[U-Boot] [PATCH 2/2] mtd: Get rid of board_mtdparts_default()

[Ladislav Michl]

Hello Boris,

On Wed, Dec 12, 2018 at 10:32:51AM +0100, Boris Brezillon wrote:
> Hi Ladislav,
> 
> On Tue, 11 Dec 2018 23:55:26 +0100
> Ladislav Michl <ladis at linux-mips.org> wrote:
> 
> > Hi Boris,
> > 
> > On Mon, Dec 10, 2018 at 04:38:50PM +0100, Boris Brezillon wrote:
> > > The only implementer of this function has been patched to use
> > > CONFIG_MTD{IDS,PARTS}_DEFAULT instead. Let's get rid of this function
> > > and the associated CONFIG_SYS_MTDPARTS_RUNTIME option.  
> > 
> > the only implementer of this fuction did so for a good reason. What is
> > a motivation to remove it?
> 
> Simplifying the code (see this discussion [1] which led me to send
> this patchset).

Thank you, makes sense.

> > The requirement is to be able to use single u-boot binary on all igep2
> > boards ever produced. These comes with various NAND and OneNAND chips
> > and  I was not able to come with single static partition configuration
> > to support them all.
> 
> That's actually the question I asked Enric in [1]. Can you list all
> the memory organization you have (for NAND and OneNAND chips)? I mean,
> the SPL part size depends on the NAND/OneNAND erase block size, and
> board vendors try to use similar flashes when they source different
> parts (same page size, same block size, ...). Assuming this is the
> case, you should always have the same layout for OneNAND/NAND devices,
> hence my proposal to define those parts statically.

First, thanks to Enric for pinging me, otherwise I would probably miss this
completely.

Now problem is that IGEPv2 comes with quite many configurations, some of
them are even customized, so static configuration is a show stopper
mainly as I do not know what devices are in field.
Another issue is how ubispl code works: It expects struct ubispl_info
filled with (among others) peb_offset of ubi partition. ubispl code counts
in terms of eraseblocks regardless of their size. So we would need to touch
this number when using static mtdparts.

> > Hence runtime detection. That code could be used
> > on all OMAP3 boards as BootROM reads up to first four sectors searching
> > for SPL (MLO).
> 
> Note that, for the nand side of things, you can also automate that using
> a u-boot script:
> 
> nand info; setexpr splsize ${nand_erasesize} * 4; setenv mtdparts mtdparts=omap2-nand:0x${splsize}(SPL),-(UBI)

That seems as a way to go!

> Shouldn't be hard to patch the onenand cmd to also expose writesize,
> erasesize and oobsize.

Side note: I never fully understand why is OneNAND using separate set of
commands.

Could you hold merging your paches until I implement above idea and test
it on a few boards? I know u-boot is now using two months merge window,
which is unfortunate, so I'll try to do it as soon as possible, but I do
not think I'll finish it till end of week.

> Regards,
> 
> Boris
> 
> [1]https://www.mail-archive.com/u-boot@lists.denx.de/msg304933.html