Re: [PATCH v4 18/29] i2c: mux: Create missing devlink between mux and adapter physical device

[Herve Codina <herve.codina@bootlin.com>]

Hi Andi,

On Thu, 30 Oct 2025 16:23:24 +0100
Andi Shyti <andi.shyti@kernel.org> wrote:

> Hi Herve,
> 
> ...
> 
> > When an i2c mux is involved in an i2c path, the struct dev topology is
> > the following:  
> 
> supernitpick: I'd leave blank line here.

Will be added.

> 
> >     +----------------+                +-------------------+
> >     | i2c controller |                |      i2c mux      |
> >     |     device     |                |      device       |
> >     |       ^        |                |                   |
> >     |       |        |                |                   |
> >     |  dev's parent  |                |                   |
> >     |       |        |                |                   |
> >     |   i2c adapter  |                | i2c adapter chanX |
> >     |     device  <---- dev's parent ------  device       |
> >     |   (no driver)  |                |    (no driver)    |
> >     +----------------+                +-------------------+
> >   
> 
> ...
> 
> > No relationship exists between the i2c mux device itself and the i2c
> > controller device (physical device) in order to have the i2c mux device
> > calling i2c_del_adapter() to remove its downtream adapters and so,  
> 
> /downtream/downstream/

Will be fixed

> 
> > release references taken to the upstream adapter.  
> 
> ...
> 
> > +	/*
> > +	 * There is no relationship set between the mux device and the physical
> > +	 * device handling the parent adapter. Create this missing relationship
> > +	 * in order to remove the i2c mux device (consumer) and so the dowstream
> > +	 * channel adapters before removing the physical device (supplier) which
> > +	 * handles the i2c mux upstream adapter.
> > +	 */
> > +	parent_physdev = i2c_get_adapter_physdev(parent);
> > +	if (!parent_physdev) {
> > +		dev_err(muxc->dev, "failed to get the parent physical device\n");
> > +		ret = -EINVAL;  
> 
> -ENODEV?

Yes, -ENODEV makes sense here. Will be changed in the next iteration.

> 
> > +		goto err_free_priv;
> > +	}
> > +	dl = device_link_add(muxc->dev, parent_physdev, DL_FLAG_AUTOREMOVE_CONSUMER);  
> 
> Not to call twice put_device, I would add it once here and then
> check for !dl.

As Andy already mentioned, we cannot do that. Indeed, dev_name(parent_physdev)
is called in the error path and so the device reference has to be kept.

> 
> > +	if (!dl) {
> > +		dev_err(muxc->dev, "failed to create device link to %s\n",
> > +			dev_name(parent_physdev));
> > +		put_device(parent_physdev);
> > +		ret = -EINVAL;  
> 
> same here, should this be -ENODEV?

For this one, I am not so sure.

The failure is related to the device link creation and probably due to some
devlink invalid internal flags or state instead of a missing device.

That's said, if you really want the -ENODEV here, let me know.

Best regards,
Hervé