On Thu, Jan 26, 2012 at 04:28:00PM +0100, Karol Lewandowski wrote: > On 25.01.2012 14:32, Mark Brown wrote: > >Well, they're different things. Device tree isn't Linux specific at > >all. > There is no official platform-agnostic regulator API, nor DT-bindings > document I'm aware of. Thus, I don't see why, while transitioning to The binding for the regulator API is supposed to be the one true platform agnostic binding, none of the device tree stuff is supposed to be Linux specific. > DT, should we lose ability to describe certain hardware configurations. We don't loose anything, a single voltage constraint that didn't set apply_uV was always meaningless. I keep meaning to make the core complain about things like that and people specifying voltage ranges without SET_VOLTAGE. > On 25.01.2012 12:22, Mark Brown wrote: > > The big problem there seems like specifying voltages in the first > > place, if we know what device it is we should already know what's > > going on. > Driver which handles said regulator might know what's going on, but > that might not be case for its consumers. Should we limit ability to > query given parameter just because its value is hardcoded in hardware? I'm sorry, this makes no sense. Setting a value in the constraints is not going to have any impact on the value reported by the driver, it never has. > Consequently, if it's property of hardware that it provides fixed > voltage somewhere shouldn't it be possible to describe this fact > in DT? If the device has a fixed voltage output the driver should just know this without having to read the information from device tree, device tree is for configuration. If the device has some hardware fixed configurability it should define this in a sensible fashion in the bindings (which may for example be a case of specifying the values of the passive components for ease of use).
Attachment:
signature.asc
Description: Digital signature
