在 2021/10/22 11:26, Andrew Morton 写道:
On Thu, 21 Oct 2021 16:07:44 +0800 Chen Wandun <chenwandun@xxxxxxxxxx> wrote:
It
What is "it"?
it == > [PATCH] mm/vmalloc: fix numa spreading for large hash tables;
will cause significant performance regressions in some situations
as Andrew mentioned in [1]. The main situation is vmalloc, vmalloc
will allocate pages with NUMA_NO_NODE by default, that will result
in alloc page one by one;
In order to solve this, __alloc_pages_bulk and mempolicy should be
considered at the same time.
1) If node is specified in memory allocation request, it will alloc
all pages by __alloc_pages_bulk.
2) If interleaving allocate memory, it will cauculate how many pages
should be allocated in each node, and use __alloc_pages_bulk to alloc
pages in each node.
This v3 patch didn't incorporate my two fixes, below. It is usual to
incorporate such fixes prior to resending. I have retained those two
fixes, now against v3.
From: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Subject: mm-vmalloc-introduce-alloc_pages_bulk_array_mempolicy-to-accelerate-memory-allocation-fix
make two functions static
Cc: Chen Wandun <chenwandun@xxxxxxxxxx>
Cc: Eric Dumazet <edumazet@xxxxxxxxxx>
Cc: Hanjun Guo <guohanjun@xxxxxxxxxx>
Cc: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
Cc: Nicholas Piggin <npiggin@xxxxxxxxx>
Cc: Shakeel Butt <shakeelb@xxxxxxxxxx>
Cc: Uladzislau Rezki (Sony) <urezki@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/mempolicy.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
--- a/mm/mempolicy.c~mm-vmalloc-introduce-alloc_pages_bulk_array_mempolicy-to-accelerate-memory-allocation-fix
+++ a/mm/mempolicy.c
@@ -2196,7 +2196,7 @@ struct page *alloc_pages(gfp_t gfp, unsi
}
EXPORT_SYMBOL(alloc_pages);
-unsigned long alloc_pages_bulk_array_interleave(gfp_t gfp,
+static unsigned long alloc_pages_bulk_array_interleave(gfp_t gfp,
struct mempolicy *pol, unsigned long nr_pages,
struct page **page_array)
{
@@ -2231,7 +2231,7 @@ unsigned long alloc_pages_bulk_array_int
return total_allocated;
}
-unsigned long alloc_pages_bulk_array_preferred_many(gfp_t gfp, int nid,
+static unsigned long alloc_pages_bulk_array_preferred_many(gfp_t gfp, int nid,
struct mempolicy *pol, unsigned long nr_pages,
struct page **page_array)
{
_
From: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Subject: mm-vmalloc-introduce-alloc_pages_bulk_array_mempolicy-to-accelerate-memory-allocation-fix-2
fix CONFIG_NUMA=n build. alloc_pages_bulk_array_mempolicy() was undefined
Cc: Chen Wandun <chenwandun@xxxxxxxxxx>
Cc: Eric Dumazet <edumazet@xxxxxxxxxx>
Cc: Hanjun Guo <guohanjun@xxxxxxxxxx>
Cc: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
Cc: Nicholas Piggin <npiggin@xxxxxxxxx>
Cc: Uladzislau Rezki (Sony) <urezki@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/vmalloc.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
--- a/mm/vmalloc.c~mm-vmalloc-introduce-alloc_pages_bulk_array_mempolicy-to-accelerate-memory-allocation-fix-2
+++ a/mm/vmalloc.c
@@ -2860,7 +2860,7 @@ vm_area_alloc_pages(gfp_t gfp, int nid,
* otherwise memory may be allocated in only one node,
* but mempolcy want to alloc memory by interleaving.
*/
- if (nid == NUMA_NO_NODE)
+ if (IS_ENABLED(CONFIG_NUMA) && nid == NUMA_NO_NODE)
nr = alloc_pages_bulk_array_mempolicy(gfp,
nr_pages_request,
pages + nr_allocated);
_
.
