Re: [bpf-next PATCH v2] bpf: Add rx_queue_mapping to bpf_sock

[Daniel Borkmann <daniel@iogearbox.net>]

On 5/22/20 8:28 PM, Amritha Nambiar wrote:
> Add "rx_queue_mapping" to bpf_sock. This gives read access for the
> existing field (sk_rx_queue_mapping) of struct sock from bpf_sock.
> Semantics for the bpf_sock rx_queue_mapping access are similar to
> sk_rx_queue_get(), i.e the value NO_QUEUE_MAPPING is not allowed
> and -1 is returned in that case.

This adds the "what this patch does" but could you also add a description for
the use-case in here?

> v2: fixed build error for CONFIG_XPS wrapping, reported by
>      kbuild test robot <lkp@intel.com>
> 
> Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
> ---
>   include/uapi/linux/bpf.h |    1 +
>   net/core/filter.c        |   18 ++++++++++++++++++
>   2 files changed, 19 insertions(+)
> 
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index 97e1fd19ff58..d2acd5aeae8d 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -3530,6 +3530,7 @@ struct bpf_sock {
>   	__u32 dst_ip4;
>   	__u32 dst_ip6[4];
>   	__u32 state;
> +	__u32 rx_queue_mapping;
>   };
>   
>   struct bpf_tcp_sock {
> diff --git a/net/core/filter.c b/net/core/filter.c
> index bd2853d23b50..c4ba92204b73 100644
> --- a/net/core/filter.c
> +++ b/net/core/filter.c
> @@ -6829,6 +6829,7 @@ bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
>   	case offsetof(struct bpf_sock, protocol):
>   	case offsetof(struct bpf_sock, dst_port):
>   	case offsetof(struct bpf_sock, src_port):
> +	case offsetof(struct bpf_sock, rx_queue_mapping):
>   	case bpf_ctx_range(struct bpf_sock, src_ip4):
>   	case bpf_ctx_range_till(struct bpf_sock, src_ip6[0], src_ip6[3]):
>   	case bpf_ctx_range(struct bpf_sock, dst_ip4):
> @@ -7872,6 +7873,23 @@ u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
>   						    skc_state),
>   				       target_size));
>   		break;
> +	case offsetof(struct bpf_sock, rx_queue_mapping):
> +#ifdef CONFIG_XPS
> +		*insn++ = BPF_LDX_MEM(
> +			BPF_FIELD_SIZEOF(struct sock, sk_rx_queue_mapping),
> +			si->dst_reg, si->src_reg,
> +			bpf_target_off(struct sock, sk_rx_queue_mapping,
> +				       sizeof_field(struct sock,
> +						    sk_rx_queue_mapping),
> +				       target_size));
> +		*insn++ = BPF_JMP_IMM(BPF_JNE, si->dst_reg, NO_QUEUE_MAPPING,
> +				      1);
> +		*insn++ = BPF_MOV64_IMM(si->dst_reg, -1);
> +#else
> +		*insn++ = BPF_MOV64_IMM(si->dst_reg, 0);

This should be -1 as queue mapping as well if XPS is not configured, no?
Otherwise, how do you tell it apart from an actual mapping to 0 if XPS is
built-in?

> +		*target_size = 2;
> +#endif
> +		break;
>   	}
>   
>   	return insn - insn_buf;
>