Re: [PATCH bpf-next v3 1/7] bpf: Add attach_type in bpf_link

[Daniel Borkmann <daniel@iogearbox.net>]

On 7/9/25 5:07 AM, Tao Chen wrote:
> Attach_type will be set when link created from user, it is better
> to record attach_type in bpf_link directly suggested by Andrii. So
> add the attach_type field in bpf_link and move the sleepable field to
> the end just to fill the byte hole.
> 
> Acked-by: Jiri Olsa <jolsa@kernel.org>
> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> ---
>   drivers/net/netkit.c           |  2 +-
>   include/linux/bpf.h            | 28 ++++++++++++++++-----------
>   kernel/bpf/bpf_iter.c          |  3 ++-
>   kernel/bpf/bpf_struct_ops.c    |  5 +++--
>   kernel/bpf/cgroup.c            |  4 ++--
>   kernel/bpf/net_namespace.c     |  2 +-
>   kernel/bpf/syscall.c           | 35 +++++++++++++++++++++-------------
>   kernel/bpf/tcx.c               |  3 ++-
>   kernel/bpf/trampoline.c        | 10 ++++++----
>   kernel/trace/bpf_trace.c       |  4 ++--
>   net/bpf/bpf_dummy_struct_ops.c |  3 ++-
>   net/core/dev.c                 |  3 ++-
>   net/core/sock_map.c            |  3 ++-
>   net/netfilter/nf_bpf_link.c    |  3 ++-
>   14 files changed, 66 insertions(+), 42 deletions(-)
[...]

> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> index 34dd90ec7fa..dd5070039de 100644
> --- a/include/linux/bpf.h
> +++ b/include/linux/bpf.h
> @@ -1729,12 +1729,10 @@ struct bpf_link {
>   	enum bpf_link_type type;
>   	const struct bpf_link_ops *ops;
>   	struct bpf_prog *prog;
> -	/* whether BPF link itself has "sleepable" semantics, which can differ
> -	 * from underlying BPF program having a "sleepable" semantics, as BPF
> -	 * link's semantics is determined by target attach hook
> -	 */
> -	bool sleepable;
> +
>   	u32 flags;
> +	enum bpf_attach_type attach_type;
> +
>   	/* rcu is used before freeing, work can be used to schedule that
>   	 * RCU-based freeing before that, so they never overlap
>   	 */
> @@ -1742,6 +1740,11 @@ struct bpf_link {
>   		struct rcu_head rcu;
>   		struct work_struct work;
>   	};
> +	/* whether BPF link itself has "sleepable" semantics, which can differ
> +	 * from underlying BPF program having a "sleepable" semantics, as BPF
> +	 * link's semantics is determined by target attach hook
> +	 */
> +	bool sleepable;
>   };

lgtm, it looks a bit weird at the end here after rcu/work but in future if other
attributes get added to the link obj and a new hole frees up we can move it up
again. Definitely good the rationale for the move is documented in the commit msg.

Acked-by: Daniel Borkmann <daniel@iogearbox.net>