[PATCH v2 18/29] nvmem: Add Realtek DHC eFuse driver

[Andreas Färber]

Hi Srini,

Am 23.06.20 um 11:32 schrieb Srinivas Kandagatla:
> On 23/06/2020 03:50, Andreas Färber wrote:
>> Implement enough of a read-only nvmem driver to easily read efuse cells.
>>
>> Cc: Cheng-Yu Lee <cylee12 at realtek.com>
>> Signed-off-by: Andreas Färber <afaerber at suse.de>
>> ---
> 
> This patch itself looks okay to me, I will apply this once DT patches 
> are Reviewed/applied by DT maintainers!

Thanks - let's give the Realtek engineers some time to review, too:

* Driver naming - new [rtk-]dhc (Stanley) vs. in-tree rtd1195 elsewhere.

* My other driver was previously reading u32-sized registers directly, 
whereas here I switched to byte-sized reads based on other in-tree nvmem 
drivers. Downstream driver seems inconsistent wrt .word_size:
https://github.com/BPI-SINOVOIP/BPI-M4-bsp/blob/master/linux-rtk/drivers/nvmem/rtk-efuse.c#L191

* RTD1619 (RTD1319, too?) may need special handling and therefore its 
own DT compatible: There's a magic OTP_CTRL register write downstream 
that I am lacking documentation w/ names&explanations and use case for.
https://github.com/BPI-SINOVOIP/BPI-M4-bsp/blob/master/linux-rtk/drivers/nvmem/rtk-efuse.c#L216
That might obviously affect the binding, too, requiring oneOf - could be 
changed in a later step though.

I would take the .dts patches through my linux-realtek.git once the 
binding is approved.

* The downstream DTs have nvmem-cells and nvmem-cell-names properties in 
the efuse node directly, which I regarded as unnecessary from reading 
the consumer binding, placing those properties into the consuming node.

* Downstream DTs have more eFuse fields declared than the one I use in 
this series [1]; they are also inconsistent in prefixing them efuse_ vs. 
otp_; in the RTD1295 datasheet the block is called eFuse, so I used 
efuse_ for consistency. I have enforced the dashes convention for nodes,
as I didn't see the node name get used anywhere.

[1] https://patchwork.kernel.org/patch/11619643/

One more comment inline...

>>   v2: New
>>   MAINTAINERS                   |  1 +
>>   drivers/nvmem/Kconfig         |  9 ++++
>>   drivers/nvmem/Makefile        |  2 +
>>   drivers/nvmem/rtk-dhc-efuse.c | 86 +++++++++++++++++++++++++++++++++++
>>   4 files changed, 98 insertions(+)
>>   create mode 100644 drivers/nvmem/rtk-dhc-efuse.c
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 1d0d6ab20451..02117fbf0e57 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -2312,6 +2312,7 @@ F:    
>> Documentation/devicetree/bindings/soc/realtek/
>>   F:    arch/arm/boot/dts/rtd*
>>   F:    arch/arm/mach-realtek/
>>   F:    arch/arm64/boot/dts/realtek/
>> +F:    drivers/nvmem/rtk-dhc-efuse.c
>>   F:    drivers/soc/realtek/
>>   ARM/RENESAS ARM64 ARCHITECTURE
[snip]

This line addition will conflict with the next line, added earlier in 
this patchset. Same for the binding patch. Do you need a v3 reordering 
them? This driver seems easier to target for 5.9 than the rest of the 
series.

If you do not intend to take the dt-bindings patch (17/29) through your 
tree, I can queue it once ack'ed by Rob and you.

Thanks for the quick review,

Andreas

-- 
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Felix Imendörffer
HRB 36809 (AG Nürnberg)