ath10k-regression due to "ath10k: fix napi_poll budget overflow" c9353bf483d3 (Was: Re: ath10k-regression in 4.14: Connections aborts with "failed to extract amsdu: -11"

[Thorsten Leemhuis]

Lo! On 03.10.2017 01:40, Ryan Hsu wrote:
> On 10/01/2017 01:59 AM, Thorsten Leemhuis wrote:
>> Lo! The wifi connection of my Dell XPS13 (9360) with its QCA6174
>> sometimes suddenly stops working since I switched to 4.14-rc2+. Every
>> time it happens, there is this error message in dmesg:
>>> ath10k_pci 0000:3a:00.0: failed to extract amsdu: -11
>> I have to switch wifi off and on with the hotkey to reconnect. I can
>> trigger the aborts by starting a big download and waiting a few minutes.
>> Sometimes the connections aborts during normal load. Installing the
>> latest firmware didn't help. The wifi works just fine with 4.13.3. While
>> investigating this I noticed a few messages in dmesg that only appear in
>> 4.14-rc (I used 35dbba31be52):
> You do run the 4.13.3 v.s 4.14-rc with the same QCA6174 firmwrae, right?
> Just want to understand the test setup here so that I could give it a try myself, and in 11ac or 11n mode you're testing?

Yup, same firmware (reproduced it with
firmware-6.bin_WLAN.RM.4.4.1-00058-QCARMSWP-1 before bisecting). And the
problem showed up with 2g and 5g networks. But while investigating it I
noticed the problem does not show up with all wifi routers. It happens
with my Fritz!Box 6490 Cable and another Fritz!Box I tried, but not with
the wifi network at work (no idea what kind of routers are installed
there; I can try to find out if it matters).

>>> 3a:00.0 Network controller [0280]: Qualcomm Atheros QCA6174 802.11ac Wireless Network Adapter [168c:003e] (rev 32)
> […] Would you mind do a bisect to locate the failure, please?

Did that yesterday and it turned out it's due to commit c9353bf483d3
(ath10k: fix napi_poll budget overflow). Reverting it on top of linux
master from yesterday made the wifi connection stable again for me.

Ciao, Thorsten

P.S.: For completeness: https://git.kernel.org/linus/c9353bf483d3

> commit c9353bf483d3724c116a9d502c0ead9cec54a61a (refs/bisect/bad)
> Author: Ryan Hsu <ryanhsu at qti.qualcomm.com>
> Date:   Tue Aug 22 14:44:02 2017 -0700
> 
>     ath10k: fix napi_poll budget overflow
>     
>     In napi_poll, the budget number is used to control the amount of packets
>     we should handle per poll to balance the resource in the system.
>     
>     In the list of the amsdu packets reception, we check if there is budget
>     count left and handle the complete list of the packets, that it will have
>     chances the very last list will over the budget leftover.
>     
>     So adding one more parameter - budget_left, this would help while
>     traversing the list to avoid handling more than the budget given.
>     
>     Reported-by: Andrey Ryabinin <aryabinin at virtuozzo.com>
>     Fix-suggested-by: Igor Mitsyanko <igor.mitsyanko.os at quantenna.com>
>     Link: https://lkml.kernel.org/r/26670dce-4dd2-f8e4-0e14-90d74257e739@virtuozzo.com
>     Signed-off-by: Ryan Hsu <ryanhsu at qti.qualcomm.com>
>     Signed-off-by: Kalle Valo <kvalo at qca.qualcomm.com>