[thirdparty/pdns.git] / pdns / pdns_recursor.cc

/*
 * This file is part of PowerDNS or dnsdist.
 * Copyright -- PowerDNS.COM B.V. and its contributors
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * In addition, for the avoidance of any doubt, permission is granted to
 * link this program with OpenSSL and to (re)distribute the binaries
 * produced as the result of such linking.
 *
 * This program 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 General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <netdb.h>
#include <sys/stat.h>
#include <unistd.h>

#include "ws-recursor.hh"
#include <pthread.h>
#include "recpacketcache.hh"
#include "utility.hh"
#include "dns_random.hh"
#ifdef HAVE_LIBSODIUM
#include <sodium.h>
#endif
#include "opensslsigners.hh"
#include <iostream>
#include <errno.h>
#include <boost/static_assert.hpp>
#include <map>
#include <set>
#include "recursor_cache.hh"
#include "cachecleaner.hh"
#include <stdio.h>
#include <signal.h>
#include <stdlib.h>
#include "misc.hh"
#include "mtasker.hh"
#include <utility>
#include "arguments.hh"
#include "syncres.hh"
#include <fcntl.h>
#include <fstream>
#include "sortlist.hh"
#include "sstuff.hh"
#include <boost/tuple/tuple.hpp>
#include <boost/tuple/tuple_comparison.hpp>
#include <boost/shared_array.hpp>
#include <boost/function.hpp>
#include <boost/algorithm/string.hpp>
#ifdef MALLOC_TRACE
#include "malloctrace.hh"
#endif
#include <netinet/tcp.h>
#include "dnsparser.hh"
#include "dnswriter.hh"
#include "dnsrecords.hh"
#include "zoneparser-tng.hh"
#include "rec_channel.hh"
#include "logger.hh"
#include "iputils.hh"
#include "mplexer.hh"
#include "config.h"
#include "lua-recursor4.hh"
#include "version.hh"
#include "responsestats.hh"
#include "secpoll-recursor.hh"
#include "dnsname.hh"
#include "filterpo.hh"
#include "rpzloader.hh"
#include "validate-recursor.hh"
#include "rec-lua-conf.hh"
#include "ednsoptions.hh"
#include "gettime.hh"

#include "rec-protobuf.hh"
#include "rec-snmp.hh"

#ifdef HAVE_SYSTEMD
#include <systemd/sd-daemon.h>
#endif

#include "namespaces.hh"

typedef map<ComboAddress, uint32_t, ComboAddress::addressOnlyLessThan> tcpClientCounts_t;

static thread_local std::shared_ptr<RecursorLua4> t_pdl;
static thread_local unsigned int t_id;
static thread_local std::shared_ptr<Regex> t_traceRegex;
static thread_local std::unique_ptr<tcpClientCounts_t> t_tcpClientCounts;

thread_local std::unique_ptr<MT_t> MT; // the big MTasker
thread_local std::unique_ptr<MemRecursorCache> t_RC;
thread_local std::unique_ptr<RecursorPacketCache> t_packetCache;
thread_local FDMultiplexer* t_fdm{nullptr};
thread_local std::unique_ptr<addrringbuf_t> t_remotes, t_servfailremotes, t_largeanswerremotes;
thread_local std::unique_ptr<boost::circular_buffer<pair<DNSName, uint16_t> > > t_queryring, t_servfailqueryring;
thread_local std::shared_ptr<NetmaskGroup> t_allowFrom;
#ifdef HAVE_PROTOBUF
thread_local std::unique_ptr<boost::uuids::random_generator> t_uuidGenerator;
#endif
__thread struct timeval g_now; // timestamp, updated (too) frequently

// for communicating with our threads
struct ThreadPipeSet
{
  int writeToThread;
  int readToThread;
  int writeFromThread;
  int readFromThread;
};

typedef vector<int> tcpListenSockets_t;
typedef map<int, ComboAddress> listenSocketsAddresses_t; // is shared across all threads right now
typedef vector<pair<int, function< void(int, any&) > > > deferredAdd_t;

static const ComboAddress g_local4("0.0.0.0"), g_local6("::");
static vector<ThreadPipeSet> g_pipes; // effectively readonly after startup
static tcpListenSockets_t g_tcpListenSockets;   // shared across threads, but this is fine, never written to from a thread. All threads listen on all sockets
static listenSocketsAddresses_t g_listenSocketsAddresses; // is shared across all threads right now
static std::unordered_map<unsigned int, deferredAdd_t> deferredAdds;
static set<int> g_fromtosockets; // listen sockets that use 'sendfromto()' mechanism
static vector<ComboAddress> g_localQueryAddresses4, g_localQueryAddresses6;
static AtomicCounter counter;
static std::shared_ptr<SyncRes::domainmap_t> g_initialDomainMap; // new threads needs this to be setup
static std::shared_ptr<NetmaskGroup> g_initialAllowFrom; // new thread needs to be setup with this
static size_t g_tcpMaxQueriesPerConn;
static uint64_t g_latencyStatSize;
static uint32_t g_disthashseed;
static unsigned int g_maxTCPPerClient;
static unsigned int g_networkTimeoutMsec;
static unsigned int g_maxMThreads;
static unsigned int g_numWorkerThreads;
static int g_tcpTimeout;
static uint16_t g_udpTruncationThreshold;
static std::atomic<bool> statsWanted;
static std::atomic<bool> g_quiet;
static bool g_logCommonErrors;
static bool g_anyToTcp;
static bool g_weDistributeQueries; // if true, only 1 thread listens on the incoming query sockets
static bool g_reusePort{false};
static bool g_useOneSocketPerThread;
static bool g_gettagNeedsEDNSOptions{false};
static time_t g_statisticsInterval;
static bool g_useIncomingECS;
std::atomic<uint32_t> g_maxCacheEntries, g_maxPacketCacheEntries;

RecursorControlChannel s_rcc; // only active in thread 0
RecursorStats g_stats;
string s_programname="pdns_recursor";
string s_pidfname;
bool g_lowercaseOutgoing;
unsigned int g_numThreads;
uint16_t g_outgoingEDNSBufsize;
bool g_logRPZChanges{false};

#define LOCAL_NETS "127.0.0.0/8, 10.0.0.0/8, 100.64.0.0/10, 169.254.0.0/16, 192.168.0.0/16, 172.16.0.0/12, ::1/128, fc00::/7, fe80::/10"
// Bad Nets taken from both:
// http://www.iana.org/assignments/iana-ipv4-special-registry/iana-ipv4-special-registry.xhtml
// and
// http://www.iana.org/assignments/iana-ipv6-special-registry/iana-ipv6-special-registry.xhtml
// where such a network may not be considered a valid destination
#define BAD_NETS   "0.0.0.0/8, 192.0.0.0/24, 192.0.2.0/24, 198.51.100.0/24, 203.0.113.0/24, 240.0.0.0/4, ::/96, ::ffff:0:0/96, 100::/64, 2001:db8::/32"
#define DONT_QUERY LOCAL_NETS ", " BAD_NETS

//! used to send information to a newborn mthread
struct DNSComboWriter {
  DNSComboWriter(const char* data, uint16_t len, const struct timeval& now) : d_mdp(true, data, len), d_now(now),
                                                                                                        d_tcp(false), d_socket(-1)
  {}
  MOADNSParser d_mdp;
  void setRemote(const ComboAddress* sa)
  {
    d_remote=*sa;
  }

  void setLocal(const ComboAddress& sa)
  {
    d_local=sa;
  }


  void setSocket(int sock)
  {
    d_socket=sock;
  }

  string getRemote() const
  {
    return d_remote.toString();
  }

  struct timeval d_now;
  ComboAddress d_remote, d_local;
#ifdef HAVE_PROTOBUF
  boost::uuids::uuid d_uuid;
  string d_requestorId;
  string d_deviceId;
#endif
  EDNSSubnetOpts d_ednssubnet;
  bool d_ecsFound{false};
  bool d_ecsParsed{false};
  bool d_tcp;
  int d_socket;
  unsigned int d_tag{0};
  uint32_t d_qhash{0};
  string d_query;
  shared_ptr<TCPConnection> d_tcpConnection;
  vector<pair<uint16_t, string> > d_ednsOpts;
  std::vector<std::string> d_policyTags;
  LuaContext::LuaObject d_data;
  uint32_t d_ttlCap{std::numeric_limits<uint32_t>::max()};
  bool d_variable{false};
};

MT_t* getMT()
{
  return MT ? MT.get() : nullptr;
}

ArgvMap &arg()
{
  static ArgvMap theArg;
  return theArg;
}

unsigned int getRecursorThreadId()
{
  return t_id;
}

int getMTaskerTID()
{
  return MT->getTid();
}

static void handleTCPClientWritable(int fd, FDMultiplexer::funcparam_t& var);

// -1 is error, 0 is timeout, 1 is success
int asendtcp(const string& data, Socket* sock)
{
  PacketID pident;
  pident.sock=sock;
  pident.outMSG=data;

  t_fdm->addWriteFD(sock->getHandle(), handleTCPClientWritable, pident);
  string packet;

  int ret=MT->waitEvent(pident, &packet, g_networkTimeoutMsec);

  if(!ret || ret==-1) { // timeout
    t_fdm->removeWriteFD(sock->getHandle());
  }
  else if(packet.size() !=data.size()) { // main loop tells us what it sent out, or empty in case of an error
    return -1;
  }
  return ret;
}

static void handleTCPClientReadable(int fd, FDMultiplexer::funcparam_t& var);

// -1 is error, 0 is timeout, 1 is success
int arecvtcp(string& data, size_t len, Socket* sock, bool incompleteOkay)
{
  data.clear();
  PacketID pident;
  pident.sock=sock;
  pident.inNeeded=len;
  pident.inIncompleteOkay=incompleteOkay;
  t_fdm->addReadFD(sock->getHandle(), handleTCPClientReadable, pident);

  int ret=MT->waitEvent(pident,&data, g_networkTimeoutMsec);
  if(!ret || ret==-1) { // timeout
    t_fdm->removeReadFD(sock->getHandle());
  }
  else if(data.empty()) {// error, EOF or other
    return -1;
  }

  return ret;
}

static void handleGenUDPQueryResponse(int fd, FDMultiplexer::funcparam_t& var)
{
  PacketID pident=*any_cast<PacketID>(&var);
  char resp[512];
  ssize_t ret=recv(fd, resp, sizeof(resp), 0);
  t_fdm->removeReadFD(fd);
  if(ret >= 0) {
    string data(resp, (size_t) ret);
    MT->sendEvent(pident, &data);
  }
  else {
    string empty;
    MT->sendEvent(pident, &empty);
    //    cerr<<"Had some kind of error: "<<ret<<", "<<strerror(errno)<<endl;
  }
}
string GenUDPQueryResponse(const ComboAddress& dest, const string& query)
{
  Socket s(dest.sin4.sin_family, SOCK_DGRAM);
  s.setNonBlocking();
  ComboAddress local = getQueryLocalAddress(dest.sin4.sin_family, 0);
  
  s.bind(local);
  s.connect(dest);
  s.send(query);

  PacketID pident;
  pident.sock=&s;
  pident.type=0;
  t_fdm->addReadFD(s.getHandle(), handleGenUDPQueryResponse, pident);

  string data;
 
  int ret=MT->waitEvent(pident,&data, g_networkTimeoutMsec);
 
  if(!ret || ret==-1) { // timeout
    t_fdm->removeReadFD(s.getHandle());
  }
  else if(data.empty()) {// error, EOF or other
    // we could special case this
    return data;
  }
  return data;
}

//! pick a random query local address
ComboAddress getQueryLocalAddress(int family, uint16_t port)
{
  ComboAddress ret;
  if(family==AF_INET) {
    if(g_localQueryAddresses4.empty())
      ret = g_local4;
    else
      ret = g_localQueryAddresses4[dns_random(g_localQueryAddresses4.size())];
    ret.sin4.sin_port = htons(port);
  }
  else {
    if(g_localQueryAddresses6.empty())
      ret = g_local6;
    else
      ret = g_localQueryAddresses6[dns_random(g_localQueryAddresses6.size())];

    ret.sin6.sin6_port = htons(port);
  }
  return ret;
}

static void handleUDPServerResponse(int fd, FDMultiplexer::funcparam_t&);

static void setSocketBuffer(int fd, int optname, uint32_t size)
{
  uint32_t psize=0;
  socklen_t len=sizeof(psize);

  if(!getsockopt(fd, SOL_SOCKET, optname, (char*)&psize, &len) && psize > size) {
    L<<Logger::Error<<"Not decreasing socket buffer size from "<<psize<<" to "<<size<<endl;
    return;
  }

  if (setsockopt(fd, SOL_SOCKET, optname, (char*)&size, sizeof(size)) < 0 )
    L<<Logger::Error<<"Unable to raise socket buffer size to "<<size<<": "<<strerror(errno)<<endl;
}


static void setSocketReceiveBuffer(int fd, uint32_t size)
{
  setSocketBuffer(fd, SO_RCVBUF, size);
}

static void setSocketSendBuffer(int fd, uint32_t size)
{
  setSocketBuffer(fd, SO_SNDBUF, size);
}


// you can ask this class for a UDP socket to send a query from
// this socket is not yours, don't even think about deleting it
// but after you call 'returnSocket' on it, don't assume anything anymore
class UDPClientSocks
{
  unsigned int d_numsocks;
public:
  UDPClientSocks() : d_numsocks(0)
  {
  }

  typedef set<int> socks_t;
  socks_t d_socks;

  // returning -2 means: temporary OS error (ie, out of files), -1 means error related to remote
  int getSocket(const ComboAddress& toaddr, int* fd)
  {
    *fd=makeClientSocket(toaddr.sin4.sin_family);
    if(*fd < 0) // temporary error - receive exception otherwise
      return -2;

    if(connect(*fd, (struct sockaddr*)(&toaddr), toaddr.getSocklen()) < 0) {
      int err = errno;
      //      returnSocket(*fd);
      try {
        closesocket(*fd);
      }
      catch(const PDNSException& e) {
        L<<Logger::Error<<"Error closing UDP socket after connect() failed: "<<e.reason<<endl;
      }

      if(err==ENETUNREACH) // Seth "My Interfaces Are Like A Yo Yo" Arnold special
        return -2;
      return -1;
    }

    d_socks.insert(*fd);
    d_numsocks++;
    return 0;
  }

  void returnSocket(int fd)
  {
    socks_t::iterator i=d_socks.find(fd);
    if(i==d_socks.end()) {
      throw PDNSException("Trying to return a socket (fd="+std::to_string(fd)+") not in the pool");
    }
    returnSocketLocked(i);
  }

  // return a socket to the pool, or simply erase it
  void returnSocketLocked(socks_t::iterator& i)
  {
    if(i==d_socks.end()) {
      throw PDNSException("Trying to return a socket not in the pool");
    }
    try {
      t_fdm->removeReadFD(*i);
    }
    catch(FDMultiplexerException& e) {
      // we sometimes return a socket that has not yet been assigned to t_fdm
    }
    try {
      closesocket(*i);
    }
    catch(const PDNSException& e) {
      L<<Logger::Error<<"Error closing returned UDP socket: "<<e.reason<<endl;
    }

    d_socks.erase(i++);
    --d_numsocks;
  }

  // returns -1 for errors which might go away, throws for ones that won't
  static int makeClientSocket(int family)
  {
    int ret=socket(family, SOCK_DGRAM, 0 ); // turns out that setting CLO_EXEC and NONBLOCK from here is not a performance win on Linux (oddly enough)

    if(ret < 0 && errno==EMFILE) // this is not a catastrophic error
      return ret;

    if(ret<0)
      throw PDNSException("Making a socket for resolver (family = "+std::to_string(family)+"): "+stringerror());

    //    setCloseOnExec(ret); // we're not going to exec

    int tries=10;
    ComboAddress sin;
    while(--tries) {
      uint16_t port;

      if(tries==1)  // fall back to kernel 'random'
        port = 0;
      else
        port = 1025 + dns_random(64510);

      sin=getQueryLocalAddress(family, port); // does htons for us

      if (::bind(ret, (struct sockaddr *)&sin, sin.getSocklen()) >= 0)
        break;
    }
    if(!tries)
      throw PDNSException("Resolver binding to local query client socket on "+sin.toString()+": "+stringerror());

    setNonBlocking(ret);
    return ret;
  }
};

static thread_local std::unique_ptr<UDPClientSocks> t_udpclientsocks;

/* these two functions are used by LWRes */
// -2 is OS error, -1 is error that depends on the remote, > 0 is success
int asendto(const char *data, size_t len, int flags,
            const ComboAddress& toaddr, uint16_t id, const DNSName& domain, uint16_t qtype, int* fd)
{

  PacketID pident;
  pident.domain = domain;
  pident.remote = toaddr;
  pident.type = qtype;

  // see if there is an existing outstanding request we can chain on to, using partial equivalence function
  pair<MT_t::waiters_t::iterator, MT_t::waiters_t::iterator> chain=MT->d_waiters.equal_range(pident, PacketIDBirthdayCompare());

  for(; chain.first != chain.second; chain.first++) {
    if(chain.first->key.fd > -1) { // don't chain onto existing chained waiter!
      /*
      cerr<<"Orig: "<<pident.domain<<", "<<pident.remote.toString()<<", id="<<id<<endl;
      cerr<<"Had hit: "<< chain.first->key.domain<<", "<<chain.first->key.remote.toString()<<", id="<<chain.first->key.id
          <<", count="<<chain.first->key.chain.size()<<", origfd: "<<chain.first->key.fd<<endl;
      */
      chain.first->key.chain.insert(id); // we can chain
      *fd=-1;                            // gets used in waitEvent / sendEvent later on
      return 1;
    }
  }

  int ret=t_udpclientsocks->getSocket(toaddr, fd);
  if(ret < 0)
    return ret;

  pident.fd=*fd;
  pident.id=id;

  t_fdm->addReadFD(*fd, handleUDPServerResponse, pident);
  ret = send(*fd, data, len, 0);

  int tmp = errno;

  if(ret < 0)
    t_udpclientsocks->returnSocket(*fd);

  errno = tmp; // this is for logging purposes only
  return ret;
}

// -1 is error, 0 is timeout, 1 is success
int arecvfrom(char *data, size_t len, int flags, const ComboAddress& fromaddr, size_t *d_len,
              uint16_t id, const DNSName& domain, uint16_t qtype, int fd, struct timeval* now)
{
  static optional<unsigned int> nearMissLimit;
  if(!nearMissLimit)
    nearMissLimit=::arg().asNum("spoof-nearmiss-max");

  PacketID pident;
  pident.fd=fd;
  pident.id=id;
  pident.domain=domain;
  pident.type = qtype;
  pident.remote=fromaddr;

  string packet;
  int ret=MT->waitEvent(pident, &packet, g_networkTimeoutMsec, now);

  if(ret > 0) {
    if(packet.empty()) // means "error"
      return -1;

    *d_len=packet.size();
    memcpy(data,packet.c_str(),min(len,*d_len));
    if(*nearMissLimit && pident.nearMisses > *nearMissLimit) {
      L<<Logger::Error<<"Too many ("<<pident.nearMisses<<" > "<<*nearMissLimit<<") bogus answers for '"<<domain<<"' from "<<fromaddr.toString()<<", assuming spoof attempt."<<endl;
      g_stats.spoofCount++;
      return -1;
    }
  }
  else {
    if(fd >= 0)
      t_udpclientsocks->returnSocket(fd);
  }
  return ret;
}

static void writePid(void)
{
  if(!::arg().mustDo("write-pid"))
    return;
  ofstream of(s_pidfname.c_str(), std::ios_base::app);
  if(of)
    of<< Utility::getpid() <<endl;
  else
    L<<Logger::Error<<"Writing pid for "<<Utility::getpid()<<" to "<<s_pidfname<<" failed: "<<strerror(errno)<<endl;
}

TCPConnection::TCPConnection(int fd, const ComboAddress& addr) : d_remote(addr), d_fd(fd)
{
  ++s_currentConnections;
  (*t_tcpClientCounts)[d_remote]++;
}

TCPConnection::~TCPConnection()
{
  try {
    if(closesocket(d_fd) < 0)
      L<<Logger::Error<<"Error closing socket for TCPConnection"<<endl;
  }
  catch(const PDNSException& e) {
    L<<Logger::Error<<"Error closing TCPConnection socket: "<<e.reason<<endl;
  }

  if(t_tcpClientCounts->count(d_remote) && !(*t_tcpClientCounts)[d_remote]--)
    t_tcpClientCounts->erase(d_remote);
  --s_currentConnections;
}

AtomicCounter TCPConnection::s_currentConnections;

static void handleRunningTCPQuestion(int fd, FDMultiplexer::funcparam_t& var);

// the idea is, only do things that depend on the *response* here. Incoming accounting is on incoming.
static void updateResponseStats(int res, const ComboAddress& remote, unsigned int packetsize, const DNSName* query, uint16_t qtype)
{
  if(packetsize > 1000 && t_largeanswerremotes)
    t_largeanswerremotes->push_back(remote);
  switch(res) {
  case RCode::ServFail:
    if(t_servfailremotes) {
      t_servfailremotes->push_back(remote);
      if(query && t_servfailqueryring) // packet cache
        t_servfailqueryring->push_back(make_pair(*query, qtype));
    }
    g_stats.servFails++;
    break;
  case RCode::NXDomain:
    g_stats.nxDomains++;
    break;
  case RCode::NoError:
    g_stats.noErrors++;
    break;
  }
}

static string makeLoginfo(DNSComboWriter* dc)
try
{
  return "("+dc->d_mdp.d_qname.toLogString()+"/"+DNSRecordContent::NumberToType(dc->d_mdp.d_qtype)+" from "+(dc->d_remote.toString())+")";
}
catch(...)
{
  return "Exception making error message for exception";
}

#ifdef HAVE_PROTOBUF
static void protobufLogQuery(const std::shared_ptr<RemoteLogger>& logger, uint8_t maskV4, uint8_t maskV6, const boost::uuids::uuid& uniqueId, const ComboAddress& remote, const ComboAddress& local, const Netmask& ednssubnet, bool tcp, uint16_t id, size_t len, const DNSName& qname, uint16_t qtype, uint16_t qclass, const std::vector<std::string>& policyTags, const std::string& requestorId, const std::string& deviceId)
{
  Netmask requestorNM(remote, remote.sin4.sin_family == AF_INET ? maskV4 : maskV6);
  const ComboAddress& requestor = requestorNM.getMaskedNetwork();
  RecProtoBufMessage message(DNSProtoBufMessage::Query, uniqueId, &requestor, &local, qname, qtype, qclass, id, tcp, len);
  message.setEDNSSubnet(ednssubnet, ednssubnet.isIpv4() ? maskV4 : maskV6);
  message.setRequestorId(requestorId);
  message.setDeviceId(deviceId);

  if (!policyTags.empty()) {
    message.setPolicyTags(policyTags);
  }

//  cerr <<message.toDebugString()<<endl;
  std::string str;
  message.serialize(str);
  logger->queueData(str);
}

static void protobufLogResponse(const std::shared_ptr<RemoteLogger>& logger, const RecProtoBufMessage& message)
{
//  cerr <<message.toDebugString()<<endl;
  std::string str;
  message.serialize(str);
  logger->queueData(str);
}
#endif

/**
 * Chases the CNAME provided by the PolicyCustom RPZ policy.
 *
 * @param spoofed: The DNSRecord that was created by the policy, should already be added to ret
 * @param qtype: The QType of the original query
 * @param sr: A SyncRes
 * @param res: An integer that will contain the RCODE of the lookup we do
 * @param ret: A vector of DNSRecords where the result of the CNAME chase should be appended to
 */
static void handleRPZCustom(const DNSRecord& spoofed, const QType& qtype, SyncRes& sr, int& res, vector<DNSRecord>& ret)
{
  if (spoofed.d_type == QType::CNAME) {
    bool oldWantsRPZ = sr.getWantsRPZ();
    sr.setWantsRPZ(false);
    vector<DNSRecord> ans;
    res = sr.beginResolve(DNSName(spoofed.d_content->getZoneRepresentation()), qtype, 1, ans);
    for (const auto& rec : ans) {
      if(rec.d_place == DNSResourceRecord::ANSWER) {
        ret.push_back(rec);
      }
    }
    // Reset the RPZ state of the SyncRes
    sr.setWantsRPZ(oldWantsRPZ);
  }
}

static bool addRecordToPacket(DNSPacketWriter& pw, const DNSRecord& rec, uint32_t& minTTL, uint32_t ttlCap, const uint16_t maxAnswerSize)
{
  pw.startRecord(rec.d_name, rec.d_type, (rec.d_ttl > ttlCap ? ttlCap : rec.d_ttl), rec.d_class, rec.d_place);

  if(rec.d_type != QType::OPT) // their TTL ain't real
    minTTL = min(minTTL, rec.d_ttl);

  rec.d_content->toPacket(pw);
  if(pw.size() > static_cast<size_t>(maxAnswerSize)) {
    pw.rollback();
    if(rec.d_place != DNSResourceRecord::ADDITIONAL) {
      pw.getHeader()->tc=1;
      pw.truncate();
    }
    return false;
  }

  return true;
}

static void startDoResolve(void *p)
{
  DNSComboWriter* dc=(DNSComboWriter *)p;
  try {
    if (t_queryring)
      t_queryring->push_back(make_pair(dc->d_mdp.d_qname, dc->d_mdp.d_qtype));

    uint16_t maxanswersize = dc->d_tcp ? 65535 : min(static_cast<uint16_t>(512), g_udpTruncationThreshold);
    EDNSOpts edo;
    bool haveEDNS=false;
    if(getEDNSOpts(dc->d_mdp, &edo)) {
      if(!dc->d_tcp) {
        /* rfc6891 6.2.3:
           "Values lower than 512 MUST be treated as equal to 512."
        */
        maxanswersize = min(static_cast<uint16_t>(edo.d_packetsize >= 512 ? edo.d_packetsize : 512), g_udpTruncationThreshold);
      }
      dc->d_ednsOpts = edo.d_options;
      haveEDNS=true;

      if (g_useIncomingECS && !dc->d_ecsParsed) {
        for (const auto& o : edo.d_options) {
          if (o.first == EDNSOptionCode::ECS) {
            dc->d_ecsFound = getEDNSSubnetOptsFromString(o.second, &dc->d_ednssubnet);
            break;
          }
        }
      }
    }
    /* perhaps there was no EDNS or no ECS but by now we looked */
    dc->d_ecsParsed = true;
    vector<DNSRecord> ret;
    vector<uint8_t> packet;

    auto luaconfsLocal = g_luaconfs.getLocal();
    // Used to tell syncres later on if we should apply NSDNAME and NSIP RPZ triggers for this query
    bool wantsRPZ(true);
    RecProtoBufMessage pbMessage(RecProtoBufMessage::Response);
#ifdef HAVE_PROTOBUF
    if (luaconfsLocal->protobufServer) {
      Netmask requestorNM(dc->d_remote, dc->d_remote.sin4.sin_family == AF_INET ? luaconfsLocal->protobufMaskV4 : luaconfsLocal->protobufMaskV6);
      const ComboAddress& requestor = requestorNM.getMaskedNetwork();
      pbMessage.update(dc->d_uuid, &requestor, &dc->d_local, dc->d_tcp, dc->d_mdp.d_header.id);
      pbMessage.setEDNSSubnet(dc->d_ednssubnet.source, dc->d_ednssubnet.source.isIpv4() ? luaconfsLocal->protobufMaskV4 : luaconfsLocal->protobufMaskV6);
      pbMessage.setQuestion(dc->d_mdp.d_qname, dc->d_mdp.d_qtype, dc->d_mdp.d_qclass);
    }
#endif /* HAVE_PROTOBUF */

    DNSPacketWriter pw(packet, dc->d_mdp.d_qname, dc->d_mdp.d_qtype, dc->d_mdp.d_qclass);

    pw.getHeader()->aa=0;
    pw.getHeader()->ra=1;
    pw.getHeader()->qr=1;
    pw.getHeader()->tc=0;
    pw.getHeader()->id=dc->d_mdp.d_header.id;
    pw.getHeader()->rd=dc->d_mdp.d_header.rd;
    pw.getHeader()->cd=dc->d_mdp.d_header.cd;

    /* This is the lowest TTL seen in the records of the response,
       so we can't cache it for longer than this value.
       If we have a TTL cap, this value can't be larger than the
       cap no matter what. */
    uint32_t minTTL = dc->d_ttlCap;

    SyncRes sr(dc->d_now);

    bool DNSSECOK=false;
    if(t_pdl) {
      sr.setLuaEngine(t_pdl);
    }
    sr.d_requestor=dc->d_remote; // ECS needs this too
    if(g_dnssecmode != DNSSECMode::Off) {
      sr.setDoDNSSEC(true);

      // Does the requestor want DNSSEC records?
      if(edo.d_Z & EDNSOpts::DNSSECOK) {
        DNSSECOK=true;
        g_stats.dnssecQueries++;
      }
    } else {
      // Ignore the client-set CD flag
      pw.getHeader()->cd=0;
    }
    sr.setDNSSECValidationRequested(g_dnssecmode == DNSSECMode::ValidateAll || g_dnssecmode==DNSSECMode::ValidateForLog || ((dc->d_mdp.d_header.ad || DNSSECOK) && g_dnssecmode==DNSSECMode::Process));

#ifdef HAVE_PROTOBUF
    sr.setInitialRequestId(dc->d_uuid);
#endif

    if (g_useIncomingECS) {
      sr.setIncomingECSFound(dc->d_ecsFound);
      if (dc->d_ecsFound) {
        sr.setIncomingECS(dc->d_ednssubnet);
      }
    }

    bool tracedQuery=false; // we could consider letting Lua know about this too
    bool variableAnswer = dc->d_variable;
    bool shouldNotValidate = false;

    /* preresolve expects res (dq.rcode) to be set to RCode::NoError by default */
    int res = RCode::NoError;
    DNSFilterEngine::Policy appliedPolicy;
    DNSRecord spoofed;
    RecursorLua4::DNSQuestion dq(dc->d_remote, dc->d_local, dc->d_mdp.d_qname, dc->d_mdp.d_qtype, dc->d_tcp, variableAnswer, wantsRPZ);
    dq.ednsFlags = &edo.d_Z;
    dq.ednsOptions = &dc->d_ednsOpts;
    dq.tag = dc->d_tag;
    dq.discardedPolicies = &sr.d_discardedPolicies;
    dq.policyTags = &dc->d_policyTags;
    dq.appliedPolicy = &appliedPolicy;
    dq.currentRecords = &ret;
    dq.dh = &dc->d_mdp.d_header;
    dq.data = dc->d_data;
#ifdef HAVE_PROTOBUF
    dq.requestorId = dc->d_requestorId;
    dq.deviceId = dc->d_deviceId;
#endif

    if(dc->d_mdp.d_qtype==QType::ANY && !dc->d_tcp && g_anyToTcp) {
      pw.getHeader()->tc = 1;
      res = 0;
      variableAnswer = true;
      goto sendit;
    }

    if(t_traceRegex && t_traceRegex->match(dc->d_mdp.d_qname.toString())) {
      sr.setLogMode(SyncRes::Store);
      tracedQuery=true;
    }


    if(!g_quiet || tracedQuery) {
      L<<Logger::Warning<<t_id<<" ["<<MT->getTid()<<"/"<<MT->numProcesses()<<"] " << (dc->d_tcp ? "TCP " : "") << "question for '"<<dc->d_mdp.d_qname<<"|"
       <<DNSRecordContent::NumberToType(dc->d_mdp.d_qtype)<<"' from "<<dc->getRemote();
      if(!dc->d_ednssubnet.source.empty()) {
        L<<" (ecs "<<dc->d_ednssubnet.source.toString()<<")";
      }
      L<<endl;
    }

    sr.setId(MT->getTid());
    if(!dc->d_mdp.d_header.rd)
      sr.setCacheOnly();

    if (t_pdl) {
      t_pdl->prerpz(dq, res);
    }

    // Check if the query has a policy attached to it
    if (wantsRPZ) {
      appliedPolicy = luaconfsLocal->dfe.getQueryPolicy(dc->d_mdp.d_qname, dc->d_remote, sr.d_discardedPolicies);
    }

    // if there is a RecursorLua active, and it 'took' the query in preResolve, we don't launch beginResolve
    if(!t_pdl || !t_pdl->preresolve(dq, res)) {

      sr.setWantsRPZ(wantsRPZ);
      if(wantsRPZ) {
        switch(appliedPolicy.d_kind) {
          case DNSFilterEngine::PolicyKind::NoAction:
            break;
          case DNSFilterEngine::PolicyKind::Drop:
            g_stats.policyDrops++;
            g_stats.policyResults[appliedPolicy.d_kind]++;
            delete dc;
            dc=0;
            return; 
          case DNSFilterEngine::PolicyKind::NXDOMAIN:
            g_stats.policyResults[appliedPolicy.d_kind]++;
            res=RCode::NXDomain;
            goto haveAnswer;
          case DNSFilterEngine::PolicyKind::NODATA:
            g_stats.policyResults[appliedPolicy.d_kind]++;
            res=RCode::NoError;
            goto haveAnswer;
          case DNSFilterEngine::PolicyKind::Custom:
            g_stats.policyResults[appliedPolicy.d_kind]++;
            res=RCode::NoError;
            spoofed=appliedPolicy.getCustomRecord(dc->d_mdp.d_qname);
            ret.push_back(spoofed);
            handleRPZCustom(spoofed, QType(dc->d_mdp.d_qtype), sr, res, ret);
            goto haveAnswer;
          case DNSFilterEngine::PolicyKind::Truncate:
            if(!dc->d_tcp) {
              g_stats.policyResults[appliedPolicy.d_kind]++;
              res=RCode::NoError;       
              pw.getHeader()->tc=1;
              goto haveAnswer;
            }
            break;
        }
      }

      // Query got not handled for QNAME Policy reasons, now actually go out to find an answer
      try {
        res = sr.beginResolve(dc->d_mdp.d_qname, QType(dc->d_mdp.d_qtype), dc->d_mdp.d_qclass, ret);
        shouldNotValidate = sr.wasOutOfBand();
      }
      catch(ImmediateServFailException &e) {
        if(g_logCommonErrors)
          L<<Logger::Notice<<"Sending SERVFAIL to "<<dc->getRemote()<<" during resolve of '"<<dc->d_mdp.d_qname<<"' because: "<<e.reason<<endl;
        res = RCode::ServFail;
      }

      dq.validationState = sr.getValidationState();

      // During lookup, an NSDNAME or NSIP trigger was hit in RPZ
      if (res == -2) { // XXX This block should be macro'd, it is repeated post-resolve.
        appliedPolicy = sr.d_appliedPolicy;
        g_stats.policyResults[appliedPolicy.d_kind]++;
        switch(appliedPolicy.d_kind) {
          case DNSFilterEngine::PolicyKind::NoAction: // This can never happen
            throw PDNSException("NoAction policy returned while a NSDNAME or NSIP trigger was hit");
          case DNSFilterEngine::PolicyKind::Drop:
            g_stats.policyDrops++;
            delete dc;
            dc=0;
            return;
          case DNSFilterEngine::PolicyKind::NXDOMAIN:
            ret.clear();
            res=RCode::NXDomain;
            goto haveAnswer;

          case DNSFilterEngine::PolicyKind::NODATA:
            ret.clear();
            res=RCode::NoError;
            goto haveAnswer;

          case DNSFilterEngine::PolicyKind::Truncate:
            if(!dc->d_tcp) {
              ret.clear();
              res=RCode::NoError;
              pw.getHeader()->tc=1;
              goto haveAnswer;
            }
            break;

          case DNSFilterEngine::PolicyKind::Custom:
            ret.clear();
            res=RCode::NoError;
            spoofed=appliedPolicy.getCustomRecord(dc->d_mdp.d_qname);
            ret.push_back(spoofed);
            handleRPZCustom(spoofed, QType(dc->d_mdp.d_qtype), sr, res, ret);
            goto haveAnswer;
        }
      }

      if (wantsRPZ) {
        appliedPolicy = luaconfsLocal->dfe.getPostPolicy(ret, sr.d_discardedPolicies);
      }

      if(t_pdl) {
        if(res == RCode::NoError) {
                auto i=ret.cbegin();
                for(; i!= ret.cend(); ++i)
                  if(i->d_type == dc->d_mdp.d_qtype && i->d_place == DNSResourceRecord::ANSWER)
                          break;
                if(i == ret.cend() && t_pdl->nodata(dq, res))
                  shouldNotValidate = true;

        }
        else if(res == RCode::NXDomain && t_pdl->nxdomain(dq, res))
          shouldNotValidate = true;

        if(t_pdl->postresolve(dq, res))
          shouldNotValidate = true;
      }

      if (wantsRPZ) { //XXX This block is repeated, see above
        g_stats.policyResults[appliedPolicy.d_kind]++;
        switch(appliedPolicy.d_kind) {
          case DNSFilterEngine::PolicyKind::NoAction:
            break;
          case DNSFilterEngine::PolicyKind::Drop:
            g_stats.policyDrops++;
            delete dc;
            dc=0;
            return; 
          case DNSFilterEngine::PolicyKind::NXDOMAIN:
            ret.clear();
            res=RCode::NXDomain;
            goto haveAnswer;

          case DNSFilterEngine::PolicyKind::NODATA:
            ret.clear();
            res=RCode::NoError;
            goto haveAnswer;

          case DNSFilterEngine::PolicyKind::Truncate:
            if(!dc->d_tcp) {
              ret.clear();
              res=RCode::NoError;
              pw.getHeader()->tc=1;
              goto haveAnswer;
            }
            break;

          case DNSFilterEngine::PolicyKind::Custom:
            ret.clear();
            res=RCode::NoError;
            spoofed=appliedPolicy.getCustomRecord(dc->d_mdp.d_qname);
            ret.push_back(spoofed);
            handleRPZCustom(spoofed, QType(dc->d_mdp.d_qtype), sr, res, ret);
            goto haveAnswer;
        }
      }
    }
  haveAnswer:;
    if(res == PolicyDecision::DROP) {
      g_stats.policyDrops++;
      delete dc;
      dc=0;
      return;
    }
    if(tracedQuery || res == -1 || res == RCode::ServFail || pw.getHeader()->rcode == RCode::ServFail)
    { 
      string trace(sr.getTrace());
      if(!trace.empty()) {
        vector<string> lines;
        boost::split(lines, trace, boost::is_any_of("\n"));
        for(const string& line : lines) {
          if(!line.empty())
            L<<Logger::Warning<< line << endl;
        }
      }
    }

    if(res == -1) {
      pw.getHeader()->rcode=RCode::ServFail;
      // no commit here, because no record
      g_stats.servFails++;
    }
    else {
      pw.getHeader()->rcode=res;

      // Does the validation mode or query demand validation?
      if(!shouldNotValidate && sr.isDNSSECValidationRequested()) {
        try {
          if(sr.doLog()) {
            L<<Logger::Warning<<"Starting validation of answer to "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" for "<<dc->d_remote.toStringWithPort()<<endl;
          }

          auto state = sr.getValidationState();

          if(state == Secure) {
            if(sr.doLog()) {
              L<<Logger::Warning<<"Answer to "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" for "<<dc->d_remote.toStringWithPort()<<" validates correctly"<<endl;
            }
            
            // Is the query source interested in the value of the ad-bit?
            if (dc->d_mdp.d_header.ad || DNSSECOK)
              pw.getHeader()->ad=1;
          }
          else if(state == Insecure) {
            if(sr.doLog()) {
              L<<Logger::Warning<<"Answer to "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" for "<<dc->d_remote.toStringWithPort()<<" validates as Insecure"<<endl;
            }
            
            pw.getHeader()->ad=0;
          }
          else if(state == Bogus) {
            if(g_dnssecLogBogus || sr.doLog() || g_dnssecmode == DNSSECMode::ValidateForLog) {
              L<<Logger::Warning<<"Answer to "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" for "<<dc->d_remote.toStringWithPort()<<" validates as Bogus"<<endl;
            }
            
            // Does the query or validation mode sending out a SERVFAIL on validation errors?
            if(!pw.getHeader()->cd && (g_dnssecmode == DNSSECMode::ValidateAll || dc->d_mdp.d_header.ad || DNSSECOK)) {
              if(sr.doLog()) {
                L<<Logger::Warning<<"Sending out SERVFAIL for "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" because recursor or query demands it for Bogus results"<<endl;
              }
              
              pw.getHeader()->rcode=RCode::ServFail;
              goto sendit;
            } else {
              if(sr.doLog()) {
                L<<Logger::Warning<<"Not sending out SERVFAIL for "<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<" Bogus validation since neither config nor query demands this"<<endl;
              }
            }
          }
        }
        catch(ImmediateServFailException &e) {
          if(g_logCommonErrors)
            L<<Logger::Notice<<"Sending SERVFAIL to "<<dc->getRemote()<<" during validation of '"<<dc->d_mdp.d_qname<<"|"<<QType(dc->d_mdp.d_qtype).getName()<<"' because: "<<e.reason<<endl;
          pw.getHeader()->rcode=RCode::ServFail;
          goto sendit;
        }
      }

      if(ret.size()) {
        orderAndShuffle(ret);
        if(auto sl = luaconfsLocal->sortlist.getOrderCmp(dc->d_remote)) {
          stable_sort(ret.begin(), ret.end(), *sl);
          variableAnswer=true;
        }
      }

      bool needCommit = false;
      for(auto i=ret.cbegin(); i!=ret.cend(); ++i) {
        if( ! DNSSECOK &&
            ( i->d_type == QType::NSEC3 ||
              (
                ( i->d_type == QType::RRSIG || i->d_type==QType::NSEC ) &&
                (
                  ( dc->d_mdp.d_qtype != i->d_type &&  dc->d_mdp.d_qtype != QType::ANY ) ||
                  i->d_place != DNSResourceRecord::ANSWER
                )
              )
            )
          ) {
          continue;
        }

        if (!addRecordToPacket(pw, *i, minTTL, dc->d_ttlCap, maxanswersize)) {
          needCommit = false;
          break;
        }
        needCommit = true;

#ifdef HAVE_PROTOBUF
        if(luaconfsLocal->protobufServer && (i->d_type == QType::A || i->d_type == QType::AAAA || i->d_type == QType::CNAME)) {
          pbMessage.addRR(*i);
        }
#endif
      }
      if(needCommit)
        pw.commit();
    }
  sendit:;

    if (haveEDNS) {
      /* we try to add the EDNS OPT RR even for truncated answers,
         as rfc6891 states:
         "The minimal response MUST be the DNS header, question section, and an
         OPT record.  This MUST also occur when a truncated response (using
         the DNS header's TC bit) is returned."
      */
      if (addRecordToPacket(pw, makeOpt(edo.d_packetsize, 0, edo.d_Z), minTTL, dc->d_ttlCap, maxanswersize)) {
        pw.commit();
      }
    }

    g_rs.submitResponse(dc->d_mdp.d_qtype, packet.size(), !dc->d_tcp);
    updateResponseStats(res, dc->d_remote, packet.size(), &dc->d_mdp.d_qname, dc->d_mdp.d_qtype);
#ifdef HAVE_PROTOBUF
    if (luaconfsLocal->protobufServer && (!luaconfsLocal->protobufTaggedOnly || (appliedPolicy.d_name && !appliedPolicy.d_name->empty()) || !dc->d_policyTags.empty())) {
      pbMessage.setBytes(packet.size());
      pbMessage.setResponseCode(pw.getHeader()->rcode);
      if (appliedPolicy.d_name) {
        pbMessage.setAppliedPolicy(*appliedPolicy.d_name);
        pbMessage.setAppliedPolicyType(appliedPolicy.d_type);
      }
      pbMessage.setPolicyTags(dc->d_policyTags);
      pbMessage.setQueryTime(dc->d_now.tv_sec, dc->d_now.tv_usec);
      pbMessage.setRequestorId(dq.requestorId);
      pbMessage.setDeviceId(dq.deviceId);
      protobufLogResponse(luaconfsLocal->protobufServer, pbMessage);
    }
#endif
    if(!dc->d_tcp) {
      struct msghdr msgh;
      struct iovec iov;
      char cbuf[256];
      fillMSGHdr(&msgh, &iov, cbuf, 0, (char*)&*packet.begin(), packet.size(), &dc->d_remote);
      msgh.msg_control=NULL;

      if(g_fromtosockets.count(dc->d_socket)) {
        addCMsgSrcAddr(&msgh, cbuf, &dc->d_local, 0);
      }
      if(sendmsg(dc->d_socket, &msgh, 0) < 0 && g_logCommonErrors) 
        L<<Logger::Warning<<"Sending UDP reply to client "<<dc->d_remote.toStringWithPort()<<" failed with: "<<strerror(errno)<<endl;

      if(!SyncRes::s_nopacketcache && !variableAnswer && !sr.wasVariable() ) {
        t_packetCache->insertResponsePacket(dc->d_tag, dc->d_qhash, dc->d_mdp.d_qname, dc->d_mdp.d_qtype, dc->d_mdp.d_qclass,
                                            string((const char*)&*packet.begin(), packet.size()),
                                            g_now.tv_sec,
                                            pw.getHeader()->rcode == RCode::ServFail ? SyncRes::s_packetcacheservfailttl :
                                            min(minTTL,SyncRes::s_packetcachettl),
                                            &pbMessage);
      }
      //      else cerr<<"Not putting in packet cache: "<<sr.wasVariable()<<endl;
    }
    else {
      char buf[2];
      buf[0]=packet.size()/256;
      buf[1]=packet.size()%256;

      Utility::iovec iov[2];

      iov[0].iov_base=(void*)buf;              iov[0].iov_len=2;
      iov[1].iov_base=(void*)&*packet.begin(); iov[1].iov_len = packet.size();

      int wret=Utility::writev(dc->d_socket, iov, 2);
      bool hadError=true;

      if(wret == 0)
        L<<Logger::Error<<"EOF writing TCP answer to "<<dc->getRemote()<<endl;
      else if(wret < 0 )
        L<<Logger::Error<<"Error writing TCP answer to "<<dc->getRemote()<<": "<< strerror(errno) <<endl;
      else if((unsigned int)wret != 2 + packet.size())
        L<<Logger::Error<<"Oops, partial answer sent to "<<dc->getRemote()<<" for "<<dc->d_mdp.d_qname<<" (size="<< (2 + packet.size()) <<", sent "<<wret<<")"<<endl;
      else
        hadError=false;

      // update tcp connection status, either by closing or moving to 'BYTE0'

      if(hadError) {
        // no need to remove us from FDM, we weren't there
        dc->d_socket = -1;
      }
      else {
        dc->d_tcpConnection->queriesCount++;
        if (g_tcpMaxQueriesPerConn && dc->d_tcpConnection->queriesCount >= g_tcpMaxQueriesPerConn) {
          dc->d_socket = -1;
        }
        else {
          dc->d_tcpConnection->state=TCPConnection::BYTE0;
          Utility::gettimeofday(&g_now, 0); // needs to be updated
          t_fdm->addReadFD(dc->d_socket, handleRunningTCPQuestion, dc->d_tcpConnection);
          t_fdm->setReadTTD(dc->d_socket, g_now, g_tcpTimeout);
        }
      }
    }
    float spent=makeFloat(sr.getNow()-dc->d_now);
    if(!g_quiet) {
      L<<Logger::Error<<t_id<<" ["<<MT->getTid()<<"/"<<MT->numProcesses()<<"] answer to "<<(dc->d_mdp.d_header.rd?"":"non-rd ")<<"question '"<<dc->d_mdp.d_qname<<"|"<<DNSRecordContent::NumberToType(dc->d_mdp.d_qtype);
      L<<"': "<<ntohs(pw.getHeader()->ancount)<<" answers, "<<ntohs(pw.getHeader()->arcount)<<" additional, took "<<sr.d_outqueries<<" packets, "<<
        sr.d_totUsec/1000.0<<" netw ms, "<< spent*1000.0<<" tot ms, "<<
        sr.d_throttledqueries<<" throttled, "<<sr.d_timeouts<<" timeouts, "<<sr.d_tcpoutqueries<<" tcp connections, rcode="<< res;

      if(!shouldNotValidate && sr.isDNSSECValidationRequested()) {
        L<< ", dnssec="<<vStates[sr.getValidationState()];
      }
        
      L<<endl;

    }

    sr.d_outqueries ? t_RC->cacheMisses++ : t_RC->cacheHits++;

    if(spent < 0.001)
      g_stats.answers0_1++;
    else if(spent < 0.010)
      g_stats.answers1_10++;
    else if(spent < 0.1)
      g_stats.answers10_100++;
    else if(spent < 1.0)
      g_stats.answers100_1000++;
    else
      g_stats.answersSlow++;

    uint64_t newLat=(uint64_t)(spent*1000000);
    newLat = min(newLat,(uint64_t)(((uint64_t) g_networkTimeoutMsec)*1000)); // outliers of several minutes exist..
    g_stats.avgLatencyUsec=(1-1.0/g_latencyStatSize)*g_stats.avgLatencyUsec + (float)newLat/g_latencyStatSize;
    // no worries, we do this for packet cache hits elsewhere

    auto ourtime = 1000.0*spent-sr.d_totUsec/1000.0; // in msec
    if(ourtime < 1)
      g_stats.ourtime0_1++;
    else if(ourtime < 2)
      g_stats.ourtime1_2++;
    else if(ourtime < 4)
      g_stats.ourtime2_4++;
    else if(ourtime < 8)
      g_stats.ourtime4_8++;
    else if(ourtime < 16)
      g_stats.ourtime8_16++;
    else if(ourtime < 32)
      g_stats.ourtime16_32++;
    else {
      //      cerr<<"SLOW: "<<ourtime<<"ms -> "<<dc->d_mdp.d_qname<<"|"<<DNSRecordContent::NumberToType(dc->d_mdp.d_qtype)<<endl;
      g_stats.ourtimeSlow++;
    }
    if(ourtime >= 0.0) {
      newLat=ourtime*1000; // usec
      g_stats.avgLatencyOursUsec=(1-1.0/g_latencyStatSize)*g_stats.avgLatencyOursUsec + (float)newLat/g_latencyStatSize;
    }
    //    cout<<dc->d_mdp.d_qname<<"\t"<<MT->getUsec()<<"\t"<<sr.d_outqueries<<endl;
    delete dc;
    dc=0;
  }
  catch(PDNSException &ae) {
    L<<Logger::Error<<"startDoResolve problem "<<makeLoginfo(dc)<<": "<<ae.reason<<endl;
    delete dc;
  }
  catch(MOADNSException& e) {
    L<<Logger::Error<<"DNS parser error "<<makeLoginfo(dc) <<": "<<dc->d_mdp.d_qname<<", "<<e.what()<<endl;
    delete dc;
  }
  catch(std::exception& e) {
    L<<Logger::Error<<"STL error "<< makeLoginfo(dc)<<": "<<e.what();

    // Luawrapper nests the exception from Lua, so we unnest it here
    try {
        std::rethrow_if_nested(e);
    } catch(const std::exception& ne) {
        L<<". Extra info: "<<ne.what();
    } catch(...) {}

    L<<endl;
    delete dc;
  }
  catch(...) {
    L<<Logger::Error<<"Any other exception in a resolver context "<< makeLoginfo(dc) <<endl;
  }

  g_stats.maxMThreadStackUsage = max(MT->getMaxStackUsage(), g_stats.maxMThreadStackUsage);
}

static void makeControlChannelSocket(int processNum=-1)
{
  string sockname=::arg()["socket-dir"]+"/"+s_programname;
  if(processNum >= 0)
    sockname += "."+std::to_string(processNum);
  sockname+=".controlsocket";
  s_rcc.listen(sockname);

  int sockowner = -1;
  int sockgroup = -1;

  if (!::arg().isEmpty("socket-group"))
    sockgroup=::arg().asGid("socket-group");
  if (!::arg().isEmpty("socket-owner"))
    sockowner=::arg().asUid("socket-owner");

  if (sockgroup > -1 || sockowner > -1) {
    if(chown(sockname.c_str(), sockowner, sockgroup) < 0) {
      unixDie("Failed to chown control socket");
    }
  }

  // do mode change if socket-mode is given
  if(!::arg().isEmpty("socket-mode")) {
    mode_t sockmode=::arg().asMode("socket-mode");
    if(chmod(sockname.c_str(), sockmode) < 0) {
      unixDie("Failed to chmod control socket");
    }
  }
}

static bool getQNameAndSubnet(const std::string& question, DNSName* dnsname, uint16_t* qtype, uint16_t* qclass, EDNSSubnetOpts* ednssubnet, std::map<uint16_t, EDNSOptionView>* options)
{
  bool found = false;
  const struct dnsheader* dh = (struct dnsheader*)question.c_str();
  size_t questionLen = question.length();
  unsigned int consumed=0;
  *dnsname=DNSName(question.c_str(), questionLen, sizeof(dnsheader), false, qtype, qclass, &consumed);

  size_t pos= sizeof(dnsheader)+consumed+4;
  /* at least OPT root label (1), type (2), class (2) and ttl (4) + OPT RR rdlen (2)
     = 11 */
  if(ntohs(dh->arcount) == 1 && questionLen > pos + 11) { // this code can extract one (1) EDNS Subnet option
    /* OPT root label (1) followed by type (2) */
    if(question.at(pos)==0 && question.at(pos+1)==0 && question.at(pos+2)==QType::OPT) {
      if (!options) {
        char* ecsStart = nullptr;
        size_t ecsLen = 0;
        int res = getEDNSOption((char*)question.c_str()+pos+9, questionLen - pos - 9, EDNSOptionCode::ECS, &ecsStart, &ecsLen);
        if (res == 0 && ecsLen > 4) {
          EDNSSubnetOpts eso;
          if(getEDNSSubnetOptsFromString(ecsStart + 4, ecsLen - 4, &eso)) {
            *ednssubnet=eso;
            found = true;
          }
        }
      }
      else {
        int res = getEDNSOptions((char*)question.c_str()+pos+9, questionLen - pos - 9, *options);
        if (res == 0) {
          const auto& it = options->find(EDNSOptionCode::ECS);
          if (it != options->end() && it->second.content != nullptr && it->second.size > 0) {
            EDNSSubnetOpts eso;
            if(getEDNSSubnetOptsFromString(it->second.content, it->second.size, &eso)) {
              *ednssubnet=eso;
              found = true;
            }
          }
        }
      }
    }
  }
  return found;
}

static void handleRunningTCPQuestion(int fd, FDMultiplexer::funcparam_t& var)
{
  shared_ptr<TCPConnection> conn=any_cast<shared_ptr<TCPConnection> >(var);

  if(conn->state==TCPConnection::BYTE0) {
    ssize_t bytes=recv(conn->getFD(), conn->data, 2, 0);
    if(bytes==1)
      conn->state=TCPConnection::BYTE1;
    if(bytes==2) {
      conn->qlen=(((unsigned char)conn->data[0]) << 8)+ (unsigned char)conn->data[1];
      conn->bytesread=0;
      conn->state=TCPConnection::GETQUESTION;
    }
    if(!bytes || bytes < 0) {
      t_fdm->removeReadFD(fd);
      return;
    }
  }
  else if(conn->state==TCPConnection::BYTE1) {
    ssize_t bytes=recv(conn->getFD(), conn->data+1, 1, 0);
    if(bytes==1) {
      conn->state=TCPConnection::GETQUESTION;
      conn->qlen=(((unsigned char)conn->data[0]) << 8)+ (unsigned char)conn->data[1];
      conn->bytesread=0;
    }
    if(!bytes || bytes < 0) {
      if(g_logCommonErrors)
        L<<Logger::Error<<"TCP client "<< conn->d_remote.toString() <<" disconnected after first byte"<<endl;
      t_fdm->removeReadFD(fd);
      return;
    }
  }
  else if(conn->state==TCPConnection::GETQUESTION) {
    ssize_t bytes=recv(conn->getFD(), conn->data + conn->bytesread, conn->qlen - conn->bytesread, 0);
    if(!bytes || bytes < 0 || bytes > std::numeric_limits<std::uint16_t>::max()) {
      L<<Logger::Error<<"TCP client "<< conn->d_remote.toString() <<" disconnected while reading question body"<<endl;
      t_fdm->removeReadFD(fd);
      return;
    }
    conn->bytesread+=(uint16_t)bytes;
    if(conn->bytesread==conn->qlen) {
      t_fdm->removeReadFD(fd); // should no longer awake ourselves when there is data to read

      DNSComboWriter* dc=nullptr;
      try {
        dc=new DNSComboWriter(conn->data, conn->qlen, g_now);
      }
      catch(MOADNSException &mde) {
        g_stats.clientParseError++;
        if(g_logCommonErrors)
          L<<Logger::Error<<"Unable to parse packet from TCP client "<< conn->d_remote.toString() <<endl;
        return;
      }
      dc->d_tcpConnection = conn; // carry the torch
      dc->setSocket(conn->getFD()); // this is the only time a copy is made of the actual fd
      dc->d_tcp=true;
      dc->setRemote(&conn->d_remote);
      ComboAddress dest;
      memset(&dest, 0, sizeof(dest));
      dest.sin4.sin_family = conn->d_remote.sin4.sin_family;
      socklen_t len = dest.getSocklen();
      getsockname(conn->getFD(), (sockaddr*)&dest, &len); // if this fails, we're ok with it
      dc->setLocal(dest);
      DNSName qname;
      uint16_t qtype=0;
      uint16_t qclass=0;
      bool needECS = false;
      string requestorId;
      string deviceId;
#ifdef HAVE_PROTOBUF
      auto luaconfsLocal = g_luaconfs.getLocal();
      if (luaconfsLocal->protobufServer) {
        needECS = true;
      }
#endif

      if(needECS || (t_pdl && (t_pdl->d_gettag_ffi || t_pdl->d_gettag))) {

        try {
          std::map<uint16_t, EDNSOptionView> ednsOptions;
          dc->d_ecsParsed = true;
          dc->d_ecsFound = getQNameAndSubnet(std::string(conn->data, conn->qlen), &qname, &qtype, &qclass, &dc->d_ednssubnet, g_gettagNeedsEDNSOptions ? &ednsOptions : nullptr);

          if(t_pdl) {
            try {
              if (t_pdl->d_gettag_ffi) {
                dc->d_tag = t_pdl->gettag_ffi(conn->d_remote, dc->d_ednssubnet.source, dest, qname, qtype, &dc->d_policyTags, dc->d_data, ednsOptions, true, requestorId, deviceId, dc->d_ttlCap, dc->d_variable);
              }
              else if (t_pdl->d_gettag) {
                dc->d_tag = t_pdl->gettag(conn->d_remote, dc->d_ednssubnet.source, dest, qname, qtype, &dc->d_policyTags, dc->d_data, ednsOptions, true, requestorId, deviceId);
              }
            }
            catch(const std::exception& e)  {
              if(g_logCommonErrors)
                L<<Logger::Warning<<"Error parsing a query packet qname='"<<qname<<"' for tag determination, setting tag=0: "<<e.what()<<endl;
            }
          }
        }
        catch(const std::exception& e)
        {
          if(g_logCommonErrors)
            L<<Logger::Warning<<"Error parsing a query packet for tag determination, setting tag=0: "<<e.what()<<endl;
        }
      }
#ifdef HAVE_PROTOBUF
      if(luaconfsLocal->protobufServer || luaconfsLocal->outgoingProtobufServer) {
        dc->d_requestorId = requestorId;
        dc->d_deviceId = deviceId;
        dc->d_uuid = (*t_uuidGenerator)();
      }

      if(luaconfsLocal->protobufServer) {
        try {
          const struct dnsheader* dh = (const struct dnsheader*) conn->data;

          if (!luaconfsLocal->protobufTaggedOnly) {
            protobufLogQuery(luaconfsLocal->protobufServer, luaconfsLocal->protobufMaskV4, luaconfsLocal->protobufMaskV6, dc->d_uuid, conn->d_remote, dest, dc->d_ednssubnet.source, true, dh->id, conn->qlen, qname, qtype, qclass, dc->d_policyTags, dc->d_requestorId, dc->d_deviceId);
          }
        }
        catch(std::exception& e) {
          if(g_logCommonErrors)
            L<<Logger::Warning<<"Error parsing a TCP query packet for edns subnet: "<<e.what()<<endl;
        }
      }
#endif
      if(dc->d_mdp.d_header.qr) {
        delete dc;
        g_stats.ignoredCount++;
        L<<Logger::Error<<"Ignoring answer from TCP client "<< conn->d_remote.toString() <<" on server socket!"<<endl;
        return;
      }
      if(dc->d_mdp.d_header.opcode) {
        delete dc;
        g_stats.ignoredCount++;
        L<<Logger::Error<<"Ignoring non-query opcode from TCP client "<< conn->d_remote.toString() <<" on server socket!"<<endl;
        return;
      }
      else {
        ++g_stats.qcounter;
        ++g_stats.tcpqcounter;
        MT->makeThread(startDoResolve, dc); // deletes dc, will set state to BYTE0 again
        return;
      }
    }
  }
}

//! Handle new incoming TCP connection
static void handleNewTCPQuestion(int fd, FDMultiplexer::funcparam_t& )
{
  ComboAddress addr;
  socklen_t addrlen=sizeof(addr);
  int newsock=accept(fd, (struct sockaddr*)&addr, &addrlen);
  if(newsock>=0) {
    if(MT->numProcesses() > g_maxMThreads) {
      g_stats.overCapacityDrops++;
      try {
        closesocket(newsock);
      }
      catch(const PDNSException& e) {
        L<<Logger::Error<<"Error closing TCP socket after an over capacity drop: "<<e.reason<<endl;
      }
      return;
    }

    if(t_remotes)
      t_remotes->push_back(addr);
    if(t_allowFrom && !t_allowFrom->match(&addr)) {
      if(!g_quiet)
        L<<Logger::Error<<"["<<MT->getTid()<<"] dropping TCP query from "<<addr.toString()<<", address not matched by allow-from"<<endl;

      g_stats.unauthorizedTCP++;
      try {
        closesocket(newsock);
      }
      catch(const PDNSException& e) {
        L<<Logger::Error<<"Error closing TCP socket after an ACL drop: "<<e.reason<<endl;
      }
      return;
    }
    if(g_maxTCPPerClient && t_tcpClientCounts->count(addr) && (*t_tcpClientCounts)[addr] >= g_maxTCPPerClient) {
      g_stats.tcpClientOverflow++;
      try {
        closesocket(newsock); // don't call TCPConnection::closeAndCleanup here - did not enter it in the counts yet!
      }
      catch(const PDNSException& e) {
        L<<Logger::Error<<"Error closing TCP socket after an overflow drop: "<<e.reason<<endl;
      }
      return;
    }

    setNonBlocking(newsock);
    std::shared_ptr<TCPConnection> tc = std::make_shared<TCPConnection>(newsock, addr);
    tc->state=TCPConnection::BYTE0;

    t_fdm->addReadFD(tc->getFD(), handleRunningTCPQuestion, tc);

    struct timeval now;
    Utility::gettimeofday(&now, 0);
    t_fdm->setReadTTD(tc->getFD(), now, g_tcpTimeout);
  }
}

static string* doProcessUDPQuestion(const std::string& question, const ComboAddress& fromaddr, const ComboAddress& destaddr, struct timeval tv, int fd)
{
  gettimeofday(&g_now, 0);
  struct timeval diff = g_now - tv;
  double delta=(diff.tv_sec*1000 + diff.tv_usec/1000.0);

  if(tv.tv_sec && delta > 1000.0) {
    g_stats.tooOldDrops++;
    return 0;
  }

  ++g_stats.qcounter;
  if(fromaddr.sin4.sin_family==AF_INET6)
     g_stats.ipv6qcounter++;

  string response;
  const struct dnsheader* dh = (struct dnsheader*)question.c_str();
  unsigned int ctag=0;
  uint32_t qhash = 0;
  bool needECS = false;
  std::vector<std::string> policyTags;
  LuaContext::LuaObject data;
  string requestorId;
  string deviceId;
#ifdef HAVE_PROTOBUF
  boost::uuids::uuid uniqueId;
  auto luaconfsLocal = g_luaconfs.getLocal();
  if (luaconfsLocal->protobufServer) {
    uniqueId = (*t_uuidGenerator)();
    needECS = true;
  } else if (luaconfsLocal->outgoingProtobufServer) {
    uniqueId = (*t_uuidGenerator)();
  }
#endif
  EDNSSubnetOpts ednssubnet;
  bool ecsFound = false;
  bool ecsParsed = false;
  uint32_t ttlCap = std::numeric_limits<uint32_t>::max();
  bool variable = false;
  try {
    DNSName qname;
    uint16_t qtype=0;
    uint16_t qclass=0;
    uint32_t age;
    bool qnameParsed=false;
#ifdef MALLOC_TRACE
    /*
    static uint64_t last=0;
    if(!last)
      g_mtracer->clearAllocators();
    cout<<g_mtracer->getAllocs()-last<<" "<<g_mtracer->getNumOut()<<" -- BEGIN TRACE"<<endl;
    last=g_mtracer->getAllocs();
    cout<<g_mtracer->topAllocatorsString()<<endl;
    g_mtracer->clearAllocators();
    */
#endif

    if(needECS || (t_pdl && (t_pdl->d_gettag || t_pdl->d_gettag_ffi))) {
      try {
        std::map<uint16_t, EDNSOptionView> ednsOptions;
        ecsFound = getQNameAndSubnet(question, &qname, &qtype, &qclass, &ednssubnet, g_gettagNeedsEDNSOptions ? &ednsOptions : nullptr);
        qnameParsed = true;
        ecsParsed = true;

        if(t_pdl) {
          try {
            if (t_pdl->d_gettag_ffi) {
              ctag = t_pdl->gettag_ffi(fromaddr, ednssubnet.source, destaddr, qname, qtype, &policyTags, data, ednsOptions, false, requestorId, deviceId, ttlCap, variable);
            }
            else if (t_pdl->d_gettag) {
              ctag=t_pdl->gettag(fromaddr, ednssubnet.source, destaddr, qname, qtype, &policyTags, data, ednsOptions, false, requestorId, deviceId);
            }
          }
          catch(const std::exception& e)  {
            if(g_logCommonErrors)
              L<<Logger::Warning<<"Error parsing a query packet qname='"<<qname<<"' for tag determination, setting tag=0: "<<e.what()<<endl;
          }
        }
      }
      catch(const std::exception& e)
      {
        if(g_logCommonErrors)
          L<<Logger::Warning<<"Error parsing a query packet for tag determination, setting tag=0: "<<e.what()<<endl;
      }
    }

    bool cacheHit = false;
    RecProtoBufMessage pbMessage(DNSProtoBufMessage::DNSProtoBufMessageType::Response);
#ifdef HAVE_PROTOBUF
    if(luaconfsLocal->protobufServer) {
      if (!luaconfsLocal->protobufTaggedOnly || !policyTags.empty()) {
        protobufLogQuery(luaconfsLocal->protobufServer, luaconfsLocal->protobufMaskV4, luaconfsLocal->protobufMaskV6, uniqueId, fromaddr, destaddr, ednssubnet.source, false, dh->id, question.size(), qname, qtype, qclass, policyTags, requestorId, deviceId);
      }
    }
#endif /* HAVE_PROTOBUF */

    /* It might seem like a good idea to skip the packet cache lookup if we know that the answer is not cacheable,
       but it means that the hash would not be computed. If some script decides at a later time to mark back the answer
       as cacheable we would cache it with a wrong tag, so better safe than sorry. */
    if (qnameParsed) {
      cacheHit = (!SyncRes::s_nopacketcache && t_packetCache->getResponsePacket(ctag, question, qname, qtype, qclass, g_now.tv_sec, &response, &age, &qhash, &pbMessage));
    }
    else {
      cacheHit = (!SyncRes::s_nopacketcache && t_packetCache->getResponsePacket(ctag, question, g_now.tv_sec, &response, &age, &qhash, &pbMessage));
    }

    if (cacheHit) {
#ifdef HAVE_PROTOBUF
      if(luaconfsLocal->protobufServer && (!luaconfsLocal->protobufTaggedOnly || !pbMessage.getAppliedPolicy().empty() || !pbMessage.getPolicyTags().empty())) {
        Netmask requestorNM(fromaddr, fromaddr.sin4.sin_family == AF_INET ? luaconfsLocal->protobufMaskV4 : luaconfsLocal->protobufMaskV6);
        const ComboAddress& requestor = requestorNM.getMaskedNetwork();
        pbMessage.update(uniqueId, &requestor, &destaddr, false, dh->id);
        pbMessage.setEDNSSubnet(ednssubnet.source, ednssubnet.source.isIpv4() ? luaconfsLocal->protobufMaskV4 : luaconfsLocal->protobufMaskV6);
        pbMessage.setQueryTime(g_now.tv_sec, g_now.tv_usec);
        pbMessage.setRequestorId(requestorId);
        pbMessage.setDeviceId(deviceId);
        protobufLogResponse(luaconfsLocal->protobufServer, pbMessage);
      }
#endif /* HAVE_PROTOBUF */
      if(!g_quiet)
        L<<Logger::Notice<<t_id<< " question answered from packet cache tag="<<ctag<<" from "<<fromaddr.toString()<<endl;

      g_stats.packetCacheHits++;
      SyncRes::s_queries++;
      ageDNSPacket(response, age);
      struct msghdr msgh;
      struct iovec iov;
      char cbuf[256];
      fillMSGHdr(&msgh, &iov, cbuf, 0, (char*)response.c_str(), response.length(), const_cast<ComboAddress*>(&fromaddr));
      msgh.msg_control=NULL;

      if(g_fromtosockets.count(fd)) {
        addCMsgSrcAddr(&msgh, cbuf, &destaddr, 0);
      }
      if(sendmsg(fd, &msgh, 0) < 0 && g_logCommonErrors)
        L<<Logger::Warning<<"Sending UDP reply to client "<<fromaddr.toStringWithPort()<<" failed with: "<<strerror(errno)<<endl;

      if(response.length() >= sizeof(struct dnsheader)) {
        struct dnsheader tmpdh;
        memcpy(&tmpdh, response.c_str(), sizeof(tmpdh));
        updateResponseStats(tmpdh.rcode, fromaddr, response.length(), 0, 0);
      }
      g_stats.avgLatencyUsec=(1-1.0/g_latencyStatSize)*g_stats.avgLatencyUsec + 0.0; // we assume 0 usec
      g_stats.avgLatencyOursUsec=(1-1.0/g_latencyStatSize)*g_stats.avgLatencyOursUsec + 0.0; // we assume 0 usec
      return 0;
    }
  }
  catch(std::exception& e) {
    L<<Logger::Error<<"Error processing or aging answer packet: "<<e.what()<<endl;
    return 0;
  }

  if(t_pdl) {
    if(t_pdl->ipfilter(fromaddr, destaddr, *dh)) {
      if(!g_quiet)
        L<<Logger::Notice<<t_id<<" ["<<MT->getTid()<<"/"<<MT->numProcesses()<<"] DROPPED question from "<<fromaddr.toStringWithPort()<<" based on policy"<<endl;
      g_stats.policyDrops++;
      return 0;
    }
  }

  if(MT->numProcesses() > g_maxMThreads) {
    if(!g_quiet)
      L<<Logger::Notice<<t_id<<" ["<<MT->getTid()<<"/"<<MT->numProcesses()<<"] DROPPED question from "<<fromaddr.toStringWithPort()<<", over capacity"<<endl;

    g_stats.overCapacityDrops++;
    return 0;
  }

  DNSComboWriter* dc = new DNSComboWriter(question.c_str(), question.size(), g_now);
  dc->setSocket(fd);
  dc->d_tag=ctag;
  dc->d_qhash=qhash;
  dc->d_query = question;
  dc->setRemote(&fromaddr);
  dc->setLocal(destaddr);
  dc->d_tcp=false;
  dc->d_policyTags = policyTags;
  dc->d_data = data;
  dc->d_ecsFound = ecsFound;
  dc->d_ecsParsed = ecsParsed;
  dc->d_ednssubnet = ednssubnet;
  dc->d_ttlCap = ttlCap;
  dc->d_variable = variable;
#ifdef HAVE_PROTOBUF
  if (luaconfsLocal->protobufServer || luaconfsLocal->outgoingProtobufServer) {
    dc->d_uuid = uniqueId;
  }
  dc->d_requestorId = requestorId;
  dc->d_deviceId = deviceId;
#endif

  MT->makeThread(startDoResolve, (void*) dc); // deletes dc
  return 0;
}


static void handleNewUDPQuestion(int fd, FDMultiplexer::funcparam_t& var)
{
  ssize_t len;
  char data[1500];
  ComboAddress fromaddr;
  struct msghdr msgh;
  struct iovec iov;
  char cbuf[256];
  bool firstQuery = true;

  fromaddr.sin6.sin6_family=AF_INET6; // this makes sure fromaddr is big enough
  fillMSGHdr(&msgh, &iov, cbuf, sizeof(cbuf), data, sizeof(data), &fromaddr);

  for(;;)
  if((len=recvmsg(fd, &msgh, 0)) >= 0) {

    firstQuery = false;

    if(t_remotes)
      t_remotes->push_back(fromaddr);

    if(t_allowFrom && !t_allowFrom->match(&fromaddr)) {
      if(!g_quiet)
        L<<Logger::Error<<"["<<MT->getTid()<<"] dropping UDP query from "<<fromaddr.toString()<<", address not matched by allow-from"<<endl;

      g_stats.unauthorizedUDP++;
      return;
    }
    BOOST_STATIC_ASSERT(offsetof(sockaddr_in, sin_port) == offsetof(sockaddr_in6, sin6_port));
    if(!fromaddr.sin4.sin_port) { // also works for IPv6
     if(!g_quiet)
        L<<Logger::Error<<"["<<MT->getTid()<<"] dropping UDP query from "<<fromaddr.toStringWithPort()<<", can't deal with port 0"<<endl;

      g_stats.clientParseError++; // not quite the best place to put it, but needs to go somewhere
      return;
    }
    try {
      dnsheader* dh=(dnsheader*)data;

      if(dh->qr) {
        g_stats.ignoredCount++;
        if(g_logCommonErrors)
          L<<Logger::Error<<"Ignoring answer from "<<fromaddr.toString()<<" on server socket!"<<endl;
      }
      else if(dh->opcode) {
        g_stats.ignoredCount++;
        if(g_logCommonErrors)
          L<<Logger::Error<<"Ignoring non-query opcode "<<dh->opcode<<" from "<<fromaddr.toString()<<" on server socket!"<<endl;
      }
      else {
        string question(data, (size_t)len);
        struct timeval tv={0,0};
        HarvestTimestamp(&msgh, &tv);
        ComboAddress dest;
        memset(&dest, 0, sizeof(dest)); // this makes sure we ignore this address if not returned by recvmsg above
        auto loc = rplookup(g_listenSocketsAddresses, fd);
        if(HarvestDestinationAddress(&msgh, &dest)) {
          // but.. need to get port too
          if(loc) 
            dest.sin4.sin_port = loc->sin4.sin_port;
        }
        else {
          if(loc) {
            dest = *loc;
          }
          else {
            dest.sin4.sin_family = fromaddr.sin4.sin_family;
            socklen_t slen = dest.getSocklen();
            getsockname(fd, (sockaddr*)&dest, &slen); // if this fails, we're ok with it
          }
        }
        if(g_weDistributeQueries)
          distributeAsyncFunction(question, boost::bind(doProcessUDPQuestion, question, fromaddr, dest, tv, fd));
        else
          doProcessUDPQuestion(question, fromaddr, dest, tv, fd);
      }
    }
    catch(MOADNSException& mde) {
      g_stats.clientParseError++;
      if(g_logCommonErrors)
        L<<Logger::Error<<"Unable to parse packet from remote UDP client "<<fromaddr.toString() <<": "<<mde.what()<<endl;
    }
    catch(std::runtime_error& e) {
      g_stats.clientParseError++;
      if(g_logCommonErrors)
        L<<Logger::Error<<"Unable to parse packet from remote UDP client "<<fromaddr.toString() <<": "<<e.what()<<endl;
    }
  }
  else {
    // cerr<<t_id<<" had error: "<<stringerror()<<endl;
    if(firstQuery && errno == EAGAIN)
      g_stats.noPacketError++;

    break;
  }
}

static void makeTCPServerSockets(unsigned int threadId)
{
  int fd;
  vector<string>locals;
  stringtok(locals,::arg()["local-address"]," ,");

  if(locals.empty())
    throw PDNSException("No local address specified");

  for(vector<string>::const_iterator i=locals.begin();i!=locals.end();++i) {
    ServiceTuple st;
    st.port=::arg().asNum("local-port");
    parseService(*i, st);

    ComboAddress sin;

    memset((char *)&sin,0, sizeof(sin));
    sin.sin4.sin_family = AF_INET;
    if(!IpToU32(st.host, (uint32_t*)&sin.sin4.sin_addr.s_addr)) {
      sin.sin6.sin6_family = AF_INET6;
      if(makeIPv6sockaddr(st.host, &sin.sin6) < 0)
        throw PDNSException("Unable to resolve local address for TCP server on '"+ st.host +"'");
    }

    fd=socket(sin.sin6.sin6_family, SOCK_STREAM, 0);
    if(fd<0)
      throw PDNSException("Making a TCP server socket for resolver: "+stringerror());

    setCloseOnExec(fd);

    int tmp=1;
    if(setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &tmp, sizeof tmp)<0) {
      L<<Logger::Error<<"Setsockopt failed for TCP listening socket"<<endl;
      exit(1);
    }
    if(sin.sin6.sin6_family == AF_INET6 && setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &tmp, sizeof(tmp)) < 0) {
      L<<Logger::Error<<"Failed to set IPv6 socket to IPv6 only, continuing anyhow: "<<strerror(errno)<<endl;
    }

#ifdef TCP_DEFER_ACCEPT
    if(setsockopt(fd, SOL_TCP, TCP_DEFER_ACCEPT, &tmp, sizeof tmp) >= 0) {
      if(i==locals.begin())
        L<<Logger::Error<<"Enabled TCP data-ready filter for (slight) DoS protection"<<endl;
    }
#endif

    if( ::arg().mustDo("non-local-bind") )
        Utility::setBindAny(AF_INET, fd);

#ifdef SO_REUSEPORT
    if(g_reusePort) {
      if(setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &tmp, sizeof(tmp)) < 0)
        throw PDNSException("SO_REUSEPORT: "+stringerror());
    }
#endif

    if (::arg().asNum("tcp-fast-open") > 0) {
#ifdef TCP_FASTOPEN
      int fastOpenQueueSize = ::arg().asNum("tcp-fast-open");
      if (setsockopt(fd, IPPROTO_TCP, TCP_FASTOPEN, &fastOpenQueueSize, sizeof fastOpenQueueSize) < 0) {
        L<<Logger::Error<<"Failed to enable TCP Fast Open for listening socket: "<<strerror(errno)<<endl;
      }
#else
      L<<Logger::Warning<<"TCP Fast Open configured but not supported for listening socket"<<endl;
#endif
    }

    sin.sin4.sin_port = htons(st.port);
    socklen_t socklen=sin.sin4.sin_family==AF_INET ? sizeof(sin.sin4) : sizeof(sin.sin6);
    if (::bind(fd, (struct sockaddr *)&sin, socklen )<0)
      throw PDNSException("Binding TCP server socket for "+ st.host +": "+stringerror());

    setNonBlocking(fd);
    setSocketSendBuffer(fd, 65000);
    listen(fd, 128);
    deferredAdds[threadId].push_back(make_pair(fd, handleNewTCPQuestion));
    g_tcpListenSockets.push_back(fd);
    // we don't need to update g_listenSocketsAddresses since it doesn't work for TCP/IP:
    //  - fd is not that which we know here, but returned from accept()
    if(sin.sin4.sin_family == AF_INET)
      L<<Logger::Error<<"Listening for TCP queries on "<< sin.toString() <<":"<<st.port<<endl;
    else
      L<<Logger::Error<<"Listening for TCP queries on ["<< sin.toString() <<"]:"<<st.port<<endl;
  }
}

static void makeUDPServerSockets(unsigned int threadId)
{
  int one=1;
  vector<string>locals;
  stringtok(locals,::arg()["local-address"]," ,");

  if(locals.empty())
    throw PDNSException("No local address specified");

  for(vector<string>::const_iterator i=locals.begin();i!=locals.end();++i) {
    ServiceTuple st;
    st.port=::arg().asNum("local-port");
    parseService(*i, st);

    ComboAddress sin;

    memset(&sin, 0, sizeof(sin));
    sin.sin4.sin_family = AF_INET;
    if(!IpToU32(st.host.c_str() , (uint32_t*)&sin.sin4.sin_addr.s_addr)) {
      sin.sin6.sin6_family = AF_INET6;
      if(makeIPv6sockaddr(st.host, &sin.sin6) < 0)
        throw PDNSException("Unable to resolve local address for UDP server on '"+ st.host +"'");
    }

    int fd=socket(sin.sin4.sin_family, SOCK_DGRAM, 0);
    if(fd < 0) {
      throw PDNSException("Making a UDP server socket for resolver: "+netstringerror());
    }
    if (!setSocketTimestamps(fd))
      L<<Logger::Warning<<"Unable to enable timestamp reporting for socket"<<endl;

    if(IsAnyAddress(sin)) {
      if(sin.sin4.sin_family == AF_INET)
        if(!setsockopt(fd, IPPROTO_IP, GEN_IP_PKTINFO, &one, sizeof(one)))     // linux supports this, so why not - might fail on other systems
          g_fromtosockets.insert(fd);
#ifdef IPV6_RECVPKTINFO
      if(sin.sin4.sin_family == AF_INET6)
        if(!setsockopt(fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, &one, sizeof(one)))
          g_fromtosockets.insert(fd);
#endif
      if(sin.sin6.sin6_family == AF_INET6 && setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &one, sizeof(one)) < 0) {
        L<<Logger::Error<<"Failed to set IPv6 socket to IPv6 only, continuing anyhow: "<<strerror(errno)<<endl;
      }
    }
    if( ::arg().mustDo("non-local-bind") )
        Utility::setBindAny(AF_INET6, fd);

    setCloseOnExec(fd);

    setSocketReceiveBuffer(fd, 250000);
    sin.sin4.sin_port = htons(st.port);

  
#ifdef SO_REUSEPORT
    if(g_reusePort) {
      if(setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &one, sizeof(one)) < 0)
        throw PDNSException("SO_REUSEPORT: "+stringerror());
    }
#endif
  socklen_t socklen=sin.getSocklen();
    if (::bind(fd, (struct sockaddr *)&sin, socklen)<0)
      throw PDNSException("Resolver binding to server socket on port "+ std::to_string(st.port) +" for "+ st.host+": "+stringerror());

    setNonBlocking(fd);

    deferredAdds[threadId].push_back(make_pair(fd, handleNewUDPQuestion));
    g_listenSocketsAddresses[fd]=sin;  // this is written to only from the startup thread, not from the workers
    if(sin.sin4.sin_family == AF_INET)
      L<<Logger::Error<<"Listening for UDP queries on "<< sin.toString() <<":"<<st.port<<endl;
    else
      L<<Logger::Error<<"Listening for UDP queries on ["<< sin.toString() <<"]:"<<st.port<<endl;
  }
}

static void daemonize(void)
{
  if(fork())
    exit(0); // bye bye

  setsid();

  int i=open("/dev/null",O_RDWR); /* open stdin */
  if(i < 0)
    L<<Logger::Critical<<"Unable to open /dev/null: "<<stringerror()<<endl;
  else {
    dup2(i,0); /* stdin */
    dup2(i,1); /* stderr */
    dup2(i,2); /* stderr */
    close(i);
  }
}

static void usr1Handler(int)
{
  statsWanted=true;
}

static void usr2Handler(int)
{
  g_quiet= !g_quiet;
  SyncRes::setDefaultLogMode(g_quiet ? SyncRes::LogNone : SyncRes::Log);
  ::arg().set("quiet")=g_quiet ? "" : "no";
}

static void doStats(void)
{
  static time_t lastOutputTime;
  static uint64_t lastQueryCount;

  uint64_t cacheHits = broadcastAccFunction<uint64_t>(pleaseGetCacheHits);
  uint64_t cacheMisses = broadcastAccFunction<uint64_t>(pleaseGetCacheMisses);

  if(g_stats.qcounter && (cacheHits + cacheMisses) && SyncRes::s_queries && SyncRes::s_outqueries) {
    L<<Logger::Notice<<"stats: "<<g_stats.qcounter<<" questions, "<<
      broadcastAccFunction<uint64_t>(pleaseGetCacheSize)<< " cache entries, "<<
      broadcastAccFunction<uint64_t>(pleaseGetNegCacheSize)<<" negative entries, "<<
      (int)((cacheHits*100.0)/(cacheHits+cacheMisses))<<"% cache hits"<<endl;

    L<<Logger::Notice<<"stats: throttle map: "
      << broadcastAccFunction<uint64_t>(pleaseGetThrottleSize) <<", ns speeds: "
      << broadcastAccFunction<uint64_t>(pleaseGetNsSpeedsSize)<<endl;
    L<<Logger::Notice<<"stats: outpacket/query ratio "<<(int)(SyncRes::s_outqueries*100.0/SyncRes::s_queries)<<"%";
    L<<Logger::Notice<<", "<<(int)(SyncRes::s_throttledqueries*100.0/(SyncRes::s_outqueries+SyncRes::s_throttledqueries))<<"% throttled, "
     <<SyncRes::s_nodelegated<<" no-delegation drops"<<endl;
    L<<Logger::Notice<<"stats: "<<SyncRes::s_tcpoutqueries<<" outgoing tcp connections, "<<
      broadcastAccFunction<uint64_t>(pleaseGetConcurrentQueries)<<" queries running, "<<SyncRes::s_outgoingtimeouts<<" outgoing timeouts"<<endl;

    //L<<Logger::Notice<<"stats: "<<g_stats.ednsPingMatches<<" ping matches, "<<g_stats.ednsPingMismatches<<" mismatches, "<<
      //g_stats.noPingOutQueries<<" outqueries w/o ping, "<< g_stats.noEdnsOutQueries<<" w/o EDNS"<<endl;

    L<<Logger::Notice<<"stats: " <<  broadcastAccFunction<uint64_t>(pleaseGetPacketCacheSize) <<
    " packet cache entries, "<<(int)(100.0*broadcastAccFunction<uint64_t>(pleaseGetPacketCacheHits)/SyncRes::s_queries) << "% packet cache hits"<<endl;

    time_t now = time(0);
    if(lastOutputTime && lastQueryCount && now != lastOutputTime) {
      L<<Logger::Notice<<"stats: "<< (SyncRes::s_queries - lastQueryCount) / (now - lastOutputTime) <<" qps (average over "<< (now - lastOutputTime) << " seconds)"<<endl;
    }
    lastOutputTime = now;
    lastQueryCount = SyncRes::s_queries;
  }
  else if(statsWanted)
    L<<Logger::Notice<<"stats: no stats yet!"<<endl;

  statsWanted=false;
}

static void houseKeeping(void *)
{
  static thread_local time_t last_stat, last_rootupdate, last_prune, last_secpoll;
  static thread_local int cleanCounter=0;
  static thread_local bool s_running;  // houseKeeping can get suspended in secpoll, and be restarted, which makes us do duplicate work
  try {
    if(s_running)
      return;
    s_running=true;

    struct timeval now;
    Utility::gettimeofday(&now, 0);

    if(now.tv_sec - last_prune > (time_t)(5 + t_id)) {
      DTime dt;
      dt.setTimeval(now);
      t_RC->doPrune(g_maxCacheEntries / g_numThreads); // this function is local to a thread, so fine anyhow
      t_packetCache->doPruneTo(g_maxPacketCacheEntries / g_numWorkerThreads);

      SyncRes::pruneNegCache(g_maxCacheEntries / (g_numWorkerThreads * 10));

      if(!((cleanCounter++)%40)) {  // this is a full scan!
        time_t limit=now.tv_sec-300;
        SyncRes::pruneNSSpeeds(limit);
      }
      last_prune=time(0);
    }

    if(now.tv_sec - last_rootupdate > 7200) {
      int res = SyncRes::getRootNS(g_now, nullptr);
      if (!res)
        last_rootupdate=now.tv_sec;
    }

    if(!t_id) {
      if(g_statisticsInterval > 0 && now.tv_sec - last_stat >= g_statisticsInterval) {
        doStats();
        last_stat=time(0);
      }

      if(now.tv_sec - last_secpoll >= 3600) {
        try {
          doSecPoll(&last_secpoll);
        }
        catch(std::exception& e)
        {
          L<<Logger::Error<<"Exception while performing security poll: "<<e.what()<<endl;
        }
        catch(PDNSException& e)
        {
          L<<Logger::Error<<"Exception while performing security poll: "<<e.reason<<endl;
        }
        catch(ImmediateServFailException &e)
        {
          L<<Logger::Error<<"Exception while performing security poll: "<<e.reason<<endl;
        }
        catch(...)
        {
          L<<Logger::Error<<"Exception while performing security poll"<<endl;
        }

      }
    }
    s_running=false;
  }
  catch(PDNSException& ae)
    {
      s_running=false;
      L<<Logger::Error<<"Fatal error in housekeeping thread: "<<ae.reason<<endl;
      throw;
    }
}

static void makeThreadPipes()
{
  for(unsigned int n=0; n < g_numThreads; ++n) {
    struct ThreadPipeSet tps;
    int fd[2];
    if(pipe(fd) < 0)
      unixDie("Creating pipe for inter-thread communications");

    tps.readToThread = fd[0];
    tps.writeToThread = fd[1];

    if(pipe(fd) < 0)
      unixDie("Creating pipe for inter-thread communications");
    tps.readFromThread = fd[0];
    tps.writeFromThread = fd[1];

    g_pipes.push_back(tps);
  }
}

struct ThreadMSG
{
  pipefunc_t func;
  bool wantAnswer;
};

void broadcastFunction(const pipefunc_t& func, bool skipSelf)
{
  unsigned int n = 0;
  for(ThreadPipeSet& tps : g_pipes)
  {
    if(n++ == t_id) {
      if(!skipSelf)
        func(); // don't write to ourselves!
      continue;
    }

    ThreadMSG* tmsg = new ThreadMSG();
    tmsg->func = func;
    tmsg->wantAnswer = true;
    if(write(tps.writeToThread, &tmsg, sizeof(tmsg)) != sizeof(tmsg)) {
      delete tmsg;
      unixDie("write to thread pipe returned wrong size or error");
    }

    string* resp;
    if(read(tps.readFromThread, &resp, sizeof(resp)) != sizeof(resp))
      unixDie("read from thread pipe returned wrong size or error");

    if(resp) {
//      cerr <<"got response: " << *resp << endl;
      delete resp;
    }
  }
}

void distributeAsyncFunction(const string& packet, const pipefunc_t& func)
{
  unsigned int hash = hashQuestion(packet.c_str(), packet.length(), g_disthashseed);
  unsigned int target = 1 + (hash % (g_pipes.size()-1));

  if(target == t_id) {
    func();
    return;
  }
  ThreadPipeSet& tps = g_pipes[target];
  ThreadMSG* tmsg = new ThreadMSG();
  tmsg->func = func;
  tmsg->wantAnswer = false;

  if(write(tps.writeToThread, &tmsg, sizeof(tmsg)) != sizeof(tmsg)) {
    delete tmsg;
    unixDie("write to thread pipe returned wrong size or error");
  }
}

static void handlePipeRequest(int fd, FDMultiplexer::funcparam_t& var)
{
  ThreadMSG* tmsg = nullptr;

  if(read(fd, &tmsg, sizeof(tmsg)) != sizeof(tmsg)) { // fd == readToThread
    unixDie("read from thread pipe returned wrong size or error");
  }

  void *resp=0;
  try {
    resp = tmsg->func();
  }
  catch(std::exception& e) {
    if(g_logCommonErrors)
      L<<Logger::Error<<"PIPE function we executed created exception: "<<e.what()<<endl; // but what if they wanted an answer.. we send 0
  }
  catch(PDNSException& e) {
    if(g_logCommonErrors)
      L<<Logger::Error<<"PIPE function we executed created PDNS exception: "<<e.reason<<endl; // but what if they wanted an answer.. we send 0
  }
  if(tmsg->wantAnswer) {
    if(write(g_pipes[t_id].writeFromThread, &resp, sizeof(resp)) != sizeof(resp)) {
      delete tmsg;
      unixDie("write to thread pipe returned wrong size or error");
    }
  }

  delete tmsg;
}

template<class T> void *voider(const boost::function<T*()>& func)
{
  return func();
}

vector<ComboAddress>& operator+=(vector<ComboAddress>&a, const vector<ComboAddress>& b)
{
  a.insert(a.end(), b.begin(), b.end());
  return a;
}

vector<pair<string, uint16_t> >& operator+=(vector<pair<string, uint16_t> >&a, const vector<pair<string, uint16_t> >& b)
{
  a.insert(a.end(), b.begin(), b.end());
  return a;
}

vector<pair<DNSName, uint16_t> >& operator+=(vector<pair<DNSName, uint16_t> >&a, const vector<pair<DNSName, uint16_t> >& b)
{
  a.insert(a.end(), b.begin(), b.end());
  return a;
}


template<class T> T broadcastAccFunction(const boost::function<T*()>& func, bool skipSelf)
{
  unsigned int n = 0;
  T ret=T();
  for(ThreadPipeSet& tps : g_pipes)
  {
    if(n++ == t_id) {
      if(!skipSelf) {
        T* resp = (T*)func(); // don't write to ourselves!
        if(resp) {
          //~ cerr <<"got direct: " << *resp << endl;
          ret += *resp;
          delete resp;
        }
      }
      continue;
    }

    ThreadMSG* tmsg = new ThreadMSG();
    tmsg->func = boost::bind(voider<T>, func);
    tmsg->wantAnswer = true;

    if(write(tps.writeToThread, &tmsg, sizeof(tmsg)) != sizeof(tmsg)) {
      delete tmsg;
      unixDie("write to thread pipe returned wrong size or error");
    }

    T* resp;
    if(read(tps.readFromThread, &resp, sizeof(resp)) != sizeof(resp))
      unixDie("read from thread pipe returned wrong size or error");

    if(resp) {
      //~ cerr <<"got response: " << *resp << endl;
      ret += *resp;
      delete resp;
    }
  }
  return ret;
}

template string broadcastAccFunction(const boost::function<string*()>& fun, bool skipSelf); // explicit instantiation
template uint64_t broadcastAccFunction(const boost::function<uint64_t*()>& fun, bool skipSelf); // explicit instantiation
template vector<ComboAddress> broadcastAccFunction(const boost::function<vector<ComboAddress> *()>& fun, bool skipSelf); // explicit instantiation
template vector<pair<DNSName,uint16_t> > broadcastAccFunction(const boost::function<vector<pair<DNSName, uint16_t> > *()>& fun, bool skipSelf); // explicit instantiation

static void handleRCC(int fd, FDMultiplexer::funcparam_t& var)
{
  string remote;
  string msg=s_rcc.recv(&remote);
  RecursorControlParser rcp;
  RecursorControlParser::func_t* command;

  string answer=rcp.getAnswer(msg, &command);

  // If we are inside a chroot, we need to strip
  if (!arg()["chroot"].empty()) {
    size_t len = arg()["chroot"].length();
    remote = remote.substr(len);
  }

  try {
    s_rcc.send(answer, &remote);
    command();
  }
  catch(std::exception& e) {
    L<<Logger::Error<<"Error dealing with control socket request: "<<e.what()<<endl;
  }
  catch(PDNSException& ae) {
    L<<Logger::Error<<"Error dealing with control socket request: "<<ae.reason<<endl;
  }
}

static void handleTCPClientReadable(int fd, FDMultiplexer::funcparam_t& var)
{
  PacketID* pident=any_cast<PacketID>(&var);
  //  cerr<<"handleTCPClientReadable called for fd "<<fd<<", pident->inNeeded: "<<pident->inNeeded<<", "<<pident->sock->getHandle()<<endl;

  shared_array<char> buffer(new char[pident->inNeeded]);

  ssize_t ret=recv(fd, buffer.get(), pident->inNeeded,0);
  if(ret > 0) {
    pident->inMSG.append(&buffer[0], &buffer[ret]);
    pident->inNeeded-=(size_t)ret;
    if(!pident->inNeeded || pident->inIncompleteOkay) {
      //      cerr<<"Got entire load of "<<pident->inMSG.size()<<" bytes"<<endl;
      PacketID pid=*pident;
      string msg=pident->inMSG;

      t_fdm->removeReadFD(fd);
      MT->sendEvent(pid, &msg);
    }
    else {
      //      cerr<<"Still have "<<pident->inNeeded<<" left to go"<<endl;
    }
  }
  else {
    PacketID tmp=*pident;
    t_fdm->removeReadFD(fd); // pident might now be invalid (it isn't, but still)
    string empty;
    MT->sendEvent(tmp, &empty); // this conveys error status
  }
}

static void handleTCPClientWritable(int fd, FDMultiplexer::funcparam_t& var)
{
  PacketID* pid=any_cast<PacketID>(&var);
  ssize_t ret=send(fd, pid->outMSG.c_str() + pid->outPos, pid->outMSG.size() - pid->outPos,0);
  if(ret > 0) {
    pid->outPos+=(ssize_t)ret;
    if(pid->outPos==pid->outMSG.size()) {
      PacketID tmp=*pid;
      t_fdm->removeWriteFD(fd);
      MT->sendEvent(tmp, &tmp.outMSG);  // send back what we sent to convey everything is ok
    }
  }
  else {  // error or EOF
    PacketID tmp(*pid);
    t_fdm->removeWriteFD(fd);
    string sent;
    MT->sendEvent(tmp, &sent);         // we convey error status by sending empty string
  }
}

// resend event to everybody chained onto it
static void doResends(MT_t::waiters_t::iterator& iter, PacketID resend, const string& content)
{
  if(iter->key.chain.empty())
    return;
  //  cerr<<"doResends called!\n";
  for(PacketID::chain_t::iterator i=iter->key.chain.begin(); i != iter->key.chain.end() ; ++i) {
    resend.fd=-1;
    resend.id=*i;
    //    cerr<<"\tResending "<<content.size()<<" bytes for fd="<<resend.fd<<" and id="<<resend.id<<endl;

    MT->sendEvent(resend, &content);
    g_stats.chainResends++;
  }
}

static void handleUDPServerResponse(int fd, FDMultiplexer::funcparam_t& var)
{
  PacketID pid=any_cast<PacketID>(var);
  ssize_t len;
  char data[g_outgoingEDNSBufsize];
  ComboAddress fromaddr;
  socklen_t addrlen=sizeof(fromaddr);

  len=recvfrom(fd, data, sizeof(data), 0, (sockaddr *)&fromaddr, &addrlen);

  if(len < (ssize_t) sizeof(dnsheader)) {
    if(len < 0)
      ; //      cerr<<"Error on fd "<<fd<<": "<<stringerror()<<"\n";
    else {
      g_stats.serverParseError++;
      if(g_logCommonErrors)
        L<<Logger::Error<<"Unable to parse packet from remote UDP server "<< fromaddr.toString() <<
          ": packet smaller than DNS header"<<endl;
    }

    t_udpclientsocks->returnSocket(fd);
    string empty;

    MT_t::waiters_t::iterator iter=MT->d_waiters.find(pid);
    if(iter != MT->d_waiters.end())
      doResends(iter, pid, empty);

    MT->sendEvent(pid, &empty); // this denotes error (does lookup again.. at least L1 will be hot)
    return;
  }

  dnsheader dh;
  memcpy(&dh, data, sizeof(dh));

  PacketID pident;
  pident.remote=fromaddr;
  pident.id=dh.id;
  pident.fd=fd;

  if(!dh.qr && g_logCommonErrors) {
    L<<Logger::Notice<<"Not taking data from question on outgoing socket from "<< fromaddr.toStringWithPort()  <<endl;
  }

  if(!dh.qdcount || // UPC, Nominum, very old BIND on FormErr, NSD
     !dh.qr) {      // one weird server
    pident.domain.clear();
    pident.type = 0;
  }
  else {
    try {
      if(len > 12)
        pident.domain=DNSName(data, len, 12, false, &pident.type); // don't copy this from above - we need to do the actual read
    }
    catch(std::exception& e) {
      g_stats.serverParseError++; // won't be fed to lwres.cc, so we have to increment
      L<<Logger::Warning<<"Error in packet from remote nameserver "<< fromaddr.toStringWithPort() << ": "<<e.what() << endl;
      return;
    }
  }
  string packet;
  packet.assign(data, len);

  MT_t::waiters_t::iterator iter=MT->d_waiters.find(pident);
  if(iter != MT->d_waiters.end()) {
    doResends(iter, pident, packet);
  }

retryWithName:

  if(!MT->sendEvent(pident, &packet)) {
    // we do a full scan for outstanding queries on unexpected answers. not too bad since we only accept them on the right port number, which is hard enough to guess
    for(MT_t::waiters_t::iterator mthread=MT->d_waiters.begin(); mthread!=MT->d_waiters.end(); ++mthread) {
      if(pident.fd==mthread->key.fd && mthread->key.remote==pident.remote &&  mthread->key.type == pident.type &&
         pident.domain == mthread->key.domain) {
        mthread->key.nearMisses++;
      }

      // be a bit paranoid here since we're weakening our matching
      if(pident.domain.empty() && !mthread->key.domain.empty() && !pident.type && mthread->key.type &&
         pident.id  == mthread->key.id && mthread->key.remote == pident.remote) {
        // cerr<<"Empty response, rest matches though, sending to a waiter"<<endl;
        pident.domain = mthread->key.domain;
        pident.type = mthread->key.type;
        goto retryWithName; // note that this only passes on an error, lwres will still reject the packet
      }
    }
    g_stats.unexpectedCount++; // if we made it here, it really is an unexpected answer
    if(g_logCommonErrors) {
      L<<Logger::Warning<<"Discarding unexpected packet from "<<fromaddr.toStringWithPort()<<": "<< (pident.domain.empty() ? "<empty>" : pident.domain.toString())<<", "<<pident.type<<", "<<MT->d_waiters.size()<<" waiters"<<endl;
    }
  }
  else if(fd >= 0) {
    t_udpclientsocks->returnSocket(fd);
  }
}

FDMultiplexer* getMultiplexer()
{
  FDMultiplexer* ret;
  for(const auto& i : FDMultiplexer::getMultiplexerMap()) {
    try {
      ret=i.second();
      return ret;
    }
    catch(FDMultiplexerException &fe) {
      L<<Logger::Error<<"Non-fatal error initializing possible multiplexer ("<<fe.what()<<"), falling back"<<endl;
    }
    catch(...) {
      L<<Logger::Error<<"Non-fatal error initializing possible multiplexer"<<endl;
    }
  }
  L<<Logger::Error<<"No working multiplexer found!"<<endl;
  exit(1);
}


static string* doReloadLuaScript()
{
  string fname= ::arg()["lua-dns-script"];
  try {
    if(fname.empty()) {
      t_pdl.reset();
      L<<Logger::Error<<t_id<<" Unloaded current lua script"<<endl;
      return new string("unloaded\n");
    }
    else {
      t_pdl = std::make_shared<RecursorLua4>(fname);
    }
  }
  catch(std::exception& e) {
    L<<Logger::Error<<t_id<<" Retaining current script, error from '"<<fname<<"': "<< e.what() <<endl;
    return new string("retaining current script, error from '"+fname+"': "+e.what()+"\n");
  }

  L<<Logger::Warning<<t_id<<" (Re)loaded lua script from '"<<fname<<"'"<<endl;
  return new string("(re)loaded '"+fname+"'\n");
}

string doQueueReloadLuaScript(vector<string>::const_iterator begin, vector<string>::const_iterator end)
{
  if(begin != end)
    ::arg().set("lua-dns-script") = *begin;

  return broadcastAccFunction<string>(doReloadLuaScript);
}

static string* pleaseUseNewTraceRegex(const std::string& newRegex)
try
{
  if(newRegex.empty()) {
    t_traceRegex.reset();
    return new string("unset\n");
  }
  else {
    t_traceRegex = std::make_shared<Regex>(newRegex);
    return new string("ok\n");
  }
}
catch(PDNSException& ae)
{
  return new string(ae.reason+"\n");
}

string doTraceRegex(vector<string>::const_iterator begin, vector<string>::const_iterator end)
{
  return broadcastAccFunction<string>(boost::bind(pleaseUseNewTraceRegex, begin!=end ? *begin : ""));
}

static void checkLinuxIPv6Limits()
{
#ifdef __linux__
  string line;
  if(readFileIfThere("/proc/sys/net/ipv6/route/max_size", &line)) {
    int lim=std::stoi(line);
    if(lim < 16384) {
      L<<Logger::Error<<"If using IPv6, please raise sysctl net.ipv6.route.max_size, currently set to "<<lim<<" which is < 16384"<<endl;
    }
  }
#endif
}
static void checkOrFixFDS()
{
  unsigned int availFDs=getFilenumLimit(); 
  unsigned int wantFDs = g_maxMThreads * g_numWorkerThreads +25; // even healthier margin then before

  if(wantFDs > availFDs) {
    unsigned int hardlimit= getFilenumLimit(true);
    if(hardlimit >= wantFDs) {
      setFilenumLimit(wantFDs);
      L<<Logger::Warning<<"Raised soft limit on number of filedescriptors to "<<wantFDs<<" to match max-mthreads and threads settings"<<endl;
    }
    else {
      int newval = (hardlimit - 25) / g_numWorkerThreads;
      L<<Logger::Warning<<"Insufficient number of filedescriptors available for max-mthreads*threads setting! ("<<hardlimit<<" < "<<wantFDs<<"), reducing max-mthreads to "<<newval<<endl;
      g_maxMThreads = newval;
      setFilenumLimit(hardlimit);
    }
  }
}

static void* recursorThread(void*);

static void* pleaseSupplantACLs(std::shared_ptr<NetmaskGroup> ng)
{
  t_allowFrom = ng;
  return nullptr;
}

int g_argc;
char** g_argv;

void parseACLs()
{
  static bool l_initialized;

  if(l_initialized) { // only reload configuration file on second call
    string configname=::arg()["config-dir"]+"/recursor.conf";
    if(::arg()["config-name"]!="") {
      configname=::arg()["config-dir"]+"/recursor-"+::arg()["config-name"]+".conf";
    }
    cleanSlashes(configname);

    if(!::arg().preParseFile(configname.c_str(), "allow-from-file"))
      throw runtime_error("Unable to re-parse configuration file '"+configname+"'");
    ::arg().preParseFile(configname.c_str(), "allow-from", LOCAL_NETS);
    ::arg().preParseFile(configname.c_str(), "include-dir");
    ::arg().preParse(g_argc, g_argv, "include-dir");

    // then process includes
    std::vector<std::string> extraConfigs;
    ::arg().gatherIncludes(extraConfigs);

    for(const std::string& fn : extraConfigs) {
      if(!::arg().preParseFile(fn.c_str(), "allow-from-file", ::arg()["allow-from-file"]))
        throw runtime_error("Unable to re-parse configuration file include '"+fn+"'");
      if(!::arg().preParseFile(fn.c_str(), "allow-from", ::arg()["allow-from"]))
        throw runtime_error("Unable to re-parse configuration file include '"+fn+"'");
    }

    ::arg().preParse(g_argc, g_argv, "allow-from-file");
    ::arg().preParse(g_argc, g_argv, "allow-from");
  }

  std::shared_ptr<NetmaskGroup> oldAllowFrom = t_allowFrom;
  std::shared_ptr<NetmaskGroup> allowFrom = std::make_shared<NetmaskGroup>();

  if(!::arg()["allow-from-file"].empty()) {
    string line;
    ifstream ifs(::arg()["allow-from-file"].c_str());
    if(!ifs) {
      throw runtime_error("Could not open '"+::arg()["allow-from-file"]+"': "+stringerror());
    }

    string::size_type pos;
    while(getline(ifs,line)) {
      pos=line.find('#');
      if(pos!=string::npos)
        line.resize(pos);
      trim(line);
      if(line.empty())
        continue;

      allowFrom->addMask(line);
    }
    L<<Logger::Warning<<"Done parsing " << allowFrom->size() <<" allow-from ranges from file '"<<::arg()["allow-from-file"]<<"' - overriding 'allow-from' setting"<<endl;
  }
  else if(!::arg()["allow-from"].empty()) {
    vector<string> ips;
    stringtok(ips, ::arg()["allow-from"], ", ");

    L<<Logger::Warning<<"Only allowing queries from: ";
    for(vector<string>::const_iterator i = ips.begin(); i!= ips.end(); ++i) {
      allowFrom->addMask(*i);
      if(i!=ips.begin())
        L<<Logger::Warning<<", ";
      L<<Logger::Warning<<*i;
    }
    L<<Logger::Warning<<endl;
  }
  else {
    if(::arg()["local-address"]!="127.0.0.1" && ::arg().asNum("local-port")==53)
      L<<Logger::Error<<"WARNING: Allowing queries from all IP addresses - this can be a security risk!"<<endl;
    allowFrom = nullptr;
  }

  g_initialAllowFrom = allowFrom;
  broadcastFunction(boost::bind(pleaseSupplantACLs, allowFrom));
  oldAllowFrom = nullptr;

  l_initialized = true;
}


static void setupDelegationOnly()
{
  vector<string> parts;
  stringtok(parts, ::arg()["delegation-only"], ", \t");
  for(const auto& p : parts) {
    SyncRes::addDelegationOnly(DNSName(p));
  }
}

static std::map<unsigned int, std::set<int> > parseCPUMap()
{
  std::map<unsigned int, std::set<int> > result;

  const std::string value = ::arg()["cpu-map"];

  if (!value.empty() && !isSettingThreadCPUAffinitySupported()) {
    L<<Logger::Warning<<"CPU mapping requested but not supported, skipping"<<endl;
    return result;
  }

  std::vector<std::string> parts;

  stringtok(parts, value, " \t");

  for(const auto& part : parts) {
    if (part.find('=') == string::npos)
      continue;

    try {
      auto headers = splitField(part, '=');
      trim(headers.first);
      trim(headers.second);

      unsigned int threadId = pdns_stou(headers.first);
      std::vector<std::string> cpus;

      stringtok(cpus, headers.second, ",");

      for(const auto& cpu : cpus) {
        int cpuId = std::stoi(cpu);

        result[threadId].insert(cpuId);
      }
    }
    catch(const std::exception& e) {
      L<<Logger::Error<<"Error parsing cpu-map entry '"<<part<<"': "<<e.what()<<endl;
    }
  }

  return result;
}

static void setCPUMap(const std::map<unsigned int, std::set<int> >& cpusMap, unsigned int n, pthread_t tid)
{
  const auto& cpuMapping = cpusMap.find(n);
  if (cpuMapping != cpusMap.cend()) {
    int rc = mapThreadToCPUList(tid, cpuMapping->second);
    if (rc == 0) {
      L<<Logger::Info<<"CPU affinity for worker "<<n<<" has been set to CPU map:";
      for (const auto cpu : cpuMapping->second) {
        L<<Logger::Info<<" "<<cpu;
      }
      L<<Logger::Info<<endl;
    }
    else {
      L<<Logger::Warning<<"Error setting CPU affinity for worker "<<n<<" to CPU map:";
      for (const auto cpu : cpuMapping->second) {
        L<<Logger::Info<<" "<<cpu;
      }
      L<<Logger::Info<<strerror(rc)<<endl;
    }
  }
}

static int serviceMain(int argc, char*argv[])
{
  L.setName(s_programname);
  L.disableSyslog(::arg().mustDo("disable-syslog"));
  L.setTimestamps(::arg().mustDo("log-timestamp"));

  if(!::arg()["logging-facility"].empty()) {
    int val=logFacilityToLOG(::arg().asNum("logging-facility") );
    if(val >= 0)
      theL().setFacility(val);
    else
      L<<Logger::Error<<"Unknown logging facility "<<::arg().asNum("logging-facility") <<endl;
  }

  showProductVersion();
  seedRandom(::arg()["entropy-source"]);

  g_disthashseed=dns_random(0xffffffff);

  checkLinuxIPv6Limits();
  try {
    vector<string> addrs;
    if(!::arg()["query-local-address6"].empty()) {
      SyncRes::s_doIPv6=true;
      L<<Logger::Warning<<"Enabling IPv6 transport for outgoing queries"<<endl;

      stringtok(addrs, ::arg()["query-local-address6"], ", ;");
      for(const string& addr : addrs) {
        g_localQueryAddresses6.push_back(ComboAddress(addr));
      }
    }
    else {
      L<<Logger::Warning<<"NOT using IPv6 for outgoing queries - set 'query-local-address6=::' to enable"<<endl;
    }
    addrs.clear();
    stringtok(addrs, ::arg()["query-local-address"], ", ;");
    for(const string& addr : addrs) {
      g_localQueryAddresses4.push_back(ComboAddress(addr));
    }
  }
  catch(std::exception& e) {
    L<<Logger::Error<<"Assigning local query addresses: "<<e.what();
    exit(99);
  }

  // keep this ABOVE loadRecursorLuaConfig!
  if(::arg()["dnssec"]=="off")
    g_dnssecmode=DNSSECMode::Off;
  else if(::arg()["dnssec"]=="process-no-validate")
    g_dnssecmode=DNSSECMode::ProcessNoValidate;
  else if(::arg()["dnssec"]=="process")
    g_dnssecmode=DNSSECMode::Process;
  else if(::arg()["dnssec"]=="validate")
    g_dnssecmode=DNSSECMode::ValidateAll;
  else if(::arg()["dnssec"]=="log-fail")
    g_dnssecmode=DNSSECMode::ValidateForLog;
  else {
    L<<Logger::Error<<"Unknown DNSSEC mode "<<::arg()["dnssec"]<<endl;
    exit(1);
  }

  g_dnssecLogBogus = ::arg().mustDo("dnssec-log-bogus");
  g_maxNSEC3Iterations = ::arg().asNum("nsec3-max-iterations");

  g_maxCacheEntries = ::arg().asNum("max-cache-entries");
  g_maxPacketCacheEntries = ::arg().asNum("max-packetcache-entries");
  
  try {
    loadRecursorLuaConfig(::arg()["lua-config-file"], ::arg().mustDo("daemon"));
  }
  catch (PDNSException &e) {
    L<<Logger::Error<<"Cannot load Lua configuration: "<<e.reason<<endl;
    exit(1);
  }

  parseACLs();
  sortPublicSuffixList();

  if(!::arg()["dont-query"].empty()) {
    vector<string> ips;
    stringtok(ips, ::arg()["dont-query"], ", ");
    ips.push_back("0.0.0.0");
    ips.push_back("::");

    L<<Logger::Warning<<"Will not send queries to: ";
    for(vector<string>::const_iterator i = ips.begin(); i!= ips.end(); ++i) {
      SyncRes::addDontQuery(*i);
      if(i!=ips.begin())
        L<<Logger::Warning<<", ";
      L<<Logger::Warning<<*i;
    }
    L<<Logger::Warning<<endl;
  }

  g_quiet=::arg().mustDo("quiet");

  g_weDistributeQueries = ::arg().mustDo("pdns-distributes-queries");
  if(g_weDistributeQueries) {
      L<<Logger::Warning<<"PowerDNS Recursor itself will distribute queries over threads"<<endl;
  }

  setupDelegationOnly();
  g_outgoingEDNSBufsize=::arg().asNum("edns-outgoing-bufsize");

  if(::arg()["trace"]=="fail") {
    SyncRes::setDefaultLogMode(SyncRes::Store);
  }
  else if(::arg().mustDo("trace")) {
    SyncRes::setDefaultLogMode(SyncRes::Log);
    ::arg().set("quiet")="no";
    g_quiet=false;
    g_dnssecLOG=true;
  }

  SyncRes::s_minimumTTL = ::arg().asNum("minimum-ttl-override");

  SyncRes::s_nopacketcache = ::arg().mustDo("disable-packetcache");

  SyncRes::s_maxnegttl=::arg().asNum("max-negative-ttl");
  SyncRes::s_maxcachettl=max(::arg().asNum("max-cache-ttl"), 15);
  SyncRes::s_packetcachettl=::arg().asNum("packetcache-ttl");
  // Cap the packetcache-servfail-ttl to the packetcache-ttl
  uint32_t packetCacheServFailTTL = ::arg().asNum("packetcache-servfail-ttl");
  SyncRes::s_packetcacheservfailttl=(packetCacheServFailTTL > SyncRes::s_packetcachettl) ? SyncRes::s_packetcachettl : packetCacheServFailTTL;
  SyncRes::s_serverdownmaxfails=::arg().asNum("server-down-max-fails");
  SyncRes::s_serverdownthrottletime=::arg().asNum("server-down-throttle-time");
  SyncRes::s_serverID=::arg()["server-id"];
  SyncRes::s_maxqperq=::arg().asNum("max-qperq");
  SyncRes::s_maxtotusec=1000*::arg().asNum("max-total-msec");
  SyncRes::s_maxdepth=::arg().asNum("max-recursion-depth");
  SyncRes::s_rootNXTrust = ::arg().mustDo( "root-nx-trust");
  if(SyncRes::s_serverID.empty()) {
    char tmp[128];
    gethostname(tmp, sizeof(tmp)-1);
    SyncRes::s_serverID=tmp;
  }

  SyncRes::s_ecsipv4limit = ::arg().asNum("ecs-ipv4-bits");
  SyncRes::s_ecsipv6limit = ::arg().asNum("ecs-ipv6-bits");

  if (!::arg().isEmpty("ecs-scope-zero-address")) {
    ComboAddress scopeZero(::arg()["ecs-scope-zero-address"]);
    SyncRes::setECSScopeZeroAddress(Netmask(scopeZero, scopeZero.isIPv4() ? 32 : 128));
  }
  else {
    bool found = false;
    for (const auto& addr : g_localQueryAddresses4) {
      if (!IsAnyAddress(addr)) {
        SyncRes::setECSScopeZeroAddress(Netmask(addr, 32));
        found = true;
        break;
      }
    }
    if (!found) {
      for (const auto& addr : g_localQueryAddresses6) {
        if (!IsAnyAddress(addr)) {
          SyncRes::setECSScopeZeroAddress(Netmask(addr, 128));
          found = true;
          break;
        }
      }
      if (!found) {
        SyncRes::setECSScopeZeroAddress(Netmask("127.0.0.1/32"));
      }
    }
  }

  g_networkTimeoutMsec = ::arg().asNum("network-timeout");

  g_initialDomainMap = parseAuthAndForwards();

  g_latencyStatSize=::arg().asNum("latency-statistic-size");

  g_logCommonErrors=::arg().mustDo("log-common-errors");
  g_logRPZChanges = ::arg().mustDo("log-rpz-changes");

  g_anyToTcp = ::arg().mustDo("any-to-tcp");
  g_udpTruncationThreshold = ::arg().asNum("udp-truncation-threshold");

  g_lowercaseOutgoing = ::arg().mustDo("lowercase-outgoing");

  g_numWorkerThreads = ::arg().asNum("threads");
  if (g_numWorkerThreads < 1) {
    L<<Logger::Warning<<"Asked to run with 0 threads, raising to 1 instead"<<endl;
    g_numWorkerThreads = 1;
  }

  g_numThreads = g_numWorkerThreads + g_weDistributeQueries;
  g_maxMThreads = ::arg().asNum("max-mthreads");

  g_gettagNeedsEDNSOptions = ::arg().mustDo("gettag-needs-edns-options");

  g_statisticsInterval = ::arg().asNum("statistics-interval");

#ifdef SO_REUSEPORT
  g_reusePort = ::arg().mustDo("reuseport");
#endif

  g_useOneSocketPerThread = (!g_weDistributeQueries && g_reusePort);

  if (g_useOneSocketPerThread) {
    for (unsigned int threadId = 0; threadId < g_numWorkerThreads; threadId++) {
      makeUDPServerSockets(threadId);
      makeTCPServerSockets(threadId);
    }
  }
  else {
    makeUDPServerSockets(0);
    makeTCPServerSockets(0);
  }

  SyncRes::parseEDNSSubnetWhitelist(::arg()["edns-subnet-whitelist"]);
  g_useIncomingECS = ::arg().mustDo("use-incoming-edns-subnet");

  int forks;
  for(forks = 0; forks < ::arg().asNum("processes") - 1; ++forks) {
    if(!fork()) // we are child
      break;
  }

  if(::arg().mustDo("daemon")) {
    L<<Logger::Warning<<"Calling daemonize, going to background"<<endl;
    L.toConsole(Logger::Critical);
    daemonize();
    loadRecursorLuaConfig(::arg()["lua-config-file"], false);
  }
  signal(SIGUSR1,usr1Handler);
  signal(SIGUSR2,usr2Handler);
  signal(SIGPIPE,SIG_IGN);

  checkOrFixFDS();

#ifdef HAVE_LIBSODIUM
  if (sodium_init() == -1) {
    L<<Logger::Error<<"Unable to initialize sodium crypto library"<<endl;
    exit(99);
  }
#endif

  openssl_thread_setup();
  openssl_seed();

  int newgid=0;
  if(!::arg()["setgid"].empty())
    newgid=Utility::makeGidNumeric(::arg()["setgid"]);
  int newuid=0;
  if(!::arg()["setuid"].empty())
    newuid=Utility::makeUidNumeric(::arg()["setuid"]);

  Utility::dropGroupPrivs(newuid, newgid);

  if (!::arg()["chroot"].empty()) {
#ifdef HAVE_SYSTEMD
     char *ns;
     ns = getenv("NOTIFY_SOCKET");
     if (ns != nullptr) {
       L<<Logger::Error<<"Unable to chroot when running from systemd. Please disable chroot= or set the 'Type' for this service to 'simple'"<<endl;
       exit(1);
     }
#endif
    if (chroot(::arg()["chroot"].c_str())<0 || chdir("/") < 0) {
      L<<Logger::Error<<"Unable to chroot to '"+::arg()["chroot"]+"': "<<strerror (errno)<<", exiting"<<endl;
      exit(1);
    }
    else
      L<<Logger::Error<<"Chrooted to '"<<::arg()["chroot"]<<"'"<<endl;
  }

  s_pidfname=::arg()["socket-dir"]+"/"+s_programname+".pid";
  if(!s_pidfname.empty())
    unlink(s_pidfname.c_str()); // remove possible old pid file
  writePid();

  makeControlChannelSocket( ::arg().asNum("processes") > 1 ? forks : -1);

  Utility::dropUserPrivs(newuid);

  makeThreadPipes();

  g_tcpTimeout=::arg().asNum("client-tcp-timeout");
  g_maxTCPPerClient=::arg().asNum("max-tcp-per-client");
  g_tcpMaxQueriesPerConn=::arg().asNum("max-tcp-queries-per-connection");

  if (::arg().mustDo("snmp-agent")) {
    g_snmpAgent = std::make_shared<RecursorSNMPAgent>("recursor", ::arg()["snmp-master-socket"]);
    g_snmpAgent->run();
  }

  const auto cpusMap = parseCPUMap();
  if(g_numThreads == 1) {
    L<<Logger::Warning<<"Operating unthreaded"<<endl;
#ifdef HAVE_SYSTEMD
    sd_notify(0, "READY=1");
#endif
    setCPUMap(cpusMap, 0, pthread_self());
    recursorThread(0);
  }
  else {
    pthread_t tid;
    L<<Logger::Warning<<"Launching "<< g_numThreads <<" threads"<<endl;
    for(unsigned int n=0; n < g_numThreads; ++n) {
      pthread_create(&tid, 0, recursorThread, (void*)(long)n);

      setCPUMap(cpusMap, n, tid);
    }
    void* res;
#ifdef HAVE_SYSTEMD
    sd_notify(0, "READY=1");
#endif
    pthread_join(tid, &res);
  }
  return 0;
}

static void* recursorThread(void* ptr)
try
{
  t_id=(int) (long) ptr;
  SyncRes tmp(g_now); // make sure it allocates tsstorage before we do anything, like primeHints or so..
  SyncRes::setDomainMap(g_initialDomainMap);
  t_allowFrom = g_initialAllowFrom;
  t_udpclientsocks = std::unique_ptr<UDPClientSocks>(new UDPClientSocks());
  t_tcpClientCounts = std::unique_ptr<tcpClientCounts_t>(new tcpClientCounts_t());
  primeHints();

  t_packetCache = std::unique_ptr<RecursorPacketCache>(new RecursorPacketCache());

#ifdef HAVE_PROTOBUF
  t_uuidGenerator = std::unique_ptr<boost::uuids::random_generator>(new boost::uuids::random_generator());
#endif
  L<<Logger::Warning<<"Done priming cache with root hints"<<endl;

  try {
    if(!::arg()["lua-dns-script"].empty()) {
      t_pdl = std::make_shared<RecursorLua4>(::arg()["lua-dns-script"]);
      L<<Logger::Warning<<"Loaded 'lua' script from '"<<::arg()["lua-dns-script"]<<"'"<<endl;
    }
  }
  catch(std::exception &e) {
    L<<Logger::Error<<"Failed to load 'lua' script from '"<<::arg()["lua-dns-script"]<<"': "<<e.what()<<endl;
    _exit(99);
  }

  unsigned int ringsize=::arg().asNum("stats-ringbuffer-entries") / g_numWorkerThreads;
  if(ringsize) {
    t_remotes = std::unique_ptr<addrringbuf_t>(new addrringbuf_t());
    if(g_weDistributeQueries)  // if so, only 1 thread does recvfrom
      t_remotes->set_capacity(::arg().asNum("stats-ringbuffer-entries"));
    else
      t_remotes->set_capacity(ringsize);
    t_servfailremotes = std::unique_ptr<addrringbuf_t>(new addrringbuf_t());
    t_servfailremotes->set_capacity(ringsize);
    t_largeanswerremotes = std::unique_ptr<addrringbuf_t>(new addrringbuf_t());
    t_largeanswerremotes->set_capacity(ringsize);

    t_queryring = std::unique_ptr<boost::circular_buffer<pair<DNSName, uint16_t> > >(new boost::circular_buffer<pair<DNSName, uint16_t> >());
    t_queryring->set_capacity(ringsize);
    t_servfailqueryring = std::unique_ptr<boost::circular_buffer<pair<DNSName, uint16_t> > >(new boost::circular_buffer<pair<DNSName, uint16_t> >());
    t_servfailqueryring->set_capacity(ringsize);
  }

  MT=std::unique_ptr<MTasker<PacketID,string> >(new MTasker<PacketID,string>(::arg().asNum("stack-size")));

  PacketID pident;

  t_fdm=getMultiplexer();
  if(!t_id) {
    if(::arg().mustDo("webserver")) {
      L<<Logger::Warning << "Enabling web server" << endl;
      try {
        new RecursorWebServer(t_fdm);
      }
      catch(PDNSException &e) {
        L<<Logger::Error<<"Exception: "<<e.reason<<endl;
        exit(99);
      }
    }
    L<<Logger::Error<<"Enabled '"<< t_fdm->getName() << "' multiplexer"<<endl;
  }

  t_fdm->addReadFD(g_pipes[t_id].readToThread, handlePipeRequest);

  if(g_useOneSocketPerThread) {
    for(deferredAdd_t::const_iterator i = deferredAdds[t_id].cbegin(); i != deferredAdds[t_id].cend(); ++i) {
      t_fdm->addReadFD(i->first, i->second);
    }
  }
  else {
    if(!g_weDistributeQueries || !t_id) { // if we distribute queries, only t_id = 0 listens
      for(deferredAdd_t::const_iterator i = deferredAdds[0].cbegin(); i != deferredAdds[0].cend(); ++i) {
        t_fdm->addReadFD(i->first, i->second);
      }
    }
  }

  registerAllStats();
  if(!t_id) {
    t_fdm->addReadFD(s_rcc.d_fd, handleRCC); // control channel
  }

  unsigned int maxTcpClients=::arg().asNum("max-tcp-clients");

  bool listenOnTCP(true);

  time_t last_carbon=0;
  time_t carbonInterval=::arg().asNum("carbon-interval");
  counter.store(0); // used to periodically execute certain tasks
  for(;;) {
    while(MT->schedule(&g_now)); // MTasker letting the mthreads do their thing

    if(!(counter%500)) {
      MT->makeThread(houseKeeping, 0);
    }

    if(!(counter%55)) {
      typedef vector<pair<int, FDMultiplexer::funcparam_t> > expired_t;
      expired_t expired=t_fdm->getTimeouts(g_now);

      for(expired_t::iterator i=expired.begin() ; i != expired.end(); ++i) {
        shared_ptr<TCPConnection> conn=any_cast<shared_ptr<TCPConnection> >(i->second);
        if(g_logCommonErrors)
          L<<Logger::Warning<<"Timeout from remote TCP client "<< conn->d_remote.toString() <<endl;
        t_fdm->removeReadFD(i->first);
      }
    }

    counter++;

    if(!t_id && statsWanted) {
      doStats();
    }

    Utility::gettimeofday(&g_now, 0);

    if(!t_id && (g_now.tv_sec - last_carbon >= carbonInterval)) {
      MT->makeThread(doCarbonDump, 0);
      last_carbon = g_now.tv_sec;
    }

    t_fdm->run(&g_now);
    // 'run' updates g_now for us

    if(!g_weDistributeQueries || !t_id) { // if pdns distributes queries, only tid 0 should do this
      if(listenOnTCP) {
        if(TCPConnection::getCurrentConnections() > maxTcpClients) {  // shutdown, too many connections
          for(tcpListenSockets_t::iterator i=g_tcpListenSockets.begin(); i != g_tcpListenSockets.end(); ++i)
            t_fdm->removeReadFD(*i);
          listenOnTCP=false;
        }
      }
      else {
        if(TCPConnection::getCurrentConnections() <= maxTcpClients) {  // reenable
          for(tcpListenSockets_t::iterator i=g_tcpListenSockets.begin(); i != g_tcpListenSockets.end(); ++i)
            t_fdm->addReadFD(*i, handleNewTCPQuestion);
          listenOnTCP=true;
        }
      }
    }
  }
}
catch(PDNSException &ae) {
  L<<Logger::Error<<"Exception: "<<ae.reason<<endl;
  return 0;
}
catch(std::exception &e) {
   L<<Logger::Error<<"STL Exception: "<<e.what()<<endl;
   return 0;
}
catch(...) {
   L<<Logger::Error<<"any other exception in main: "<<endl;
   return 0;
}


int main(int argc, char **argv)
{
  g_argc = argc;
  g_argv = argv;
  g_stats.startupTime=time(0);
  versionSetProduct(ProductRecursor);
  reportBasicTypes();
  reportOtherTypes();

  int ret = EXIT_SUCCESS;

  try {
    ::arg().set("stack-size","stack size per mthread")="200000";
    ::arg().set("soa-minimum-ttl","Don't change")="0";
    ::arg().set("no-shuffle","Don't change")="off";
    ::arg().set("local-port","port to listen on")="53";
    ::arg().set("local-address","IP addresses to listen on, separated by spaces or commas. Also accepts ports.")="127.0.0.1";
    ::arg().setSwitch("non-local-bind", "Enable binding to non-local addresses by using FREEBIND / BINDANY socket options")="no";
    ::arg().set("trace","if we should output heaps of logging. set to 'fail' to only log failing domains")="off";
    ::arg().set("dnssec", "DNSSEC mode: off/process-no-validate (default)/process/log-fail/validate")="process-no-validate";
    ::arg().set("dnssec-log-bogus", "Log DNSSEC bogus validations")="no";
    ::arg().set("daemon","Operate as a daemon")="no";
    ::arg().setSwitch("write-pid","Write a PID file")="yes";
    ::arg().set("loglevel","Amount of logging. Higher is more. Do not set below 3")="6";
    ::arg().set("disable-syslog","Disable logging to syslog, useful when running inside a supervisor that logs stdout")="no";
    ::arg().set("log-timestamp","Print timestamps in log lines, useful to disable when running with a tool that timestamps stdout already")="yes";
    ::arg().set("log-common-errors","If we should log rather common errors")="no";
    ::arg().set("chroot","switch to chroot jail")="";
    ::arg().set("setgid","If set, change group id to this gid for more security")="";
    ::arg().set("setuid","If set, change user id to this uid for more security")="";
    ::arg().set("network-timeout", "Wait this number of milliseconds for network i/o")="1500";
    ::arg().set("threads", "Launch this number of threads")="2";
    ::arg().set("processes", "Launch this number of processes (EXPERIMENTAL, DO NOT CHANGE)")="1"; // if we un-experimental this, need to fix openssl rand seeding for multiple PIDs!
    ::arg().set("config-name","Name of this virtual configuration - will rename the binary image")="";
    ::arg().set("api-config-dir", "Directory where REST API stores config and zones") = "";
    ::arg().set("api-key", "Static pre-shared authentication key for access to the REST API") = "";
    ::arg().set("api-logfile", "Location of the server logfile (used by the REST API)") = "/var/log/pdns.log";
    ::arg().set("api-readonly", "Disallow data modification through the REST API when set") = "no";
    ::arg().setSwitch("webserver", "Start a webserver (for REST API)") = "no";
    ::arg().set("webserver-address", "IP Address of webserver to listen on") = "127.0.0.1";
    ::arg().set("webserver-port", "Port of webserver to listen on") = "8082";
    ::arg().set("webserver-password", "Password required for accessing the webserver") = "";
    ::arg().set("webserver-allow-from","Webserver access is only allowed from these subnets")="127.0.0.1,::1";
    ::arg().set("carbon-ourname", "If set, overrides our reported hostname for carbon stats")="";
    ::arg().set("carbon-server", "If set, send metrics in carbon (graphite) format to this server IP address")="";
    ::arg().set("carbon-interval", "Number of seconds between carbon (graphite) updates")="30";
    ::arg().set("statistics-interval", "Number of seconds between printing of recursor statistics, 0 to disable")="1800";
    ::arg().set("quiet","Suppress logging of questions and answers")="";
    ::arg().set("logging-facility","Facility to log messages as. 0 corresponds to local0")="";
    ::arg().set("config-dir","Location of configuration directory (recursor.conf)")=SYSCONFDIR;
    ::arg().set("socket-owner","Owner of socket")="";
    ::arg().set("socket-group","Group of socket")="";
    ::arg().set("socket-mode", "Permissions for socket")="";

    ::arg().set("socket-dir",string("Where the controlsocket will live, ")+LOCALSTATEDIR+" when unset and not chrooted" )="";
    ::arg().set("delegation-only","Which domains we only accept delegations from")="";
    ::arg().set("query-local-address","Source IP address for sending queries")="0.0.0.0";
    ::arg().set("query-local-address6","Source IPv6 address for sending queries. IF UNSET, IPv6 WILL NOT BE USED FOR OUTGOING QUERIES")="";
    ::arg().set("client-tcp-timeout","Timeout in seconds when talking to TCP clients")="2";
    ::arg().set("max-mthreads", "Maximum number of simultaneous Mtasker threads")="2048";
    ::arg().set("max-tcp-clients","Maximum number of simultaneous TCP clients")="128";
    ::arg().set("server-down-max-fails","Maximum number of consecutive timeouts (and unreachables) to mark a server as down ( 0 => disabled )")="64";
    ::arg().set("server-down-throttle-time","Number of seconds to throttle all queries to a server after being marked as down")="60";
    ::arg().set("hint-file", "If set, load root hints from this file")="";
    ::arg().set("max-cache-entries", "If set, maximum number of entries in the main cache")="1000000";
    ::arg().set("max-negative-ttl", "maximum number of seconds to keep a negative cached entry in memory")="3600";
    ::arg().set("max-cache-ttl", "maximum number of seconds to keep a cached entry in memory")="86400";
    ::arg().set("packetcache-ttl", "maximum number of seconds to keep a cached entry in packetcache")="3600";
    ::arg().set("max-packetcache-entries", "maximum number of entries to keep in the packetcache")="500000";
    ::arg().set("packetcache-servfail-ttl", "maximum number of seconds to keep a cached servfail entry in packetcache")="60";
    ::arg().set("server-id", "Returned when queried for 'id.server' TXT or NSID, defaults to hostname")="";
    ::arg().set("stats-ringbuffer-entries", "maximum number of packets to store statistics for")="10000";
    ::arg().set("version-string", "string reported on version.pdns or version.bind")=fullVersionString();
    ::arg().set("allow-from", "If set, only allow these comma separated netmasks to recurse")=LOCAL_NETS;
    ::arg().set("allow-from-file", "If set, load allowed netmasks from this file")="";
    ::arg().set("entropy-source", "If set, read entropy from this file")="/dev/urandom";
    ::arg().set("dont-query", "If set, do not query these netmasks for DNS data")=DONT_QUERY;
    ::arg().set("max-tcp-per-client", "If set, maximum number of TCP sessions per client (IP address)")="0";
    ::arg().set("max-tcp-queries-per-connection", "If set, maximum number of TCP queries in a TCP connection")="0";
    ::arg().set("spoof-nearmiss-max", "If non-zero, assume spoofing after this many near misses")="20";
    ::arg().set("single-socket", "If set, only use a single socket for outgoing queries")="off";
    ::arg().set("auth-zones", "Zones for which we have authoritative data, comma separated domain=file pairs ")="";
    ::arg().set("lua-config-file", "More powerful configuration options")="";

    ::arg().set("forward-zones", "Zones for which we forward queries, comma separated domain=ip pairs")="";
    ::arg().set("forward-zones-recurse", "Zones for which we forward queries with recursion bit, comma separated domain=ip pairs")="";
    ::arg().set("forward-zones-file", "File with (+)domain=ip pairs for forwarding")="";
    ::arg().set("export-etc-hosts", "If we should serve up contents from /etc/hosts")="off";
    ::arg().set("export-etc-hosts-search-suffix", "Also serve up the contents of /etc/hosts with this suffix")="";
    ::arg().set("etc-hosts-file", "Path to 'hosts' file")="/etc/hosts";
    ::arg().set("serve-rfc1918", "If we should be authoritative for RFC 1918 private IP space")="yes";
    ::arg().set("lua-dns-script", "Filename containing an optional 'lua' script that will be used to modify dns answers")="";
    ::arg().set("latency-statistic-size","Number of latency values to calculate the qa-latency average")="10000";
    ::arg().setSwitch( "disable-packetcache", "Disable packetcache" )= "no";
    ::arg().set("ecs-ipv4-bits", "Number of bits of IPv4 address to pass for EDNS Client Subnet")="24";
    ::arg().set("ecs-ipv6-bits", "Number of bits of IPv6 address to pass for EDNS Client Subnet")="56";
    ::arg().set("edns-subnet-whitelist", "List of netmasks and domains that we should enable EDNS subnet for")="";
    ::arg().set("ecs-scope-zero-address", "Address to send to whitelisted authoritative servers for incoming queries with ECS prefix-length source of 0")="";
    ::arg().setSwitch( "use-incoming-edns-subnet", "Pass along received EDNS Client Subnet information")="no";
    ::arg().setSwitch( "pdns-distributes-queries", "If PowerDNS itself should distribute queries over threads")="yes";
    ::arg().setSwitch( "root-nx-trust", "If set, believe that an NXDOMAIN from the root means the TLD does not exist")="yes";
    ::arg().setSwitch( "any-to-tcp","Answer ANY queries with tc=1, shunting to TCP" )="no";
    ::arg().setSwitch( "lowercase-outgoing","Force outgoing questions to lowercase")="no";
    ::arg().setSwitch("gettag-needs-edns-options", "If EDNS Options should be extracted before calling the gettag() hook")="no";
    ::arg().set("udp-truncation-threshold", "Maximum UDP response size before we truncate")="1680";
    ::arg().set("edns-outgoing-bufsize", "Outgoing EDNS buffer size")="1680";
    ::arg().set("minimum-ttl-override", "Set under adverse conditions, a minimum TTL")="0";
    ::arg().set("max-qperq", "Maximum outgoing queries per query")="50";
    ::arg().set("max-total-msec", "Maximum total wall-clock time per query in milliseconds, 0 for unlimited")="7000";
    ::arg().set("max-recursion-depth", "Maximum number of internal recursion calls per query, 0 for unlimited")="40";

    ::arg().set("include-dir","Include *.conf files from this directory")="";
    ::arg().set("security-poll-suffix","Domain name from which to query security update notifications")="secpoll.powerdns.com.";
    
    ::arg().setSwitch("reuseport","Enable SO_REUSEPORT allowing multiple recursors processes to listen to 1 address")="no";

    ::arg().setSwitch("snmp-agent", "If set, register as an SNMP agent")="no";
    ::arg().set("snmp-master-socket", "If set and snmp-agent is set, the socket to use to register to the SNMP master")="";

    ::arg().set("tcp-fast-open", "Enable TCP Fast Open support on the listening sockets, using the supplied numerical value as the queue size")="0";
    ::arg().set("nsec3-max-iterations", "Maximum number of iterations allowed for an NSEC3 record")="2500";

    ::arg().set("cpu-map", "Thread to CPU mapping, space separated thread-id=cpu1,cpu2..cpuN pairs")="";

    ::arg().setSwitch("log-rpz-changes", "Log additions and removals to RPZ zones at Info level")="no";

    ::arg().setCmd("help","Provide a helpful message");
    ::arg().setCmd("version","Print version string");
    ::arg().setCmd("config","Output blank configuration");
    L.toConsole(Logger::Info);
    ::arg().laxParse(argc,argv); // do a lax parse

    string configname=::arg()["config-dir"]+"/recursor.conf";
    if(::arg()["config-name"]!="") {
      configname=::arg()["config-dir"]+"/recursor-"+::arg()["config-name"]+".conf";
      s_programname+="-"+::arg()["config-name"];
    }
    cleanSlashes(configname);

    if(::arg().mustDo("config")) {
      cout<<::arg().configstring()<<endl;
      exit(0);
    }

    if(!::arg().file(configname.c_str()))
      L<<Logger::Warning<<"Unable to parse configuration file '"<<configname<<"'"<<endl;

    ::arg().parse(argc,argv);

    if( !::arg()["chroot"].empty() && !::arg()["api-config-dir"].empty() && !::arg().mustDo("api-readonly") )  {
      L<<Logger::Error<<"Using chroot and a writable API is not possible"<<endl;
      exit(EXIT_FAILURE);
    }

    if (::arg()["socket-dir"].empty()) {
      if (::arg()["chroot"].empty())
        ::arg().set("socket-dir") = LOCALSTATEDIR;
      else
        ::arg().set("socket-dir") = "/";
    }

    ::arg().set("delegation-only")=toLower(::arg()["delegation-only"]);

    if(::arg().asNum("threads")==1)
      ::arg().set("pdns-distributes-queries")="no";

    if(::arg().mustDo("help")) {
      cout<<"syntax:"<<endl<<endl;
      cout<<::arg().helpstring(::arg()["help"])<<endl;
      exit(0);
    }
    if(::arg().mustDo("version")) {
      showProductVersion();
      showBuildConfiguration();
      exit(0);
    }

    Logger::Urgency logUrgency = (Logger::Urgency)::arg().asNum("loglevel");

    if (logUrgency < Logger::Error)
      logUrgency = Logger::Error;
    if(!g_quiet && logUrgency < Logger::Info) { // Logger::Info=6, Logger::Debug=7
      logUrgency = Logger::Info;                // if you do --quiet=no, you need Info to also see the query log
    }
    L.setLoglevel(logUrgency);
    L.toConsole(logUrgency);

    serviceMain(argc, argv);
  }
  catch(PDNSException &ae) {
    L<<Logger::Error<<"Exception: "<<ae.reason<<endl;
    ret=EXIT_FAILURE;
  }
  catch(std::exception &e) {
    L<<Logger::Error<<"STL Exception: "<<e.what()<<endl;
    ret=EXIT_FAILURE;
  }
  catch(...) {
    L<<Logger::Error<<"any other exception in main: "<<endl;
    ret=EXIT_FAILURE;
  }

  return ret;
}