usr.sbin/ppp/ip.c

/*
 *		PPP IP Protocol Interface
 *
 *	    Written by Toshiharu OHNO (tony-o@iij.ad.jp)
 *
 *   Copyright (C) 1993, Internet Initiative Japan, Inc. All rights reserverd.
 *
 * Redistribution and use in source and binary forms are permitted
 * provided that the above copyright notice and this paragraph are
 * duplicated in all such forms and that any documentation,
 * advertising materials, and other materials related to such
 * distribution and use acknowledge that the software was developed
 * by the Internet Initiative Japan.  The name of the
 * IIJ may not be used to endorse or promote products derived
 * from this software without specific prior written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
 * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 *
 * $Id: ip.c,v 1.65 1999/07/27 23:43:59 brian Exp $
 *
 *	TODO:
 *		o Return ICMP message for filterd packet
 *		  and optionaly record it into log.
 */
#include <sys/param.h>
#if defined(__OpenBSD__) || defined(__NetBSD__)
#include <sys/socket.h>
#endif
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/ip_icmp.h>
#include <netinet/udp.h>
#include <netinet/tcp.h>
#include <arpa/inet.h>
#include <sys/un.h>

#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <termios.h>
#include <unistd.h>

#include "layer.h"
#include "proto.h"
#include "mbuf.h"
#include "log.h"
#include "defs.h"
#include "timer.h"
#include "fsm.h"
#include "lqr.h"
#include "hdlc.h"
#include "throughput.h"
#include "iplist.h"
#include "slcompress.h"
#include "ipcp.h"
#include "filter.h"
#include "descriptor.h"
#include "lcp.h"
#include "ccp.h"
#include "link.h"
#include "mp.h"
#ifndef NORADIUS
#include "radius.h"
#endif
#include "bundle.h"
#include "vjcomp.h"
#include "tun.h"
#include "ip.h"

static const u_short interactive_ports[32] = {
  544, 513, 514, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  80, 81, 0, 0, 0, 21, 22, 23, 0, 0, 0, 0, 0, 0, 0, 543,
};

#define	INTERACTIVE(p)	(interactive_ports[(p) & 0x1F] == (p))

static const char *TcpFlags[] = { "FIN", "SYN", "RST", "PSH", "ACK", "URG" };

static __inline int
PortMatch(int op, u_short pport, u_short rport)
{
  switch (op) {
  case OP_EQ:
    return (pport == rport);
  case OP_GT:
    return (pport > rport);
  case OP_LT:
    return (pport < rport);
  default:
    return (0);
  }
}

/*
 *  Check a packet against a defined filter
 *  Returns 0 to accept the packet, non-zero to drop the packet
 *
 *  If filtering is enabled, the initial fragment of a datagram must
 *  contain the complete protocol header, and subsequent fragments
 *  must not attempt to over-write it.
 */
static int
FilterCheck(const struct ip *pip, const struct filter *filter)
{
  int gotinfo;			/* true if IP payload decoded */
  int cproto;			/* P_* protocol type if (gotinfo) */
  int estab, syn, finrst;	/* TCP state flags if (gotinfo) */
  u_short sport, dport;		/* src, dest port from packet if (gotinfo) */
  int n;			/* filter rule to process */
  int len;			/* bytes used in dbuff */
  int didname;			/* true if filter header printed */
  int match;			/* true if condition matched */
  const struct filterent *fp = filter->rule;
  char dbuff[100];

  if (fp->f_action == A_NONE)
    return (0);		/* No rule is given. Permit this packet */

  /* Deny any packet fragment that tries to over-write the header.
   * Since we no longer have the real header available, punt on the
   * largest normal header - 20 bytes for TCP without options, rounded
   * up to the next possible fragment boundary.  Since the smallest
   * `legal' MTU is 576, and the smallest recommended MTU is 296, any
   * fragmentation within this range is dubious at best */
  len = ntohs(pip->ip_off) & IP_OFFMASK;	/* fragment offset */
  if (len > 0) {		/* Not first fragment within datagram */
    if (len < (24 >> 3))	/* don't allow fragment to over-write header */
      return (1);
    /* permit fragments on in and out filter */
    return (filter->fragok);
  }

  cproto = gotinfo = estab = syn = finrst = didname = 0;
  sport = dport = 0;
  for (n = 0; n < MAXFILTERS; ) {
    if (fp->f_action == A_NONE) {
      n++;
      fp++;
      continue;
    }

    if (!didname) {
      log_Printf(LogDEBUG, "%s filter:\n", filter->name);
      didname = 1;
    }

    match = 0;
    if (!((pip->ip_src.s_addr ^ fp->f_src.ipaddr.s_addr) &
	  fp->f_src.mask.s_addr) &&
	!((pip->ip_dst.s_addr ^ fp->f_dst.ipaddr.s_addr) &
	  fp->f_dst.mask.s_addr)) {
      if (fp->f_proto != P_NONE) {
	if (!gotinfo) {
	  const char *ptop = (const char *) pip + (pip->ip_hl << 2);
	  const struct tcphdr *th;
	  const struct udphdr *uh;
	  const struct icmp *ih;
	  int datalen;	/* IP datagram length */

	  datalen = ntohs(pip->ip_len) - (pip->ip_hl << 2);
	  switch (pip->ip_p) {
	  case IPPROTO_ICMP:
	    cproto = P_ICMP;
	    if (datalen < 8)	/* ICMP must be at least 8 octets */
	      return (1);
	    ih = (const struct icmp *) ptop;
	    sport = ih->icmp_type;
	    estab = syn = finrst = -1;
	    if (log_IsKept(LogDEBUG))
	      snprintf(dbuff, sizeof dbuff, "sport = %d", sport);
	    break;
	  case IPPROTO_IGMP:
	    cproto = P_IGMP;
	    if (datalen < 8)	/* IGMP uses 8-octet messages */
	      return (1);
	    estab = syn = finrst = -1;
	    sport = ntohs(0);
	    break;
	  case IPPROTO_OSPFIGP:
	    cproto = P_OSPF;
	    if (datalen < 8)	/* IGMP uses 8-octet messages */
	      return (1);
	    estab = syn = finrst = -1;
	    sport = ntohs(0);
	    break;
	  case IPPROTO_UDP:
	  case IPPROTO_IPIP:
	    cproto = P_UDP;
	    if (datalen < 8)	/* UDP header is 8 octets */
	      return (1);
	    uh = (const struct udphdr *) ptop;
	    sport = ntohs(uh->uh_sport);
	    dport = ntohs(uh->uh_dport);
	    estab = syn = finrst = -1;
	    if (log_IsKept(LogDEBUG))
	      snprintf(dbuff, sizeof dbuff, "sport = %d, dport = %d",
		       sport, dport);
	    break;
	  case IPPROTO_TCP:
	    cproto = P_TCP;
	    th = (const struct tcphdr *) ptop;
	    /* TCP headers are variable length.  The following code
	     * ensures that the TCP header length isn't de-referenced if
	     * the datagram is too short
	     */
	    if (datalen < 20 || datalen < (th->th_off << 2))
	      return (1);
	    sport = ntohs(th->th_sport);
	    dport = ntohs(th->th_dport);
	    estab = (th->th_flags & TH_ACK);
	    syn = (th->th_flags & TH_SYN);
	    finrst = (th->th_flags & (TH_FIN|TH_RST));
	    if (log_IsKept(LogDEBUG)) {
	      if (!estab)
		snprintf(dbuff, sizeof dbuff,
			 "flags = %02x, sport = %d, dport = %d",
			 th->th_flags, sport, dport);
	      else
		*dbuff = '\0';
	    }
	    break;
	  default:
	    return (1);	/* We'll block unknown type of packet */
	  }

	  if (log_IsKept(LogDEBUG)) {
	    if (estab != -1) {
	      len = strlen(dbuff);
	      snprintf(dbuff + len, sizeof dbuff - len,
		       ", estab = %d, syn = %d, finrst = %d",
		       estab, syn, finrst);
	    }
	    log_Printf(LogDEBUG, " Filter: proto = %s, %s\n",
		       filter_Proto2Nam(cproto), dbuff);
	  }
	  gotinfo = 1;
	}
	if (log_IsKept(LogDEBUG)) {
	  if (fp->f_srcop != OP_NONE) {
	    snprintf(dbuff, sizeof dbuff, ", src %s %d",
		     filter_Op2Nam(fp->f_srcop), fp->f_srcport);
	    len = strlen(dbuff);
	  } else
	    len = 0;
	  if (fp->f_dstop != OP_NONE) {
	    snprintf(dbuff + len, sizeof dbuff - len,
		     ", dst %s %d", filter_Op2Nam(fp->f_dstop),
		     fp->f_dstport);
	  } else if (!len)
	    *dbuff = '\0';

	  log_Printf(LogDEBUG, "  rule = %d: Address match, "
		     "check against proto %s%s, action = %s\n",
		     n, filter_Proto2Nam(fp->f_proto),
		     dbuff, filter_Action2Nam(fp->f_action));
	}

	if (cproto == fp->f_proto) {
	  if ((fp->f_srcop == OP_NONE ||
	       PortMatch(fp->f_srcop, sport, fp->f_srcport)) &&
	      (fp->f_dstop == OP_NONE ||
	       PortMatch(fp->f_dstop, dport, fp->f_dstport)) &&
	      (fp->f_estab == 0 || estab) &&
	      (fp->f_syn == 0 || syn) &&
	      (fp->f_finrst == 0 || finrst)) {
	    match = 1;
	  }
	}
      } else {
	/* Address is matched and no protocol specified. Make a decision. */
	log_Printf(LogDEBUG, "  rule = %d: Address match, action = %s\n", n,
		   filter_Action2Nam(fp->f_action));
	match = 1;
      }
    } else
      log_Printf(LogDEBUG, "  rule = %d: Address mismatch\n", n);

    if (match != fp->f_invert) {
      /* Take specified action */
      if (fp->f_action < A_NONE)
	fp = &filter->rule[n = fp->f_action];
      else
	return (fp->f_action != A_PERMIT);
    } else {
      n++;
      fp++;
    }
  }
  return (1);		/* No rule is mached. Deny this packet */
}

#ifdef notdef
static void
IcmpError(struct ip *pip, int code)
{
  struct mbuf *bp;

  if (pip->ip_p != IPPROTO_ICMP) {
    bp = mbuf_Alloc(cnt, MB_IPIN);
    memcpy(MBUF_CTOP(bp), ptr, cnt);
    vj_SendFrame(bp);
    ipcp_AddOutOctets(cnt);
  }
}
#endif

/*
 *  For debugging aid.
 */
int
PacketCheck(struct bundle *bundle, char *cp, int nb, struct filter *filter)
{
  struct ip *pip;
  struct tcphdr *th;
  struct udphdr *uh;
  struct icmp *icmph;
  char *ptop;
  int mask, len, n;
  int pri = PRI_NORMAL;
  int logit, loglen;
  char logbuf[200];

  logit = log_IsKept(LogTCPIP) && filter->logok;
  loglen = 0;

  pip = (struct ip *) cp;

  if (logit && loglen < sizeof logbuf) {
    snprintf(logbuf + loglen, sizeof logbuf - loglen, "%s ", filter->name);
    loglen += strlen(logbuf + loglen);
  }
  ptop = (cp + (pip->ip_hl << 2));

  switch (pip->ip_p) {
  case IPPROTO_ICMP:
    if (logit && loglen < sizeof logbuf) {
      icmph = (struct icmp *) ptop;
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	     "ICMP: %s:%d ---> ", inet_ntoa(pip->ip_src), icmph->icmp_type);
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s:%d", inet_ntoa(pip->ip_dst), icmph->icmp_type);
      loglen += strlen(logbuf + loglen);
    }
    break;
  case IPPROTO_UDP:
    if (logit && loglen < sizeof logbuf) {
      uh = (struct udphdr *) ptop;
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	   "UDP: %s:%d ---> ", inet_ntoa(pip->ip_src), ntohs(uh->uh_sport));
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s:%d", inet_ntoa(pip->ip_dst), ntohs(uh->uh_dport));
      loglen += strlen(logbuf + loglen);
    }
    break;
  case IPPROTO_OSPFIGP:
    if (logit && loglen < sizeof logbuf) {
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	   "OSPF: %s ---> ", inet_ntoa(pip->ip_src));
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s", inet_ntoa(pip->ip_dst));
      loglen += strlen(logbuf + loglen);
    }
    break;
  case IPPROTO_IPIP:
    if (logit && loglen < sizeof logbuf) {
      uh = (struct udphdr *) ptop;
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	   "IPIP: %s:%d ---> ", inet_ntoa(pip->ip_src), ntohs(uh->uh_sport));
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s:%d", inet_ntoa(pip->ip_dst), ntohs(uh->uh_dport));
      loglen += strlen(logbuf + loglen);
    }
    break;
  case IPPROTO_IGMP:
    if (logit && loglen < sizeof logbuf) {
      uh = (struct udphdr *) ptop;
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	   "IGMP: %s:%d ---> ", inet_ntoa(pip->ip_src), ntohs(uh->uh_sport));
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s:%d", inet_ntoa(pip->ip_dst), ntohs(uh->uh_dport));
      loglen += strlen(logbuf + loglen);
    }
    break;
  case IPPROTO_TCP:
    th = (struct tcphdr *) ptop;
    if (pip->ip_tos == IPTOS_LOWDELAY)
      pri = PRI_FAST;
    else if ((ntohs(pip->ip_off) & IP_OFFMASK) == 0) {
      if (INTERACTIVE(ntohs(th->th_sport)) || INTERACTIVE(ntohs(th->th_dport)))
	pri = PRI_FAST;
    }
    if (logit && loglen < sizeof logbuf) {
      len = ntohs(pip->ip_len) - (pip->ip_hl << 2) - (th->th_off << 2);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	   "TCP: %s:%d ---> ", inet_ntoa(pip->ip_src), ntohs(th->th_sport));
      loglen += strlen(logbuf + loglen);
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "%s:%d", inet_ntoa(pip->ip_dst), ntohs(th->th_dport));
      loglen += strlen(logbuf + loglen);
      n = 0;
      for (mask = TH_FIN; mask != 0x40; mask <<= 1) {
	if (th->th_flags & mask) {
	  snprintf(logbuf + loglen, sizeof logbuf - loglen, " %s", TcpFlags[n]);
	  loglen += strlen(logbuf + loglen);
	}
	n++;
      }
      snprintf(logbuf + loglen, sizeof logbuf - loglen,
	       "  seq:%lx  ack:%lx (%d/%d)",
	       (u_long)ntohl(th->th_seq), (u_long)ntohl(th->th_ack), len, nb);
      loglen += strlen(logbuf + loglen);
      if ((th->th_flags & TH_SYN) && nb > 40) {
	u_short *sp;

	ptop += 20;
	sp = (u_short *) ptop;
	if (ntohs(sp[0]) == 0x0204) {
	  snprintf(logbuf + loglen, sizeof logbuf - loglen,
		   " MSS = %d", ntohs(sp[1]));
	  loglen += strlen(logbuf + loglen);
	}
      }
    }
    break;
  }

  if (FilterCheck(pip, filter)) {
    if (logit)
      log_Printf(LogTCPIP, "%s - BLOCKED\n", logbuf);
#ifdef notdef
    if (direction == 0)
      IcmpError(pip, pri);
#endif
    return (-1);
  } else {
    /* Check Keep Alive filter */
    if (logit) {
      if (FilterCheck(pip, &bundle->filter.alive))
        log_Printf(LogTCPIP, "%s - NO KEEPALIVE\n", logbuf);
      else
        log_Printf(LogTCPIP, "%s\n", logbuf);
    }
    return (pri);
  }
}

struct mbuf *
ip_Input(struct bundle *bundle, struct link *l, struct mbuf *bp)
{
  int nb, nw;
  struct tun_data tun;
  struct ip *pip;

  if (bundle->ncp.ipcp.fsm.state != ST_OPENED) {
    log_Printf(LogWARN, "ip_Input: IPCP not open - packet dropped\n");
    mbuf_Free(bp);
    return NULL;
  }

  mbuf_SetType(bp, MB_IPIN);
  tun_fill_header(tun, AF_INET);
  nb = mbuf_Length(bp);
  if (nb > sizeof tun.data) {
    log_Printf(LogWARN, "ip_Input: %s: Packet too large (got %d, max %d)\n",
               l->name, nb, (int)(sizeof tun.data));
    mbuf_Free(bp);
    return NULL;
  }
  mbuf_Read(bp, tun.data, nb);

  if (PacketCheck(bundle, tun.data, nb, &bundle->filter.in) < 0)
    return NULL;

  pip = (struct ip *)tun.data;
  if (!FilterCheck(pip, &bundle->filter.alive))
    bundle_StartIdleTimer(bundle);

  ipcp_AddInOctets(&bundle->ncp.ipcp, nb);

  nb += sizeof tun - sizeof tun.data;
  nw = write(bundle->dev.fd, &tun, nb);
  if (nw != nb) {
    if (nw == -1)
      log_Printf(LogERROR, "ip_Input: %s: wrote %d, got %s\n",
                 l->name, nb, strerror(errno));
    else
      log_Printf(LogERROR, "ip_Input: %s: wrote %d, got %d\n", l->name, nb, nw);
  }

  return NULL;
}

void
ip_Enqueue(struct ipcp *ipcp, int pri, char *ptr, int count)
{
  struct mbuf *bp;

  if (pri < 0 || pri > sizeof ipcp->Queue / sizeof ipcp->Queue[0])
    log_Printf(LogERROR, "Can't store in ip queue %d\n", pri);
  else {
    /*
     * We allocate an extra 6 bytes, four at the front and two at the end.
     * This is an optimisation so that we need to do less work in
     * mbuf_Prepend() in acf_LayerPush() and proto_LayerPush() and
     * appending in hdlc_LayerPush().
     */
    bp = mbuf_Alloc(count + 6, MB_IPOUT);
    bp->offset += 4;
    bp->cnt -= 6;
    memcpy(MBUF_CTOP(bp), ptr, count);
    mbuf_Enqueue(&ipcp->Queue[pri], bp);
  }
}

void
ip_DeleteQueue(struct ipcp *ipcp)
{
  struct mqueue *queue;

  for (queue = ipcp->Queue; queue < ipcp->Queue + PRI_MAX; queue++)
    while (queue->top)
      mbuf_Free(mbuf_Dequeue(queue));
}

int
ip_QueueLen(struct ipcp *ipcp)
{
  struct mqueue *queue;
  int result = 0;

  for (queue = ipcp->Queue; queue < ipcp->Queue + PRI_MAX; queue++)
    result += queue->qlen;

  return result;
}

int
ip_PushPacket(struct link *l, struct bundle *bundle)
{
  struct ipcp *ipcp = &bundle->ncp.ipcp;
  struct mqueue *queue;
  struct mbuf *bp;
  struct ip *pip;
  int cnt;

  if (ipcp->fsm.state != ST_OPENED)
    return 0;

  for (queue = &ipcp->Queue[PRI_FAST]; queue >= ipcp->Queue; queue--)
    if (queue->top) {
      bp = mbuf_Contiguous(mbuf_Dequeue(queue));
      cnt = mbuf_Length(bp);
      pip = (struct ip *)MBUF_CTOP(bp);
      if (!FilterCheck(pip, &bundle->filter.alive))
        bundle_StartIdleTimer(bundle);
      link_PushPacket(l, bp, bundle, PRI_NORMAL, PROTO_IP);
      ipcp_AddOutOctets(ipcp, cnt);
      return 1;
    }

  return 0;
}