This patch set adds a DT binding documentation for new 'clock-parents' and 'clock-rates' DT properties and a helper to parse DT and set the specified clocks configuration. The helper is now called in the platform bus before driver probing. Changes since v4: - added note explaining how to skip setting parent and rate of a clock, - added missing call to of_node_put(), - moved of_clk_dev_init() calls to the platform bus, - dropped debug traces. Thanks for the reviews. Currently clock parent/rate setting is only handled automatically for devices on Linux platform bus; I didn't add it to other busses this time as the original problem this patch series is trying to solve is related to the platform devices. However, in general it might make sense to also add it to other busses, like: amba, I2C, SPI, etc. The other issue is that in case of deferred probe of_clk_dev_init() could be called multiple times. To avoid this a field could be added to struct device or struct platform_device, etc. so a device can be marked as initialized WRT clocks and initialization could be skipped before any subsequent deferred probe calls. I wasn't sure if I should bother generic device data structures with something like this so I just didn't add it now. Changes since v3: - improved documentation of the DT binding, - fixed build errors for !CONFIG_OF, the parsing helpers are only compiled in if CONFIG_OF is set. Changes since v2: - code reordering to ensure there is no build errors, the clock configuration code moved to a separate file, - introduced an 'assigned-clocks' DT node which is supposed to contain clocks, clock-parents, clock-rates properties and be child node a clock provider node, and a code parsing it called from of_clk_init(); It's for clocks which are not directly connected to consumer devices. An alternative would be to list such assigned clocks in 'clocks' property, along with "proper" parent clocks, but then there would be phandles in clocks property of a node pointing to itself and it would require proper handling in of_clock_init(). I actually tried it but it looked a bit ugly and chose this time to use an extra subnode. Changes since v1: - updated DT binding documentation, - dropped the platform bus notifier, the clock setup routine is now being called directly from the driver core before a driver probe() call; this has an advantage such as all bus types are handled and any errors are propagated, so that, for instance a driver probe() can be deferred also when resources specified by clock-parents/clock-rates properties are not yet available; an alternative would be to let drivers call of_clk_device_setup() directly, - dropped the patch adding a macro definition for maximum DT property name length for now. Open issues: - handling of errors from of_clk_get_by_property() could be improved, currently ENOENT is returned by this function not only for a null entry. This series has been tested on ARM, on Exynos4412 Trats2 board. Sylwester Nawrocki (2): clk: Add function parsing arbitrary clock list DT property clk: Add handling of clk parent and rate assigned from DT .../devicetree/bindings/clock/clock-bindings.txt | 44 ++++++++++ drivers/base/platform.c | 5 ++ drivers/clk/Makefile | 3 + drivers/clk/clk-conf.c | 85 ++++++++++++++++++++ drivers/clk/clk.c | 12 ++- drivers/clk/clkdev.c | 25 +++++- include/linux/clk.h | 7 ++ include/linux/clk/clk-conf.h | 19 +++++ 8 files changed, 195 insertions(+), 5 deletions(-) create mode 100644 drivers/clk/clk-conf.c create mode 100644 include/linux/clk/clk-conf.h -- 1.7.9.5 -- 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
