Re: [PATCH v2] kasan: fix more unit tests with CONFIG_UBSAN_LOCAL_BOUNDS enabled

From: Andrey Konovalov
Date: Tue Mar 08 2022 - 08:45:13 EST


On Thu, Feb 24, 2022 at 1:20 AM Peter Collingbourne <pcc@xxxxxxxxxx> wrote:
>
> This is a followup to commit f649dc0e0d7b ("kasan: fix unit tests
> with CONFIG_UBSAN_LOCAL_BOUNDS enabled") that fixes tests that fail
> as a result of __alloc_size annotations being added to the kernel
> allocator functions.
>
> Link: https://linux-review.googlesource.com/id/I4334cafc5db600fda5cebb851b2ee9fd09fb46cc
> Signed-off-by: Peter Collingbourne <pcc@xxxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> # 5.16.x
> Fixes: c37495d6254c ("slab: add __alloc_size attributes for better bounds checking")
> ---
> v2:
> - use OPTIMIZER_HIDE_VAR instead of volatile
>
> lib/test_kasan.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/lib/test_kasan.c b/lib/test_kasan.c
> index 26a5c9007653..7c3dfb569445 100644
> --- a/lib/test_kasan.c
> +++ b/lib/test_kasan.c
> @@ -185,6 +185,7 @@ static void kmalloc_pagealloc_oob_right(struct kunit *test)
> ptr = kmalloc(size, GFP_KERNEL);
> KUNIT_ASSERT_NOT_ERR_OR_NULL(test, ptr);
>
> + OPTIMIZER_HIDE_VAR(ptr);
> KUNIT_EXPECT_KASAN_FAIL(test, ptr[size + OOB_TAG_OFF] = 0);
>
> kfree(ptr);
> @@ -295,6 +296,7 @@ static void krealloc_more_oob_helper(struct kunit *test,
> KUNIT_EXPECT_KASAN_FAIL(test, ptr2[size2] = 'x');
>
> /* For all modes first aligned offset after size2 must be inaccessible. */
> + OPTIMIZER_HIDE_VAR(ptr2);
> KUNIT_EXPECT_KASAN_FAIL(test,
> ptr2[round_up(size2, KASAN_GRANULE_SIZE)] = 'x');
>
> @@ -319,6 +321,8 @@ static void krealloc_less_oob_helper(struct kunit *test,
> /* Must be accessible for all modes. */
> ptr2[size2 - 1] = 'x';
>
> + OPTIMIZER_HIDE_VAR(ptr2);
> +
> /* Generic mode is precise, so unaligned size2 must be inaccessible. */
> if (IS_ENABLED(CONFIG_KASAN_GENERIC))
> KUNIT_EXPECT_KASAN_FAIL(test, ptr2[size2] = 'x');
> --
> 2.35.1.473.g83b2b277ed-goog
>

Acked-by: Andrey Konovalov <andreyknvl@xxxxxxxxx>

This patch seems to be in partial conflict with the "kasan: test:
Silence allocation warnings from GCC 12" patch by Kees, which is
already in mm.

Thanks!