Re: [PATCH v2 01/13] HID: playstation: initial DualSense USB support.

Hi Barnabas,

Thanks for your suggestions.

On Thu, Jan 7, 2021 at 9:14 AM Barnabás Pőcze [off-list ref] wrote:

Hi


I have just a couple minor comments.


2021. január 2., szombat 23:30 keltezéssel, Roderick Colenbrander írta:

quoted

From: Roderick Colenbrander roderick.colenbrander@sony.com

Implement support for PlayStation DualSense gamepad in USB mode.
Support features include buttons and sticks, which adhere to the
Linux gamepad spec.

Signed-off-by: Roderick Colenbrander roderick.colenbrander@sony.com
[...]
+/* Common gamepad buttons across DualShock 3 / 4 and DualSense.
+ * Note: for device with a touchpad, touchpad button is not included
+ *        as it will be part of the touchpad device.
+ */
+static const int ps_gamepad_buttons[] = {
+     BTN_WEST, /* Square */
+     BTN_NORTH, /* Triangle */
+     BTN_EAST, /* Circle */
+     BTN_SOUTH, /* Cross */
+     BTN_TL, /* L1 */
+     BTN_TR, /* R1 */
+     BTN_TL2, /* L2 */
+     BTN_TR2, /* R2 */
+     BTN_SELECT, /* Create (PS5) / Share (PS4) */
+     BTN_START, /* Option */
+     BTN_THUMBL, /* L3 */
+     BTN_THUMBR, /* R3 */
+     BTN_MODE, /* PS Home */
+};
+
+static const struct {int x; int y; } ps_gamepad_hat_mapping[] = {
+     {0, -1}, {1, -1}, {1, 0}, {1, 1}, {0, 1}, {-1, 1}, {-1, 0}, {-1, -1},
+     {0, 0}
+};

I believe the preferred way is to have a comma after each array/enum/etc. element
unless it is a terminating entry.

quoted

+
+static struct input_dev *ps_allocate_input_dev(struct hid_device *hdev, const char *name_suffix)
+{
+     struct input_dev *input_dev;
+
+     input_dev = devm_input_allocate_device(&hdev->dev);
+     if (!input_dev)
+             return ERR_PTR(-ENOMEM);
+
+     input_dev->id.bustype = hdev->bus;
+     input_dev->id.vendor = hdev->vendor;
+     input_dev->id.product = hdev->product;
+     input_dev->id.version = hdev->version;
+     input_dev->uniq = hdev->uniq;
+
+     if (name_suffix) {
+             input_dev->name = devm_kasprintf(&hdev->dev, GFP_KERNEL, "%s %s", hdev->name,
+                             name_suffix);
+             if (!input_dev->name)
+                     return ERR_PTR(-ENOMEM);
+     } else {
+             input_dev->name = hdev->name;
+     }
+
+     input_set_drvdata(input_dev, hdev);
+
+     return input_dev;
+}
+
+static struct input_dev *ps_gamepad_create(struct hid_device *hdev)
+{
+     struct input_dev *gamepad;
+     unsigned int i;
+     int ret;
+
+     gamepad = ps_allocate_input_dev(hdev, NULL);
+     if (IS_ERR(gamepad))
+             return ERR_PTR(-ENOMEM);

I know that at the moment ENOMEM is the only possible error, but I believe
`return ERR_CAST(gamepad);` would be better. (Or even just `return gamepad;`.)

quoted

+
+     input_set_abs_params(gamepad, ABS_X, 0, 255, 0, 0);
+     input_set_abs_params(gamepad, ABS_Y, 0, 255, 0, 0);
+     input_set_abs_params(gamepad, ABS_Z, 0, 255, 0, 0);
+     input_set_abs_params(gamepad, ABS_RX, 0, 255, 0, 0);
+     input_set_abs_params(gamepad, ABS_RY, 0, 255, 0, 0);
+     input_set_abs_params(gamepad, ABS_RZ, 0, 255, 0, 0);
+
+     input_set_abs_params(gamepad, ABS_HAT0X, -1, 1, 0, 0);
+     input_set_abs_params(gamepad, ABS_HAT0Y, -1, 1, 0, 0);
+
+     for (i = 0; i < ARRAY_SIZE(ps_gamepad_buttons); i++)
+             input_set_capability(gamepad, EV_KEY, ps_gamepad_buttons[i]);
+
+     ret = input_register_device(gamepad);
+     if (ret)
+             return ERR_PTR(ret);
+
+     return gamepad;
+}
+
+static int dualsense_parse_report(struct ps_device *ps_dev, struct hid_report *report,
+             u8 *data, int size)
+{
+     struct hid_device *hdev = ps_dev->hdev;
+     struct dualsense *ds = container_of(ps_dev, struct dualsense, base);
+     struct dualsense_input_report *ds_report;
+     uint8_t value;
+

I think `size` should be checked somewhere around here.

quoted

+     /* DualSense in USB uses the full HID report for reportID 1, but
+      * Bluetooth uses a minimal HID report for reportID 1 and reports
+      * the full report using reportID 49.
+      */
+     if (report->id == DS_INPUT_REPORT_USB && hdev->bus == BUS_USB) {
+             ds_report = (struct dualsense_input_report *)&data[1];
+     } else {
+             hid_err(hdev, "Unhandled reportID=%d\n", report->id);
+             return -1;
+     }
+
+     input_report_abs(ds->gamepad, ABS_X,  ds_report->x);
+     input_report_abs(ds->gamepad, ABS_Y,  ds_report->y);
+     input_report_abs(ds->gamepad, ABS_RX, ds_report->rx);
+     input_report_abs(ds->gamepad, ABS_RY, ds_report->ry);
+     input_report_abs(ds->gamepad, ABS_Z,  ds_report->z);
+     input_report_abs(ds->gamepad, ABS_RZ, ds_report->rz);
+
+     value = ds_report->buttons[0] & DS_BUTTONS0_HAT_SWITCH;
+     if (value > 7)
+             value = 8; /* center */

This seems a bit flimsy to me, it relies on a different part of the code
being in a certain way that is not enforced by anything

What do you mean with not enforced? I'm not saying I'm a big fan of
the code, but HATs seem to work like this. The DualShock4/DualSense
describe it in their HID descriptors with a logical minimum value of 0
and a max value of 7.

The code is very similar to hid-input.c:

static const struct {
__s32 x;
__s32 y;
}  hid_hat_to_axis[] = {{ 0, 0}, { 0,-1}, { 1,-1}, { 1, 0}, { 1, 1}, {
0, 1}, {-1, 1}, {-1, 0}, {-1,-1}};

int hat_dir = usage->hat_dir;
if (!hat_dir)
    hat_dir = (value - usage->hat_min) * 8 / (usage->hat_max -
usage->hat_min + 1) + 1;
if (hat_dir < 0 || hat_dir > 8) hat_dir = 0;
input_event(input, usage->type, usage->code    , hid_hat_to_axis[hat_dir].x);
input_event(input, usage->type, usage->code + 1, hid_hat_to_axis[hat_dir].y);

Main difference seems to be that this code places {0, 0} at the start
and adds a "+1" to avoid having to set the value to "8" when out of
range.

 I'd probably do something

like this:

enum {
  HAT_DIR_W = 0,
  HAT_DIR_NW,
  ...
  HAT_DIR_SW,
  HAT_DIR_NONE,
};

static const struct {int x; int y; } ps_gamepad_hat_mapping[] = {
  [HAT_DIR_W] = {0, -1},
  ...
  [HAT_DIR_NONE] = {0, 0},
};

and then

if (value >= ARRAY_SIZE(ps_gamepad_hat_mapping))
  value = HAT_DIR_NONE;

Please consider it. By the way, are values 9..15 actually sent by the controller?

See above. They are not sent. The Hat Switch in the report descriptor
is reported with a logical minimum of 0 and a max of 8.

quoted

+     input_report_abs(ds->gamepad, ABS_HAT0X, ps_gamepad_hat_mapping[value].x);
+     input_report_abs(ds->gamepad, ABS_HAT0Y, ps_gamepad_hat_mapping[value].y);
+
+     input_report_key(ds->gamepad, BTN_WEST,   ds_report->buttons[0] & DS_BUTTONS0_SQUARE);
+     input_report_key(ds->gamepad, BTN_SOUTH,  ds_report->buttons[0] & DS_BUTTONS0_CROSS);
+     input_report_key(ds->gamepad, BTN_EAST,   ds_report->buttons[0] & DS_BUTTONS0_CIRCLE);
+     input_report_key(ds->gamepad, BTN_NORTH,  ds_report->buttons[0] & DS_BUTTONS0_TRIANGLE);
+     input_report_key(ds->gamepad, BTN_TL,     ds_report->buttons[1] & DS_BUTTONS1_L1);
+     input_report_key(ds->gamepad, BTN_TR,     ds_report->buttons[1] & DS_BUTTONS1_R1);
+     input_report_key(ds->gamepad, BTN_TL2,    ds_report->buttons[1] & DS_BUTTONS1_L2);
+     input_report_key(ds->gamepad, BTN_TR2,    ds_report->buttons[1] & DS_BUTTONS1_R2);
+     input_report_key(ds->gamepad, BTN_SELECT, ds_report->buttons[1] & DS_BUTTONS1_CREATE);
+     input_report_key(ds->gamepad, BTN_START,  ds_report->buttons[1] & DS_BUTTONS1_OPTIONS);
+     input_report_key(ds->gamepad, BTN_THUMBL, ds_report->buttons[1] & DS_BUTTONS1_L3);
+     input_report_key(ds->gamepad, BTN_THUMBR, ds_report->buttons[1] & DS_BUTTONS1_R3);
+     input_report_key(ds->gamepad, BTN_MODE,   ds_report->buttons[2] & DS_BUTTONS2_PS_HOME);
+     input_sync(ds->gamepad);
+
+     return 0;
+}
+
+static struct ps_device *dualsense_create(struct hid_device *hdev)
+{
+     struct dualsense *ds;
+     int ret;
+
+     ds = devm_kzalloc(&hdev->dev, sizeof(*ds), GFP_KERNEL);
+     if (!ds)
+             return ERR_PTR(-ENOMEM);
+
+     /* Patch version to allow userspace to distinguish between
+      * hid-generic vs hid-playstation axis and button mapping.
+      */
+     hdev->version |= HID_PLAYSTATION_VERSION_PATCH;
+
+     ds->base.hdev = hdev;
+     ds->base.parse_report = dualsense_parse_report;
+     hid_set_drvdata(hdev, ds);
+
+     ds->gamepad = ps_gamepad_create(hdev);
+     if (IS_ERR(ds->gamepad)) {
+             ret = PTR_ERR(ds->gamepad);
+             goto err;
+     }
+
+     return &ds->base;
+
+err:
+     return ERR_PTR(ret);
+}
[...]
+static int ps_probe(struct hid_device *hdev, const struct hid_device_id *id)
+{
+     struct ps_device *dev;
+     int ret;
+
+     ret = hid_parse(hdev);
+     if (ret) {
+             hid_err(hdev, "parse failed\n");
+             return ret;
+     }
+
+     ret = hid_hw_start(hdev, HID_CONNECT_HIDRAW);
+     if (ret) {
+             hid_err(hdev, "hw start failed\n");
+             return ret;
+     }
+
+     ret = hid_hw_open(hdev);
+     if (ret) {
+             hid_err(hdev, "hw open failed\n");
+             goto err_stop;
+     }
+
+     if (hdev->product == USB_DEVICE_ID_SONY_PS5_CONTROLLER) {

I'm still not fully seeing the purpose of this `if`. The probe should not be
called for devices not in the id_table, so this seems to me to be a long way
of writing `if (true)`. Or am I missing something?

It is not used. It more there for the future, when we will add
DualShock 4 and perhaps some other devices here.

quoted

+             dev = dualsense_create(hdev);
+             if (IS_ERR(dev)) {
+                     hid_err(hdev, "Failed to create dualsense.\n");

I think it'd be preferable if all log messages would either be lowercase or
uppercase, not a mix of both. Same for punctuation. This applies to all patches.

quoted

+                     ret = PTR_ERR(dev);
+                     goto err_close;
+             }
+     }
+
+     return ret;
+
+err_close:
+     hid_hw_close(hdev);
+err_stop:
+     hid_hw_stop(hdev);
+     return ret;
+}
[...]

Regards,
Barnabás Pőcze

Regards,
Roderick Colenbrander