On Fri, Jul 25, 2014 at 2:23 AM, Stephen Boyd <sboyd@xxxxxxxxxxxxxx> wrote: > On 07/24/14 08:40, Linus Walleij wrote: >> On Thu, Jul 24, 2014 at 1:47 AM, Stephen Boyd <sboyd@xxxxxxxxxxxxxx> wrote: >> >>>> Please add these constants to the table of valid power-source values and use >>>> something like I did to translate them to register values - it makes the DT >>>> much more readable. >>> The DT could be similarly readable if we had a bunch of #defines for the >>> different VIN settings that resolved to the final register value for >>> that pmic. Something like PM8921_GPIO1_14_VPH, PM8921_GPIO19_36_VPH, >>> etc. There would be a lot of them, but then the driver could be really >>> simple and just jam whatever value is in the DT into the register >>> without having to bounce through a mapping table in software to figure >>> out the register value. If we did this for the functions also then I >>> believe we achieve readability without requiring a bunch of drivers for >>> each and every single pmic? >> Not sure but it sounds like you want to make the device tree a jam table, >> (know about individual register offsets, sequences etc). That has been >> throrougly NACKed in the past, because DT is not Open Firmware. >> >> The exception is pinctrl-single which is restricted to single register >> per pin use cases and is still a point of contention... >> > > I'm not proposing a jam table. I'm proposing that we make the > function/source property convenient to the driver by having the actual > function field value encoded there instead of some string that has to be > translated through a table in a driver. There's still going to be > shifting and masking of bits in the driver to put the value in the right > place in the register, but we avoid needing N number of drivers for each > pmic just to translate strings into integers (for functions) and > integers into other integers (for the power source). From what I can > tell there isn't any benefit to having the function property be a string > vs. a #define number besides having a human readable string in pinctrl > debugfs. Is there some other benefit? OK I get it ... I think. One good reason to use strings is that it apart from debugfs makes for quite readable debug messages if used the right way. I feel sort of lukewarm on the issue, so I'd let the driver author decide the most elegant way to deal with this from an end-user point of view. If it helps people configure and debug their board set-ups is a crucial factor to me, and I suspect both Ivan and Björn has some experience with this. Yours, Linus Walleij -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html