On Fri, 22 Dec 2023 20:28:05 +0000 "Matthew Wilcox (Oracle)" <willy@xxxxxxxxxxxxx> wrote:
> Add folio_alloc_node() to replace alloc_pages_node() and then use
> folio APIs throughout instead of converting back to pages.
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> ---
> include/linux/gfp.h | 9 +++++++++
> mm/slub.c | 15 +++++++--------
This depends on changes which are in Vlastimil's tree and linux-next.
So I reworked it to not do that, which means there will be a resolution
for Linus to do, which Stephen will tell us about. It's simple, just
from code motion.
Maybe mm.git should include the slab tree, I haven't really considered
what would be the implications of that.
include/linux/gfp.h | 9 +++++++++
mm/slab_common.c | 15 +++++++--------
2 files changed, 16 insertions(+), 8 deletions(-)
--- a/include/linux/gfp.h~slab-convert-__kmalloc_large_node-and-free_large_kmalloc-to-use-folios
+++ a/include/linux/gfp.h
@@ -247,6 +247,15 @@ struct folio *__folio_alloc_node(gfp_t g
return __folio_alloc(gfp, order, nid, NULL);
}
+static inline
+struct folio *folio_alloc_node(gfp_t gfp, unsigned int order, int nid)
+{
+ if (nid == NUMA_NO_NODE)
+ nid = numa_mem_id();
+
+ return __folio_alloc_node(gfp, order, nid);
+}
+
/*
* Allocate pages, preferring the node given as nid. When nid == NUMA_NO_NODE,
* prefer the current CPU's closest node. Otherwise node must be valid and
--- a/mm/slab_common.c~slab-convert-__kmalloc_large_node-and-free_large_kmalloc-to-use-folios
+++ a/mm/slab_common.c
@@ -979,9 +979,9 @@ void free_large_kmalloc(struct folio *fo
kasan_kfree_large(object);
kmsan_kfree_large(object);
- mod_lruvec_page_state(folio_page(folio, 0), NR_SLAB_UNRECLAIMABLE_B,
+ lruvec_stat_mod_folio(folio, NR_SLAB_UNRECLAIMABLE_B,
-(PAGE_SIZE << order));
- __free_pages(folio_page(folio, 0), order);
+ folio_put(folio);
}
static void *__kmalloc_large_node(size_t size, gfp_t flags, int node);
@@ -1137,18 +1137,17 @@ gfp_t kmalloc_fix_flags(gfp_t flags)
static void *__kmalloc_large_node(size_t size, gfp_t flags, int node)
{
- struct page *page;
+ struct folio *folio;
void *ptr = NULL;
unsigned int order = get_order(size);
if (unlikely(flags & GFP_SLAB_BUG_MASK))
flags = kmalloc_fix_flags(flags);
- flags |= __GFP_COMP;
- page = alloc_pages_node(node, flags, order);
- if (page) {
- ptr = page_address(page);
- mod_lruvec_page_state(page, NR_SLAB_UNRECLAIMABLE_B,
+ folio = folio_alloc_node(flags, order, node);
+ if (folio) {
+ ptr = folio_address(folio);
+ lruvec_stat_mod_folio(folio, NR_SLAB_UNRECLAIMABLE_B,
PAGE_SIZE << order);
}
_
