On 21/10/2022 09:05, Tilki, Ibrahim wrote: >> On 21/10/2022 11:44:12+0000, Tilki, Ibrahim wrote: >>>>> + interrupt-names: >>>>> + description: | >>>>> + Name of the interrupt pin of the RTC used for IRQ. Not required for >>>>> + RTCs that only have single interrupt pin available. Some of the RTCs >>>>> + share interrupt pins with clock input/output pins. >>>>> + minItems: 1 >>>>> + items: >>>>> + - enum: [INTA, INTB] >>>>> + - enum: [INTA, INTB] >>>>> + >>>> >>>> I don't think this is right, what this is doing is essentially pinmuxing >>>> interrupts versus clocks. What happens if you want INTB but this goes >>>> directly to a PMIC instead of the SoC? >>>> It is not something you can express with your current bindings. >>>> >>> >>> Why would a user want INTB when it is not connected to SoC? >>> User can specify none, either one or both of the interrupt pins. >>> I don't see what the problem here is. >>> >> >> the interrupt pin may be connected to a PMIC that is able to start or >> wake up the platform. In that case, the user would not have any >> interrupt-names and your driver will fail to mux the interrupt on INTB. >> Please fix. > > Interrupt muxing depends on the clock configuration, not the interrupt-names property. > Devices don't support muxing the alarm interrupt independently. > > For example in the case of max31329, alarm interrupt is muxed into INTA by default. > Alarm interrupt is muxed into INTB pin if and only if clkin is enabled.\ Just to be sure: are you now describing hardware or Linux driver behavior? > > This means that if a user wants the alarm interrupt on INTB pin, they have to > provide a clock input through "clocks" property. If that is the case, they can provide > an interrupt for INTB pin, otherwise the alarm feature of the rtc gets disabled. > > Side note: Some devices have 2 Alarms but Alarm2 does not have a register for > matching "seconds" so only Alarm1 is used by the driver. Best regards, Krzysztof
