BPF(9) Kernel Developer's Manual BPF(9)

bpf
Berkeley Packet Filter kernel interface

#include <sys/mbuf.h>
#include <net/bpf.h>

bpf is called via an operations vector described by the following struct:
struct bpf_ops {
        void (*bpf_attach)(struct ifnet *, u_int, u_int, struct bpf_if **);
        void (*bpf_detach)(struct ifnet *);
        void (*bpf_change_type)(struct ifnet *, u_int, u_int);

        void (*bpf_mtap)(struct bpf_if *, struct mbuf *);
        void (*bpf_mtap2)(struct bpf_if *, void *, u_int, struct mbuf *);
        void (*bpf_mtap_af)(struct bpf_if *, uint32_t, struct mbuf *);
        void (*bpf_mtap_et)(struct bpf_if *, uint16_t, struct mbuf *);
        void (*bpf_mtap_sl_in)(struct bpf_if *, u_char *, struct mbuf **);
        void (*bpf_mtap_sl_out)(struct bpf_if *, u_char *, struct mbuf *);
};

This vector is found from struct bpf_ops * bpf_ops. It is either a “stub” implementation or a real implementation, depending on whether bpf is absent or present in the kernel, respectively.

The bpf implementation is located in sys/net/bpf.c, the stub implementation is in sys/net/bpf_stub.c, and sys/net/bpf.h describes the interface.
June 25, 2018 NetBSD 9.0