On Tue, 17 Dec 2013, Paul Walmsley wrote: > On Tue, 26 Nov 2013, Tero Kristo wrote: > > > Current clock wrappers assume simple and direct mapped hardware register > > access. Improve this support by adding functionality for registering > > platform specific clock I/O wrappers, which can be used to support > > various features needed like endianess conversions, indexed regmap support, > > etc. Default I/O wrapper provided also which uses the existing direct > > I/O mapped behavior. > > I think it makes more sense to define per-SoC register read and write > functions, Sorry, this should have read "per-IP block register read and write functions". > for the same reason that I think it makes more sense to define clock > data as part of the IP blocks that contain the clocks' registers. The > register read and write functions should be implemented by the drivers > for the IP blocks that the clocks are contained in. That way the > register read and write functions can take whatever steps are necessary > to ensure that the IP block registers are accessible before performing > the read/write; can implement non-MMIO register accesses if needed; and > can operate on register offsets, rather than absolute addresses. > > Something like the following draft implementation. Also needed would be a > way for clock providers that rely on the common clock providers (e.g., > divider, mux, etc.) to pass in the struct clk_ll_ops pointer. I guess the > simplest way to implement this would be to add another set of registration > functions - e.g., clk_register_divider_ipb() or clk_register_divider2() or > something similar. These functions would be similar to > clk_register_divider(), but would take a pointer to a struct clk_ll_ops > and pass that along to the CCF core. - Paul -- 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
