| Message ID | 20230516184626.154892-1-afd@ti.com |
|---|---|
| State | Accepted |
| Commit | 595409cf07ade54d6265942f25a3d33b0d76a17c |
| Headers | show |
| Series | [1/2] clk: keystone: syscon-clk: Allow the clock node to not be of type syscon | expand |
On 5/16/23 11:36 PM, Vignesh Raghavendra wrote: > > > On 17/05/23 00:16, Andrew Davis wrote: >> This node's register space is not accessed by any other node, which >> is the traditional use for the "syscon" hint. > > Unfortunately that's not the case across SoCs. Eg AM65x See TRM section > Table 5-582. CTRLMMR_EPWM0_CTRL Register Field Descriptions > Not sure what version of the TRM you have, latest (Rev. E) has this register as Table 5-636.. but I found it and see your point here. > TB_CLKEN is clubbed with SYNCIN_SEL and ePWM tripzone configuration > signals which may require register to be shared with other drivers in future > This looks to only be a problem in AM65x, all later devices we have fixed the issue and now group the clock enable bits all together. Do we actually expect this to be an issue and have a user of these other bits? If so then we modeled this region wrong in AM65x DT, these registers are not "tbclk gate registers" any more then they are to the other functions they provide. These registers should be a syscon node and then each function within should be a child node. syscon@4140 { compatible = "ti,am654-epwm-crtl", "syscon"; reg = <0x4140 0x18>; ehrpwm_tbclk: clock { compatible = "ti,am654-ehrpwm-tbclk"; #clock-cells = <1>; }; pwm_mux: mux-controller { compatible = "mmio-mux"; #mux-control-cells = <1>; }; }; Something like that. That way we do not give preference to one device and have to have it give out shared registers. Either that or split the binding compatible, one for AM65x with syscon and one for all later device compatibles that do not share the register: compatible: oneOf: - items: - const: ti,am654-ehrpwm-tbclk - const: syscon - items: - enum: - ti,am64-epwm-tbclk - ti,am62-epwm-tbclk Would rather the first option. Andrew > >> It looks to have been >> added here to make use of a Linux kernel helper syscon_node_to_regmap(). >> The Linux driver now uses a more appropriate helper that does not >> require the hint, so let's remove it from the binding. >> >> Signed-off-by: Andrew Davis <afd@ti.com> >> --- >> .../devicetree/bindings/clock/ti,am654-ehrpwm-tbclk.yaml | 5 ++--- >> 1 file changed, 2 insertions(+), 3 deletions(-) >> >> diff --git a/Documentation/devicetree/bindings/clock/ti,am654-ehrpwm-tbclk.yaml b/Documentation/devicetree/bindings/clock/ti,am654-ehrpwm-tbclk.yaml >> index 66765116aff5..64b8bce5962c 100644 >> --- a/Documentation/devicetree/bindings/clock/ti,am654-ehrpwm-tbclk.yaml >> +++ b/Documentation/devicetree/bindings/clock/ti,am654-ehrpwm-tbclk.yaml >> @@ -16,7 +16,6 @@ properties: >> - ti,am654-ehrpwm-tbclk >> - ti,am64-epwm-tbclk >> - ti,am62-epwm-tbclk >> - - const: syscon >> >> "#clock-cells": >> const: 1 >> @@ -33,8 +32,8 @@ additionalProperties: false >> >> examples: >> - | >> - ehrpwm_tbclk: syscon@4140 { >> - compatible = "ti,am654-ehrpwm-tbclk", "syscon"; >> + ehrpwm_tbclk: clock@4140 { >> + compatible = "ti,am654-ehrpwm-tbclk"; >> reg = <0x4140 0x18>; >> #clock-cells = <1>; >> }; >
Quoting Andrew Davis (2023-05-16 11:46:25) > There is a helper device_node_to_regmap() we can use that does not force > this clock DT node to be a "syscon" node. It should work the same in > this case but allow us to remove the unneeded "syscon" compatible. > > Signed-off-by: Andrew Davis <afd@ti.com> > --- Applied to clk-next
diff --git a/drivers/clk/keystone/syscon-clk.c b/drivers/clk/keystone/syscon-clk.c index 5d7cc83682da..bd5cec0bd12d 100644 --- a/drivers/clk/keystone/syscon-clk.c +++ b/drivers/clk/keystone/syscon-clk.c @@ -101,10 +101,10 @@ static int ti_syscon_gate_clk_probe(struct platform_device *pdev) if (!data) return -EINVAL; - regmap = syscon_node_to_regmap(dev->of_node); + regmap = device_node_to_regmap(dev->of_node); if (IS_ERR(regmap)) return dev_err_probe(dev, PTR_ERR(regmap), - "failed to find parent regmap\n"); + "failed to get regmap\n"); num_clks = 0; for (p = data; p->name; p++)
There is a helper device_node_to_regmap() we can use that does not force this clock DT node to be a "syscon" node. It should work the same in this case but allow us to remove the unneeded "syscon" compatible. Signed-off-by: Andrew Davis <afd@ti.com> --- drivers/clk/keystone/syscon-clk.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)