Hi, Andrew Lunn <and...@lunn.ch> writes:
> On Wed, Jun 08, 2016 at 08:44:52PM -0400, Vivien Didelot wrote: >> The MDIO device probe and remove functions are respectively incrementing >> and decrementing the bus refcount themselves. Since these bus level >> actions are out of the device scope, remove them. > > I agree with the patch. But have you checked the mdio layer is doing > the right thing? If not, we should fix that first. So I added some printing after incrementing/decrementing the refcount in get_device/put_device to track &ps->bus->dev, which name is "0.1". Regardless having this patch or not, the refcount of the 0.1 mii_bus device is 5 before loading the mv88e6xxx module on vf610-zii-dev-rev-b. Below is a portion of dmesg: [ 8.921647] get_device: 400d1000.etherne refcount: 4 [ 8.926225] get_device: 0.1 refcount: 2 [ 8.929561] get_device: mdio-mux refcount: 5 [ 8.934076] get_device: 0.1 refcount: 3 [ 8.937446] get_device: 0.1 refcount: 4 [ 8.940792] put_device: 0.1 refcount: 3 [ 8.944181] libphy: mdio_mux: probed [ 8.947885] mdio_bus 0.1:00: mdio_device_register [ 8.952649] get_device: 0.1:00 refcount: 2 [ 8.956283] get_device: 0.1 refcount: 4 [ 8.959838] get_device: 0.1:00 refcount: 3 [ 8.963991] get_device: 0.1:00 refcount: 4 [ 8.967598] put_device: 0.1:00 refcount: 3 [ 8.971298] get_device: 0.2 refcount: 2 [ 8.974687] get_device: mdio-mux refcount: 7 So it seems like of_ is managing the bus refcount on events such as a new child node (0.1:00). Thanks, Vivien