Messages in this thread Patch in this message | | | From | Mel Gorman <> | Subject | [PATCH 6/9] Allow HugeTLB allocations to use ZONE_EASYRCLM | Date | Thu, 26 Jan 2006 18:45:05 +0000 (GMT) |
| |
On ppc64 at least, a HugeTLB is the same size as a memory section. Hence, it causes no fragmentation that is worth caring about because a section can still be offlined.
Once HugeTLB is allowed to use ZONE_EASYRCLM, the size of the zone becomes a "soft" area where HugeTLB allocations may be satisified. For example, take a situation where a system administrator is not willing to reserve HugeTLB pages at boot time. In this case, he can use kernelcore to size the EasyRclm zone which is still usable by normal processes. If a job starts that need HugeTLB pages, one could dd a file the size of physical memory, delete it and have a good chance of getting a number of HugeTLB pages. To get all of EasyRclm as HugeTLB pages, the ability to drain per-cpu pages is required.
Signed-off-by: Mel Gorman <mel@csn.ul.ie> diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.16-rc1-mm3-106_zonechoose/mm/hugetlb.c linux-2.6.16-rc1-mm3-107_hugetlb_use_easyrclm/mm/hugetlb.c --- linux-2.6.16-rc1-mm3-106_zonechoose/mm/hugetlb.c 2006-01-17 07:44:47.000000000 +0000 +++ linux-2.6.16-rc1-mm3-107_hugetlb_use_easyrclm/mm/hugetlb.c 2006-01-26 18:13:43.000000000 +0000 @@ -49,7 +49,7 @@ static struct page *dequeue_huge_page(st for (z = zonelist->zones; *z; z++) { nid = (*z)->zone_pgdat->node_id; - if (cpuset_zone_allowed(*z, GFP_HIGHUSER) && + if (cpuset_zone_allowed(*z, GFP_RCLMUSER) && !list_empty(&hugepage_freelists[nid])) break; } @@ -68,7 +68,7 @@ static struct page *alloc_fresh_huge_pag { static int nid = 0; struct page *page; - page = alloc_pages_node(nid, GFP_HIGHUSER|__GFP_COMP|__GFP_NOWARN, + page = alloc_pages_node(nid, GFP_RCLMUSER|__GFP_COMP|__GFP_NOWARN, HUGETLB_PAGE_ORDER); nid = (nid + 1) % num_online_nodes(); if (page) { diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.16-rc1-mm3-106_zonechoose/mm/mempolicy.c linux-2.6.16-rc1-mm3-107_hugetlb_use_easyrclm/mm/mempolicy.c --- linux-2.6.16-rc1-mm3-106_zonechoose/mm/mempolicy.c 2006-01-25 13:42:46.000000000 +0000 +++ linux-2.6.16-rc1-mm3-107_hugetlb_use_easyrclm/mm/mempolicy.c 2006-01-26 18:13:43.000000000 +0000 @@ -1169,7 +1169,7 @@ struct zonelist *huge_zonelist(struct vm unsigned nid; nid = interleave_nid(pol, vma, addr, HPAGE_SHIFT); - return NODE_DATA(nid)->node_zonelists + gfp_zone(GFP_HIGHUSER); + return NODE_DATA(nid)->node_zonelists + gfp_zone(GFP_RCLMUSER); } return zonelist_policy(GFP_HIGHUSER, pol); } - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |