gvpe/src/protocol.C

/*
    protocol.C -- handle the protocol, encryption, handshaking etc.
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    (at your option) any later version.
 
    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. 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/

#include "config.h"

#include <list>

#include <cstdlib>
#include <cstring>
#include <cstdio>

#include <sys/types.h>
#include <sys/socket.h>
#include <sys/poll.h>
#include <sys/wait.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <errno.h>
#include <time.h>
#include <unistd.h>

#include <openssl/rand.h>
#include <openssl/hmac.h>
#include <openssl/evp.h>
#include <openssl/rsa.h>
#include <openssl/err.h>

extern "C" {
# include "lzf/lzf.h"
}

#include "gettext.h"
#include "pidfile.h"

#include "conf.h"
#include "slog.h"
#include "device.h"
#include "protocol.h"

#if !HAVE_RAND_PSEUDO_BYTES
# define  RAND_pseudo_bytes RAND_bytes
#endif

static time_t next_timecheck;

#define MAGIC "vped\xbd\xc6\xdb\x82"    // 8 bytes of magic

static u8
best_protocol (u8 protset)
{
  if (protset & PROT_IPv4)
    return PROT_IPv4;

  return PROT_UDPv4;
}

struct crypto_ctx
  {
    EVP_CIPHER_CTX cctx;
    HMAC_CTX hctx;

    crypto_ctx (const rsachallenge &challenge, int enc);
    ~crypto_ctx ();
  };

crypto_ctx::crypto_ctx (const rsachallenge &challenge, int enc)
{
  EVP_CIPHER_CTX_init (&cctx);
  EVP_CipherInit_ex (&cctx, CIPHER, 0, &challenge[CHG_CIPHER_KEY], 0, enc);
  HMAC_CTX_init (&hctx);
  HMAC_Init_ex (&hctx, &challenge[CHG_HMAC_KEY], HMAC_KEYLEN, DIGEST, 0);
}

crypto_ctx::~crypto_ctx ()
{
  EVP_CIPHER_CTX_cleanup (&cctx);
  HMAC_CTX_cleanup (&hctx);
}

static void
rsa_hash (const rsaid &id, const rsachallenge &chg, rsaresponse &h)
{
  EVP_MD_CTX ctx;

  EVP_MD_CTX_init (&ctx);
  EVP_DigestInit (&ctx, RSA_HASH);
  EVP_DigestUpdate(&ctx, &chg, sizeof chg);
  EVP_DigestUpdate(&ctx, &id, sizeof id);
  EVP_DigestFinal (&ctx, (unsigned char *)&h, 0);
  EVP_MD_CTX_cleanup (&ctx);
}

struct rsa_entry {
  tstamp expire;
  rsaid id;
  rsachallenge chg;
};

struct rsa_cache : list<rsa_entry>
{
  void cleaner_cb (tstamp &ts); time_watcher cleaner;
  
  bool find (const rsaid &id, rsachallenge &chg)
    {
      for (iterator i = begin (); i != end (); ++i)
        {
          if (!memcmp (&id, &i->id, sizeof id) && i->expire > NOW)
            {
              memcpy (&chg, &i->chg, sizeof chg);

              erase (i);
              return true;
            }
        }

      if (cleaner.at < NOW)
        cleaner.start (NOW + RSA_TTL);

      return false;
    }

  void gen (rsaid &id, rsachallenge &chg)
    {
      rsa_entry e;

      RAND_bytes ((unsigned char *)&id,  sizeof id);
      RAND_bytes ((unsigned char *)&chg, sizeof chg);

      e.expire = NOW + RSA_TTL;
      e.id = id;
      memcpy (&e.chg, &chg, sizeof chg);

      push_back (e);

      if (cleaner.at < NOW)
        cleaner.start (NOW + RSA_TTL);
    }

  rsa_cache ()
    : cleaner (this, &rsa_cache::cleaner_cb)
    { }

} rsa_cache;

void rsa_cache::cleaner_cb (tstamp &ts)
{
  if (empty ())
    ts = TSTAMP_CANCEL;
  else
    {
      ts = NOW + RSA_TTL;

      for (iterator i = begin (); i != end (); )
        if (i->expire <= NOW)
          i = erase (i);
        else
          ++i;
    }
}

typedef callback<const char *, int> run_script_cb;

// run a shell script (or actually an external program).
static void
run_script (const run_script_cb &cb, bool wait)
{
  int pid;

  if ((pid = fork ()) == 0)
    {
      char *filename;
      asprintf (&filename, "%s/%s", confbase, cb(0));
      execl (filename, filename, (char *) 0);
      exit (255);
    }
  else if (pid > 0)
    {
      if (wait)
        {
          waitpid (pid, 0, 0);
          /* TODO: check status */
        }
    }
}

//////////////////////////////////////////////////////////////////////////////

void pkt_queue::put (tap_packet *p)
{
  if (queue[i])
    {
      delete queue[i];
      j = (j + 1) % QUEUEDEPTH;
    }

  queue[i] = p;

  i = (i + 1) % QUEUEDEPTH;
}

tap_packet *pkt_queue::get ()
{
  tap_packet *p = queue[j];

  if (p)
    {
      queue[j] = 0;
      j = (j + 1) % QUEUEDEPTH;
    }

  return p;
}

pkt_queue::pkt_queue ()
{
  memset (queue, 0, sizeof (queue));
  i = 0;
  j = 0;
}

pkt_queue::~pkt_queue ()
{
  for (i = QUEUEDEPTH; --i > 0; )
    delete queue[i];
}

struct net_rateinfo {
  u32    host;
  double pcnt, diff;
  tstamp last;
};

// only do action once every x seconds per host whole allowing bursts.
// this implementation ("splay list" ;) is inefficient,
// but low on resources.
struct net_rate_limiter : list<net_rateinfo>
{
  static const double ALPHA  = 1. - 1. / 90.; // allow bursts
  static const double CUTOFF = 20.;           // one event every CUTOFF seconds
  static const double EXPIRE = CUTOFF * 30.;  // expire entries after this time

  bool can (const sockinfo &si) { return can((u32)si.host);             }
  bool can (u32 host);
};

net_rate_limiter auth_rate_limiter, reset_rate_limiter;

bool net_rate_limiter::can (u32 host)
{
  iterator i;

  for (i = begin (); i != end (); )
    if (i->host == host)
      break;
    else if (i->last < NOW - EXPIRE)
      i = erase (i);
    else
      i++;

  if (i == end ())
    {
      net_rateinfo ri;

      ri.host = host;
      ri.pcnt = 1.;
      ri.diff = CUTOFF * (1. / (1. - ALPHA));
      ri.last = NOW;

      push_front (ri);

      return true;
    }
  else
    {
      net_rateinfo ri (*i);
      erase (i);

      ri.pcnt = ri.pcnt * ALPHA;
      ri.diff = ri.diff * ALPHA + (NOW - ri.last);

      ri.last = NOW;

      bool send = ri.diff / ri.pcnt > CUTOFF;

      if (send)
        ri.pcnt++;

      push_front (ri);

      return send;
    }
}

/////////////////////////////////////////////////////////////////////////////

static void next_wakeup (time_t next)
{
  if (next_timecheck > next)
    next_timecheck = next;
}

static unsigned char hmac_digest[EVP_MAX_MD_SIZE];

struct hmac_packet:net_packet
{
  u8 hmac[HMACLENGTH];          // each and every packet has a hmac field, but that is not (yet) checked everywhere

  void hmac_set (crypto_ctx * ctx);
  bool hmac_chk (crypto_ctx * ctx);

private:
  void hmac_gen (crypto_ctx * ctx)
  {
    unsigned int xlen;
    HMAC_CTX *hctx = &ctx->hctx;

    HMAC_Init_ex (hctx, 0, 0, 0, 0);
    HMAC_Update (hctx, ((unsigned char *) this) + sizeof (hmac_packet),
                 len - sizeof (hmac_packet));
    HMAC_Final (hctx, (unsigned char *) &hmac_digest, &xlen);
  }
};

void
hmac_packet::hmac_set (crypto_ctx * ctx)
{
  hmac_gen (ctx);

  memcpy (hmac, hmac_digest, HMACLENGTH);
}

bool
hmac_packet::hmac_chk (crypto_ctx * ctx)
{
  hmac_gen (ctx);

  return !memcmp (hmac, hmac_digest, HMACLENGTH);
}

struct vpn_packet : hmac_packet
  {
    enum ptype
    {
      PT_RESET = 0,
      PT_DATA_UNCOMPRESSED,
      PT_DATA_COMPRESSED,
      PT_PING, PT_PONG, // wasting namespace space? ;)
      PT_AUTH_REQ,      // authentification request
      PT_AUTH_RES,      // authentification response
      PT_CONNECT_REQ,   // want other host to contact me
      PT_CONNECT_INFO,  // request connection to some node
      PT_MAX
    };

    u8 type;
    u8 srcdst, src1, dst1;

    void set_hdr (ptype type, unsigned int dst);

    unsigned int src () const
    {
      return src1 | ((srcdst >> 4) << 8);
    }

    unsigned int dst () const
    {
      return dst1 | ((srcdst & 0xf) << 8);
    }

    ptype typ () const
    {
      return (ptype) type;
    }
  };

void vpn_packet::set_hdr (ptype type, unsigned int dst)
{
  this->type = type;

  int src = THISNODE->id;

  src1 = src;
  srcdst = ((src >> 8) << 4) | (dst >> 8);
  dst1 = dst;
}

#define MAXVPNDATA (MAX_MTU - 6 - 6)
#define DATAHDR (sizeof (u32) + RAND_SIZE)

struct vpndata_packet:vpn_packet
  {
    u8 data[MAXVPNDATA + DATAHDR]; // seqno

    void setup (connection *conn, int dst, u8 *d, u32 len, u32 seqno);
    tap_packet *unpack (connection *conn, u32 &seqno);
private:

    const u32 data_hdr_size () const
    {
      return sizeof (vpndata_packet) - sizeof (net_packet) - MAXVPNDATA - DATAHDR;
    }
  };

void
vpndata_packet::setup (connection *conn, int dst, u8 *d, u32 l, u32 seqno)
{
  EVP_CIPHER_CTX *cctx = &conn->octx->cctx;
  int outl = 0, outl2;
  ptype type = PT_DATA_UNCOMPRESSED;

#if ENABLE_COMPRESSION
  u8 cdata[MAX_MTU];
  u32 cl;

  cl = lzf_compress (d, l, cdata + 2, (l - 2) & ~7);
  if (cl)
    {
      type = PT_DATA_COMPRESSED;
      d = cdata;
      l = cl + 2;

      d[0] = cl >> 8;
      d[1] = cl;
    }
#endif

  EVP_EncryptInit_ex (cctx, 0, 0, 0, 0);

  struct {
#if RAND_SIZE
    u8 rnd[RAND_SIZE];
#endif
    u32 seqno;
  } datahdr;

  datahdr.seqno = ntohl (seqno);
#if RAND_SIZE
  RAND_pseudo_bytes ((unsigned char *) datahdr.rnd, RAND_SIZE);
#endif

  EVP_EncryptUpdate (cctx,
                     (unsigned char *) data + outl, &outl2,
                     (unsigned char *) &datahdr, DATAHDR);
  outl += outl2;

  EVP_EncryptUpdate (cctx,
                     (unsigned char *) data + outl, &outl2,
                     (unsigned char *) d, l);
  outl += outl2;

  EVP_EncryptFinal_ex (cctx, (unsigned char *) data + outl, &outl2);
  outl += outl2;

  len = outl + data_hdr_size ();

  set_hdr (type, dst);

  hmac_set (conn->octx);
}

tap_packet *
vpndata_packet::unpack (connection *conn, u32 &seqno)
{
  EVP_CIPHER_CTX *cctx = &conn->ictx->cctx;
  int outl = 0, outl2;
  tap_packet *p = new tap_packet;
  u8 *d;
  u32 l = len - data_hdr_size ();

  EVP_DecryptInit_ex (cctx, 0, 0, 0, 0);

#if ENABLE_COMPRESSION
  u8 cdata[MAX_MTU];

  if (type == PT_DATA_COMPRESSED)
    d = cdata;
  else
#endif
    d = &(*p)[6 + 6 - DATAHDR];

  /* this overwrites part of the src mac, but we fix that later */
  EVP_DecryptUpdate (cctx,
                     d, &outl2,
                     (unsigned char *)&data, len - data_hdr_size ());
  outl += outl2;

  EVP_DecryptFinal_ex (cctx, (unsigned char *)d + outl, &outl2);
  outl += outl2;
  
  seqno = ntohl (*(u32 *)(d + RAND_SIZE));

  id2mac (dst () ? dst() : THISNODE->id, p->dst);
  id2mac (src (),                        p->src);

#if ENABLE_COMPRESSION
  if (type == PT_DATA_COMPRESSED)
    {
      u32 cl = (d[DATAHDR] << 8) | d[DATAHDR + 1];

      p->len = lzf_decompress (d + DATAHDR + 2, cl < MAX_MTU ? cl : 0,
                               &(*p)[6 + 6], MAX_MTU)
               + 6 + 6;
    }
  else
    p->len = outl + (6 + 6 - DATAHDR);
#endif

  return p;
}

struct ping_packet : vpn_packet
{
  void setup (int dst, ptype type)
  {
    set_hdr (type, dst);
    len = sizeof (*this) - sizeof (net_packet);
  }
};

struct config_packet : vpn_packet
{
  // actually, hmaclen cannot be checked because the hmac
  // field comes before this data, so peers with other
  // hmacs simply will not work.
  u8 prot_major, prot_minor, randsize, hmaclen;
  u8 flags, challengelen, pad2, pad3;
  u32 cipher_nid, digest_nid, hmac_nid;

  const u8 curflags () const
  {
    return 0x80
           | (ENABLE_COMPRESSION ? 0x01 : 0x00);
  }

  void setup (ptype type, int dst);
  bool chk_config () const;
};

void config_packet::setup (ptype type, int dst)
{
  prot_major = PROTOCOL_MAJOR;
  prot_minor = PROTOCOL_MINOR;
  randsize = RAND_SIZE;
  hmaclen = HMACLENGTH;
  flags = curflags ();
  challengelen = sizeof (rsachallenge);

  cipher_nid = htonl (EVP_CIPHER_nid (CIPHER));
  digest_nid = htonl (EVP_MD_type (RSA_HASH));
  hmac_nid   = htonl (EVP_MD_type (DIGEST));

  len = sizeof (*this) - sizeof (net_packet);
  set_hdr (type, dst);
}

bool config_packet::chk_config () const
{
  return prot_major == PROTOCOL_MAJOR
         && randsize == RAND_SIZE
         && hmaclen == HMACLENGTH
         && flags == curflags ()
         && challengelen == sizeof (rsachallenge)
         && cipher_nid == htonl (EVP_CIPHER_nid (CIPHER))
         && digest_nid == htonl (EVP_MD_type (RSA_HASH))
         && hmac_nid   == htonl (EVP_MD_type (DIGEST));
}

struct auth_req_packet : config_packet
{
  char magic[8];
  u8 initiate, can_recv; // false if this is just an automatic reply
  u8 pad2, pad3;
  rsaid id;
  rsaencrdata encr;

  auth_req_packet (int dst, bool initiate_, u8 can_recv_)
  {
    config_packet::setup (PT_AUTH_REQ, dst);
    strncpy (magic, MAGIC, 8);
    initiate = !!initiate_;
    can_recv = can_recv_;

    len = sizeof (*this) - sizeof (net_packet);
  }
};

struct auth_res_packet : config_packet
{
  rsaid id;
  u8 pad1, pad2, pad3;
  u8 response_len; // encrypted length
  rsaresponse response;

  auth_res_packet (int dst)
  {
    config_packet::setup (PT_AUTH_RES, dst);

    len = sizeof (*this) - sizeof (net_packet);
  }
};

struct connect_req_packet : vpn_packet
{
  u8 id;
  u8 pad1, pad2, pad3;

  connect_req_packet (int dst, int id_)
  {
    id = id_;
    set_hdr (PT_CONNECT_REQ, dst);
    len = sizeof (*this) - sizeof (net_packet);
  }
};

struct connect_info_packet : vpn_packet
{
  u8 id, can_recv;
  u8 pad1, pad2;
  sockinfo si;

  connect_info_packet (int dst, int id_, sockinfo &si_, u8 can_recv_)
  {
    id = id_;
    can_recv = can_recv_;
    si = si_;
    set_hdr (PT_CONNECT_INFO, dst);

    len = sizeof (*this) - sizeof (net_packet);
  }
};

/////////////////////////////////////////////////////////////////////////////

void
connection::reset_dstaddr ()
{
  si.set (conf);
}

void
connection::send_ping (const sockinfo &si, u8 pong)
{
  ping_packet *pkt = new ping_packet;

  pkt->setup (conf->id, pong ? ping_packet::PT_PONG : ping_packet::PT_PING);
  send_vpn_packet (pkt, si, IPTOS_LOWDELAY);

  delete pkt;
}

void
connection::send_reset (const sockinfo &si)
{
  if (reset_rate_limiter.can (si) && connectmode != conf_node::C_DISABLED)
    {
      config_packet *pkt = new config_packet;

      pkt->setup (vpn_packet::PT_RESET, conf->id);
      send_vpn_packet (pkt, si, IPTOS_MINCOST);

      delete pkt;
    }
}

void
connection::send_auth_request (const sockinfo &si, bool initiate)
{
  auth_req_packet *pkt = new auth_req_packet (conf->id, initiate, THISNODE->can_recv);

  // the next line is very conservative
  prot_send = best_protocol (THISNODE->can_send & THISNODE->can_recv & conf->can_recv);

  rsachallenge chg;

  rsa_cache.gen (pkt->id, chg);

  if (0 > RSA_public_encrypt (sizeof chg,
                              (unsigned char *)&chg, (unsigned char *)&pkt->encr,
                              conf->rsa_key, RSA_PKCS1_OAEP_PADDING))
    fatal ("RSA_public_encrypt error");

  slog (L_TRACE, ">>%d PT_AUTH_REQ [%s]", conf->id, (const char *)si);

  send_vpn_packet (pkt, si, IPTOS_RELIABILITY); // rsa is very very costly

  delete pkt;
}

void
connection::send_auth_response (const sockinfo &si, const rsaid &id, const rsachallenge &chg)
{
  auth_res_packet *pkt = new auth_res_packet (conf->id);

  pkt->id = id;

  rsa_hash (id, chg, pkt->response);

  pkt->hmac_set (octx);

  slog (L_TRACE, ">>%d PT_AUTH_RES [%s]", conf->id, (const char *)si);

  send_vpn_packet (pkt, si, IPTOS_RELIABILITY); // rsa is very very costly

  delete pkt;
}

void
connection::establish_connection_cb (tstamp &ts)
{
  if (ictx || conf == THISNODE
      || connectmode == conf_node::C_NEVER
      || connectmode == conf_node::C_DISABLED)
    ts = TSTAMP_CANCEL;
  else if (ts <= NOW)
    {
      double retry_int = double (retry_cnt & 3 ? (retry_cnt & 3) : 1 << (retry_cnt >> 2)) * 0.6;

      if (retry_int < 3600 * 8)
        retry_cnt++;

      ts = NOW + retry_int;

      if (conf->hostname)
        {
          reset_dstaddr ();
          if (si.host && auth_rate_limiter.can (si))
           {
            if (retry_cnt < 4)
              send_auth_request (si, true);
            else
              send_ping (si, 0);
           }
        }
      else
        vpn->connect_request (conf->id);
    }
}

void
connection::reset_connection ()
{
  if (ictx && octx)
    {
      slog (L_INFO, _("%s(%s): connection lost"),
            conf->nodename, (const char *)si);

      if (::conf.script_node_down)
        run_script (run_script_cb (this, &connection::script_node_down), false);
    }

  delete ictx; ictx = 0;
  delete octx; octx = 0;

  si.host= 0;

  last_activity = 0;
  retry_cnt = 0;

  rekey.reset ();
  keepalive.reset ();
  establish_connection.reset ();
}

void
connection::shutdown ()
{
  if (ictx && octx)
    send_reset (si);

  reset_connection ();
}

void
connection::rekey_cb (tstamp &ts)
{
  ts = TSTAMP_CANCEL;

  reset_connection ();
  establish_connection ();
}

void
connection::send_data_packet (tap_packet *pkt, bool broadcast)
{
  vpndata_packet *p = new vpndata_packet;
  int tos = 0;

  if (conf->inherit_tos
      && (*pkt)[12] == 0x08 && (*pkt)[13] == 0x00 // IP
      && ((*pkt)[14] & 0xf0) == 0x40)             // IPv4
    tos = (*pkt)[15] & IPTOS_TOS_MASK;

  p->setup (this, broadcast ? 0 : conf->id, &((*pkt)[6 + 6]), pkt->len - 6 - 6, ++oseqno); // skip 2 macs
  send_vpn_packet (p, si, tos);

  delete p;

  if (oseqno > MAX_SEQNO)
    rekey ();
}

void
connection::inject_data_packet (tap_packet *pkt, bool broadcast)
{
  if (ictx && octx)
    send_data_packet (pkt, broadcast);
  else
    {
      if (!broadcast)//DDDD
        queue.put (new tap_packet (*pkt));

      establish_connection ();
    }
}

void
connection::recv_vpn_packet (vpn_packet *pkt, const sockinfo &rsi)
{
  last_activity = NOW;

  slog (L_NOISE, "<<%d received packet type %d from %d to %d", 
        conf->id, pkt->typ (), pkt->src (), pkt->dst ());

  switch (pkt->typ ())
    {
    case vpn_packet::PT_PING:
      // we send pings instead of auth packets after some retries,
      // so reset the retry counter and establish a connection
      // when we receive a ping.
      if (!ictx)
        {
          if (auth_rate_limiter.can (rsi))
            send_auth_request (rsi, true);
        }
      else
        send_ping (rsi, 1); // pong

      break;

    case vpn_packet::PT_PONG:
      break;

    case vpn_packet::PT_RESET:
      {
        reset_connection ();

        config_packet *p = (config_packet *) pkt;

        if (!p->chk_config ())
          {
            slog (L_WARN, _("%s(%s): protocol mismatch, disabling node"),
                  conf->nodename, (const char *)rsi);
            connectmode = conf_node::C_DISABLED;
          }
        else if (connectmode == conf_node::C_ALWAYS)
          establish_connection ();
      }
      break;

    case vpn_packet::PT_AUTH_REQ:
      if (auth_rate_limiter.can (rsi))
        {
          auth_req_packet *p = (auth_req_packet *) pkt;

          slog (L_TRACE, "<<%d PT_AUTH_REQ(%d)", conf->id, p->initiate);

          if (p->chk_config () && !strncmp (p->magic, MAGIC, 8))
            {
              if (p->prot_minor != PROTOCOL_MINOR)
                slog (L_INFO, _("%s(%s): protocol minor version mismatch: ours is %d, %s's is %d."),
                      conf->nodename, (const char *)rsi,
                      PROTOCOL_MINOR, conf->nodename, p->prot_minor);

              if (p->initiate)
                send_auth_request (rsi, false);

              rsachallenge k;

              if (0 > RSA_private_decrypt (sizeof (p->encr),
                                           (unsigned char *)&p->encr, (unsigned char *)&k,
                                           ::conf.rsa_key, RSA_PKCS1_OAEP_PADDING))
                slog (L_ERR, _("%s(%s): challenge illegal or corrupted"),
                      conf->nodename, (const char *)rsi);
              else
                {
                  retry_cnt = 0;
                  establish_connection.set (NOW + 8); //? ;)
                  keepalive.reset ();
                  rekey.reset ();

                  delete ictx;
                  ictx = 0;

                  delete octx;

                  octx   = new crypto_ctx (k, 1);
                  oseqno = ntohl (*(u32 *)&k[CHG_SEQNO]) & 0x7fffffff;

                  conf->can_recv = p->can_recv;
                  send_auth_response (rsi, p->id, k);

                  break;
                }
            }

          send_reset (rsi);
        }

      break;

    case vpn_packet::PT_AUTH_RES:
      {
        auth_res_packet *p = (auth_res_packet *) pkt;

        slog (L_TRACE, "<<%d PT_AUTH_RES", conf->id);

        if (p->chk_config ())
          {
            if (p->prot_minor != PROTOCOL_MINOR)
              slog (L_INFO, _("%s(%s): protocol minor version mismatch: ours is %d, %s's is %d."),
                    conf->nodename, (const char *)rsi,
                    PROTOCOL_MINOR, conf->nodename, p->prot_minor);

            rsachallenge chg;

            if (!rsa_cache.find (p->id, chg))
              slog (L_ERR, _("%s(%s): unrequested auth response"),
                    conf->nodename, (const char *)rsi);
            else
              {
                crypto_ctx *cctx = new crypto_ctx (chg, 0);

                if (!p->hmac_chk (cctx))
                  slog (L_ERR, _("%s(%s): hmac authentication error on auth response, received invalid packet\n"
                                 "could be an attack, or just corruption or an synchronization error"),
                        conf->nodename, (const char *)rsi);
                else
                  {
                    rsaresponse h;

                    rsa_hash (p->id, chg, h);

                    if (!memcmp ((u8 *)&h, (u8 *)p->response, sizeof h))
                      {
                        prot_minor = p->prot_minor;

                        delete ictx; ictx = cctx;

                        iseqno.reset (ntohl (*(u32 *)&chg[CHG_SEQNO]) & 0x7fffffff); // at least 2**31 sequence numbers are valid

                        si = rsi;

                        rekey.set (NOW + ::conf.rekey);
                        keepalive.set (NOW + ::conf.keepalive);

                        // send queued packets
                        while (tap_packet *p = queue.get ())
                          {
                            send_data_packet (p);
                            delete p;
                          }

                        connectmode = conf->connectmode;

                        slog (L_INFO, _("%s(%s): connection established, protocol version %d.%d"),
                              conf->nodename, (const char *)rsi,
                              p->prot_major, p->prot_minor);

                        if (::conf.script_node_up)
                          run_script (run_script_cb (this, &connection::script_node_up), false);

                        break;
                      }
                    else
                      slog (L_ERR, _("%s(%s): sent and received challenge do not match"),
                            conf->nodename, (const char *)rsi);
                  }

                delete cctx;
              }
          }
      }

      send_reset (rsi);
      break;

    case vpn_packet::PT_DATA_COMPRESSED:
#if !ENABLE_COMPRESSION
      send_reset (rsi);
      break;
#endif

    case vpn_packet::PT_DATA_UNCOMPRESSED:

      if (ictx && octx)
        {
          vpndata_packet *p = (vpndata_packet *)pkt;

          if (rsi == si)
            {
              if (!p->hmac_chk (ictx))
                slog (L_ERR, _("%s(%s): hmac authentication error, received invalid packet\n"
                               "could be an attack, or just corruption or an synchronization error"),
                      conf->nodename, (const char *)rsi);
              else
                {
                  u32 seqno;
                  tap_packet *d = p->unpack (this, seqno);

                  if (iseqno.recv_ok (seqno))
                    {
                      vpn->tap->send (d);

                      if (p->dst () == 0) // re-broadcast
                        for (vpn::conns_vector::iterator i = vpn->conns.begin (); i != vpn->conns.end (); ++i)
                          {
                            connection *c = *i;

                            if (c->conf != THISNODE && c->conf != conf)
                              c->inject_data_packet (d);
                          }

                      delete d;

                      break;
                    }
                }
            }
          else
            slog (L_ERR,  _("received data packet from unknown source %s"), (const char *)rsi);
        }

      send_reset (rsi);
      break;

    case vpn_packet::PT_CONNECT_REQ:
      if (ictx && octx && rsi == si && pkt->hmac_chk (ictx))
        {
          connect_req_packet *p = (connect_req_packet *) pkt;

          assert (p->id > 0 && p->id <= vpn->conns.size ()); // hmac-auth does not mean we accept anything

          connection *c = vpn->conns[p->id - 1];

          slog (L_TRACE, "<<%d PT_CONNECT_REQ(%d) [%d]\n",
                         conf->id, p->id, c->ictx && c->octx);

          if (c->ictx && c->octx)
            {
              // send connect_info packets to both sides, in case one is
              // behind a nat firewall (or both ;)
              {
                slog (L_TRACE, ">>%d PT_CONNECT_INFO(%d,%s)\n",
                               c->conf->id, conf->id, (const char *)si);

                connect_info_packet *r = new connect_info_packet (c->conf->id, conf->id, si, conf->can_recv);

                r->hmac_set (c->octx);
                send_vpn_packet (r, c->si);

                delete r;
              }

              {
                slog (L_TRACE, ">>%d PT_CONNECT_INFO(%d,%s)\n",
                               conf->id, c->conf->id, (const char *)c->si);

                connect_info_packet *r = new connect_info_packet (conf->id, c->conf->id, c->si, c->conf->can_recv);

                r->hmac_set (octx);
                send_vpn_packet (r, si);

                delete r;
              }
            }
        }

      break;

    case vpn_packet::PT_CONNECT_INFO:
      if (ictx && octx && rsi == si && pkt->hmac_chk (ictx))
        {
          connect_info_packet *p = (connect_info_packet *) pkt;

          assert (p->id > 0 && p->id <= vpn->conns.size ()); // hmac-auth does not mean we accept anything

          connection *c = vpn->conns[p->id - 1];

          slog (L_TRACE, "<<%d PT_CONNECT_INFO(%d,%s) (%d)",
                         conf->id, p->id, (const char *)p->si, !c->ictx && !c->octx);

          c->conf->can_recv = p->can_recv;
          c->send_auth_request (p->si, true);
        }

      break;

    default:
      send_reset (rsi);
      break;

    }
}

void connection::keepalive_cb (tstamp &ts)
{
  if (NOW >= last_activity + ::conf.keepalive + 30)
    {
      reset_connection ();
      establish_connection ();
    }
  else if (NOW < last_activity + ::conf.keepalive)
    ts = last_activity + ::conf.keepalive;
  else if (conf->connectmode != conf_node::C_ONDEMAND
           || THISNODE->connectmode != conf_node::C_ONDEMAND)
    {
      send_ping (si);
      ts = NOW + 5;
    }
  else
    reset_connection ();
}

void connection::connect_request (int id)
{
  connect_req_packet *p = new connect_req_packet (conf->id, id);

  slog (L_TRACE, ">>%d PT_CONNECT_REQ(%d)", id, conf->id);
  p->hmac_set (octx);
  send_vpn_packet (p, si);

  delete p;
}

void connection::script_node ()
{
  vpn->script_if_up (0);

  char *env;
  asprintf (&env, "DESTID=%d",   conf->id); putenv (env);
  asprintf (&env, "DESTNODE=%s", conf->nodename); putenv (env);
  asprintf (&env, "DESTIP=%s",   si.ntoa ()); putenv (env);
  asprintf (&env, "DESTPORT=%d", ntohs (si.port)); putenv (env);
}

const char *connection::script_node_up (int)
{
  script_node ();

  putenv ("STATE=up");

  return ::conf.script_node_up ? ::conf.script_node_up : "node-up";
}

const char *connection::script_node_down (int)
{
  script_node ();

  putenv ("STATE=down");

  return ::conf.script_node_up ? ::conf.script_node_down : "node-down";
}

// send a vpn packet out to other hosts
void
connection::send_vpn_packet (vpn_packet *pkt, const sockinfo &si, int tos)
{
  if (prot_send & PROT_IPv4)
    vpn->send_ipv4_packet (pkt, si, tos);
  else
    vpn->send_udpv4_packet (pkt, si, tos);
}

connection::connection(struct vpn *vpn_)
: vpn(vpn_)
, rekey (this, &connection::rekey_cb)
, keepalive (this, &connection::keepalive_cb)
, establish_connection (this, &connection::establish_connection_cb)
{
  octx = ictx = 0;
  retry_cnt = 0;

  connectmode = conf_node::C_ALWAYS; // initial setting
  reset_connection ();
}

connection::~connection ()
{
  shutdown ();
}

/////////////////////////////////////////////////////////////////////////////

const char *vpn::script_if_up (int)
{
  // the tunnel device mtu should be the physical mtu - overhead
  // the tricky part is rounding to the cipher key blocksize
  int mtu = conf.mtu - ETH_OVERHEAD - VPE_OVERHEAD - MAX_OVERHEAD;
  mtu += ETH_OVERHEAD - 6 - 6; // now we have the data portion
  mtu -= mtu % EVP_CIPHER_block_size (CIPHER); // round
  mtu -= ETH_OVERHEAD - 6 - 6; // and get interface mtu again

  char *env;
  asprintf (&env, "CONFBASE=%s", confbase);
  putenv (env);
  asprintf (&env, "NODENAME=%s", THISNODE->nodename);
  putenv (env);
  asprintf (&env, "NODEID=%d", THISNODE->id);
  putenv (env);
  asprintf (&env, "IFNAME=%s", tap->interface ());
  putenv (env);
  asprintf (&env, "MTU=%d", mtu);
  putenv (env);
  asprintf (&env, "MAC=%02x:%02x:%02x:%02x:%02x:%02x",
            0xfe, 0xfd, 0x80, 0x00, THISNODE->id >> 8,
            THISNODE->id & 0xff);
  putenv (env);

  return ::conf.script_if_up ? ::conf.script_if_up : "if-up";
}

int
vpn::setup ()
{
  u8 prots = 0;

  for (configuration::node_vector::iterator i = conf.nodes.begin ();
       i != conf.nodes.end (); ++i)
    prots |= (*i)->can_send | (*i)->can_recv;

  sockinfo si;

  si.set (THISNODE);

  udpv4_fd = -1;

  if (prots & PROT_UDPv4)
    {
      udpv4_fd = socket (PF_INET, SOCK_DGRAM, IPPROTO_UDP);

      if (udpv4_fd < 0)
        return -1;

      if (bind (udpv4_fd, si.sav4 (), si.salenv4 ()))
        {
          slog (L_ERR, _("can't bind udpv4 to %s: %s"), (const char *)si, strerror (errno));
          exit (1);
        }

#ifdef IP_MTU_DISCOVER
      // this I really consider a linux bug. I am neither connected
      // nor do I fragment myself. Linux still sets DF and doesn't
      // fragment for me sometimes.
      {
        int oval = IP_PMTUDISC_DONT;
        setsockopt (udpv4_fd, SOL_IP, IP_MTU_DISCOVER, &oval, sizeof oval);
      }
#endif

      // standard daemon practise...
      {
        int oval = 1;
        setsockopt (udpv4_fd, SOL_SOCKET, SO_REUSEADDR, &oval, sizeof oval);
      }

      udpv4_ev_watcher.start (udpv4_fd, POLLIN);
    }

  ipv4_fd = -1;
  if (prots & PROT_IPv4)
    {
      ipv4_fd = socket (PF_INET, SOCK_RAW, ::conf.ip_proto);

      if (ipv4_fd < 0)
        return -1;

      if (bind (ipv4_fd, si.sav4 (), si.salenv4 ()))
        {
          slog (L_ERR, _("can't bind ipv4 socket to %s: %s"), (const char *)si, strerror (errno));
          exit (1);
        }

#ifdef IP_MTU_DISCOVER
      // this I really consider a linux bug. I am neither connected
      // nor do I fragment myself. Linux still sets DF and doesn't
      // fragment for me sometimes.
      {
        int oval = IP_PMTUDISC_DONT;
        setsockopt (ipv4_fd, SOL_IP, IP_MTU_DISCOVER, &oval, sizeof oval);
      }
#endif

      ipv4_ev_watcher.start (ipv4_fd, POLLIN);
    }

  tap = new tap_device ();
  if (!tap) //D this, of course, never catches
    {
      slog (L_ERR, _("cannot create network interface '%s'"), conf.ifname);
      exit (1);
    }
  
  run_script (run_script_cb (this, &vpn::script_if_up), true);

  tap_ev_watcher.start (tap->fd, POLLIN);

  reconnect_all ();

  return 0;
}

void
vpn::send_ipv4_packet (vpn_packet *pkt, const sockinfo &si, int tos)
{
  setsockopt (ipv4_fd, SOL_IP, IP_TOS, &tos, sizeof tos);
  sendto (ipv4_fd, &((*pkt)[0]), pkt->len, 0, si.sav4 (), si.salenv4 ());
}

void
vpn::send_udpv4_packet (vpn_packet *pkt, const sockinfo &si, int tos)
{
  setsockopt (udpv4_fd, SOL_IP, IP_TOS, &tos, sizeof tos);
  sendto (udpv4_fd, &((*pkt)[0]), pkt->len, 0, si.sav4 (), si.salenv4 ());
}

void
vpn::recv_vpn_packet (vpn_packet *pkt, const sockinfo &rsi)
{
  unsigned int src = pkt->src ();
  unsigned int dst = pkt->dst ();

  slog (L_NOISE, _("<<?/%s received possible vpn packet type %d from %d to %d, length %d"),
        (const char *)rsi, pkt->typ (), pkt->src (), pkt->dst (), pkt->len);

  if (dst > conns.size () || pkt->typ () >= vpn_packet::PT_MAX)
    slog (L_WARN, _("<<? received CORRUPTED packet type %d from %d to %d"),
          pkt->typ (), pkt->src (), pkt->dst ());
  else if (dst == 0 && !THISNODE->routerprio)
    slog (L_WARN, _("<<%d received broadcast, but we are no router"), dst);
  else if (dst != 0 && dst != THISNODE->id)
    slog (L_WARN,
         _("received frame for node %d ('%s') from %s, but this is node %d ('%s')"),
         dst, conns[dst - 1]->conf->nodename,
         (const char *)rsi,
         THISNODE->id, THISNODE->nodename);
  else if (src == 0 || src > conns.size ())
    slog (L_WARN, _("received frame from unknown node %d (%s)"),
          src, (const char *)rsi);
  else
    conns[src - 1]->recv_vpn_packet (pkt, rsi);
}

void
vpn::udpv4_ev (short revents)
{
  if (revents & (POLLIN | POLLERR))
    {
      vpn_packet *pkt = new vpn_packet;
      struct sockaddr_in sa;
      socklen_t sa_len = sizeof (sa);
      int len;

      len = recvfrom (udpv4_fd, &((*pkt)[0]), MAXSIZE, 0, (sockaddr *)&sa, &sa_len);

      sockinfo si(sa);

      if (len > 0)
        {
          pkt->len = len;

          recv_vpn_packet (pkt, si);
        }
      else
        {
          // probably ECONNRESET or somesuch
          slog (L_DEBUG, _("%s: %s"), (const char *)si, strerror (errno));
        }

      delete pkt;
    }
  else if (revents & POLLHUP)
    {
      // this cannot ;) happen on udp sockets
      slog (L_ERR, _("FATAL: POLLHUP on udp v4 fd, terminating."));
      exit (1);
    }
  else
    {
      slog (L_ERR,
              _("FATAL: unknown revents %08x in socket, terminating\n"),
              revents);
      exit (1);
    }
}

void
vpn::ipv4_ev (short revents)
{
  if (revents & (POLLIN | POLLERR))
    {
      vpn_packet *pkt = new vpn_packet;
      struct sockaddr_in sa;
      socklen_t sa_len = sizeof (sa);
      int len;

      len = recvfrom (ipv4_fd, &((*pkt)[0]), MAXSIZE, 0, (sockaddr *)&sa, &sa_len);

      sockinfo si(sa, PROT_IPv4);

      if (len > 0)
        {
          pkt->len = len;

          // raw sockets deliver the ipv4, but don't expect it on sends
          // this is slow, but...
          pkt->skip_hdr (IP_OVERHEAD);

          recv_vpn_packet (pkt, si);
        }
      else
        {
          // probably ECONNRESET or somesuch
          slog (L_DEBUG, _("%s: %s"), (const char *)si, strerror (errno));
        }

      delete pkt;
    }
  else if (revents & POLLHUP)
    {
      // this cannot ;) happen on udp sockets
      slog (L_ERR, _("FATAL: POLLHUP on ipv4 fd, terminating."));
      exit (1);
    }
  else
    {
      slog (L_ERR,
              _("FATAL: unknown revents %08x in socket, terminating\n"),
              revents);
      exit (1);
    }
}

void
vpn::tap_ev (short revents)
{
  if (revents & POLLIN)
    {
      /* process data */
      tap_packet *pkt;

      pkt = tap->recv ();

      int dst = mac2id (pkt->dst);
      int src = mac2id (pkt->src);

      if (src != THISNODE->id)
        {
          slog (L_ERR, _("FATAL: tap packet not originating on current node received, terminating."));
          exit (1);
        }

      if (dst == THISNODE->id)
        {
          slog (L_ERR, _("FATAL: tap packet destined for current node received, terminating."));
          exit (1);
        }

      if (dst > conns.size ())
        slog (L_ERR, _("tap packet for unknown node %d received, ignoring."), dst);
      else
        {
          if (dst)
            {
              // unicast
              if (dst != THISNODE->id)
                conns[dst - 1]->inject_data_packet (pkt);
            }
          else
            {
              // broadcast, first check router, then self, then english
              connection *router = find_router ();

              if (router)
                router->inject_data_packet (pkt, true);
              else
                for (conns_vector::iterator c = conns.begin (); c != conns.end (); ++c)
                  if ((*c)->conf != THISNODE)
                    (*c)->inject_data_packet (pkt);
            }
        }

      delete pkt;
    }
  else if (revents & (POLLHUP | POLLERR))
    {
      slog (L_ERR, _("FATAL: POLLHUP or POLLERR on network device fd, terminating."));
      exit (1);
    }
  else
    abort ();
}

void
vpn::event_cb (tstamp &ts)
{
  if (events)
    {
      if (events & EVENT_SHUTDOWN)
        {
          slog (L_INFO, _("preparing shutdown..."));

          shutdown_all ();

          remove_pid (pidfilename);

          slog (L_INFO, _("terminating"));

          exit (0);
        }

      if (events & EVENT_RECONNECT)
        {
          slog (L_INFO, _("forced reconnect"));

          reconnect_all ();
        }

      events = 0;
    }

  ts = TSTAMP_CANCEL;
}

void
vpn::shutdown_all ()
{
  for (conns_vector::iterator c = conns.begin (); c != conns.end (); ++c)
    (*c)->shutdown ();
}

void
vpn::reconnect_all ()
{
  for (conns_vector::iterator c = conns.begin (); c != conns.end (); ++c)
    delete *c;

  conns.clear ();

  auth_rate_limiter.clear ();
  reset_rate_limiter.clear ();

  for (configuration::node_vector::iterator i = conf.nodes.begin ();
       i != conf.nodes.end (); ++i)
    {
      connection *conn = new connection (this);

      conn->conf = *i;
      conns.push_back (conn);

      conn->establish_connection ();
    }
}

connection *vpn::find_router ()
{
  u32 prio = 0;
  connection *router = 0;

  for (conns_vector::iterator i = conns.begin (); i != conns.end (); ++i)
    {
      connection *c = *i;

      if (c->conf->routerprio > prio
          && c->connectmode == conf_node::C_ALWAYS
          && c->conf != THISNODE
          && c->ictx && c->octx)
        {
          prio = c->conf->routerprio;
          router = c;
        }
    }

  return router;
}

void vpn::connect_request (int id)
{
  connection *c = find_router ();

  if (c)
    c->connect_request (id);
  //else // does not work, because all others must connect to the same router
  //  // no router found, aggressively connect to all routers
  //  for (conns_vector::iterator i = conns.begin (); i != conns.end (); ++i)
  //    if ((*i)->conf->routerprio)
  //      (*i)->establish_connection ();
}

void
connection::dump_status ()
{
  slog (L_NOTICE, _("node %s (id %d)"), conf->nodename, conf->id);
  slog (L_NOTICE, _("  connectmode %d (%d) / sockaddr %s / minor %d"),
        connectmode, conf->connectmode, (const char *)si, (int)prot_minor);
  slog (L_NOTICE, _("  ictx/octx %08lx/%08lx / oseqno %d / retry_cnt %d"),
        (long)ictx, (long)octx, (int)oseqno, (int)retry_cnt);
  slog (L_NOTICE, _("  establish_conn %ld / rekey %ld / keepalive %ld"),
        (long)(establish_connection.at), (long)(rekey.at), (long)(keepalive.at));
}

void
vpn::dump_status ()
{
  slog (L_NOTICE, _("BEGIN status dump (%ld)"), (long)NOW);

  for (conns_vector::iterator c = conns.begin (); c != conns.end (); ++c)
    (*c)->dump_status ();

  slog (L_NOTICE, _("END status dump"));
}

vpn::vpn (void)
: udpv4_ev_watcher(this, &vpn::udpv4_ev)
, ipv4_ev_watcher(this, &vpn::ipv4_ev)
, tap_ev_watcher(this, &vpn::tap_ev)
, event(this, &vpn::event_cb)
{
}

vpn::~vpn ()
{
}

Revision:	1.23
Committed:	Fri Mar 28 05:40:54 2003 UTC (21 years, 2 months ago) by pcg
Content type:	text/plain
Branch:	MAIN
Changes since 1.22:	+1 -2 lines
Log Message:	* empty log message *