On Thu, May 3, 2018 at 11:40 AM, Boris Brezillon
<boris.brezillon@xxxxxxxxxxx> wrote:
> The device might be described in the device tree but not connected to
> the I2C bus. Update the status property so that the DRM panel logic
> returns -ENODEV when someone tries to get the panel attached to this
> DT node.
>
> Signed-off-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxx>
> ---
> .../gpu/drm/panel/panel-raspberrypi-touchscreen.c | 35 ++++++++++++++++++++++
> 1 file changed, 35 insertions(+)
>
> diff --git a/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c b/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c
> index 2c9c9722734f..b8fcb1acef75 100644
> --- a/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c
> +++ b/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c
> @@ -358,6 +358,39 @@ static const struct drm_panel_funcs rpi_touchscreen_funcs = {
> .get_modes = rpi_touchscreen_get_modes,
> };
>
> +static void rpi_touchscreen_set_status_fail(struct i2c_client *i2c)
> +{
> + struct property *newprop;
> +
> + newprop = kzalloc(sizeof(*newprop), GFP_KERNEL);
> + if (!newprop)
> + return;
> +
> + newprop->name = kstrdup("status", GFP_KERNEL);
> + if (!newprop->name)
> + goto err;
> +
> + newprop->value = kstrdup("fail", GFP_KERNEL);
> + if (!newprop->value)
> + goto err;
> +
> + newprop->length = sizeof("fail");
> +
> + if (of_update_property(i2c->dev.of_node, newprop))
> + goto err;
> +
As I mentioned on irc, can you make this a common DT function.
I'm not sure if it matters that we set status to fail vs. disabled. I
somewhat prefer the latter as we already have other cases and I'd
rather the api not pass a string in. I can't think of any reason to
distinguish the difference between fail and disabled.
> + /* We intentionally leak the memory we allocate here, because the new
> + * OF property might live longer than the underlying dev, so no way
> + * we can use devm_kzalloc() here.
> + */
> + return;
> +
> +err:
> + kfree(newprop->value);
> + kfree(newprop->name);
> + kfree(newprop);
> +}
> +
> static int rpi_touchscreen_probe(struct i2c_client *i2c,
> const struct i2c_device_id *id)
> {
> @@ -382,6 +415,7 @@ static int rpi_touchscreen_probe(struct i2c_client *i2c,
>
> ver = rpi_touchscreen_i2c_read(ts, REG_ID);
> if (ver < 0) {
> + rpi_touchscreen_set_status_fail(i2c);
I've thought some more about this and I still think this should be
handled in the driver core or i2c core.
The reason is simple. I think the state of the system should be the
same after this as if you booted with 'status = "disabled"' for this
node. And that means the device should be removed completely because
we don't create struct device's for disabled nodes.
Rob
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
