Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 5ced66c901f1cf0b684feb15c2cd8b126e263d07

Authored by Andi Kleen
Committed by Linus Torvalds
1 parent a343787016
Exists in master and in 7 other branches imx_3.0.35_4.1.0, imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga, smarc-l5.0.0_1.0.0-ga

hugetlb: abstract numa round robin selection 

Need this as a separate function for a future patch.

No behaviour change.

Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 1 changed file with 22 additions and 15 deletions Side-by-side Diff

... ... @@ -565,6 +565,27 @@
565 565 return page;
566 566 }
567 567  
  568 +/*
  569 + * Use a helper variable to find the next node and then
  570 + * copy it back to hugetlb_next_nid afterwards:
  571 + * otherwise there's a window in which a racer might
  572 + * pass invalid nid MAX_NUMNODES to alloc_pages_node.
  573 + * But we don't need to use a spin_lock here: it really
  574 + * doesn't matter if occasionally a racer chooses the
  575 + * same nid as we do. Move nid forward in the mask even
  576 + * if we just successfully allocated a hugepage so that
  577 + * the next caller gets hugepages on the next node.
  578 + */
  579 +static int hstate_next_node(struct hstate *h)
  580 +{
  581 + int next_nid;
  582 + next_nid = next_node(h->hugetlb_next_nid, node_online_map);
  583 + if (next_nid == MAX_NUMNODES)
  584 + next_nid = first_node(node_online_map);
  585 + h->hugetlb_next_nid = next_nid;
  586 + return next_nid;
  587 +}
  588 +
568 589 static int alloc_fresh_huge_page(struct hstate *h)
569 590 {
570 591 struct page *page;
... ... @@ -578,21 +599,7 @@
578 599 page = alloc_fresh_huge_page_node(h, h->hugetlb_next_nid);
579 600 if (page)
580 601 ret = 1;
581   - /*
582   - * Use a helper variable to find the next node and then
583   - * copy it back to hugetlb_next_nid afterwards:
584   - * otherwise there's a window in which a racer might
585   - * pass invalid nid MAX_NUMNODES to alloc_pages_node.
586   - * But we don't need to use a spin_lock here: it really
587   - * doesn't matter if occasionally a racer chooses the
588   - * same nid as we do. Move nid forward in the mask even
589   - * if we just successfully allocated a hugepage so that
590   - * the next caller gets hugepages on the next node.
591   - */
592   - next_nid = next_node(h->hugetlb_next_nid, node_online_map);
593   - if (next_nid == MAX_NUMNODES)
594   - next_nid = first_node(node_online_map);
595   - h->hugetlb_next_nid = next_nid;
  602 + next_nid = hstate_next_node(h);
596 603 } while (!page && h->hugetlb_next_nid != start_nid);
597 604  
598 605 if (ret)