Re: [PATCH net-next v3 04/12] net: ethernet: oa_tc6: implement software reset

[<Parthiban.Veerasooran@microchip.com>]

Hi Andrew,

On 07/03/24 6:05 am, Andrew Lunn wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
>> +/* Status Register #0 */
>> +#define OA_TC6_REG_STATUS0                   0x0008
>> +#define STATUS0_RESETC                               BIT(6)  /* Reset Complete */
>> +
>>   /* Control command header */
>>   #define OA_TC6_CTRL_HEADER_DATA_NOT_CTRL     BIT(31)
>>   #define OA_TC6_CTRL_HEADER_WRITE             BIT(29)
>> @@ -24,6 +34,8 @@
>>                                                (OA_TC6_CTRL_MAX_REGISTERS *\
>>                                                OA_TC6_CTRL_REG_VALUE_SIZE) +\
>>                                                OA_TC6_CTRL_IGNORED_SIZE)
>> +#define STATUS0_RESETC_POLL_DELAY            5
>> +#define STATUS0_RESETC_POLL_TIMEOUT          100
>>
>>   /* Internal structure for MAC-PHY drivers */
>>   struct oa_tc6 {
>> @@ -279,6 +291,39 @@ int oa_tc6_write_register(struct oa_tc6 *tc6, u32 address, u32 value)
>>   }
>>   EXPORT_SYMBOL_GPL(oa_tc6_write_register);
>>
>> +static int oa_tc6_read_sw_reset_status(struct oa_tc6 *tc6)
>> +{
>> +     u32 regval;
>> +     int ret;
>> +
>> +     ret = oa_tc6_read_register(tc6, OA_TC6_REG_STATUS0, &regval);
>> +     if (ret)
>> +             return 0;
>> +
>> +     return regval;
> 
> The function name does not really fit what the function does. The
> function returns OA_TC6_REG_STATUS0. I assume it has more bits in it
> than just STATUS0_RESETC. So either this function should be called
> oa_tc6_read_status0, or you should mask regval with STATUS0_RESETC, so
> that it does actually return the sw reset status.
Ok, as we do the masking in the calling function readx_poll_timeout, as 
you suggested I will change the function name as oa_tc6_read_status0 in 
the next version.
> 
>> +static int oa_tc6_sw_reset_macphy(struct oa_tc6 *tc6)
>> +{
>> +     u32 regval = RESET_SWRESET;
>> +     int ret;
>> +
>> +     ret = oa_tc6_write_register(tc6, OA_TC6_REG_RESET, regval);
>> +     if (ret)
>> +             return ret;
>> +
>> +     /* Poll for soft reset complete for every 5us until 100us timeout */
> 
> Is this 100us defined in the standard? It is pretty short. If it is
> not part of the standard, maybe set it to something much bigger?
No, the standard doesn't define this and it is not part of the standard.
> 
> Also, polling every 5us is pretty quick. I doubt most systems can even
> sleep that short a time. So maybe 1ms between polls, and 1 second
> before -ETIMEDOUT?
Ok, I agree with your proposal and will change the poll delay as 1ms 
with the poll timeout of 1 second.

Best regards,
Parthiban V
> 
>> +     ret = readx_poll_timeout(oa_tc6_read_sw_reset_status, tc6, regval,
>> +                              regval & STATUS0_RESETC,
>> +                              STATUS0_RESETC_POLL_DELAY,
>> +                              STATUS0_RESETC_POLL_TIMEOUT);
>> +     if (ret)
>> +             return -ENODEV;
>> +
>> +     /* Clear the reset complete status */
>> +     return oa_tc6_write_register(tc6, OA_TC6_REG_STATUS0, regval);
> 
>          Andrew
>