Re: [RFC PATCH 0/7] mtd: nand: Abstract away the NAND interface type

[Brian Norris <computersforpeace@gmail.com>]

On Thu, Sep 22, 2016 at 12:12:53PM +0200, Boris Brezillon wrote:
> Hi,
> 
> This series is aiming at providing a generic NAND layer to share code
> between different NAND based devices.
> 
> We currently have 3 different interfaces to interact with NANDs:
> - Raw NANDs
> - OneNANDs
> - SPI NANDs
> 
> Apart from the way these NAND devices are accessed they have a lot
> in common, like the way the memory is organized, or their constraints.
> This is usually a good sign that some work should be done to factorize
> the code.
> 
> This work has been started by Peter who wanted to re-use the BBT
> code for its SPI-NAND driver. But I think we can push it further
> other stuff (the software ECC implementation, or the way offsets are
> converted to block/page number).
> 
> Before I continue in this direction, I'd like to get some feedback
> from Peter and those who reviewed his initial submission (Brian,
> Ezequiel) [1], or anyone who is interested in this topic.

My eyes are bleeding for two reasons:

(1) You haven't used any of git's nice rename detection for your patches
(2) The 'rawnand' naming seems a bit much for me. If we really need to
    reorganize everything, keeping the name shorter, like just 'raw'
    might be fine -- at least wherever we're already namespaces as
    "nand". e.g., 'drivers/mtd/nand/raw/' and 'mtd: nand: raw:
    commit subject', but you might still keep 'rawnand.h' (unless you
    want to move things into an include/linux/mtd/nand/ subdirectory).

You'll also need to update MAINTAINERS for the header pattern.

Haven't looked too closely at the code yet.

Brian