On 15/04/16 18:49, Laxman Dewangan wrote: > > On Friday 15 April 2016 11:14 PM, Jon Hunter wrote: >> On 15/04/16 17:41, Laxman Dewangan wrote: >>> On Friday 15 April 2016 09:15 PM, Jon Hunter wrote: >>>> On 15/04/16 16:14, Laxman Dewangan wrote: >>>>> I used pins as this is the property from pincon generic so that I can >>>>> use the generic implementation. >>>>> >>>>> Here, I will not go to the pin level control as HW does not support >>>>> pin >>>>> level control. >>>>> >>>>> I will say the unit should be interface level. Should we say >>>>> IO_GROUP_CSIA, IO_GROUP_CSIB etc? >>>> So we need to reflect the hardware in device-tree and although yes the >>>> power-down for the CSI_x_xxx pads are all controlled together as a >>>> single group, it does not feel right that we add a pseudo pin called >>>> csix to represent these. >>>> >>>> The CSI_x_xxx pads are already in device-tree and so why not add a >>>> property to each of these pads which has the IO rail information for >>>> power-down and voltage-select? >>> Which dt binding docs have these? >>> I looked for nvidia,tegra210-pinmux.txt and not able to find csi_xxx. >> For CSI you are right they are not included by the current DT binding >> docs, however, the sdmmc1/3 pads are. So that makes things a bit more >> messy as some are and some are not. > > Yaah and so lets have the names in new dt files. Names may be same but > define all possible names f groups in dt binding and need not to refer > from other file which does not have all. I still do not like that. In the case of sdmmc we now have two pinctrl drivers to deal with for a single set of pins. That does not seem correct IMO. >>> Here I dont want to refer the individual pins as control should be as >>> group. >> I understand, however, at least for power-down control I don't see why >> we cannot refer to the individual pins and once all are inactive then >> the rail can be powered down. >> >> For switching the voltage it is a bit more complex, but may be we could >> still look-up the IO rail based upon the pads the device uses. >> > > Yes, it can be done with ref count also for power down. Exactly. > But interfaces are complex. As a client, it is easy to say power down > SDMMC1 IO interface rather than saying power down 10 pins (names) of > that group. Right and like I said, we could always look up the IO rail from the pins associated once at probe time and then control it from there. Cheers Jon -- 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