On Thu, Jan 16, 2025 at 05:01:03PM +0100, Maxime Ripard wrote:

Hi Dmitry,

On Wed, Jan 15, 2025 at 12:21:39PM +0200, Dmitry Baryshkov wrote:

quoted

On Tue, Jan 14, 2025 at 10:38:53PM +0100, Sasha Finkelstein via B4 Relay wrote:

quoted

+static int adp_dsi_host_attach(struct mipi_dsi_host *host,
+			       struct mipi_dsi_device *dev)
+{
+	struct adp_mipi_drv_private *adp = mipi_to_adp(host);
+	struct drm_bridge *next;
+
+	next = devm_drm_of_get_bridge(adp->dsi.dev, adp->dsi.dev->of_node, 1, 0);

select DRM_PANEL_BRIDGE for this API to be available.
Also there is an issue of lifetimes. The bridge will be destroyed only
when MIPI driver is unbound, however the panel can be attached and
detached several times.

Can you expand on that one a bit more? AFAIK, it's always been unsafe
and we don't have a good mitigation at the moment, so it's unclear to me
what change you want here.

I was thinking about using drmm_ here, as the DRM device is also created
and destroyed each time. But I might be mistaken here.

-- 
With best wishes
Dmitry