Support MSM8916 USB PHY + UDC in qcom_defconfig
Subsystems (2)
v2 cover Support MSM8916 USB PHY + UDC in qcom_defconfig
The MSM8916 SoC includes a ChipIdea UDC. The U-Boot driver for this UDC is quite old, and did not support USB gadget driver model. As a result, it could not be built alongside the DWC3 driver. This meant that qcom_defconfig could not include support for USB on MSM8916.
This series adds support to build ChipIdea UDC with DM_USB_GADGET and ensures that the newly introduced driver is registered from the MSM EHCI glue driver.
The ChipIdea UDC driver continues to work without DM_USB_GADGET. I tested these changes by building dragonboard410c_defconfig (which does not enable DM_USB_GADGET), booting it, and then chain-booting (via fastboot) another u-boot build built with qcom_defconfig. USB gadget mode worked in both builds.
---
Changes in v2 (thanks Stephan):
- Deduplicate CI UDC setup/teardown paths
- Formatting fix
- Regenerate qcom_defconfig with savedefconfig
- Link to v1: https://lore.kernel.org/r/[email protected]
---
Sam Day (3):
usb: udc: ci: support USB gadget driver model
usb: host: ehci-msm: Register ChipIdea UDC from glue wrapper
qcom_defconfig: Enable MSM8916 USB PHY + ChipIdea UDC
configs/qcom_defconfig | 4 +- drivers/usb/gadget/Kconfig | 1 - drivers/usb/gadget/ci_udc.c | 118 ++++++++++++++++++++++++++++++++++++-------- drivers/usb/host/ehci-msm.c | 7 +++ 4 files changed, 107 insertions(+), 23 deletions(-)
---
base-commit: 987907ae4bcc5d6055bdf7d318a3edf53e14d5fa
change-id: 20250412-msm8916-usb-a1dc93c01bd5
Best regards,
Signature · 1 line
Sam Day <[email protected]>
PATCH v2 1/3 usb: udc: ci: support USB gadget driver model
When CONFIG_DM_USB_GADGET is enabled, a UCLASS_USB_GADGET_GENERIC driver will be defined that wraps the ChipIdea UDC operations. The (dm_)?usb_gadget_.* symbols will no longer be defined (as these are now handled by the UDC uclass).
If CONFIG_DM_USB_GADGET is not enabled, this driver behaves as it previously did.
This new driver does not declare any compatibles of its own. It requires a glue driver to bind it. The ehci_msm driver will be updated in the following commit to do exactly that.
--- drivers/usb/gadget/Kconfig | 1 - drivers/usb/gadget/ci_udc.c | 118 ++++++++++++++++++++++++++++++++++++-------- 2 files changed, 97 insertions(+), 22 deletions(-)
drivers/usb/gadget/Kconfig +0 -1
@@ -153,7 +153,6 @@ config USB_GADGET_OS_DESCRIPTORS config CI_UDC bool "ChipIdea device controller" - depends on !DM_USB_GADGET select USB_GADGET_DUALSPEED help Say Y here to enable device controller functionality of the
drivers/usb/gadget/ci_udc.c +97 -21
178 lines not on this page · raw message
Signature · 1 line
2.54.0
Re: [PATCH v2 1/3] usb: udc: ci: support USB gadget driver model
Hi Sam,
Thank you for the patch.
Sorry for the delays in the review for this one.
Quoted 80 lines · Sam Day
When
CONFIG_DM_USB_GADGETis enabled, aUCLASS_USB_GADGET_GENERICdriver will be defined that wraps the ChipIdea UDC operations. The (dm_)?usb_gadget_.* symbols will no longer be defined (as these are now handled by the UDC uclass).If
CONFIG_DM_USB_GADGETis not enabled, this driver behaves as it previously did.This new driver does not declare any compatibles of its own. It requires a glue driver to bind it. The
ehci_msmdriver will be updated in the following commit to do exactly that.Signed-off-by Sam Day <[email protected]>--- drivers/usb/gadget/Kconfig | 1 - drivers/usb/gadget/ci_udc.c | 118 ++++++++++++++++++++++++++++++++++++-------- 2 files changed, 97 insertions(+), 22 deletions(-)drivers/usb/gadget/Kconfig +0
-1@@ -153,7 +153,6 @@ config USB_GADGET_OS_DESCRIPTORS config CI_UDC bool "ChipIdea device controller" - depends on !DM_USB_GADGET select USB_GADGET_DUALSPEED help Say Y here to enable device controller functionality of thedrivers/usb/gadget/ci_udc.c +12
-2@@ -10,6 +10,7 @@#include <command.h> #include <config.h> #include <cpu_func.h> +#include <dm.h> #include <net.h> #include <malloc.h> #include <wait_bit.h> @@ -94,8 +95,17 @@ static struct usb_request * ci_ep_alloc_request(struct usb_ep *ep, unsigned int gfp_flags);static void ci_ep_free_request (struct usb_ep *ep,struct usb_request *_req); +#if CONFIG_IS_ENABLED (DM_USB_GADGET ) +static int ci_udc_gadget_start (struct usb_gadget *g,struct usb_gadget_driver *driver); +static int ci_udc_gadget_stop (struct usb_gadget *g); +#endif +static const struct usb_gadget_ops ci_udc_ops = { .pullup = ci_pullup, +#if CONFIG_IS_ENABLED(DM_USB_GADGET) + .udc_start = ci_udc_gadget_start, + .udc_stop = ci_udc_gadget_stop, +#endif };static const struct usb_ep_ops ci_ep_ops = {@@ -978,7 +988,7 @@ void udc_irq(void) } } -int dm_usb_gadget_handle_interrupts (struct udevice *dev) +static int ci_udc_handle_interrupts (struct udevice *dev) {struct ci_udc *udc = (struct ci_udc *)controller.ctrl->hcor;u32 value;@@ -1037,7 +1047,7 @@ static int ci_pullup(struct usb_gadget *gadget, int is_on)static int ci_udc_probe (void ) {struct ept_queue_head *head; -int i; +int i, ret;const int num =2 *NUM_ENDPOINTS ;@@ -1046,6 +1056,14 @@ static int ci_udc_probe(void)const int eplist_raw_sz = num *sizeof (struct ept_queue_head );
drivers/usb/gadget/ci_udc.c static int ci_udc_probe(void) · line 1066
Quoted 10 lines
drivers/usb/gadget/ci_udc.c +8
-0const int eplist_sz =roundup (eplist_raw_sz,ARCH_DMA_MINALIGN ); +if (IS_ENABLED (CONFIG_DM_USB )) + ret =usb_setup_ehci_gadget (&controller.ctrl); +else + ret =usb_lowlevel_init (0 ,USB_INIT_DEVICE , (void **)&controller.ctrl); + +if (ret) +return ret; +
What happens if later down in ci_udc_probe() we fail to finish the function sucessfully?
For example, memalign(ILIST_ALIGN, ILIST_SZ) can return -ENOMEM.
Doesn't that leaves us with an usb that's halfway configured? Shouldn't we add some error handling / do some cleanup in that case?
Quoted 12 lines
drivers/usb/gadget/ci_udc.c +5
-0/* The QH list must be aligned to 4096 bytes. */ controller.epts =memalign (eplist_align, eplist_sz);if (!controller.epts)@@ -1123,6 +1141,73 @@ static int ci_udc_probe(void)return 0 ; } +static void ci_udc_remove (void ) +{ +if (IS_ENABLED (CONFIG_DM_USB )) { +usb_remove_ehci_gadget (&controller.ctrl); + }else {
drivers/usb/gadget/ci_udc.c static int ci_udc_probe(void) · line 1158
Quoted 10 lines
drivers/usb/gadget/ci_udc.c +10
-0+usb_lowlevel_stop (0 ); + controller.ctrl =NULL ; + } +} + +static void ci_udc_stop (void ) +{ +ci_ep_free_request (&controller.ep[0 ].ep, &controller.ep0_req->req); +free (controller.items_mem); +free (controller.epts);
It feels a bit weird to free the endpoint request in stop(). Shouldn't we do this in the remove() function to match the ci_ep_alloc_request() which is done in ci_udc_probe()?
drivers/usb/gadget/ci_udc.c static int ci_udc_probe(void) · line 1165
Quoted 7 lines
drivers/usb/gadget/ci_udc.c +7
-0+} + +static int ci_udc_gadget_start (struct usb_gadget *g,struct usb_gadget_driver *driver) +{ + controller.driver = driver; +return 0 ; +}
Nitpick: could we please move this next to ci_udc_gadget_stop() ? I know this would force you to write add controller.driver = driver in usb_gadget_register_driver() but I don't this it's a problem given that we do something similar in usb_gadget_unregister_driver().
Quoted 26 lines
drivers/usb/gadget/ci_udc.c +26
-0+ +#if CONFIG_IS_ENABLED (DM_USB_GADGET ) +static int ci_udc_generic_probe (struct udevice *dev) +{ +int ret =ci_udc_probe (); + +if (ret) +return ret; + +return usb_add_gadget_udc ((struct device *)dev, &controller.gadget); +} + +static int ci_udc_generic_remove (struct udevice *dev) +{ +usb_del_gadget_udc (&controller.gadget); + +ci_udc_remove (); + +return 0 ; +} + +static const struct usb_gadget_generic_ops ci_udc_generic_ops = { + .handle_interrupts = ci_udc_handle_interrupts, +}; + +U_BOOT_DRIVER(ci_udc_generic) = {
drivers/usb/gadget/ci_udc.c static int ci_udc_probe(void) · line 1201
Quoted 10 lines
drivers/usb/gadget/ci_udc.c +10
-0+ .name = "ci-udc", + .id = UCLASS_USB_GADGET_GENERIC, + .ops = &ci_udc_generic_ops, + .probe = ci_udc_generic_probe, + .remove = ci_udc_generic_remove, +}; + +static int ci_udc_gadget_stop (struct usb_gadget *g) +{ +/* Avoid calling disconnect() twice; the UDC uclass already did it. */
There is no disconnect in udc-uclass.c? Did you mean "UDC core" here instead of "UDC uclass" ?
Quoted 69 lines
drivers/usb/gadget/ci_udc.c +19
-19+ controller.driver =NULL ; +udc_disconnect (); + +ci_udc_stop (); + + return 0; +} + +#else int usb_gadget_register_driver (struct usb_gadget_driver *driver) {int ret;@@ -1132,14 +1217,6 @@ int usb_gadget_register_driver(struct usb_gadget_driver *driver)if (!driver->bind || !driver->setup || !driver->disconnect)return -EINVAL ; -#if CONFIG_IS_ENABLED (DM_USB ) - ret =usb_setup_ehci_gadget (&controller.ctrl); -#else - ret =usb_lowlevel_init (0 ,USB_INIT_DEVICE , (void **)&controller.ctrl); -#endif -if (ret) -return ret; - ret =ci_udc_probe ();if (ret) {DBG ("udc probe failed, returned %d\n" , ret);@@ -1151,7 +1228,8 @@ int usb_gadget_register_driver(struct usb_gadget_driver *driver)DBG ("driver->bind() returned %d\n" , ret);return ret; } - controller.driver = driver; + +ci_udc_gadget_start (&controller.gadget, driver);return 0 ; }@@ -1163,20 +1241,18 @@ int usb_gadget_unregister_driver(struct usb_gadget_driver *driver) driver->unbind (&controller.gadget); controller.driver =NULL ; -ci_ep_free_request (&controller.ep[0 ].ep, &controller.ep0_req->req); -free (controller.items_mem); -free (controller.epts); - -#if CONFIG_IS_ENABLED (DM_USB ) -usb_remove_ehci_gadget (&controller.ctrl); -#else -usb_lowlevel_stop (0 ); - controller.ctrl =NULL ; -#endif +ci_udc_stop (); +ci_udc_remove (); return 0; } +int dm_usb_gadget_handle_interrupts (struct udevice *dev) +{ +return ci_udc_handle_interrupts (dev); +} +#endif +bool dfu_usb_get_reset (void ) {struct ci_udc *udc = (struct ci_udc *)controller.ctrl->hcor;Signature · 2 lines
-- 2.54.0
PATCH v2 2/3 usb: host: ehci-msm: Register ChipIdea UDC from glue wrapper
In reply to Sam DayWhen CONFIG_DM_USB_GADGET is enabled, the CI UDC driver needs to be explicitly bound. So we do this from the newly introduced glue driver.
--- drivers/usb/host/ehci-msm.c | 7 +++++++ 1 file changed, 7 insertions(+)
drivers/usb/host/ehci-msm.c +7 -0
@@ -161,6 +161,13 @@ static int qcom_ci_hdrc_bind(struct udevice *dev)if (ret)return ret; +if (IS_ENABLED (CONFIG_DM_USB_GADGET ) &&IS_ENABLED (CONFIG_CI_UDC )) { + ret =device_bind_driver_to_node (dev,"ci-udc" ,"ci-udc" , +dev_ofnode (dev),NULL ); +if (ret) +return ret; + } +if (!ofnode_valid (ulpi_node))return 0 ;
Signature · 1 line
2.54.0
Re: [PATCH v2 2/3] usb: host: ehci-msm: Register ChipIdea UDC from glue wrapper
Hi Sam,
Thank you for the patch.
Quoted 4 lines · Sam Day
When
CONFIG_DM_USB_GADGETis enabled, the CI UDC driver needs to be explicitly bound. So we do this from the newly introduced glue driver.Signed-off-by Sam Day <[email protected]>
Quoted 1 line
---
PATCH v2 3/3 qcom_defconfig: Enable MSM8916 USB PHY + ChipIdea UDC
In reply to Sam DayNow that the ChipIdea UDC driver supports USB Gadget driver model, we can enable it alongside the dwc3 generic driver.
--- configs/qcom_defconfig | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
configs/qcom_defconfig +3 -1
@@ -108,7 +108,6 @@ CONFIG_DWC_ETH_QOS_QCOM=y CONFIG_RGMII=y CONFIG_PHY_MICREL=y CONFIG_PHY_MICREL_KSZ90X1=y -CONFIG_PHY=y CONFIG_PHY_QCOM_QMP_UFS=y CONFIG_PHY_QCOM_QUSB2=y CONFIG_PHY_QCOM_USB_SNPS_FEMTO_V2=y
@@ -139,9 +138,12 @@ CONFIG_SYSRESET_QCOM_PSHOLD=y CONFIG_USB=y CONFIG_USB_XHCI_HCD=y CONFIG_USB_XHCI_DWC3=y +CONFIG_USB_EHCI_HCD=y +CONFIG_USB_EHCI_MSM=y CONFIG_USB_DWC3=y CONFIG_USB_DWC3_GENERIC=y CONFIG_USB_GADGET=y +CONFIG_CI_UDC=y CONFIG_UFS=y CONFIG_UFS_QCOM=y CONFIG_VIDEO=y
Signature · 1 line
2.54.0
Re: [PATCH v2 3/3] qcom_defconfig: Enable MSM8916 USB PHY + ChipIdea UDC
Hi Sam,
Thank you for the patch.
Quoted 4 lines · Sam Day
Now that the ChipIdea UDC driver supports USB Gadget driver model, we can enable it alongside the dwc3 generic driver.
Signed-off-by Sam Day <[email protected]>
Quoted 1 line
---
Re: [PATCH v2 3/3] qcom_defconfig: Enable MSM8916 USB PHY + ChipIdea UDC
In reply to Sam DayHi Sam,
Thanks for the patch.
Quoted 13 lines · Sam Day
Now that the ChipIdea UDC driver supports USB Gadget driver model, we can enable it alongside the dwc3 generic driver.
Signed-off-by Sam Day <[email protected]>--- configs/qcom_defconfig | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)configs/qcom_defconfig +0
-0
configs/qcom_defconfig CONFIG_DWC_ETH_QOS_QCOM=y · line 111
Quoted 4 lines
configs/qcom_defconfig +0
-1@@ -108,7 +108,6 @@ CONFIG_DWC_ETH_QOS_QCOM=y CONFIG_RGMII=y CONFIG_PHY_MICREL=y CONFIG_PHY_MICREL_KSZ90X1=y -CONFIG_PHY=y
Is this change related to the goal of this patch? Maybe we can have this as seperate patch?
Regards,
Balaji
Quoted 17 lines
configs/qcom_defconfig +3
-0CONFIG_PHY_QCOM_QMP_UFS=y CONFIG_PHY_QCOM_QUSB2=y CONFIG_PHY_QCOM_USB_SNPS_FEMTO_V2=y@@ -139,9 +138,12 @@ CONFIG_SYSRESET_QCOM_PSHOLD=y CONFIG_USB=y CONFIG_USB_XHCI_HCD=y CONFIG_USB_XHCI_DWC3=y +CONFIG_USB_EHCI_HCD=y +CONFIG_USB_EHCI_MSM=y CONFIG_USB_DWC3=y CONFIG_USB_DWC3_GENERIC=y CONFIG_USB_GADGET=y +CONFIG_CI_UDC=y CONFIG_UFS=y CONFIG_UFS_QCOM=y CONFIG_VIDEO=y