Re: [PATCH] kcsan: test: Adjust "expect" allocation type for kmalloc_obj

[Kees Cook <kees@kernel.org>]

On Tue, Feb 24, 2026 at 11:39:45PM +0100, Marco Elver wrote:
> On Tue, 24 Feb 2026 at 22:48, Kees Cook <kees@kernel.org> wrote:
> >
> > On Tue, Feb 24, 2026 at 11:09:44AM +0100, Marco Elver wrote:
> > > On Mon, 23 Feb 2026 at 23:22, Kees Cook <kees@kernel.org> wrote:
> > > >
> > > > Instead of depending on the implicit case between a pointer to pointers
> > > > and pointer to arrays, use the assigned variable type for the allocation
> > > > type so they correctly match. Solves 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);
> > > >       |                ^
> > > >
> > > > 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>
> > > > ---
> > > > Cc: Marco Elver <elver@google.com>
> > > > Cc: Dmitry Vyukov <dvyukov@google.com>
> > > > Cc: <kasan-dev@googlegroups.com>
> > > > ---
> > > >  kernel/kcsan/kcsan_test.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/kernel/kcsan/kcsan_test.c b/kernel/kcsan/kcsan_test.c
> > > > index 79e655ea4ca1..056fa859ad9a 100644
> > > > --- a/kernel/kcsan/kcsan_test.c
> > > > +++ b/kernel/kcsan/kcsan_test.c
> > > > @@ -168,7 +168,7 @@ static bool __report_matches(const struct expect_report *r)
> > > >         if (!report_available())
> > > >                 return false;
> > > >
> > > > -       expect = kmalloc_obj(observed.lines);
> > > > +       expect = kmalloc_obj(*expect);
> > >
> > > This is wrong. Instead of allocating 3x512 bytes it's now only
> > > allocating 512 bytes, so we get OOB below with this change. 'expect'
> > > is a pointer to a 3-dimensional array of 512-char arrays (matching
> > > observed.lines).
> >
> > Why did running the kunit test not trip over this? :(
> >
> > Hmpf, getting arrays allocated without an explicit cast seems to be
> > impossible. How about this:
> >
> >
> > diff --git a/kernel/kcsan/kcsan_test.c b/kernel/kcsan/kcsan_test.c
> > index 056fa859ad9a..ae758150ccb9 100644
> > --- a/kernel/kcsan/kcsan_test.c
> > +++ b/kernel/kcsan/kcsan_test.c
> > @@ -168,7 +168,7 @@ static bool __report_matches(const struct expect_report *r)
> >         if (!report_available())
> >                 return false;
> >
> > -       expect = kmalloc_obj(*expect);
> > +       expect = (typeof(expect))kmalloc_obj(observed.lines);
> 
> That works - or why not revert it back to normal kmalloc? There's
> marginal benefit for kmalloc_obj() in this case, and this really is
> just a bunch of char buffers - not a complex object. If there's still
> a benefit to be had from kmalloc_obj() here, I'm fine with the typeof
> cast.

Honestly... it's because I can't figure out how to make a exclusion for
this (nor how to get multidimensional array types) in Coccinelle to
avoid this case. (So re-running the conversion script will keep trying
to change this case.) And it's the only place in the kernel doing this
kind of thing. :P

I've sent v2 with the cast and a better commit log describing what's
happening.

-- 
Kees Cook