ath12k_pci errors and loss of connectivity in 6.12.y branch

[Baochen Qiang]

On 7/2/2025 10:53 PM, Matt Mower wrote:
>> Matt, could you help enable verbose ath12k log to verify my guess?
> 
> Here are kernel logs with ath12k debugging enabled:

Thanks Matt.

I see firmware crash before IOMMU fault in both logs, which verifies my guess.

> 1. WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
>    https://cmphys.com/ath12k/dmesg-6.12.35-ath12kdebug-fw0x100301e1-20250702.log

[   91.625809] ath12k_pci 0000:c2:00.0: mhi notify status reason MHI_CB_EE_RDDM
[   91.625916] ath12k_pci 0000:c2:00.0: reset starting
[   91.674375] ath12k_pci 0000:c2:00.0: waiting recovery start...
[   91.679445] ath12k_pci 0000:c2:00.0: setting mhi state: POWER_OFF(3)
[   91.680721] ath12k_pci 0000:c2:00.0: qmi wifi fw del server
[   91.680754] ath12k_pci 0000:c2:00.0: setting mhi state: DEINIT(1)
[   91.681842] ath12k_pci 0000:c2:00.0: cookie:0x0
[   91.681858] ath12k_pci 0000:c2:00.0: WLAON_WARM_SW_ENTRY 0x14c4e54
[   91.687109] ath12k_pci 0000:c2:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0010
address=0xfe980000 flags=0x0020]


> 2. WLAN.HMT.1.1.c5-00284.1-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
>    https://cmphys.com/ath12k/dmesg-6.12.35-ath12kdebug-fw0x1108811c-20250702.log
> 

[  113.621429] ath12k_pci 0000:c2:00.0: mhi notify status reason MHI_CB_EE_RDDM
[  113.621794] ath12k_pci 0000:c2:00.0: reset starting
[  113.670134] ath12k_pci 0000:c2:00.0: waiting recovery start...
[  113.675177] ath12k_pci 0000:c2:00.0: setting mhi state: POWER_OFF(3)
[  113.676331] ath12k_pci 0000:c2:00.0: setting mhi state: DEINIT(1)
[  113.676581] ath12k_pci 0000:c2:00.0: qmi wifi fw del server
[  113.676874] ath12k_pci 0000:c2:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0010
address=0xfea50000 flags=0x0020]

> I captured these after setting CONFIG_ATH12K_DEBUG=y and running "echo
> 0xffffffff > /sys/module/ath12k/parameters/debug_mask" during boot
> (using @reboot in crontab).

Unfortunately I can not tell the root cause to the firmware crash from host log.
Internally I will try to repro this issue, in the meanwhile, Matt, could you help do some
more work to narrow down the problematic change?