Re: [PATCH 2/3] HID: i2c-hid: Read wMaxInputLength of input report instead of bufsize

[Benjamin Tissoires <benjamin.tissoires@redhat.com>]

On Feb 23 2015 or thereabouts, Mika Westerberg wrote:
> Wacom digitizer (WCOM0008) on Lenovo Thinkpad 10 seems to allow reading of
> subsequent reports (or part of them) if we read more than wMaxInputLength
> of data at the time. This data is not always aligned so that the next
> report would start right after another.
> 
> For example we might get following sequence:
> 
>  i2c_hid i2c-WCOM0008:00: input: 0a 00 02 21 20 17 ad 22 11 03
>  i2c_hid i2c-WCOM0008:00: input: ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03 0a 00 02 21 20 17
>                                  ad 22 11 03
>  i2c_hid i2c-WCOM0008:00: input: 02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17 ad 22 11 03 0a 00
>                                  02 21 20 17
>  i2c_hid i2c-WCOM0008:00: i2c_hid_get_input: incomplete report (76/8450)
> 
> The bufsize is 76 and wMaxInputLength is 10. In above example the first
> read gets right amount of data. The second and third reads get full bufsize
> (76 bytes) but the report is missing a start already. This causes the
> driver to reject the report because we got less than was expected by the
> report length (0x2102 = 8450).
> 
> If we read only wMaxInputLength at the time this does not happen at all and
> the digitizer works fine.
> 
> Based on this change the driver to read wMaxInputLength bytes instead of
> bufsize if the value looks sane.
> 
> Reported-by: Sébastien Bourdeauducq <sb@m-labs.hk>
> Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
> ---

We can actually drop this one. Jiri already applied a patch sent last
week which does the exact same thing:
https://patchwork.kernel.org/patch/5857521/ (should appear shortly in
Jiri's tree I guess).

But thanks for confirming that it was really needed for other devices.

Cheers,
Benjamin

>  drivers/hid/i2c-hid/i2c-hid.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/drivers/hid/i2c-hid/i2c-hid.c b/drivers/hid/i2c-hid/i2c-hid.c
> index 8f1dfc5c5d9c..404ccde49acd 100644
> --- a/drivers/hid/i2c-hid/i2c-hid.c
> +++ b/drivers/hid/i2c-hid/i2c-hid.c
> @@ -369,9 +369,17 @@ static int i2c_hid_hwreset(struct i2c_client *client)
>  
>  static void i2c_hid_get_input(struct i2c_hid *ihid)
>  {
> +	unsigned max_input_length = le16_to_cpu(ihid->hdesc.wMaxInputLength);
>  	int ret, ret_size;
>  	int size = ihid->bufsize;
>  
> +	/*
> +	 * Take the input length from HID descriptor if it is available and
> +	 * looks reasonable.
> +	 */
> +	if (max_input_length > 0)
> +		size = min_t(unsigned, ihid->bufsize, max_input_length);
> +
>  	ret = i2c_master_recv(ihid->client, ihid->inbuf, size);
>  	if (ret != size) {
>  		if (ret < 0)
> -- 
> 2.1.4
>