[PATCH] mtd: Introduce CONFIG_MTD_RESERVE_END

[Brian Norris]

On Fri, May 01, 2015 at 01:57:13PM +0200, Jonas Gorski wrote:
> On Thu, Apr 30, 2015 at 7:36 PM, Ben Shelton <ben.shelton at ni.com> wrote:
> > The reason for doing this as a Kconfig option rather than with an additional
> > partition is that we use the same .itb boot image (and kernel arguments) for
> > a series of embedded controllers that have different NAND flash sizes, and we
> > use the '-' command line parameter to give the root partition all the available
> > space after the other partitions.
> 
> Wouldn't it make more sense to make cmdlineparts to recognize if it is
> run on a nand flash that has on-flash BBT enabled, and then reduce the
> SIZE_REMAINING partition's size by the amount of nand_bbt_descr's
> maxblocks * erase block size?

That's a possibility, but that seems like a bit too much hidden policy
to add on top of the cmdlineparts format, which is pretty precise.

A better improvement, if you're dead set on using cmdlineparts rather
than some other better parser, might be to support something like:

  mtdparts=name:-(main),bbt(1M)

So the "remaining space" partition will allow for following partitions.
IOW, you would need to handle this case, which is currently an error:

        /* test if more partitions are following */
        if (*s == ',') {
                if (size == SIZE_REMAINING) {
                        printk(KERN_ERR ERRP "no partitions allowed after a fill-up partition\n");
                        return ERR_PTR(-EINVAL);
                }
		...
	}

> Currently your proposed solution would break if boards have differing
> erase block sizes, or if some have NOR flash, which makes it an option
> for a rather narrow use case IMHO.

Yeah, consider this a NAK for the current patch.

Brian