wcn6855 2.1 - slow download speed on any pairwise cipher except CCMP

[Baochen Qiang]

On 6/4/2024 5:29 AM, Yaroslav Isakov wrote:
> Hello! Thank you for patch! I tried to apply this patch on top of 6.9.3,
> but it failed to apply - what kernel version should I use, to test it?
> 
That patch is based on ath-202405301555 tag of Kalle's ath.git:

https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/

but I suppose it could also be applied cleanly on ToT of ath.git.


> On Fri, 31 May 2024, 04:18 Baochen Qiang, <quic_bqiang at quicinc.com> wrote:
> 
>>
>>
>> On 4/15/2024 7:26 PM, Yaroslav Isakov wrote:
>>> Hello!
>>>
>>> вс, 14 апр. 2024 г. в 10:12, Yaroslav Isakov <yaroslav.isakov at gmail.com
>>> :
>>>>
>>>> Hello!
>>>>
>>>> вт, 2 апр. 2024 г. в 02:35, Baochen Qiang <quic_bqiang at quicinc.com>:
>>>>>
>>>>>
>>>>>
>>>>> On 4/1/2024 10:53 PM, Jeff Johnson wrote:
>>>>>> On 3/31/2024 8:49 AM, Yaroslav Isakov wrote:
>>>>>>> Hello! When I tried to switch my OpenWRT router from default CCMP
>>>>>>> pairwise cipher to any other, supported by both my router (Asus
>>>>>>> TUF-AX6000) and my laptop (Thinkpat T14 Gen 4 AMD with WCN6855 2.1),
>> I
>>>>>>> saw that download dropped from 1Gbit/s to 100 MBit/s. When I changed
>>>>>>> back to CCMP, download speed recovered back to 1Gbit/s. Upload speed
>>>>>>> is good, though, for all ciphers - 1Gbit/s.
>>>>>>>
>>>>>>> Initially, I thought, that issue with either router hardware, or
>>>>>>> openwrt driver - but, I tested the same settings on another laptop -
>>>>>>> T14 Gen3 Intel - and with Intel WIFI, both download and upload speed
>>>>>>> was 1Gbit/s on any pairwise cipher.
>>>>>>>
>>>>>>> This issue is 100% reproducible for me, it happens all the times I'm
>>>>>>> switching to any pairwise cipher other than CCMP.
>>>>>>>
>>>>>>> I'm on Linux 6.7.6, compiled from sources on Gentoo.
>>>>> I'm not familiar with Gentoo, is there any private changes in it
>>>>> compared to upstream ath11k? If there indeed is, could you try Kalle's
>>>>> ath11k tree?
>>>>>
>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/
>>>>>
>>>> Sorry, missed the message. Gentoo kernel I'm using is very close to
>>>> upstream. I will try ath11.git, and will report there, thank you.
>>>>
>>>
>>> I checked ath.git, master branch - it has the exact same behavior as
>>> kernel I used - 1Gbit download with CCMP, 100Mbit download with GCMP.
>>> For me, this issue is easily reproducible, it happens 100% if I'm
>>> switching to pairwise_cipher (and therefore, groupwise, too) other
>>> than CCMP.
>> I just submitted a patch which may fix this issue, could you merge it and
>> try again?
>>
>>
>> https://patchwork.kernel.org/project/linux-wireless/patch/20240531021228.2846-1-quic_bqiang@quicinc.com/
>>
>>>>>>>
>>>>>>> Here is my dmesg for ath11k:
>>>>>>> [   13.092519] ath11k_pci 0000:02:00.0: BAR 0: assigned [mem
>>>>>>> 0x78600000-0x787fffff 64bit]
>>>>>>> [   13.093400] ath11k_pci 0000:02:00.0: enabling device (0000 ->
>> 0002)
>>>>>>> [   13.095723] ath11k_pci 0000:02:00.0: MSI vectors: 32
>>>>>>> [   13.096260] ath11k_pci 0000:02:00.0: wcn6855 hw2.1
>>>>>>> [   13.974707] ath11k_pci 0000:02:00.0: chip_id 0x12 chip_family 0xb
>>>>>>> board_id 0xff soc_id 0x400c1211
>>>>>>> [   13.974732] ath11k_pci 0000:02:00.0: fw_version 0x1106196e
>>>>>>> fw_build_timestamp 2024-01-12 11:30 fw_build_id
>>>>>>> WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
>>>>>>> [   14.315774] ath11k_pci 0000:02:00.0: Failed to set the requested
>>>>>>> Country regulatory setting
>>>>>>> [   14.316046] ath11k_pci 0000:02:00.0: Failed to set the requested
>>>>>>> Country regulatory setting
>>>>>>> [  337.889222] ath11k_pci 0000:02:00.0: failed to enqueue rx buf: -28
>>>>>>>
>>>>>>
>>>>>> There was another reporting of that error in this thread:
>>>>>>
>> https://lore.kernel.org/all/5c63a3a2-29fe-444c-96f1-f87c89d7af39@gmail.com/
>>>>>>
>>>>> I don't think this is the same issue as reported by James there
>> although
>>>>> same error message. For now I don't know the root cause to that message
>>>>> but it's unlikely it causes some tput issue.
>>>>>
>>>>> Let's wait for Yasoslav verifying it on latest ath.git.
>>>>>
>>>>>> +Baochen Qiang for awareness
>>>>>>
>>>>>> /jeff
>>
>