The patch titled
Subject: kmsan: introduce test_memcpy_initialized_gap()
has been added to the -mm mm-unstable branch. Its filename is
kmsan-introduce-test_memcpy_initialized_gap.patch
This patch will shortly appear at
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/kmsan-introduce-test_memcpy_initialized_gap.patch
This patch will later appear in the mm-unstable branch at
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next via the mm-everything
branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there every 2-3 working days
------------------------------------------------------
From: Alexander Potapenko <glider@xxxxxxxxxx>
Subject: kmsan: introduce test_memcpy_initialized_gap()
Date: Mon, 11 Sep 2023 16:57:02 +0200
Add a regression test for the special case where memcpy() previously
failed to correctly set the origins: if upon memcpy() four aligned
initialized bytes with a zero origin value ended up split between two
aligned four-byte chunks, one of those chunks could've received the zero
origin value even despite it contained uninitialized bytes from other
writes.
Link: https://lkml.kernel.org/r/20230911145702.2663753-4-glider@xxxxxxxxxx
Signed-off-by: Alexander Potapenko <glider@xxxxxxxxxx>
Suggested-by: Marco Elver <elver@xxxxxxxxxx>
Acked-by: Marco Elver <elver@xxxxxxxxxx>
Cc: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/kmsan/kmsan_test.c | 53 ++++++++++++++++++++++++++++++++++++++++
1 file changed, 53 insertions(+)
--- a/mm/kmsan/kmsan_test.c~kmsan-introduce-test_memcpy_initialized_gap
+++ a/mm/kmsan/kmsan_test.c
@@ -486,6 +486,58 @@ static void test_memcpy_aligned_to_unali
KUNIT_EXPECT_TRUE(test, report_matches(&expect));
}
+/*
+ * Test case: ensure that origin slots do not accidentally get overwritten with
+ * zeroes during memcpy().
+ *
+ * Previously, when copying memory from an aligned buffer to an unaligned one,
+ * if there were zero origins corresponding to zero shadow values in the source
+ * buffer, they could have ended up being copied to nonzero shadow values in the
+ * destination buffer:
+ *
+ * memcpy(0xffff888080a00000, 0xffff888080900002, 8)
+ *
+ * src (0xffff888080900002): ..xx .... xx..
+ * src origins: o111 0000 o222
+ * dst (0xffff888080a00000): xx.. ..xx
+ * dst origins: o111 0000
+ * (or 0000 o222)
+ *
+ * (here . stands for an initialized byte, and x for an uninitialized one.
+ *
+ * Ensure that this does not happen anymore, and for both destination bytes
+ * the origin is nonzero (i.e. KMSAN reports an error).
+ */
+static void test_memcpy_initialized_gap(struct kunit *test)
+{
+ EXPECTATION_UNINIT_VALUE_FN(expect, "test_memcpy_initialized_gap");
+ volatile char uninit_src[12];
+ volatile char dst[8] = { 0 };
+
+ kunit_info(
+ test,
+ "unaligned 4-byte initialized value gets a nonzero origin after memcpy() - (2 UMR reports)\n");
+
+ uninit_src[0] = 42;
+ uninit_src[1] = 42;
+ uninit_src[4] = 42;
+ uninit_src[5] = 42;
+ uninit_src[6] = 42;
+ uninit_src[7] = 42;
+ uninit_src[10] = 42;
+ uninit_src[11] = 42;
+ memcpy_noinline((void *)&dst[0], (void *)&uninit_src[2], 8);
+
+ kmsan_check_memory((void *)&dst[0], 4);
+ KUNIT_EXPECT_TRUE(test, report_matches(&expect));
+ report_reset();
+ kmsan_check_memory((void *)&dst[2], 4);
+ KUNIT_EXPECT_FALSE(test, report_matches(&expect));
+ report_reset();
+ kmsan_check_memory((void *)&dst[4], 4);
+ KUNIT_EXPECT_TRUE(test, report_matches(&expect));
+}
+
/* Generate test cases for memset16(), memset32(), memset64(). */
#define DEFINE_TEST_MEMSETXX(size) \
static void test_memset##size(struct kunit *test) \
@@ -579,6 +631,7 @@ static struct kunit_case kmsan_test_case
KUNIT_CASE(test_init_memcpy),
KUNIT_CASE(test_memcpy_aligned_to_aligned),
KUNIT_CASE(test_memcpy_aligned_to_unaligned),
+ KUNIT_CASE(test_memcpy_initialized_gap),
KUNIT_CASE(test_memset16),
KUNIT_CASE(test_memset32),
KUNIT_CASE(test_memset64),
_
Patches currently in -mm which might be from glider@xxxxxxxxxx are
kmsan-simplify-kmsan_internal_memmove_metadata.patch
kmsan-prevent-optimizations-in-memcpy-tests.patch
kmsan-merge-test_memcpy_aligned_to_unaligned2-together.patch
kmsan-introduce-test_memcpy_initialized_gap.patch
