RE: [PATCH v3 24/33] radix-tree: Add radix_tree_split

[Matthew Wilcox]

From: Randy Dunlap [mailto:rdunlap@xxxxxxxxxxxxx]
> > +void radix_tree_iter_replace(struct radix_tree_root *,
> > +		const struct radix_tree_iter *, void **slot, void *item);
> 
> > +int radix_tree_split(struct radix_tree_root *, unsigned long index,
> > +			unsigned new_order);
> 
> and above:
> 
> As indicated in CodingStyle:
> In function prototypes, include parameter names with their data types.
> Although this is not required by the C language, it is preferred in Linux
> because it is a simple way to add valuable information for the reader.

I think the rule here should be a bit more nuanced.  I think it is positively criminal to have an unnamed 'unsigned long' or 'bool' in a function prototype.  But what extra information is communicated by adding 'root' after 'struct radix_tree_root *'?  I know it's a root, you told me that with the structure name!

Obviously sometimes it would be useful, for example if we had a function to move an entry from one radix tree to another, you might want to have 'struct radix_tree_root *old, struct radix_tree_root *new' as two of your parameters.

> >  int radix_tree_join(struct radix_tree_root *, unsigned long index,
> >  			unsigned new_order, void *);
> 
> Yes, the source file already omits some function prototype parameter names,
> so these patches just follow that tradition.  It's weird (to me) though that
> the existing code even mixes this style in one function prototype (see
> immed. above).

I agree that void * should probably be named (as new_entry).  Naming index and new_order is correct.  But again, there's no useful information given by naming root.