Hi Laurent,
On 04/01/2021 07:39, Laurent Pinchart wrote:
> Add a new V4L2 subdev driver for the OmniVision OV10633 and OV10635
> camera sensors. The work is based on the driver from the TI BSP, itself
> based on original work by Phil Edworthy posted to the linux-media
> mailing list ([1]). This version of the code is a large rewrite of many
> parts.
<snip>
> +static int ov1063x_tpg_setup(struct ov1063x_priv *priv, struct v4l2_ctrl *ctrl)
> +{
> + const struct ov1063x_tpg_config *cfg;
> + int ret = 0;
> +
> + if (!ctrl->val)
> + return ov1063x_write_array(priv, ov1063x_regs_colorbar_disable,
> + ARRAY_SIZE(ov1063x_regs_colorbar_disable));
> +
> + if (!ctrl->cur.val) {
> + /*
> + * Only write the full settings when the test pattern was
> + * disabled, not when we're just changing the test pattern type.
> + */
> + ret = ov1063x_write_array(priv, ov1063x_regs_colorbar_enable,
> + ARRAY_SIZE(ov1063x_regs_colorbar_enable));
> + if (ret < 0)
> + return ret;
> + }
> +
> + cfg = &ov1063x_tpg_configs[ctrl->val - 1];
> +
> + /* TODO: Add support for the moving bar overlay. */
> + ov1063x_write(priv, OV1063X_ISP_CTRL3D, cfg->ctrl3d, &ret);
> + ov1063x_write(priv, OV1063X_ISP_CTRL3E, cfg->ctrl3e, &ret);
> +
> + return ret;
> +}
I didn't go through the code in detail, but looks like the color bar
settings from ov1063x_regs_colorbar_enable are lost if capture is
stopped and started again. The color bar itself stays, but it's
distorted due to all the ISP processing being done.
Tomi
