On Wed, 29 Mar 2017 00:42:08 -0700
Dmitry Torokhov <dmitry.torok...@gmail.com> wrote:
> The data read from the device is 3 little-endian words, so let's annotate
> them as such and use le16_to_cpu() to convert them to host endianness - it
> might turn out to be a bit more performant, and it expresses the conversion
> more clearly.
Agreed
>
> Signed-off-by: Dmitry Torokhov <dmitry.torok...@gmail.com>
> ---
> drivers/auxdisplay/ht16k33.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/auxdisplay/ht16k33.c b/drivers/auxdisplay/ht16k33.c
> index f66b45b235b0..61af52a7afd5 100644
> --- a/drivers/auxdisplay/ht16k33.c
> +++ b/drivers/auxdisplay/ht16k33.c
> @@ -254,18 +254,19 @@ static bool ht16k33_keypad_scan(struct ht16k33_keypad
> *keypad)
> {
> const unsigned short *keycodes = keypad->dev->keycode;
> u16 new_state[HT16K33_MATRIX_KEYPAD_MAX_COLS];
> - u8 data[HT16K33_MATRIX_KEYPAD_MAX_COLS * 2];
> + __le16 data[HT16K33_MATRIX_KEYPAD_MAX_COLS];
> unsigned long bits_changed;
> int row, col, code;
> bool pressed = false;
>
> - if (i2c_smbus_read_i2c_block_data(keypad->client, 0x40, 6, data) != 6) {
> + if (i2c_smbus_read_i2c_block_data(keypad->client, 0x40,
> + sizeof(data), (u8 *)data) != 6) {
Why use sizeof(data) for length but check against the hardcoded 6?
> dev_err(&keypad->client->dev, "Failed to read key data\n");
> return false;
> }
>
> for (col = 0; col < keypad->cols; col++) {
> - new_state[col] = (data[col * 2 + 1] << 8) | data[col * 2];
> + new_state[col] = le16_to_cpu(data[col]);
nice!
> if (new_state[col])
> pressed = true;
> bits_changed = keypad->last_key_state[col] ^ new_state[col];
Robin
