[PATCH 0/3] usb: dwc3-generic: improve DT parsing and support qcom flat DT

[PATCH 0/3] usb: dwc3-generic: improve DT parsing and support qcom flat DT

[Casey Connolly]

Add support for the new flat DT format being adopted by Qualcomm boards,
this merges the dwc3 and glue nodes together.

Correctly default dr_mode to OTG when it isn't set explicitly, and
support the usb-role-switch and role-switch-default-mode properties so
that boards can express a preference when OTG is supported.

---
Casey Connolly (3):
      usb: common: default dr_mode to OTG when not set
      usb: dwc3-generic: support Qualcomm flattened DT
      usb: dwc3-generic: respect role-switch-default-mode

 drivers/usb/common/common.c     |  3 ++-
 drivers/usb/dwc3/dwc3-generic.c | 26 ++++++++++++++++++++------
 2 files changed, 22 insertions(+), 7 deletions(-)
---
base-commit: 03c01c69806d4265070f14e7a7cbf617fcbb8f6a

// Casey (she/they)

[PATCH 1/3] usb: common: default dr_mode to OTG when not set

[Casey Connolly]

DT bindings dictate that dr_mode being unset on a DRD capable USB
controller means we should default to OTG mode. Adjust usb_get_dr_mode()
to reflect this.

Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
---
 drivers/usb/common/common.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/common/common.c b/drivers/usb/common/common.c
index 13e9a61072a9..e3bd7339e68e 100644
--- a/drivers/usb/common/common.c
+++ b/drivers/usb/common/common.c
@@ -29,9 +29,10 @@ enum usb_dr_mode usb_get_dr_mode(ofnode node)
 
 	dr_mode = ofnode_read_string(node, "dr_mode");
 	if (!dr_mode) {
 		pr_debug("usb dr_mode not found\n");
-		return USB_DR_MODE_UNKNOWN;
+		/* Default to OTG mode as per DT bindings */
+		return USB_DR_MODE_OTG;
 	}
 
 	for (i = 0; i < ARRAY_SIZE(usb_dr_modes); i++)
 		if (!strcmp(dr_mode, usb_dr_modes[i]))

-- 
2.51.0

Re: [PATCH 1/3] usb: common: default dr_mode to OTG when not set

[Neil Armstrong]

On 1/14/26 15:13, Casey Connolly wrote:
> DT bindings dictate that dr_mode being unset on a DRD capable USB
> controller means we should default to OTG mode. Adjust usb_get_dr_mode()
> to reflect this.
> 
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
> ---
>   drivers/usb/common/common.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/usb/common/common.c b/drivers/usb/common/common.c
> index 13e9a61072a9..e3bd7339e68e 100644
> --- a/drivers/usb/common/common.c
> +++ b/drivers/usb/common/common.c
> @@ -29,9 +29,10 @@ enum usb_dr_mode usb_get_dr_mode(ofnode node)
>   
>   	dr_mode = ofnode_read_string(node, "dr_mode");
>   	if (!dr_mode) {
>   		pr_debug("usb dr_mode not found\n");
> -		return USB_DR_MODE_UNKNOWN;
> +		/* Default to OTG mode as per DT bindings */
> +		return USB_DR_MODE_OTG;
>   	}
>   
>   	for (i = 0; i < ARRAY_SIZE(usb_dr_modes); i++)
>   		if (!strcmp(dr_mode, usb_dr_modes[i]))
> 

Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>

Thanks,
Neil

Re: [PATCH 1/3] usb: common: default dr_mode to OTG when not set

[Mattijs Korpershoek]

Hi Casey,

Thank you for the patch.

On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:

> DT bindings dictate that dr_mode being unset on a DRD capable USB

For the curious, this is defined in https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/usb/usb-drd.yaml#n23

> controller means we should default to OTG mode. Adjust usb_get_dr_mode()
> to reflect this.
>
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
> ---
>  drivers/usb/common/common.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/common/common.c b/drivers/usb/common/common.c
> index 13e9a61072a9..e3bd7339e68e 100644
> --- a/drivers/usb/common/common.c
> +++ b/drivers/usb/common/common.c
> @@ -29,9 +29,10 @@ enum usb_dr_mode usb_get_dr_mode(ofnode node)
>  
>  	dr_mode = ofnode_read_string(node, "dr_mode");
>  	if (!dr_mode) {
>  		pr_debug("usb dr_mode not found\n");
> -		return USB_DR_MODE_UNKNOWN;
> +		/* Default to OTG mode as per DT bindings */

Nitpick: can we change the comment to:
/* Default to OTG mode as per devicetree/bindings/usb/usb-drd.yaml */

This tells the same but is more precise and is imho, a more helpful
comment.

Reviewed-by: Mattijs Korpershoek <mkorpershoek@kernel.org>


> +		return USB_DR_MODE_OTG;
>  	}
>  
>  	for (i = 0; i < ARRAY_SIZE(usb_dr_modes); i++)
>  		if (!strcmp(dr_mode, usb_dr_modes[i]))
>
> -- 
> 2.51.0

Re: [PATCH 1/3] usb: common: default dr_mode to OTG when not set

[Mattijs Korpershoek]

On Thu, Jan 15, 2026 at 10:17, Mattijs Korpershoek <mkorpershoek@kernel.org> wrote:

> Hi Casey,
>
> Thank you for the patch.
>
> On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:
>
>> DT bindings dictate that dr_mode being unset on a DRD capable USB
>
> For the curious, this is defined in https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/usb/usb-drd.yaml#n23
>
>> controller means we should default to OTG mode. Adjust usb_get_dr_mode()
>> to reflect this.
>>
>> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
>> ---
>>  drivers/usb/common/common.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/usb/common/common.c b/drivers/usb/common/common.c
>> index 13e9a61072a9..e3bd7339e68e 100644
>> --- a/drivers/usb/common/common.c
>> +++ b/drivers/usb/common/common.c
>> @@ -29,9 +29,10 @@ enum usb_dr_mode usb_get_dr_mode(ofnode node)
>>  
>>  	dr_mode = ofnode_read_string(node, "dr_mode");
>>  	if (!dr_mode) {
>>  		pr_debug("usb dr_mode not found\n");
>> -		return USB_DR_MODE_UNKNOWN;
>> +		/* Default to OTG mode as per DT bindings */
>
> Nitpick: can we change the comment to:
> /* Default to OTG mode as per devicetree/bindings/usb/usb-drd.yaml */
>
> This tells the same but is more precise and is imho, a more helpful
> comment.
>
> Reviewed-by: Mattijs Korpershoek <mkorpershoek@kernel.org>

Also, I just noticed, but there is another similar change here:
https://lore.kernel.org/all/20260108-usb-dwc3-exynos7870-v3-1-343fa1e81e48@disroot.org/

Casey, can you see if the above is useful to you?

>
>
>> +		return USB_DR_MODE_OTG;
>>  	}
>>  
>>  	for (i = 0; i < ARRAY_SIZE(usb_dr_modes); i++)
>>  		if (!strcmp(dr_mode, usb_dr_modes[i]))
>>
>> -- 
>> 2.51.0

Re: [PATCH 1/3] usb: common: default dr_mode to OTG when not set

[Casey Connolly]

On 1/16/26 13:30, Mattijs Korpershoek wrote:
> On Thu, Jan 15, 2026 at 10:17, Mattijs Korpershoek <mkorpershoek@kernel.org> wrote:
> 
>> Hi Casey,
>>
>> Thank you for the patch.
>>
>> On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:
>>
>>> DT bindings dictate that dr_mode being unset on a DRD capable USB
>>
>> For the curious, this is defined in https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/usb/usb-drd.yaml#n23
>>
>>> controller means we should default to OTG mode. Adjust usb_get_dr_mode()
>>> to reflect this.
>>>
>>> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
>>> ---
>>>   drivers/usb/common/common.c | 3 ++-
>>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/usb/common/common.c b/drivers/usb/common/common.c
>>> index 13e9a61072a9..e3bd7339e68e 100644
>>> --- a/drivers/usb/common/common.c
>>> +++ b/drivers/usb/common/common.c
>>> @@ -29,9 +29,10 @@ enum usb_dr_mode usb_get_dr_mode(ofnode node)
>>>   
>>>   	dr_mode = ofnode_read_string(node, "dr_mode");
>>>   	if (!dr_mode) {
>>>   		pr_debug("usb dr_mode not found\n");
>>> -		return USB_DR_MODE_UNKNOWN;
>>> +		/* Default to OTG mode as per DT bindings */
>>
>> Nitpick: can we change the comment to:
>> /* Default to OTG mode as per devicetree/bindings/usb/usb-drd.yaml */
>>
>> This tells the same but is more precise and is imho, a more helpful
>> comment.
>>
>> Reviewed-by: Mattijs Korpershoek <mkorpershoek@kernel.org>
> 
> Also, I just noticed, but there is another similar change here:
> https://lore.kernel.org/all/20260108-usb-dwc3-exynos7870-v3-1-343fa1e81e48@disroot.org/
> 
> Casey, can you see if the above is useful to you?

Oh yep, same patch basically. I'll drop my version in the next revision.

> 
>>
>>
>>> +		return USB_DR_MODE_OTG;
>>>   	}
>>>   
>>>   	for (i = 0; i < ARRAY_SIZE(usb_dr_modes); i++)
>>>   		if (!strcmp(dr_mode, usb_dr_modes[i]))
>>>
>>> -- 
>>> 2.51.0

[PATCH 2/3] usb: dwc3-generic: support Qualcomm flattened DT

[Casey Connolly]

Qualcomm devicetrees are moving away from having a glue node with dwc3
as a subnode and now may just have a single flattened node.

Rockchip already have a glue_get_ctrl_dev op which returns the node for
the glue device itself, commonise this and reuse it for the new Qualcomm
node.

Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
---
 drivers/usb/dwc3/dwc3-generic.c | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
index 02c02b1d80e4..7044696ced0c 100644
--- a/drivers/usb/dwc3/dwc3-generic.c
+++ b/drivers/usb/dwc3/dwc3-generic.c
@@ -6,8 +6,10 @@
  *
  * Based on dwc3-omap.c.
  */
 
+#define LOG_DEBUG
+
 #include <dm.h>
 #include <reset.h>
 #include <asm/gpio.h>
 #include <dm/device_compat.h>
@@ -488,23 +490,29 @@ static void dwc3_qcom_glue_configure(struct udevice *dev, int index,
 	if (mode != USB_DR_MODE_HOST)
 		dwc3_qcom_vbus_override_enable(qscratch_base, true);
 }
 
-struct dwc3_glue_ops qcom_ops = {
-	.glue_configure = dwc3_qcom_glue_configure,
-};
-
-static int dwc3_rk_glue_get_ctrl_dev(struct udevice *dev, ofnode *node)
+/* In cases where there is no dwc3 node and it's flattened into the glue node */
+static int dwc3_flat_dt_get_ctrl_dev(struct udevice *dev, ofnode *node)
 {
 	*node = dev_ofnode(dev);
 	if (!ofnode_valid(*node))
 		return -EINVAL;
 
 	return 0;
 }
 
+struct dwc3_glue_ops qcom_ops = {
+	.glue_configure = dwc3_qcom_glue_configure,
+};
+
+struct dwc3_glue_ops qcom_flat_dt_ops = {
+	.glue_configure = dwc3_qcom_glue_configure,
+	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
+};
+
 struct dwc3_glue_ops rk_ops = {
-	.glue_get_ctrl_dev = dwc3_rk_glue_get_ctrl_dev,
+	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
 };
 
 static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
 {
@@ -711,8 +719,9 @@ static const struct udevice_id dwc3_glue_ids[] = {
 	{ .compatible = "rockchip,rk3568-dwc3", .data = (ulong)&rk_ops },
 	{ .compatible = "rockchip,rk3576-dwc3", .data = (ulong)&rk_ops },
 	{ .compatible = "rockchip,rk3588-dwc3", .data = (ulong)&rk_ops },
 	{ .compatible = "qcom,dwc3", .data = (ulong)&qcom_ops },
+	{ .compatible = "qcom,snps-dwc3", .data = (ulong)&qcom_flat_dt_ops },
 	{ .compatible = "fsl,imx8mp-dwc3", .data = (ulong)&imx8mp_ops },
 	{ .compatible = "fsl,imx8mq-dwc3" },
 	{ .compatible = "intel,tangier-dwc3" },
 	{ .compatible = "samsung,exynos7870-dwusb3" },

-- 
2.51.0

Re: [PATCH 2/3] usb: dwc3-generic: support Qualcomm flattened DT

[Neil Armstrong]

On 1/14/26 15:13, Casey Connolly wrote:
> Qualcomm devicetrees are moving away from having a glue node with dwc3
> as a subnode and now may just have a single flattened node.
> 
> Rockchip already have a glue_get_ctrl_dev op which returns the node for
> the glue device itself, commonise this and reuse it for the new Qualcomm
> node.
> 
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
> ---
>   drivers/usb/dwc3/dwc3-generic.c | 21 +++++++++++++++------
>   1 file changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
> index 02c02b1d80e4..7044696ced0c 100644
> --- a/drivers/usb/dwc3/dwc3-generic.c
> +++ b/drivers/usb/dwc3/dwc3-generic.c
> @@ -6,8 +6,10 @@
>    *
>    * Based on dwc3-omap.c.
>    */
>   
> +#define LOG_DEBUG

You should drop this

> +
>   #include <dm.h>
>   #include <reset.h>
>   #include <asm/gpio.h>
>   #include <dm/device_compat.h>
> @@ -488,23 +490,29 @@ static void dwc3_qcom_glue_configure(struct udevice *dev, int index,
>   	if (mode != USB_DR_MODE_HOST)
>   		dwc3_qcom_vbus_override_enable(qscratch_base, true);
>   }
>   
> -struct dwc3_glue_ops qcom_ops = {
> -	.glue_configure = dwc3_qcom_glue_configure,
> -};
> -
> -static int dwc3_rk_glue_get_ctrl_dev(struct udevice *dev, ofnode *node)
> +/* In cases where there is no dwc3 node and it's flattened into the glue node */
> +static int dwc3_flat_dt_get_ctrl_dev(struct udevice *dev, ofnode *node)
>   {
>   	*node = dev_ofnode(dev);
>   	if (!ofnode_valid(*node))
>   		return -EINVAL;
>   
>   	return 0;
>   }
>   
> +struct dwc3_glue_ops qcom_ops = {
> +	.glue_configure = dwc3_qcom_glue_configure,
> +};
> +
> +struct dwc3_glue_ops qcom_flat_dt_ops = {
> +	.glue_configure = dwc3_qcom_glue_configure,
> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
> +};
> +
>   struct dwc3_glue_ops rk_ops = {
> -	.glue_get_ctrl_dev = dwc3_rk_glue_get_ctrl_dev,
> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
>   };
>   
>   static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
>   {
> @@ -711,8 +719,9 @@ static const struct udevice_id dwc3_glue_ids[] = {
>   	{ .compatible = "rockchip,rk3568-dwc3", .data = (ulong)&rk_ops },
>   	{ .compatible = "rockchip,rk3576-dwc3", .data = (ulong)&rk_ops },
>   	{ .compatible = "rockchip,rk3588-dwc3", .data = (ulong)&rk_ops },
>   	{ .compatible = "qcom,dwc3", .data = (ulong)&qcom_ops },
> +	{ .compatible = "qcom,snps-dwc3", .data = (ulong)&qcom_flat_dt_ops },
>   	{ .compatible = "fsl,imx8mp-dwc3", .data = (ulong)&imx8mp_ops },
>   	{ .compatible = "fsl,imx8mq-dwc3" },
>   	{ .compatible = "intel,tangier-dwc3" },
>   	{ .compatible = "samsung,exynos7870-dwusb3" },
> 

With that fixed:
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>

Re: [PATCH 2/3] usb: dwc3-generic: support Qualcomm flattened DT

[Mattijs Korpershoek]

Hi Casey,

Thank you for the patch.

On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:

> Qualcomm devicetrees are moving away from having a glue node with dwc3
> as a subnode and now may just have a single flattened node.

In the commit message, could you provide a link to an example dt who has
this single flattened node?

>
> Rockchip already have a glue_get_ctrl_dev op which returns the node for
> the glue device itself, commonise this and reuse it for the new Qualcomm
> node.
>
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
> ---
>  drivers/usb/dwc3/dwc3-generic.c | 21 +++++++++++++++------
>  1 file changed, 15 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
> index 02c02b1d80e4..7044696ced0c 100644
> --- a/drivers/usb/dwc3/dwc3-generic.c
> +++ b/drivers/usb/dwc3/dwc3-generic.c
> @@ -6,8 +6,10 @@
>   *
>   * Based on dwc3-omap.c.
>   */
>  
> +#define LOG_DEBUG
> +

Debug leftover? Please drop for v2.

>  #include <dm.h>
>  #include <reset.h>
>  #include <asm/gpio.h>
>  #include <dm/device_compat.h>
> @@ -488,23 +490,29 @@ static void dwc3_qcom_glue_configure(struct udevice *dev, int index,
>  	if (mode != USB_DR_MODE_HOST)
>  		dwc3_qcom_vbus_override_enable(qscratch_base, true);
>  }
>  
> -struct dwc3_glue_ops qcom_ops = {
> -	.glue_configure = dwc3_qcom_glue_configure,
> -};
> -
> -static int dwc3_rk_glue_get_ctrl_dev(struct udevice *dev, ofnode *node)
> +/* In cases where there is no dwc3 node and it's flattened into the glue node */
> +static int dwc3_flat_dt_get_ctrl_dev(struct udevice *dev, ofnode *node)
>  {
>  	*node = dev_ofnode(dev);
>  	if (!ofnode_valid(*node))
>  		return -EINVAL;
>  
>  	return 0;
>  }
>  
> +struct dwc3_glue_ops qcom_ops = {
> +	.glue_configure = dwc3_qcom_glue_configure,
> +};
> +
> +struct dwc3_glue_ops qcom_flat_dt_ops = {
> +	.glue_configure = dwc3_qcom_glue_configure,
> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
> +};
> +
>  struct dwc3_glue_ops rk_ops = {
> -	.glue_get_ctrl_dev = dwc3_rk_glue_get_ctrl_dev,
> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
>  };
>  
>  static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
>  {
> @@ -711,8 +719,9 @@ static const struct udevice_id dwc3_glue_ids[] = {
>  	{ .compatible = "rockchip,rk3568-dwc3", .data = (ulong)&rk_ops },
>  	{ .compatible = "rockchip,rk3576-dwc3", .data = (ulong)&rk_ops },
>  	{ .compatible = "rockchip,rk3588-dwc3", .data = (ulong)&rk_ops },
>  	{ .compatible = "qcom,dwc3", .data = (ulong)&qcom_ops },
> +	{ .compatible = "qcom,snps-dwc3", .data = (ulong)&qcom_flat_dt_ops },
>  	{ .compatible = "fsl,imx8mp-dwc3", .data = (ulong)&imx8mp_ops },
>  	{ .compatible = "fsl,imx8mq-dwc3" },
>  	{ .compatible = "intel,tangier-dwc3" },
>  	{ .compatible = "samsung,exynos7870-dwusb3" },
>
> -- 
> 2.51.0

Re: [PATCH 2/3] usb: dwc3-generic: support Qualcomm flattened DT

[Casey Connolly]

Hi Mattijs,

On 1/15/26 10:21, Mattijs Korpershoek wrote:
> Hi Casey,
> 
> Thank you for the patch.
> 
> On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:
> 
>> Qualcomm devicetrees are moving away from having a glue node with dwc3
>> as a subnode and now may just have a single flattened node.
> 
> In the commit message, could you provide a link to an example dt who has
> this single flattened node?

Sorry I just saw this after posting v2... You can grep for the new 
compatible string in upstream DT, sc7280, sm8650 and a few other 
platforms are already switched over, eventually all the used/maintained 
platforms will get switched over I suspect.

> 
>>
>> Rockchip already have a glue_get_ctrl_dev op which returns the node for
>> the glue device itself, commonise this and reuse it for the new Qualcomm
>> node.
>>
>> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
>> ---
>>   drivers/usb/dwc3/dwc3-generic.c | 21 +++++++++++++++------
>>   1 file changed, 15 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
>> index 02c02b1d80e4..7044696ced0c 100644
>> --- a/drivers/usb/dwc3/dwc3-generic.c
>> +++ b/drivers/usb/dwc3/dwc3-generic.c
>> @@ -6,8 +6,10 @@
>>    *
>>    * Based on dwc3-omap.c.
>>    */
>>   
>> +#define LOG_DEBUG
>> +
> 
> Debug leftover? Please drop for v2.
> 
>>   #include <dm.h>
>>   #include <reset.h>
>>   #include <asm/gpio.h>
>>   #include <dm/device_compat.h>
>> @@ -488,23 +490,29 @@ static void dwc3_qcom_glue_configure(struct udevice *dev, int index,
>>   	if (mode != USB_DR_MODE_HOST)
>>   		dwc3_qcom_vbus_override_enable(qscratch_base, true);
>>   }
>>   
>> -struct dwc3_glue_ops qcom_ops = {
>> -	.glue_configure = dwc3_qcom_glue_configure,
>> -};
>> -
>> -static int dwc3_rk_glue_get_ctrl_dev(struct udevice *dev, ofnode *node)
>> +/* In cases where there is no dwc3 node and it's flattened into the glue node */
>> +static int dwc3_flat_dt_get_ctrl_dev(struct udevice *dev, ofnode *node)
>>   {
>>   	*node = dev_ofnode(dev);
>>   	if (!ofnode_valid(*node))
>>   		return -EINVAL;
>>   
>>   	return 0;
>>   }
>>   
>> +struct dwc3_glue_ops qcom_ops = {
>> +	.glue_configure = dwc3_qcom_glue_configure,
>> +};
>> +
>> +struct dwc3_glue_ops qcom_flat_dt_ops = {
>> +	.glue_configure = dwc3_qcom_glue_configure,
>> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
>> +};
>> +
>>   struct dwc3_glue_ops rk_ops = {
>> -	.glue_get_ctrl_dev = dwc3_rk_glue_get_ctrl_dev,
>> +	.glue_get_ctrl_dev = dwc3_flat_dt_get_ctrl_dev,
>>   };
>>   
>>   static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
>>   {
>> @@ -711,8 +719,9 @@ static const struct udevice_id dwc3_glue_ids[] = {
>>   	{ .compatible = "rockchip,rk3568-dwc3", .data = (ulong)&rk_ops },
>>   	{ .compatible = "rockchip,rk3576-dwc3", .data = (ulong)&rk_ops },
>>   	{ .compatible = "rockchip,rk3588-dwc3", .data = (ulong)&rk_ops },
>>   	{ .compatible = "qcom,dwc3", .data = (ulong)&qcom_ops },
>> +	{ .compatible = "qcom,snps-dwc3", .data = (ulong)&qcom_flat_dt_ops },
>>   	{ .compatible = "fsl,imx8mp-dwc3", .data = (ulong)&imx8mp_ops },
>>   	{ .compatible = "fsl,imx8mq-dwc3" },
>>   	{ .compatible = "intel,tangier-dwc3" },
>>   	{ .compatible = "samsung,exynos7870-dwusb3" },
>>
>> -- 
>> 2.51.0

Re: [PATCH 2/3] usb: dwc3-generic: support Qualcomm flattened DT

[Mattijs Korpershoek]

On Fri, Jan 16, 2026 at 18:54, Casey Connolly <casey.connolly@linaro.org> wrote:

> Hi Mattijs,
>
> On 1/15/26 10:21, Mattijs Korpershoek wrote:
>> Hi Casey,
>> 
>> Thank you for the patch.
>> 
>> On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:
>> 
>>> Qualcomm devicetrees are moving away from having a glue node with dwc3
>>> as a subnode and now may just have a single flattened node.
>> 
>> In the commit message, could you provide a link to an example dt who has
>> this single flattened node?
>
> Sorry I just saw this after posting v2... You can grep for the new 

No worries.

> compatible string in upstream DT, sc7280, sm8650 and a few other 
> platforms are already switched over, eventually all the used/maintained 
> platforms will get switched over I suspect.

Thanks for the hints. That's good enough for me!

>
>> 
>>>

[PATCH 3/3] usb: dwc3-generic: respect role-switch-default-mode

[Casey Connolly]

Respect the default mode configured in DT if it is set and dr_mode is
OTG. This allows the board to prefer host mode even on a DRD capable
controller.

Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
---
 drivers/usb/dwc3/dwc3-generic.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
index 7044696ced0c..1aa9fc933282 100644
--- a/drivers/usb/dwc3/dwc3-generic.c
+++ b/drivers/usb/dwc3/dwc3-generic.c
@@ -528,8 +528,13 @@ static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
 	dr_mode = usb_get_dr_mode(dev_ofnode(parent));
 	if (!dr_mode)
 		dr_mode = usb_get_dr_mode(node);
 
+	/* check default role for DRD controllers */
+	if ((dr_mode == USB_DR_MODE_OTG || dr_mode == USB_DR_MODE_UNKNOWN) &&
+	    ofnode_read_bool(node, "usb-role-switch"))
+		dr_mode = usb_get_role_switch_default_mode(node);
+
 	/* usb mode must fallback to peripheral if not known */
 	if (dr_mode == USB_DR_MODE_UNKNOWN)
 		dr_mode = USB_DR_MODE_OTG;
 

-- 
2.51.0

Re: [PATCH 3/3] usb: dwc3-generic: respect role-switch-default-mode

[Neil Armstrong]

On 1/14/26 15:13, Casey Connolly wrote:
> Respect the default mode configured in DT if it is set and dr_mode is
> OTG. This allows the board to prefer host mode even on a DRD capable
> controller.
> 
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
> ---
>   drivers/usb/dwc3/dwc3-generic.c | 5 +++++
>   1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-generic.c
> index 7044696ced0c..1aa9fc933282 100644
> --- a/drivers/usb/dwc3/dwc3-generic.c
> +++ b/drivers/usb/dwc3/dwc3-generic.c
> @@ -528,8 +528,13 @@ static int dwc3_glue_bind_common(struct udevice *parent, ofnode node)
>   	dr_mode = usb_get_dr_mode(dev_ofnode(parent));
>   	if (!dr_mode)
>   		dr_mode = usb_get_dr_mode(node);
>   
> +	/* check default role for DRD controllers */
> +	if ((dr_mode == USB_DR_MODE_OTG || dr_mode == USB_DR_MODE_UNKNOWN) &&
> +	    ofnode_read_bool(node, "usb-role-switch"))
> +		dr_mode = usb_get_role_switch_default_mode(node);
> +
>   	/* usb mode must fallback to peripheral if not known */
>   	if (dr_mode == USB_DR_MODE_UNKNOWN)
>   		dr_mode = USB_DR_MODE_OTG;
>   
> 

So it means that,on phone platforms, the controller will default to host ?

Neil

Re: [PATCH 3/3] usb: dwc3-generic: respect role-switch-default-mode

[Casey Connolly]

On 14/01/2026 15:39, Neil Armstrong wrote:
> On 1/14/26 15:13, Casey Connolly wrote:
>> Respect the default mode configured in DT if it is set and dr_mode is
>> OTG. This allows the board to prefer host mode even on a DRD capable
>> controller.
>>
>> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>
>> ---
>>   drivers/usb/dwc3/dwc3-generic.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/usb/dwc3/dwc3-generic.c b/drivers/usb/dwc3/dwc3-
>> generic.c
>> index 7044696ced0c..1aa9fc933282 100644
>> --- a/drivers/usb/dwc3/dwc3-generic.c
>> +++ b/drivers/usb/dwc3/dwc3-generic.c
>> @@ -528,8 +528,13 @@ static int dwc3_glue_bind_common(struct udevice
>> *parent, ofnode node)
>>       dr_mode = usb_get_dr_mode(dev_ofnode(parent));
>>       if (!dr_mode)
>>           dr_mode = usb_get_dr_mode(node);
>>   +    /* check default role for DRD controllers */
>> +    if ((dr_mode == USB_DR_MODE_OTG || dr_mode ==
>> USB_DR_MODE_UNKNOWN) &&
>> +        ofnode_read_bool(node, "usb-role-switch"))
>> +        dr_mode = usb_get_role_switch_default_mode(node);
>> +
>>       /* usb mode must fallback to peripheral if not known */
>>       if (dr_mode == USB_DR_MODE_UNKNOWN)
>>           dr_mode = USB_DR_MODE_OTG;
>>  
> 
> So it means that,on phone platforms, the controller will default to host ?

No, it will default to whatever role-switch-default-mode is if it's set,
phones will have this set to peripheral if it's set at all.

> 
> Neil

-- 
// Casey (she/her)

Re: [PATCH 3/3] usb: dwc3-generic: respect role-switch-default-mode

[Mattijs Korpershoek]

Hi Casey,

Thank you for the patch.

On Wed, Jan 14, 2026 at 15:13, Casey Connolly <casey.connolly@linaro.org> wrote:

> Respect the default mode configured in DT if it is set and dr_mode is
> OTG. This allows the board to prefer host mode even on a DRD capable
> controller.
>
> Signed-off-by: Casey Connolly <casey.connolly@linaro.org>

Reviewed-by: Mattijs Korpershoek <mkorpershoek@kernel.org>

> ---
>  drivers/usb/dwc3/dwc3-generic.c | 5 +++++
>  1 file changed, 5 insertions(+)
>