mm, slab: Check GFP_SLAB_BUG_MASK before alloc_pages in kmalloc_order

PATCH Superseded Long Li 2 participants 2 messages Jul 4, 2020

Subsystems (2)

Memory management Slab allocator

lore

Share & tools

lore

Long Li

author

2020-07-04 02:26 UTC #0 lore raw

PATCH v5 1/1 mm, slab: Check GFP_SLAB_BUG_MASK before alloc_pages in kmalloc_order

kmalloc cannot allocate memory from HIGHMEM. Allocating large amounts of memory currently bypasses the check and will simply leak the memory when page_address() returns NULL. To fix this, factor the GFP_SLAB_BUG_MASK check out of slab & slub, and call it from kmalloc_order() as well. In order to make the code clear, the warning message is put in one place.

Reviewed-by Matthew Wilcox (Oracle) <[email protected]>

Reviewed-by Pekka Enberg <[email protected]>

Cc Christoph Lameter <[email protected]>

Cc David Rientjes <[email protected]>

Cc Joonsoo Kim <[email protected]>

Cc Andrew Morton <[email protected]>

Signed-off-by Long Li <[email protected]>

---

changes in v5:
-Change the check function name to kmalloc_fix_flags(), This name may be more appropriate.

changes in V4:
-Change the check function name to kmalloc_check_flags() -Put the flags check into the kmalloc_check_flags()

changes in V3:
-Put the warning message in one place
-updage the change log to be clear

 mm/slab.c        |  8 +-------
 mm/slab.h        |  1 +
 mm/slab_common.c | 18 +++++++++++++++++-
 mm/slub.c        |  8 +-------
 4 files changed, 20 insertions(+), 15 deletions(-)

mm/slab.c +1 -7

index ac7a223d9ac3..f2f150bd180b 100644
@@ -2573,13 +2573,7 @@ static struct page *cache_grow_begin(struct kmem_cache *cachep,
 	 * Be lazy and only check for valid flags here,  keeping it out of the
 	 * critical path in kmem_cache_alloc().
 	 */
-	if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
-		gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
-		flags &= ~GFP_SLAB_BUG_MASK;
-		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
-				invalid_mask, &invalid_mask, flags, &flags);
-		dump_stack();
-	}
+	flags = kmalloc_fix_flags(flags);
 	WARN_ON_ONCE(cachep->ctor && (flags & __GFP_ZERO));
 	local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);

mm/slab.h +1 -0

index a06f3313e4a0..8cd2bf391725 100644
@@ -90,6 +90,7 @@ void create_kmalloc_caches(slab_flags_t);
 struct kmem_cache *kmalloc_slab(size_t, gfp_t);
 #endif
 
+gfp_t kmalloc_fix_flags(gfp_t flags);
 
 /* Functions provided by the slab allocators */
 int __kmem_cache_create(struct kmem_cache *, slab_flags_t flags);

mm/slab_common.c +17 -1

index a143a8c8f874..16d63f6dad05 100644
@@ -26,6 +26,8 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/kmem.h>
 
+#include "internal.h"
+
 #include "slab.h"
 
 enum slab_state slab_state;

@@ -805,6 +807,20 @@ void __init create_kmalloc_caches(slab_flags_t flags)
 }
 #endif /* !CONFIG_SLOB */
 
+gfp_t kmalloc_fix_flags(gfp_t flags)
+{
+	if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
+		gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
+
+		flags &= ~GFP_SLAB_BUG_MASK;
+		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
+				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();
+	}
+
+	return flags;
+}
+
 /*
  * To avoid unnecessary overhead, we pass through large allocation requests
  * directly to the page allocator. We use __GFP_COMP, because we will need to

@@ -815,7 +831,7 @@ void *kmalloc_order(size_t size, gfp_t flags, unsigned int order)
 	void *ret = NULL;
 	struct page *page;
 
-	flags |= __GFP_COMP;
+	flags = kmalloc_fix_flags(flags) | __GFP_COMP;
 	page = alloc_pages(flags, order);
 	if (likely(page)) {
 		ret = page_address(page);

mm/slub.c +1 -7

index 62d2de56549e..dfaad93163d5 100644
@@ -1817,13 +1817,7 @@ static struct page *allocate_slab(struct kmem_cache *s, gfp_t flags, int node)
 
 static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 {
-	if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
-		gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
-		flags &= ~GFP_SLAB_BUG_MASK;
-		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
-				invalid_mask, &invalid_mask, flags, &flags);
-		dump_stack();
-	}
+	flags = kmalloc_fix_flags(flags);
 
 	return allocate_slab(s,
 		flags & (GFP_RECLAIM_MASK | GFP_CONSTRAINT_MASK), node);

Signature · 1 line

2.17.1

Matthew Wilcox

2020-07-04 02:32 UTC #1 lore raw

Quoted 21 lines · Long Li

kmalloc cannot allocate memory from HIGHMEM. Allocating large amounts of memory currently bypasses the check and will simply leak the memory when page_address() returns NULL. To fix this, factor the GFP_SLAB_BUG_MASK check out of slab & slub, and call it from kmalloc_order() as well. In order to make the code clear, the warning message is put in one place.
Reviewed-by Matthew Wilcox (Oracle) <[email protected]>
Reviewed-by Pekka Enberg <[email protected]>
Cc Christoph Lameter <[email protected]>
Cc David Rientjes <[email protected]>
Cc Joonsoo Kim <[email protected]>
Cc Andrew Morton <[email protected]>
Signed-off-by Long Li <[email protected]>
---
changes in v5:
-Change the check function name to kmalloc_fix_flags(), This name may be more appropriate.
changes in V4:

Review

Quoted 2 lines

(quoted diff) +0 -2

-Change the check function name to kmalloc_check_flags()
-Put the flags check into the kmalloc_check_flags()

No. As I said:

The point of not doing that was that this is unlikely(). With your change there is now a function call to check something that's (extremely) unlikely().