Hi Alexander,
On Fri, Jul 15, 2022 at 09:48:58AM +0200, Alexander Stein wrote:
> Apparently the Vision Components model (VC MIPI OV9281) does not support
> address auto-increment, so probe fails with:
> ov9282 2-0060: chip id mismatch: 9281!=92ff
> Instead do two 1 byte reads and combine the result.
>
> Signed-off-by: Alexander Stein <alexander.stein@xxxxxxxxxxxxxxx>
> ---
> Changes in v2:
> * Fix commit message
> * Add comment about prevented auto-increment
> * Return early if reading ID register failed
> * Reorder ID registers, smaller register number first
>
> drivers/media/i2c/ov9282.c | 15 ++++++++++++++-
> 1 file changed, 14 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/i2c/ov9282.c b/drivers/media/i2c/ov9282.c
> index 263cdffc558f..532791304c3d 100644
> --- a/drivers/media/i2c/ov9282.c
> +++ b/drivers/media/i2c/ov9282.c
> @@ -761,11 +761,24 @@ static int ov9282_set_stream(struct v4l2_subdev *sd, int enable)
> static int ov9282_detect(struct ov9282 *ov9282)
> {
> int ret;
> + u32 id[2];
> u32 val;
>
> - ret = ov9282_read_reg(ov9282, OV9282_REG_ID, 2, &val);
> + /*
> + * Some vendors prevent auto-increment, so each register has to
> + * be read separately
> + */
I suppose it still works for writes I guess? Would be nice to have this in
the comment.
> + ret = ov9282_read_reg(ov9282, OV9282_REG_ID,
> + 1, &id[0]);
Fits on one line. Same below.
> if (ret)
> return ret;
> + ret = ov9282_read_reg(ov9282, OV9282_REG_ID + 1,
> + 1, &id[1]);
> + if (ret)
> + return ret;
> +
> + val = id[1];
> + val |= (id[0] << 8);
>
> if (val != OV9282_ID) {
> dev_err(ov9282->dev, "chip id mismatch: %x!=%x",
--
Sakari Ailus
