projects / unix-history / blobdiff
? search:
summary | tags | clone url | shortlog | log | commit | commitdiff | tree
raw | inline | side by side
localize header files
[unix-history] / usr / src / sys / netinet / ip_input.c
diff --git a/usr/src/sys/netinet/ip_input.c b/usr/src/sys/netinet/ip_input.c
index 9808dc1..cf9864f 100644 (file)
--- a/usr/src/sys/netinet/ip_input.c
+++ b/usr/src/sys/netinet/ip_input.c
@@ -1,23 +1,28 @@
-/* ip_input.c 1.15 81/11/18 */
+/*     ip_input.c      1.51    82/10/09        */
 
 #include "../h/param.h"
 #include "../h/systm.h"
 
 #include "../h/param.h"
 #include "../h/systm.h"
-#include "../h/clock.h"
 #include "../h/mbuf.h"
 #include "../h/protosw.h"
 #include "../h/socket.h"
 #include "../h/mbuf.h"
 #include "../h/protosw.h"
 #include "../h/socket.h"
-#include "../net/inet.h"
-#include "../net/inet_systm.h"
-#include "../net/imp.h"
-#include "../net/ip.h"                 /* belongs before inet.h */
-#include "../net/ip_var.h"
-#include "../net/ip_icmp.h"
-#include "../net/tcp.h"
+#include "../netinet/in.h"
+#include "../netinet/in_systm.h"
+#include "../net/if.h"
+#include "../netinet/ip.h"                     /* belongs before in.h */
+#include "../netinet/ip_var.h"
+#include "../netinet/ip_icmp.h"
+#include "../netinet/tcp.h"
+#include <time.h>
+#include "../h/kernel.h"
+#include <errno.h>
 
 u_char ip_protox[IPPROTO_MAX];
 
 u_char ip_protox[IPPROTO_MAX];
+int    ipqmaxlen = IFQ_MAXLEN;
+struct ifnet *ifinet;                  /* first inet interface */
 
 /*
 
 /*
- * Ip initialization.
+ * IP initialization: fill in IP protocol switch table.
+ * All protocols not implemented in kernel go to raw IP protocol handler.
  */
 ip_init()
 {
  */
 ip_init()
 {
@@ -34,52 +39,63 @@ ip_init()
                    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
                        ip_protox[pr->pr_protocol] = pr - protosw;
        ipq.next = ipq.prev = &ipq;
                    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
                        ip_protox[pr->pr_protocol] = pr - protosw;
        ipq.next = ipq.prev = &ipq;
-       ip_id = time & 0xffff;
+       ip_id = time.tv_sec & 0xffff;
+       ipintrq.ifq_maxlen = ipqmaxlen;
+       ifinet = if_ifwithaf(AF_INET);
 }
 
 u_char ipcksum = 1;
 struct ip *ip_reass();
 }
 
 u_char ipcksum = 1;
 struct ip *ip_reass();
-
-/*
- * Ip input routines.
- */
+struct sockaddr_in ipaddr = { AF_INET };
 
 /*
  * Ip input routine.  Checksum and byte swap header.  If fragmented
  * try to reassamble.  If complete and fragment queue exists, discard.
  * Process options.  Pass to next level.
  */
 
 /*
  * Ip input routine.  Checksum and byte swap header.  If fragmented
  * try to reassamble.  If complete and fragment queue exists, discard.
  * Process options.  Pass to next level.
  */
-ip_input(m0)
-       struct mbuf *m0;
+ipintr()
 {
        register struct ip *ip;
 {
        register struct ip *ip;
-       register struct mbuf *m = m0;
+       register struct mbuf *m;
+       struct mbuf *m0, *mopt;
        register int i;
        register struct ipq *fp;
        register int i;
        register struct ipq *fp;
-       int hlen;
+       int hlen, s;
 
 
-COUNT(IP_INPUT);
+next:
        /*
        /*
-        * Check header and byteswap.
+        * Get next datagram off input queue and get IP header
+        * in first mbuf.
         */
         */
+       s = splimp();
+       IF_DEQUEUE(&ipintrq, m);
+       splx(s);
+       if (m == 0)
+               return;
+       if ((m->m_off > MMAXOFF || m->m_len < sizeof (struct ip)) &&
+           (m = m_pullup(m, sizeof (struct ip))) == 0)
+               return;
        ip = mtod(m, struct ip *);
        if ((hlen = ip->ip_hl << 2) > m->m_len) {
        ip = mtod(m, struct ip *);
        if ((hlen = ip->ip_hl << 2) > m->m_len) {
-               printf("ip hdr ovflo\n");
-               m_freem(m);
-               return;
-       }
-       ip->ip_sum = inet_cksum(m, hlen);
-       if (ip->ip_sum) {
-               printf("ip_sum %x\n", ip->ip_sum);
-               ipstat.ips_badsum++;
-               if (ipcksum) {
-                       m_freem(m);
+               if ((m = m_pullup(m, hlen)) == 0)
                        return;
                        return;
-               }
+               ip = mtod(m, struct ip *);
        }
        }
+       if (ipcksum)
+               if (ip->ip_sum = in_cksum(m, hlen)) {
+                       printf("ip_sum %x\n", ip->ip_sum);      /* XXX */
+                       ipstat.ips_badsum++;
+                       goto bad;
+               }
+
+#if vax
+       /*
+        * Convert fields to host representation.
+        */
        ip->ip_len = ntohs((u_short)ip->ip_len);
        ip->ip_id = ntohs(ip->ip_id);
        ip->ip_len = ntohs((u_short)ip->ip_len);
        ip->ip_id = ntohs(ip->ip_id);
-       ip->ip_off = ntohs(ip->ip_off);
+       ip->ip_off = ntohs((u_short)ip->ip_off);
+#endif
 
        /*
         * Check that the amount of data in the buffers
 
        /*
         * Check that the amount of data in the buffers
@@ -87,34 +103,57 @@ COUNT(IP_INPUT);
         * Trim mbufs if longer than we expect.
         * Drop packet if shorter than we expect.
         */
         * Trim mbufs if longer than we expect.
         * Drop packet if shorter than we expect.
         */
-       i = 0;
-       for (; m != NULL; m = m->m_next)
+       i = -ip->ip_len;
+       m0 = m;
+       for (;;) {
                i += m->m_len;
                i += m->m_len;
-       m = m0;
-       if (i != ip->ip_len) {
-               if (i < ip->ip_len) {
-                       printf("ip_input: short packet\n");
-                       m_freem(m);
-                       return;
+               if (m->m_next == 0)
+                       break;
+               m = m->m_next;
+       }
+       if (i != 0) {
+               if (i < 0) {
+                       ipstat.ips_tooshort++;
+                       goto bad;
                }
                }
-               m_adj(m, ip->ip_len - i);
+               if (i <= m->m_len)
+                       m->m_len -= i;
+               else
+                       m_adj(m0, -i);
        }
        }
+       m = m0;
 
        /*
         * Process options and, if not destined for us,
 
        /*
         * Process options and, if not destined for us,
-        * ship it on.
+        * ship it on.  ip_dooptions returns 1 when an
+        * error was detected (causing an icmp message
+        * to be sent).
         */
         */
-       if (hlen > sizeof (struct ip))
-               ip_dooptions(ip);
-       if (ip->ip_dst.s_addr != n_lhost.s_addr) {
-               if (--ip->ip_ttl == 0) {
-                       icmp_error(ip, ICMP_TIMXCEED, 0);
-                       return;
-               }
-               ip_output(dtom(ip));
-               return;
+       if (hlen > sizeof (struct ip) && ip_dooptions(ip))
+               goto next;
+
+       /*
+        * Fast check on the first internet
+        * interface in the list.
+        */
+       if (ifinet) {
+               struct sockaddr_in *sin;
+
+               sin = (struct sockaddr_in *)&ifinet->if_addr;
+               if (sin->sin_addr.s_addr == ip->ip_dst.s_addr)
+                       goto ours;
+               sin = (struct sockaddr_in *)&ifinet->if_broadaddr;
+               if ((ifinet->if_flags & IFF_BROADCAST) &&
+                   sin->sin_addr.s_addr == ip->ip_dst.s_addr)
+                       goto ours;
+       }
+       ipaddr.sin_addr = ip->ip_dst;
+       if (if_ifwithaddr((struct sockaddr *)&ipaddr) == 0) {
+               ip_forward(ip);
+               goto next;
        }
 
        }
 
+ours:
        /*
         * Look for queue of fragments
         * of this datagram.
        /*
         * Look for queue of fragments
         * of this datagram.
@@ -147,18 +186,26 @@ found:
        if (((struct ipasfrag *)ip)->ipf_mff || ip->ip_off) {
                ip = ip_reass((struct ipasfrag *)ip, fp);
                if (ip == 0)
        if (((struct ipasfrag *)ip)->ipf_mff || ip->ip_off) {
                ip = ip_reass((struct ipasfrag *)ip, fp);
                if (ip == 0)
-                       return;
+                       goto next;
                hlen = ip->ip_hl << 2;
                m = dtom(ip);
        } else
                if (fp)
                        (void) ip_freef(fp);
                hlen = ip->ip_hl << 2;
                m = dtom(ip);
        } else
                if (fp)
                        (void) ip_freef(fp);
+
+       /*
+        * Switch out to protocol's input routine.
+        */
        (*protosw[ip_protox[ip->ip_p]].pr_input)(m);
        (*protosw[ip_protox[ip->ip_p]].pr_input)(m);
+       goto next;
+bad:
+       m_freem(m);
+       goto next;
 }
 
 /*
  * Take incoming datagram fragment and try to
 }
 
 /*
  * Take incoming datagram fragment and try to
- * reassamble it into whole datagram.  If a chain for
+ * reassemble it into whole datagram.  If a chain for
  * reassembly of this datagram already exists, then it
  * is given as fp; otherwise have to make a chain.
  */
  * reassembly of this datagram already exists, then it
  * is given as fp; otherwise have to make a chain.
  */
@@ -184,9 +231,8 @@ ip_reass(ip, fp)
         * If first fragment to arrive, create a reassembly queue.
         */
        if (fp == 0) {
         * If first fragment to arrive, create a reassembly queue.
         */
        if (fp == 0) {
-               if ((t = m_get(1)) == NULL)
+               if ((t = m_get(M_WAIT)) == NULL)
                        goto dropfrag;
                        goto dropfrag;
-               t->m_off = MMINOFF;
                fp = mtod(t, struct ipq *);
                insque(fp, &ipq);
                fp->ipq_ttl = IPFRAGTTL;
                fp = mtod(t, struct ipq *);
                insque(fp, &ipq);
                fp->ipq_ttl = IPFRAGTTL;
@@ -195,6 +241,8 @@ ip_reass(ip, fp)
                fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
                fp->ipq_src = ((struct ip *)ip)->ip_src;
                fp->ipq_dst = ((struct ip *)ip)->ip_dst;
                fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
                fp->ipq_src = ((struct ip *)ip)->ip_src;
                fp->ipq_dst = ((struct ip *)ip)->ip_dst;
+               q = (struct ipasfrag *)fp;
+               goto insert;
        }
 
        /*
        }
 
        /*
@@ -228,6 +276,7 @@ ip_reass(ip, fp)
                i = (ip->ip_off + ip->ip_len) - q->ip_off;
                if (i < q->ip_len) {
                        q->ip_len -= i;
                i = (ip->ip_off + ip->ip_len) - q->ip_off;
                if (i < q->ip_len) {
                        q->ip_len -= i;
+                       q->ip_off += i;
                        m_adj(dtom(q), i);
                        break;
                }
                        m_adj(dtom(q), i);
                        break;
                }
@@ -236,6 +285,7 @@ ip_reass(ip, fp)
                ip_deq(q->ipf_prev);
        }
 
                ip_deq(q->ipf_prev);
        }
 
+insert:
        /*
         * Stick new segment in its place;
         * check for complete reassembly.
        /*
         * Stick new segment in its place;
         * check for complete reassembly.
@@ -258,8 +308,12 @@ ip_reass(ip, fp)
        t = m->m_next;
        m->m_next = 0;
        m_cat(m, t);
        t = m->m_next;
        m->m_next = 0;
        m_cat(m, t);
-       while ((q = q->ipf_next) != (struct ipasfrag *)fp)
-               m_cat(m, dtom(q));
+       q = q->ipf_next;
+       while (q != (struct ipasfrag *)fp) {
+               t = dtom(q);
+               q = q->ipf_next;
+               m_cat(m, t);
+       }
 
        /*
         * Create header for new ip packet by
 
        /*
         * Create header for new ip packet by
@@ -310,7 +364,6 @@ ip_freef(fp)
 ip_enq(p, prev)
        register struct ipasfrag *p, *prev;
 {
 ip_enq(p, prev)
        register struct ipasfrag *p, *prev;
 {
-COUNT(IP_ENQ);
 
        p->ipf_prev = prev;
        p->ipf_next = prev->ipf_next;
 
        p->ipf_prev = prev;
        p->ipf_next = prev->ipf_next;
@@ -324,7 +377,6 @@ COUNT(IP_ENQ);
 ip_deq(p)
        register struct ipasfrag *p;
 {
 ip_deq(p)
        register struct ipasfrag *p;
 {
-COUNT(IP_DEQ);
 
        p->ipf_prev->ipf_next = p->ipf_next;
        p->ipf_next->ipf_prev = p->ipf_prev;
 
        p->ipf_prev->ipf_next = p->ipf_next;
        p->ipf_next->ipf_prev = p->ipf_prev;
@@ -339,9 +391,13 @@ ip_slowtimo()
 {
        register struct ipq *fp;
        int s = splnet();
 {
        register struct ipq *fp;
        int s = splnet();
-COUNT(IP_SLOWTIMO);
 
 
-       for (fp = ipq.next; fp != &ipq; )
+       fp = ipq.next;
+       if (fp == 0) {
+               splx(s);
+               return;
+       }
+       while (fp != &ipq)
                if (--fp->ipq_ttl == 0)
                        fp = ip_freef(fp);
                else
                if (--fp->ipq_ttl == 0)
                        fp = ip_freef(fp);
                else
@@ -349,9 +405,14 @@ COUNT(IP_SLOWTIMO);
        splx(s);
 }
 
        splx(s);
 }
 
+/*
+ * Drain off all datagram fragments.
+ */
 ip_drain()
 {
 
 ip_drain()
 {
 
+       while (ipq.next != &ipq)
+               (void) ip_freef(ipq.next);
 }
 
 /*
 }
 
 /*
@@ -363,9 +424,11 @@ ip_dooptions(ip)
        struct ip *ip;
 {
        register u_char *cp;
        struct ip *ip;
 {
        register u_char *cp;
-       int opt, optlen, cnt;
+       int opt, optlen, cnt, code, type;
        struct in_addr *sin;
        register struct ip_timestamp *ipt;
        struct in_addr *sin;
        register struct ip_timestamp *ipt;
+       register struct ifnet *ifp;
+       struct in_addr t;
 
        cp = (u_char *)(ip + 1);
        cnt = (ip->ip_hl << 2) - sizeof (struct ip);
 
        cp = (u_char *)(ip + 1);
        cnt = (ip->ip_hl << 2) - sizeof (struct ip);
@@ -382,22 +445,41 @@ ip_dooptions(ip)
                default:
                        break;
 
                default:
                        break;
 
+               /*
+                * Source routing with record.
+                * Find interface with current destination address.
+                * If none on this machine then drop if strictly routed,
+                * or do nothing if loosely routed.
+                * Record interface address and bring up next address
+                * component.  If strictly routed make sure next
+                * address on directly accessible net.
+                */
                case IPOPT_LSRR:
                case IPOPT_SSRR:
                        if (cp[2] < 4 || cp[2] > optlen - (sizeof (long) - 1))
                                break;
                        sin = (struct in_addr *)(cp + cp[2]);
                case IPOPT_LSRR:
                case IPOPT_SSRR:
                        if (cp[2] < 4 || cp[2] > optlen - (sizeof (long) - 1))
                                break;
                        sin = (struct in_addr *)(cp + cp[2]);
-                       if (n_lhost.s_addr == *(u_long *)sin) {
-                               if (opt == IPOPT_SSRR) {
-                                       /* MAKE SURE *SP DIRECTLY ACCESSIBLE */
-                               }
-                               ip->ip_dst = *sin;
-                               *sin = n_lhost;
-                               cp[2] += 4;
+                       ipaddr.sin_addr = *sin;
+                       ifp = if_ifwithaddr((struct sockaddr *)&ipaddr);
+                       type = ICMP_UNREACH, code = ICMP_UNREACH_SRCFAIL;
+                       if (ifp == 0) {
+                               if (opt == IPOPT_SSRR)
+                                       goto bad;
+                               break;
                        }
                        }
+                       t = ip->ip_dst; ip->ip_dst = *sin; *sin = t;
+                       cp[2] += 4;
+                       if (cp[2] > optlen - (sizeof (long) - 1))
+                               break;
+                       ip->ip_dst = sin[1];
+                       if (opt == IPOPT_SSRR &&
+                           if_ifonnetof(in_netof(ip->ip_dst)) == 0)
+                               goto bad;
                        break;
 
                case IPOPT_TS:
                        break;
 
                case IPOPT_TS:
+                       code = cp - (u_char *)ip;
+                       type = ICMP_PARAMPROB;
                        ipt = (struct ip_timestamp *)cp;
                        if (ipt->ipt_len < 5)
                                goto bad;
                        ipt = (struct ip_timestamp *)cp;
                        if (ipt->ipt_len < 5)
                                goto bad;
@@ -415,12 +497,15 @@ ip_dooptions(ip)
                        case IPOPT_TS_TSANDADDR:
                                if (ipt->ipt_ptr + 8 > ipt->ipt_len)
                                        goto bad;
                        case IPOPT_TS_TSANDADDR:
                                if (ipt->ipt_ptr + 8 > ipt->ipt_len)
                                        goto bad;
-                               *(struct in_addr *)sin++ = n_lhost;
+                               if (ifinet == 0)
+                                       goto bad;       /* ??? */
+                               *sin++ = ((struct sockaddr_in *)&ifinet->if_addr)->sin_addr;
                                break;
 
                        case IPOPT_TS_PRESPEC:
                                break;
 
                        case IPOPT_TS_PRESPEC:
-                               if (*(u_long *)sin != n_lhost.s_addr)
-                                       break;
+                               ipaddr.sin_addr = *sin;
+                               if (!if_ifwithaddr((struct sockaddr *)&ipaddr))
+                                       continue;
                                if (ipt->ipt_ptr + 8 > ipt->ipt_len)
                                        goto bad;
                                ipt->ipt_ptr += 4;
                                if (ipt->ipt_ptr + 8 > ipt->ipt_len)
                                        goto bad;
                                ipt->ipt_ptr += 4;
@@ -433,27 +518,141 @@ ip_dooptions(ip)
                        ipt->ipt_ptr += 4;
                }
        }
                        ipt->ipt_ptr += 4;
                }
        }
-       return;
+       return (0);
 bad:
 bad:
-       /* SHOULD FORCE ICMP MESSAGE */
-       return;
+       icmp_error(ip, type, code);
+       return (1);
 }
 
 /*
 }
 
 /*
- * Strip out IP options, e.g. before passing
- * to higher level protocol in the kernel.
+ * Strip out IP options, at higher
+ * level protocol in the kernel.
+ * Second argument is buffer to which options
+ * will be moved, and return value is their length.
  */
  */
-ip_stripoptions(ip)
+ip_stripoptions(ip, mopt)
        struct ip *ip;
        struct ip *ip;
+       struct mbuf *mopt;
 {
        register int i;
        register struct mbuf *m;
        int olen;
 {
        register int i;
        register struct mbuf *m;
        int olen;
-COUNT(IP_OPT);
 
        olen = (ip->ip_hl<<2) - sizeof (struct ip);
 
        olen = (ip->ip_hl<<2) - sizeof (struct ip);
-       m = dtom(++ip);
+       m = dtom(ip);
+       ip++;
+       if (mopt) {
+               mopt->m_len = olen;
+               mopt->m_off = MMINOFF;
+               bcopy((caddr_t)ip, mtod(m, caddr_t), (unsigned)olen);
+       }
        i = m->m_len - (sizeof (struct ip) + olen);
        bcopy((caddr_t)ip+olen, (caddr_t)ip, (unsigned)i);
        i = m->m_len - (sizeof (struct ip) + olen);
        bcopy((caddr_t)ip+olen, (caddr_t)ip, (unsigned)i);
-       m->m_len -= i;
+       m->m_len -= olen;
+}
+
+u_char inetctlerrmap[] = {
+       ECONNABORTED,   ECONNABORTED,   0,              0,
+       0,              0,
+       EHOSTDOWN,      EHOSTUNREACH,   ENETUNREACH,    EHOSTUNREACH,
+       ECONNREFUSED,   ECONNREFUSED,   EMSGSIZE,       0,
+       0,              0,              0,              0
+};
+
+ip_ctlinput(cmd, arg)
+       int cmd;
+       caddr_t arg;
+{
+       struct in_addr *sin;
+       int tcp_abort(), udp_abort();
+       extern struct inpcb tcb, udb;
+
+       if (cmd < 0 || cmd > PRC_NCMDS)
+               return;
+       if (inetctlerrmap[cmd] == 0)
+               return;         /* XXX */
+       if (cmd == PRC_IFDOWN)
+               sin = &((struct sockaddr_in *)arg)->sin_addr;
+       else if (cmd == PRC_HOSTDEAD || cmd == PRC_HOSTUNREACH)
+               sin = (struct in_addr *)arg;
+       else
+               sin = &((struct icmp *)arg)->icmp_ip.ip_dst;
+       in_pcbnotify(&tcb, sin, inetctlerrmap[cmd], tcp_abort);
+       in_pcbnotify(&udb, sin, inetctlerrmap[cmd], udp_abort);
+}
+
+int    ipprintfs = 0;
+int    ipforwarding = 1;
+/*
+ * Forward a packet.  If some error occurs return the sender
+ * and icmp packet.  Note we can't always generate a meaningful
+ * icmp message because icmp doesn't have a large enough repetoire
+ * of codes and types.
+ */
+ip_forward(ip)
+       register struct ip *ip;
+{
+       register int error, type, code;
+       struct mbuf *mopt, *mcopy;
+
+       if (ipprintfs)
+               printf("forward: src %x dst %x ttl %x\n", ip->ip_src,
+                       ip->ip_dst, ip->ip_ttl);
+       if (ipforwarding == 0) {
+               /* can't tell difference between net and host */
+               type = ICMP_UNREACH, code = ICMP_UNREACH_NET;
+               goto sendicmp;
+       }
+       if (ip->ip_ttl < IPTTLDEC) {
+               type = ICMP_TIMXCEED, code = ICMP_TIMXCEED_INTRANS;
+               goto sendicmp;
+       }
+       ip->ip_ttl -= IPTTLDEC;
+       mopt = m_get(M_DONTWAIT);
+       if (mopt == 0) {
+               m_freem(dtom(ip));
+               return;
+       }
+
+       /*
+        * Save at most 64 bytes of the packet in case
+        * we need to generate an ICMP message to the src.
+        */
+       mcopy = m_copy(dtom(ip), 0, imin(ip->ip_len, 64));
+       ip_stripoptions(ip, mopt);
+
+       /* last 0 here means no directed broadcast */
+       if ((error = ip_output(dtom(ip), mopt, 0, 0)) == 0) {
+               if (mcopy)
+                       m_freem(mcopy);
+               return;
+       }
+       ip = mtod(mcopy, struct ip *);
+       type = ICMP_UNREACH, code = 0;          /* need ``undefined'' */
+       switch (error) {
+
+       case ENETUNREACH:
+       case ENETDOWN:
+               code = ICMP_UNREACH_NET;
+               break;
+
+       case EMSGSIZE:
+               code = ICMP_UNREACH_NEEDFRAG;
+               break;
+
+       case EPERM:
+               code = ICMP_UNREACH_PORT;
+               break;
+
+       case ENOBUFS:
+               type = ICMP_SOURCEQUENCH;
+               break;
+
+       case EHOSTDOWN:
+       case EHOSTUNREACH:
+               code = ICMP_UNREACH_HOST;
+               break;
+       }
+sendicmp:
+       icmp_error(ip, type, code);
 }
 }
Continuous source code history from original PDP-7 UNIX to FreeBSD 2, the first fully unencumbered FreeBSD release.