Re: [PATCH v2 2/4] mtd: nand: Allow MTD_NAND_BRCMNAND to be selected for ARM64

[Florian Fainelli <f.fainelli@gmail.com>]

On 16/10/15 12:25, Arnd Bergmann wrote:
> On Friday 16 October 2015 09:54:45 Brian Norris wrote:
>>> MTD_NAND_BRCMNAND uses the generic nand functions, but is currently allowed
>>> to be built without CONFIG_MTD_NAND, which results in a link error:
>>>
>>> drivers/built-in.o: In function `brcmnand_remove':
>>> coresight-replicator.c:(.text+0x17ae6c): undefined reference to `nand_release'
>>> drivers/built-in.o: In function `brcmnand_probe':
>>> coresight-replicator.c:(.text+0x17d4b4): undefined reference to `nand_scan_ident'
>>> coresight-replicator.c:(.text+0x17d948): undefined reference to `nand_scan_tail'
>>
>> How did you get this? MTD_NAND_BRCMNAND is surrounded in the
>> 'if MTD_NAND' block, which implicitly generates a MTD_NAND dependency.
>> And I can confirm that in menuconfig, I see this when I disable MTD_NAND
>> and search for BRCMNAND:
>>
>>    Symbol: MTD_NAND_BRCMNAND [=n]
>>    Type  : tristate
>>    Prompt: Broadcom STB NAND controller
>>      Location:
>>        -> Device Drivers
>>          -> Memory Technology Device (MTD) support (MTD [=y])
>>    (1)     -> NAND Device Support (MTD_NAND [=n])
>>      Defined at drivers/mtd/nand/Kconfig:394
>>      Depends on: MTD [=y] && MTD_NAND [=n] && (ARM [=y] || ARM64 || MIPS)
> 
> Sorry, can't reproduce it any more. My patch is dated June 4, so it was
> probably broken then but got fixed since. I normally try to verify that
> the patches are still needed before I send them, but this time I only
> saw the current discussion and remembered something vague about it
> and sent what I had in my backlog of the randconfig-fixes series.

It may have been a problem before this patch:

d80d942bcc8e1555a76774d20be9800cfef2d415 ("ARM: BCM: Do not select
CONFIG_MTD_NAND_BRCMNAND")
-- 
Florian