diff mbox series

[v2,2/2] kasan: Add strscpy() test to trigger tag fault on arm64

Message ID 20250318214035.481950-3-pcc@google.com (mailing list archive)
State Superseded
Headers show
Series string: Add load_unaligned_zeropad() code path to sized_strscpy() | expand

Commit Message

Peter Collingbourne March 18, 2025, 9:40 p.m. UTC 
From: Vincenzo Frascino <vincenzo.frascino@arm.com>

When we invoke strscpy() with a maximum size of N bytes, it assumes
that:
- It can always read N bytes from the source.
- It always write N bytes (zero-padded) to the destination.

On aarch64 with Memory Tagging Extension enabled if we pass an N that is
bigger then the source buffer, it triggers an MTE fault.

Implement a KASAN KUnit test that triggers the issue with the current
implementation of read_word_at_a_time() on aarch64 with MTE enabled.

Cc: Will Deacon <will@kernel.org>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Co-developed-by: Peter Collingbourne <pcc@google.com>
Signed-off-by: Peter Collingbourne <pcc@google.com>
Link: https://linux-review.googlesource.com/id/If88e396b9e7c058c1a4b5a252274120e77b1898a
---
v2:
- rebased
- fixed test failure

 mm/kasan/kasan_test_c.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)

Comments

Andrey Konovalov March 20, 2025, 5:25 p.m. UTC | #1 
On Tue, Mar 18, 2025 at 10:41 PM Peter Collingbourne <pcc@google.com> wrote:
>
> From: Vincenzo Frascino <vincenzo.frascino@arm.com>
>
> When we invoke strscpy() with a maximum size of N bytes, it assumes
> that:
> - It can always read N bytes from the source.
> - It always write N bytes (zero-padded) to the destination.
>
> On aarch64 with Memory Tagging Extension enabled if we pass an N that is
> bigger then the source buffer, it triggers an MTE fault.
>
> Implement a KASAN KUnit test that triggers the issue with the current
> implementation of read_word_at_a_time() on aarch64 with MTE enabled.
>
> Cc: Will Deacon <will@kernel.org>
> Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> Co-developed-by: Peter Collingbourne <pcc@google.com>
> Signed-off-by: Peter Collingbourne <pcc@google.com>
> Link: https://linux-review.googlesource.com/id/If88e396b9e7c058c1a4b5a252274120e77b1898a
> ---
> v2:
> - rebased
> - fixed test failure
>
>  mm/kasan/kasan_test_c.c | 31 ++++++++++++++++++++++++++++++-
>  1 file changed, 30 insertions(+), 1 deletion(-)
>
> diff --git a/mm/kasan/kasan_test_c.c b/mm/kasan/kasan_test_c.c
> index 59d673400085f..c4bb3ee497b54 100644
> --- a/mm/kasan/kasan_test_c.c
> +++ b/mm/kasan/kasan_test_c.c
> @@ -1570,7 +1570,9 @@ static void kasan_memcmp(struct kunit *test)
>  static void kasan_strings(struct kunit *test)
>  {
>         char *ptr;
> -       size_t size = 24;
> +       char *src, *src2;
> +       u8 tag;
> +       size_t size = 2 * KASAN_GRANULE_SIZE;
>
>         /*
>          * str* functions are not instrumented with CONFIG_AMD_MEM_ENCRYPT.
> @@ -1581,6 +1583,33 @@ static void kasan_strings(struct kunit *test)
>         ptr = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
>         KUNIT_ASSERT_NOT_ERR_OR_NULL(test, ptr);
>
> +       src = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
> +       strscpy(src, "f0cacc1a00000000f0cacc1a00000000", size);
> +
> +       tag = get_tag(src);
> +
> +       src2 = src + KASAN_GRANULE_SIZE;
> +
> +       /*
> +        * Shorten string and poison the granule after it so that the unaligned
> +        * read in strscpy() triggers a tag mismatch.
> +        */
> +       src[KASAN_GRANULE_SIZE - 1] = '\0';
> +       kasan_poison(src2, KASAN_GRANULE_SIZE, tag + 1, false);
> +
> +       /*
> +        * The expected size does not include the terminator '\0'
> +        * so it is (KASAN_GRANULE_SIZE - 2) ==
> +        * KASAN_GRANULE_SIZE - ("initial removed character" + "\0").
> +        */
> +       KUNIT_EXPECT_EQ(test, KASAN_GRANULE_SIZE - 2,
> +                       strscpy(ptr, src + 1, size));
> +
> +       /* Undo operations above. */
> +       src[KASAN_GRANULE_SIZE - 1] = '0';
> +       kasan_poison(src2, KASAN_GRANULE_SIZE, tag, false);
> +
> +       kfree(src);

I have trouble understanding what this code is doing...

So the goal is to call strcpy with such an address, that the first 8
bytes (partially) cover 2 granules, one accessible and the other is
not?

If so, can we not do something like:

src = kmalloc(KASAN_GRANULE_SIZE, GFP_KERNEL | __GFP_ZERO);
strscpy(src, "aabbcceeddeeffg\0", size);
strscpy(ptr, src + KASAN_GRANULE_SIZE - 2, sizeof(unsigned long));

Otherwise, this code needs more explanatory comments and it's probably
better to move it out to a helper function.

>         kfree(ptr);
>
>         /*
> --
> 2.49.0.395.g12beb8f557-goog
>
Peter Collingbourne March 21, 2025, 2:41 a.m. UTC | #2 
On Thu, Mar 20, 2025 at 10:25 AM Andrey Konovalov <andreyknvl@gmail.com> wrote:
>
> On Tue, Mar 18, 2025 at 10:41 PM Peter Collingbourne <pcc@google.com> wrote:
> >
> > From: Vincenzo Frascino <vincenzo.frascino@arm.com>
> >
> > When we invoke strscpy() with a maximum size of N bytes, it assumes
> > that:
> > - It can always read N bytes from the source.
> > - It always write N bytes (zero-padded) to the destination.
> >
> > On aarch64 with Memory Tagging Extension enabled if we pass an N that is
> > bigger then the source buffer, it triggers an MTE fault.
> >
> > Implement a KASAN KUnit test that triggers the issue with the current
> > implementation of read_word_at_a_time() on aarch64 with MTE enabled.
> >
> > Cc: Will Deacon <will@kernel.org>
> > Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
> > Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> > Co-developed-by: Peter Collingbourne <pcc@google.com>
> > Signed-off-by: Peter Collingbourne <pcc@google.com>
> > Link: https://linux-review.googlesource.com/id/If88e396b9e7c058c1a4b5a252274120e77b1898a
> > ---
> > v2:
> > - rebased
> > - fixed test failure
> >
> >  mm/kasan/kasan_test_c.c | 31 ++++++++++++++++++++++++++++++-
> >  1 file changed, 30 insertions(+), 1 deletion(-)
> >
> > diff --git a/mm/kasan/kasan_test_c.c b/mm/kasan/kasan_test_c.c
> > index 59d673400085f..c4bb3ee497b54 100644
> > --- a/mm/kasan/kasan_test_c.c
> > +++ b/mm/kasan/kasan_test_c.c
> > @@ -1570,7 +1570,9 @@ static void kasan_memcmp(struct kunit *test)
> >  static void kasan_strings(struct kunit *test)
> >  {
> >         char *ptr;
> > -       size_t size = 24;
> > +       char *src, *src2;
> > +       u8 tag;
> > +       size_t size = 2 * KASAN_GRANULE_SIZE;
> >
> >         /*
> >          * str* functions are not instrumented with CONFIG_AMD_MEM_ENCRYPT.
> > @@ -1581,6 +1583,33 @@ static void kasan_strings(struct kunit *test)
> >         ptr = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
> >         KUNIT_ASSERT_NOT_ERR_OR_NULL(test, ptr);
> >
> > +       src = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
> > +       strscpy(src, "f0cacc1a00000000f0cacc1a00000000", size);
> > +
> > +       tag = get_tag(src);
> > +
> > +       src2 = src + KASAN_GRANULE_SIZE;
> > +
> > +       /*
> > +        * Shorten string and poison the granule after it so that the unaligned
> > +        * read in strscpy() triggers a tag mismatch.
> > +        */
> > +       src[KASAN_GRANULE_SIZE - 1] = '\0';
> > +       kasan_poison(src2, KASAN_GRANULE_SIZE, tag + 1, false);
> > +
> > +       /*
> > +        * The expected size does not include the terminator '\0'
> > +        * so it is (KASAN_GRANULE_SIZE - 2) ==
> > +        * KASAN_GRANULE_SIZE - ("initial removed character" + "\0").
> > +        */
> > +       KUNIT_EXPECT_EQ(test, KASAN_GRANULE_SIZE - 2,
> > +                       strscpy(ptr, src + 1, size));
> > +
> > +       /* Undo operations above. */
> > +       src[KASAN_GRANULE_SIZE - 1] = '0';
> > +       kasan_poison(src2, KASAN_GRANULE_SIZE, tag, false);
> > +
> > +       kfree(src);
>
> I have trouble understanding what this code is doing...
>
> So the goal is to call strcpy with such an address, that the first 8
> bytes (partially) cover 2 granules, one accessible and the other is
> not?

The first 16 bytes, but yes.

> If so, can we not do something like:
>
> src = kmalloc(KASAN_GRANULE_SIZE, GFP_KERNEL | __GFP_ZERO);
> strscpy(src, "aabbcceeddeeffg\0", size);
> strscpy(ptr, src + KASAN_GRANULE_SIZE - 2, sizeof(unsigned long));

Yes, something like that should work as well. Let me send a v3.

Peter

> Otherwise, this code needs more explanatory comments and it's probably
> better to move it out to a helper function.
>
> >         kfree(ptr);
> >
> >         /*
> > --
> > 2.49.0.395.g12beb8f557-goog
> >
diff mbox series

Patch

diff --git a/mm/kasan/kasan_test_c.c b/mm/kasan/kasan_test_c.c
index 59d673400085f..c4bb3ee497b54 100644
--- a/mm/kasan/kasan_test_c.c
+++ b/mm/kasan/kasan_test_c.c
@@ -1570,7 +1570,9 @@  static void kasan_memcmp(struct kunit *test)
 static void kasan_strings(struct kunit *test)
 {
 	char *ptr;
-	size_t size = 24;
+	char *src, *src2;
+	u8 tag;
+	size_t size = 2 * KASAN_GRANULE_SIZE;
 
 	/*
 	 * str* functions are not instrumented with CONFIG_AMD_MEM_ENCRYPT.
@@ -1581,6 +1583,33 @@  static void kasan_strings(struct kunit *test)
 	ptr = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
 	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, ptr);
 
+	src = kmalloc(size, GFP_KERNEL | __GFP_ZERO);
+	strscpy(src, "f0cacc1a00000000f0cacc1a00000000", size);
+
+	tag = get_tag(src);
+
+	src2 = src + KASAN_GRANULE_SIZE;
+
+	/*
+	 * Shorten string and poison the granule after it so that the unaligned
+	 * read in strscpy() triggers a tag mismatch.
+	 */
+	src[KASAN_GRANULE_SIZE - 1] = '\0';
+	kasan_poison(src2, KASAN_GRANULE_SIZE, tag + 1, false);
+
+	/*
+	 * The expected size does not include the terminator '\0'
+	 * so it is (KASAN_GRANULE_SIZE - 2) ==
+	 * KASAN_GRANULE_SIZE - ("initial removed character" + "\0").
+	 */
+	KUNIT_EXPECT_EQ(test, KASAN_GRANULE_SIZE - 2,
+			strscpy(ptr, src + 1, size));
+
+	/* Undo operations above. */
+	src[KASAN_GRANULE_SIZE - 1] = '0';
+	kasan_poison(src2, KASAN_GRANULE_SIZE, tag, false);
+
+	kfree(src);
 	kfree(ptr);
 
 	/*