Re: [Bugme-new] [Bug 2019] New: Bug from the mm subsystem involving X (fwd)

[Dave Hansen <haveblue@us.ibm.com>]

On Thu, 2004-02-05 at 23:17, Martin J. Bligh wrote:
> +#ifdef CONFIG_NUMA
> +	#ifdef CONFIG_X86_NUMAQ
> +		#include <asm/numaq.h>
> +	#else	/* summit or generic arch */
> +		#include <asm/srat.h>
> +	#endif
> +#else /* !CONFIG_NUMA */
> +	#define get_memcfg_numa get_memcfg_numa_flat
> +	#define get_zholes_size(n) (0)
> +#endif /* CONFIG_NUMA */

We ran into a bug with #ifdefs like this before.  It was fixed in some
of the code that you're trying to remove.

It's not safe to assume that NUMA && !NUMAQ means SUMMIT.  Remember the
linking errors we got when we turned CONFIG_NUMA on with the regular PC
config?  The generic arch wasn't a problem because it sets
CONFIG_X86_SUMMIT and compiles in the summit code, but the regular PC
code doesn't.  

Also, I don't think we need the #ifdef CONFIG_NUMA around the whole
block.  How about something like this?

#ifdef CONFIG_X86_NUMAQ
	#include <asm/numaq.h>
#elif CONFIG_X86_SUMMIT
	#include <asm/srat.h>
#else
	#define get_memcfg_numa get_memcfg_numa_flat
	#define get_zholes_size(n) (0)
#endif /* CONFIG_NUMA */


> +static inline int pfn_to_nid(unsigned long pfn)
> +{
> +#ifdef CONFIG_NUMA
> +	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
> +#else
> +	return 0;
> +#endif
> +}

Looks like somebody pasted that in from a macro. "(pfn)" :)

--dave