Re: mmotm 2008-12-03-01-50 uploaded (sdhci)

[Randy Dunlap <randy.dunlap@oracle.com>]

Pierre Ossman wrote:
> On Wed, 03 Dec 2008 12:53:24 -0800
> Randy Dunlap <randy.dunlap@oracle.com> wrote:
> 
>> akpm@linux-foundation.org wrote:
>>> The mm-of-the-moment snapshot 2008-12-03-01-50 has been uploaded to
>>>
>>>    http://userweb.kernel.org/~akpm/mmotm/
>>>
>>> It contains the following patches against 2.6.28-rc7:
>>
>> drivers/built-in.o: In function `sdhci_remove_host':
>> (.text+0xd246a): undefined reference to `led_classdev_unregister'
>> drivers/built-in.o: In function `sdhci_add_host':
>> (.text+0xd5378): undefined reference to `led_classdev_register'
>> make[1]: *** [.tmp_vmlinux1] Error 1
>>
>> sdhci.c tries to handle the availability of led_()* without using
>> Kconfig attributes.  In this case, MMC_SDHCI=y and LEDS_CLASS=m,
>> which sdhci.c doesn't handle.
>>
> 
> I'm not sure how to properly solve this. I'm not a big fan of adding a
> specific Kconfig option for each and every optional feature. I'd prefer
> if we get the correct thing<tm> based on just the MMC_SDHCI and
> LEDS_CLASS options.
> 
> -#if defined(CONFIG_LEDS_CLASS) || defined(CONFIG_LEDS_CLASS_MODULE)
> +#if defined(CONFIG_LEDS_CLASS) || (defined(CONFIG_LEDS_CLASS_MODULE)
>  	&& !defined(CONFIG_MMC_SDHCI_MODULE))
> 
> perhaps?

Hi Pierre,

If I change the last line above to
	&& defined(CONFIG_MMC_SDHCI_MODULE)
then it builds OK.  Using !defined still has build errors...
and I think that my small change is also correct.

However, I'd prefer to see this done with Kconfig and then use
empty stubs for the functions that wouldn't be called.


I think that you should audit the other uses of
#if defined(CONFIG_LEDS_CLASS) || defined(CONFIG_LEDS_CLASS_MODULE)

in sdhci.c as well.

Also, in sdhci_add_host(), why is the call to free_irq() inside the
CONFIG_LEDS block:

#if defined(CONFIG_LEDS_CLASS) || defined(CONFIG_LEDS_CLASS_MODULE)
reset:
	sdhci_reset(host, SDHCI_RESET_ALL);
	free_irq(host->irq, host);
#endif


HTH.
~Randy