]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
kcsan: test: Adjust "expect" allocation type for kmalloc_obj
authorKees Cook <kees@kernel.org>
Mon, 23 Feb 2026 22:01:56 +0000 (14:01 -0800)
committerKees Cook <kees@kernel.org>
Thu, 26 Feb 2026 17:54:08 +0000 (09:54 -0800)
commit795469820c638b4449f3bb90ee5e98ebccfbc480
tree2c6cc478d742d5b8869af8f710ff392ed1cfdbe7
parent4b44cbb264d0ed3f2f2bc2659db6ce45882f4670
kcsan: test: Adjust "expect" allocation type for kmalloc_obj

The call to kmalloc_obj(observed.lines) returns "char (*)[3][512]",
a pointer to the whole 2D array. But "expect" wants to be "char (*)[512]",
the decayed pointer type, as if it were observed.lines itself (though
without the "3" bounds). This produces the following build error:

../kernel/kcsan/kcsan_test.c: In function '__report_matches':
../kernel/kcsan/kcsan_test.c:171:16: error: assignment to 'char (*)[512]' from incompatible pointer type 'char (*)[3][512]'
[-Wincompatible-pointer-types]
  171 |         expect = kmalloc_obj(observed.lines);
      |                ^

Instead of changing the "expect" type to "char (*)[3][512]" and
requiring a dereference at each use (e.g. "(expect*)[0]"), just
explicitly cast the return to the desired type.

Note that I'm intentionally not switching back to byte-based "kmalloc"
here because I cannot find a way for the Coccinelle script (which will
be used going forward to catch future conversions) to exclude this case.

Tested with:

$ ./tools/testing/kunit/kunit.py run \
--kconfig_add CONFIG_DEBUG_KERNEL=y \
--kconfig_add CONFIG_KCSAN=y \
--kconfig_add CONFIG_KCSAN_KUNIT_TEST=y \
--arch=x86_64 --qemu_args '-smp 2' kcsan

Reported-by: Nathan Chancellor <nathan@kernel.org>
Fixes: 69050f8d6d07 ("treewide: Replace kmalloc with kmalloc_obj for non-scalar types")
Signed-off-by: Kees Cook <kees@kernel.org>
kernel/kcsan/kcsan_test.c