[PATCH v3 1/2] usb: dwc3: core: add support for remapping global register start address

[PATCH v3 1/2] usb: dwc3: core: add support for remapping global register start address

[Stanley Chang]

The RTK DHC SoCs were designed the global register address offset at
0x8100. The default address offset is constant at DWC3_GLOBALS_REGS_START
(0xc100). Therefore, add the property of device-tree to adjust this
address offset.

Signed-off-by: Stanley Chang <stanley_chang@realtek.com>
---
 v2 to v3 change:
1.  Fix the dtschema validation error.

 v1 to v2 change:
1. Change the name of the property "snps,global-regs-starting-offset".
2. Adjust the format of comment.
3. Add initial value of the global_regs_starting_offset
4. Remove the log of dev_info.
---
 drivers/usb/dwc3/core.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
index 476b63618511..8c1d1afbdc65 100644
--- a/drivers/usb/dwc3/core.c
+++ b/drivers/usb/dwc3/core.c
@@ -1785,6 +1785,21 @@ static int dwc3_probe(struct platform_device *pdev)
 	dwc_res = *res;
 	dwc_res.start += DWC3_GLOBALS_REGS_START;
 
+	/*
+	 * For some dwc3 controller, the dwc3 global register start address is
+	 * not at DWC3_GLOBALS_REGS_START (0xc100).
+	 */
+	if (dev->of_node) {
+		int global_regs_starting_offset = 0;
+
+		device_property_read_u32(dev, "snps,global-regs-starting-offset",
+			    &global_regs_starting_offset);
+		if (global_regs_starting_offset) {
+			dwc_res.start -= DWC3_GLOBALS_REGS_START;
+			dwc_res.start += global_regs_starting_offset;
+		}
+	}
+
 	regs = devm_ioremap_resource(dev, &dwc_res);
 	if (IS_ERR(regs))
 		return PTR_ERR(regs);
-- 
2.34.1

[PATCH v3 2/2] dt-bindings: usb: snps,dwc3: Add 'snps,global-regs-starting-offset' quirk

[Stanley Chang]

Add a new 'snps,global-regs-starting-offset' DT to dwc3 core to remap
the global register start address

The RTK DHC SoCs were designed the global register address offset at
0x8100. The default address offset is constant at DWC3_GLOBALS_REGS_START
(0xc100). Therefore, add the property of device-tree to adjust this
address offset.

Signed-off-by: Stanley Chang <stanley_chang@realtek.com>
---
 v2 to v3 change:
1.  Fix the dtschema validation error.

 v1 to v2 change:
1. Change the name of the property "snps,global-regs-starting-offset".
2. Adjust the format of comment.
3. Add initial value of the global_regs_starting_offset
4. Remove the log of dev_info.
---
 Documentation/devicetree/bindings/usb/snps,dwc3.yaml | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
index be36956af53b..4f83fa8cb6cb 100644
--- a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
+++ b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
@@ -359,6 +359,14 @@ properties:
     items:
       enum: [1, 4, 8, 16, 32, 64, 128, 256]
 
+  snps,global-regs-starting-offset:
+    description:
+      value for remapping global register start address. For some dwc3
+      controller, the dwc3 global register start address is not at
+      default DWC3_GLOBALS_REGS_START (0xc100). This property is added to
+      adjust the address.
+    $ref: '/schemas/types.yaml#/definitions/uint32'
+
   port:
     $ref: /schemas/graph.yaml#/properties/port
     description:
-- 
2.34.1

Re: [PATCH v3 2/2] dt-bindings: usb: snps,dwc3: Add 'snps,global-regs-starting-offset' quirk

[Rob Herring]

On Fri, Apr 21, 2023 at 1:18 AM Stanley Chang <stanley_chang@realtek.com> wrote:
>
> Add a new 'snps,global-regs-starting-offset' DT to dwc3 core to remap
> the global register start address
>
> The RTK DHC SoCs were designed the global register address offset at
> 0x8100. The default address offset is constant at DWC3_GLOBALS_REGS_START
> (0xc100). Therefore, add the property of device-tree to adjust this
> address offset.
>
> Signed-off-by: Stanley Chang <stanley_chang@realtek.com>
> ---
>  v2 to v3 change:
> 1.  Fix the dtschema validation error.
>
>  v1 to v2 change:
> 1. Change the name of the property "snps,global-regs-starting-offset".
> 2. Adjust the format of comment.
> 3. Add initial value of the global_regs_starting_offset
> 4. Remove the log of dev_info.
> ---
>  Documentation/devicetree/bindings/usb/snps,dwc3.yaml | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
> index be36956af53b..4f83fa8cb6cb 100644
> --- a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
> +++ b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
> @@ -359,6 +359,14 @@ properties:
>      items:
>        enum: [1, 4, 8, 16, 32, 64, 128, 256]
>
> +  snps,global-regs-starting-offset:
> +    description:
> +      value for remapping global register start address. For some dwc3
> +      controller, the dwc3 global register start address is not at
> +      default DWC3_GLOBALS_REGS_START (0xc100). This property is added to
> +      adjust the address.
> +    $ref: '/schemas/types.yaml#/definitions/uint32'

Again, we're not going to keep adding properties for every DWC3
variation. If it is board specific, then yes a property is
appropriate. If it is SoC specific, then imply it from the compatible.
Or in this case, you could possibly add another reg entry.

Rob

RE: [PATCH v3 2/2] dt-bindings: usb: snps,dwc3: Add 'snps,global-regs-starting-offset' quirk

[Stanley Chang[昌育德]]

Hi Rob,

> Again, we're not going to keep adding properties for every DWC3 variation. If
> it is board specific, then yes a property is appropriate. If it is SoC specific, then
> imply it from the compatible.
> Or in this case, you could possibly add another reg entry.
> 
> Rob
> 
Let me try to understand your concerns.

The device-tree property should work for all dwc3 IPs and can be specified by different boards.
For a SoC specific, it should use a compatible or registry entry to change it.
So you think we can't use a property to specify this offset.

Is my understanding correct?

If it is right, using property to solve this problem was my mistake.
I will think a new method to resolve it.