RE: [BUG] rtw89: 8852CE BT A2DP stuttering with WiFi traffic, "timed out to flush queues", bogus -110 dBm RSSI (ThinkPad E16 Gen 2 AMD)

[Ping-Ke Shih <pkshih@realtek.com>]

Mauro <mauro.mossino91@gmail.com> wrote:
> Hello,
> 
> on a ThinkPad E16 Gen 2 AMD with RTL8852CE, Bluetooth A2DP audio stutters
> whenever WiFi carries traffic, even on 5 GHz and with all power saving
> disabled.
> 
> System
>   Machine:   ThinkPad E16 Gen 2 AMD (21M5CTO1WW), Ryzen 7 7735HS
>   BIOS:      R2KET37W (1.26)
>   Kernel:    7.2.9-200.fc44.x86_64 (Fedora 44)
>   linux-fw:  20260916
>   PipeWire 1.6.9, WirePlumber 0.5.18, BlueZ 5.87
> 
> WiFi
>   10ec:c852, subsystem 17aa:5852, driver rtw89_8852ce
>   Firmware 0.27.129.4 (3f1a5302), rtw89/rtw8852c_fw-2.bin
>   chip info CID: 0, CV: 1, AID: 0, ACV: 0, RFE: 1
>   Link: 5200 MHz, 80 MHz, -45 dBm, VHT-MCS 9 NSS 2
> 
> Bluetooth
>   USB 0bda:5852, rtl_bt/rtl8852cu_fw_v2.bin + rtl8852cu_config.bin
>   fw version 0x040d7225, lmp_subver 0x8852
>   Codec in use: <SBC/AAC>
> 
> Settings tried (no improvement):
>   options rtw89_core disable_ps_mode=y
>   options rtw89_pci disable_aspm_l1=y disable_aspm_l1ss=y disable_clkreq=y
>   options btusb enable_autosuspend=n
>   options usbcore autosuspend=-1
>   NetworkManager wifi.powersave=2 (iw: Power save: off)

Thanks. So the problem isn't because of power save.

> 
> Note: ASPM is firmware-controlled on this machine ("OS doesn't have ASPM
> control"), so the rtw89_pci ASPM options may have no effect.
> 
> Observations
> 1) During traffic the driver logs:
>      rtw89_8852ce 0000:03:00.0: timed out to flush queues

Can you describe how the traffic is?

Normally driver needs to flush queues for scanning or rekeying. 
Not sure why it depends on traffic. 

> 2) On a previous boot the log was flooded with:
>      wlp3s0 (phy #0): CQM event: RSSI (-110 dBm) went above threshold
>    -110 dBm looks like a placeholder value, not a real measurement
>    (real signal -45..-62 dBm).

I don't see the event in attachment. Where did you find it?

No clear idea why somehow hardware reported abnormal RSSI.


> 3) rtw89 IRQ rate: ~460/s under normal use, no interrupt storm.
> 
> Possibly relevant: the TSC is marked unstable at boot on this machine
> (firmware issue, reported to platform-driver-x86) and the clocksource is
> HPET, which may make the timing-sensitive coex path worse.
> 
> Similar report: https://discussion.fedoraproject.org/t/142165

It looks like there are many tests in the thread. Can you bisect the
cause in your side like the thread did? 

Ping-Ke