Re: [PATCH libbpf] libbpf: check if pin_path was set even map fd exist

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Hangbin Liu <liuhangbin@gmail.com>
Cc: bpf <bpf@vger.kernel.org>, "Alexei Starovoitov" <ast@kernel.org>,
	"Daniel Borkmann" <daniel@iogearbox.net>,
	"Song Liu" <songliubraving@fb.com>, "Yonghong Song" <yhs@fb.com>,
	"Andrii Nakryiko" <andriin@fb.com>,
	Networking <netdev@vger.kernel.org>,
	"Toke Høiland-Jørgensen" <toke@redhat.com>
Subject: Re: [PATCH libbpf] libbpf: check if pin_path was set even map fd exist
Date: Fri, 2 Oct 2020 11:11:54 -0700	[thread overview]
Message-ID: <CAEf4BzZTBxCKp2JT0yuwBwWX-EuqdSAOM7Duz7ifkRzeYKbKJw@mail.gmail.com> (raw)
In-Reply-To: <20201002075750.1978298-1-liuhangbin@gmail.com>

On Fri, Oct 2, 2020 at 12:58 AM Hangbin Liu <liuhangbin@gmail.com> wrote:
>
> Say a user reuse map fd after creating a map manually and set the
> pin_path, then load the object via libbpf.
>
> In libbpf bpf_object__create_maps(), bpf_object__reuse_map() will
> return 0 if there is no pinned map in map->pin_path. Then after
> checking if map fd exist, we should also check if pin_path was set
> and do bpf_map__pin() instead of continue the loop.
>
> Fix it by creating map if fd not exist and continue checking pin_path
> after that.
>
> Suggested-by: Andrii Nakryiko <andrii.nakryiko@gmail.com>
> Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
> ---
>  tools/lib/bpf/libbpf.c | 75 +++++++++++++++++++++---------------------
>  1 file changed, 37 insertions(+), 38 deletions(-)
>

Please add a selftests that validates the logic you are going to rely on.

> +                                       targ_map = map->init_slots[j];
> +                                       fd = bpf_map__fd(targ_map);
> +                                       err = bpf_map_update_elem(map->fd, &j, &fd, 0);
> +                                       if (err) {
> +                                               err = -errno;
> +                                               pr_warn("map '%s': failed to initialize slot [%d] to map '%s' fd=%d: %d\n",
> +                                                       map->name, j, targ_map->name,
> +                                                       fd, err);

I just noticed that we don't zclose(map->fd) here, can you please fix
it with a separate patch along these changes? Thanks!

> +                                               goto err_out;
> +                                       }
> +                                       pr_debug("map '%s': slot [%d] set to map '%s' fd=%d\n",
> +                                               map->name, j, targ_map->name, fd);
> +                               }
> +                               zfree(&map->init_slots);
> +                               map->init_slots_sz = 0;

Let's move this slot initting logic into a helper function
(init_map_slots() or something like that? doesn't have to use
"bpf_object__" prefix as it is internal static function), that will
simplify overall flow.

> +                       }
> +               } else {
> +                       pr_debug("map '%s': skipping creation (preset fd=%d)\n",
> +                                map->name, map->fd);

to make diff a bit smaller, maybe let's keep the original order, but
do if/else instead of continuing:

if (map->fd >= 0) {
    pr_debug("skipping...");
} else {
   /* do the creation here */
}

/* pinning logic here */

>                 }
>
>                 if (map->pin_path && !map->pinned) {
> --
> 2.25.4
>