On Tue, Aug 29, 2023 at 3:19 AM Daan De Meyer <daan.j.demeyer@xxxxxxxxx> wrote:
>
> As prep for adding unix socket support to the cgroup sockaddr hooks,
> let's add a kfunc bpf_sock_addr_set() that allows modifying a sockaddr
> from bpf. While this is already possible for AF_INET and AF_INET6, we'll
> need this kfunc when we add unix socket support since modifying the
> address for those requires modifying both the address and the sockaddr
> length.
>
> We also add the necessary hook to make the new kfunc work properly.
>
> Signed-off-by: Daan De Meyer <daan.j.demeyer@xxxxxxxxx>
> ---
> kernel/bpf/btf.c | 3 +++
> net/core/filter.c | 42 +++++++++++++++++++++++++++++++++++++++++-
> 2 files changed, 44 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index 249657c466dd..157342eaa2bb 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -217,6 +217,7 @@ enum btf_kfunc_hook {
> m,
> BTF_KFUNC_HOOK_LWT,
> BTF_KFUNC_HOOK_NETFILTER,
> + BTF_KFUNC_HOOK_SOCK_ADDR,
Do we really need a new kfunc category?
Can BTF_KFUNC_HOOK_CGROUP_SKB be reused here?
or even BTF_KFUNC_HOOK_COMMON ?
struct bpf_sock_addr_kern * type of bpf_sock_addr_set_addr()
will prevent any other type being passed in here
and bpf_sock_addr_kern type is local to prog
run via __cgroup_bpf_run_filter_sock_addr.
