On 09/18/2012 03:32 PM, Tom Warren wrote: ...
As far as I'm aware, an FDT pinmux for Tegra (which is on my plate, but quite a bit behind T30) would be essentially the same deal - one large list of mux settings per build/board.
Initializing pinmux from DT can work either way, depending on how the DT author wrote the DT file.
The pinmux DT node can contain a pinmux configuration which is applied as soon as the pinmux driver loads. This configuration can contain as little as you want (even nothing) all the way through to containing the entire board's static pinmux configuration.
For portions of the pinmux settings which the pinmux driver's own DT node doesn't configure (if any, based on the above), the relevant individual driver DT node can configure the pinmux as required, and that configuration would be applied when the relevant driver loads and parses its DT node.
In practice, so far, all the kernel board files for Tegra almost exclusively use static muxing in the pinmux controller's own DT node. However, there are a couple small dynamic cases (e.g. Seaboard/Springbank's pinctrl-based I2C mux for example).