mm: fix set pageblock migratetype when boot

PATCH In review Xishi Qiu 2 participants 3 messages Jun 26, 2015

Subsystems (2)

Memory management Page allocator

lore

Share & tools

lore

Xishi Qiu

author

2015-06-26 10:09 UTC #0 lore raw

PATCH 1/1 mm: fix set pageblock migratetype when boot

memmap_init_zone()

	...
	if ((z->zone_start_pfn <= pfn)
	    && (pfn < zone_end_pfn(z))
	    && !(pfn & (pageblock_nr_pages - 1)))
		set_pageblock_migratetype(page, MIGRATE_MOVABLE);
	...

If the pfn does not align to pageblock, it will not init the migratetype. So call it for every page, it will takes more time, but it doesn't matter, this function will be called only in boot or hotadd memory.

e.g.

[  223.679446]   node   0: [mem 0x00001000-0x00099fff]
[  223.679449]   node   0: [mem 0x00100000-0xbf78ffff]
[  223.680486]   node   0: [mem 0x100000000-0x27fffffff]

Signed-off-by Xishi Qiu <[email protected]>

---
 mm/page_alloc.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

mm/page_alloc.c +1 -2

index ebffa0e..a1df227 100644
@@ -4210,8 +4210,7 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 		 * pfn out of zone.
 		 */
 		if ((z->zone_start_pfn <= pfn)
-		    && (pfn < zone_end_pfn(z))
-		    && !(pfn & (pageblock_nr_pages - 1)))
+		    && (pfn < zone_end_pfn(z)))
 			set_pageblock_migratetype(page, MIGRATE_MOVABLE);
 
 		INIT_LIST_HEAD(&page->lru);

Signature · 1 line

1.7.1

Mel Gorman

2015-06-26 11:04 UTC #1 lore raw

Quoted 9 lines · Xishi Qiu

memmap_init_zone()
	...
	if ((z->zone_start_pfn <= pfn)
	    && (pfn < zone_end_pfn(z))
	    && !(pfn & (pageblock_nr_pages - 1)))
		set_pageblock_migratetype(page, MIGRATE_MOVABLE);
	...
If the pfn does not align to pageblock, it will not init the migratetype.

What important impact does that have? It should leave a partial pageblock as MIGRATE_UNMOVABLE which is fine by me.

Quoted 3 lines

So call it for every page, it will takes more time, but it doesn't matter, this function will be called only in boot or hotadd memory.

It's a lot of additional overhead to add to memory initialisation. It would need to be for an excellent reason with no alternative solution.

Signature · 8 lines

Mel Gorman
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Xishi Qiu

author

2015-06-26 11:25 UTC #2 lore raw

Quoted 9 lines · Mel Gorman

Quoted 8 lines · Xishi Qiu

memmap_init_zone()

	...
	if ((z->zone_start_pfn <= pfn)
	    && (pfn < zone_end_pfn(z))
	    && !(pfn & (pageblock_nr_pages - 1)))
		set_pageblock_migratetype(page, MIGRATE_MOVABLE);
	...

Quoted 5 lines

Quoted 1 line
If the pfn does not align to pageblock, it will not init the migratetype.
What important impact does that have? It should leave a partial pageblock as MIGRATE_UNMOVABLE which is fine by me.

Hi Mel,

The impact is less, it's OK to ignore it.

Thanks,
Xishi Qiu

Quoted 7 lines

Quoted 3 lines
So call it for every page, it will takes more time, but it doesn't matter, this function will be called only in boot or hotadd memory.
It's a lot of additional overhead to add to memory initialisation. It would need to be for an excellent reason with no alternative solution.