Re: [PATCH] mtd: rawnand: driver for Mediatek MT7621 SoC NAND flash controller

[Boris Brezillon]

On Thu, 7 Nov 2019 09:40:07 +0100
Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:

> On Thu, Nov 07, 2019 at 05:35:21PM +1000, gerg@xxxxxxxxxx wrote:
> > From: Greg Ungerer <gerg@xxxxxxxxxx>
> > 
> > Add a driver to support the NAND flash controller of the MediaTek MT7621
> > System-on-Chip device. (This one is the MIPS based parts from Mediatek).
> > 
> > This code is a re-working of the earlier patches for this hardware that
> > have been floating around the internet for years:
> > 
> > https://github.com/ReclaimYourPrivacy/cloak/blob/master/target/linux/ramips/patches-3.18/0045-mtd-add-mt7621-nand-support.patch
> > 
> > This is a much cleaned up version, put in staging to start with.
> > It does still have some problems, mainly that it still uses a lot of the
> > mtd raw nand legacy support.  
> 
> Is that an issue?

Well, yes that's an issue since we decided that all new drivers should
implement ->exec_op() instead of the legacy hooks. But that would be an
issue even if we were merging the driver in staging.

> Why not just put it in the "real" part of the kernel
> then, if those apis are still in use?
> 
> > The driver not only compiles, but it works well on the small range of
> > hardware platforms that it has been used on so far. I have been using
> > for quite a while now, cleaning up as I get time.
> > 
> > So... I am looking for comments on the best approach forward with this.
> > At least in staging it can get some more eyeballs going over it.  
> 
> staging will just nit-pick it to death for coding style issues, it's not
> going to be get any major api changes/cleanups there usually.  I'd
> recommend just merging this to the "real" part of the kernel now if it's
> working for you.

I agree on that point: we should merge this driver directly in the NAND
framework after it's been reworked to implement ->exec_op() instead of
the legacy hooks.