[PATCH] ath12k: add module param to control MSI mode and address width

[Jeff Johnson]

On 6/24/2025 9:50 AM, Govind Singh wrote:
> This patch adds two module parameters to the ath12k PCI driver.

Please use imperative voice.
And it is preferential to first describe the problem, and then describe the
proposed solution.

> 
>  - ath12k_msi_mode: Allows selecting the MSI allocation mode.
>    * 0 = multi-vector MSI (default)
>    * 1 = single-vector MSI
> 
>  - ath12k_32bit_msi: Forces the use of 32-bit MSI addressing
>    by setting pdev->no_64bit_msi = 1 before IRQ vector allocation.
> 
> These options are useful for working around hardware or platform
> limitations(ex:  i.MX 8M Plus) where 64-bit MSI or multi-vector MSI
> allocations fails or cause stability issues.

So there is no way to detect this?

> 
> The original MSI allocation logic remains unchanged unless the
> parameters are explicitly set via modprobe or kernel boot options.
> 
> Tested-on: QCN9274 hw2.0 WLAN.WBE.1.4.1-00103-QCAHKSWPL_SILICONZ-1 with
>            i.MX 8M Plus SOC

Don't split tags, and need a blank line after Tested-on since it is an
ath-only tag, not an upstream tag

> Signed-off-by: Govind Singh <govind.sk85 at gmail.com>
> ---
>  drivers/net/wireless/ath/ath12k/pci.c | 25 +++++++++++++++++++++----
>  drivers/net/wireless/ath/ath12k/pci.h |  5 +++++
>  2 files changed, 26 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath12k/pci.c b/drivers/net/wireless/ath/ath12k/pci.c
> index 1f3cfd9b89fd..4f58cb1e03f6 100644
> --- a/drivers/net/wireless/ath/ath12k/pci.c
> +++ b/drivers/net/wireless/ath/ath12k/pci.c
> @@ -45,6 +45,14 @@
>  #define DOMAIN_NUMBER_MASK		GENMASK(7, 4)
>  #define BUS_NUMBER_MASK			GENMASK(3, 0)
>  
> +int ath12k_msi_mode;

should be static, make W=1 reports:
drivers/net/wireless/ath/ath12k/pci.c:48:5: warning: symbol 'ath12k_msi_mode'
was not declared. Should it be static?

> +module_param(ath12k_msi_mode, int, 0644);

why is this writable via sysfs?

> +MODULE_PARM_DESC(ath12k_msi_mode, "MSI mode: 0 = multi-vector (default), 1 = single-vector");
> +
> +bool ath12k_32bit_msi;

should be static, make W=1 reports:
drivers/net/wireless/ath/ath12k/pci.c:52:6: warning: symbol 'ath12k_32bit_msi'
was not declared. Should it be static?

> +module_param(ath12k_32bit_msi, bool, 0644);

why is this writable via sysfs?

> +MODULE_PARM_DESC(ath12k_32bit_msi, "Force 32-bit MSI addressing");
> +
>  static const struct pci_device_id ath12k_pci_id_table[] = {
>  	{ PCI_VDEVICE(QCOM, QCN9274_DEVICE_ID) },
>  	{ PCI_VDEVICE(QCOM, WCN7850_DEVICE_ID) },
> @@ -773,10 +781,19 @@ static int ath12k_pci_msi_alloc(struct ath12k_pci *ab_pci)
>  	int num_vectors;
>  	int ret;
>  
> -	num_vectors = pci_alloc_irq_vectors(ab_pci->pdev,
> -					    msi_config->total_vectors,
> -					    msi_config->total_vectors,
> -					    PCI_IRQ_MSI);
> +	/* Set 32-bit MSI flag early if requested */
> +	if (ath12k_32bit_msi)
> +		ab_pci->pdev->no_64bit_msi = 1;
> +
> +	/* Force single MSI mode if requested */
> +	if (ath12k_msi_mode == ATH12K_MSI_VEC_SINGLE) {
> +		num_vectors = -EINVAL; /* Force fallback path */
> +	} else {
> +		num_vectors = pci_alloc_irq_vectors(ab_pci->pdev,
> +						    msi_config->total_vectors,
> +						    msi_config->total_vectors,
> +						    PCI_IRQ_MSI);
> +	}
>  
>  	if (num_vectors == msi_config->total_vectors) {
>  		set_bit(ATH12K_PCI_FLAG_MULTI_MSI_VECTORS, &ab_pci->flags);
> diff --git a/drivers/net/wireless/ath/ath12k/pci.h b/drivers/net/wireless/ath/ath12k/pci.h
> index d1ec8aad7f6c..4fa8bb619cd9 100644
> --- a/drivers/net/wireless/ath/ath12k/pci.h
> +++ b/drivers/net/wireless/ath/ath12k/pci.h
> @@ -92,6 +92,11 @@ enum ath12k_pci_flags {
>  	ATH12K_PCI_FLAG_MULTI_MSI_VECTORS,
>  };
>  
> +enum ath12k_msi_mode {
> +	ATH12K_MSI_VEC_AUTO = 0,
> +	ATH12K_MSI_VEC_SINGLE = 1,
> +};

I'd put this in pci.c since this is not part of an external driver interface
to pci.c

> +
>  struct ath12k_pci_ops {
>  	int (*wakeup)(struct ath12k_base *ab);
>  	void (*release)(struct ath12k_base *ab);