From: Bill Wendling <morbo@google.com>
To: Andrey Ryabinin <ryabinin.a.a@gmail.com>,
Andrew Morton <akpm@linux-foundation.org>
Cc: Alexander Potapenko <glider@google.com>,
Andrey Konovalov <andreyknvl@gmail.com>,
Dmitry Vyukov <dvyukov@google.com>,
Vincenzo Frascino <vincenzo.frascino@arm.com>,
Kees Cook <kees@kernel.org>,
"Gustavo A. R. Silva" <gustavoars@kernel.org>,
kasan-dev@googlegroups.com, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, linux-hardening@vger.kernel.org,
thomas.weissschuh@linutronix.de,
Bill Wendling <morbo@google.com>
Subject: [PATCH v3] fortify: add KUnit tests for __counted_by and __counted_by_ptr
Date: Tue, 29 Sep 2026 07:20:16 +0000 [thread overview]
Message-ID: <20260929072016.2360806-1-morbo@google.com> (raw)
In-Reply-To: <20260929010031.2186255-1-morbo@google.com>
The '__counted_by' and '__counted_by_ptr' attributes associate a
flexible array member or pointer member with a struct field that holds
its element count. Supporting compilers use these annotations to
compute dynamic object sizes via '__builtin_dynamic_object_size()' for
runtime bounds checking with CONFIG_FORTIFY_SOURCE.
Add KUnit tests ('fortify_test_counted_by_flex' and
'fortify_test_counted_by_ptr', guarded by CONFIG_CC_HAS_COUNTED_BY and
CONFIG_CC_HAS_COUNTED_BY_PTR respectively) to verify that:
- '__builtin_dynamic_object_size()', if available, returns the expected
logical byte size for annotated flexible array and pointer members.
- Fortified operations ('memset()' and 'memchr()') succeed within the
logical bounds and detect out-of-bounds read and write accesses
beyond the annotated count.
Allocate the test buffers with extra physical capacity (2 * size) in
'noinline' helpers and hide the returned pointers with
OPTIMIZER_HIDE_VAR() so allocation-size attributes, physical slab
bounds, and compiler optimizations do not mask the '__counted_by' and
'__counted_by_ptr' checks.
Signed-off-by: Bill Wendling <morbo@google.com>
---
v3: - Use "IS_ENABLED(CONFIG...)" instead of "#ifdefs". It's a lot cleaner
and documents better when skipped.
- Use "kunit_kzalloc" and "struct_size" for the flexible array member.
- Use KUNIT_EXPECT_BDOS which skips the test if BDOS isn't available.
v2: - Move tests to the 'fortify' KUnit tests. It uses UBSAN, which is
what gets triggered by 'counted_by'.
---
lib/tests/fortify_kunit.c | 114 ++++++++++++++++++++++++++++++++++++++
1 file changed, 114 insertions(+)
diff --git a/lib/tests/fortify_kunit.c b/lib/tests/fortify_kunit.c
index 413cdbf3dc0d..8baf6dc96bab 100644
--- a/lib/tests/fortify_kunit.c
+++ b/lib/tests/fortify_kunit.c
@@ -1011,6 +1011,118 @@ static void fortify_test_kmemdup(struct kunit *test)
kfree(copy);
}
+struct counted_by_flex_struct {
+ size_t size;
+ int array[] __counted_by(size);
+};
+
+/*
+ * Allocate the struct out-of-line with extra physical capacity so that
+ * __alloc_size() and physical slab bounds do not mask the __counted_by()
+ * logical bounds check.
+ */
+static noinline struct counted_by_flex_struct *
+alloc_counted_by_flex_struct(struct kunit *test, size_t size)
+{
+ struct counted_by_flex_struct *s;
+
+ s = kunit_kzalloc(test, struct_size(s, array, 2 * size), GFP_KERNEL);
+ KUNIT_ASSERT_NOT_ERR_OR_NULL(test, s);
+
+ /* Intentionally fake the size so that we can trigger a trap. */
+ s->size = size;
+ return s;
+}
+
+static void fortify_test_counted_by_flex(struct kunit *test)
+{
+ size_t size = 128;
+ struct counted_by_flex_struct *s;
+ size_t elem_bytes = size * sizeof(s->array[0]);
+
+ if (!IS_ENABLED(CONFIG_CC_HAS_COUNTED_BY))
+ kunit_skip(test, "requires CONFIG_CC_HAS_COUNTED_BY");
+
+ s = alloc_counted_by_flex_struct(test, size);
+
+ OPTIMIZER_HIDE_VAR(s);
+ OPTIMIZER_HIDE_VAR(elem_bytes);
+
+ /* __builtin_dynamic_object_size() should return the logical length. */
+ KUNIT_EXPECT_BDOS(test, s->array, elem_bytes,
+ "struct counted_by_flex_struct");
+
+ /* Within-bounds write and read succeed. */
+ memset(s->array, 0x42, elem_bytes);
+ KUNIT_EXPECT_EQ(test, fortify_write_overflows, 0);
+ KUNIT_EXPECT_NOT_NULL(test, memchr(s->array, 0x42, elem_bytes));
+ KUNIT_EXPECT_EQ(test, fortify_read_overflows, 0);
+
+ /* Out-of-bounds write and read past logical size are caught. */
+ memset(s->array, 0x42, elem_bytes + 1);
+ KUNIT_EXPECT_EQ(test, fortify_write_overflows, 1);
+ KUNIT_EXPECT_NULL(test, memchr(s->array, 0x42, elem_bytes + 1));
+ KUNIT_EXPECT_EQ(test, fortify_read_overflows, 1);
+}
+
+struct counted_by_ptr_struct {
+ char *ptr __counted_by_ptr(size);
+ size_t size;
+};
+
+/*
+ * Allocate the struct out-of-line with extra physical capacity so that
+ * __alloc_size() and physical slab bounds do not mask the __counted_by_ptr()
+ * logical bounds check.
+ */
+static noinline struct counted_by_ptr_struct *
+alloc_counted_by_ptr_struct(struct kunit *test, size_t size)
+{
+ struct counted_by_ptr_struct *s;
+
+ s = kmalloc_obj(struct counted_by_ptr_struct);
+ KUNIT_ASSERT_NOT_ERR_OR_NULL(test, s);
+
+ /* Intentionally fake the size so that we can trigger a trap. */
+ s->size = size;
+ s->ptr = kzalloc(2 * size, GFP_KERNEL);
+ KUNIT_ASSERT_NOT_ERR_OR_NULL(test, s->ptr);
+
+ return s;
+}
+
+static void fortify_test_counted_by_ptr(struct kunit *test)
+{
+ size_t size = 128;
+ struct counted_by_ptr_struct *s;
+
+ if (!IS_ENABLED(CONFIG_CC_HAS_COUNTED_BY_PTR))
+ kunit_skip(test, "requires CONFIG_CC_HAS_COUNTED_BY_PTR");
+
+ s = alloc_counted_by_ptr_struct(test, size);
+
+ OPTIMIZER_HIDE_VAR(s);
+ OPTIMIZER_HIDE_VAR(size);
+
+ /* __builtin_dynamic_object_size() should return the logical length. */
+ KUNIT_EXPECT_BDOS(test, s->ptr, size, "struct counted_by_ptr_struct");
+
+ /* Within-bounds write and read succeed. */
+ memset(s->ptr, 0x42, size);
+ KUNIT_EXPECT_EQ(test, fortify_write_overflows, 0);
+ KUNIT_EXPECT_NOT_NULL(test, memchr(s->ptr, 0x42, size));
+ KUNIT_EXPECT_EQ(test, fortify_read_overflows, 0);
+
+ /* Out-of-bounds write and read past logical size are caught. */
+ memset(s->ptr, 0x42, size + 1);
+ KUNIT_EXPECT_EQ(test, fortify_write_overflows, 1);
+ KUNIT_EXPECT_NULL(test, memchr(s->ptr, 0x42, size + 1));
+ KUNIT_EXPECT_EQ(test, fortify_read_overflows, 1);
+
+ kfree(s->ptr);
+ kfree(s);
+}
+
static int fortify_test_init(struct kunit *test)
{
if (!IS_ENABLED(CONFIG_FORTIFY_SOURCE))
@@ -1054,6 +1166,8 @@ static struct kunit_case fortify_test_cases[] = {
KUNIT_CASE(fortify_test_memchr_inv),
KUNIT_CASE(fortify_test_memcmp),
KUNIT_CASE(fortify_test_kmemdup),
+ KUNIT_CASE(fortify_test_counted_by_flex),
+ KUNIT_CASE(fortify_test_counted_by_ptr),
{}
};
--
2.56.0.rc1.315.gc6ed9934b7-goog
prev parent reply other threads:[~2026-09-29 7:20 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-28 23:17 [PATCH] kasan: " Bill Wendling
2026-09-28 23:42 ` Andrew Morton
2026-09-29 0:59 ` Bill Wendling
2026-09-29 1:00 ` [PATCH] fortify: " Bill Wendling
2026-09-29 6:23 ` Thomas Weißschuh
2026-09-29 6:53 ` Bill Wendling
2026-09-29 7:20 ` Bill Wendling [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260929072016.2360806-1-morbo@google.com \
--to=morbo@google.com \
--cc=akpm@linux-foundation.org \
--cc=andreyknvl@gmail.com \
--cc=dvyukov@google.com \
--cc=glider@google.com \
--cc=gustavoars@kernel.org \
--cc=kasan-dev@googlegroups.com \
--cc=kees@kernel.org \
--cc=linux-hardening@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=ryabinin.a.a@gmail.com \
--cc=thomas.weissschuh@linutronix.de \
--cc=vincenzo.frascino@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®