hugetlb: add allocate function for hugepage migration
authorNaoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Wed, 8 Sep 2010 01:19:33 +0000 (10:19 +0900)
committerAndi Kleen <ak@linux.intel.com>
Fri, 8 Oct 2010 07:32:44 +0000 (09:32 +0200)
We can't use existing hugepage allocation functions to allocate hugepage
for page migration, because page migration can happen asynchronously with
the running processes and page migration users should call the allocation
function with physical addresses (not virtual addresses) as arguments.

ChangeLog since v3:
- unify alloc_buddy_huge_page() and alloc_buddy_huge_page_node()

ChangeLog since v2:
- remove unnecessary get/put_mems_allowed() (thanks to David Rientjes)

ChangeLog since v1:
- add comment on top of alloc_huge_page_no_vma()

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
include/linux/hugetlb.h
mm/hugetlb.c

index f479700df61b186118d607a0db86b1e47ffa8294..0b73c536afd2d3002c533d71457d792413d666e8 100644 (file)
@@ -228,6 +228,8 @@ struct huge_bootmem_page {
        struct hstate *hstate;
 };
 
+struct page *alloc_huge_page_node(struct hstate *h, int nid);
+
 /* arch callback */
 int __init alloc_bootmem_huge_page(struct hstate *h);
 
@@ -303,6 +305,7 @@ static inline struct hstate *page_hstate(struct page *page)
 
 #else
 struct hstate {};
+#define alloc_huge_page_node(h, nid) NULL
 #define alloc_bootmem_huge_page(h) NULL
 #define hstate_file(f) NULL
 #define hstate_vma(v) NULL
index bd031a4c738e7b1083bb18453109cfb2a62e7087..83fa0c3b6e2b14b5dc520cc58f607bf5c95c44ef 100644 (file)
@@ -466,11 +466,23 @@ static void enqueue_huge_page(struct hstate *h, struct page *page)
        h->free_huge_pages_node[nid]++;
 }
 
+static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
+{
+       struct page *page;
+
+       if (list_empty(&h->hugepage_freelists[nid]))
+               return NULL;
+       page = list_entry(h->hugepage_freelists[nid].next, struct page, lru);
+       list_del(&page->lru);
+       h->free_huge_pages--;
+       h->free_huge_pages_node[nid]--;
+       return page;
+}
+
 static struct page *dequeue_huge_page_vma(struct hstate *h,
                                struct vm_area_struct *vma,
                                unsigned long address, int avoid_reserve)
 {
-       int nid;
        struct page *page = NULL;
        struct mempolicy *mpol;
        nodemask_t *nodemask;
@@ -496,19 +508,13 @@ static struct page *dequeue_huge_page_vma(struct hstate *h,
 
        for_each_zone_zonelist_nodemask(zone, z, zonelist,
                                                MAX_NR_ZONES - 1, nodemask) {
-               nid = zone_to_nid(zone);
-               if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask) &&
-                   !list_empty(&h->hugepage_freelists[nid])) {
-                       page = list_entry(h->hugepage_freelists[nid].next,
-                                         struct page, lru);
-                       list_del(&page->lru);
-                       h->free_huge_pages--;
-                       h->free_huge_pages_node[nid]--;
-
-                       if (!avoid_reserve)
-                               decrement_hugepage_resv_vma(h, vma);
-
-                       break;
+               if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask)) {
+                       page = dequeue_huge_page_node(h, zone_to_nid(zone));
+                       if (page) {
+                               if (!avoid_reserve)
+                                       decrement_hugepage_resv_vma(h, vma);
+                               break;
+                       }
                }
        }
 err:
@@ -770,11 +776,10 @@ static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
        return ret;
 }
 
-static struct page *alloc_buddy_huge_page(struct hstate *h,
-                       struct vm_area_struct *vma, unsigned long address)
+static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
 {
        struct page *page;
-       unsigned int nid;
+       unsigned int r_nid;
 
        if (h->order >= MAX_ORDER)
                return NULL;
@@ -812,9 +817,14 @@ static struct page *alloc_buddy_huge_page(struct hstate *h,
        }
        spin_unlock(&hugetlb_lock);
 
-       page = alloc_pages(htlb_alloc_mask|__GFP_COMP|
-                                       __GFP_REPEAT|__GFP_NOWARN,
-                                       huge_page_order(h));
+       if (nid == NUMA_NO_NODE)
+               page = alloc_pages(htlb_alloc_mask|__GFP_COMP|
+                                  __GFP_REPEAT|__GFP_NOWARN,
+                                  huge_page_order(h));
+       else
+               page = alloc_pages_exact_node(nid,
+                       htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
+                       __GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
 
        if (page && arch_prepare_hugepage(page)) {
                __free_pages(page, huge_page_order(h));
@@ -829,13 +839,13 @@ static struct page *alloc_buddy_huge_page(struct hstate *h,
                 */
                put_page_testzero(page);
                VM_BUG_ON(page_count(page));
-               nid = page_to_nid(page);
+               r_nid = page_to_nid(page);
                set_compound_page_dtor(page, free_huge_page);
                /*
                 * We incremented the global counters already
                 */
-               h->nr_huge_pages_node[nid]++;
-               h->surplus_huge_pages_node[nid]++;
+               h->nr_huge_pages_node[r_nid]++;
+               h->surplus_huge_pages_node[r_nid]++;
                __count_vm_event(HTLB_BUDDY_PGALLOC);
        } else {
                h->nr_huge_pages--;
@@ -847,6 +857,25 @@ static struct page *alloc_buddy_huge_page(struct hstate *h,
        return page;
 }
 
+/*
+ * This allocation function is useful in the context where vma is irrelevant.
+ * E.g. soft-offlining uses this function because it only cares physical
+ * address of error page.
+ */
+struct page *alloc_huge_page_node(struct hstate *h, int nid)
+{
+       struct page *page;
+
+       spin_lock(&hugetlb_lock);
+       page = dequeue_huge_page_node(h, nid);
+       spin_unlock(&hugetlb_lock);
+
+       if (!page)
+               page = alloc_buddy_huge_page(h, nid);
+
+       return page;
+}
+
 /*
  * Increase the hugetlb pool such that it can accomodate a reservation
  * of size 'delta'.
@@ -871,7 +900,7 @@ static int gather_surplus_pages(struct hstate *h, int delta)
 retry:
        spin_unlock(&hugetlb_lock);
        for (i = 0; i < needed; i++) {
-               page = alloc_buddy_huge_page(h, NULL, 0);
+               page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
                if (!page) {
                        /*
                         * We were not able to allocate enough pages to
@@ -1052,7 +1081,7 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
        spin_unlock(&hugetlb_lock);
 
        if (!page) {
-               page = alloc_buddy_huge_page(h, vma, addr);
+               page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
                if (!page) {
                        hugetlb_put_quota(inode->i_mapping, chg);
                        return ERR_PTR(-VM_FAULT_SIGBUS);