*   - PACKET_FANOUT_CPU
  *   - PACKET_FANOUT_ROLLOVER
  *   - PACKET_FANOUT_CBPF
+ *   - PACKET_FANOUT_EBPF
  *
  * Todo:
  * - functionality: PACKET_FANOUT_FLAG_DEFRAG
 #include <arpa/inet.h>
 #include <errno.h>
 #include <fcntl.h>
+#include <linux/unistd.h>      /* for __NR_bpf */
 #include <linux/filter.h>
+#include <linux/bpf.h>
 #include <linux/if_packet.h>
 #include <net/ethernet.h>
 #include <netinet/ip.h>
        return fd;
 }
 
+static void sock_fanout_set_ebpf(int fd)
+{
+       const int len_off = __builtin_offsetof(struct __sk_buff, len);
+       struct bpf_insn prog[] = {
+               { BPF_ALU64 | BPF_MOV | BPF_X,   6, 1, 0, 0 },
+               { BPF_LDX   | BPF_W   | BPF_MEM, 0, 6, len_off, 0 },
+               { BPF_JMP   | BPF_JGE | BPF_K,   0, 0, 1, DATA_LEN },
+               { BPF_JMP   | BPF_JA  | BPF_K,   0, 0, 4, 0 },
+               { BPF_LD    | BPF_B   | BPF_ABS, 0, 0, 0, 0x50 },
+               { BPF_JMP   | BPF_JEQ | BPF_K,   0, 0, 2, DATA_CHAR },
+               { BPF_JMP   | BPF_JEQ | BPF_K,   0, 0, 1, DATA_CHAR_1 },
+               { BPF_ALU   | BPF_MOV | BPF_K,   0, 0, 0, 0 },
+               { BPF_JMP   | BPF_EXIT,          0, 0, 0, 0 }
+       };
+       char log_buf[512];
+       union bpf_attr attr;
+       int pfd;
+
+       memset(&attr, 0, sizeof(attr));
+       attr.prog_type = BPF_PROG_TYPE_SOCKET_FILTER;
+       attr.insns = (unsigned long) prog;
+       attr.insn_cnt = sizeof(prog) / sizeof(prog[0]);
+       attr.license = (unsigned long) "GPL";
+       attr.log_buf = (unsigned long) log_buf,
+       attr.log_size = sizeof(log_buf),
+       attr.log_level = 1,
+
+       pfd = syscall(__NR_bpf, BPF_PROG_LOAD, &attr, sizeof(attr));
+       if (pfd < 0) {
+               perror("bpf");
+               fprintf(stderr, "bpf verifier:\n%s\n", log_buf);
+               exit(1);
+       }
+
+       if (setsockopt(fd, SOL_PACKET, PACKET_FANOUT_DATA, &pfd, sizeof(pfd))) {
+               perror("fanout data ebpf");
+               exit(1);
+       }
+
+       if (close(pfd)) {
+               perror("close ebpf");
+               exit(1);
+       }
+}
+
 static char *sock_fanout_open_ring(int fd)
 {
        struct tpacket_req req = {
        }
        if (type == PACKET_FANOUT_CBPF)
                sock_setfilter(fds[0], SOL_PACKET, PACKET_FANOUT_DATA);
+       else if (type == PACKET_FANOUT_EBPF)
+               sock_fanout_set_ebpf(fds[0]);
 
        rings[0] = sock_fanout_open_ring(fds[0]);
        rings[1] = sock_fanout_open_ring(fds[1]);
                             port_off, expect_lb[0], expect_lb[1]);
        ret |= test_datapath(PACKET_FANOUT_ROLLOVER,
                             port_off, expect_rb[0], expect_rb[1]);
+
        ret |= test_datapath(PACKET_FANOUT_CBPF,
                             port_off, expect_bpf[0], expect_bpf[1]);
+       ret |= test_datapath(PACKET_FANOUT_EBPF,
+                            port_off, expect_bpf[0], expect_bpf[1]);
 
        set_cpuaffinity(0);
        ret |= test_datapath(PACKET_FANOUT_CPU, port_off,