Re: [Question] should {read,write}_page hooks moved to nand_controller_ops ?

[Boris Brezillon <bbrezillon@xxxxxxxxxx>]

Hi Masahiro,

On Fri, 25 Jan 2019 17:23:28 +0900
Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx> wrote:

> Hello.
> 
> 
> Many hooks in nand_ecc_ctrl, such as read_page, write_page, etc.
> look controller-specific instead of per-chip.
> 
> Will they be moved to nand_controller_op or somewhere
> in the future?

Not exactly :-). I don't know if I should say it, but Miquel is working
on a generic NAND ECC engine framework which we'll try to re-use at the
raw NAND layer. The idea is to dissociate the ECC engine concept from
the NAND controller one. I imagine something like this:

	ecc_engine->prepare_page_io_req(nand, ioreq);
	nand_controller->do_page_io(nand, ioreq);
	ecc_engine->finish_page_io_req(nand, ioreq);

If we go for this approach, the nand controller should only have one
hook: ->do_page_io(), and this hook would indeed be placed in
nand_controller_ops. Once we have that in place, we can deprecate the
ecc field in nand_chip (by moving it to nand_legacy).

Of course, this implies making the raw NAND framework use the generic
NAND layer (which I started to work on here [1]).

> 
> Boris and Miquel are doing a great job
> for the NAND framework improvement.

Thanks.

> So, maybe it is already in their TODO list.

It is, but, given the amount of work needed, it's unlikely to pop up in
a near future.

Regards,

Boris

[1]https://github.com/bbrezillon/linux/commits/nand/cleanup

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/