Rumble doesn't work on Stadia controller

[asweintraub]

BlueRetro firmware version

1.9.1 and 24.04 beta 21

BlueRetro firmware specification

HW1

BlueRetro firmware variant

Universal

BlueRetro hardware type

External adapter with detachable cord

Manufacturer

blue-retro.com

System used

Nintendo 64

Bluetooth controller brand & name

Google Stadia

What is problem? (only list ONE problem per report)

Vibrations don't work with the Stadia controller.

Previous bugs for this issue were closed, so reopening here.

I don't have a linux machine set up to pull the HID report from the controller, but assuming this HID report is correct, the issue seems to be that USAGE_PAGE16 is not handled in a way that supports the Stadia controller's HID report.

In particular, it looks like the Stadia controller is sending
0x06, 0x0F, 0x00, // Usage Page (PID Page)
instead of
0x05, 0x0F, // Usage Page (PID Page)
which would be properly handled with the existing code.

I'm going to see if I can get a dev environment set up to test this, but I think changing the USAGE_PAGE16 code to the following might fix the problem:

hid_stack[hid_stack_idx].usage_page = *desc;
desc += 2;

What did you expect to happen?

I would expect vibrations to work with the Stadia controller.

Attach files like logs or Bluetooth traces here

No response

[darthcloud]

I'm very open to receiving PRs ;) Hopefully this help a bit:

Usage page is only stored as a 8bits value in the hid_parser stack:

BlueRetro/main/adapter/hid_parser.c

Lines 16 to 22 in 3c45c15

	struct hid_stack_element {
	uint32_t report_size;
	uint32_t report_cnt;
	int32_t logical_min;
	int32_t logical_max;
	uint8_t usage_page;
	};

Would make sense to upgrade it to 32bits since everything else is that, including the global structure used in all other module:

BlueRetro/main/adapter/adapter.h

Lines 469 to 477 in 3c45c15

	struct hid_usage {
	uint32_t usage_page;
	uint32_t usage;
	uint32_t flags;
	uint32_t bit_offset;
	uint32_t bit_size;
	int32_t logical_min;
	int32_t logical_max;
	};

You would need to cast the pointer to uint16_t * to grab the full value.

hid_stack[hid_stack_idx].usage_page = *(uint16_t *)desc;;
desc += 2;

Then after that you would likely need to edit the generic rumble code to support that usage_page/usage combination:

BlueRetro/main/adapter/wireless/hid_generic.c

Lines 700 to 786 in 3c45c15

	void hid_fb_from_generic(struct generic_fb *fb_data, struct bt_data *bt_data) {
	struct generic_rumble *rumble = (struct generic_rumble *)bt_data->base.output;
	rumble->report_size = 0;
	uint32_t bytes_count = 0;
	uint32_t tmp_value = 0;
	uint32_t offset = 0;
	uint32_t counter = 0;
	bool is_rumble_usage = false;
	for (uint32_t i = 0; i < bt_data->reports[RUMBLE].usage_cnt; i++)
	{
	is_rumble_usage = false;
	switch (bt_data->reports[RUMBLE].usages[i].usage)
	{
	case 0x50: /* Duration */
	bytes_count = (bt_data->reports[RUMBLE].usages[i].bit_size + 7) / 8;
	rumble->report_size += bytes_count;
	if (fb_data->state) {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_max;
	}
	else {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_min;
	}
	is_rumble_usage = true;
	break;
	case 0x70: /* Magnitude */
	case 0x97: /* Enable Actuators */
	bytes_count = (bt_data->reports[RUMBLE].usages[i].bit_size + 7) / 8;
	rumble->report_size += bytes_count;
	if (fb_data->state) {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_max * RUMBLE_ON_MULTIPLIER;
	}
	else {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_min;
	}
	is_rumble_usage = true;
	break;
	case 0x7C: /* Loop Count */
	bytes_count = (bt_data->reports[RUMBLE].usages[i].bit_size + 7) / 8;
	rumble->report_size += bytes_count;
	if (fb_data->state) {
	if (fb_data->cycles) {
	tmp_value = fb_data->cycles;
	}
	else {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_max;
	}
	}
	else {
	tmp_value = bt_data->reports[RUMBLE].usages[i].logical_min;
	}
	is_rumble_usage = true;
	break;
	case 0xA7: /* Start Delay */
	bytes_count = (bt_data->reports[RUMBLE].usages[i].bit_size + 7) / 8;
	rumble->report_size += bytes_count;
	tmp_value = fb_data->start;
	is_rumble_usage = true;
	break;
	}
	if (is_rumble_usage) {
	counter = 0;
	while(tmp_value)
	{
	rumble->state[offset++] = tmp_value;
	tmp_value >>= 8;
	counter++;
	}
	for (uint32_t refill = counter; refill < bytes_count; refill++) {
	rumble->state[offset++] = 0;
	}
	}
	}
	rumble->report_id = bt_data->reports[RUMBLE].id;
	}

[asweintraub]

Thanks for the pointers! I probably won't have time to get around to this for a few weeks, since I don't have the ESP-IDF environment set up and will be quite busy. That said, it looks like a good idea to support 32-bit usage pages, so I'll give that a shot when I have the chance.

Either way, it looks as if the values are little-endian, so 0x06, 0x0F, 0x00 should be synonymous with 0x05, 0x0F, meaning we shouldn't have to modify anything else for this to work with Stadia. In fact, the largest usage page I can find in the HID docs is 0x91, so switching this to 32-bits (instead of dropping the extra bytes) is good for future-proofing, but otherwise unnecessary.

[darthcloud]

Yes indeed it was simply the same value using a bigger sized var.

Rumble.work now!

https://github.com/darthcloud/BlueRetro/actions/runs/8808673421

[asweintraub]

Nice! I didn't realize you had already merged the fix in. I modified the code earlier today but wasn't able to test it until just now. Thank you!

[darthcloud]

Fixed in v24.04