diff mbox series

[RFC,06/14] mm/slub: pass additional args to alloc_slab_page

Message ID 20230915105933.495735-7-matteorizzo@google.com (mailing list archive)
State New
Headers show
Series Prevent cross-cache attacks in the SLUB allocator | expand

Commit Message

Matteo Rizzo Sept. 15, 2023, 10:59 a.m. UTC 
From: Jann Horn <jannh@google.com>

This is refactoring in preparation for SLAB_VIRTUAL.

The implementation of SLAB_VIRTUAL needs access to struct kmem_cache in
alloc_slab_page in order to take unused slabs from the slab freelist,
which is per-cache.

In addition to that it passes two different sets of GFP flags.
meta_gfp_flags is used for the memory backing the metadata region and
page tables, and gfp_flags for the data memory.

Signed-off-by: Jann Horn <jannh@google.com>
Co-developed-by: Matteo Rizzo <matteorizzo@google.com>
Signed-off-by: Matteo Rizzo <matteorizzo@google.com>
---
 mm/slub.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

Comments

Kees Cook Sept. 15, 2023, 9:03 p.m. UTC | #1 
On Fri, Sep 15, 2023 at 10:59:25AM +0000, Matteo Rizzo wrote:
> From: Jann Horn <jannh@google.com>
> 
> This is refactoring in preparation for SLAB_VIRTUAL.
> 
> The implementation of SLAB_VIRTUAL needs access to struct kmem_cache in
> alloc_slab_page in order to take unused slabs from the slab freelist,
> which is per-cache.
> 
> In addition to that it passes two different sets of GFP flags.
> meta_gfp_flags is used for the memory backing the metadata region and
> page tables, and gfp_flags for the data memory.
> 
> Signed-off-by: Jann Horn <jannh@google.com>

Reviewed-by: Kees Cook <keescook@chromium.org>
diff mbox series

Patch

diff --git a/mm/slub.c b/mm/slub.c
index 9b87afade125..eaa1256aff89 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -1869,7 +1869,8 @@  static void folio_clear_slab(struct folio *folio, struct slab *slab)
 	__folio_clear_slab(folio);
 }
 
-static inline struct slab *alloc_slab_page(gfp_t flags, int node,
+static inline struct slab *alloc_slab_page(struct kmem_cache *s,
+		gfp_t meta_flags, gfp_t flags, int node,
 		struct kmem_cache_order_objects oo)
 {
 	struct folio *folio;
@@ -2020,7 +2021,7 @@  static struct slab *allocate_slab(struct kmem_cache *s, gfp_t flags, int node)
 	if ((alloc_gfp & __GFP_DIRECT_RECLAIM) && oo_order(oo) > oo_order(s->min))
 		alloc_gfp = (alloc_gfp | __GFP_NOMEMALLOC) & ~__GFP_RECLAIM;
 
-	slab = alloc_slab_page(alloc_gfp, node, oo);
+	slab = alloc_slab_page(s, flags, alloc_gfp, node, oo);
 	if (unlikely(!slab)) {
 		oo = s->min;
 		alloc_gfp = flags;
@@ -2028,7 +2029,7 @@  static struct slab *allocate_slab(struct kmem_cache *s, gfp_t flags, int node)
 		 * Allocation may have failed due to fragmentation.
 		 * Try a lower order alloc if possible
 		 */
-		slab = alloc_slab_page(alloc_gfp, node, oo);
+		slab = alloc_slab_page(s, flags, alloc_gfp, node, oo);
 		if (unlikely(!slab))
 			return NULL;
 		stat(s, ORDER_FALLBACK);