[PATCH bpf-next] bpf: fix static checker warning

[PATCH bpf-next] bpf: fix static checker warning

[Alexei Starovoitov]

kernel/bpf/btf.c:4023 btf_distill_func_proto()
        error: potentially dereferencing uninitialized 't'.

kernel/bpf/btf.c
  4012          nargs = btf_type_vlen(func);
  4013          if (nargs >= MAX_BPF_FUNC_ARGS) {
  4014                  bpf_log(log,
  4015                          "The function %s has %d arguments. Too many.\n",
  4016                          tname, nargs);
  4017                  return -EINVAL;
  4018          }
  4019          ret = __get_type_size(btf, func->type, &t);
                                                       ^^
t isn't initialized for the first -EINVAL return

This is unlikely path, since BTF should have been validated at this point.
Fix it by returning 'void' BTF.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 kernel/bpf/btf.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 40efde5eedcb..bd5e11881ba3 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -3976,8 +3976,10 @@ static int __get_type_size(struct btf *btf, u32 btf_id,
 	t = btf_type_by_id(btf, btf_id);
 	while (t && btf_type_is_modifier(t))
 		t = btf_type_by_id(btf, t->type);
-	if (!t)
+	if (!t) {
+		*bad_type = btf->types[0];
 		return -EINVAL;
+	}
 	if (btf_type_is_ptr(t))
 		/* kernel size of pointer. Not BPF's size of pointer*/
 		return sizeof(void *);
-- 
2.23.0

Re: [PATCH bpf-next] bpf: fix static checker warning

[Daniel Borkmann]

On 11/27/19 12:01 AM, Alexei Starovoitov wrote:
> kernel/bpf/btf.c:4023 btf_distill_func_proto()
>          error: potentially dereferencing uninitialized 't'.
> 
> kernel/bpf/btf.c
>    4012          nargs = btf_type_vlen(func);
>    4013          if (nargs >= MAX_BPF_FUNC_ARGS) {
>    4014                  bpf_log(log,
>    4015                          "The function %s has %d arguments. Too many.\n",
>    4016                          tname, nargs);
>    4017                  return -EINVAL;
>    4018          }
>    4019          ret = __get_type_size(btf, func->type, &t);
>                                                         ^^
> t isn't initialized for the first -EINVAL return
> 
> This is unlikely path, since BTF should have been validated at this point.
> Fix it by returning 'void' BTF.
> 
> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Applied, thanks!