Re: [PATCH 2/8] Fragmentation Avoidance V17: 002_usemap

[Joel Schopp <jschopp@austin.ibm.com>]

>>>@@ -473,6 +491,15 @@ extern struct pglist_data contig_page_da
>>> #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
>>> #error Allocator MAX_ORDER exceeds SECTION_SIZE
>>> #endif
>>>+#if ((SECTION_SIZE_BITS - MAX_ORDER) * BITS_PER_RCLM_TYPE) > 64
>>>+#error free_area_usemap is not big enough
>>>+#endif
>>
>>Every time I look at these patches, I see this #if, and I don't remember
>>what that '64' means.  Can it please get a real name?
>>
> 
> 
> Joel, suggestions?

Oh yeah, blame it on me just because I wrote that bit of code.  How about
#define FREE_AREA_USEMAP_SIZE 64

> 
> 
>>>+/* Usemap initialisation */
>>>+#ifdef CONFIG_SPARSEMEM
>>>+static inline void setup_usemap(struct pglist_data *pgdat,
>>>+				struct zone *zone, unsigned long zonesize) {}
>>>+#endif /* CONFIG_SPARSEMEM */
>>>
>>> struct page;
>>> struct mem_section {
>>>@@ -485,6 +512,7 @@ struct mem_section {
>>> 	 * before using it wrong.
>>> 	 */
>>> 	unsigned long section_mem_map;
>>>+	DECLARE_BITMAP(free_area_usemap,64);
>>> };
>>
>>There's that '64' again!  You need a space after the comma, too.

Ditto.

>>>+ * RCLM_SHIFT is the number of bits that a gfp_mask has to be shifted right
>>>+ * to have just the __GFP_USER and __GFP_KERNRCLM bits. The static check is
>>>+ * made afterwards in case the GFP flags are not updated without updating
>>>+ * this number
>>>+ */
>>>+#define RCLM_SHIFT 19
>>>+#if (__GFP_USER >> RCLM_SHIFT) != RCLM_USER
>>>+#error __GFP_USER not mapping to RCLM_USER
>>>+#endif
>>>+#if (__GFP_KERNRCLM >> RCLM_SHIFT) != RCLM_KERN
>>>+#error __GFP_KERNRCLM not mapping to RCLM_KERN
>>>+#endif
>>
>>Should this really be in page_alloc.c, or should it be close to the
>>RCLM_* definitions?

I had the same first impression, but concluded this was the best place.  The 
compile time checks should keep things from getting out of sync.