On 07/14/2016 05:24 AM, Michal Simek wrote:
Simple version of clk_get_by_index() added by: "dm: clk: Add a simple version of clk_get_by_index()" (sha1: a4b10c088c4f6ef2e2bba33e8cfea369bcbbce44) is not sufficient if you use multiple clocks in the system because clk->id is phandle id which for example fixed-clock is not able to handle. Use the same implementation as is used in full version.
It took me a while to work out what failure case you were describing. It might be worth more explicitly pointing out that the existing simple implementation fails in any case where #clock-cells=<0>, or for larger #clock-cells, where the clock ID isn't in the first cell.
To be honest, I'd be inclined to always include the real clk_get_by_name() in SPL builds too. If it's never called, the function will be dropped by the linker. If it is called, the dummy implementation probably actively causes failures that we should avoid by using the real implementation. I'm not sure why the original SPL-specific code existed, unless the 773 byte code increase you mention is actually problematic for some specific build?
Signed-off-by: Michal Simek michal.simek@xilinx.com
Diff for ZynqMP is 773 B based on buildman.