl9170-fw im using. The problem 
has been there since i were able to compile compat-wireless for 2.6.37 
which i cant recall when. The problems appears either way, ht disabled 
or enabled. Tried to do some "debugging" by enabling 
CONFIG_MAC80211_HT_DEBUG=y....etc switches from the kernel but nothing 
strange seems to appear in dmesg. I never had this problem with 
ar9170usb driver for a note, so i think its not hardware problem exact.

Re: 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Christian Lamparter]

On Tuesday 05 April 2011 14:13:50 Kimmo Huoso wrote:
> Im having vast amount of  Invalid misc, Tx excessive retries. my 
> hardware is 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB 
> Adapter(rev.A) [Atheros AR9001U-(2)NG].
There are two "rev A", hope you got the second revision.

But back to the topic: these softmac numbers/statistics are bogus...
It might sound a bit weird, but in fact they are not high enough.

To make a fair comparison you would have to check the hardware
counters [RX_PHY_MISC_ERROR & friends] instead. carl9170 already
has a debugfs interface for this reason [hw_[r|t]x_tally], but
ar9170usb lacks this feature.

Regards,
	Chr

Re: 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Kimmo Huoso]

Thanks for answering.

Im trying now to get the carl9170 debug working. I didnt see any 
parameter for debug while running modinfo carl9170. But some sort of 
debug code seems to be in place in compat-wireless under carl9170. What 
should i do to enable it. And second how can i get readings from 
/sys/kernel/debug to be exact. I dont know much of what comes to kernel 
debugging. And for third even the softmac numbers/statistics might be 
bogus, im still having problems with the driver while trying to run 
programs that are strict for smooth connection like online games.

The card is 7d1:3c10 which has been marked as Rev.A1 and sticker at the 
back says so too.

Thanks Kimmo

06.04.2011 15:38, Christian Lamparter kirjoitti:
> On Tuesday 05 April 2011 14:13:50 Kimmo Huoso wrote:
>> Im having vast amount of  Invalid misc, Tx excessive retries. my
>> hardware is 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB
>> Adapter(rev.A) [Atheros AR9001U-(2)NG].
> There are two "rev A", hope you got the second revision.
>
> But back to the topic: these softmac numbers/statistics are bogus...
> It might sound a bit weird, but in fact they are not high enough.
>
> To make a fair comparison you would have to check the hardware
> counters [RX_PHY_MISC_ERROR&  friends] instead. carl9170 already
> has a debugfs interface for this reason [hw_[r|t]x_tally], but
> ar9170usb lacks this feature.
>
> Regards,
> 	Chr
>

Re: 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Christian Lamparter]

On Wednesday 06 April 2011 16:18:54 Kimmo Huoso wrote:
> And for third even the softmac numbers/statistics might be 
> bogus, im still having problems with the driver while trying to run 
> programs that are strict for smooth connection like online games.
ah, that's why you are so concerned about excessive tx retries & invalid
misc numbers. Well, one of the worst enemy of programs that rely on low latency,
time-critical connection is "bufferbloat". However, by design 11n aggregation
needs to do a fair amount of additional caching & reordering so it's not
really suited for rt data anyway... Given that you have already ruled out your
router/AP and your environment, etc.. as a culprit, the good news is that
someone has already thought about this [QoS + the fact that aggregation is
not enabled for AC_VO]. The bad news is that ar9170's QoS [note: I'm talking
about the device and not the driver] does not work every well unless 11n is
completely disabled [module parameter noht=1 does that].

Regards,
	Chr

Re: 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Kimmo Huoso]

Yep. Thats the reason im in concern, and for a note i did disable the ht 
by adding this noht=1 to module and problem seems to still persist. My 
router is running currently openrwrt Backfire (10.03.1-rc4, r24045) 
which is in dir-825 hardware. There are no problems with router that i 
have noticed. So only solution for me for now is to wait that someone 
adds better code for QoS. Right? At the moment im sticking with 
lts-kernel and ar9170usb.

06.04.2011 18:16, Christian Lamparter kirjoitti:
> On Wednesday 06 April 2011 16:18:54 Kimmo Huoso wrote:
>> And for third even the softmac numbers/statistics might be
>> bogus, im still having problems with the driver while trying to run
>> programs that are strict for smooth connection like online games.
> ah, that's why you are so concerned about excessive tx retries&  invalid
> misc numbers. Well, one of the worst enemy of programs that rely on low latency,
> time-critical connection is "bufferbloat". However, by design 11n aggregation
> needs to do a fair amount of additional caching&  reordering so it's not
> really suited for rt data anyway... Given that you have already ruled out your
> router/AP and your environment, etc.. as a culprit, the good news is that
> someone has already thought about this [QoS + the fact that aggregation is
> not enabled for AC_VO]. The bad news is that ar9170's QoS [note: I'm talking
> about the device and not the driver] does not work every well unless 11n is
> completely disabled [module parameter noht=1 does that].
>
> Regards,
> 	Chr
>

Re: 7d1:3c10 D-Link System DWA-160 Xtreme N Dual Band USB Adapter(rev.A) [Atheros AR9001U-(2)NG] Invalid misc and Tx excessive retries

[Christian Lamparter]

On Wednesday 06 April 2011 17:35:02 Kimmo Huoso wrote:
> So only solution for me for now is to wait that someone 
> adds better code for QoS. Right? At the moment im sticking with 
> lts-kernel and ar9170usb.
good luck with that.