Hi Steve, Btw. I think probably a smaller list of recipients would be just fine on the next version. On Fri, Feb 28, 2020 at 10:16:06AM -0800, Steve Longerbeam wrote: > Hi Sakari, > > On 2/25/20 7:07 AM, Sakari Ailus wrote: > > Hi Steve, > > > > On Sat, Feb 15, 2020 at 11:41:21AM -0800, Steve Longerbeam wrote: > > > Instead of allocating a notifier in v4l2_async_register_fwnode_subdev(), > > > have the caller provide one. This allows the caller to implement > > > notifier ops (bind, unbind). > > > > > > The caller is now responsible for first initializing its notifier with a > > > call to v4l2_async_notifier_init(). > > > > > > Signed-off-by: Steve Longerbeam <slongerbeam@xxxxxxxxx> > > Instead of improving v4l2_async_register_fwnode_subdev(), could you convert > > the users (IMX driver in this case) to call the preferred APIs instead? As > > the lines below show, v4l2_async_register_fwnode_subdev() has only two > > users left --- the other one of which is the IMX driver. After converting > > these two, we could just remove this API. > > > > See e.g. drivers/media/pci/intel/ipu3/ipu3-cio2.c and > > drivers/media/platform/omap3isp/isp.c for examples. > > Shouldn't v4l2_async_notifier_add_fwnode_remote_subdev() check for the > availability of the remote before adding it to the notifier's asd list, as > in: > > diff --git a/drivers/media/v4l2-core/v4l2-async.c > b/drivers/media/v4l2-core/v4l2-async.c > index 8bde33c21ce4..b48ed68c6c6c 100644 > --- a/drivers/media/v4l2-core/v4l2-async.c > +++ b/drivers/media/v4l2-core/v4l2-async.c > @@ -615,7 +615,7 @@ v4l2_async_notifier_add_fwnode_remote_subdev(struct > v4l2_async_notifier *notif, > int ret; > > remote = fwnode_graph_get_remote_port_parent(endpoint); > - if (!remote) > + if (!remote || !fwnode_device_is_available(remote)) > return -ENOTCONN; > > asd->match_type = V4L2_ASYNC_MATCH_FWNODE; > > > Otherwise we are back to the problem that the notifier will never complete > because the remote's driver is not probed. fwnode_graph_get_endpoint_by_id() only gives you endpoints that belong to an enabled device (unless requested otherwise). So the there's need to check the same in v4l2-fwnode.c. -- Regards, Sakari Ailus sakari.ailus@xxxxxxxxxxxxxxx
