[PATCH] nand: omap2: fix building with CONFIG_MTD_NAND_OMAP_BCH=m

On 10/01/2014 12:56 PM, Arnd Bergmann wrote:

On Wednesday 01 October 2014 12:32:09 Roger Quadros wrote:

quoted

With this patch NAND probe on DRA7xx fails like so

[    2.077313] omap-gpmc 50000000.gpmc: GPMC revision 6.0
[    2.083842] nand: device found, Manufacturer ID: 0x2c, Chip ID: 0xca
[    2.090524] nand: Micron MT29F2G16ABAEAWP
[    2.094728] nand: 256MiB, SLC, page size: 2048, OOB size: 64
[    2.100745] nand: using OMAP_ECC_BCH8_CODE_HW ECC scheme
[    2.109764] omap2-nand: probe of omap2-nand.0 failed with error -38

OMAP NAND driver is the only user of the ELM module and we want it
to be usable in all possible configurations when enabled.

I don't understand. Is the BCH driver optional or not?

It is optional. If it is disabled we error out on platforms that have
the ELM IP. But if it is enabled, we don't want to fail probe on such platforms.

The help text says:

          This config enables the ELM hardware engine, which can be used to
          locate and correct errors when using BCH ECC scheme. This offloads
          the cpu from doing ECC error searching and correction. However some
          legacy OMAP families like OMAP2xxx, OMAP3xxx do not have ELM engine
          so they should not enable this config symbol.

which adds more to the confusion. The help text sounds like everything should
work even if ELM is disabled (which contradicts your finding above) but you
must not enable the driver if you are on an older machine (which would
break multiplatform builds).

I understand. We need to update that description. It is no longer the case that
we have to disable ELM engine to work on older platforms.

quoted

Let's pick either one of the below patches instead

http://article.gmane.org/gmane.linux.ports.arm.omap/118488

This doesn't let you have the BCH driver as a module, which seems
wrong.

quoted

http://article.gmane.org/gmane.linux.ports.arm.omap/118847

Looks good, although I think you can simplify this to

config MTD_NAND_OMAP_BCH_BUILD
	def_tristate MTD_NAND_OMAP2 && MTD_NAND_OMAP_BCH

which makes it 'm' if MTD_NAND_OMAP_BCH is set to m.

Right. I will resend this patch with this fixup
and update the Kconfig description as well.

cheers,
-roger