Re: [PATCH v2 09/12] i2c: riic: Add support for fast mode plus

[claudiu beznea]

Hi, Geert,

On 11.07.2024 10:53, Geert Uytterhoeven wrote:
> Hi Claudiu,
> 
> On Wed, Jul 10, 2024 at 4:20 PM claudiu beznea <claudiu.beznea@xxxxxxxxx> wrote:
>> On 28.06.2024 12:22, Geert Uytterhoeven wrote:
>>> On Tue, Jun 25, 2024 at 2:14 PM Claudiu <claudiu.beznea@xxxxxxxxx> wrote:
>>>> From: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
>>>>
>>>> Fast mode plus is available on most of the IP variants that RIIC driver
>>>> is working with. The exception is (according to HW manuals of the SoCs
>>>> where this IP is available) the Renesas RZ/A1H. For this, patch
>>>> introduces the struct riic_of_data::fast_mode_plus.
>>>>
>>>> Fast mode plus was tested on RZ/G3S, RZ/G2{L,UL,LC}, RZ/Five by
>>>> instantiating the RIIC frequency to 1MHz and issuing i2c reads on the
>>>> fast mode plus capable devices (and the i2c clock frequency was checked on
>>>> RZ/G3S).
>>>>
>>>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
>>>
>>> Thanks for your patch!
>>>
>>>> --- a/drivers/i2c/busses/i2c-riic.c
>>>> +++ b/drivers/i2c/busses/i2c-riic.c
>>>> @@ -407,6 +413,9 @@ static int riic_init_hw(struct riic_dev *riic)
>>>>         riic_writeb(riic, 0, RIIC_ICSER);
>>>>         riic_writeb(riic, ICMR3_ACKWP | ICMR3_RDRFS, RIIC_ICMR3);
>>>>
>>>> +       if (info->fast_mode_plus && t->bus_freq_hz == I2C_MAX_FAST_MODE_PLUS_FREQ)
>>>> +               riic_clear_set_bit(riic, 0, ICFER_FMPE, RIIC_ICFER);
>>>
>>> Unless FM+ is specified, RIIC_ICFER is never written to.
>>> Probably the register should always be initialized, also to make sure
>>> the FMPE bit is cleared when it was set by the boot loader, but FM+
>>> is not to be used.
>>
>> Instead of clearing only this bit, what do you think about using
>> reset_control_reset() instead of reset_control_deassert() in riic_i2c_probe()?
>>
>> HW manuals for all the devices listed in
>> Documentation/devicetree/bindings/i2c/renesas,riic.yaml specifies that
>> ICFER_FMPE register is initialized with a default value by reset. All the
>> other registers are initialized with default values at reset (according to
>> HW manuals). I've checked it on RZ/G3S and it behaves like this.
> 
> RZ/A1 and RZ/A2M do not have reset controller support yet, so calling
> reset_control_reset() is a no-op on these SoCs.
> 
> However, I overlooked that riic_init_hw() does an internal reset first
> by setting the ICCR1_IICRST bit in RIIC_ICCR1.
> Is that sufficient to reset the FMPE bit?

Yes, that also restores the content of RIIC_ICFER register to the default
value (including clearing the FMPE bit). Thanks for pointing it, I
overlooked it, too.

Thank you,
Claudiu Beznea

> 
> Gr{oetje,eeting}s,
> 
>                         Geert
>