/src/ndpi/example/reader_util.c

Source
/*
 * reader_util.c
 *
 * Copyright (C) 2011-25 - ntop.org
 *
 * This file is part of nDPI, an open source deep packet inspection
 * library based on the OpenDPI and PACE technology by ipoque GmbH
 *
 * nDPI is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * nDPI is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with nDPI.  If not, see <http://www.gnu.org/licenses/>.
 *
 */

#include "ndpi_config.h"
#include "ndpi_api.h"

#include <stdlib.h>
#include <math.h>
#include <float.h>

#ifdef WIN32
#include <winsock2.h> /* winsock.h is included automatically */
#include <windows.h>
#include <ws2tcpip.h>
#include <process.h>
#include <io.h>
#ifndef DISABLE_NPCAP
#include <ip6_misc.h>
#endif
#else
#include <unistd.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#endif
#include <assert.h>
#include <sys/stat.h>

#include "reader_util.h"

#define SNAP                   0XAA
#define BSTP                   0x42     /* Bridge Spanning Tree Protocol */

/* Keep last 32 packets */
#define DATA_ANALUYSIS_SLIDING_WINDOW    32

/* mask for FCF */
#define WIFI_DATA                        0x2    /* 0000 0010 */
#define FCF_TYPE(fc)     (((fc) >> 2) & 0x3)    /* 0000 0011 = 0x3 */
#define FCF_SUBTYPE(fc)  (((fc) >> 4) & 0xF)    /* 0000 1111 = 0xF */
#define FCF_TO_DS(fc)        ((fc) & 0x0100)
#define FCF_FROM_DS(fc)      ((fc) & 0x0200)

/* mask for Bad FCF presence */
#define BAD_FCS                         0x50    /* 0101 0000 */

#define GTP_U_V1_PORT                  2152
#define NDPI_CAPWAP_DATA_PORT          5247
#define TZSP_PORT                      37008

#ifndef DLT_LINUX_SLL
#define DLT_LINUX_SLL  113
#endif

#include "ndpi_main.h"
#include "reader_util.h"
#include "ndpi_classify.h"

extern u_int8_t enable_flow_stats, enable_payload_analyzer;
extern u_int8_t verbose, human_readeable_string_len;
extern u_int8_t max_num_udp_dissected_pkts /* 24 */, max_num_tcp_dissected_pkts /* 80 */;
static u_int32_t flow_id = 0;
extern FILE *fingerprint_fp;
extern char *addr_dump_path;
extern u_int8_t enable_doh_dot_detection;
extern int malloc_size_stats;
extern int monitoring_enabled;

/* ****************************************************** */

struct flow_id_stats {
  u_int32_t flow_id;
  UT_hash_handle hh;   /* makes this structure hashable */
};

struct packet_id_stats {
  u_int32_t packet_id;
  UT_hash_handle hh;   /* makes this structure hashable */
};

struct payload_stats {
  u_int8_t *pattern;
  u_int8_t pattern_len;
  u_int16_t num_occurrencies;
  struct flow_id_stats *flows;
  struct packet_id_stats *packets;
  UT_hash_handle hh;   /* makes this structure hashable */
};


struct payload_stats *pstats = NULL;
u_int32_t max_num_packets_per_flow      = 10; /* ETTA requires min 10 pkts for record. */
u_int32_t max_packet_payload_dissection = 128;
u_int32_t max_num_reported_top_payloads = 25;
u_int16_t min_pattern_len               = 4;
u_int16_t max_pattern_len               = 8;


/* *********************************************************** */

int ndpi_analyze_payload(struct ndpi_flow_info *flow,
       u_int8_t *payload,
       u_int16_t payload_len,
       u_int32_t packet_id) {
  struct payload_stats *ret, *ret_found;
  struct flow_id_stats *f, *f_found;
  struct packet_id_stats *p, *p_found;

#ifdef DEBUG_PAYLOAD
  u_int16_t i;
  for(i=0; i<payload_len; i++)
    printf("%c", ndpi_isprint(payload[i]) ? payload[i] : '.');
  printf("\n");
#endif

  HASH_FIND(hh, pstats, payload, payload_len, ret);
  if(ret == NULL) {
    if((ret = (struct payload_stats*)ndpi_calloc(1, sizeof(struct payload_stats))) == NULL)
      return -1; /* OOM */

    if((ret->pattern = (u_int8_t*)ndpi_malloc(payload_len)) == NULL) {
      ndpi_free(ret);
      return -1;
    }

    memcpy(ret->pattern, payload, payload_len);
    ret->pattern_len = payload_len;
    ret->num_occurrencies = 1;

    HASH_ADD(hh, pstats, pattern[0], payload_len, ret);

    HASH_FIND(hh, pstats, payload, payload_len, ret_found);
    if(ret_found == NULL) { /* The insertion failed (because of a memory allocation error) */
      ndpi_free(ret->pattern);
      ndpi_free(ret);
      return -1;
    }

#ifdef DEBUG_PAYLOAD
    printf("Added element [total: %u]\n", HASH_COUNT(pstats));
#endif
  } else {
    ret->num_occurrencies++;
    // printf("==> %u\n", ret->num_occurrencies);
  }

  HASH_FIND_INT(ret->flows, &flow->flow_id, f);
  if(f == NULL) {
    if((f = (struct flow_id_stats*)ndpi_calloc(1, sizeof(struct flow_id_stats))) == NULL)
      return -1; /* OOM */

    f->flow_id = flow->flow_id;
    HASH_ADD_INT(ret->flows, flow_id, f);

    HASH_FIND_INT(ret->flows, &flow->flow_id, f_found);
    if(f_found == NULL) { /* The insertion failed (because of a memory allocation error) */
      ndpi_free(f);
      return -1;
    }
  }

  HASH_FIND_INT(ret->packets, &packet_id, p);
  if(p == NULL) {
    if((p = (struct packet_id_stats*)ndpi_calloc(1, sizeof(struct packet_id_stats))) == NULL)
      return -1; /* OOM */
    p->packet_id = packet_id;

    HASH_ADD_INT(ret->packets, packet_id, p);

    HASH_FIND_INT(ret->packets, &packet_id, p_found);
    if(p_found == NULL) { /* The insertion failed (because of a memory allocation error) */
      ndpi_free(p);
    }
  }
  return 0;
}

/* *********************************************************** */

void ndpi_payload_analyzer(struct ndpi_flow_info *flow,
         u_int8_t *payload, u_int16_t payload_len,
         u_int32_t packet_id) {
  u_int16_t i, j;
  u_int16_t scan_len = ndpi_min(max_packet_payload_dissection, payload_len);

  if((flow->src2dst_packets+flow->dst2src_packets) <= max_num_packets_per_flow) {
#ifdef DEBUG_PAYLOAD
    printf("[hashval: %u][proto: %u][vlan: %u][%s:%u <-> %s:%u][direction: %s][payload_len: %u]\n",
     flow->hashval, flow->protocol, flow->vlan_id,
     flow->src_name ? flow->src_name : "",
     flow->src_port,
     flow->dst_name ? flow->dst_name : "",
     flow->dst_port,
     src_to_dst_direction ? "s2d" : "d2s",
     payload_len);
#endif
  } else
    return;

  for(i=0; i<scan_len; i++) {
    for(j=min_pattern_len; j <= max_pattern_len; j++) {
      if((i+j) < payload_len) {
  if(ndpi_analyze_payload(flow, &payload[i], j, packet_id) == -1) {
          LOG(NDPI_LOG_ERROR, "Error ndpi_analyze_payload (allocation failure)\n");
  }
      }
    }
  }
}

/* ***************************************************** */

static int payload_stats_sort_asc(void *_a, void *_b) {
  struct payload_stats *a = (struct payload_stats *)_a;
  struct payload_stats *b = (struct payload_stats *)_b;

  //return(a->num_occurrencies - b->num_occurrencies);
  return(b->num_occurrencies - a->num_occurrencies);
}

/* ***************************************************** */

static void print_payload_stat(struct payload_stats *p, FILE *out) {
  u_int i;
  struct flow_id_stats *s, *tmp;
  struct packet_id_stats *s1, *tmp1;

  fprintf(out, "\t[");

  for(i=0; i<p->pattern_len; i++) {
    fprintf(out, "%c", ndpi_isprint(p->pattern[i]) ? p->pattern[i] : '.');
  }

  fprintf(out, "]");
  for(; i<16; i++) fprintf(out, " ");
  fprintf(out, "[");

  for(i=0; i<p->pattern_len; i++) {
    fprintf(out, "%s%02X", (i > 0) ? " " : "", ndpi_isprint(p->pattern[i]) ? p->pattern[i] : '.');
  }

  fprintf(out, "]");

  for(; i<16; i++) fprintf(out, "  ");
  for(i=p->pattern_len; i<max_pattern_len; i++) fprintf(out, " ");

  fprintf(out, "[len: %u][num_occurrencies: %u][flowId: ",
    p->pattern_len, p->num_occurrencies);

  i = 0;
  HASH_ITER(hh, p->flows, s, tmp) {
    fprintf(out, "%s%u", (i > 0) ? " " : "", s->flow_id);
    i++;
  }

  fprintf(out, "][packetIds: ");

  /* ******************************** */

  i = 0;
  HASH_ITER(hh, p->packets, s1, tmp1) {
    fprintf(out, "%s%u", (i > 0) ? " " : "", s1->packet_id);
    i++;
  }

  fprintf(out, "]\n");


}

/* ***************************************************** */

void ndpi_report_payload_stats(FILE *out) {
  struct payload_stats *p, *tmp;
  u_int num = 0;

  if(out)
    fprintf(out, "\n\nPayload Analysis\n");

  HASH_SORT(pstats, payload_stats_sort_asc);

  HASH_ITER(hh, pstats, p, tmp) {
    if(out && num <= max_num_reported_top_payloads)
      print_payload_stat(p, out);

    ndpi_free(p->pattern);

    {
      struct flow_id_stats *p1, *tmp1;

      HASH_ITER(hh, p->flows, p1, tmp1) {
  HASH_DEL(p->flows, p1);
  ndpi_free(p1);
      }
    }

    {
      struct packet_id_stats *p1, *tmp1;

      HASH_ITER(hh, p->packets, p1, tmp1) {
  HASH_DEL(p->packets, p1);
  ndpi_free(p1);
      }
    }

    HASH_DEL(pstats, p);
    ndpi_free(p);
    num++;
  }
}

/* ***************************************************** */

void ndpi_free_flow_info_half(struct ndpi_flow_info *flow) {
  if(flow->ndpi_flow) { ndpi_flow_free(flow->ndpi_flow); flow->ndpi_flow = NULL; }
}

/* ***************************************************** */

void ndpi_stats_free(ndpi_stats_t *s) {
  if (s->protocol_counter)           ndpi_free(s->protocol_counter);
  if (s->protocol_counter_bytes)     ndpi_free(s->protocol_counter_bytes);
  if (s->protocol_flows)             ndpi_free(s->protocol_flows);
  if (s->fpc_protocol_counter)       ndpi_free(s->fpc_protocol_counter);
  if (s->fpc_protocol_counter_bytes) ndpi_free(s->fpc_protocol_counter_bytes);
  if (s->fpc_protocol_flows)         ndpi_free(s->fpc_protocol_flows);

  s->num_protocols = 0;
}

int ndpi_stats_init(ndpi_stats_t *s, uint32_t num_protocols) {
  memset(s, 0, sizeof(*s));
  s->num_protocols = num_protocols;

  s->protocol_counter           = ndpi_calloc(num_protocols, sizeof(u_int64_t));
  s->protocol_counter_bytes     = ndpi_calloc(num_protocols, sizeof(u_int64_t));
  s->protocol_flows             = ndpi_calloc(num_protocols, sizeof(u_int32_t));
  s->fpc_protocol_counter       = ndpi_calloc(num_protocols, sizeof(u_int64_t));
  s->fpc_protocol_counter_bytes = ndpi_calloc(num_protocols, sizeof(u_int64_t));
  s->fpc_protocol_flows         = ndpi_calloc(num_protocols, sizeof(u_int32_t));

  if(!s->protocol_counter || !s->protocol_counter_bytes || !s->protocol_flows ||
     !s->fpc_protocol_counter || !s->fpc_protocol_counter_bytes || !s->fpc_protocol_flows) {

    LOG(NDPI_LOG_ERROR, "[NDPI] %s: error allocating memory for ndpi_stats\n", __FUNCTION__);
    return 0;
  }
  return 1;
}

void ndpi_stats_reset(ndpi_stats_t *s) {
  memset(s->flow_count, 0, sizeof(s->flow_count));
  s->guessed_flow_protocols = 0;
  s->raw_packet_count = 0;
  s->ip_packet_count = 0;
  s->total_wire_bytes = 0;
  s->total_ip_bytes = 0;
  s->total_discarded_bytes = 0;
  s->ndpi_flow_count = 0;
  s->tcp_count = 0;
  s->udp_count = 0;
  s->mpls_count = 0;
  s->pppoe_count = 0;
  s->vlan_count = 0;
  s->fragmented_count = 0;
  s->max_packet_len = 0;
  s->num_dissector_calls = 0;

  memset(s->packet_len, 0, sizeof(s->packet_len));
  memset(s->dpi_packet_count, 0, sizeof(s->dpi_packet_count));
  memset(s->flow_confidence, 0, sizeof(s->flow_confidence));
  memset(s->fpc_flow_confidence, 0, sizeof(s->fpc_flow_confidence));
  memset(s->category_counter, 0, sizeof(s->category_counter));
  memset(s->category_counter_bytes, 0, sizeof(s->category_counter_bytes));
  memset(s->category_flows, 0, sizeof(s->category_flows));
  memset(s->lru_stats, 0, sizeof(s->lru_stats));
  memset(s->automa_stats, 0, sizeof(s->automa_stats));
  memset(s->patricia_stats, 0, sizeof(s->patricia_stats));
  memset(s->hash_stats, 0, sizeof(s->hash_stats));

  if (s->protocol_counter)           memset(s->protocol_counter,           0, sizeof(u_int64_t) * s->num_protocols);
  if (s->protocol_counter_bytes)     memset(s->protocol_counter_bytes,     0, sizeof(u_int64_t) * s->num_protocols);
  if (s->protocol_flows)             memset(s->protocol_flows,             0, sizeof(u_int32_t) * s->num_protocols);
  if (s->fpc_protocol_counter)       memset(s->fpc_protocol_counter,       0, sizeof(u_int64_t) * s->num_protocols);
  if (s->fpc_protocol_counter_bytes) memset(s->fpc_protocol_counter_bytes, 0, sizeof(u_int64_t) * s->num_protocols);
  if (s->fpc_protocol_flows)         memset(s->fpc_protocol_flows,         0, sizeof(u_int32_t) * s->num_protocols);
}

/* ***************************************************** */

struct ndpi_workflow* ndpi_workflow_init(const struct ndpi_workflow_prefs * prefs,
           pcap_t * pcap_handle, int do_init_flows_root,
           ndpi_serialization_format serialization_format,
           struct ndpi_global_context *g_ctx) {
  struct ndpi_detection_module_struct * module;
  struct ndpi_workflow * workflow;

  module = ndpi_init_detection_module(g_ctx);

  if(module == NULL) {
    LOG(NDPI_LOG_ERROR, "global structure initialization failed\n");
    return NULL;
  }

  workflow = ndpi_calloc(1, sizeof(struct ndpi_workflow));
  if(workflow == NULL) {
    LOG(NDPI_LOG_ERROR, "global structure initialization failed\n");
    ndpi_exit_detection_module(module);
    return NULL;
  }

  workflow->pcap_handle = pcap_handle;
  workflow->prefs       = *prefs;
  workflow->ndpi_struct = module;

  ndpi_set_user_data(module, workflow);

  if(do_init_flows_root) {
    workflow->ndpi_flows_root = ndpi_calloc(workflow->prefs.num_roots, sizeof(void *));

    if(!workflow->ndpi_flows_root) {
      ndpi_exit_detection_module(module);
      ndpi_free(workflow);
      return NULL;
    }
  }

  workflow->ndpi_serialization_format = serialization_format;

  return workflow;
}

/* ***************************************************** */

void ndpi_flow_info_freer(void *node) {
  struct ndpi_flow_info *flow = (struct ndpi_flow_info*)node;

  ndpi_flow_info_free_data(flow);
  ndpi_free(flow);
}

/* ***************************************************** */

static void ndpi_free_flow_tls_data(struct ndpi_flow_info *flow) {
  if(flow->dhcp_fingerprint) {
    ndpi_free(flow->dhcp_fingerprint);
    flow->dhcp_fingerprint = NULL;
  }

  if(flow->dhcp_class_ident) {
    ndpi_free(flow->dhcp_class_ident);
    flow->dhcp_class_ident = NULL;
  }

  if(flow->server_hostname) {
    ndpi_free(flow->server_hostname);
    flow->server_hostname = NULL;
  }

  if(flow->bittorent_hash) {
    ndpi_free(flow->bittorent_hash);
    flow->bittorent_hash = NULL;
  }

  if(flow->telnet.username) {
    ndpi_free(flow->telnet.username);
    flow->telnet.username = NULL;
  }

  if(flow->telnet.password) {
    ndpi_free(flow->telnet.password);
    flow->telnet.password = NULL;
  }

  if(flow->ssh_tls.server_names) {
    ndpi_free(flow->ssh_tls.server_names);
    flow->ssh_tls.server_names = NULL;
  }

  if(flow->ssh_tls.advertised_alpns) {
    ndpi_free(flow->ssh_tls.advertised_alpns);
    flow->ssh_tls.advertised_alpns = NULL;
  }

  if(flow->ssh_tls.negotiated_alpn) {
    ndpi_free(flow->ssh_tls.negotiated_alpn);
    flow->ssh_tls.negotiated_alpn = NULL;
  }

  if(flow->ssh_tls.tls_supported_versions) {
    ndpi_free(flow->ssh_tls.tls_supported_versions);
    flow->ssh_tls.tls_supported_versions = NULL;
  }

  if(flow->ssh_tls.tls_issuerDN) {
    ndpi_free(flow->ssh_tls.tls_issuerDN);
    flow->ssh_tls.tls_issuerDN = NULL;
  }

  if(flow->ssh_tls.tls_subjectDN) {
    ndpi_free(flow->ssh_tls.tls_subjectDN);
    flow->ssh_tls.tls_subjectDN = NULL;
  }

  if(flow->ssh_tls.ja4_client_raw) {
    ndpi_free(flow->ssh_tls.ja4_client_raw);
    flow->ssh_tls.ja4_client_raw = NULL;
  }

  if(flow->ndpi_fingerprint) {
    ndpi_free(flow->ndpi_fingerprint);
    flow->ndpi_fingerprint = NULL;
  }

  if(flow->stun.mapped_address.aps) {
    ndpi_free(flow->stun.mapped_address.aps);
    flow->stun.mapped_address.aps = NULL;
  }
  if(flow->stun.other_address.aps) {
    ndpi_free(flow->stun.other_address.aps);
    flow->stun.other_address.aps = NULL;
  }
  if(flow->stun.peer_address.aps) {
    ndpi_free(flow->stun.peer_address.aps);
    flow->stun.peer_address.aps = NULL;
  }
  if(flow->stun.relayed_address.aps) {
    ndpi_free(flow->stun.relayed_address.aps);
    flow->stun.relayed_address.aps = NULL;
  }
  if(flow->stun.response_origin.aps) {
    ndpi_free(flow->stun.response_origin.aps);
    flow->stun.response_origin.aps = NULL;
  }
}

/* ***************************************************** */

static void ndpi_free_flow_data_analysis(struct ndpi_flow_info *flow) {
  if(flow->iat_c_to_s) ndpi_free_data_analysis(flow->iat_c_to_s, 1);
  if(flow->iat_s_to_c) ndpi_free_data_analysis(flow->iat_s_to_c, 1);

  if(flow->pktlen_c_to_s) ndpi_free_data_analysis(flow->pktlen_c_to_s, 1);
  if(flow->pktlen_s_to_c) ndpi_free_data_analysis(flow->pktlen_s_to_c, 1);

  if(flow->iat_flow) ndpi_free_data_analysis(flow->iat_flow, 1);

  if(flow->entropy) ndpi_free(flow->entropy);
  if(flow->last_entropy) ndpi_free(flow->last_entropy);
}

/* ***************************************************** */

void ndpi_flow_info_free_data(struct ndpi_flow_info *flow) {

  ndpi_free_flow_info_half(flow);
  ndpi_term_serializer(&flow->ndpi_flow_serializer);
  ndpi_free_flow_data_analysis(flow);
  ndpi_free_flow_tls_data(flow);

#ifdef DIRECTION_BINS
  ndpi_free_bin(&flow->payload_len_bin_src2dst);
  ndpi_free_bin(&flow->payload_len_bin_dst2src);
#else
  ndpi_free_bin(&flow->payload_len_bin);
#endif

  if(flow->src_name)        ndpi_free(flow->src_name);
  if(flow->dst_name)        ndpi_free(flow->dst_name);
  if(flow->tcp_fingerprint) ndpi_free(flow->tcp_fingerprint);
  if(flow->risk_str)        ndpi_free(flow->risk_str);
  if(flow->flow_payload)    ndpi_free(flow->flow_payload);
}

/* ***************************************************** */

void ndpi_workflow_free(struct ndpi_workflow * workflow) {
  u_int i;

  for(i=0; i<workflow->prefs.num_roots; i++)
    ndpi_tdestroy(workflow->ndpi_flows_root[i], ndpi_flow_info_freer);

  if(addr_dump_path != NULL)
    ndpi_cache_address_dump(workflow->ndpi_struct, addr_dump_path, 0);

  ndpi_exit_detection_module(workflow->ndpi_struct);
  ndpi_free(workflow->ndpi_flows_root);

  ndpi_stats_free(&workflow->stats);

  ndpi_free(workflow);
}

static inline int cmp_n32(uint32_t a,uint32_t b) {
  return a == b ? 0 : ntohl(a) < ntohl(b) ? -1:1;
}
static inline int cmp_n16(uint16_t a,uint16_t b) {
  return a == b ? 0 : ntohs(a) < ntohs(b) ? -1:1;
}

/* ***************************************************** */

int ndpi_workflow_node_cmp(const void *a, const void *b) {
  const struct ndpi_flow_info *fa = (const struct ndpi_flow_info*)a;
  const struct ndpi_flow_info *fb = (const struct ndpi_flow_info*)b;

  if(fa->hashval < fb->hashval) return(-1); else if(fa->hashval > fb->hashval) return(1);

  /* Flows have the same hash */

  if(fa->vlan_id   < fb->vlan_id   ) return(-1); else { if(fa->vlan_id    > fb->vlan_id   ) return(1); }
  if(fa->protocol  < fb->protocol  ) return(-1); else { if(fa->protocol   > fb->protocol  ) return(1); }

  int r;
  r = cmp_n32(fa->src_ip, fb->src_ip); if(r) return r;
  r = cmp_n16(fa->src_port, fb->src_port) ; if(r) return r;
  r = cmp_n32(fa->dst_ip, fb->dst_ip); if(r) return r;
  r = cmp_n16(fa->dst_port, fb->dst_port);

  return(r);
}

/* ***************************************************** */

/**
 * \brief Update the byte count for the flow record.
 * \param f Flow data
 * \param x Data to use for update
 * \param len Length of the data (in bytes)
 * \return none
 */
static void
ndpi_flow_update_byte_count(struct ndpi_flow_info *flow, const void *x,
                            unsigned int len, u_int8_t src_to_dst_direction) {
  /*
   * implementation note: The spec says that 4000 octets is enough of a
   * sample size to accurately reflect the byte distribution. Also, to avoid
   * wrapping of the byte count at the 16-bit boundry, we stop counting once
   * the 4000th octet has been seen for a flow.
   */

  if((flow->entropy->src2dst_pkt_count+flow->entropy->dst2src_pkt_count) <= max_num_packets_per_flow) {
    /* octet count was already incremented before processing this payload */
    u_int32_t current_count;

    if(src_to_dst_direction) {
      current_count = flow->entropy->src2dst_l4_bytes - len;
    } else {
      current_count = flow->entropy->dst2src_l4_bytes - len;
    }

    if(current_count < ETTA_MIN_OCTETS) {
      u_int32_t i;
      const unsigned char *data = x;

      for(i=0; i<len; i++) {
        if(src_to_dst_direction) {
          flow->entropy->src2dst_byte_count[data[i]]++;
        } else {
          flow->entropy->dst2src_byte_count[data[i]]++;
        }
        current_count++;
        if(current_count >= ETTA_MIN_OCTETS) {
          break;
        }
      }
    }
  }
}

/* ***************************************************** */

/**
 * \brief Update the byte distribution mean for the flow record.
 * \param f Flow record
 * \param x Data to use for update
 * \param len Length of the data (in bytes)
 * \return none
 */
static void
ndpi_flow_update_byte_dist_mean_var(ndpi_flow_info_t *flow, const void *x,
                                    unsigned int len, u_int8_t src_to_dst_direction) {
  const unsigned char *data = x;

  if((flow->entropy->src2dst_pkt_count+flow->entropy->dst2src_pkt_count) <= max_num_packets_per_flow) {
    unsigned int i;

    for(i=0; i<len; i++) {
      double delta;

      if(src_to_dst_direction) {
        flow->entropy->src2dst_num_bytes += 1;
        delta = ((double)data[i] - flow->entropy->src2dst_bd_mean);
        flow->entropy->src2dst_bd_mean += delta/((double)flow->entropy->src2dst_num_bytes);
        flow->entropy->src2dst_bd_variance += delta*((double)data[i] - flow->entropy->src2dst_bd_mean);
      } else {
        flow->entropy->dst2src_num_bytes += 1;
        delta = ((double)data[i] - flow->entropy->dst2src_bd_mean);
        flow->entropy->dst2src_bd_mean += delta/((double)flow->entropy->dst2src_num_bytes);
        flow->entropy->dst2src_bd_variance += delta*((double)data[i] - flow->entropy->dst2src_bd_mean);
      }
    }
  }
}

/* ***************************************************** */

static struct ndpi_flow_info *get_ndpi_flow_info(struct ndpi_workflow * workflow,
             const u_int8_t version,
             u_int16_t vlan_id,
             ndpi_packet_tunnel tunnel_type,
             const struct ndpi_iphdr *iph,
             const struct ndpi_ipv6hdr *iph6,
             u_int16_t ipsize,
             u_int16_t l4_packet_len,
             u_int16_t l4_offset,
             struct ndpi_tcphdr **tcph,
             struct ndpi_udphdr **udph,
             u_int16_t *sport, u_int16_t *dport,
             u_int8_t *proto,
             u_int8_t **payload,
             u_int16_t *payload_len,
             u_int8_t *src_to_dst_direction,
                                                 pkt_timeval when) {
  u_int32_t idx, hashval;
  struct ndpi_flow_info flow;
  void *ret;
  const u_int8_t *l3, *l4;
  u_int32_t l4_data_len = 0XFEEDFACE;

  /*
    Note: to keep things simple (ndpiReader is just a demo app)
    we handle IPv6 a-la-IPv4.
  */
  if(version == IPVERSION) {
    if(ipsize < 20)
      return NULL;

    if((iph->ihl * 4) > ipsize || ipsize < ntohs(iph->tot_len)
       /* || (iph->frag_off & htons(0x1FFF)) != 0 */)
      return NULL;

    l3 = (const u_int8_t*)iph;
  } else {
    if(l4_offset > ipsize)
      return NULL;

    l3 = (const u_int8_t*)iph6;
  }
  if(ipsize < l4_offset + l4_packet_len)
    return NULL;

  *proto = iph->protocol;

  if(l4_packet_len < 64)
    workflow->stats.packet_len[0]++;
  else if(l4_packet_len >= 64 && l4_packet_len < 128)
    workflow->stats.packet_len[1]++;
  else if(l4_packet_len >= 128 && l4_packet_len < 256)
    workflow->stats.packet_len[2]++;
  else if(l4_packet_len >= 256 && l4_packet_len < 1024)
    workflow->stats.packet_len[3]++;
  else if(l4_packet_len >= 1024 && l4_packet_len < 1500)
    workflow->stats.packet_len[4]++;
  else if(l4_packet_len >= 1500)
    workflow->stats.packet_len[5]++;

  if(l4_packet_len > workflow->stats.max_packet_len)
    workflow->stats.max_packet_len = l4_packet_len;

  l4 =& ((const u_int8_t *) l3)[l4_offset];

  if(*proto == IPPROTO_TCP && l4_packet_len >= sizeof(struct ndpi_tcphdr)) {
    u_int tcp_len;

    // TCP
    workflow->stats.tcp_count++;
    *tcph = (struct ndpi_tcphdr *)l4;
    *sport = ntohs((*tcph)->source), *dport = ntohs((*tcph)->dest);
    tcp_len = ndpi_min(4*(*tcph)->doff, l4_packet_len);
    *payload = (u_int8_t*)&l4[tcp_len];
    *payload_len = ndpi_max(0, l4_packet_len-4*(*tcph)->doff);
    l4_data_len = l4_packet_len - sizeof(struct ndpi_tcphdr);
  } else if(*proto == IPPROTO_UDP && l4_packet_len >= sizeof(struct ndpi_udphdr)) {
    // UDP
    workflow->stats.udp_count++;
    *udph = (struct ndpi_udphdr *)l4;
    *sport = ntohs((*udph)->source), *dport = ntohs((*udph)->dest);
    *payload = (u_int8_t*)&l4[sizeof(struct ndpi_udphdr)];
    *payload_len = (l4_packet_len > sizeof(struct ndpi_udphdr)) ? l4_packet_len-sizeof(struct ndpi_udphdr) : 0;
    l4_data_len = l4_packet_len - sizeof(struct ndpi_udphdr);
  } else if(*proto == IPPROTO_ICMP) {
    *payload = (u_int8_t*)&l4[sizeof(struct ndpi_icmphdr )];
    *payload_len = (l4_packet_len > sizeof(struct ndpi_icmphdr)) ? l4_packet_len-sizeof(struct ndpi_icmphdr) : 0;
    l4_data_len = l4_packet_len - sizeof(struct ndpi_icmphdr);
    *sport = *dport = 0;
  } else if(*proto == IPPROTO_ICMPV6) {
    *payload = (u_int8_t*)&l4[sizeof(struct ndpi_icmp6hdr)];
    *payload_len = (l4_packet_len > sizeof(struct ndpi_icmp6hdr)) ? l4_packet_len-sizeof(struct ndpi_icmp6hdr) : 0;
    l4_data_len = l4_packet_len - sizeof(struct ndpi_icmp6hdr);
    *sport = *dport = 0;
  } else {
    // non tcp/udp protocols
    *sport = *dport = 0;
    l4_data_len = 0;
  }

  flow.protocol = iph->protocol, flow.vlan_id = vlan_id;
  flow.src_ip = iph->saddr, flow.dst_ip = iph->daddr;
  flow.src_port = htons(*sport), flow.dst_port = htons(*dport);
  flow.hashval = hashval = flow.protocol + ntohl(flow.src_ip) + ntohl(flow.dst_ip)
    + ntohs(flow.src_port) + ntohs(flow.dst_port);

#if 0
  {
  char ip1[48],ip2[48];
       inet_ntop(AF_INET, &flow.src_ip, ip1, sizeof(ip1));
       inet_ntop(AF_INET, &flow.dst_ip, ip2, sizeof(ip2));
  printf("hashval=%u [%u][%u][%s:%u][%s:%u]\n", hashval, flow.protocol, flow.vlan_id,
        ip1, ntohs(flow.src_port),  ip2, ntohs(flow.dst_port));
  }
#endif

  idx = hashval % workflow->prefs.num_roots;
  ret = ndpi_tfind(&flow, &workflow->ndpi_flows_root[idx], ndpi_workflow_node_cmp);

  /* to avoid two nodes in one binary tree for a flow */
  int is_changed = 0;
  if(ret == NULL) {
    u_int32_t orig_src_ip = flow.src_ip;
    u_int16_t orig_src_port = flow.src_port;
    u_int32_t orig_dst_ip = flow.dst_ip;
    u_int16_t orig_dst_port = flow.dst_port;

    flow.src_ip = orig_dst_ip;
    flow.src_port = orig_dst_port;
    flow.dst_ip = orig_src_ip;
    flow.dst_port = orig_src_port;

    is_changed = 1;

    ret = ndpi_tfind(&flow, &workflow->ndpi_flows_root[idx], ndpi_workflow_node_cmp);
  }

  if(ret == NULL) {
    if(workflow->stats.ndpi_flow_count == workflow->prefs.max_ndpi_flows) {
      LOG(NDPI_LOG_ERROR,
         "maximum flow count (%u) has been exceeded\n",
         workflow->prefs.max_ndpi_flows);
      return NULL;
    } else {
      struct ndpi_flow_info *newflow = (struct ndpi_flow_info*)ndpi_malloc(sizeof(struct ndpi_flow_info));

      if(newflow == NULL) {
  LOG(NDPI_LOG_ERROR, "[NDPI] %s(1): not enough memory\n", __FUNCTION__);
  return(NULL);
      } else
        workflow->num_allocated_flows++;

      memset(newflow, 0, sizeof(struct ndpi_flow_info));
      newflow->flow_id = flow_id++;
      newflow->hashval = hashval;
      newflow->tunnel_type = tunnel_type;
      newflow->protocol = iph->protocol, newflow->vlan_id = vlan_id;
      newflow->src_ip = iph->saddr, newflow->dst_ip = iph->daddr;
      newflow->src_port = htons(*sport), newflow->dst_port = htons(*dport);
      newflow->ip_version = version;
      newflow->iat_c_to_s = ndpi_alloc_data_analysis(DATA_ANALUYSIS_SLIDING_WINDOW),
  newflow->iat_s_to_c =  ndpi_alloc_data_analysis(DATA_ANALUYSIS_SLIDING_WINDOW);
      newflow->pktlen_c_to_s = ndpi_alloc_data_analysis(DATA_ANALUYSIS_SLIDING_WINDOW),
  newflow->pktlen_s_to_c =  ndpi_alloc_data_analysis(DATA_ANALUYSIS_SLIDING_WINDOW),
  newflow->iat_flow = ndpi_alloc_data_analysis(DATA_ANALUYSIS_SLIDING_WINDOW);

#ifdef DIRECTION_BINS
      ndpi_init_bin(&newflow->payload_len_bin_src2dst, ndpi_bin_family8, PLEN_NUM_BINS);
      ndpi_init_bin(&newflow->payload_len_bin_dst2src, ndpi_bin_family8, PLEN_NUM_BINS);
#else
      ndpi_init_bin(&newflow->payload_len_bin, ndpi_bin_family8, PLEN_NUM_BINS);
#endif

      if (version == 4 || version == 6) {
        uint16_t inet_addrlen = (version == 4) ? INET_ADDRSTRLEN : INET6_ADDRSTRLEN;
        newflow->src_name = ndpi_malloc(inet_addrlen);
        newflow->dst_name = ndpi_malloc(inet_addrlen);

        if(version == 4) {
          if (newflow->src_name)
            inet_ntop(AF_INET, &newflow->src_ip, newflow->src_name, inet_addrlen);
          if (newflow->dst_name)
            inet_ntop(AF_INET, &newflow->dst_ip, newflow->dst_name, inet_addrlen);
        } else if (version == 6) {
          newflow->src_ip6 = *(struct ndpi_in6_addr *)&iph6->ip6_src;
          newflow->dst_ip6 = *(struct ndpi_in6_addr *)&iph6->ip6_dst;

          if (newflow->src_name)
            inet_ntop(AF_INET6, &newflow->src_ip6, newflow->src_name, inet_addrlen);
          if (newflow->dst_name)
            inet_ntop(AF_INET6, &newflow->dst_ip6, newflow->dst_name, inet_addrlen);

          /* For consistency across platforms replace :0: with :: */
          if (newflow->src_name) ndpi_patchIPv6Address(newflow->src_name);
          if (newflow->dst_name) ndpi_patchIPv6Address(newflow->dst_name);
        }
      }

      if((newflow->ndpi_flow = ndpi_flow_malloc(SIZEOF_FLOW_STRUCT)) == NULL) {
  LOG(NDPI_LOG_ERROR, "[NDPI] %s(2): not enough memory\n", __FUNCTION__);
  ndpi_flow_info_free_data(newflow);
  ndpi_free(newflow);
  return(NULL);
      } else
  memset(newflow->ndpi_flow, 0, SIZEOF_FLOW_STRUCT);

    if (workflow->ndpi_serialization_format != ndpi_serialization_format_unknown)
    {
      if (ndpi_init_serializer(&newflow->ndpi_flow_serializer,
                               workflow->ndpi_serialization_format) != 0)
      {
        LOG(NDPI_LOG_ERROR, "ndpi serializer init failed\n");
        ndpi_flow_info_free_data(newflow);
        ndpi_free(newflow);
        return(NULL);
      }
    }

      if(ndpi_tsearch(newflow, &workflow->ndpi_flows_root[idx], ndpi_workflow_node_cmp) == NULL) { /* Add */
        ndpi_flow_info_free_data(newflow);
        ndpi_free(newflow);
        return(NULL);
      }
      workflow->stats.ndpi_flow_count++;
      if(*proto == IPPROTO_TCP)
        workflow->stats.flow_count[0]++;
      else if(*proto == IPPROTO_UDP)
        workflow->stats.flow_count[1]++;
      else
        workflow->stats.flow_count[2]++;

      if(enable_flow_stats) {
        newflow->entropy = ndpi_calloc(1, sizeof(struct ndpi_entropy));
        newflow->last_entropy = ndpi_calloc(1, sizeof(struct ndpi_entropy));
        if(!newflow->entropy || !newflow->last_entropy) {
          ndpi_tdelete(newflow, &workflow->ndpi_flows_root[idx], ndpi_workflow_node_cmp);
          ndpi_flow_info_free_data(newflow);
          ndpi_free(newflow);
          return(NULL);
        }
        newflow->entropy->src2dst_pkt_len[newflow->entropy->src2dst_pkt_count] = l4_data_len;
        newflow->entropy->src2dst_pkt_time[newflow->entropy->src2dst_pkt_count] = when;
        if(newflow->entropy->src2dst_pkt_count == 0) {
          newflow->entropy->src2dst_start = when;
        }
        newflow->entropy->src2dst_pkt_count++;
        // Non zero app data.
        if(l4_data_len != 0XFEEDFACE && l4_data_len != 0) {
          newflow->entropy->src2dst_opackets++;
          newflow->entropy->src2dst_l4_bytes += l4_data_len;
        }
      }
      return newflow;
    }
  } else {
    struct ndpi_flow_info *rflow = *(struct ndpi_flow_info**)ret;

    if(is_changed) {
  *src_to_dst_direction = 0, rflow->bidirectional |= 1;
    }
    else {
  *src_to_dst_direction = 1;
    }
    if(enable_flow_stats) {
      if(*src_to_dst_direction) {
        if(rflow->entropy->src2dst_pkt_count < max_num_packets_per_flow) {
          rflow->entropy->src2dst_pkt_len[rflow->entropy->src2dst_pkt_count] = l4_data_len;
          rflow->entropy->src2dst_pkt_time[rflow->entropy->src2dst_pkt_count] = when;
          rflow->entropy->src2dst_l4_bytes += l4_data_len;
          rflow->entropy->src2dst_pkt_count++;
        }
        // Non zero app data.
        if(l4_data_len != 0XFEEDFACE && l4_data_len != 0) {
          rflow->entropy->src2dst_opackets++;
        }
      } else {
        if(rflow->entropy->dst2src_pkt_count < max_num_packets_per_flow) {
          rflow->entropy->dst2src_pkt_len[rflow->entropy->dst2src_pkt_count] = l4_data_len;
          rflow->entropy->dst2src_pkt_time[rflow->entropy->dst2src_pkt_count] = when;
          if(rflow->entropy->dst2src_pkt_count == 0) {
            rflow->entropy->dst2src_start = when;
          }
          rflow->entropy->dst2src_l4_bytes += l4_data_len;
          rflow->entropy->dst2src_pkt_count++;
        }
        // Non zero app data.
        if(l4_data_len != 0XFEEDFACE && l4_data_len != 0) {
          rflow->entropy->dst2src_opackets++;
        }
      }
    }

    return(rflow);
  }
}

/* ****************************************************** */

static struct ndpi_flow_info *get_ndpi_flow_info6(struct ndpi_workflow * workflow,
              u_int16_t vlan_id,
              ndpi_packet_tunnel tunnel_type,
              const struct ndpi_ipv6hdr *iph6,
              u_int16_t ipsize,
              struct ndpi_tcphdr **tcph,
              struct ndpi_udphdr **udph,
              u_int16_t *sport, u_int16_t *dport,
              u_int8_t *proto,
              u_int8_t **payload,
              u_int16_t *payload_len,
              u_int8_t *src_to_dst_direction,
                                                  pkt_timeval when) {
  struct ndpi_iphdr iph;

  if(ipsize < 40)
    return(NULL);
  memset(&iph, 0, sizeof(iph));
  iph.version = IPVERSION;
  iph.saddr = iph6->ip6_src.u6_addr.u6_addr32[2] + iph6->ip6_src.u6_addr.u6_addr32[3];
  iph.daddr = iph6->ip6_dst.u6_addr.u6_addr32[2] + iph6->ip6_dst.u6_addr.u6_addr32[3];
  u_int8_t l4proto = iph6->ip6_hdr.ip6_un1_nxt;
  u_int16_t ip_len = ntohs(iph6->ip6_hdr.ip6_un1_plen);
  const u_int8_t *l4ptr = (((const u_int8_t *) iph6) + sizeof(struct ndpi_ipv6hdr));
  if(ipsize < sizeof(struct ndpi_ipv6hdr) + ip_len)
    return(NULL);
  if(ndpi_handle_ipv6_extension_headers(ipsize - sizeof(struct ndpi_ipv6hdr), &l4ptr, &ip_len, &l4proto) != 0) {
    return(NULL);
  }
  iph.protocol = l4proto;

  return(get_ndpi_flow_info(workflow, 6, vlan_id, tunnel_type,
          &iph, iph6, ipsize,
          ip_len, l4ptr - (const u_int8_t *)iph6,
          tcph, udph, sport, dport,
          proto, payload,
          payload_len, src_to_dst_direction, when));
}

/* ****************************************************** */

u_int8_t is_ndpi_proto(struct ndpi_flow_info *flow, u_int16_t id) {
  if((flow->detected_protocol.proto.master_protocol == id)
     || (flow->detected_protocol.proto.app_protocol == id))
    return(1);
  else
    return(0);
}

/* ****************************************************** */

void correct_csv_data_field(char* data) {
  /* Replace , with ; to avoid issues with CSVs */
  u_int i;
  for(i=0; data[i] != '\0'; i++) if(data[i] == ',') data[i] = ';';
}

/* ****************************************************** */

u_int8_t plen2slot(u_int16_t plen) {
  /*
     Slots [32 bytes lenght]
     0..31, 32..63 ...
  */

  if(plen > PLEN_MAX)
    return(PLEN_NUM_BINS-1);
  else
    return(plen/PLEN_BIN_LEN);
}

/* ****************************************************** */

static void dump_flow_fingerprint(struct ndpi_workflow * workflow,
          struct ndpi_flow_info *flow) {
  ndpi_serializer serializer;
  bool rc;

  if(ndpi_init_serializer(&serializer, ndpi_serialization_format_json) == -1)
    return;

  ndpi_serialize_start_of_block(&serializer, "fingerprint");
  rc = ndpi_serialize_flow_fingerprint(workflow->ndpi_struct, flow->ndpi_flow, &serializer);
  ndpi_serialize_end_of_block(&serializer);

  if(rc) {
    char buf[64], *buffer;
    u_int32_t buffer_len;

    ndpi_serialize_string_uint32(&serializer, "proto", flow->protocol);
    ndpi_serialize_string_string(&serializer, "cli_ip", flow->src_name ? flow->src_name : "");
    ndpi_serialize_string_uint32(&serializer, "cli_port", ntohs(flow->src_port));
    ndpi_serialize_string_string(&serializer, "srv_ip", flow->dst_name ? flow->dst_name : "");
    ndpi_serialize_string_uint32(&serializer, "srv_port", ntohs(flow->dst_port));
    ndpi_serialize_string_string(&serializer, "proto",
         ndpi_protocol2name(workflow->ndpi_struct,
                flow->detected_protocol.proto,
                buf, sizeof(buf)));

    if(flow->server_hostname)
      ndpi_serialize_string_string(&serializer, "server_hostname", flow->server_hostname);

    buffer = ndpi_serializer_get_buffer(&serializer, &buffer_len);
    fprintf(fingerprint_fp, "%s\n", buffer);
  }

  ndpi_term_serializer(&serializer);
}


static void add_to_address_port_list(ndpi_address_port_list *list, ndpi_address_port *ap)
{
  int new_num;
  void *new_buf;
  unsigned int i;

  if(ap->port == 0)
    return;

  /* Avoid saving duplicates */
  for(i = 0; i < list->num_aps; i++)
    if(memcmp(&list->aps[i], ap, sizeof(*ap)) == 0)
      return;

  if(list->num_aps == list->num_aps_allocated) {
    new_num = 1 + list->num_aps_allocated * 2;
    new_buf = ndpi_realloc(list->aps,
                     new_num * sizeof(ndpi_address_port));
    if(!new_buf)
      return;
    list->aps = new_buf;
    list->num_aps_allocated = new_num;
  }
  memcpy(&list->aps[list->num_aps++], ap, sizeof(ndpi_address_port));
}

/* ****************************************************** */

static void process_ndpi_monitoring_info(struct ndpi_flow_info *flow) {
  if(!flow->ndpi_flow || !flow->ndpi_flow->monit)
    return;

  if(flow->monitoring_state == 0 &&
     flow->ndpi_flow->state == NDPI_STATE_MONITORING) {
    /* We just moved to monitoring state */
    flow->monitoring_state = 1;
    flow->num_packets_before_monitoring = flow->ndpi_flow->packet_direction_complete_counter[0] + flow->ndpi_flow->packet_direction_complete_counter[1];
  }

  /* In theory, we should check only for STUN.
     However since we sometimes might not have STUN in protocol classification
     (because we have only two protocols in flow->ndpi_flow->detected_protocol_stack[])
     we need to check also for the other "master" protocols set by STUN dissector
     See at the beginning of the STUN c file for further details
   */
  if(flow->detected_protocol.proto.app_protocol == NDPI_PROTOCOL_STUN ||
     flow->detected_protocol.proto.master_protocol == NDPI_PROTOCOL_STUN ||
     flow->detected_protocol.proto.app_protocol == NDPI_PROTOCOL_DTLS ||
     flow->detected_protocol.proto.master_protocol == NDPI_PROTOCOL_DTLS ||
     flow->detected_protocol.proto.app_protocol == NDPI_PROTOCOL_SRTP ||
     flow->detected_protocol.proto.master_protocol == NDPI_PROTOCOL_SRTP) {

    add_to_address_port_list(&flow->stun.mapped_address, &flow->ndpi_flow->monit->protos.dtls_stun_rtp.mapped_address);
    add_to_address_port_list(&flow->stun.other_address, &flow->ndpi_flow->monit->protos.dtls_stun_rtp.other_address);
    add_to_address_port_list(&flow->stun.peer_address, &flow->ndpi_flow->monit->protos.dtls_stun_rtp.peer_address);
    add_to_address_port_list(&flow->stun.relayed_address, &flow->ndpi_flow->monit->protos.dtls_stun_rtp.relayed_address);
    add_to_address_port_list(&flow->stun.response_origin, &flow->ndpi_flow->monit->protos.dtls_stun_rtp.response_origin);
    flow->multimedia_flow_types |= flow->ndpi_flow->flow_multimedia_types;

    flow->stun.rtp_counters[0] = flow->ndpi_flow->stun.rtp_counters[0];
    flow->stun.rtp_counters[1] = flow->ndpi_flow->stun.rtp_counters[1];
  }
}

/* ****************************************************** */

static void serialize_monitoring_metadata(struct ndpi_flow_info *flow)
{
  unsigned int i;
  char buf[64];

  if(!flow->ndpi_flow->monit)
    return;

  ndpi_serialize_start_of_block(&flow->ndpi_flow_serializer, "monitoring");

  switch(flow->detected_protocol.proto.master_protocol ? flow->detected_protocol.proto.master_protocol : flow->detected_protocol.proto.app_protocol) {
    case NDPI_PROTOCOL_STUN:
    case NDPI_PROTOCOL_DTLS:
    case NDPI_PROTOCOL_SRTP:
      ndpi_serialize_start_of_block(&flow->ndpi_flow_serializer, "stun");

      if(flow->stun.mapped_address.num_aps > 0) {
        ndpi_serialize_start_of_list(&flow->ndpi_flow_serializer, "mapped_address");
        for(i = 0; i < flow->stun.mapped_address.num_aps; i++) {
          if(flow->stun.mapped_address.aps[i].port > 0) {
            ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "mapped_address",
                                         print_ndpi_address_port(&flow->stun.mapped_address.aps[i], buf, sizeof(buf)));
          }
        }
        ndpi_serialize_end_of_list(&flow->ndpi_flow_serializer);
      }

      if(flow->stun.other_address.num_aps > 0) {
        ndpi_serialize_start_of_list(&flow->ndpi_flow_serializer, "other_address");
        for(i = 0; i < flow->stun.other_address.num_aps; i++) {
          if(flow->stun.other_address.aps[i].port > 0) {
            ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "other_address",
                                         print_ndpi_address_port(&flow->stun.other_address.aps[i], buf, sizeof(buf)));
          }
        }
        ndpi_serialize_end_of_list(&flow->ndpi_flow_serializer);
      }

      if(flow->stun.peer_address.num_aps > 0) {
        ndpi_serialize_start_of_list(&flow->ndpi_flow_serializer, "peer_address");
        for(i = 0; i < flow->stun.peer_address.num_aps; i++) {
          if(flow->stun.peer_address.aps[i].port > 0) {
            ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "peer_address",
                                         print_ndpi_address_port(&flow->stun.peer_address.aps[i], buf, sizeof(buf)));
          }
        }
        ndpi_serialize_end_of_list(&flow->ndpi_flow_serializer);
      }

      if(flow->stun.relayed_address.num_aps > 0) {
        ndpi_serialize_start_of_list(&flow->ndpi_flow_serializer, "relayed_address");
        for(i = 0; i < flow->stun.relayed_address.num_aps; i++) {
          if(flow->stun.relayed_address.aps[i].port > 0) {
            ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "relayed_address",
                                         print_ndpi_address_port(&flow->stun.relayed_address.aps[i], buf, sizeof(buf)));
          }
        }
        ndpi_serialize_end_of_list(&flow->ndpi_flow_serializer);
      }

      if(flow->stun.response_origin.num_aps > 0) {
        ndpi_serialize_start_of_list(&flow->ndpi_flow_serializer, "response_origin");
        for(i = 0; i < flow->stun.response_origin.num_aps; i++) {
          if(flow->stun.response_origin.aps[i].port > 0) {
            ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "response_origin",
                                         print_ndpi_address_port(&flow->stun.response_origin.aps[i], buf, sizeof(buf)));
          }
        }
        ndpi_serialize_end_of_list(&flow->ndpi_flow_serializer);
      }

      ndpi_serialize_end_of_block(&flow->ndpi_flow_serializer); /* stun */

      break;
  }

  ndpi_serialize_end_of_block(&flow->ndpi_flow_serializer);
}

/* ****************************************************** */

void process_ndpi_collected_info(struct ndpi_workflow * workflow, struct ndpi_flow_info *flow) {
  u_int i;
  char out[128], *s;

  if(!flow->ndpi_flow) return;

  flow->info_type = INFO_INVALID;

  s = ndpi_get_flow_risk_info(flow->ndpi_flow, out, sizeof(out), 0 /* text */);

  if(s != NULL)
    flow->risk_str = ndpi_strdup(s);

  flow->confidence = flow->ndpi_flow->confidence;

  flow->num_dissector_calls = flow->ndpi_flow->num_dissector_calls;

  ndpi_snprintf(flow->host_server_name, sizeof(flow->host_server_name), "%s",
    flow->ndpi_flow->host_server_name);

  if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_MINING)) {
    ndpi_snprintf(flow->mining.currency, sizeof(flow->mining.currency), "%s",
      flow->ndpi_flow->protos.mining.currency);
  }

  flow->risk = flow->ndpi_flow->risk;

  if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_DHCP)) {
    if(flow->ndpi_flow->protos.dhcp.fingerprint[0] != '\0')
      flow->dhcp_fingerprint = ndpi_strdup(flow->ndpi_flow->protos.dhcp.fingerprint);

    if(flow->ndpi_flow->protos.dhcp.class_ident[0] != '\0')
      flow->dhcp_class_ident = ndpi_strdup(flow->ndpi_flow->protos.dhcp.class_ident);
  } else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_BITTORRENT) &&
            !ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_DNS) &&
            !ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_TLS)) {
    u_int j;

    if(flow->ndpi_flow->protos.bittorrent.hash[0] != '\0') {
      u_int avail = sizeof(flow->ndpi_flow->protos.bittorrent.hash) * 2 + 1;
      flow->bittorent_hash = ndpi_malloc(avail);

      if(flow->bittorent_hash) {
        for(i=0, j = 0; i < sizeof(flow->ndpi_flow->protos.bittorrent.hash); i++) {
          snprintf(&flow->bittorent_hash[j], avail-j, "%02x",
            flow->ndpi_flow->protos.bittorrent.hash[i]);

          j += 2;
        }

        flow->bittorent_hash[j] = '\0';
      }
    }
  }
  /* TIVOCONNECT */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_TIVOCONNECT)) {
    flow->info_type = INFO_TIVOCONNECT;
    ndpi_snprintf(flow->tivoconnect.identity_uuid, sizeof(flow->tivoconnect.identity_uuid),
                  "%s", flow->ndpi_flow->protos.tivoconnect.identity_uuid);
    ndpi_snprintf(flow->tivoconnect.machine, sizeof(flow->tivoconnect.machine),
                  "%s", flow->ndpi_flow->protos.tivoconnect.machine);
    ndpi_snprintf(flow->tivoconnect.platform, sizeof(flow->tivoconnect.platform),
                  "%s", flow->ndpi_flow->protos.tivoconnect.platform);
    ndpi_snprintf(flow->tivoconnect.services, sizeof(flow->tivoconnect.services),
                  "%s", flow->ndpi_flow->protos.tivoconnect.services);
  }
  /* SOFTETHER */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_SOFTETHER) &&
      !ndpi_stack_contains(&flow->detected_protocol.protocol_stack,  NDPI_PROTOCOL_HTTP)) {
    flow->info_type = INFO_SOFTETHER;
    ndpi_snprintf(flow->softether.ip, sizeof(flow->softether.ip), "%s",
                  flow->ndpi_flow->protos.softether.ip);
    ndpi_snprintf(flow->softether.port, sizeof(flow->softether.port), "%s",
                  flow->ndpi_flow->protos.softether.port);
    ndpi_snprintf(flow->softether.hostname, sizeof(flow->softether.hostname), "%s",
                  flow->ndpi_flow->protos.softether.hostname);
    ndpi_snprintf(flow->softether.fqdn, sizeof(flow->softether.fqdn), "%s",
                  flow->ndpi_flow->protos.softether.fqdn);
  }
  /* SERVICE_LOCATION */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_SERVICE_LOCATION)) {
    size_t i;

    flow->info_type = INFO_GENERIC;
    flow->info[0] = 0;
    if (flow->ndpi_flow->protos.slp.url_count > 0)
      strncat(flow->info, "URL(s): ", sizeof(flow->info)-1);

    for (i = 0; i < flow->ndpi_flow->protos.slp.url_count; ++i) {
      size_t length = strlen(flow->info);

      strncat(flow->info + length, flow->ndpi_flow->protos.slp.url[i],
              sizeof(flow->info) - length);
      length = strlen(flow->info);

      if (i < (size_t)flow->ndpi_flow->protos.slp.url_count - 1)
        strncat(flow->info + length, ", ", sizeof(flow->info) - length);
    }
  }
  /* NATPMP */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_NATPMP)) {
    flow->info_type = INFO_NATPMP;
    flow->natpmp.result_code = flow->ndpi_flow->protos.natpmp.result_code;
    flow->natpmp.internal_port = flow->ndpi_flow->protos.natpmp.internal_port;
    flow->natpmp.external_port = flow->ndpi_flow->protos.natpmp.external_port;
    inet_ntop(AF_INET, &flow->ndpi_flow->protos.natpmp.external_address.ipv4, &flow->natpmp.ip[0], sizeof(flow->natpmp.ip));
  }
  /* DISCORD */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_DISCORD) &&
          !ndpi_stack_is_tls_like(&flow->detected_protocol.protocol_stack) &&
          flow->ndpi_flow->protos.discord.client_ip[0] != '\0') {
    flow->info_type = INFO_GENERIC;
    ndpi_snprintf(flow->info, sizeof(flow->info), "Client IP: %s",
                  flow->ndpi_flow->protos.discord.client_ip);
  }
  /* DNS */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_DNS)) {
    if(flow->ndpi_flow->protos.dns.is_rsp_addr_ipv6[0] == 0)
    {
      flow->info_type = INFO_GENERIC;
      inet_ntop(AF_INET, &flow->ndpi_flow->protos.dns.rsp_addr[0].ipv4, flow->info, sizeof(flow->info));
    } else {
      flow->info_type = INFO_GENERIC;
      inet_ntop(AF_INET6, &flow->ndpi_flow->protos.dns.rsp_addr[0].ipv6, flow->info, sizeof(flow->info));

      /* For consistency across platforms replace :0: with :: */
      ndpi_patchIPv6Address(flow->info);
    }

    if(flow->ndpi_flow->protos.dns.geolocation_iata_code[0] != '\0')
      strcpy(flow->dns.geolocation_iata_code, flow->ndpi_flow->protos.dns.geolocation_iata_code);

    if(flow->ndpi_flow->protos.dns.ptr_domain_name[0] != '\0')
      strcpy(flow->dns.ptr_domain_name, flow->ndpi_flow->protos.dns.ptr_domain_name);

    flow->dns.transaction_id = flow->ndpi_flow->protos.dns.transaction_id;

#if 0
    if(0) {
      u_int8_t i;

      for(i=0; i<flow->ndpi_flow->protos.dns.num_rsp_addr; i++) {
  char buf[64];

  if(flow->ndpi_flow->protos.dns.is_rsp_addr_ipv6[i] == 0) {
    inet_ntop(AF_INET, &flow->ndpi_flow->protos.dns.rsp_addr[i].ipv4, buf, sizeof(buf));
  } else {
    inet_ntop(AF_INET6, &flow->ndpi_flow->protos.dns.rsp_addr[i].ipv6, buf, sizeof(buf));
  }

  printf("(%s) %s [ttl: %u]\n", flow->host_server_name, buf, flow->ndpi_flow->protos.dns.rsp_addr_ttl[i]);
      }
    }
#endif
  }
  /* MDNS */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_MDNS)) {
    flow->info_type = INFO_GENERIC;
    ndpi_snprintf(flow->info, sizeof(flow->info), "%s", flow->ndpi_flow->host_server_name);
  }
  /* UBNTAC2 */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_UBNTAC2)) {
    flow->info_type = INFO_GENERIC;
    ndpi_snprintf(flow->info, sizeof(flow->info), "%s", flow->ndpi_flow->protos.ubntac2.version);
  }
  /* FTP, IMAP, SMTP, POP3 */
  else if(!ndpi_stack_is_tls_like(&flow->detected_protocol.protocol_stack) &&
          (ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_FTP_CONTROL) ||
     ndpi_stack_contains(&flow->detected_protocol.protocol_stack,  NDPI_PROTOCOL_MAIL_IMAP) ||
     ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_MAIL_POP) ||
     ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_MAIL_SMTP))) {
    flow->info_type = INFO_FTP_IMAP_POP_SMTP;
    ndpi_snprintf(flow->ftp_imap_pop_smtp.username,
                  sizeof(flow->ftp_imap_pop_smtp.username),
                  "%s", flow->ndpi_flow->l4.tcp.ftp_imap_pop_smtp.username);
    ndpi_snprintf(flow->ftp_imap_pop_smtp.password,
                  sizeof(flow->ftp_imap_pop_smtp.password),
                  "%s", flow->ndpi_flow->l4.tcp.ftp_imap_pop_smtp.password);
    flow->ftp_imap_pop_smtp.auth_failed =
      flow->ndpi_flow->l4.tcp.ftp_imap_pop_smtp.auth_failed;
  }
  /* TFTP */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_TFTP)) {
    flow->info_type = INFO_GENERIC;
    if(flow->ndpi_flow->protos.tftp.filename[0] != '\0')
      ndpi_snprintf(flow->info, sizeof(flow->info), "Filename: %s",
                    flow->ndpi_flow->protos.tftp.filename);
  }
  /* KERBEROS */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_KERBEROS)) {
    flow->info_type = INFO_KERBEROS;
    ndpi_snprintf(flow->kerberos.domain,
                  sizeof(flow->kerberos.domain),
                  "%s", flow->ndpi_flow->protos.kerberos.domain);
    ndpi_snprintf(flow->kerberos.hostname,
                  sizeof(flow->kerberos.hostname),
                  "%s", flow->ndpi_flow->protos.kerberos.hostname);
    ndpi_snprintf(flow->kerberos.username,
                  sizeof(flow->kerberos.username),
                  "%s", flow->ndpi_flow->protos.kerberos.username);
  /* COLLECTD */
  } else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_COLLECTD)) {
    flow->info_type = INFO_GENERIC;
    if(flow->ndpi_flow->protos.collectd.client_username[0] != '\0')
      ndpi_snprintf(flow->info, sizeof(flow->info), "Username: %s",
                    flow->ndpi_flow->protos.collectd.client_username);
  }
  /* SIP */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_SIP)) {
    flow->info_type = INFO_SIP;
    if(flow->ndpi_flow->protos.sip.from)
      ndpi_snprintf(flow->sip.from, sizeof(flow->sip.from), "%s", flow->ndpi_flow->protos.sip.from);
    if(flow->ndpi_flow->protos.sip.from_imsi[0] != '\0')
      ndpi_snprintf(flow->sip.from_imsi, sizeof(flow->sip.from_imsi), "%s", flow->ndpi_flow->protos.sip.from_imsi);
    if(flow->ndpi_flow->protos.sip.to)
      ndpi_snprintf(flow->sip.to, sizeof(flow->sip.to), "%s", flow->ndpi_flow->protos.sip.to);
    if(flow->ndpi_flow->protos.sip.to_imsi[0] != '\0')
      ndpi_snprintf(flow->sip.to_imsi, sizeof(flow->sip.to_imsi), "%s", flow->ndpi_flow->protos.sip.to_imsi);
  }
  /* BFCP */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_BFCP)) {
    flow->info_type = INFO_BFCP;
    flow->bfcp.conference_id = flow->ndpi_flow->protos.bfcp.conference_id;
    flow->bfcp.user_id = flow->ndpi_flow->protos.bfcp.user_id;
  }
  /* TELNET */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_TELNET)) {
    if(flow->ndpi_flow->protos.telnet.username[0] != '\0')
      flow->telnet.username = ndpi_strdup(flow->ndpi_flow->protos.telnet.username);
    if(flow->ndpi_flow->protos.telnet.password[0] != '\0')
      flow->telnet.password = ndpi_strdup(flow->ndpi_flow->protos.telnet.password);
  } else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_SSH)) {
    ndpi_snprintf(flow->host_server_name,
       sizeof(flow->host_server_name), "%s",
       flow->ndpi_flow->protos.ssh.client_signature);
    ndpi_snprintf(flow->ssh_tls.server_info, sizeof(flow->ssh_tls.server_info), "%s",
       flow->ndpi_flow->protos.ssh.server_signature);
    ndpi_snprintf(flow->ssh_tls.client_hassh, sizeof(flow->ssh_tls.client_hassh), "%s",
       flow->ndpi_flow->protos.ssh.hassh_client);
    ndpi_snprintf(flow->ssh_tls.server_hassh, sizeof(flow->ssh_tls.server_hassh), "%s",
       flow->ndpi_flow->protos.ssh.hassh_server);
  }
  /* TLS/QUIC/DTLS/MAIL_S/FTPS */
  else if(ndpi_stack_is_tls_like(&flow->detected_protocol.protocol_stack)) {
    flow->ssh_tls.ssl_version = flow->ndpi_flow->protos.tls_quic.ssl_version;
    flow->ssh_tls.quic_version = flow->ndpi_flow->protos.tls_quic.quic_version;

    if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_QUIC))
      flow->idle_timeout_sec = flow->ndpi_flow->protos.tls_quic.quic_idle_timeout_sec;

    if(flow->ndpi_flow->protos.tls_quic.server_names_len > 0 && flow->ndpi_flow->protos.tls_quic.server_names)
      flow->ssh_tls.server_names = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.server_names);

    flow->ssh_tls.notBefore = flow->ndpi_flow->protos.tls_quic.notBefore;
    flow->ssh_tls.notAfter = flow->ndpi_flow->protos.tls_quic.notAfter;
    ndpi_snprintf(flow->ssh_tls.ja4_client, sizeof(flow->ssh_tls.ja4_client), "%s",
       flow->ndpi_flow->protos.tls_quic.ja4_client);

    if(flow->ndpi_flow->ndpi.fingerprint)
      flow->ndpi_fingerprint = ndpi_strdup(flow->ndpi_flow->ndpi.fingerprint);

    if(flow->ndpi_flow->protos.tls_quic.ja4_client_raw)
      flow->ssh_tls.ja4_client_raw = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.ja4_client_raw);

    ndpi_snprintf(flow->ssh_tls.ja3_server, sizeof(flow->ssh_tls.ja3_server), "%s",
            flow->ndpi_flow->protos.tls_quic.ja3_server);
    flow->ssh_tls.server_unsafe_cipher = flow->ndpi_flow->protos.tls_quic.server_unsafe_cipher;
    flow->ssh_tls.server_cipher = flow->ndpi_flow->protos.tls_quic.server_cipher;

    if(flow->ndpi_flow->protos.tls_quic.fingerprint_set) {
      memcpy(flow->ssh_tls.sha1_cert_fingerprint,
       flow->ndpi_flow->protos.tls_quic.sha1_certificate_fingerprint, 20);
      flow->ssh_tls.sha1_cert_fingerprint_set = 1;
    }

    flow->ssh_tls.browser_heuristics = flow->ndpi_flow->protos.tls_quic.browser_heuristics;

    if(flow->ndpi_flow->protos.tls_quic.issuerDN)
      flow->ssh_tls.tls_issuerDN = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.issuerDN);

    if(flow->ndpi_flow->protos.tls_quic.subjectDN)
      flow->ssh_tls.tls_subjectDN = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.subjectDN);

    flow->ssh_tls.encrypted_ch.version = flow->ndpi_flow->protos.tls_quic.encrypted_ch.version;

    if(flow->ndpi_flow->protos.tls_quic.tls_supported_versions) {
      if((flow->ssh_tls.tls_supported_versions = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.tls_supported_versions)) != NULL)
  correct_csv_data_field(flow->ssh_tls.tls_supported_versions);
    }

    if(flow->ndpi_flow->protos.tls_quic.advertised_alpns) {
      if((flow->ssh_tls.advertised_alpns = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.advertised_alpns)) != NULL)
  correct_csv_data_field(flow->ssh_tls.advertised_alpns);
    }

    if(flow->ndpi_flow->protos.tls_quic.negotiated_alpn) {
      if((flow->ssh_tls.negotiated_alpn = ndpi_strdup(flow->ndpi_flow->protos.tls_quic.negotiated_alpn)) != NULL)
  correct_csv_data_field(flow->ssh_tls.negotiated_alpn);
    }

    if(flow->protocol == IPPROTO_TCP) {
      if(enable_doh_dot_detection) {
  /* For TLS we use TLS block lenght instead of payload lenght */
  ndpi_reset_bin(&flow->payload_len_bin);
  
  for(i=0; i<flow->ndpi_flow->l4.tcp.tls.num_tls_blocks; i++) {
    u_int16_t len = abs(flow->ndpi_flow->l4.tcp.tls.tls_blocks[i].len);
    
    /* printf("[TLS_LEN] %u\n", len); */
    ndpi_inc_bin(&flow->payload_len_bin, plen2slot(len), 1);
  }
      }
      
      flow->ssh_tls.num_blocks = flow->ndpi_flow->l4.tcp.tls.num_tls_blocks;
      memcpy(flow->ssh_tls.blocks, flow->ndpi_flow->l4.tcp.tls.tls_blocks, sizeof(flow->ndpi_flow->l4.tcp.tls.tls_blocks));
    }
  }
  /* FASTCGI */
  else if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_FASTCGI)) {
    flow->info_type = INFO_FASTCGI;
    flow->fast_cgi.method = flow->ndpi_flow->protos.fast_cgi.method;
    ndpi_snprintf(flow->fast_cgi.user_agent, sizeof(flow->fast_cgi.user_agent), "%s", flow->ndpi_flow->protos.fast_cgi.user_agent);
    ndpi_snprintf(flow->fast_cgi.url, sizeof(flow->fast_cgi.url), "%s", flow->ndpi_flow->protos.fast_cgi.url);
  }

  if(!monitoring_enabled) {
    add_to_address_port_list(&flow->stun.mapped_address, &flow->ndpi_flow->stun.mapped_address);
    add_to_address_port_list(&flow->stun.peer_address, &flow->ndpi_flow->stun.peer_address);
    add_to_address_port_list(&flow->stun.relayed_address, &flow->ndpi_flow->stun.relayed_address);
    add_to_address_port_list(&flow->stun.response_origin, &flow->ndpi_flow->stun.response_origin);
    add_to_address_port_list(&flow->stun.other_address, &flow->ndpi_flow->stun.other_address);
  }

  flow->multimedia_flow_types |= flow->ndpi_flow->flow_multimedia_types;

  if(flow->ndpi_flow->tcp.fingerprint) {
    char buf[128];

    snprintf(buf, sizeof(buf), "%s/%s", flow->ndpi_flow->tcp.fingerprint,
       ndpi_print_os_hint(flow->ndpi_flow->tcp.os_hint));
    flow->tcp_fingerprint = ndpi_strdup(buf);
  }

  /* HTTP metadata are "global" not in `flow->ndpi_flow->protos` union; for example, we can have
     HTTP/BitTorrent and in that case we want to export also HTTP attributes */
  if(ndpi_stack_is_http_like(&flow->detected_protocol.protocol_stack)) { /* HTTP, HTTP_PROXY, HTTP_CONNECT */
    if(flow->ndpi_flow->http.url != NULL) {
      ndpi_snprintf(flow->http.url, sizeof(flow->http.url), "%s", flow->ndpi_flow->http.url);
    }

    flow->http.response_status_code = flow->ndpi_flow->http.response_status_code;
    ndpi_snprintf(flow->http.content_type, sizeof(flow->http.content_type), "%s", flow->ndpi_flow->http.content_type ? flow->ndpi_flow->http.content_type : "");
    ndpi_snprintf(flow->http.server, sizeof(flow->http.server), "%s", flow->ndpi_flow->http.server ? flow->ndpi_flow->http.server : "");
    ndpi_snprintf(flow->http.request_content_type, sizeof(flow->http.request_content_type), "%s", flow->ndpi_flow->http.request_content_type ? flow->ndpi_flow->http.request_content_type : "");
    ndpi_snprintf(flow->http.nat_ip, sizeof(flow->http.nat_ip), "%s", flow->ndpi_flow->http.nat_ip ? flow->ndpi_flow->http.nat_ip : "");
    ndpi_snprintf(flow->http.filename, sizeof(flow->http.filename), "%s", flow->ndpi_flow->http.filename ? flow->ndpi_flow->http.filename : "");
    ndpi_snprintf(flow->http.username, sizeof(flow->http.username), "%s", flow->ndpi_flow->http.username ? flow->ndpi_flow->http.username : "");
    ndpi_snprintf(flow->http.password, sizeof(flow->http.password), "%s", flow->ndpi_flow->http.password ? flow->ndpi_flow->http.password : "");
  }

  if(ndpi_stack_contains(&flow->detected_protocol.protocol_stack, NDPI_PROTOCOL_RTP))
    memcpy(&flow->rtp, &flow->ndpi_flow->rtp, sizeof(flow->rtp));

  ndpi_snprintf(flow->http.user_agent,
                sizeof(flow->http.user_agent),
                "%s", (flow->ndpi_flow->http.user_agent ? flow->ndpi_flow->http.user_agent : ""));

  {
    ndpi_ip_addr_t ip_addr;
    struct ndpi_address_cache_item *c;

    memset(&ip_addr, 0, sizeof(ip_addr));

    if(flow->ip_version == 4)
      ip_addr.ipv4 = flow->dst_ip;
    else
      memcpy(&ip_addr.ipv6, &flow->dst_ip6, sizeof(struct ndpi_in6_addr));

    c = ndpi_cache_address_find(workflow->ndpi_struct, ip_addr);

    if(c) {
      flow->server_hostname = ndpi_strdup(c->hostname);
    }
  }

  if (workflow->ndpi_serialization_format != ndpi_serialization_format_unknown) {
    if (ndpi_flow2json(workflow->ndpi_struct, flow->ndpi_flow,
                       flow->ip_version, flow->protocol,
           flow->vlan_id,
                       flow->src_ip, flow->dst_ip,
                       &flow->src_ip6, &flow->dst_ip6,
                       flow->src_port, flow->dst_port,
                       flow->detected_protocol,
                       &flow->ndpi_flow_serializer) != 0) {
      LOG(NDPI_LOG_ERROR, "flow2json failed\n");
      return;
    }

    ndpi_serialize_string_uint32(&flow->ndpi_flow_serializer, "detection_completed", flow->detection_completed);
    ndpi_serialize_string_uint32(&flow->ndpi_flow_serializer, "check_extra_packets", flow->check_extra_packets);

    if(flow->ndpi_flow->state == NDPI_STATE_MONITORING) {
      serialize_monitoring_metadata(flow);
    }

    if(flow->server_hostname)
      ndpi_serialize_string_string(&flow->ndpi_flow_serializer, "server_hostname", flow->server_hostname);
  }

  if(flow->detection_completed && (!flow->check_extra_packets)) {
    flow->flow_payload = flow->ndpi_flow->flow_payload, flow->flow_payload_len = flow->ndpi_flow->flow_payload_len;
    flow->ndpi_flow->flow_payload = NULL; /* We'll free the memory */

    if(workflow->flow_callback != NULL)
      workflow->flow_callback(workflow, flow, workflow->flow_callback_userdata);

    if(fingerprint_fp)
      dump_flow_fingerprint(workflow, flow);

    ndpi_free_flow_info_half(flow);
  }
}

/* ****************************************************** */

/**
 * @brief Clear entropy stats if it meets prereq.
 */
static void
ndpi_clear_entropy_stats(struct ndpi_flow_info *flow) {
  if(enable_flow_stats) {
    if(flow->entropy->src2dst_pkt_count + flow->entropy->dst2src_pkt_count == max_num_packets_per_flow) {
      memcpy(flow->last_entropy, flow->entropy,  sizeof(struct ndpi_entropy));
      memset(flow->entropy, 0x00, sizeof(struct ndpi_entropy));
    }
  }
}

void update_tcp_flags_count(struct ndpi_flow_info* flow, struct ndpi_tcphdr* tcp, u_int8_t src_to_dst_direction){
  if(tcp->cwr){
    flow->cwr_count++;
    src_to_dst_direction ? flow->src2dst_cwr_count++ : flow->dst2src_cwr_count++;
  }
  if(tcp->ece){
    flow->ece_count++;
    src_to_dst_direction ? flow->src2dst_ece_count++ : flow->dst2src_ece_count++;
  }
  if(tcp->rst){
    flow->rst_count++;
    src_to_dst_direction ? flow->src2dst_rst_count++ : flow->dst2src_rst_count++;
  }
  if(tcp->ack){
    flow->ack_count++;
    src_to_dst_direction ? flow->src2dst_ack_count++ : flow->dst2src_ack_count++;
  }
  if(tcp->fin){
    flow->fin_count++;
    src_to_dst_direction ? flow->src2dst_fin_count++ : flow->dst2src_fin_count++;
  }
  if(tcp->syn){
    flow->syn_count++;
    src_to_dst_direction ? flow->src2dst_syn_count++ : flow->dst2src_syn_count++;
  }
  if(tcp->psh){
    flow->psh_count++;
    src_to_dst_direction ? flow->src2dst_psh_count++ : flow->dst2src_psh_count++;
  }
  if(tcp->urg){
    flow->urg_count++;
    src_to_dst_direction ? flow->src2dst_urg_count++ : flow->dst2src_urg_count++;
  }
}

/* ****************************************************** */

/**
   Function to process the packet:
   determine the flow of a packet and try to decode it
   @return: 0 if success; else != 0

   @Note: ipsize = header->len - ip_offset ; rawsize = header->len
*/
static struct ndpi_proto packet_processing(struct ndpi_workflow * workflow,
             const u_int64_t time_ms,
             u_int16_t vlan_id,
             ndpi_packet_tunnel tunnel_type,
             const struct ndpi_iphdr *iph,
             struct ndpi_ipv6hdr *iph6,
             u_int16_t ipsize, u_int16_t rawsize,
             const struct pcap_pkthdr *header,
             const u_char *packet,
             pkt_timeval when,
             ndpi_risk *flow_risk,
             struct ndpi_flow_info **flow_ext) {
  struct ndpi_flow_info *flow = NULL;
  struct ndpi_flow_struct *ndpi_flow = NULL;
  u_int8_t proto;
  struct ndpi_tcphdr *tcph = NULL;
  struct ndpi_udphdr *udph = NULL;
  u_int16_t sport, dport, payload_len = 0;
  u_int8_t *payload;
  u_int8_t src_to_dst_direction = 1;
  u_int8_t begin_or_end_tcp = 0;
  struct ndpi_proto nproto;

  memset(&nproto, '\0', sizeof(nproto));

  if(workflow->prefs.ignore_vlanid)
    vlan_id = 0;

  if(iph)
    flow = get_ndpi_flow_info(workflow, IPVERSION, vlan_id,
            tunnel_type, iph, NULL,
            ipsize,
            ntohs(iph->tot_len) ? (ntohs(iph->tot_len) - (iph->ihl * 4)) : ipsize - (iph->ihl * 4) /* TSO */,
            iph->ihl * 4,
            &tcph, &udph, &sport, &dport,
            &proto,
            &payload, &payload_len, &src_to_dst_direction, when);
  else
    flow = get_ndpi_flow_info6(workflow, vlan_id,
             tunnel_type, iph6, ipsize,
             &tcph, &udph, &sport, &dport,
             &proto,
             &payload, &payload_len, &src_to_dst_direction, when);

  if(flow != NULL) {
    pkt_timeval tdiff;

    workflow->stats.ip_packet_count++;
    workflow->stats.total_wire_bytes += rawsize + 24 /* CRC etc */,
      workflow->stats.total_ip_bytes += rawsize;
    ndpi_flow = flow->ndpi_flow;

    if(tcph != NULL){
      update_tcp_flags_count(flow, tcph, src_to_dst_direction);
      if(tcph->syn && !flow->src2dst_bytes){
  flow->c_to_s_init_win = rawsize;
      }else if(tcph->syn && tcph->ack && flow->src2dst_bytes == flow->c_to_s_init_win){
  flow->s_to_c_init_win = rawsize;
      }
    }

    if((tcph != NULL) && (tcph->fin || tcph->rst || tcph->syn))
      begin_or_end_tcp = 1;

    if(flow->flow_last_pkt_time.tv_sec) {
      ndpi_timer_sub(&when, &flow->flow_last_pkt_time, &tdiff);

      if(flow->iat_flow
   && (tdiff.tv_sec >= 0) /* Discard backward time */
   ) {
  u_int64_t ms = ndpi_timeval_to_milliseconds(tdiff);

  if(ms > 0)
    ndpi_data_add_value(flow->iat_flow, ms);
      }
    }

    memcpy(&flow->flow_last_pkt_time, &when, sizeof(when));

    if(src_to_dst_direction) {
      if(flow->src2dst_last_pkt_time.tv_sec) {
  ndpi_timer_sub(&when, &flow->src2dst_last_pkt_time, &tdiff);

  if(flow->iat_c_to_s
     && (tdiff.tv_sec >= 0) /* Discard backward time */
     ) {
    u_int64_t ms = ndpi_timeval_to_milliseconds(tdiff);

    ndpi_data_add_value(flow->iat_c_to_s, ms);
  }
      }

      ndpi_data_add_value(flow->pktlen_c_to_s, rawsize);
      flow->src2dst_packets++, flow->src2dst_bytes += rawsize, flow->src2dst_goodput_bytes += payload_len;
      memcpy(&flow->src2dst_last_pkt_time, &when, sizeof(when));

#ifdef DIRECTION_BINS
      if(payload_len && (flow->src2dst_packets < MAX_NUM_BIN_PKTS))
  ndpi_inc_bin(&flow->payload_len_bin_src2dst, plen2slot(payload_len));
#endif
    } else {
      if(flow->dst2src_last_pkt_time.tv_sec && (!begin_or_end_tcp)) {
  ndpi_timer_sub(&when, &flow->dst2src_last_pkt_time, &tdiff);

  if(flow->iat_s_to_c) {
    u_int64_t ms = ndpi_timeval_to_milliseconds(tdiff);

    ndpi_data_add_value(flow->iat_s_to_c, ms);
  }
      }
      ndpi_data_add_value(flow->pktlen_s_to_c, rawsize);
      flow->dst2src_packets++, flow->dst2src_bytes += rawsize, flow->dst2src_goodput_bytes += payload_len;
      flow->risk &= ~(1ULL << NDPI_UNIDIRECTIONAL_TRAFFIC); /* Clear bit */
      memcpy(&flow->dst2src_last_pkt_time, &when, sizeof(when));

#ifdef DIRECTION_BINS
      if(payload_len && (flow->dst2src_packets < MAX_NUM_BIN_PKTS))
  ndpi_inc_bin(&flow->payload_len_bin_dst2src, plen2slot(payload_len));
#endif
    }

#ifndef DIRECTION_BINS
    if(payload_len && ((flow->src2dst_packets+flow->dst2src_packets) < MAX_NUM_BIN_PKTS)) {
#if 0
      /* Discard packets until the protocol is detected */
      if(flow->detected_protocol.proto.app_protocol != NDPI_PROTOCOL_UNKNOWN)
#endif
  ndpi_inc_bin(&flow->payload_len_bin, plen2slot(payload_len), 1);
    }
#endif

    if(enable_payload_analyzer && (payload_len > 0))
      ndpi_payload_analyzer(flow,
          payload, payload_len,
          workflow->stats.ip_packet_count);

    if(enable_flow_stats) {
      /* Update BD, distribution and mean. */
      ndpi_flow_update_byte_count(flow, payload, payload_len, src_to_dst_direction);
      ndpi_flow_update_byte_dist_mean_var(flow, payload, payload_len, src_to_dst_direction);
      /* Update SPLT scores for first 32 packets. */
      if((flow->entropy->src2dst_pkt_count+flow->entropy->dst2src_pkt_count) <= max_num_packets_per_flow) {
        if(flow->bidirectional)
          flow->entropy->score = ndpi_classify(flow->entropy->src2dst_pkt_len, flow->entropy->src2dst_pkt_time,
                flow->entropy->dst2src_pkt_len, flow->entropy->dst2src_pkt_time,
                flow->entropy->src2dst_start, flow->entropy->dst2src_start,
                max_num_packets_per_flow, ntohs(flow->src_port), ntohs(flow->dst_port),
                flow->src2dst_packets, flow->dst2src_packets,
                flow->entropy->src2dst_opackets, flow->entropy->dst2src_opackets,
                flow->entropy->src2dst_l4_bytes, flow->entropy->dst2src_l4_bytes, 1,
                flow->entropy->src2dst_byte_count, flow->entropy->dst2src_byte_count);
  else
    flow->entropy->score = ndpi_classify(flow->entropy->src2dst_pkt_len, flow->entropy->src2dst_pkt_time,
                NULL, NULL, flow->entropy->src2dst_start, flow->entropy->src2dst_start,
                max_num_packets_per_flow, ntohs(flow->src_port), ntohs(flow->dst_port),
                flow->src2dst_packets, 0,
                flow->entropy->src2dst_opackets, 0,
                flow->entropy->src2dst_l4_bytes, 0, 1,
                flow->entropy->src2dst_byte_count, NULL);
      }
    }

    if(flow->first_seen_ms == 0)
      flow->first_seen_ms = time_ms;

    flow->last_seen_ms = time_ms;

    /* Copy packets entropy if num packets count == 10 */
    ndpi_clear_entropy_stats(flow);
    /* Reset IAT reeference times (see https://github.com/ntop/nDPI/pull/1316) */
    if(((flow->src2dst_packets + flow->dst2src_packets) % max_num_packets_per_flow) == 0) {
      memset(&flow->src2dst_last_pkt_time, '\0', sizeof(flow->src2dst_last_pkt_time));
      memset(&flow->dst2src_last_pkt_time, '\0', sizeof(flow->dst2src_last_pkt_time));
      memset(&flow->flow_last_pkt_time, '\0', sizeof(flow->flow_last_pkt_time));
    }

    if((human_readeable_string_len != 0) && (!flow->has_human_readeable_strings)) {
      u_int8_t skip = 0;

      if(proto == IPPROTO_TCP &&
   (is_ndpi_proto(flow, NDPI_PROTOCOL_TLS) ||
    is_ndpi_proto(flow, NDPI_PROTOCOL_SSH))) {
  if((flow->src2dst_packets+flow->dst2src_packets) < 10 /* MIN_NUM_ENCRYPT_SKIP_PACKETS */)
    skip = 1; /* Skip initial negotiation packets */
      }

      if((!skip) && ((flow->src2dst_packets+flow->dst2src_packets) < 100)) {
  if(ndpi_has_human_readable_string((char*)packet, header->caplen,
             human_readeable_string_len,
             flow->human_readeable_string_buffer,
             sizeof(flow->human_readeable_string_buffer)) == 1)
    flow->has_human_readeable_strings = 1;
      }
    } else {
      if(proto == IPPROTO_TCP &&
         (is_ndpi_proto(flow, NDPI_PROTOCOL_TLS) ||
          is_ndpi_proto(flow, NDPI_PROTOCOL_SSH)))
  flow->has_human_readeable_strings = 0;
    }
  } else { // flow is NULL
    workflow->stats.total_discarded_bytes += header->len;
    return(nproto);
  }

  if(!flow->detection_completed) {
    struct ndpi_flow_input_info input_info;

    u_int enough_packets =
      ((proto == IPPROTO_UDP && (max_num_udp_dissected_pkts > 0 && flow->src2dst_packets + flow->dst2src_packets >= max_num_udp_dissected_pkts)) ||
       (proto == IPPROTO_TCP && (max_num_tcp_dissected_pkts > 0 && flow->src2dst_packets + flow->dst2src_packets >= max_num_tcp_dissected_pkts))) ? 1 : 0;

#if 0
    printf("%s()\n", __FUNCTION__);
#endif

    if(proto == IPPROTO_TCP)
      workflow->stats.dpi_packet_count[0]++;
    else if(proto == IPPROTO_UDP)
      workflow->stats.dpi_packet_count[1]++;
    else
      workflow->stats.dpi_packet_count[2]++;
    flow->dpi_packets++;

    memset(&input_info, '\0', sizeof(input_info)); /* To be sure to set to "unknown" any fields */
    /* Set here any information (easily) available; in this trivial example we don't have any */
    input_info.in_pkt_dir = NDPI_IN_PKT_DIR_UNKNOWN;
    input_info.seen_flow_beginning = NDPI_FLOW_BEGINNING_UNKNOWN;
    malloc_size_stats = 1;
    flow->detected_protocol = ndpi_detection_process_packet(workflow->ndpi_struct, ndpi_flow,
                  iph ? (uint8_t *)iph : (uint8_t *)iph6,
                  ipsize, time_ms, &input_info);
    if(monitoring_enabled)
      process_ndpi_monitoring_info(flow);
    if(flow->detected_protocol.state == NDPI_STATE_CLASSIFIED ||
       enough_packets) {

      flow->detection_completed = 1;

      if(flow->detected_protocol.state != NDPI_STATE_CLASSIFIED) {
         flow->detected_protocol = ndpi_detection_giveup(workflow->ndpi_struct, flow->ndpi_flow);
      }

      if(flow->ndpi_flow->protocol_was_guessed) workflow->stats.guessed_flow_protocols++;
      process_ndpi_collected_info(workflow, flow);
    }

    /* Let's try to save client-server direction */
    flow->current_pkt_from_client_to_server = input_info.in_pkt_dir;

    malloc_size_stats = 0;
  } else {
    flow->current_pkt_from_client_to_server = NDPI_IN_PKT_DIR_UNKNOWN; /* Unknown */
  }

#if 0
  if(flow->risk != 0) {
    FILE *r = fopen("/tmp/e", "a");

    if(r) {
      fprintf(r, "->>> %u [%08X]\n", flow->risk, flow->risk);
      fclose(r);
    }
  }
#endif

  *flow_risk = flow->risk;
  *flow_ext = flow;

  return(flow->detected_protocol);
}

/* ****************************************************** */

int ndpi_is_datalink_supported(int datalink_type) {
  /* Keep in sync with the similar switch in ndpi_workflow_process_packet */
  switch(datalink_type) {
  case DLT_NULL:
  case DLT_PPP_SERIAL:
  case DLT_C_HDLC:
  case DLT_PPP:
#ifdef DLT_IPV4
  case DLT_IPV4:
#endif
#ifdef DLT_IPV6
  case DLT_IPV6:
#endif
  case DLT_EN10MB:
  case DLT_LINUX_SLL:
  case DLT_IEEE802_11_RADIO:
  case DLT_RAW:
  case DLT_PPI:
  case LINKTYPE_LINUX_SLL2:
    return 1;
  default:
    return 0;
  }
}

static bool ndpi_is_valid_vxlan(const struct pcap_pkthdr *header, const u_char *packet, u_int16_t ip_offset, u_int16_t ip_len){
  if(header->caplen < ip_offset + ip_len + sizeof(struct ndpi_udphdr) + sizeof(struct ndpi_vxlanhdr)) {
    return false;
  }
  u_int32_t vxlan_dst_port  = ntohs(4789);
  struct ndpi_udphdr *udp = (struct ndpi_udphdr *)&packet[ip_offset+ip_len];
  u_int offset = ip_offset + ip_len + sizeof(struct ndpi_udphdr);
  /**
   * rfc-7348
   *    VXLAN Header:  This is an 8-byte field that has:

    - Flags (8 bits): where the I flag MUST be set to 1 for a valid
      VXLAN Network ID (VNI).  The other 7 bits (designated "R") are
      reserved fields and MUST be set to zero on transmission and
      ignored on receipt.

    - VXLAN Segment ID/VXLAN Network Identifier (VNI): this is a
      24-bit value used to designate the individual VXLAN overlay
      network on which the communicating VMs are situated.  VMs in
      different VXLAN overlay networks cannot communicate with each
      other.

    - Reserved fields (24 bits and 8 bits): MUST be set to zero on
      transmission and ignored on receipt.
         VXLAN Header:
    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
    |R|R|R|R|I|R|R|R|            Reserved                           |
    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
    |                VXLAN Network Identifier (VNI) |   Reserved    |
    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  */
  if((udp->dest == vxlan_dst_port || udp->source == vxlan_dst_port) &&
    (packet[offset] == 0x8) &&
    (packet[offset + 1] == 0x0) &&
    (packet[offset + 2] == 0x0) &&
    (packet[offset + 3] == 0x0) &&
    (packet[offset + 7] ==  0x0)) {
    return true;
    }
  return false;
}

static inline u_int ndpi_skip_vxlan(u_int16_t ip_offset, u_int16_t ip_len){
  return ip_offset + ip_len + sizeof(struct ndpi_udphdr) + sizeof(struct ndpi_vxlanhdr);
}

static uint32_t ndpi_is_valid_gre_tunnel(const struct pcap_pkthdr *header,
                const u_char *packet, const u_int16_t ip_offset,
                const u_int16_t ip_len) {
  uint32_t offset = ip_offset + ip_len;
  struct ndpi_gre_basehdr *grehdr = (struct ndpi_gre_basehdr*)&packet[offset];
  offset += sizeof(struct ndpi_gre_basehdr);
  /*
    The GRE flags are encoded in the first two octets.  Bit 0 is the
    most significant bit, bit 15 is the least significant bit.  Bits
    13 through 15 are reserved for the Version field.  Bits 9 through
    12 are reserved for future use and MUST be transmitted as zero.
  */
  if(NDPI_GRE_IS_FLAGS(grehdr->flags))
    return 0;
  if(NDPI_GRE_IS_REC(grehdr->flags))
    return 0;
  /*GRE rfc 2890 that update 1701*/
  if(NDPI_GRE_IS_VERSION_0(grehdr->flags)) {
    if(NDPI_GRE_IS_CSUM(grehdr->flags)) {
      if(header->caplen < offset + 4)
        return 0;
      /*checksum field and offset field*/
      offset += 4;
    }
    if(NDPI_GRE_IS_KEY(grehdr->flags)) {
      if(header->caplen < offset + 4)
        return 0;
      offset += 4;
    }
    if(NDPI_GRE_IS_SEQ(grehdr->flags)) {
      if(header->caplen < offset + 4)
        return 0;
      offset += 4;
    }
  } else if(NDPI_GRE_IS_VERSION_1(grehdr->flags)) { /*rfc-2637 section 4.1 enhanced gre*/
    if(NDPI_GRE_IS_CSUM(grehdr->flags))
      return 0;
    if(NDPI_GRE_IS_ROUTING(grehdr->flags))
      return 0;
    if(!NDPI_GRE_IS_KEY(grehdr->flags))
      return 0;
    if(NDPI_GRE_IS_STRICT(grehdr->flags))
      return 0;
    if(grehdr->protocol != NDPI_GRE_PROTO_PPP)
      return 0;
    /*key field*/
    if(header->caplen < offset + 4)
      return 0;
    offset += 4;
    if(NDPI_GRE_IS_SEQ(grehdr->flags)) {
      if(header->caplen < offset + 4)
        return 0;
      offset += 4;
    }
    if(NDPI_GRE_IS_ACK(grehdr->flags)) {
      if(header->caplen < offset + 4)
        return 0;
      offset += 4;
    }
  } else { /*support only ver 0, 1*/
    return 0;
  }
  return offset;
}

/* ****************************************************** */

struct ndpi_proto ndpi_workflow_process_packet(struct ndpi_workflow * workflow,
                 const struct pcap_pkthdr *header,
                 const u_char *packet,
                 ndpi_risk *flow_risk,
                 struct ndpi_flow_info **flow) {
  /*
   * Declare pointers to packet headers
   */
  /* --- Ethernet header --- */
  const struct ndpi_ethhdr *ethernet;
  /* --- LLC header --- */
  const struct ndpi_llc_header_snap *llc;

  /* --- Cisco HDLC header --- */
  const struct ndpi_chdlc *chdlc;

  /* --- Radio Tap header --- */
  const struct ndpi_radiotap_header *radiotap;
  /* --- Wifi header --- */
  const struct ndpi_wifi_header *wifi;

  /* --- MPLS header --- */
  union mpls {
    uint32_t u32;
    struct ndpi_mpls_header mpls;
  } mpls;

  /** --- IP header --- **/
  struct ndpi_iphdr *iph;
  /** --- IPv6 header --- **/
  struct ndpi_ipv6hdr *iph6;

  struct ndpi_proto nproto;
  ndpi_packet_tunnel tunnel_type = ndpi_no_tunnel;

  /* lengths and offsets */
  u_int32_t eth_offset = 0, dlt;
  u_int16_t radio_len, header_length;
  u_int16_t fc;
  u_int16_t type = 0;
  int wifi_len = 0;
  int pyld_eth_len = 0;
  int check;
  u_int64_t time_ms;
  u_int16_t ip_offset = 0, ip_len;
  u_int16_t frag_off = 0, vlan_id = 0;
  u_int8_t proto = 0, recheck_type;
  u_int8_t ip_ver, ppp_type;
  /*u_int32_t label;*/

  /* counters */
  u_int8_t vlan_packet = 0;

  *flow_risk = 0 /* NDPI_NO_RISK */;
  *flow = NULL;

  memset(&nproto, '\0', sizeof(nproto));

  if((addr_dump_path != NULL) && (workflow->stats.raw_packet_count == 0)) {
    /* At the first packet flush expired cached addresses */
    ndpi_cache_address_flush_expired(workflow->ndpi_struct, header->ts.tv_sec);
  }

  /* Increment raw packet counter */
  workflow->stats.raw_packet_count++;

  /* setting time */
  time_ms = ((uint64_t) header->ts.tv_sec) * TICK_RESOLUTION + header->ts.tv_usec / (1000000 / TICK_RESOLUTION);

  /* safety check */
  if(workflow->last_time > time_ms) {
    /* printf("\nWARNING: timestamp bug in the pcap file (ts delta: %llu, repairing)\n", ndpi_thread_info[thread_id].last_time - time); */
    time_ms = workflow->last_time;
  }
  /* update last time value */
  workflow->last_time = time_ms;

  /*** check Data Link type ***/
  int datalink_type;

#ifdef USE_DPDK
  datalink_type = DLT_EN10MB;
#else
  datalink_type = (int)pcap_datalink(workflow->pcap_handle);
#endif

 datalink_check:
  // 20 for min iph and 8 for min UDP
  if(header->caplen < eth_offset + 28)
    return(nproto); /* Too short */

  /* Keep in sync with ndpi_is_datalink_supported() */
  switch(datalink_type) {
  case DLT_NULL:
    if(ntohl(*((u_int32_t*)&packet[eth_offset])) == 2)
      type = ETH_P_IP;
    else
      type = ETH_P_IPV6;

    ip_offset = 4 + eth_offset;
    break;

    /* Cisco PPP in HDLC-like framing - 50 */
  case DLT_PPP_SERIAL:
    chdlc = (struct ndpi_chdlc *) &packet[eth_offset];
    ip_offset = eth_offset + sizeof(struct ndpi_chdlc); /* CHDLC_OFF = 4 */
    type = ntohs(chdlc->proto_code);
    break;

    /* Cisco PPP - 9 or 104 */
  case DLT_C_HDLC:
  case DLT_PPP:
    if(packet[0] == 0x0f || packet[0] == 0x8f) {
      chdlc = (struct ndpi_chdlc *) &packet[eth_offset];
      ip_offset = eth_offset + sizeof(struct ndpi_chdlc); /* CHDLC_OFF = 4 */
      type = ntohs(chdlc->proto_code);
    } else {
      ip_offset = eth_offset + 2;
      ppp_type = ntohs(*((u_int16_t*)&packet[eth_offset]));
      if(ppp_type == 0x0021)
        type = ETH_P_IP;
      else if(ppp_type == 0x0057)
        type = ETH_P_IPV6;
      else
        return(nproto);
    }
    break;

#ifdef DLT_IPV4
  case DLT_IPV4:
    type = ETH_P_IP;
    ip_offset = eth_offset;
    break;
#endif

#ifdef DLT_IPV6
  case DLT_IPV6:
    type = ETH_P_IPV6;
    ip_offset = eth_offset;
    break;
#endif

    /* IEEE 802.3 Ethernet - 1 */
  case DLT_EN10MB:
    ethernet = (struct ndpi_ethhdr *) &packet[eth_offset];
    ip_offset = sizeof(struct ndpi_ethhdr) + eth_offset;
    check = ntohs(ethernet->h_proto);

    if(check <= 1500)
      pyld_eth_len = check;
    else if(check >= 1536)
      type = check;

    if(pyld_eth_len != 0) {
      llc = (struct ndpi_llc_header_snap *)(&packet[ip_offset]);
      /* check for LLC layer with SNAP extension */
      if(llc->dsap == SNAP || llc->ssap == SNAP) {
  type = llc->snap.proto_ID;
  ip_offset += + 8;
      }
      /* No SNAP extension - Spanning Tree pkt must be discarted */
      else if(llc->dsap == BSTP || llc->ssap == BSTP) {
  goto v4_warning;
      }
    }
    break;

    /* Linux Cooked Capture - 113 */
  case DLT_LINUX_SLL:
    type = (packet[eth_offset+14] << 8) + packet[eth_offset+15];
    ip_offset = 16 + eth_offset;
    break;

    /* Linux Cooked Capture v2 - 276 */
  case LINKTYPE_LINUX_SLL2:
    type = (packet[eth_offset+10] << 8) + packet[eth_offset+11];
    ip_offset = 20 + eth_offset;
    break;

    /* Radiotap link-layer - 127 */
  case DLT_IEEE802_11_RADIO:
    radiotap = (struct ndpi_radiotap_header *) &packet[eth_offset];
    radio_len = radiotap->len;

    /* Check Bad FCS presence */
    if((radiotap->flags & BAD_FCS) == BAD_FCS) {
      workflow->stats.total_discarded_bytes +=  header->len;
      return(nproto);
    }

    if(header->caplen < (eth_offset + radio_len + sizeof(struct ndpi_wifi_header)))
      return(nproto);

    /* Calculate 802.11 header length (variable) */
    wifi = (struct ndpi_wifi_header*)( packet + eth_offset + radio_len);
    fc = wifi->fc;

    /* check wifi data presence */
    if(FCF_TYPE(fc) == WIFI_DATA) {
      if((FCF_TO_DS(fc) && FCF_FROM_DS(fc) == 0x0) ||
   (FCF_TO_DS(fc) == 0x0 && FCF_FROM_DS(fc)))
  wifi_len = 26; /* + 4 byte fcs */
    } else   /* no data frames */
      return(nproto);

    /* Check ether_type from LLC */
    if(header->caplen < (eth_offset + wifi_len + radio_len + sizeof(struct ndpi_llc_header_snap)))
      return(nproto);
    llc = (struct ndpi_llc_header_snap*)(packet + eth_offset + wifi_len + radio_len);
    if(llc->dsap == SNAP)
      type = ntohs(llc->snap.proto_ID);

    /* Set IP header offset */
    ip_offset = wifi_len + radio_len + sizeof(struct ndpi_llc_header_snap) + eth_offset;
    break;

  case DLT_RAW:
    ip_offset = eth_offset;
    /* Heuristic: no explicit field with next protocol */
    ip_ver = (packet[ip_offset] & 0xF0) >> 4;
    if(ip_ver == 4)
      type = ETH_P_IP;
    else if(ip_ver == 6)
      type = ETH_P_IPV6;
    else
      return(nproto);

    break;

  case DLT_PPI:
    header_length = le16toh(*(u_int16_t *)&packet[eth_offset + 2]);
    dlt = le32toh(*(u_int32_t *)&packet[eth_offset + 4]);
    if(dlt != DLT_EN10MB) /* Handle only standard ethernet, for the time being */
      return(nproto);
    datalink_type = DLT_EN10MB;
    eth_offset += header_length;
    goto datalink_check;

  default:
    /*
     * We shoudn't be here, because we already checked that this datalink is supported.
     * Should ndpi_is_datalink_supported() be updated?
     */
    printf("Unknown datalink %d\n", datalink_type);
    return(nproto);
  }

 ether_type_check:
  recheck_type = 0;

  /* check ether type */
  switch(type) {
  case ETH_P_VLAN:
    if(ip_offset+4 >= (int)header->caplen)
      return(nproto);
    vlan_id = ((packet[ip_offset] << 8) + packet[ip_offset+1]) & 0xFFF;
    type = (packet[ip_offset+2] << 8) + packet[ip_offset+3];
    ip_offset += 4;
    vlan_packet = 1;

    // double tagging for 802.1Q
    while((type == 0x8100) && (((bpf_u_int32)ip_offset+4) < header->caplen)) {
      vlan_id = ((packet[ip_offset] << 8) + packet[ip_offset+1]) & 0xFFF;
      type = (packet[ip_offset+2] << 8) + packet[ip_offset+3];
      ip_offset += 4;
    }
    recheck_type = 1;
    break;

  case ETH_P_MPLS_UNI:
  case ETH_P_MPLS_MULTI:
    if(ip_offset+4 >= (int)header->caplen)
      return(nproto);
    mpls.u32 = *((uint32_t *) &packet[ip_offset]);
    mpls.u32 = ntohl(mpls.u32);
    workflow->stats.mpls_count++;
    type = ETH_P_IP, ip_offset += 4;

    while(!mpls.mpls.s && (((bpf_u_int32)ip_offset) + 4 < header->caplen)) {
      mpls.u32 = *((uint32_t *) &packet[ip_offset]);
      mpls.u32 = ntohl(mpls.u32);
      ip_offset += 4;
    }
    recheck_type = 1;
    break;

  case ETH_P_PPPoE:
    workflow->stats.pppoe_count++;
    type = ETH_P_IP;
    ip_offset += 8;
    recheck_type = 1;
    break;

  case ETH_P_IP:
  case ETH_P_IPV6:
    /* Good let's keep decoding */
    break;

  default:
    return(nproto);
  }

  if(recheck_type)
    goto ether_type_check;

  workflow->stats.vlan_count += vlan_packet;

 iph_check:
  /* Check and set IP header size and total packet length */
  if(header->caplen < ip_offset + sizeof(struct ndpi_iphdr))
    return(nproto); /* Too short for next IP header*/

  iph = (struct ndpi_iphdr *) &packet[ip_offset];

  /* just work on Ethernet packets that contain IP */
  if(type == ETH_P_IP && header->caplen >= ip_offset) {
    frag_off = ntohs(iph->frag_off);

    proto = iph->protocol;
    if(header->caplen < header->len) {
      static u_int8_t cap_warning_used = 0;

      if(cap_warning_used == 0) {
  if(!workflow->prefs.quiet_mode)
    LOG(NDPI_LOG_DEBUG,
       "\n\nWARNING: packet capture size is smaller than packet size, DETECTION MIGHT NOT WORK CORRECTLY\n\n");
  cap_warning_used = 1;
      }
    }
  }

  if(iph->version == IPVERSION) {
    ip_len = ((u_int16_t)iph->ihl * 4);
    iph6 = NULL;

    if(iph->protocol == IPPROTO_IPV6
       || iph->protocol == NDPI_IPIP_PROTOCOL_TYPE
       ) {
      ip_offset += ip_len;
      if(ip_len > 0)
        goto iph_check;
    }

    if((frag_off & 0x1FFF) != 0) {
      static u_int8_t ipv4_frags_warning_used = 0;
      workflow->stats.fragmented_count++;

      if(ipv4_frags_warning_used == 0) {
  if(!workflow->prefs.quiet_mode)
    LOG(NDPI_LOG_DEBUG, "\n\nWARNING: IPv4 fragments are not handled by this demo (nDPI supports them)\n");
  ipv4_frags_warning_used = 1;
      }

      workflow->stats.total_discarded_bytes +=  header->len;
      return(nproto);
    }
  } else if(iph->version == 6) {
    if(header->caplen < ip_offset + sizeof(struct ndpi_ipv6hdr))
      return(nproto); /* Too short for IPv6 header*/

    iph6 = (struct ndpi_ipv6hdr *)&packet[ip_offset];
    proto = iph6->ip6_hdr.ip6_un1_nxt;
    ip_len = ntohs(iph6->ip6_hdr.ip6_un1_plen);

    if(header->caplen < (ip_offset + sizeof(struct ndpi_ipv6hdr) + ntohs(iph6->ip6_hdr.ip6_un1_plen)))
      return(nproto); /* Too short for IPv6 payload*/

    const u_int8_t *l4ptr = (((const u_int8_t *) iph6) + sizeof(struct ndpi_ipv6hdr));
    u_int16_t ipsize = header->caplen - ip_offset;

    if(ndpi_handle_ipv6_extension_headers(ipsize - sizeof(struct ndpi_ipv6hdr), &l4ptr, &ip_len, &proto) != 0) {
      return(nproto);
    }

    if(proto == IPPROTO_IPV6
       || proto == NDPI_IPIP_PROTOCOL_TYPE
       ) {
      if(l4ptr > packet) { /* Better safe than sorry */
        ip_offset = (l4ptr - packet);
        goto iph_check;
      }
    }

    iph = NULL;
  } else {
    static u_int8_t ipv4_warning_used = 0;

  v4_warning:
    if(ipv4_warning_used == 0) {
      if(!workflow->prefs.quiet_mode)
        LOG(NDPI_LOG_DEBUG,
     "\n\nWARNING: only IPv4/IPv6 packets are supported in this demo (nDPI supports both IPv4 and IPv6), all other packets will be discarded\n\n");
      ipv4_warning_used = 1;
    }

    workflow->stats.total_discarded_bytes +=  header->len;
    return(nproto);
  }

  if(workflow->prefs.decode_tunnels && (proto == IPPROTO_UDP)) {
    if(header->caplen < ip_offset + ip_len + sizeof(struct ndpi_udphdr))
      return(nproto); /* Too short for UDP header*/
    else {
      struct ndpi_udphdr *udp = (struct ndpi_udphdr *)&packet[ip_offset+ip_len];
      u_int16_t sport = ntohs(udp->source), dport = ntohs(udp->dest);

      if(((sport == GTP_U_V1_PORT) || (dport == GTP_U_V1_PORT)) &&
         (ip_offset + ip_len + sizeof(struct ndpi_udphdr) + 8 /* Minimum GTPv1 header len */ < header->caplen)) {
  /* Check if it's GTPv1 */
  u_int offset = ip_offset+ip_len+sizeof(struct ndpi_udphdr);
  u_int8_t flags = packet[offset];
  u_int8_t message_type = packet[offset+1];
  u_int8_t exts_parsing_error = 0;

  if((((flags & 0xE0) >> 5) == 1 /* GTPv1 */) &&
     (message_type == 0xFF /* T-PDU */)) {

    offset += 8; /* GTPv1 header len */
    if(flags & 0x07)
      offset += 4; /* sequence_number + pdu_number + next_ext_header fields */
    /* Extensions parsing */
    if(flags & 0x04) {
      unsigned int ext_length = 0;

      while(offset < header->caplen) {
        ext_length = packet[offset] << 2;
        offset += ext_length;
        if(offset >= header->caplen || ext_length == 0) {
          exts_parsing_error = 1;
          break;
        }
        if(packet[offset - 1] == 0)
          break;
      }
    }

    if(offset < header->caplen && !exts_parsing_error) {
      /* Ok, valid GTP-U */
      tunnel_type = ndpi_gtp_tunnel;
      ip_offset = offset;
      iph = (struct ndpi_iphdr *)&packet[ip_offset];
      if(iph->version == 6) {
        iph6 = (struct ndpi_ipv6hdr *)&packet[ip_offset];
        iph = NULL;
              if(header->caplen < ip_offset + sizeof(struct ndpi_ipv6hdr))
          return(nproto);
      } else if(iph->version != IPVERSION) {
        // printf("WARNING: not good (packet_id=%u)!\n", (unsigned int)workflow->stats.raw_packet_count);
        goto v4_warning;
      } else {
              if(header->caplen < ip_offset + sizeof(struct ndpi_iphdr))
          return(nproto);
      }
    }
  }
      } else if((sport == TZSP_PORT) || (dport == TZSP_PORT)) {
  /* https://en.wikipedia.org/wiki/TZSP */
  if(header->caplen < ip_offset + ip_len + sizeof(struct ndpi_udphdr) + 4)
    return(nproto); /* Too short for TZSP*/

  u_int offset           = ip_offset+ip_len+sizeof(struct ndpi_udphdr);
  u_int8_t version       = packet[offset];
  u_int8_t ts_type       = packet[offset+1];
  u_int16_t encapsulates = ntohs(*((u_int16_t*)&packet[offset+2]));

  tunnel_type = ndpi_tzsp_tunnel;

  if((version == 1) && (ts_type == 0) && (encapsulates == 1)) {
    u_int8_t stop = 0;

    offset += 4;

    while((!stop) && (offset < header->caplen)) {
      u_int8_t tag_type = packet[offset];
      u_int8_t tag_len;

      switch(tag_type) {
      case 0: /* PADDING Tag */
        tag_len = 1;
        break;
      case 1: /* END Tag */
        tag_len = 1, stop = 1;
        break;
      default:
        if(offset + 1 >= header->caplen)
          return(nproto); /* Invalid packet */
        tag_len = packet[offset+1];
        break;
      }

      offset += tag_len;

      if(offset >= header->caplen)
        return(nproto); /* Invalid packet */
      else {
        eth_offset = offset;
        goto datalink_check;
      }
    }
  }
      } else if((sport == NDPI_CAPWAP_DATA_PORT) || (dport == NDPI_CAPWAP_DATA_PORT)) {
  /* We dissect ONLY CAPWAP traffic */
  u_int offset           = ip_offset+ip_len+sizeof(struct ndpi_udphdr);

  if((offset+1) < header->caplen) {
    uint8_t preamble = packet[offset];

    if((preamble & 0x0F) == 0) { /* CAPWAP header */
      u_int16_t msg_len = (packet[offset+1] & 0xF8) >> 1;

      offset += msg_len;

      if((offset + 32 < header->caplen) &&
         (packet[offset + 1] == 0x08)) {
        /* IEEE 802.11 Data */
        offset += 24;
        /* LLC header is 8 bytes */
        type = ntohs((u_int16_t)*((u_int16_t*)&packet[offset+6]));

        ip_offset = offset + 8;

        tunnel_type = ndpi_capwap_tunnel;
        goto iph_check;
      }
    }
  }
      }else if(ndpi_is_valid_vxlan(header, packet, ip_offset, ip_len)){
        tunnel_type = ndpi_vxlan_tunnel;
        eth_offset = ndpi_skip_vxlan(ip_offset, ip_len);
        goto datalink_check;
      }
    }
  } else if(workflow->prefs.decode_tunnels && (proto == IPPROTO_GRE)) {
    if(header->caplen < ip_offset + ip_len + sizeof(struct ndpi_gre_basehdr))
      return(nproto); /* Too short for GRE header*/
    u_int32_t offset = 0;
    if((offset = ndpi_is_valid_gre_tunnel(header, packet, ip_offset, ip_len))) {
      tunnel_type = ndpi_gre_tunnel;
      struct ndpi_gre_basehdr *grehdr = (struct ndpi_gre_basehdr*)&packet[ip_offset + ip_len];
      if(grehdr->protocol == ntohs(ETH_P_IP) || grehdr->protocol == ntohs(ETH_P_IPV6)) {
        ip_offset = offset;
        goto iph_check;
      } else if(grehdr->protocol ==  NDPI_GRE_PROTO_PPP) {  // ppp protocol
        ip_offset = offset + NDPI_PPP_HDRLEN;
        goto iph_check;
      } else {
        eth_offset = offset;
        goto datalink_check;
      }
    } else {
      return(nproto);
    }
  }

  /* process the packet */
  return(packet_processing(workflow, time_ms, vlan_id, tunnel_type, iph, iph6,
         header->caplen - ip_offset,
         header->caplen, header, packet, header->ts,
         flow_risk, flow));
}

/* *********************************************** */

#ifdef USE_DPDK

#include <rte_version.h>
#include <rte_ether.h>

static const struct rte_eth_conf port_conf_default = {
#if(RTE_VERSION < RTE_VERSION_NUM(19, 8, 0, 0))
                  .rxmode = { .max_rx_pkt_len = ETHER_MAX_LEN }
#else
                  .rxmode = { .max_rx_pkt_len = RTE_ETHER_MAX_LEN }
#endif
};

/* ************************************ */

int dpdk_port_init(int port, struct rte_mempool *mbuf_pool) {
  struct rte_eth_conf port_conf = port_conf_default;
  const u_int16_t rx_rings = 1, tx_rings = 1;
  int retval;
  u_int16_t q;

  /* 1 RX queue */
  retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);

  if(retval != 0)
    return retval;

  for(q = 0; q < rx_rings; q++) {
    retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE, rte_eth_dev_socket_id(port), NULL, mbuf_pool);
    if(retval < 0)
      return retval;
  }

  for(q = 0; q < tx_rings; q++) {
    retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE, rte_eth_dev_socket_id(port), NULL);
    if(retval < 0)
      return retval;
  }

  retval = rte_eth_dev_start(port);

  if(retval < 0)
    return retval;

  rte_eth_promiscuous_enable(port);

  return 0;
}

int dpdk_port_deinit(int port) {
  rte_eth_dev_stop(port);
  rte_eth_dev_close(port);
  return 0;
}

#endif

Coverage Report

Created: 2025-11-20 06:48