[26/31] kasan, page_alloc: allow skipping unpoisoning for HW_TAGS

Commit Message

[e-mail] Nov. 30, 2021, 10:08 p.m. UTC

From: Andrey Konovalov <[e-mail]>

This patch add a new GFP flag __GFP_SKIP_KASAN_UNPOISON that allows
skipping KASAN poisoning for page_alloc allocations. The flag is only
effective with HW_TAGS KASAN.

This flag will be used by vmalloc code for page_alloc allocations
backing vmalloc() mappings in the following patch. The reason to skip
KASAN poisoning for these pages in page_alloc is because vmalloc code
will be poisoning them instead.

Signed-off-by: Andrey Konovalov <[e-mail]>
---
 include/linux/gfp.h | 13 +++++++++----
 mm/page_alloc.c     | 24 +++++++++++++++++-------
 2 files changed, 26 insertions(+), 11 deletions(-)

Patch

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index dddd7597689f..a4c8ff3fbed1 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -54,9 +54,10 @@  struct vm_area_struct;
 #define ___GFP_THISNODE		0x200000u
 #define ___GFP_ACCOUNT		0x400000u
 #define ___GFP_ZEROTAGS		0x800000u
-#define ___GFP_SKIP_KASAN_POISON	0x1000000u
+#define ___GFP_SKIP_KASAN_UNPOISON	0x1000000u
+#define ___GFP_SKIP_KASAN_POISON	0x2000000u
 #ifdef CONFIG_LOCKDEP
-#define ___GFP_NOLOCKDEP	0x2000000u
+#define ___GFP_NOLOCKDEP	0x4000000u
 #else
 #define ___GFP_NOLOCKDEP	0
 #endif
@@ -235,6 +236,9 @@  struct vm_area_struct;
  * %__GFP_ZEROTAGS zeroes memory tags at allocation time if the memory itself
  * is being zeroed (either via __GFP_ZERO or via init_on_alloc).
  *
+ * %__GFP_SKIP_KASAN_UNPOISON skips KASAN unpoisoning on page allocation.
+ * Currently only has an effect in HW tags mode.
+ *
  * %__GFP_SKIP_KASAN_POISON returns a page which does not need to be poisoned
  * on deallocation. Typically used for userspace pages. Currently only has an
  * effect in HW tags mode.
@@ -243,13 +247,14 @@  struct vm_area_struct;
 #define __GFP_COMP	((__force gfp_t)___GFP_COMP)
 #define __GFP_ZERO	((__force gfp_t)___GFP_ZERO)
 #define __GFP_ZEROTAGS	((__force gfp_t)___GFP_ZEROTAGS)
-#define __GFP_SKIP_KASAN_POISON	((__force gfp_t)___GFP_SKIP_KASAN_POISON)
+#define __GFP_SKIP_KASAN_UNPOISON ((__force gfp_t)___GFP_SKIP_KASAN_UNPOISON)
+#define __GFP_SKIP_KASAN_POISON   ((__force gfp_t)___GFP_SKIP_KASAN_POISON)
 
 /* Disable lockdep for GFP context tracking */
 #define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
 
 /* Room for N __GFP_FOO bits */
-#define __GFP_BITS_SHIFT (25 + IS_ENABLED(CONFIG_LOCKDEP))
+#define __GFP_BITS_SHIFT (26 + IS_ENABLED(CONFIG_LOCKDEP))
 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 
 /**
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 4eb341351124..3afebc037fcd 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2381,6 +2381,21 @@  static bool check_new_pages(struct page *page, unsigned int order)
 	return false;
 }
 
+static inline bool should_skip_kasan_unpoison(gfp_t flags, bool init_tags)
+{
+	/* Don't skip if a software KASAN mode is enabled. */
+	if (!IS_ENABLED(CONFIG_KASAN_HW_TAGS))
+		return false;
+
+	/*
+	 * For hardware tag-based KASAN, skip if either:
+	 *
+	 * 1. Memory tags have already been cleared via tag_clear_highpage().
+	 * 2. Skipping has been requested via __GFP_SKIP_KASAN_UNPOISON.
+	 */
+	return init_tags || (flags & __GFP_SKIP_KASAN_UNPOISON);
+}
+
 inline void post_alloc_hook(struct page *page, unsigned int order,
 				gfp_t gfp_flags)
 {
@@ -2420,13 +2435,8 @@  inline void post_alloc_hook(struct page *page, unsigned int order,
 		/* Note that memory is already initialized by the loop above. */
 		init = false;
 	}
-	/*
-	 * If either a software KASAN mode is enabled, or,
-	 * in the case of hardware tag-based KASAN,
-	 * if memory tags have not been cleared via tag_clear_highpage().
-	 */
-	if (!IS_ENABLED(CONFIG_KASAN_HW_TAGS) || !init_tags) {
-		/* Mark shadow memory or set memory tags. */
+	if (!should_skip_kasan_unpoison(gfp_flags, init_tags)) {
+		/* Unpoison shadow memory or set memory tags. */
 		kasan_unpoison_pages(page, order, init);
 
 		/* Note that memory is already initialized by KASAN. */