On 16/10/2018 16:44, Rob Herring wrote: > On Mon, Oct 15, 2018 at 1:01 PM Daniel Lezcano > <daniel.lezcano@xxxxxxxxxx> wrote: >> >> >> Hi Rob, >> >> thanks for the review. >> >> On 15/10/2018 18:28, Rob Herring wrote: >>> On Tue, Sep 25, 2018 at 11:03:06AM +0200, Daniel Lezcano wrote: >>>> Add the interrupt names for the sensors, so the code can rely on them >>>> instead of dealing with index which are prone to error. > > [...] > >>>> + interrupt-names = "tsensor_intr"; >>> >>> That name seems pretty pointless. >>> >>>> clocks = <&sys_ctrl HI6220_TSENSOR_CLK>; >>>> clock-names = "thermal_clk"; >>>> #thermal-sensor-cells = <1>; >>>> @@ -28,5 +30,6 @@ for Hi3660: >>>> compatible = "hisilicon,hi3660-tsensor"; >>>> reg = <0x0 0xfff30000 0x0 0x1000>; >>>> interrupts = <GIC_SPI 145 IRQ_TYPE_LEVEL_HIGH>; >>>> + interrupt-names = "tsensor_a73"; >>> >>> Just 'a73' is sufficient. >> >> This is the name defined in the board documentation to give a name to >> the interrupt when requesting it. This one appears in /proc/interrupts. >> >> I can replace the 'tsensor_intr' by 'tsensor', but if 'tsensor_a73' is >> replaced by 'a73' that may looks odd in /proc/interrupts to see an >> interrupts line with the 'a73' name. > > Sounds like a Linux problem, not DT... > > The thing is that any *-names property is supposed to be a local to > block name. For example, a UART may have a TX irq. It's local name is > just "TX IRQ", but then upstream at the interrupt controller it may be > "UART1 TX IRQ". If you have multiple instances, the local names are > the same. > > If we lose the node name from /proc/interrupts when interrupt-names is > used, then we should fix that. Ah, ok. It makes more sense for me now. Thanks for the explanation, . -- <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook | <http://twitter.com/#!/linaroorg> Twitter | <http://www.linaro.org/linaro-blog/> Blog
