Re: [PATCH net-next v2 7/9] net: ethernet: ti: cpsw_ale: add policer save restore for PM sleep

[Roger Quadros <rogerq@kernel.org>]

On 08/05/2025 16:49, Paolo Abeni wrote:
> On 5/5/25 6:26 PM, Roger Quadros wrote:
>> diff --git a/drivers/net/ethernet/ti/cpsw_ale.c b/drivers/net/ethernet/ti/cpsw_ale.c
>> index ce216606d915..4e29702b86ea 100644
>> --- a/drivers/net/ethernet/ti/cpsw_ale.c
>> +++ b/drivers/net/ethernet/ti/cpsw_ale.c
>> @@ -1823,3 +1823,45 @@ int cpsw_ale_policer_set_entry(struct cpsw_ale *ale, u32 policer_idx,
>>  
>>  	return 0;
>>  }
>> +
>> +void cpsw_ale_policer_save(struct cpsw_ale *ale, u32 *data)
>> +{
>> +	int i, idx;
>> +
>> +	for (idx = 0; idx < ale->params.num_policers; idx++) {
>> +		cpsw_ale_policer_read_idx(ale, idx);
>> +
>> +		for (i = 0; i < CPSW_ALE_POLICER_ENTRY_WORDS; i++)
>> +			data[i] = readl_relaxed(ale->params.ale_regs +
>> +						ALE_POLICER_PORT_OUI + 4 * i);
>> +
>> +		regmap_field_write(ale->fields[ALE_THREAD_CLASS_INDEX], idx);
>> +		data[i++] = readl_relaxed(ale->params.ale_regs +
>> +					ALE_THREAD_VAL);
>> +		data += i * 4;
> 
> I'm confused by the '* 4' part. I think that you just need:
> 		data += i
> 
> to move to the next policer context. I *think* the current code causes
> OoB?!?

Good catch. Since data is u32 pointer we should not need '* 4'

> 
> /P
> 

-- 
cheers,
-roger