* Re: [RFC v2 01/13] mm/kfence: Add a new kunit test test_use_after_free_read_nofault()
2024-09-19 2:55 ` [RFC v2 01/13] mm/kfence: Add a new kunit test test_use_after_free_read_nofault() Ritesh Harjani (IBM)
@ 2024-10-03 5:06 ` Ritesh Harjani
0 siblings, 0 replies; 2+ messages in thread
From: Ritesh Harjani @ 2024-10-03 5:06 UTC (permalink / raw)
To: linuxppc-dev, kasan-dev
Cc: Michael Ellerman, Nicholas Piggin, Madhavan Srinivasan,
Christophe Leroy, Hari Bathini, Aneesh Kumar K . V, Donet Tom,
Pavithra Prakash, Nirjhar Roy, LKML, Alexander Potapenko,
linux-mm, Heiko Carstens
Hello Kasan/kfence-devs,
Wanted your inputs on this kfence kunit test [PATCH-1] and it's respective
powerpc fix [Patch-2]. The commit msgs has a good description of it. I
see that the same problem was noticed on s390 as well [1] a while ago.
So that makes me believe that maybe we should have a kunit test for the
same to make sure all architectures handles this properly.
Thoughts?
[1]: https://lore.kernel.org/all/20230213183858.1473681-1-hca@linux.ibm.com/
-ritesh
"Ritesh Harjani (IBM)" <ritesh.list@gmail.com> writes:
> From: Nirjhar Roy <nirjhar@linux.ibm.com>
>
> Faults from copy_from_kernel_nofault() needs to be handled by fixup
> table and should not be handled by kfence. Otherwise while reading
> /proc/kcore which uses copy_from_kernel_nofault(), kfence can generate
> false negatives. This can happen when /proc/kcore ends up reading an
> unmapped address from kfence pool.
>
> Let's add a testcase to cover this case.
>
> Co-developed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
> Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
> Signed-off-by: Nirjhar Roy <nirjhar@linux.ibm.com>
> Cc: kasan-dev@googlegroups.com
> Cc: Alexander Potapenko <glider@google.com>
> Cc: linux-mm@kvack.org
> ---
> mm/kfence/kfence_test.c | 17 +++++++++++++++++
> 1 file changed, 17 insertions(+)
>
> diff --git a/mm/kfence/kfence_test.c b/mm/kfence/kfence_test.c
> index 00fd17285285..f65fb182466d 100644
> --- a/mm/kfence/kfence_test.c
> +++ b/mm/kfence/kfence_test.c
> @@ -383,6 +383,22 @@ static void test_use_after_free_read(struct kunit *test)
> KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> }
>
> +static void test_use_after_free_read_nofault(struct kunit *test)
> +{
> + const size_t size = 32;
> + char *addr;
> + char dst;
> + int ret;
> +
> + setup_test_cache(test, size, 0, NULL);
> + addr = test_alloc(test, size, GFP_KERNEL, ALLOCATE_ANY);
> + test_free(addr);
> + /* Use after free with *_nofault() */
> + ret = copy_from_kernel_nofault(&dst, addr, 1);
> + KUNIT_EXPECT_EQ(test, ret, -EFAULT);
> + KUNIT_EXPECT_FALSE(test, report_available());
> +}
> +
> static void test_double_free(struct kunit *test)
> {
> const size_t size = 32;
> @@ -780,6 +796,7 @@ static struct kunit_case kfence_test_cases[] = {
> KFENCE_KUNIT_CASE(test_out_of_bounds_read),
> KFENCE_KUNIT_CASE(test_out_of_bounds_write),
> KFENCE_KUNIT_CASE(test_use_after_free_read),
> + KFENCE_KUNIT_CASE(test_use_after_free_read_nofault),
> KFENCE_KUNIT_CASE(test_double_free),
> KFENCE_KUNIT_CASE(test_invalid_addr_free),
> KFENCE_KUNIT_CASE(test_corruption),
> --
> 2.46.0
^ permalink raw reply [flat|nested] 2+ messages in thread