回复：[PATCH v22 net-next 02/12] net/nebula-matrix: add core driver architecture and HW layer initialization

[Illusion Wang]

(https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260723040110.91410-1-illusion.wang%40nebula-matrix.com)
One comment asks: 
>> +void nbl_hw_remove_leonis(struct nbl_adapter *adapter)
>> +{
>> +	/* All BAR mappings & PCI regions are managed by pcim/devres,
>> +	 * no manual iounmap / release required
>> +	 */
>> +}

>since nbl_hw_remove_leonis is an empty stub function with no
>implementation, is there a valid reason to retain it (along with its prototype
>in nbl_def_h.h), or should we delete both until a subsequent patch adds actual
>cleanup logic to it?

(https://lore.kernel.org/netdev/20260412125728.662867-1-horms@xxxxxxxxxx/)
The other comment questions whether our initialization and teardown paths should
maintain strict symmetry—every successfully initialized layer must have a matching
remove callback invoked both in runtime device removal (nbl_core_remove) and all
early error unwind branches of the init flow.


I'm a bit confused about the two Sashiko review comments on Patch 2 of this patchset:
Which piece of feedback should I follow?