On Fri, May 04, 2018 at 01:58:20PM +0200, Boris Brezillon wrote:
On Fri, 4 May 2018 12:18:52 +0200 Thierry Reding thierry.reding@gmail.com wrote:
On Thu, May 03, 2018 at 06:40:05PM +0200, Boris Brezillon wrote:
[...]
return mdp4_lvds_connector->panel ? connector_status_connected : diff --git a/drivers/gpu/drm/msm/dsi/dsi_host.c b/drivers/gpu/drm/msm/dsi/dsi_host.c index 7a03a9489708..fffc80b73966 100644 --- a/drivers/gpu/drm/msm/dsi/dsi_host.c +++ b/drivers/gpu/drm/msm/dsi/dsi_host.c @@ -1881,7 +1881,7 @@ int msm_dsi_host_register(struct mipi_dsi_host *host, bool check_defer) * output */ if (check_defer && msm_host->device_node) {
if (!of_drm_find_panel(msm_host->device_node))
if (IS_ERR(of_drm_find_panel(msm_host->device_node))) if (!of_drm_find_bridge(msm_host->device_node)) return -EPROBE_DEFER;Again, pretty weird stuff going on here, prior to the patch. But I think this needs to be changed to take the -ENODEV into account in the next patch. As it is, this will continue to defer probe even if the panel node is disabled.
Not sure this is a problem. I mean, the code was working before, and we had no way to differentiate the -ENODEV vs -EPROBE_DEFER, which in turn means that, any driver that assumed that the device would appear at some point were just as broken as they are after this patch when the node they're pointing to has its status set to "disabled".
I'm not trying to patch all drivers to take the return code into account, just those that might take advantage of it.
Okay, fair enough.
Thierry