On 3/22/23 16:42, Andrii Nakryiko wrote:
On Tue, Mar 21, 2023 at 4:28 PM Kui-Feng Lee <kuifeng@xxxxxxxx> wrote:Introduce bpf_link__update_map(), which allows to atomically update underlying struct_ops implementation for given struct_ops BPF link Signed-off-by: Kui-Feng Lee <kuifeng@xxxxxxxx> --- tools/lib/bpf/bpf.h | 5 ++++- tools/lib/bpf/libbpf.c | 35 +++++++++++++++++++++++++++++++++++ tools/lib/bpf/libbpf.h | 1 + tools/lib/bpf/libbpf.map | 1 + 4 files changed, 41 insertions(+), 1 deletion(-) diff --git a/tools/lib/bpf/bpf.h b/tools/lib/bpf/bpf.h index f0f786373238..4fae4e698a8e 100644 --- a/tools/lib/bpf/bpf.h +++ b/tools/lib/bpf/bpf.h @@ -335,7 +335,10 @@ LIBBPF_API int bpf_link_detach(int link_fd); struct bpf_link_update_opts { size_t sz; /* size of this struct for forward/backward compatibility */ __u32 flags; /* extra flags */ - __u32 old_prog_fd; /* expected old program FD */ + union { + __u32 old_prog_fd; /* expected old program FD */ + __u32 old_map_fd; /* expected old map FD */ + };so for these low-level wrappers in libbpf with OPTS we've been trying to avoid unnecessary unions. If you look at bpf_link_create and bpf_link_create_ops some fields that are in a union in kernel UAPI are actually listed as separate fields, and libbpf makes sure that both fields are not specified at the same time (like iter_info_len and target_btf_id, for instance). So let's do the same here, instead of making a union, let's have __u32 old_prog_fd; __u32 old_map_fd; and then in bpf_link_update() implementation make sure that both can't be set at the same time.
Got it! Thanks!
The rest of the patch looks good to me, thanks....
