[2/4] kmsan: another take at fixing memcpy tests

  5478afc55a21 ("kmsan: fix memcpy tests") uses OPTIMIZER_HIDE_VAR() to
hide the uninitialized var from the compiler optimizations.

However OPTIMIZER_HIDE_VAR(uninit) enforces an immediate check of
@uninit, so memcpy tests did not actually check the behavior of memcpy(),
because they always contained a KMSAN report.

Replace OPTIMIZER_HIDE_VAR() with a file-local asm macro that just
clobbers the memory, and add a test case for memcpy() that does not
expect an error report.

Also reflow kmsan_test.c with clang-format.

Signed-off-by: Alexander Potapenko <glider@google.com>
---
 mm/kmsan/kmsan_test.c | 43 +++++++++++++++++++++++++++++++++++++------
 1 file changed, 37 insertions(+), 6 deletions(-)
  

On Wed, 1 Mar 2023 at 15:39, 'Alexander Potapenko' via kasan-dev
<kasan-dev@googlegroups.com> wrote:
>
> 5478afc55a21 ("kmsan: fix memcpy tests") uses OPTIMIZER_HIDE_VAR() to
> hide the uninitialized var from the compiler optimizations.
>
> However OPTIMIZER_HIDE_VAR(uninit) enforces an immediate check of
> @uninit, so memcpy tests did not actually check the behavior of memcpy(),
> because they always contained a KMSAN report.
>
> Replace OPTIMIZER_HIDE_VAR() with a file-local asm macro that just
> clobbers the memory, and add a test case for memcpy() that does not
> expect an error report.
>
> Also reflow kmsan_test.c with clang-format.
>
> Signed-off-by: Alexander Potapenko <glider@google.com>
> ---
>  mm/kmsan/kmsan_test.c | 43 +++++++++++++++++++++++++++++++++++++------
>  1 file changed, 37 insertions(+), 6 deletions(-)
>
> diff --git a/mm/kmsan/kmsan_test.c b/mm/kmsan/kmsan_test.c
> index 088e21a48dc4b..cc98a3f4e0899 100644
> --- a/mm/kmsan/kmsan_test.c
> +++ b/mm/kmsan/kmsan_test.c
> @@ -407,6 +407,36 @@ static void test_printk(struct kunit *test)
>         KUNIT_EXPECT_TRUE(test, report_matches(&expect));
>  }
>
> +/*
> + * Prevent the compiler from optimizing @var away. Without this, Clang may
> + * notice that @var is uninitialized and drop memcpy() calls that use it.
> + *
> + * There is OPTIMIZER_HIDE_VAR() in linux/compier.h that we cannot use here,
> + * because it is implemented as inline assembly receiving @var as a parameter
> + * and will enforce a KMSAN check.
> + */
> +#define DO_NOT_OPTIMIZE(var) asm("" ::: "memory")

That's just a normal "barrier()" - use that instead?

> +/*
> + * Test case: ensure that memcpy() correctly copies initialized values.
> + */
> +static void test_init_memcpy(struct kunit *test)
> +{
> +       EXPECTATION_NO_REPORT(expect);
> +       volatile int src;
> +       volatile int dst = 0;
> +
> +       // Ensure DO_NOT_OPTIMIZE() does not cause extra checks.

^^ this comment seems redundant now, given DO_NOT_OPTIMIZE() has a
comment (it's also using //-style comment).

> +       DO_NOT_OPTIMIZE(src);
> +       src = 1;
> +       kunit_info(
> +               test,
> +               "memcpy()ing aligned initialized src to aligned dst (no reports)\n");
> +       memcpy((void *)&dst, (void *)&src, sizeof(src));
> +       kmsan_check_memory((void *)&dst, sizeof(dst));
> +       KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> +}
> +
>  /*
>   * Test case: ensure that memcpy() correctly copies uninitialized values between
>   * aligned `src` and `dst`.
> @@ -420,7 +450,7 @@ static void test_memcpy_aligned_to_aligned(struct kunit *test)
>         kunit_info(
>                 test,
>                 "memcpy()ing aligned uninit src to aligned dst (UMR report)\n");
> -       OPTIMIZER_HIDE_VAR(uninit_src);
> +       DO_NOT_OPTIMIZE(uninit_src);
>         memcpy((void *)&dst, (void *)&uninit_src, sizeof(uninit_src));
>         kmsan_check_memory((void *)&dst, sizeof(dst));
>         KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> @@ -443,7 +473,7 @@ static void test_memcpy_aligned_to_unaligned(struct kunit *test)
>         kunit_info(
>                 test,
>                 "memcpy()ing aligned uninit src to unaligned dst (UMR report)\n");
> -       OPTIMIZER_HIDE_VAR(uninit_src);
> +       DO_NOT_OPTIMIZE(uninit_src);
>         memcpy((void *)&dst[1], (void *)&uninit_src, sizeof(uninit_src));
>         kmsan_check_memory((void *)dst, 4);
>         KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> @@ -467,13 +497,14 @@ static void test_memcpy_aligned_to_unaligned2(struct kunit *test)
>         kunit_info(
>                 test,
>                 "memcpy()ing aligned uninit src to unaligned dst - part 2 (UMR report)\n");
> -       OPTIMIZER_HIDE_VAR(uninit_src);
> +       DO_NOT_OPTIMIZE(uninit_src);
>         memcpy((void *)&dst[1], (void *)&uninit_src, sizeof(uninit_src));
>         kmsan_check_memory((void *)&dst[4], sizeof(uninit_src));
>         KUNIT_EXPECT_TRUE(test, report_matches(&expect));
>  }
>
> -static noinline void fibonacci(int *array, int size, int start) {
> +static noinline void fibonacci(int *array, int size, int start)
> +{
>         if (start < 2 || (start == size))
>                 return;
>         array[start] = array[start - 1] + array[start - 2];
> @@ -482,8 +513,7 @@ static noinline void fibonacci(int *array, int size, int start) {
>
>  static void test_long_origin_chain(struct kunit *test)
>  {
> -       EXPECTATION_UNINIT_VALUE_FN(expect,
> -                                   "test_long_origin_chain");
> +       EXPECTATION_UNINIT_VALUE_FN(expect, "test_long_origin_chain");
>         /* (KMSAN_MAX_ORIGIN_DEPTH * 2) recursive calls to fibonacci(). */
>         volatile int accum[KMSAN_MAX_ORIGIN_DEPTH * 2 + 2];
>         int last = ARRAY_SIZE(accum) - 1;
> @@ -515,6 +545,7 @@ static struct kunit_case kmsan_test_cases[] = {
>         KUNIT_CASE(test_uaf),
>         KUNIT_CASE(test_percpu_propagate),
>         KUNIT_CASE(test_printk),
> +       KUNIT_CASE(test_init_memcpy),
>         KUNIT_CASE(test_memcpy_aligned_to_aligned),
>         KUNIT_CASE(test_memcpy_aligned_to_unaligned),
>         KUNIT_CASE(test_memcpy_aligned_to_unaligned2),
> --
> 2.39.2.722.g9855ee24e9-goog
>
> --
> You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/20230301143933.2374658-2-glider%40google.com.
  

> > +#define DO_NOT_OPTIMIZE(var) asm("" ::: "memory")
>
> That's just a normal "barrier()" - use that instead?

Ok, will do (I still think I'd better hide it behind a macro so that
we can change the implementation of DO_NOT_OPTIMIZE in the future if
the compiler starts outsmarting us again.

> > +/*
> > + * Test case: ensure that memcpy() correctly copies initialized values.
> > + */
> > +static void test_init_memcpy(struct kunit *test)
> > +{
> > +       EXPECTATION_NO_REPORT(expect);
> > +       volatile int src;
> > +       volatile int dst = 0;
> > +
> > +       // Ensure DO_NOT_OPTIMIZE() does not cause extra checks.
>
> ^^ this comment seems redundant now, given DO_NOT_OPTIMIZE() has a
> comment (it's also using //-style comment).

Moved it to the test description:

/*
 * Test case: ensure that memcpy() correctly copies initialized values.
 * Also serves as a regression test to ensure DO_NOT_OPTIMIZE() does not cause
 * extra checks.
 */

I think it's still relevant here.