On 08/07/2024 16:52, Maxime Ripard wrote: > On Mon, Jul 08, 2024 at 04:04:21PM GMT, Krzysztof Kozlowski wrote: >> On 08/07/2024 08:40, Liu Ying wrote: >>>>> + >>>>> + "^framegen@[0-9a-f]+$": >>>>> + type: object >>>>> + additionalProperties: true >>>>> + >>>>> + properties: >>>>> + compatible: >>>>> + const: fsl,imx8qxp-dc-framegen >>>>> + >>>>> + "^gammacor@[0-9a-f]+$": >>>> >>>> This looks like you are organizing bindings per your driver architecture. >>> >>> As I mentioned in cover letter, this series addresses Maxime's >>> comment for the previous series - split the display controller >>> into multiple internal devices. Maxime insisted on doing this. >> >> But these are not separate devices. Look: >> 1. parent DC: >> reg = <0x56180000 0x40000>; >> >> 2. child interrupt controller: >> reg = <0x56180040 0x60>; >> >> That address is within parent. >> >> 3. Then we go to things like: >> reg = <0x5618b400 0x14>, <0x5618b800 0x1c00>; >> >> Still within parent's range and just few words in address range. That's >> a clear indication that you choose few registers and call it a "device". > > That's never really been a metric though? > > If not, one could just create a "soc" device node covering the entire > register map, and since it would overlap despite clearly defined > features, you would claim it's a single device? Since I do not create such one-address-soc devices, I claim I have separate devices in the SoC. Here is not the case: there is a device covering entire address space. Soc is a good example, because components/blocks of the SoC are being re-used among different SoCs. Is the case here? BTW, it could be that some of the sub-devices here are worth to be devices, I agree. Best regards, Krzysztof
