gtp_itf.cpp

#include <map>
using namespace std;

#ifdef __cplusplus
extern "C" {
#endif
#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <sys/types.h>
#include <netdb.h>

#include <openair2/COMMON/platform_types.h>
#include <openair3/UTILS/conversions.h>
#include "common/utils/LOG/log.h"
#include <common/utils/ocp_itti/intertask_interface.h>
#include <openair2/COMMON/gtpv1_u_messages_types.h>
#include <openair3/ocp-gtpu/gtp_itf.h>
#include <openair2/LAYER2/PDCP_v10.1.0/pdcp.h>
#include <openair2/LAYER2/nr_pdcp/nr_pdcp_oai_api.h>
#include <openair2/LAYER2/nr_rlc/nr_rlc_oai_api.h>
#include "openair2/SDAP/nr_sdap/nr_sdap.h"
#include "sim.h"

#pragma pack(1)

typedef struct Gtpv1uMsgHeader {
  uint8_t PN:1;
  uint8_t S:1;
  uint8_t E:1;
  uint8_t spare:1;
  uint8_t PT:1;
  uint8_t version:3;
  uint8_t msgType;
  uint16_t msgLength;
  teid_t teid;
} __attribute__((packed)) Gtpv1uMsgHeaderT;

//TS 38.425, Figure 5.5.2.2-1
typedef struct DlDataDeliveryStatus_flags {
  uint8_t LPR:1;                    //Lost packet report
  uint8_t FFI:1;                    //Final Frame Ind
  uint8_t deliveredPdcpSn:1;        //Highest Delivered NR PDCP SN Ind
  uint8_t transmittedPdcpSn:1;      //Highest Transmitted NR PDCP SN Ind
  uint8_t pduType:4;                //PDU type
  uint8_t CR:1;                     //Cause Report
  uint8_t deliveredReTxPdcpSn:1;    //Delivered retransmitted NR PDCP SN Ind
  uint8_t reTxPdcpSn:1;             //Retransmitted NR PDCP SN Ind
  uint8_t DRI:1;                    //Data Rate Indication
  uint8_t deliveredPdcpSnRange:1;   //Delivered NR PDCP SN Range Ind
  uint8_t spare:3;
  uint32_t drbBufferSize;            //Desired buffer size for the data radio bearer
} __attribute__((packed)) DlDataDeliveryStatus_flagsT;

typedef struct Gtpv1uMsgHeaderOptFields {
  uint8_t seqNum1Oct;
  uint8_t seqNum2Oct;
  uint8_t NPDUNum;
  uint8_t NextExtHeaderType;    
} __attribute__((packed)) Gtpv1uMsgHeaderOptFieldsT;

#define DL_PDU_SESSION_INFORMATION 0
#define UL_PDU_SESSION_INFORMATION 1

  typedef struct PDUSessionContainer {
  uint8_t spare:4;
  uint8_t PDU_type:4;
  uint8_t QFI:6;
  uint8_t Reflective_QoS_activation:1;
  uint8_t Paging_Policy_Indicator:1;
} __attribute__((packed)) PDUSessionContainerT;

typedef struct Gtpv1uExtHeader {
  uint8_t ExtHeaderLen;
  PDUSessionContainerT pdusession_cntr;
  uint8_t NextExtHeaderType;
}__attribute__((packed)) Gtpv1uExtHeaderT;

#pragma pack()

// TS 29.281, fig 5.2.1-3
#define PDU_SESSION_CONTAINER       (0x85)
#define NR_RAN_CONTAINER            (0x84)

// TS 29.281, 5.2.1
#define EXT_HDR_LNTH_OCTET_UNITS    (4)
#define NO_MORE_EXT_HDRS            (0)

// TS 29.060, table 7.1 defines the possible message types
// here are all the possible messages (3GPP R16)
#define GTP_ECHO_REQ                                         (1)
#define GTP_ECHO_RSP                                         (2)
#define GTP_ERROR_INDICATION                                 (26)
#define GTP_SUPPORTED_EXTENSION_HEADER_INDICATION            (31)
#define GTP_END_MARKER                                       (254)
#define GTP_GPDU                                             (255)

typedef struct gtpv1u_bearer_s {
  /* TEID used in dl and ul */
  teid_t          teid_incoming;                ///< eNB TEID
  teid_t          teid_outgoing;                ///< Remote TEID
  in_addr_t       outgoing_ip_addr;
  struct in6_addr outgoing_ip6_addr;
  tcp_udp_port_t  outgoing_port;
  uint16_t        seqNum;
  uint8_t         npduNum;
  int outgoing_qfi;
} gtpv1u_bearer_t;

typedef struct {
  map<ue_id_t, gtpv1u_bearer_t> bearers;
  teid_t outgoing_teid;
} teidData_t;

typedef struct {
  ue_id_t ue_id;
  ebi_t incoming_rb_id;
  gtpCallback callBack;
  teid_t outgoing_teid;
  gtpCallbackSDAP callBackSDAP;
  int pdusession_id;
} ueidData_t;

class gtpEndPoint {
 public:
  openAddr_t addr;
  uint8_t foundAddr[20];
  int foundAddrLen;
  int ipVersion;
  map<uint64_t, teidData_t> ue2te_mapping;
  // we use the same port number for source and destination address
  // this allow using non standard gtp port number (different from 2152)
  // and so, for example tu run 4G and 5G cores on one system
  tcp_udp_port_t get_dstport() {
    return (tcp_udp_port_t)atol(addr.destinationService);
  }
};

class gtpEndPoints {
 public:
  pthread_mutex_t gtp_lock=PTHREAD_MUTEX_INITIALIZER;
  // the instance id will be the Linux socket handler, as this is uniq
  map<uint64_t, gtpEndPoint> instances;
  map<uint64_t, ueidData_t> te2ue_mapping;
  gtpEndPoints() {
    unsigned int seed;
    fill_random(&seed, sizeof(seed));
    srandom(seed);
  }

  ~gtpEndPoints() {
    // automatically close all sockets on quit
    for (const auto &p : instances)
      close(p.first);
  }
};

gtpEndPoints globGtp;

// note TEid 0 is reserved for specific usage: echo req/resp, error and supported extensions
static  teid_t gtpv1uNewTeid(void) {
#ifdef GTPV1U_LINEAR_TEID_ALLOCATION
  g_gtpv1u_teid = g_gtpv1u_teid + 1;
  return g_gtpv1u_teid;
#else
  return random() + random() % (RAND_MAX - 1) + 1;
#endif
}

instance_t legacyInstanceMapping=0;
#define compatInst(a) ((a)==0 || (a)==INSTANCE_DEFAULT ? legacyInstanceMapping:a)

#define getInstRetVoid(insT)                                    \
  auto instChk=globGtp.instances.find(compatInst(insT));    \
  if (instChk == globGtp.instances.end()) {                        \
    LOG_E(GTPU,"try to get a gtp-u not existing output\n");     \
    pthread_mutex_unlock(&globGtp.gtp_lock);                    \
    return;                                                     \
  }                                                             \
  gtpEndPoint * inst=&instChk->second;
  
#define getInstRetInt(insT)                                    \
  auto instChk=globGtp.instances.find(compatInst(insT));    \
  if (instChk == globGtp.instances.end()) {                        \
    LOG_E(GTPU,"try to get a gtp-u not existing output\n");     \
    pthread_mutex_unlock(&globGtp.gtp_lock);                    \
    return GTPNOK;                                                     \
  }                                                             \
  gtpEndPoint * inst=&instChk->second;


#define getUeRetVoid(insT, Ue)                                            \
    auto ptrUe=insT->ue2te_mapping.find(Ue);                        \
                                                                        \
  if (  ptrUe==insT->ue2te_mapping.end() ) {                          \
    LOG_E(GTPU, "[%ld] gtpv1uSend failed: while getting ue id %ld in hashtable ue_mapping\n", instance, ue_id); \
    pthread_mutex_unlock(&globGtp.gtp_lock);                            \
    return;                                                             \
  }
  
#define getUeRetInt(insT, Ue)                                            \
    auto ptrUe=insT->ue2te_mapping.find(Ue);                        \
                                                                        \
  if (  ptrUe==insT->ue2te_mapping.end() ) {                          \
    LOG_E(GTPU, "[%ld] gtpv1uSend failed: while getting ue id %ld in hashtable ue_mapping\n", instance, ue_id); \
    pthread_mutex_unlock(&globGtp.gtp_lock);                            \
    return GTPNOK;                                                             \
  }
  
#define HDR_MAX 256 // 256 is supposed to be larger than any gtp header
static int gtpv1uCreateAndSendMsg(int h,
                                  uint32_t peerIp,
                                  uint16_t peerPort,
                                  int msgType,
                                  teid_t teid,
                                  uint8_t *Msg,
                                  int msgLen,
                                  bool seqNumFlag,
                                  bool npduNumFlag,
                                  int seqNum,
                                  int npduNum,
                                  int extHdrType,
                                  uint8_t *extensionHeader_buffer,
                                  uint8_t extensionHeader_length) {
  LOG_D(GTPU, "Peer IP:%u peer port:%u outgoing teid:%u \n", peerIp, peerPort, teid);

  uint8_t buffer[msgLen+HDR_MAX]; 
  uint8_t *curPtr=buffer;
  Gtpv1uMsgHeaderT      *msgHdr = (Gtpv1uMsgHeaderT *)buffer ;
  // N should be 0 for us (it was used only in 2G and 3G)
  msgHdr->PN=npduNumFlag;
  msgHdr->S=seqNumFlag;
  msgHdr->E = extHdrType != NO_MORE_EXT_HDRS;
  msgHdr->spare=0;
  //PT=0 is for GTP' TS 32.295 (charging)
  msgHdr->PT=1;
  msgHdr->version=1;
  msgHdr->msgType=msgType;
  msgHdr->teid=htonl(teid);

  curPtr+=sizeof(Gtpv1uMsgHeaderT);

  if (seqNumFlag || (extHdrType != NO_MORE_EXT_HDRS) || npduNumFlag) {
    *(uint16_t *)curPtr = seqNumFlag ? seqNum : 0x0000;
    curPtr+=sizeof(uint16_t);
    *(uint8_t *)curPtr = npduNumFlag ? npduNum : 0x00;
    curPtr++;
    *(uint8_t *)curPtr = extHdrType;
    curPtr++;
  }

  // Bug: if there is more than one extension, infinite loop on extensionHeader_buffer
  while (extHdrType != NO_MORE_EXT_HDRS) {
    if (extensionHeader_length > 0) {
      memcpy(curPtr, extensionHeader_buffer, extensionHeader_length);
      curPtr += extensionHeader_length;
      LOG_D(GTPU, "Extension Header for DDD added. The length is: %d, extension header type is: %x \n", extensionHeader_length, *((uint8_t *)(buffer + 11)));
      extHdrType = extensionHeader_buffer[extensionHeader_length - 1];
      LOG_D(GTPU, "Next extension header type is: %x \n", *((uint8_t *)(buffer + 11)));
    } else {
      LOG_W(GTPU, "Extension header type not supported, returning... \n");
    }
  }

  if (Msg!= NULL){
    memcpy(curPtr, Msg, msgLen);
    curPtr+=msgLen;
  }

  msgHdr->msgLength = htons(curPtr-(buffer+sizeof(Gtpv1uMsgHeaderT)));
  AssertFatal(curPtr-(buffer+msgLen) < HDR_MAX, "fixed max size of all headers too short");
  // Fix me: add IPv6 support, using flag ipVersion
  struct sockaddr_in to= {0};
  to.sin_family      = AF_INET;
  to.sin_port        = htons(peerPort);
  to.sin_addr.s_addr = peerIp ;
  LOG_D(GTPU,"sending packet size: %ld to %s\n",curPtr-buffer, inet_ntoa(to.sin_addr) );
  int ret;

  if ((ret=sendto(h, (void *)buffer, curPtr-buffer, 0,(struct sockaddr *)&to, sizeof(to) )) != curPtr-buffer ) {
    LOG_E(GTPU, "[SD %d] Failed to send data to " IPV4_ADDR " on port %d, buffer size %lu, ret: %d, errno: %d\n",
          h, IPV4_ADDR_FORMAT(peerIp), peerPort, curPtr-buffer, ret, errno);
    return GTPNOK;
  }

  return  !GTPNOK;
}

static void gtpv1uSend(instance_t instance, gtpv1u_tunnel_data_req_t *req, bool seqNumFlag, bool npduNumFlag) {
  uint8_t *buffer=req->buffer+req->offset;
  size_t length=req->length;
  ue_id_t ue_id=req->ue_id;
  int  bearer_id=req->bearer_id;
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetVoid(compatInst(instance));
  getUeRetVoid(inst, ue_id);

  auto ptr2=ptrUe->second.bearers.find(bearer_id);

  if ( ptr2 == ptrUe->second.bearers.end() ) {
    LOG_E(GTPU,"[%ld] GTP-U instance: sending a packet to a non existant UE:RAB: %lx/%x\n", instance, ue_id, bearer_id);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return;
  }

  LOG_D(GTPU,"[%ld] sending a packet to UE:RAB:teid %lx/%x/%x, len %lu, oldseq %d, oldnum %d\n",
        instance, ue_id, bearer_id,ptr2->second.teid_outgoing,length, ptr2->second.seqNum,ptr2->second.npduNum );

  if(seqNumFlag)
    ptr2->second.seqNum++;

  if(npduNumFlag)
    ptr2->second.npduNum++;

  // copy to release the mutex
  gtpv1u_bearer_t tmp=ptr2->second;
  pthread_mutex_unlock(&globGtp.gtp_lock);

  if (tmp.outgoing_qfi != -1) {
    Gtpv1uExtHeaderT ext = { 0 };
    ext.ExtHeaderLen = 1; // in quad bytes  EXT_HDR_LNTH_OCTET_UNITS
    ext.pdusession_cntr.spare = 0;
    ext.pdusession_cntr.PDU_type = UL_PDU_SESSION_INFORMATION;
    ext.pdusession_cntr.QFI = tmp.outgoing_qfi;
    ext.pdusession_cntr.Reflective_QoS_activation = false;
    ext.pdusession_cntr.Paging_Policy_Indicator = false;
    ext.NextExtHeaderType = NO_MORE_EXT_HDRS;

    gtpv1uCreateAndSendMsg(compatInst(instance),
                           tmp.outgoing_ip_addr,
                           tmp.outgoing_port,
                           GTP_GPDU,
                           tmp.teid_outgoing,
                           buffer,
                           length,
                           seqNumFlag,
                           npduNumFlag,
                           tmp.seqNum,
                           tmp.npduNum,
                           PDU_SESSION_CONTAINER,
                           (uint8_t *)&ext,
                           sizeof(ext));
  } else {
    gtpv1uCreateAndSendMsg(
        compatInst(instance), tmp.outgoing_ip_addr, tmp.outgoing_port, GTP_GPDU, tmp.teid_outgoing, buffer, length, seqNumFlag, npduNumFlag, tmp.seqNum, tmp.npduNum, NO_MORE_EXT_HDRS, NULL, 0);
  }
}

static void fillDlDeliveryStatusReport(extensionHeader_t *extensionHeader, uint32_t RLC_buffer_availability, uint32_t NR_PDCP_PDU_SN){

  extensionHeader->buffer[0] = (1+sizeof(DlDataDeliveryStatus_flagsT)+(NR_PDCP_PDU_SN>0?3:0)+(NR_PDCP_PDU_SN>0?1:0)+1)/4;
  DlDataDeliveryStatus_flagsT DlDataDeliveryStatus;
  DlDataDeliveryStatus.deliveredPdcpSn = 0;
  DlDataDeliveryStatus.transmittedPdcpSn= NR_PDCP_PDU_SN>0?1:0;
  DlDataDeliveryStatus.pduType = 1;
  DlDataDeliveryStatus.drbBufferSize = htonl(RLC_buffer_availability);
  memcpy(extensionHeader->buffer+1, &DlDataDeliveryStatus, sizeof(DlDataDeliveryStatus_flagsT));
  uint8_t offset = sizeof(DlDataDeliveryStatus_flagsT)+1;

  if(NR_PDCP_PDU_SN>0){
    extensionHeader->buffer[offset] =   (NR_PDCP_PDU_SN >> 16) & 0xff;
    extensionHeader->buffer[offset+1] = (NR_PDCP_PDU_SN >> 8) & 0xff;
    extensionHeader->buffer[offset+2] = NR_PDCP_PDU_SN & 0xff;
    LOG_D(GTPU, "Octets reporting NR_PDCP_PDU_SN, extensionHeader->buffer[offset]: %u, extensionHeader->buffer[offset+1]:%u, extensionHeader->buffer[offset+2]:%u \n", extensionHeader->buffer[offset], extensionHeader->buffer[offset+1],extensionHeader->buffer[offset+2]);
    extensionHeader->buffer[offset+3] = 0x00; //Padding octet
    offset = offset+3;
  }
  extensionHeader->buffer[offset] = 0x00; //No more extension headers
  /*Total size of DDD_status PDU = size of mandatory part +
   * 3 octets for highest transmitted/delivered PDCP SN +
   * 1 octet for padding + 1 octet for next extension header type,
   * according to TS 38.425: Fig. 5.5.2.2-1 and section 5.5.3.24*/
  extensionHeader->length  = 1+sizeof(DlDataDeliveryStatus_flagsT)+
                              (NR_PDCP_PDU_SN>0?3:0)+
                              (NR_PDCP_PDU_SN>0?1:0)+1;
}

static void gtpv1uSendDlDeliveryStatus(instance_t instance, gtpv1u_DU_buffer_report_req_t *req){
  ue_id_t ue_id=req->ue_id;
  int  bearer_id=req->pdusession_id;
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetVoid(compatInst(instance));
  getUeRetVoid(inst, ue_id);

  auto ptr2=ptrUe->second.bearers.find(bearer_id);

  if ( ptr2 == ptrUe->second.bearers.end() ) {
    LOG_D(GTPU,"GTP-U instance: %ld sending a packet to a non existant UE ID:RAB: %lu/%x\n", instance, ue_id, bearer_id);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return;
  }

  extensionHeader_t *extensionHeader;
  extensionHeader = (extensionHeader_t *) calloc(1, sizeof(extensionHeader_t));
  fillDlDeliveryStatusReport(extensionHeader, req->buffer_availability,0);

  LOG_I(GTPU,"[%ld] GTP-U sending DL Data Delivery status to UE ID:RAB:teid %lu/%x/%x, oldseq %d, oldnum %d\n",
        instance, ue_id, bearer_id,ptr2->second.teid_outgoing, ptr2->second.seqNum,ptr2->second.npduNum );
  // copy to release the mutex
  gtpv1u_bearer_t tmp=ptr2->second;
  pthread_mutex_unlock(&globGtp.gtp_lock);
  gtpv1uCreateAndSendMsg(
      compatInst(instance), tmp.outgoing_ip_addr, tmp.outgoing_port, GTP_GPDU, tmp.teid_outgoing, NULL, 0, false, false, 0, 0, NR_RAN_CONTAINER, extensionHeader->buffer, extensionHeader->length);
}

static void gtpv1uEndTunnel(instance_t instance, gtpv1u_tunnel_data_req_t *req) {
  ue_id_t ue_id=req->ue_id;
  int  bearer_id=req->bearer_id;
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetVoid(compatInst(instance));
  getUeRetVoid(inst, ue_id);

  auto ptr2=ptrUe->second.bearers.find(bearer_id);

  if ( ptr2 == ptrUe->second.bearers.end() ) {
    LOG_E(GTPU,"[%ld] GTP-U sending a packet to a non existant UE:RAB: %lx/%x\n", instance, ue_id, bearer_id);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return;
  }

  LOG_D(GTPU,"[%ld] sending a end packet packet to UE:RAB:teid %lx/%x/%x\n",
        instance, ue_id, bearer_id,ptr2->second.teid_outgoing);
  gtpv1u_bearer_t tmp=ptr2->second;
  pthread_mutex_unlock(&globGtp.gtp_lock);
  Gtpv1uMsgHeaderT  msgHdr;
  // N should be 0 for us (it was used only in 2G and 3G)
  msgHdr.PN=0;
  msgHdr.S=0;
  msgHdr.E=0;
  msgHdr.spare=0;
  //PT=0 is for GTP' TS 32.295 (charging)
  msgHdr.PT=1;
  msgHdr.version=1;
  msgHdr.msgType=GTP_END_MARKER;
  msgHdr.msgLength=htons(0);
  msgHdr.teid=htonl(tmp.teid_outgoing);
  // Fix me: add IPv6 support, using flag ipVersion
  static struct sockaddr_in to= {0};
  to.sin_family      = AF_INET;
  to.sin_port        = htons(tmp.outgoing_port);
  to.sin_addr.s_addr = tmp.outgoing_ip_addr;
  char ip4[INET_ADDRSTRLEN];
  //char ip6[INET6_ADDRSTRLEN];
  LOG_D(GTPU,"[%ld] sending end packet to %s\n", instance, inet_ntoa(to.sin_addr) );

  if (sendto(compatInst(instance), (void *)&msgHdr, sizeof(msgHdr), 0,(struct sockaddr *)&to, sizeof(to) ) !=  sizeof(msgHdr)) {
    LOG_E(GTPU,
          "[%ld] Failed to send data to %s on port %d, buffer size %lu\n",
          compatInst(instance), inet_ntop(AF_INET, &tmp.outgoing_ip_addr, ip4, INET_ADDRSTRLEN), tmp.outgoing_port, sizeof(msgHdr));
  }
}

static  int udpServerSocket(openAddr_s addr) {
  LOG_I(GTPU, "Initializing UDP for local address %s with port %s\n", addr.originHost, addr.originService);
  int status;
  struct addrinfo hints= {0}, *servinfo, *p;
  hints.ai_family = AF_UNSPEC;
  hints.ai_socktype = SOCK_DGRAM;
  hints.ai_flags = AI_PASSIVE;

  if ((status = getaddrinfo(addr.originHost, addr.originService, &hints, &servinfo)) != 0) {
    LOG_E(GTPU,"getaddrinfo error: %s\n", gai_strerror(status));
    return -1;
  }

  int sockfd=-1;

  // loop through all the results and bind to the first we can
  for(p = servinfo; p != NULL; p = p->ai_next) {
    if ((sockfd = socket(p->ai_family, p->ai_socktype,
                         p->ai_protocol)) == -1) {
      LOG_W(GTPU,"socket: %s\n", strerror(errno));
      continue;
    }

    if (bind(sockfd, p->ai_addr, p->ai_addrlen) == -1) {
      close(sockfd);
      LOG_W(GTPU,"bind: %s\n", strerror(errno));
      continue;
    } else {
      // We create the gtp instance on the socket
      globGtp.instances[sockfd].addr=addr;

      if (p->ai_family == AF_INET) {
        struct sockaddr_in *ipv4=(struct sockaddr_in *)p->ai_addr;
        memcpy(globGtp.instances[sockfd].foundAddr,
               &ipv4->sin_addr.s_addr, sizeof(ipv4->sin_addr.s_addr));
        globGtp.instances[sockfd].foundAddrLen=sizeof(ipv4->sin_addr.s_addr);
        globGtp.instances[sockfd].ipVersion=4;
        break;
      } else if (p->ai_family == AF_INET6) {
        LOG_W(GTPU,"Local address is IP v6\n");
        struct sockaddr_in6 *ipv6=(struct sockaddr_in6 *)p->ai_addr;
        memcpy(globGtp.instances[sockfd].foundAddr,
               &ipv6->sin6_addr.s6_addr, sizeof(ipv6->sin6_addr.s6_addr));
        globGtp.instances[sockfd].foundAddrLen=sizeof(ipv6->sin6_addr.s6_addr);
        globGtp.instances[sockfd].ipVersion=6;
      } else
        AssertFatal(false,"Local address is not IPv4 or IPv6");
    }

    break; // if we get here, we must have connected successfully
  }

  if (p == NULL) {
    // looped off the end of the list with no successful bind
    LOG_E(GTPU,"failed to bind socket: %s %s \n", addr.originHost, addr.originService);
    return -1;
  }

  freeaddrinfo(servinfo); // all done with this structure

  if (strlen(addr.destinationHost)>1) {
    struct addrinfo hints;
    memset(&hints,0,sizeof(hints));
    hints.ai_family=AF_UNSPEC;
    hints.ai_socktype=SOCK_DGRAM;
    hints.ai_protocol=0;
    hints.ai_flags=AI_PASSIVE|AI_ADDRCONFIG;
    struct addrinfo *res=0;
    int err=getaddrinfo(addr.destinationHost,addr.destinationService,&hints,&res);

    if (err==0) {
      for(p = res; p != NULL; p = p->ai_next) {
        if ((err=connect(sockfd,  p->ai_addr, p->ai_addrlen))==0)
          break;
      }
    }

    if (err)
      LOG_E(GTPU,"Can't filter remote host: %s, %s\n", addr.destinationHost,addr.destinationService);
  }

  int sendbuff = 1000*1000*10;
  AssertFatal(0==setsockopt(sockfd, SOL_SOCKET, SO_SNDBUF, &sendbuff, sizeof(sendbuff)),"");
  LOG_D(GTPU,"[%d] Created listener for paquets to: %s:%s, send buffer size: %d\n", sockfd, addr.originHost, addr.originService,sendbuff);
  return sockfd;
}

instance_t gtpv1Init(openAddr_t context) {
  pthread_mutex_lock(&globGtp.gtp_lock);
  int id=udpServerSocket(context);

  if (id>=0) {
    itti_subscribe_event_fd(TASK_GTPV1_U, id);
  } else
    LOG_E(GTPU,"can't create GTP-U instance\n");

  pthread_mutex_unlock(&globGtp.gtp_lock);
  LOG_I(GTPU, "Created gtpu instance id: %d\n", id);
  return id;
}

void GtpuUpdateTunnelOutgoingAddressAndTeid(instance_t instance, ue_id_t ue_id, ebi_t bearer_id, in_addr_t newOutgoingAddr, teid_t newOutgoingTeid) {
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetVoid(compatInst(instance));
  getUeRetVoid(inst, ue_id);

  auto ptr2=ptrUe->second.bearers.find(bearer_id);

  if ( ptr2 == ptrUe->second.bearers.end() ) {
    LOG_E(GTPU,"[%ld] Update tunnel for a existing ue id %lu, but wrong bearer_id %u\n", instance, ue_id, bearer_id);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return;
  }

  ptr2->second.outgoing_ip_addr = newOutgoingAddr;
  ptr2->second.teid_outgoing = newOutgoingTeid;
  LOG_I(GTPU, "[%ld] Tunnel Outgoing TEID updated to %x and address to %x\n", instance, ptr2->second.teid_outgoing, ptr2->second.outgoing_ip_addr);
  pthread_mutex_unlock(&globGtp.gtp_lock);
  return;
}

teid_t newGtpuCreateTunnel(instance_t instance,
                           ue_id_t ue_id,
                           int incoming_bearer_id,
                           int outgoing_bearer_id,
                           teid_t outgoing_teid,
                           int outgoing_qfi,
                           transport_layer_addr_t remoteAddr,
                           int port,
                           gtpCallback callBack,
                           gtpCallbackSDAP callBackSDAP) {
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));
  auto it=inst->ue2te_mapping.find(ue_id);

  if ( it != inst->ue2te_mapping.end() &&  it->second.bearers.find(outgoing_bearer_id) != it->second.bearers.end()) {
    LOG_W(GTPU,"[%ld] Create a config for a already existing GTP tunnel (ue id %lu)\n", instance, ue_id);
    inst->ue2te_mapping.erase(it);
  }

  teid_t incoming_teid=gtpv1uNewTeid();

  while (globGtp.te2ue_mapping.find(incoming_teid) != globGtp.te2ue_mapping.end()) {
    LOG_W(GTPU, "[%ld] generated a random Teid that exists, re-generating (%x)\n", instance, incoming_teid);
    incoming_teid=gtpv1uNewTeid();
  };

  globGtp.te2ue_mapping[incoming_teid].ue_id = ue_id;
  globGtp.te2ue_mapping[incoming_teid].incoming_rb_id = incoming_bearer_id;
  globGtp.te2ue_mapping[incoming_teid].outgoing_teid = outgoing_teid;
  globGtp.te2ue_mapping[incoming_teid].callBack = callBack;
  globGtp.te2ue_mapping[incoming_teid].callBackSDAP = callBackSDAP;
  globGtp.te2ue_mapping[incoming_teid].pdusession_id = (uint8_t)outgoing_bearer_id;

  gtpv1u_bearer_t *tmp=&inst->ue2te_mapping[ue_id].bearers[outgoing_bearer_id];

  int addrs_length_in_bytes = remoteAddr.length / 8;

  switch (addrs_length_in_bytes) {
    case 4:
      memcpy(&tmp->outgoing_ip_addr,remoteAddr.buffer,4);
      break;

    case 16:
      memcpy(tmp->outgoing_ip6_addr.s6_addr,remoteAddr.buffer,
             16);
      break;

    case 20:
      memcpy(&tmp->outgoing_ip_addr,remoteAddr.buffer,4);
      memcpy(tmp->outgoing_ip6_addr.s6_addr,
             remoteAddr.buffer+4,
             16);

    default:
      AssertFatal(false, "SGW Address size impossible");
  }

  tmp->teid_incoming = incoming_teid;
  tmp->outgoing_port=port;
  tmp->teid_outgoing= outgoing_teid;
  tmp->outgoing_qfi=outgoing_qfi;
  pthread_mutex_unlock(&globGtp.gtp_lock);
  char ip4[INET_ADDRSTRLEN];
  char ip6[INET6_ADDRSTRLEN];
  LOG_I(GTPU,
        "[%ld] Created tunnel for UE ID %lu, teid for incoming: %x, teid for outgoing %x to remote IPv4: %s, IPv6 %s\n",
        instance,
        ue_id,
        tmp->teid_incoming,
        tmp->teid_outgoing,
        inet_ntop(AF_INET, (void *)&tmp->outgoing_ip_addr, ip4, INET_ADDRSTRLEN),
        inet_ntop(AF_INET6, (void *)&tmp->outgoing_ip6_addr.s6_addr, ip6, INET6_ADDRSTRLEN));
  return incoming_teid;
}

int gtpv1u_create_s1u_tunnel(instance_t instance,
                             const gtpv1u_enb_create_tunnel_req_t  *create_tunnel_req,
                             gtpv1u_enb_create_tunnel_resp_t *create_tunnel_resp,
                             gtpCallback callBack)
{
  LOG_D(GTPU, "[%ld] Start create tunnels for UE ID %u, num_tunnels %d, sgw_S1u_teid %x\n",
        instance,
        create_tunnel_req->rnti,
        create_tunnel_req->num_tunnels,
        create_tunnel_req->sgw_S1u_teid[0]);
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));
  
  tcp_udp_port_t dstport=inst->get_dstport();
  uint8_t addr[inst->foundAddrLen];
  memcpy(addr, inst->foundAddr, inst->foundAddrLen);
  pthread_mutex_unlock(&globGtp.gtp_lock);
  
  for (int i = 0; i < create_tunnel_req->num_tunnels; i++) {
    AssertFatal(create_tunnel_req->eps_bearer_id[i] > 4,
                "From legacy code not clear, seems impossible (bearer=%d)\n",
                create_tunnel_req->eps_bearer_id[i]);
    int incoming_rb_id=create_tunnel_req->eps_bearer_id[i]-4;
    teid_t teid = newGtpuCreateTunnel(compatInst(instance),
                                      create_tunnel_req->rnti,
                                      incoming_rb_id,
                                      create_tunnel_req->eps_bearer_id[i],
                                      create_tunnel_req->sgw_S1u_teid[i],
                                      -1, // no pdu session in 4G
                                      create_tunnel_req->sgw_addr[i],
                                      dstport,
                                      callBack,
                                      NULL);
    create_tunnel_resp->status=0;
    create_tunnel_resp->rnti=create_tunnel_req->rnti;
    create_tunnel_resp->num_tunnels=create_tunnel_req->num_tunnels;
    create_tunnel_resp->enb_S1u_teid[i]=teid;
    create_tunnel_resp->eps_bearer_id[i] = create_tunnel_req->eps_bearer_id[i];
    memcpy(create_tunnel_resp->enb_addr.buffer,addr,sizeof(addr));
    create_tunnel_resp->enb_addr.length= sizeof(addr);
  }

  return !GTPNOK;
}

int gtpv1u_update_s1u_tunnel(
  const instance_t                              instance,
  const gtpv1u_enb_create_tunnel_req_t *const   create_tunnel_req,
  const rnti_t                                  prior_rnti
) {
  LOG_D(GTPU, "[%ld] Start update tunnels for old RNTI %x, new RNTI %x, num_tunnels %d, sgw_S1u_teid %x, eps_bearer_id %x\n",
        instance,
        prior_rnti,
        create_tunnel_req->rnti,
        create_tunnel_req->num_tunnels,
        create_tunnel_req->sgw_S1u_teid[0],
        create_tunnel_req->eps_bearer_id[0]);
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));

  if ( inst->ue2te_mapping.find(create_tunnel_req->rnti) == inst->ue2te_mapping.end() ) {
    LOG_E(GTPU,"[%ld] Update not already existing tunnel (new rnti %x, old rnti %x)\n", 
          instance, create_tunnel_req->rnti, prior_rnti);
  }

  auto it=inst->ue2te_mapping.find(prior_rnti);

  if ( it != inst->ue2te_mapping.end() ) {
    pthread_mutex_unlock(&globGtp.gtp_lock);
    AssertFatal(false, "logic bug: update of non-existing tunnel (new ue id %u, old ue id %u)\n", create_tunnel_req->rnti, prior_rnti);
    /* we don't know if we need 4G or 5G PDCP and can therefore not create a
     * new tunnel */
    return 0;
  }

  inst->ue2te_mapping[create_tunnel_req->rnti]=it->second;
  inst->ue2te_mapping.erase(it);
  pthread_mutex_unlock(&globGtp.gtp_lock);
  return 0;
}

int gtpv1u_create_ngu_tunnel(const instance_t instance,
                             const gtpv1u_gnb_create_tunnel_req_t *const create_tunnel_req,
                             gtpv1u_gnb_create_tunnel_resp_t *const create_tunnel_resp,
                             gtpCallback callBack,
                             gtpCallbackSDAP callBackSDAP)
{
  LOG_D(GTPU, "[%ld] Start create tunnels for ue id %lu, num_tunnels %d, sgw_S1u_teid %x\n",
        instance,
        create_tunnel_req->ue_id,
        create_tunnel_req->num_tunnels,
        create_tunnel_req->outgoing_teid[0]);
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));

  tcp_udp_port_t dstport = inst->get_dstport();
  uint8_t addr[inst->foundAddrLen];
  memcpy(addr, inst->foundAddr, inst->foundAddrLen);
  pthread_mutex_unlock(&globGtp.gtp_lock);
  for (int i = 0; i < create_tunnel_req->num_tunnels; i++) {
    teid_t teid = newGtpuCreateTunnel(instance,
                                      create_tunnel_req->ue_id,
                                      create_tunnel_req->incoming_rb_id[i],
                                      create_tunnel_req->pdusession_id[i],
                                      create_tunnel_req->outgoing_teid[i],
                                      create_tunnel_req->outgoing_qfi[i],
                                      create_tunnel_req->dst_addr[i],
                                      dstport,
                                      callBack,
                                      callBackSDAP);
    create_tunnel_resp->status=0;
    create_tunnel_resp->ue_id=create_tunnel_req->ue_id;
    create_tunnel_resp->num_tunnels=create_tunnel_req->num_tunnels;
    create_tunnel_resp->gnb_NGu_teid[i]=teid;
    memcpy(create_tunnel_resp->gnb_addr.buffer,addr,sizeof(addr));
    create_tunnel_resp->gnb_addr.length= sizeof(addr);
    create_tunnel_resp->pdusession_id[i] = create_tunnel_req->pdusession_id[i];
  }

  return !GTPNOK;
}

int gtpv1u_update_ue_id(const instance_t instanceP, ue_id_t old_ue_id, ue_id_t new_ue_id)
{
  pthread_mutex_lock(&globGtp.gtp_lock);

  auto inst = &globGtp.instances[compatInst(instanceP)];
  auto it = inst->ue2te_mapping.find(old_ue_id);
  if (it == inst->ue2te_mapping.end()) {
    LOG_W(GTPU, "[%ld] Update GTP tunnels for UEid: %lx, but no tunnel exits\n", instanceP, old_ue_id);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return GTPNOK;
  }

  for (unsigned i = 0; i < it->second.bearers.size(); ++i) {
    teid_t incoming_teid = inst->ue2te_mapping[old_ue_id].bearers[i].teid_incoming;
    if (globGtp.te2ue_mapping[incoming_teid].ue_id == old_ue_id) {
      globGtp.te2ue_mapping[incoming_teid].ue_id = new_ue_id;
    }
  }

  inst->ue2te_mapping[new_ue_id] = it->second;
  inst->ue2te_mapping.erase(it);

  pthread_mutex_unlock(&globGtp.gtp_lock);

  LOG_I(GTPU, "[%ld] Updated tunnels from UEid %lx to UEid %lx\n", instanceP, old_ue_id, new_ue_id);
  return !GTPNOK;
}

int gtpv1u_create_x2u_tunnel(
  const instance_t instanceP,
  const gtpv1u_enb_create_x2u_tunnel_req_t   *const create_tunnel_req_pP,
  gtpv1u_enb_create_x2u_tunnel_resp_t *const create_tunnel_resp_pP) {
  AssertFatal( false, "to be developped\n");
}

int newGtpuDeleteAllTunnels(instance_t instance, ue_id_t ue_id) {
  LOG_D(GTPU, "[%ld] Start delete tunnels for ue id %lu\n",
        instance, ue_id);
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));
  getUeRetInt(inst, ue_id);

  int nb=0;

  for (auto j=ptrUe->second.bearers.begin();
       j!=ptrUe->second.bearers.end();
       ++j) {
    globGtp.te2ue_mapping.erase(j->second.teid_incoming);
    nb++;
  }

  inst->ue2te_mapping.erase(ptrUe);
  pthread_mutex_unlock(&globGtp.gtp_lock);
  LOG_I(GTPU, "[%ld] Deleted all tunnels for ue id %ld (%d tunnels deleted)\n", instance, ue_id, nb);
  return !GTPNOK;
}

int gtpv1u_delete_s1u_tunnel( const instance_t instance,
                              const gtpv1u_enb_delete_tunnel_req_t *const req_pP) {
  LOG_D(GTPU, "[%ld] Start delete tunnels for RNTI %x\n", instance, req_pP->rnti);
  pthread_mutex_lock(&globGtp.gtp_lock);
  auto inst = &globGtp.instances[compatInst(instance)];
  auto ptrRNTI = inst->ue2te_mapping.find(req_pP->rnti);
  if (ptrRNTI == inst->ue2te_mapping.end()) {
    LOG_W(GTPU, "[%ld] Delete Released GTP tunnels for rnti: %x, but no tunnel exits\n", instance, req_pP->rnti);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return -1;
  }

  int nb = 0;

  for (int i = 0; i < req_pP->num_erab; i++) {
    auto ptr2 = ptrRNTI->second.bearers.find(req_pP->eps_bearer_id[i]);
    if (ptr2 == ptrRNTI->second.bearers.end()) {
      LOG_E(GTPU,
            "[%ld] GTP-U instance: delete of not existing tunnel RNTI:RAB: %x/%x\n",
            instance,
            req_pP->rnti,
            req_pP->eps_bearer_id[i]);
    } else {
      globGtp.te2ue_mapping.erase(ptr2->second.teid_incoming);
      nb++;
    }
  }

  if (ptrRNTI->second.bearers.size() == 0)
    // no tunnels on this rnti, erase the ue entry
    inst->ue2te_mapping.erase(ptrRNTI);

  pthread_mutex_unlock(&globGtp.gtp_lock);
  LOG_I(GTPU, "[%ld] Deleted released tunnels for RNTI %x (%d tunnels deleted)\n", instance, req_pP->rnti, nb);
  return !GTPNOK;
}

// Legacy delete tunnel finish by deleting all the ue id
int gtpv1u_delete_all_s1u_tunnel(const instance_t instance, const rnti_t rnti)
{
  return newGtpuDeleteAllTunnels(instance, rnti);
}

int newGtpuDeleteTunnels(instance_t instance, ue_id_t ue_id, int nbTunnels, pdusessionid_t *pdusession_id) {
  LOG_D(GTPU, "[%ld] Start delete tunnels for ue id %lu\n",
        instance, ue_id);
  pthread_mutex_lock(&globGtp.gtp_lock);
  getInstRetInt(compatInst(instance));
  getUeRetInt(inst, ue_id);
  int nb=0;

  for (int i=0; i<nbTunnels; i++) {
    auto ptr2=ptrUe->second.bearers.find(pdusession_id[i]);

    if ( ptr2 == ptrUe->second.bearers.end() ) {
      LOG_E(GTPU,"[%ld] GTP-U instance: delete of not existing tunnel UE ID:RAB: %ld/%x\n", instance, ue_id, pdusession_id[i]);
    } else {
      globGtp.te2ue_mapping.erase(ptr2->second.teid_incoming);
      nb++;
    }
  }

  if (ptrUe->second.bearers.size() == 0 )
    // no tunnels on this ue id, erase the ue entry
    inst->ue2te_mapping.erase(ptrUe);

  pthread_mutex_unlock(&globGtp.gtp_lock);
  LOG_I(GTPU, "[%ld] Deleted all tunnels for ue id %lu (%d tunnels deleted)\n", instance, ue_id, nb);
  return !GTPNOK;
}

int gtpv1u_delete_x2u_tunnel( const instance_t instanceP,
                              const gtpv1u_enb_delete_tunnel_req_t *const req_pP) {
  LOG_E(GTPU,"x2 tunnel not implemented\n");
  return 0;
}

int gtpv1u_delete_ngu_tunnel( const instance_t instance,
                              gtpv1u_gnb_delete_tunnel_req_t *req) {
  return  newGtpuDeleteTunnels(instance, req->ue_id, req->num_pdusession, req->pdusession_id);
}

static int Gtpv1uHandleEchoReq(int h,
                               uint8_t *msgBuf,
                               uint32_t msgBufLen,
                               uint16_t peerPort,
                               uint32_t peerIp) {
  Gtpv1uMsgHeaderT      *msgHdr = (Gtpv1uMsgHeaderT *) msgBuf;

  if ( msgHdr->version != 1 ||  msgHdr->PT != 1 ) {
    LOG_E(GTPU, "[%d] Received a packet that is not GTP header\n", h);
    return GTPNOK;
  }

  if ( msgHdr->S != 1 ) {
    LOG_E(GTPU, "[%d] Received a echo request packet with no sequence number \n", h);
    return GTPNOK;
  }

  uint16_t seq=ntohs(*(uint16_t *)(msgHdr+1));
  LOG_D(GTPU, "[%d] Received a echo request, TEID: %d, seq: %hu\n", h, msgHdr->teid, seq);
  uint8_t recovery[2]= {14,0};
  return gtpv1uCreateAndSendMsg(h, peerIp, peerPort, GTP_ECHO_RSP, ntohl(msgHdr->teid), recovery, sizeof recovery, true, false, seq, 0, NO_MORE_EXT_HDRS, NULL, 0);
}

static int Gtpv1uHandleError(int h,
                             uint8_t *msgBuf,
                             uint32_t msgBufLen,
                             uint16_t peerPort,
                             uint32_t peerIp) {
  LOG_E(GTPU,"Hadle error to be dev\n");
  int rc = GTPNOK;
  return rc;
}

static int Gtpv1uHandleSupportedExt(int h,
                                    uint8_t *msgBuf,
                                    uint32_t msgBufLen,
                                    uint16_t peerPort,
                                    uint32_t peerIp) {
  LOG_E(GTPU,"Supported extensions to be dev\n");
  int rc = GTPNOK;
  return rc;
}

// When end marker arrives, we notify the client with buffer size = 0
// The client will likely call "delete tunnel"
// nevertheless we don't take the initiative
static int Gtpv1uHandleEndMarker(int h,
                                 uint8_t *msgBuf,
                                 uint32_t msgBufLen,
                                 uint16_t peerPort,
                                 uint32_t peerIp) {
  Gtpv1uMsgHeaderT      *msgHdr = (Gtpv1uMsgHeaderT *) msgBuf;

  if ( msgHdr->version != 1 ||  msgHdr->PT != 1 ) {
    LOG_E(GTPU, "[%d] Received a packet that is not GTP header\n", h);
    return GTPNOK;
  }

  pthread_mutex_lock(&globGtp.gtp_lock);
  // the socket Linux file handler is the instance id
  getInstRetInt(h);

  auto tunnel = globGtp.te2ue_mapping.find(ntohl(msgHdr->teid));

  if (tunnel == globGtp.te2ue_mapping.end()) {
    LOG_E(GTPU,"[%d] Received a incoming packet on unknown teid (%x) Dropping!\n", h, msgHdr->teid);
    pthread_mutex_unlock(&globGtp.gtp_lock);
    return GTPNOK;
  }

  // This context is not good for gtp
  // frame, ... has no meaning
  // manyother attributes may come from create tunnel
  protocol_ctxt_t ctxt;
  ctxt.module_id = 0;
  ctxt.enb_flag = 1;
  ctxt.instance = inst->addr.originInstance;
  ctxt.rntiMaybeUEid = tunnel->second.ue_id;
  ctxt.frame = 0;
  ctxt.subframe = 0;
  ctxt.eNB_index = 0;
  ctxt.brOption = 0;
  const srb_flag_t     srb_flag=SRB_FLAG_NO;
  const rb_id_t        rb_id=tunnel->second.incoming_rb_id;
  const mui_t          mui=RLC_MUI_UNDEFINED;
  const confirm_t      confirm=RLC_SDU_CONFIRM_NO;
  const pdcp_transmission_mode_t mode=PDCP_TRANSMISSION_MODE_DATA;
  const uint32_t sourceL2Id=0;
  const uint32_t destinationL2Id=0;
  pthread_mutex_unlock(&globGtp.gtp_lock);