[PATCH] wifi: ath11k: pci: Fix msi_irq crash on driver unload with QCN9074 PCIe WiFi 6 modules

[Baochen Qiang]

On 4/16/2025 6:09 PM, balsam.chihi at moment.tech wrote:
> From: Balsam CHIHI <balsam.chihi at moment.tech>
> 
> This patch addresses a crash issue that occurs when unloading the
> ath11k_pci driver with QCN9074 PCIe WiFi 6 modules.
> The crash is caused by the driver attempting to perform reset
> operations during unload, leading to a synchronous external abort

Do we know the root cause of the synchronous external abort?

> and kernel panic, as indicated by the error log:
> 
> [ 5615.902985] Internal error: synchronous external abort: 0000000096000210 [#1] SMP
> ...
> [ 5616.056382] CPU: 7 PID: 12605 Comm: procd Tainted: G O 6.6.73 #0
> ...
> [ 5616.069876] pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> [ 5616.076841] pc : ath11k_pci_get_msi_irq+0x18b4/0x1914 [ath11k_pci]
> [ 5616.083035] lr : ath11k_pcic_init_msi_config+0x98/0xc4 [ath11k]
> [ 5616.163712] Call trace:
> [ 5616.166153] ath11k_pci_get_msi_irq+0x18b4/0x1914 [ath11k_pci]
> [ 5616.171993] ath11k_pcic_init_msi_config+0x98/0xc4 [ath11k]
> [ 5616.177583] ath11k_pcic_read32+0x30/0xb4 [ath11k]
> [ 5616.182391] ath11k_pci_get_msi_irq+0x528/0x1914 [ath11k_pci]
> [ 5616.188143] ath11k_pci_get_msi_irq+0x147c/0x1914 [ath11k_pci]
> [ 5616.193983] ath11k_pci_get_msi_irq+0x1764/0x1914 [ath11k_pci]
> [ 5616.199822] pci_device_shutdown+0x34/0x44
> [ 5616.203923] device_shutdown+0x160/0x268
> [ 5616.207847] kernel_restart+0x40/0xc0
> [ 5616.211512] __do_sys_reboot+0x104/0x23c
> [ 5616.215436] __arm64_sys_reboot+0x24/0x30
> [ 5616.219447] do_el0_svc+0x6c/0xfc
> [ 5616.222761] el0_svc+0x28/0x9c
> [ 5616.225817] el0t_64_sync_handler+0x120/0x12c
> [ 5616.230174] el0t_64_sy
> [ 5616.233839] Code: f94e0a80 92404a73 91420273 8b130013 (b9400273)
> [ 5616.239932] ---[ end trace 0000000000000000 ]---
> [ 5616.244547] Kernel panic - not syncing: synchronous external abort: Fatal exception in interrupt
> [ 5616.253343] Kernel Offset: disabled
> [ 5616.256827] CPU features: 0x0,00000000,00020000,1000400b
> [ 5616.262138] Memory Limit: none
> [ 5616.265188] Rebooting in 3 seconds..
> [ 5620.268926] Unable to restart system
> [ 5620.272503] Reboot failed -- System halted
> 
> The fix involves adding a conditional check for the power state before
> performing the reset operations in the ath11k_pci_sw_reset function.
> This ensures that the reset functions are only called when loading the driver,

The ath11k_pci_soc_global_reset() is called to make sure the device is in the expected
state, such that after machine restart, the device can be successfully enumerated. If
skipped, the chip may be not detected. At least this is the case for WCN6855, for others I
am not sure.


> preventing the crash during driver unload.
> 
> Signed-off-by: Balsam CHIHI <balsam.chihi at moment.tech>
> ---
>  drivers/net/wireless/ath/ath11k/pci.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/pci.c b/drivers/net/wireless/ath/ath11k/pci.c
> index 78444f8ea1535..b9c8963c17047 100644
> --- a/drivers/net/wireless/ath/ath11k/pci.c
> +++ b/drivers/net/wireless/ath/ath11k/pci.c
> @@ -368,9 +368,12 @@ static void ath11k_pci_sw_reset(struct ath11k_base *ab, bool power_on)
>  	}
>  
>  	ath11k_mhi_clear_vector(ab);
> -	ath11k_pci_clear_dbg_registers(ab);
> -	ath11k_pci_soc_global_reset(ab);
> -	ath11k_mhi_set_mhictrl_reset(ab);
> +
> +	if (power_on) {
> +		ath11k_pci_clear_dbg_registers(ab);
> +		ath11k_pci_soc_global_reset(ab);
> +		ath11k_mhi_set_mhictrl_reset(ab);
> +	}
>  }
>  
>  static void ath11k_pci_init_qmi_ce_config(struct ath11k_base *ab)