Re: sparsemem usage

From: Andy Whitcroft
Date: Wed Aug 02 2006 - 10:25:19 EST


moreau francis wrote:
My board has a really weird mem mapping.

MEM1: 0xc000 0000 - 32 Mo
MEM2: 0xd000 0000 - 8 Mo
MEM3: 0xd800 0000 - 128 Ko

MEM3 has interesting properties, such as speed and security,
and I really need to use it.

I think that sparsemem can deal with such mapping. But I
encounter an issue when choosing the section bit size. I choose
SECTION_SIZE_BITS = 17. Therefore the section size is
equal to the smallest size of my memories. But I get a
compilation error which is due to this:

#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
#error Allocator MAX_ORDER exceeds SECTION_SIZE
#endif

I'm not sure to understand why there's such check. To fix this
I should change MAX_ORDER to 6.

Is it the only way to fix that ?

The memory allocator buddy location algorithm has an implicit assumption that the memory map will be contigious and valid out to MAX_ORDER. ie that we can do relative arithmetic on a page* for a page to find its buddy at all times. The allocator never looks outside a MAX_ORDER block, aligned to MAX_ORDER in physical pages. SPARSEMEM's implementation by it nature breaks up the mem_map at the section size. Thus for the buddy to work a section must be >= MAX_ORDER in size to maintain the contiguity constraint.

However, just because you have a small memory block in your memory map doesn't mean that the sparsemem section size needs to be that small to match. If there is any valid memory in any section that section will be instantiated and the valid memory marked within it, any invalid memory is marked reserved. The section size bounds the amount of internal fragmentation we can have in the mem_map. SPARSEMEM as its name suggests wins biggest when memory is very sparsly populate. If I am reading correctly your memory is actually contigious.

-apw

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/