[ath9k-devel] EnGenius EMP9602 mini-PCI Wireless Card

[ath9k-devel] EnGenius EMP9602 mini-PCI Wireless Card

[Cristian Talle]

Hi,

I'm trying to get the EnGenius EMP9602 mini-PCI Wireless Card (Atheros 
AR9160 + AR 9103) working on a Micrel KS8695P system (arm 922T) with 
kernel 2.6.32.5.
The card runs fine except for the messages I'm getting in the system log :

Feb 2 17:02:33 (none) user.debug kernel: ath9k: received PCI FATAL 
interrupt...
Feb 2 17:02:33 (none) user.debug kernel: ath9k: Failed to stop TX DMA in 
100 ms...
Feb 2 17:02:33 (none) user.debug kernel: ath9k: Unable to stop TxDMA. 
Reset HAL...

The driver doesn't crash (it's built in the kernel) but downloading 
through the wireless link is VERY unreliable (100-200kbps) while the 
upload is constant (so far I got 2Mbps but it was limited by the other 
endpoint).

We have the PCI bus running @ 33MHz with the CPU DMA cache size of 3KBytes.

I've seen a lot of similar posts about the same type of error and all 
the replies I found say - it's harmless, the system is recovering.. etc.
It's true that the driver seems to be fairly stable but the _cause_ of 
this problem needs to be found so that performance is not impacted.

I also tried the latest version of the kernel (2.6.33-rc6) but I get the 
same error, on top of the fact that ioctl seems to have been changed (I 
get Operation Not Supported when trying to add wlan0 to a bridge).

Is there anyone out there that has ath9k working reliably with AR9160 or 
is the Tx DMA problem common across the board?

Thank you,
Cristian

[ath9k-devel] EnGenius EMP9602 mini-PCI Wireless Card

[rootkit85 at yahoo.it]

On Wed, Feb 3, 2010 at 8:48 PM, Cristian Talle <ctalle@voiceway.ca> wrote:
> Hi,
>
> I'm trying to get the EnGenius EMP9602 mini-PCI Wireless Card (Atheros
> AR9160 + AR 9103) working on a Micrel KS8695P system (arm 922T) with
> kernel 2.6.32.5.
> The card runs fine except for the messages I'm getting in the system log :
>
> Feb 2 17:02:33 (none) user.debug kernel: ath9k: received PCI FATAL
> interrupt...
> Feb 2 17:02:33 (none) user.debug kernel: ath9k: Failed to stop TX DMA in
> 100 ms...
> Feb 2 17:02:33 (none) user.debug kernel: ath9k: Unable to stop TxDMA.
> Reset HAL...
>
> The driver doesn't crash (it's built in the kernel) but downloading
> through the wireless link is VERY unreliable (100-200kbps) while the
> upload is constant (so far I got 2Mbps but it was limited by the other
> endpoint).
>
> We have the PCI bus running @ 33MHz with the CPU DMA cache size of 3KBytes.
>
> I've seen a lot of similar posts about the same type of error and all
> the replies I found say - it's harmless, the system is recovering.. etc.
> It's true that the driver seems to be fairly stable but the _cause_ of
> this problem needs to be found so that performance is not impacted.
>
> I also tried the latest version of the kernel (2.6.33-rc6) but I get the
> same error, on top of the fact that ioctl seems to have been changed (I
> get Operation Not Supported when trying to add wlan0 to a bridge).
>
> Is there anyone out there that has ath9k working reliably with AR9160 or
> is the Tx DMA problem common across the board?
>
> Thank you,
> Cristian
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel
>

I had such a similar throughput too, but I solved disabling powersave
in the driver.
Let me know if it helps for you

[ath9k-devel] EnGenius EMP9602 mini-PCI Wireless Card

[vladi470]

 <rootkit85 <at> yahoo.it> writes:

> 
> On Wed, Feb 3, 2010 at 8:48 PM, Cristian Talle <ctalle <at> voiceway.ca> 
wrote:
> > I'm trying to get the EnGenius EMP9602 mini-PCI Wireless Card (Atheros
> > AR9160 + AR 9103) working on a Micrel KS8695P system (arm 922T) with
> > kernel 2.6.32.5.
> > The card runs fine except for the messages I'm getting in the system log :
> >
> > Feb 2 17:02:33 (none) user.debug kernel: ath9k: received PCI FATAL
> > interrupt...
> > Feb 2 17:02:33 (none) user.debug kernel: ath9k: Failed to stop TX DMA in
> > 100 ms...
> > Feb 2 17:02:33 (none) user.debug kernel: ath9k: Unable to stop TxDMA.
> > Reset HAL...
> > ...
> > Thank you,
> > Cristian
> > _______________________________________________
> > ath9k-devel mailing list
> > ath9k-devel <at> lists.ath9k.org
> > https://lists.ath9k.org/mailman/listinfo/ath9k-devel
> >
> 
> I had such a similar throughput too, but I solved disabling powersave
> in the driver.
> Let me know if it helps for you
> 

Hi Cristian, All
I have same issue on another board and it is very annoying:
In my case it is not recoverable, means there is no further traffic.
Could you advise what can be a solution?

Thanks,
Vladimir

[ath9k-devel] EnGenius EMP9602 mini-PCI Wireless Card

[Adrian Chadd]

A PCI fatal interrupt means something about the PCI bus transaction
annoyed the NIC PCI glue hardware and it decided that was it. If you
add some patches to ath9k to print out the value of the sync and async
interrupt cause registers whenever a fatal interrupt occurs, I can try
and decode the error for you.

I'm far from being cluey in this particular area, but you may want to
double and triple check all the PCI bus code is all up to snuff. Maybe
some of the latency/burst size values (that are written into the
atheros EEPROM and programmed into the PCI BAR at hard-reset) are of
sufficient values to annoy the PCI bus on the ARM.

Unfortunately this may be one of those things you have to consult a
guru who has inner clue about the ARM PCI bus hardware, the ARM SoC
and general PCI bus timing clue.

Good luck,


Adrian