On 24/05/2019 16.27, Andrew Lunn wrote:
>> @@ -4841,6 +4910,10 @@ static const struct of_device_id mv88e6xxx_of_match[]
>> = {
>> .compatible = "marvell,mv88e6190",
>> .data = &mv88e6xxx_table[MV88E6190],
>> },
>> + {
>> + .compatible = "marvell,mv88e6250",
>> + .data = &mv88e6xxx_table[MV88E6250],
>> + },
>> { /* sentinel */ },
>> };
>
> Ah, yes. I had not thought about that. A device at address 0 would be
> found, but a device at address 16 would be missed.
Eh, no? The port registers are at offset 0x8, i.e. at either SMI address
8 or 24, so I don't think a 6250 at address 0 could be detected using
either of the existing families?
> Please add this compatible string to
> Documentation/devicetree/bindings/net/dsa/marvell.txt
Will do.
>> +++ b/drivers/net/dsa/mv88e6xxx/global1.c
>> @@ -182,6 +182,25 @@ int mv88e6185_g1_reset(struct mv88e6xxx_chip *chip)
>> return mv88e6185_g1_wait_ppu_polling(chip);
>> }
>>
>> +int mv88e6250_g1_reset(struct mv88e6xxx_chip *chip)
>> +{
>> + u16 val;
>> + int err;
>> +
>> + /* Set the SWReset bit 15 */
>> + err = mv88e6xxx_g1_read(chip, MV88E6XXX_G1_CTL1, &val);
>> + if (err)
>> + return err;
>> +
>> + val |= MV88E6XXX_G1_CTL1_SW_RESET;
>> +
>> + err = mv88e6xxx_g1_write(chip, MV88E6XXX_G1_CTL1, val);
>> + if (err)
>> + return err;
>> +
>> + return mv88e6xxx_g1_wait_init_ready(chip);
>> +}
>
> It looks like you could refactor mv88e6352_g1_reset() to call
> this function, and then mv88e6352_g1_wait_ppu_polling(chip);
Yes, I actually deliberately moved the 6250 reset function further up in
v2 to allow that. I'll add that refactoring as a separate patch.
Thanks,
Rasmus
