uac2/hid gadget issues on Win10 hosts

["N. Chen" <takhv1@xxxxxxxxx>]

I'm attempting to create a composite gadget device that uses uac2 and
hid functions (speakerphone), using the latest kernel branch
(rpi-5.14.y) on a rpi4 device.
Windows10 host does not recognize the device as an audio usb device.
In order to get it recognized, I had to use c_sync in adaptive mode,
as well as make the following adjustments:

diff --git a/drivers/usb/gadget/function/f_uac2.c
b/drivers/usb/gadget/function/f_uac2.c
index ae29ff2b2b68..74a221939ca0 100644
--- a/drivers/usb/gadget/function/f_uac2.c
+++ b/drivers/usb/gadget/function/f_uac2.c
@@ -177,7 +177,7 @@ static struct uac2_input_terminal_descriptor
io_in_it_desc = {

        .bDescriptorSubtype = UAC_INPUT_TERMINAL,
        /* .bTerminalID = DYNAMIC */
-       .wTerminalType = cpu_to_le16(UAC_INPUT_TERMINAL_UNDEFINED),
+       .wTerminalType = cpu_to_le16(UAC_INPUT_TERMINAL_MICROPHONE),
        .bAssocTerminal = 0,
        /* .bCSourceID = DYNAMIC */
        .iChannelNames = 0,
@@ -205,7 +205,7 @@ static struct uac2_output_terminal_descriptor
io_out_ot_desc = {

        .bDescriptorSubtype = UAC_OUTPUT_TERMINAL,
        /* .bTerminalID = DYNAMIC */
-       .wTerminalType = cpu_to_le16(UAC_OUTPUT_TERMINAL_UNDEFINED),
+       .wTerminalType = cpu_to_le16(UAC_OUTPUT_TERMINAL_SPEAKER),
        .bAssocTerminal = 0,
        /* .bSourceID = DYNAMIC */
        /* .bCSourceID = DYNAMIC */
@@ -216,7 +216,7 @@ static struct uac2_ac_header_descriptor ac_hdr_desc = {
        .bLength = sizeof ac_hdr_desc,
        .bDescriptorType = USB_DT_CS_INTERFACE,

-       .bDescriptorSubtype = UAC_MS_HEADER,
+       .bDescriptorSubtype = UAC_HEADER,
        .bcdADC = cpu_to_le16(0x200),
        .bCategory = UAC2_FUNCTION_IO_BOX,
        /* .wTotalLength = DYNAMIC */
@@ -400,7 +400,7 @@ static struct usb_endpoint_descriptor fs_epin_desc = {
        .bDescriptorType = USB_DT_ENDPOINT,

        .bEndpointAddress = USB_DIR_IN,
-       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_ASYNC,
+       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_SYNC,
        /* .wMaxPacketSize = DYNAMIC */
        .bInterval = 1,
 };
@@ -409,7 +409,7 @@ static struct usb_endpoint_descriptor hs_epin_desc = {
        .bLength = USB_DT_ENDPOINT_SIZE,
        .bDescriptorType = USB_DT_ENDPOINT,

-       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_ASYNC,
+       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_SYNC,
        /* .wMaxPacketSize = DYNAMIC */
        .bInterval = 4,
 };
@@ -419,7 +419,7 @@ static struct usb_endpoint_descriptor ss_epin_desc = {
        .bDescriptorType = USB_DT_ENDPOINT,

        .bEndpointAddress = USB_DIR_IN,
-       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_ASYNC,
+       .bmAttributes = USB_ENDPOINT_XFER_ISOC | USB_ENDPOINT_SYNC_SYNC,
        /* .wMaxPacketSize = DYNAMIC */
        .bInterval = 4,
 };
--

For the hid function report descriptor, I used the telephony led page
with off-hook, mute and ring. I was not able to get any hid packets
when tested using Teams. I made the following change, and it allowed
me to get the ring hid alone, but not the led and the off-hook, so I
guess this hack is incorrect:

diff --git a/drivers/usb/gadget/function/f_hid.c
b/drivers/usb/gadget/function/f_hid.c
index bb476e121eae..10882e18cb88 100644
--- a/drivers/usb/gadget/function/f_hid.c
+++ b/drivers/usb/gadget/function/f_hid.c
@@ -549,7 +549,9 @@ static int hidg_setup(struct usb_function *f,
        case ((USB_DIR_OUT | USB_TYPE_CLASS | USB_RECIP_INTERFACE) << 8
                  | HID_REQ_SET_REPORT):
                VDBG(cdev, "set_report | wLength=%d\n", ctrl->wLength);
-               goto stall;
+               req->context  = hidg;
+               req->complete = hidg_set_report_complete;
+               goto respond;
                break;

        case ((USB_DIR_OUT | USB_TYPE_CLASS | USB_RECIP_INTERFACE) << 8
--

I'm aware that the aforementioned changes to uac2 function might be
incorrect as well, but it allowed me to get through the first hurdle.
I have no objections to making it more correct.
I especially value an opinion on this matter from someone that has
worked with Windows10 hosts before, as it is a specific issue with
this OS (it works well for Ubuntu hosts, yet to be tested on Macs).

Thanks,
Tak