On Wed, Mar 13, 2024 at 10:24 AM Quentin Monnet <quentin@xxxxxxxxxxxxx> wrote:
>
> This reverts commit df7c3f7d3a3ddab31ca8cfa9b86a8729ec43fd2e.
>
> There's no guarantee that obj->btf is non-NULL when passing it to
> btf__fd(), and this function doesn't perform any check before
> dereferencing its argument. As a consequence, this commit caused
> segmentation faults in bpftool (for example) when trying to load
> programs that come without BTF info.
>
> Signed-off-by: Quentin Monnet <quentin@xxxxxxxxxxxxx>
> ---
> tools/lib/bpf/libbpf.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index efab29b8935b..0f9a4d232468 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -7343,7 +7343,7 @@ static int bpf_object_load_prog(struct bpf_object *obj, struct bpf_program *prog
> load_attr.prog_ifindex = prog->prog_ifindex;
>
> /* specify func_info/line_info only if kernel supports them */
> - btf_fd = btf__fd(obj->btf);
> + btf_fd = bpf_object__btf_fd(obj);
oh, my bad, sorry about that. But I'd still like to stick to using
btf__fd(), so maybe let's fix forward by dropping btf_fd variable here
and doing what bpf_object__create_map() does:
if (obj->btf && btf__fd(obj->btf) >= 0 && kernel_supports(...)) {
load_attr.prog_btf_fd = btf__fd(obj->btf);
...
}
?
> if (btf_fd >= 0 && kernel_supports(obj, FEAT_BTF_FUNC)) {
> load_attr.prog_btf_fd = btf_fd;
> load_attr.func_info = prog->func_info;
> --
> 2.34.1
>
