[ath9k-devel] Weird Atheros 9485 http connecting problems

[Arkh4mKn1ght <arkh4mkn1ght@gmail.com>]

Ok i tried the patch but the issue happened again :(

After resuming my laptop from hibernation i saw in 
/var/log/kernel/errors.log these messages again:

May 21 09:20:10 localhost kernel: mei_me 0000:00:16.0: suspend
May 21 09:20:10 localhost kernel: ath: phy1: RX failed to go idle in 10 
ms RXSM=0xdeadbeef
May 21 09:20:10 localhost kernel: ath: phy1: DMA failed to stop in 10 ms 
AR_CR=0xdeadbeef AR_DIAG_SW=0xdeadbeef DMADBG_7=0xdeadbeef
May 21 09:20:11 localhost kernel: dpm_run_callback(): 
usb_dev_restore+0x0/0x20 [usbcore] returns -107
May 21 09:20:11 localhost kernel: PM: Device 1-1 failed to restore 
async: error -107

ASPM was disabled on root complex and atheros card.powersave was off.
After 1 hour and 10 minutes of those messages that problem arise. I have 
noticed that the problem arises always after some time of resuming the 
laptop from hibernation.

Its really odd, because when it does happen, browsing the web is 
painfully slow with the "connecting to domain.com..." message in the 
browser status bar... it does load the website but after some minutes..
Meanwhile at the same time if i try to download a torrent using a 
bittorrent client i can do it without problems at full speed. If i pause 
the torrent, browsing websites still behaves slowly until i reboot the 
laptop.

At first i thought that i have a defective atheros card, but why i can 
download torrents at full speed when this problem arises?
What options do i have now? i have tested all combinations of ASPM 
enabled/disabled and powersave on/off.

I have seen that there is a BIOS update for my laptop, but im unsure it 
could help for this at all (besides that i would have to install windows 
8 to install the friggin bios update)

Greetings

Oskar



On 05/19/2014 09:32 PM, Adrian Chadd wrote:
> Right;
>
>> powersave is nothing but 802.11 protocol implementation to put the
>> station into network sleep when it is idle and wake it up whenever it is
>> active. IIRC it is not related to ASPM. So yes powersave would work event
>> without ASPM.
>>
>> Since the issue is happending only when ASPM is enabled, please test the
>> patch with ASPM.
> Just to be clear - the powersave support when loading ath9k also puts
> the MAC/PHY to sleep and wakes it up as appropriate.
>
> You're seeing 0xdeadc0de which the internal AHB glue reports if
> there's a transaction timeout (ie, the MAC/PHY is asleep or powered
> down when you went to access it.)
>
> ASPM has to do with the PHYs at both ends being awake (or not.) I
> think you'd either trigger some pcie error or read 0xffffffff if
> things were screwed up there.
>
> So I'd try with ASPM enabled/disabled as well as ps_enable on or off.
> It could be an interplay between the two as well.
>
>
> -a
>