On Thu, Apr 30, 2020 at 3:24 AM Eelco Chaudron <echaudro@xxxxxxxxxx> wrote:
>
> When the probe code was failing for any reason ENOTSUP was returned, even
> if this was due to no having enough lock space. This patch fixes this by
> returning EPERM to the user application, so it can respond and increase
> the RLIMIT_MEMLOCK size.
>
> Signed-off-by: Eelco Chaudron <echaudro@xxxxxxxxxx>
> ---
> tools/lib/bpf/libbpf.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 8f480e29a6b0..a62388a151d4 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -3381,8 +3381,13 @@ bpf_object__probe_caps(struct bpf_object *obj)
>
> for (i = 0; i < ARRAY_SIZE(probe_fn); i++) {
> ret = probe_fn[i](obj);
> - if (ret < 0)
> + if (ret < 0) {
> pr_debug("Probe #%d failed with %d.\n", i, ret);
> + if (ret == -EPERM) {
> + pr_perm_msg(ret);
> + return ret;
I think this is dangerous to do. This detection loop is not supposed
to return error to user if any of the features are missing. I'd feel
more comfortable if we split bpf_object__probe_name() into two tests:
one testing trivial program and another testing same program with
name. If the first one fails with EPERM -- then we can return error to
user. If anything else fails -- that's ok. Thoughts?
> + }
> + }
> }
>
> return 0;
>
