Re: [PATCH 01/17] mtd: prepare to convert of_mtd_parse_partitions to partition parser

[Artem Bityutskiy <dedekind1@gmail.com>]

On Thu, 2011-06-09 at 19:08 +0400, Dmitry Eremin-Solenikov wrote:
> >>>> +#ifdef CONFIG_OF
> >>>> +	if (master->node)
> >>>> +		of_node_get(master->node);
> >>>> +#endif
> >>>
> >>> These ifdefs are not very nice, do you have ideas how to avoid them?
> >>> Ideally, mtdcore should not know or bother about OF things. All
> >>> OF-specific things should be done in ofpart.c...
> >>
> >> I know they aren't nice. OTOH ofpart.c also seems a bit non-logical: one
> >> can have of node in the MTD, but doesn't (strangely) want to compile in
> >> ofpart.c. Of course I can add separate small of-handling functions (to
> >> do OF handling) to mtdcore.c to be replaced by empty functions in the
> >> absence of CONFIG_OF, but this also look like overhead for me.
> >
> > How about turning the "origin" argument into "void *private" and
> > declaring that this is "parser-specific info". It then can become
> > "origin" for the RedBoot parser and the OF node pointer for the ofpart
> > parser?
> 
> And what will happen when ixp4xx (the only user of redboot "exception") 
> will get OF support?

Hmm, may be introducing something like:

/**
 * struct mtd_part_parser_data - used to pass data to MTD partition parsers.
 * @origin: blah blah, RedBoot-specific
 * @of_node: points to the OF node describing the partitions, ofpart-specific
 */
struct mtd_part_parser_data {
	unsigned long origin;
	struct device_node *of_node;
};

And change the current "origin" argument with a
"struct mtd_part_parser_data *data" pointer?

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)