Skip to contentSkip to conversation

arm64: improve efficiency of setting tags for user pages

PATCH Superseded Peter Collingbourne 2 participants 7 messages May 11, 2021 – May 12, 2021

Subsystems (7)
Memory management MM core arm64 Tracing +3 more subsystems

v2 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Peter Collingbourne

author
#0 lore raw

v2 cover arm64: improve efficiency of setting tags for user pages

Currently we can end up touching PROT_MTE user pages twice on fault and once on unmap. On fault, with KASAN disabled we first clear data and then set tags to 0, and with KASAN enabled we simultaneously clear data and set tags to the KASAN random tag, and then set tags again to 0. On unmap, we poison the page by setting tags, but this is less likely to find a bug than poisoning kernel pages.

This patch series fixes these inefficiencies by only touching the pages once on fault using the DC GZVA instruction to clear both data and tags, and providing the option to avoid poisoning user pages on free.

Peter Collingbourne (3):
kasan: use separate (un)poison implementation for integrated init
arm64: mte: handle tags zeroing at page allocation time
kasan: allow freed user page poisoning to be disabled with HW tags

 arch/arm64/include/asm/mte.h   |  4 ++
 arch/arm64/include/asm/page.h  |  9 ++++-
 arch/arm64/lib/mte.S           | 20 ++++++++++
 arch/arm64/mm/fault.c          | 25 +++++++++++++
 arch/arm64/mm/proc.S           | 10 +++--
 include/linux/gfp.h            | 18 +++++++--
 include/linux/highmem.h        |  8 ++++
 include/linux/kasan.h          | 66 ++++++++++++++++++++-------------
 include/linux/page-flags.h     |  9 +++++
 include/trace/events/mmflags.h |  9 ++++-
 mm/kasan/common.c              |  4 +-
 mm/kasan/hw_tags.c             | 31 ++++++++++++++++
 mm/mempool.c                   |  6 ++-
 mm/page_alloc.c                | 67 ++++++++++++++++++++--------------
 14 files changed, 221 insertions(+), 65 deletions(-)
Signature · 1 line
2.31.1.607.g51e8a6a459-goog
author
#1 lore raw

PATCH v2 1/3 kasan: use separate (un)poison implementation for integrated init

Currently with integrated init page_alloc.c needs to know whether kasan_alloc_pages() will zero initialize memory, but this will start becoming more complicated once we start adding tag initialization support for user pages. To avoid page_alloc.c needing to know more details of what integrated init will do, move the unpoisoning logic for integrated init into the HW tags implementation. Currently the logic is identical but it will diverge in subsequent patches.

For symmetry do the same for poisoning although this logic will be unaffected by subsequent patches.

Signed-off-by Peter Collingbourne <[email protected]>
Link https://linux-review.googlesource.com/id/I2c550234c6c4a893c48c18ff0c6ce658c7c67056

---
v2:
- fix build with KASAN disabled

 include/linux/kasan.h | 66 +++++++++++++++++++++++++++----------------
 mm/kasan/common.c     |  4 +--
 mm/kasan/hw_tags.c    | 14 +++++++++
 mm/mempool.c          |  6 ++--
 mm/page_alloc.c       | 56 +++++++++++++++++++-----------------
 5 files changed, 91 insertions(+), 55 deletions(-)
include/linux/kasan.h +41 -25
index b1678a61e6a7..e35fa301d3db 100644
@@ -2,6 +2,7 @@
 #ifndef _LINUX_KASAN_H
 #define _LINUX_KASAN_H
 
+#include <linux/bug.h>
 #include <linux/static_key.h>
 #include <linux/types.h>
 

@@ -79,14 +80,6 @@ static inline void kasan_disable_current(void) {}
 
 #endif /* CONFIG_KASAN_GENERIC || CONFIG_KASAN_SW_TAGS */
 
-#ifdef CONFIG_KASAN
-
-struct kasan_cache {
-	int alloc_meta_offset;
-	int free_meta_offset;
-	bool is_kmalloc;
-};
-
 #ifdef CONFIG_KASAN_HW_TAGS
 
 DECLARE_STATIC_KEY_FALSE(kasan_flag_enabled);

@@ -101,11 +94,18 @@ static inline bool kasan_has_integrated_init(void)
 	return kasan_enabled();
 }
 
+void kasan_alloc_pages(struct page *page, unsigned int order, gfp_t flags);
+void kasan_free_pages(struct page *page, unsigned int order);
+
 #else /* CONFIG_KASAN_HW_TAGS */
 
 static inline bool kasan_enabled(void)
 {
+#ifdef CONFIG_KASAN
 	return true;
+#else
+	return false;
+#endif
 }
 
 static inline bool kasan_has_integrated_init(void)

@@ -113,8 +113,30 @@ static inline bool kasan_has_integrated_init(void)
 	return false;
 }
 
+static __always_inline void kasan_alloc_pages(struct page *page,
+					      unsigned int order, gfp_t flags)
+{
+	/* Only available for integrated init. */
+	BUG();
+}
+
+static __always_inline void kasan_free_pages(struct page *page,
+					     unsigned int order)
+{
+	/* Only available for integrated init. */
+	BUG();
+}
+
 #endif /* CONFIG_KASAN_HW_TAGS */
 
+#ifdef CONFIG_KASAN
+
+struct kasan_cache {
+	int alloc_meta_offset;
+	int free_meta_offset;
+	bool is_kmalloc;
+};
+
 slab_flags_t __kasan_never_merge(void);
 static __always_inline slab_flags_t kasan_never_merge(void)
 {

@@ -130,20 +152,20 @@ static __always_inline void kasan_unpoison_range(const void *addr, size_t size)
 		__kasan_unpoison_range(addr, size);
 }
 
-void __kasan_alloc_pages(struct page *page, unsigned int order, bool init);
-static __always_inline void kasan_alloc_pages(struct page *page,
+void __kasan_poison_pages(struct page *page, unsigned int order, bool init);
+static __always_inline void kasan_poison_pages(struct page *page,
 						unsigned int order, bool init)
 {
 	if (kasan_enabled())
-		__kasan_alloc_pages(page, order, init);
+		__kasan_poison_pages(page, order, init);
 }
 
-void __kasan_free_pages(struct page *page, unsigned int order, bool init);
-static __always_inline void kasan_free_pages(struct page *page,
-						unsigned int order, bool init)
+void __kasan_unpoison_pages(struct page *page, unsigned int order, bool init);
+static __always_inline void kasan_unpoison_pages(struct page *page,
+						 unsigned int order, bool init)
 {
 	if (kasan_enabled())
-		__kasan_free_pages(page, order, init);
+		__kasan_unpoison_pages(page, order, init);
 }
 
 void __kasan_cache_create(struct kmem_cache *cache, unsigned int *size,

@@ -285,21 +307,15 @@ void kasan_restore_multi_shot(bool enabled);
 
 #else /* CONFIG_KASAN */
 
-static inline bool kasan_enabled(void)
-{
-	return false;
-}
-static inline bool kasan_has_integrated_init(void)
-{
-	return false;
-}
 static inline slab_flags_t kasan_never_merge(void)
 {
 	return 0;
 }
 static inline void kasan_unpoison_range(const void *address, size_t size) {}
-static inline void kasan_alloc_pages(struct page *page, unsigned int order, bool init) {}
-static inline void kasan_free_pages(struct page *page, unsigned int order, bool init) {}
+static inline void kasan_poison_pages(struct page *page, unsigned int order,
+				      bool init) {}
+static inline void kasan_unpoison_pages(struct page *page, unsigned int order,
+					bool init) {}
 static inline void kasan_cache_create(struct kmem_cache *cache,
 				      unsigned int *size,
 				      slab_flags_t *flags) {}
mm/kasan/common.c +2 -2
index 6bb87f2acd4e..0ecd293af344 100644
@@ -97,7 +97,7 @@ slab_flags_t __kasan_never_merge(void)
 	return 0;
 }
 
-void __kasan_alloc_pages(struct page *page, unsigned int order, bool init)
+void __kasan_unpoison_pages(struct page *page, unsigned int order, bool init)
 {
 	u8 tag;
 	unsigned long i;

@@ -111,7 +111,7 @@ void __kasan_alloc_pages(struct page *page, unsigned int order, bool init)
 	kasan_unpoison(page_address(page), PAGE_SIZE << order, init);
 }
 
-void __kasan_free_pages(struct page *page, unsigned int order, bool init)
+void __kasan_poison_pages(struct page *page, unsigned int order, bool init)
 {
 	if (likely(!PageHighMem(page)))
 		kasan_poison(page_address(page), PAGE_SIZE << order,
mm/kasan/hw_tags.c +14 -0
index 4004388b4e4b..45e552cb9172 100644
@@ -238,6 +238,20 @@ struct kasan_track *kasan_get_free_track(struct kmem_cache *cache,
 	return &alloc_meta->free_track[0];
 }
 
+void kasan_alloc_pages(struct page *page, unsigned int order, gfp_t flags)
+{
+	bool init = !want_init_on_free() && want_init_on_alloc(flags);
+
+	kasan_unpoison_pages(page, order, init);
+}
+
+void kasan_free_pages(struct page *page, unsigned int order)
+{
+	bool init = want_init_on_free();
+
+	kasan_poison_pages(page, order, init);
+}
+
 #if IS_ENABLED(CONFIG_KASAN_KUNIT_TEST)
 
 void kasan_set_tagging_report_once(bool state)
mm/mempool.c +4 -2
index a258cf4de575..0b8afbec3e35 100644
@@ -106,7 +106,8 @@ static __always_inline void kasan_poison_element(mempool_t *pool, void *element)
 	if (pool->alloc == mempool_alloc_slab || pool->alloc == mempool_kmalloc)
 		kasan_slab_free_mempool(element);
 	else if (pool->alloc == mempool_alloc_pages)
-		kasan_free_pages(element, (unsigned long)pool->pool_data, false);
+		kasan_poison_pages(element, (unsigned long)pool->pool_data,
+				   false);
 }
 
 static void kasan_unpoison_element(mempool_t *pool, void *element)

@@ -114,7 +115,8 @@ static void kasan_unpoison_element(mempool_t *pool, void *element)
 	if (pool->alloc == mempool_alloc_slab || pool->alloc == mempool_kmalloc)
 		kasan_unpoison_range(element, __ksize(element));
 	else if (pool->alloc == mempool_alloc_pages)
-		kasan_alloc_pages(element, (unsigned long)pool->pool_data, false);
+		kasan_unpoison_pages(element, (unsigned long)pool->pool_data,
+				     false);
 }
 
 static __always_inline void add_element(mempool_t *pool, void *element)
mm/page_alloc.c +30 -26

100 lines not on this page · raw message

Signature · 1 line
2.31.1.607.g51e8a6a459-goog
maintainer
#2 lore raw

Re: [PATCH v2 1/3] kasan: use separate (un)poison implementation for integrated init

Quoted 4 lines · Peter Collingbourne
include/linux/kasan.h +0 -0
index b1678a61e6a7..e35fa301d3db 100644

[...]

Quoted 7 lines
include/linux/kasan.h +3 -0
@@ -113,8 +113,30 @@ static inline bool kasan_has_integrated_init(void)
 	return false;
 }
 
+static __always_inline void kasan_alloc_pages(struct page *page,
+					      unsigned int order, gfp_t flags)
+{

include/linux/kasan.h static inline bool kasan_has_integrated_init(void) · line 128

Quoted 10 lines
include/linux/kasan.h +10 -0
+	/* Only available for integrated init. */
+	BUG();
+}
+
+static __always_inline void kasan_free_pages(struct page *page,
+					     unsigned int order)
+{
+	/* Only available for integrated init. */
+	BUG();
+}

Would these two work as BUILD_BUG()?

Signature · 1 line
Catalin
author
#3 lore raw

Re: [PATCH v2 1/3] kasan: use separate (un)poison implementation for integrated init

Quoted 11 lines · Catalin Marinas
Quoted 4 lines · Peter Collingbourne
include/linux/kasan.h +0 -0
index b1678a61e6a7..e35fa301d3db 100644

[...]

Quoted 4 lines
include/linux/kasan.h +0 -0
@@ -113,8 +113,30 @@ static inline bool kasan_has_integrated_init(void)
      return false;
 }
Quoted 15 lines

include/linux/kasan.h static inline bool kasan_has_integrated_init(void) · line 128

Quoted 13 lines
include/linux/kasan.h +13 -0
+static __always_inline void kasan_alloc_pages(struct page *page,
+                                           unsigned int order, gfp_t flags)
+{
+     /* Only available for integrated init. */
+     BUG();
+}
+
+static __always_inline void kasan_free_pages(struct page *page,
+                                          unsigned int order)
+{
+     /* Only available for integrated init. */
+     BUG();
+}

Would these two work as BUILD_BUG()?

It looks like it. This creates a dependency on the compiler being powerful enough to optimize away "if (0) undefined_function();" after inlining, which I'm not really a big fan of, but it looks like we're already doing similar things elsewhere in the kernel, so we might as well do it here as well.

Peter

author
#4 lore raw

PATCH v2 2/3 arm64: mte: handle tags zeroing at page allocation time

In reply to Peter Collingbourne

Currently, on an anonymous page fault, the kernel allocates a zeroed page and maps it in user space. If the mapping is tagged (PROT_MTE), set_pte_at() additionally clears the tags. It is, however, more efficient to clear the tags at the same time as zeroing the data on allocation. To avoid clearing the tags on any page (which may not be mapped as tagged), only do this if the vma flags contain VM_MTE. This requires introducing a new GFP flag that is used to determine whether to clear the tags.

The DC GZVA instruction with a 0 top byte (and 0 tag) requires top-byte-ignore. Set the TCR_EL1.{TBI1,TBID1} bits irrespective of whether KASAN_HW is enabled.

Signed-off-by Peter Collingbourne <[email protected]>
Co-developed-by Catalin Marinas <[email protected]>
Signed-off-by Catalin Marinas <[email protected]>
Link https://linux-review.googlesource.com/id/Id46dc94e30fe11474f7e54f5d65e7658dbdddb26

---
v2:
- remove want_zero_tags_on_free()

 arch/arm64/include/asm/mte.h  |  4 ++++
 arch/arm64/include/asm/page.h |  9 +++++++--
 arch/arm64/lib/mte.S          | 20 ++++++++++++++++++++
 arch/arm64/mm/fault.c         | 25 +++++++++++++++++++++++++
 arch/arm64/mm/proc.S          | 10 +++++++---
 include/linux/gfp.h           |  9 +++++++--
 include/linux/highmem.h       |  8 ++++++++
 mm/kasan/hw_tags.c            |  9 ++++++++-
 mm/page_alloc.c               | 13 ++++++++++---
 9 files changed, 96 insertions(+), 11 deletions(-)
arch/arm64/include/asm/mte.h +4 -0
index bc88a1ced0d7..67bf259ae768 100644
@@ -37,6 +37,7 @@ void mte_free_tag_storage(char *storage);
 /* track which pages have valid allocation tags */
 #define PG_mte_tagged	PG_arch_2
 
+void mte_zero_clear_page_tags(void *addr);
 void mte_sync_tags(pte_t *ptep, pte_t pte);
 void mte_copy_page_tags(void *kto, const void *kfrom);
 void mte_thread_init_user(void);

@@ -53,6 +54,9 @@ int mte_ptrace_copy_tags(struct task_struct *child, long request,
 /* unused if !CONFIG_ARM64_MTE, silence the compiler */
 #define PG_mte_tagged	0
 
+static inline void mte_zero_clear_page_tags(void *addr)
+{
+}
 static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
 {
 }
arch/arm64/include/asm/page.h +7 -2
index 012cffc574e8..448e14071d13 100644
@@ -13,6 +13,7 @@
 #ifndef __ASSEMBLY__
 
 #include <linux/personality.h> /* for READ_IMPLIES_EXEC */
+#include <linux/types.h> /* for gfp_t */
 #include <asm/pgtable-types.h>
 
 struct page;

@@ -28,10 +29,14 @@ void copy_user_highpage(struct page *to, struct page *from,
 void copy_highpage(struct page *to, struct page *from);
 #define __HAVE_ARCH_COPY_HIGHPAGE
 
-#define __alloc_zeroed_user_highpage(movableflags, vma, vaddr) \
-	alloc_page_vma(GFP_HIGHUSER | __GFP_ZERO | movableflags, vma, vaddr)
+struct page *__alloc_zeroed_user_highpage(gfp_t movableflags,
+					  struct vm_area_struct *vma,
+					  unsigned long vaddr);
 #define __HAVE_ARCH_ALLOC_ZEROED_USER_HIGHPAGE
 
+void tag_clear_highpage(struct page *to);
+#define __HAVE_ARCH_TAG_CLEAR_HIGHPAGE
+
 #define clear_user_page(page, vaddr, pg)	clear_page(page)
 #define copy_user_page(to, from, vaddr, pg)	copy_page(to, from)
arch/arm64/lib/mte.S +20 -0
index 351537c12f36..e83643b3995f 100644
@@ -36,6 +36,26 @@ SYM_FUNC_START(mte_clear_page_tags)
 	ret
 SYM_FUNC_END(mte_clear_page_tags)
 
+/*
+ * Zero the page and tags at the same time
+ *
+ * Parameters:
+ *	x0 - address to the beginning of the page
+ */
+SYM_FUNC_START(mte_zero_clear_page_tags)
+	mrs	x1, dczid_el0
+	and	w1, w1, #0xf
+	mov	x2, #4
+	lsl	x1, x2, x1
+	and	x0, x0, #(1 << MTE_TAG_SHIFT) - 1	// clear the tag
+
+1:	dc	gzva, x0
+	add	x0, x0, x1
+	tst	x0, #(PAGE_SIZE - 1)
+	b.ne	1b
+	ret
+SYM_FUNC_END(mte_zero_clear_page_tags)
+
 /*
  * Copy the tags from the source page to the destination one
  *   x0 - address of the destination page
arch/arm64/mm/fault.c +25 -0
index 871c82ab0a30..8127e0c0b8fb 100644
@@ -921,3 +921,28 @@ void do_debug_exception(unsigned long addr_if_watchpoint, unsigned int esr,
 	debug_exception_exit(regs);
 }
 NOKPROBE_SYMBOL(do_debug_exception);
+
+/*
+ * Used during anonymous page fault handling.
+ */
+struct page *__alloc_zeroed_user_highpage(gfp_t flags,
+					  struct vm_area_struct *vma,
+					  unsigned long vaddr)
+{
+	/*
+	 * If the page is mapped with PROT_MTE, initialise the tags at the
+	 * point of allocation and page zeroing as this is usually faster than
+	 * separate DC ZVA and STGM.
+	 */
+	if (vma->vm_flags & VM_MTE)
+		flags |= __GFP_ZEROTAGS;
+
+	return alloc_page_vma(GFP_HIGHUSER | __GFP_ZERO | flags, vma, vaddr);
+}
+
+void tag_clear_highpage(struct page *page)
+{
+	mte_zero_clear_page_tags(page_address(page));
+	page_kasan_tag_reset(page);
+	set_bit(PG_mte_tagged, &page->flags);
+}
arch/arm64/mm/proc.S +7 -3
index 0a48191534ff..a27c77dbe91c 100644
@@ -46,9 +46,13 @@
 #endif
 
 #ifdef CONFIG_KASAN_HW_TAGS
-#define TCR_KASAN_HW_FLAGS SYS_TCR_EL1_TCMA1 | TCR_TBI1 | TCR_TBID1
+#define TCR_MTE_FLAGS SYS_TCR_EL1_TCMA1 | TCR_TBI1 | TCR_TBID1
 #else
-#define TCR_KASAN_HW_FLAGS 0
+/*
+ * The mte_zero_clear_page_tags() implementation uses DC GZVA, which relies on
+ * TBI being enabled at EL1.
+ */
+#define TCR_MTE_FLAGS TCR_TBI1 | TCR_TBID1
 #endif
 
 /*

@@ -452,7 +456,7 @@ SYM_FUNC_START(__cpu_setup)
 	msr_s	SYS_TFSRE0_EL1, xzr
 
 	/* set the TCR_EL1 bits */
-	mov_q	x10, TCR_KASAN_HW_FLAGS
+	mov_q	x10, TCR_MTE_FLAGS
 	orr	tcr, tcr, x10
 1:
 #endif
include/linux/gfp.h +7 -2
index 11da8af06704..68ba237365dc 100644
@@ -53,8 +53,9 @@ struct vm_area_struct;
 #define ___GFP_HARDWALL		0x100000u
 #define ___GFP_THISNODE		0x200000u
 #define ___GFP_ACCOUNT		0x400000u
+#define ___GFP_ZEROTAGS		0x800000u
 #ifdef CONFIG_LOCKDEP
-#define ___GFP_NOLOCKDEP	0x800000u
+#define ___GFP_NOLOCKDEP	0x1000000u
 #else
 #define ___GFP_NOLOCKDEP	0
 #endif

@@ -229,16 +230,20 @@ struct vm_area_struct;
  * %__GFP_COMP address compound page metadata.
  *
  * %__GFP_ZERO returns a zeroed page on success.
+ *
+ * %__GFP_ZEROTAGS returns a page with zeroed memory tags on success, if
+ * __GFP_ZERO is set.
  */
 #define __GFP_NOWARN	((__force gfp_t)___GFP_NOWARN)
 #define __GFP_COMP	((__force gfp_t)___GFP_COMP)
 #define __GFP_ZERO	((__force gfp_t)___GFP_ZERO)
+#define __GFP_ZEROTAGS	((__force gfp_t)___GFP_ZEROTAGS)
 
 /* Disable lockdep for GFP context tracking */
 #define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
 
 /* Room for N __GFP_FOO bits */
-#define __GFP_BITS_SHIFT (23 + IS_ENABLED(CONFIG_LOCKDEP))
+#define __GFP_BITS_SHIFT (24 + IS_ENABLED(CONFIG_LOCKDEP))
 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 
 /**
include/linux/highmem.h +8 -0
index 832b49b50c7b..caaa62e1dd24 100644
@@ -204,6 +204,14 @@ static inline void clear_highpage(struct page *page)
 	kunmap_atomic(kaddr);
 }
 
+#ifndef __HAVE_ARCH_TAG_CLEAR_HIGHPAGE
+
+static inline void tag_clear_highpage(struct page *page)
+{
+}
+
+#endif
+
 /*
  * If we pass in a base or tail page, we can zero up to PAGE_SIZE.
  * If we pass in a head page, we can zero up to the size of the compound page.
mm/kasan/hw_tags.c +8 -1
index 45e552cb9172..34362c8d0955 100644
@@ -242,7 +242,14 @@ void kasan_alloc_pages(struct page *page, unsigned int order, gfp_t flags)
 {
 	bool init = !want_init_on_free() && want_init_on_alloc(flags);
 
-	kasan_unpoison_pages(page, order, init);
+	if (flags & __GFP_ZEROTAGS) {
+		int i;
+
+		for (i = 0; i != 1 << order; ++i)
+			tag_clear_highpage(page + i);
+	} else {
+		kasan_unpoison_pages(page, order, init);
+	}
 }
 
 void kasan_free_pages(struct page *page, unsigned int order)
mm/page_alloc.c +10 -3
index 6e82a7f6fd6f..24e6f668ef73 100644
@@ -1219,10 +1219,16 @@ static int free_tail_pages_check(struct page *head_page, struct page *page)
 	return ret;
 }
 
-static void kernel_init_free_pages(struct page *page, int numpages)
+static void kernel_init_free_pages(struct page *page, int numpages, bool zero_tags)
 {
 	int i;
 
+	if (zero_tags) {
+		for (i = 0; i < numpages; i++)
+			tag_clear_highpage(page + i);
+		return;
+	}
+
 	/* s390's use of memset() could override KASAN redzones. */
 	kasan_disable_current();
 	for (i = 0; i < numpages; i++) {

@@ -1314,7 +1320,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
 		bool init = want_init_on_free();
 
 		if (init)
-			kernel_init_free_pages(page, 1 << order);
+			kernel_init_free_pages(page, 1 << order, false);
 		if (!skip_kasan_poison)
 			kasan_poison_pages(page, order, init);
 	}

@@ -2350,7 +2356,8 @@ inline void post_alloc_hook(struct page *page, unsigned int order,
 
 		kasan_unpoison_pages(page, order, init);
 		if (init)
-			kernel_init_free_pages(page, 1 << order);
+			kernel_init_free_pages(page, 1 << order,
+					       gfp_flags & __GFP_ZEROTAGS);
 	}
 
 	set_page_owner(page, order, gfp_flags);
Signature · 1 line
2.31.1.607.g51e8a6a459-goog
maintainer Reviewed-by on v2 2/3
#5 lore raw

Re: [PATCH v2 2/3] arm64: mte: handle tags zeroing at page allocation time

Quoted 9 lines · Peter Collingbourne

Currently, on an anonymous page fault, the kernel allocates a zeroed page and maps it in user space. If the mapping is tagged (PROT_MTE), set_pte_at() additionally clears the tags. It is, however, more efficient to clear the tags at the same time as zeroing the data on allocation. To avoid clearing the tags on any page (which may not be mapped as tagged), only do this if the vma flags contain VM_MTE. This requires introducing a new GFP flag that is used to determine whether to clear the tags.

Quoted 8 lines

The DC GZVA instruction with a 0 top byte (and 0 tag) requires top-byte-ignore. Set the TCR_EL1.{TBI1,TBID1} bits irrespective of whether KASAN_HW is enabled.

Signed-off-by Peter Collingbourne <[email protected]>
Co-developed-by Catalin Marinas <[email protected]>
Signed-off-by Catalin Marinas <[email protected]>
Link https://linux-review.googlesource.com/id/Id46dc94e30fe11474f7e54f5d65e7658dbdddb26
Reviewed-by Catalin Marinas <[email protected]>
author
#6 lore raw

PATCH v2 3/3 kasan: allow freed user page poisoning to be disabled with HW tags

In reply to Peter Collingbourne

Poisoning freed pages protects against kernel use-after-free. The likelihood of such a bug involving kernel pages is significantly higher than that for user pages. At the same time, poisoning freed pages can impose a significant performance cost, which cannot always be justified for user pages given the lower probability of finding a bug. Therefore, make it possible to configure the kernel to disable freed user page poisoning when using HW tags via the new kasan.skip_user_poison_on_free command line option.

Signed-off-by Peter Collingbourne <[email protected]>
Link https://linux-review.googlesource.com/id/I716846e2de8ef179f44e835770df7e6307be96c9
---
 include/linux/gfp.h            | 13 ++++++++++---
 include/linux/page-flags.h     |  9 +++++++++
 include/trace/events/mmflags.h |  9 ++++++++-
 mm/kasan/hw_tags.c             | 10 ++++++++++
 mm/page_alloc.c                | 12 +++++++-----
 5 files changed, 44 insertions(+), 9 deletions(-)
include/linux/gfp.h +10 -3
index 68ba237365dc..9a77e5660b07 100644
@@ -54,8 +54,9 @@ struct vm_area_struct;
 #define ___GFP_THISNODE		0x200000u
 #define ___GFP_ACCOUNT		0x400000u
 #define ___GFP_ZEROTAGS		0x800000u
+#define ___GFP_SKIP_KASAN_POISON	0x1000000u
 #ifdef CONFIG_LOCKDEP
-#define ___GFP_NOLOCKDEP	0x1000000u
+#define ___GFP_NOLOCKDEP	0x2000000u
 #else
 #define ___GFP_NOLOCKDEP	0
 #endif

@@ -233,17 +234,22 @@ struct vm_area_struct;
  *
  * %__GFP_ZEROTAGS returns a page with zeroed memory tags on success, if
  * __GFP_ZERO is set.
+ *
+ * %__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, and only if a command line option is set.
  */
 #define __GFP_NOWARN	((__force gfp_t)___GFP_NOWARN)
 #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)
 
 /* Disable lockdep for GFP context tracking */
 #define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
 
 /* Room for N __GFP_FOO bits */
-#define __GFP_BITS_SHIFT (24 + IS_ENABLED(CONFIG_LOCKDEP))
+#define __GFP_BITS_SHIFT (25 + IS_ENABLED(CONFIG_LOCKDEP))
 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 
 /**

@@ -320,7 +326,8 @@ struct vm_area_struct;
 #define GFP_NOWAIT	(__GFP_KSWAPD_RECLAIM)
 #define GFP_NOIO	(__GFP_RECLAIM)
 #define GFP_NOFS	(__GFP_RECLAIM | __GFP_IO)
-#define GFP_USER	(__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
+#define GFP_USER	(__GFP_RECLAIM | __GFP_IO | __GFP_FS | \
+			 __GFP_HARDWALL | __GFP_SKIP_KASAN_POISON)
 #define GFP_DMA		__GFP_DMA
 #define GFP_DMA32	__GFP_DMA32
 #define GFP_HIGHUSER	(GFP_USER | __GFP_HIGHMEM)
include/linux/page-flags.h +9 -0
index 04a34c08e0a6..40e2c5000585 100644
@@ -137,6 +137,9 @@ enum pageflags {
 #endif
 #ifdef CONFIG_64BIT
 	PG_arch_2,
+#endif
+#ifdef CONFIG_KASAN_HW_TAGS
+	PG_skip_kasan_poison,
 #endif
 	__NR_PAGEFLAGS,
 

@@ -443,6 +446,12 @@ TESTCLEARFLAG(Young, young, PF_ANY)
 PAGEFLAG(Idle, idle, PF_ANY)
 #endif
 
+#ifdef CONFIG_KASAN_HW_TAGS
+PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD)
+#else
+PAGEFLAG_FALSE(SkipKASanPoison)
+#endif
+
 /*
  * PageReported() is used to track reported free pages within the Buddy
  * allocator. We can use the non-atomic version of the test and set
include/trace/events/mmflags.h +8 -1
index 629c7a0eaff2..390270e00a1d 100644
@@ -85,6 +85,12 @@
 #define IF_HAVE_PG_ARCH_2(flag,string)
 #endif
 
+#ifdef CONFIG_KASAN_HW_TAGS
+#define IF_HAVE_PG_SKIP_KASAN_POISON(flag,string) ,{1UL << flag, string}
+#else
+#define IF_HAVE_PG_SKIP_KASAN_POISON(flag,string)
+#endif
+
 #define __def_pageflag_names						\
 	{1UL << PG_locked,		"locked"	},		\
 	{1UL << PG_waiters,		"waiters"	},		\

@@ -112,7 +118,8 @@ IF_HAVE_PG_UNCACHED(PG_uncached,	"uncached"	)		\
 IF_HAVE_PG_HWPOISON(PG_hwpoison,	"hwpoison"	)		\
 IF_HAVE_PG_IDLE(PG_young,		"young"		)		\
 IF_HAVE_PG_IDLE(PG_idle,		"idle"		)		\
-IF_HAVE_PG_ARCH_2(PG_arch_2,		"arch_2"	)
+IF_HAVE_PG_ARCH_2(PG_arch_2,		"arch_2"	)		\
+IF_HAVE_PG_SKIP_KASAN_POISON(PG_skip_kasan_poison, "skip_kasan_poison")
 
 #define show_page_flags(flags)						\
 	(flags) ? __print_flags(flags, "|",				\
mm/kasan/hw_tags.c +10 -0
index 34362c8d0955..954d5c2f7683 100644
@@ -238,10 +238,20 @@ struct kasan_track *kasan_get_free_track(struct kmem_cache *cache,
 	return &alloc_meta->free_track[0];
 }
 
+static bool skip_user_poison_on_free;
+static int __init skip_user_poison_on_free_param(char *buf)
+{
+	return kstrtobool(buf, &skip_user_poison_on_free);
+}
+early_param("kasan.skip_user_poison_on_free", skip_user_poison_on_free_param);
+
 void kasan_alloc_pages(struct page *page, unsigned int order, gfp_t flags)
 {
 	bool init = !want_init_on_free() && want_init_on_alloc(flags);
 
+	if (skip_user_poison_on_free && (flags & __GFP_SKIP_KASAN_POISON))
+		SetPageSkipKASanPoison(page);
+
 	if (flags & __GFP_ZEROTAGS) {
 		int i;
mm/page_alloc.c +7 -5
index 24e6f668ef73..2c3ac15ddd54 100644
@@ -394,11 +394,12 @@ static DEFINE_STATIC_KEY_TRUE(deferred_pages);
  * on-demand allocation and then freed again before the deferred pages
  * initialization is done, but this is not likely to happen.
  */
-static inline bool should_skip_kasan_poison(fpi_t fpi_flags)
+static inline bool should_skip_kasan_poison(struct page *page, fpi_t fpi_flags)
 {
 	return static_branch_unlikely(&deferred_pages) ||
 	       (!IS_ENABLED(CONFIG_KASAN_GENERIC) &&
-		(fpi_flags & FPI_SKIP_KASAN_POISON));
+		(fpi_flags & FPI_SKIP_KASAN_POISON)) ||
+	       PageSkipKASanPoison(page);
 }
 
 /* Returns true if the struct page for the pfn is uninitialised */

@@ -449,10 +450,11 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn)
 	return false;
 }
 #else
-static inline bool should_skip_kasan_poison(fpi_t fpi_flags)
+static inline bool should_skip_kasan_poison(struct page *page, fpi_t fpi_flags)
 {
 	return (!IS_ENABLED(CONFIG_KASAN_GENERIC) &&
-		(fpi_flags & FPI_SKIP_KASAN_POISON));
+		(fpi_flags & FPI_SKIP_KASAN_POISON)) ||
+	       PageSkipKASanPoison(page);
 }
 
 static inline bool early_page_uninitialised(unsigned long pfn)

@@ -1244,7 +1246,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
 			unsigned int order, bool check_free, fpi_t fpi_flags)
 {
 	int bad = 0;
-	bool skip_kasan_poison = should_skip_kasan_poison(fpi_flags);
+	bool skip_kasan_poison = should_skip_kasan_poison(page, fpi_flags);
 
 	VM_BUG_ON_PAGE(PageTail(page), page);
Signature · 1 line
2.31.1.607.g51e8a6a459-goog