[PATCH] hid: sony: Drop invalid Sixaxis input reports

[PATCH] hid: sony: Drop invalid Sixaxis input reports

[Frank Praznik]

When connected via Bluetooth the sixaxis periodically sends reports with
an ID of 1, the second byte 0xff and the rest zeroed.  These reports are
not related to the controller state and must be dropped to avoid generating
false input events.

Signed-off-by: Frank Praznik <frank.praznik@oh.rr.com>
---
 drivers/hid/hid-sony.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/hid/hid-sony.c b/drivers/hid/hid-sony.c
index ed2f008..beb2b02 100644
--- a/drivers/hid/hid-sony.c
+++ b/drivers/hid/hid-sony.c
@@ -1270,6 +1270,17 @@ static int sony_raw_event(struct hid_device *hdev, struct hid_report *report,
 	 * has to be BYTE_SWAPPED before passing up to joystick interface
 	 */
 	if ((sc->quirks & SIXAXIS_CONTROLLER) && rd[0] == 0x01 && size == 49) {
+		/*
+		 * When connected via Bluetooth the Sixaxis occasionally sends
+		 * a report with the second byte 0xff and the rest zeroed.
+		 *
+		 * This report does not reflect the actual state of the
+		 * controller must be ignored to avoid generating false input
+		 * events.
+		 */
+		if (rd[1] == 0xff)
+			return -EINVAL;
+
 		swap(rd[41], rd[42]);
 		swap(rd[43], rd[44]);
 		swap(rd[45], rd[46]);
-- 
2.4.3

Re: [PATCH] hid: sony: Drop invalid Sixaxis input reports

[simon]

> When connected via Bluetooth the sixaxis periodically sends reports with
> an ID of 1, the second byte 0xff and the rest zeroed.  These reports are
> not related to the controller state and must be dropped to avoid
> generating false input events.

These 'problem reports' were reported (*) over the on Bluetooth list, and
appear in both the HID and the HCI streams (suggesting that they are
really sent from device). They are not seen on the USB connection.

Dropping them in 'raw_event()' seems a sensible approach,
Simon.

* http://www.spinics.net/lists/linux-bluetooth/msg63028.html


>
> Signed-off-by: Frank Praznik <frank.praznik@oh.rr.com>
> ---
>  drivers/hid/hid-sony.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
>
> diff --git a/drivers/hid/hid-sony.c b/drivers/hid/hid-sony.c
> index ed2f008..beb2b02 100644
> --- a/drivers/hid/hid-sony.c
> +++ b/drivers/hid/hid-sony.c
> @@ -1270,6 +1270,17 @@ static int sony_raw_event(struct hid_device *hdev,
> struct hid_report *report,
>  	 * has to be BYTE_SWAPPED before passing up to joystick interface
>  	 */
>  	if ((sc->quirks & SIXAXIS_CONTROLLER) && rd[0] == 0x01 && size == 49) {
> +		/*
> +		 * When connected via Bluetooth the Sixaxis occasionally sends
> +		 * a report with the second byte 0xff and the rest zeroed.
> +		 *
> +		 * This report does not reflect the actual state of the
> +		 * controller must be ignored to avoid generating false input
> +		 * events.
> +		 */
> +		if (rd[1] == 0xff)
> +			return -EINVAL;
> +
>  		swap(rd[41], rd[42]);
>  		swap(rd[43], rd[44]);
>  		swap(rd[45], rd[46]);
> --
> 2.4.3
>
>

Re: [PATCH] hid: sony: Drop invalid Sixaxis input reports

[Jiri Kosina]

On Thu, 23 Jul 2015, Frank Praznik wrote:

> When connected via Bluetooth the sixaxis periodically sends reports with
> an ID of 1, the second byte 0xff and the rest zeroed.  These reports are
> not related to the controller state and must be dropped to avoid generating
> false input events.
> 
> Signed-off-by: Frank Praznik <frank.praznik@oh.rr.com>
> ---
>  drivers/hid/hid-sony.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/drivers/hid/hid-sony.c b/drivers/hid/hid-sony.c
> index ed2f008..beb2b02 100644
> --- a/drivers/hid/hid-sony.c
> +++ b/drivers/hid/hid-sony.c
> @@ -1270,6 +1270,17 @@ static int sony_raw_event(struct hid_device *hdev, struct hid_report *report,
>  	 * has to be BYTE_SWAPPED before passing up to joystick interface
>  	 */
>  	if ((sc->quirks & SIXAXIS_CONTROLLER) && rd[0] == 0x01 && size == 49) {
> +		/*
> +		 * When connected via Bluetooth the Sixaxis occasionally sends
> +		 * a report with the second byte 0xff and the rest zeroed.
> +		 *
> +		 * This report does not reflect the actual state of the
> +		 * controller must be ignored to avoid generating false input
> +		 * events.
> +		 */
> +		if (rd[1] == 0xff)
> +			return -EINVAL;
> +

Given that this is not a regression and given the current development 
phase, I've applied it to for-4.3/sony.

Thanks,

-- 
Jiri Kosina
SUSE Labs