[PATCH v7 4/9] ath11k: Add register access logic for WCN6750

[Manikanta Pubbisetty]

On 5/26/2022 12:16 AM, Maxime Bizon wrote:
> 
> On Fri, 2022-04-29 at 22:34 +0530, Manikanta Pubbisetty wrote:
> 
> Hello Manikanta,
> 
>> Tested-on: QCN9074 hw1.0 PCI WLAN.HK.2.5.0.1-01100-
>> QCAHKSWPL_SILICONZ-1
> 
> Nope your patch breaks QCN9074:
> 
> [   13.660963] ath11k_pci 0000:03:00.0: failed to set pcie link register 0x01e0e0a8: 0xffffffff != 0x00000010
> [   13.675994] ath11k_pci 0000:03:00.0: failed to set sysclk: -110
> 
> device still seem to work though

Thanks for letting me know about this, IIRC I don't remember 
encountering this problem in my testing. Just for my understanding, have 
you reverted this change and confirmed that these errors go away ?

> 
>> @@ -134,16 +134,13 @@ EXPORT_SYMBOL(ath11k_pcic_init_msi_config);
>>   static inline u32 ath11k_pcic_get_window_start(struct ath11k_base *ab,
>>   					       u32 offset)
>>   {
>> -	u32 window_start;
>> +	u32 window_start = 0;
>>   
>> -	/* If offset lies within DP register range, use 3rd window */
>>   	if ((offset ^ HAL_SEQ_WCSS_UMAC_OFFSET) < ATH11K_PCI_WINDOW_RANGE_MASK)
>> -		window_start = 3 * ATH11K_PCI_WINDOW_START;
>> -	/* If offset lies within CE register range, use 2nd window */
>> -	else if ((offset ^ HAL_CE_WFSS_CE_REG_BASE) < ATH11K_PCI_WINDOW_RANGE_MASK)
>> -		window_start = 2 * ATH11K_PCI_WINDOW_START;
>> -	else
>> -		window_start = ATH11K_PCI_WINDOW_START;
>> +		window_start = ab->hw_params.dp_window_idx * ATH11K_PCI_WINDOW_START;
>> +	else if ((offset ^ HAL_SEQ_WCSS_UMAC_CE0_SRC_REG(ab)) <
>> +		 ATH11K_PCI_WINDOW_RANGE_MASK)
>> +		window_start = ab->hw_params.ce_window_idx * ATH11K_PCI_WINDOW_START;
>>   
>>   	return window_start;
>>   }
> 
> 
> for some offsets, previous code could return ATH11K_PCI_WINDOW_START,
> whereas new code now returns 0
> 
> 
>> @@ -162,19 +159,12 @@ void ath11k_pcic_write32(struct ath11k_base *ab, u32 offset, u32 value)
>>   
>>   	if (offset < ATH11K_PCI_WINDOW_START) {
>>   		iowrite32(value, ab->mem  + offset);
>> -	} else {
>> -		if (ab->hw_params.static_window_map)
>> -			window_start = ath11k_pcic_get_window_start(ab, offset);
>> -		else
>> -			window_start = ATH11K_PCI_WINDOW_START;
>> -
>> -		if (window_start == ATH11K_PCI_WINDOW_START &&
>> -		    ab->pci.ops->window_write32) {
>> -			ab->pci.ops->window_write32(ab, offset, value);
>> -		} else {
>> -			iowrite32(value, ab->mem + window_start +
>> -				  (offset & ATH11K_PCI_WINDOW_RANGE_MASK));
>> -		}
>> +	} else if (ab->hw_params.static_window_map) {
>> +		window_start = ath11k_pcic_get_window_start(ab, offset);
>> +		iowrite32(value, ab->mem + window_start +
>> +			  (offset & ATH11K_PCI_WINDOW_RANGE_MASK));
>> +	} else if (ab->pci.ops->window_write32) {
>> +		ab->pci.ops->window_write32(ab, offset, value);
>>   	}
>>
> 
> with previous code on QCN9074, when ath11k_pcic_get_window_start()
> returned ATH11K_PCI_WINDOW_START, then it would call window_write32()
> 
> with new code on QCN9074, static_window_map is true, so window_write32
> will never be called.
> 
>>   u32 ath11k_pcic_read32(struct ath11k_base *ab, u32 offset)
> 
> ditto here
> 

If you could please confirm that reverting this change helps, I can send 
a patch to fix this.

Thanks,
Manikanta