Re: [PATCH v23 08/13] mfd: core: Add firmware-node support to MFD cells

On Tue, 21 Jul 2026 14:48:46 +0200, Shivendra Pratap
<[email protected]> said:
>>
>>> +		struct device_node *named_np;
>>> +
>>> +		fwnode = device_get_named_child_node(parent, cell->named_fwnode);
>>> +		if (!fwnode) {
>>> +			ret = -ENODEV;
>>> +			goto fail_alias;
>>> +		}
>>> +
>>> +		named_np = to_of_node(fwnode);
>>> +		if (named_np) {
>>> +			ret = mfd_claim_of_node_to_dev(pdev, named_np);
>>> +			fwnode_handle_put(fwnode);
>>> +			if (ret == -EAGAIN)
>>> +				ret = -EBUSY;
>>> +			if (ret)
>>> +				goto fail_alias;
>>> +		} else {
>>> +			ret = mfd_claim_named_fwnode(pdev, fwnode);
>>> +			if (ret) {
>>> +				fwnode_handle_put(fwnode);
>>> +				if (ret == -EAGAIN)
>>> +					ret = -EBUSY;
>>> +				goto fail_alias;
>>> +			}
>>> +			device_set_node(&pdev->dev, fwnode);
>>> +			fwnode_handle_put(fwnode);
>>> +		}
>>
>> What is the reason to have this split into OF and fwnode functions? I don't
>> really see why you would need to do it twice, isn't fwnode sufficient for the
>> OF node case too?
>
> We do a duplicate check here before adding the named_fwnode and split it
> for two reasons:
> 1. of_node list pre-exists. We add a new list of fwnode. of_node based
> duplicate check go wrong, if its checked with the fwnode list.
> 2. of_node need to go to its respective list as fwnode list, release
> path, is different for us.
>

I see. This could use a unification to fwnodes as well but that's out of scope
for this series.

Bart