Hi, On 12/19/2018 10:48 PM, Felipe Balbi wrote: > Hi, > > Thinh Nguyen <thinh.nguyen@xxxxxxxxxxxx> writes: >>>> On 12/18/2018 8:39 AM, Rob Herring wrote: >>>>> On Fri, Dec 07, 2018 at 06:27:30PM -0800, Thinh Nguyen wrote: >>>>>> This patch introduces property "snps,refclk-period-ns" to inform the >>>>>> controller of the reference clock period. If the reference clock period >>>>>> is different from the default Core Consultant setting, then this >>>>>> property can be set to the reference clock period. >>>>>> >>>>>> This property does not control the reference clock rate. The controller >>>>>> uses this value to perform internal timing calculations that are based >>>>>> on the reference clock. >>>>>> >>>>>> Signed-off-by: Thinh Nguyen <thinhn@xxxxxxxxxxxx> >>>>>> --- >>>>>> Changes in v2: >>>>>> - Split from "usb: dwc3: Add reference clock properties" >>>>>> - Revise commit message and property description >>>>>> >>>>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ >>>>>> 1 file changed, 2 insertions(+) >>>>>> >>>>>> diff --git a/Documentation/devicetree/bindings/usb/dwc3.txt b/Documentation/devicetree/bindings/usb/dwc3.txt >>>>>> index 8e5265e9f658..b7e67edff9b2 100644 >>>>>> --- a/Documentation/devicetree/bindings/usb/dwc3.txt >>>>>> +++ b/Documentation/devicetree/bindings/usb/dwc3.txt >>>>>> @@ -99,6 +99,8 @@ Optional properties: >>>>>> this and tx-thr-num-pkt-prd to a valid, non-zero value >>>>>> 1-16 (DWC_usb31 programming guide section 1.2.3) to >>>>>> enable periodic ESS TX threshold. >>>>>> + - snps,refclk-period-ns: if set, this value informs the controller of the >>>>>> + reference clock period in nanoseconds. >>>>> Shouldn't you be able to retrieve the refclk frequency and then >>>>> calculate the period? >>>> The thing is we cannot determine the ref_clk frequency for some devices >>>> that don't specify their clocks. So I think we should have an option to >>>> inform the controller of the ref_clk period for those devices. >>> Specifying the clock should be mandatory (if you want/need this >>> feature). It just requires a fixed-clock node at a minimum. >> Depending on the design of the controller, the ref_clk frequency is not >> something that the OS can read/control. So we cannot make it mandatory >> for every device to have a clock node. > We can make it mandatory to everyone who wants to use the feature. It's > no different than making snps,refclk-period-ns mandatory to everyone who > wants to use the feature you're introducing. > But not every design has access to the clock with an actual address for the OS to read. Only the developers will know the frequency of the ref_clk, and they can inform the controller via this property. We cannot force the developers to change their design requirement simply to inform the controller of the ref_clk period. Thinh
