BSD 4_3_Reno release
[unix-history] / usr / src / sys / net / route.c
index 54654e4..676a6e6 100644 (file)
-/*     route.c 4.8     82/05/11        */
-
-#include "../h/param.h"
-#include "../h/systm.h"
-#include "../h/mbuf.h"
-#include "../h/protosw.h"
-#include "../h/socket.h"
-#include "../h/ioctl.h"
-#include "../net/in.h"
-#include "../net/in_systm.h"
-#include "../net/if.h"
-#include "../net/af.h"
-#include "../net/route.h"
-#include <errno.h>
+/*
+ * Copyright (c) 1980, 1986 Regents of the University of California.
+ * All rights reserved.
+ *
+ * Redistribution is only permitted until one year after the first shipment
+ * of 4.4BSD by the Regents.  Otherwise, redistribution and use in source and
+ * binary forms are permitted provided that: (1) source distributions retain
+ * this entire copyright notice and comment, and (2) distributions including
+ * binaries display the following acknowledgement:  This product includes
+ * software developed by the University of California, Berkeley and its
+ * contributors'' in the documentation or other materials provided with the
+ * distribution and in all advertising materials mentioning features or use
+ * of this software.  Neither the name of the University nor the names of
+ * its contributors may 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
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ *     @(#)route.c     7.17 (Berkeley) 6/28/90
+ */
+#include "machine/reg.h"
+#include "param.h"
+#include "systm.h"
+#include "user.h"
+#include "proc.h"
+#include "mbuf.h"
+#include "socket.h"
+#include "socketvar.h"
+#include "domain.h"
+#include "protosw.h"
+#include "errno.h"
+#include "ioctl.h"
+
+#include "if.h"
+#include "af.h"
+#include "route.h"
+#include "raw_cb.h"
+#include "../netinet/in.h"
+#include "../netinet/in_var.h"
+
+#include "../netns/ns.h"
+#include "machine/mtpr.h"
+#include "netisr.h"
+#define        SA(p) ((struct sockaddr *)(p))
+
+int    rttrash;                /* routes not in table but not freed */
+struct sockaddr wildcard;      /* zero valued cookie for wildcard searches */
+int    rthashsize = RTHASHSIZ; /* for netstat, etc. */
+
+static int rtinits_done = 0;
+struct radix_node_head *ns_rnhead, *in_rnhead;
+struct radix_node *rn_match(), *rn_delete(), *rn_addroute();
+rtinitheads()
+{
+       if (rtinits_done == 0 &&
+           rn_inithead(&ns_rnhead, 16, AF_NS) &&
+           rn_inithead(&in_rnhead, 32, AF_INET))
+               rtinits_done = 1;
+}
 
 /*
  * Packet routing routines.
  */
 
 /*
  * Packet routing routines.
  */
-
 rtalloc(ro)
        register struct route *ro;
 {
 rtalloc(ro)
        register struct route *ro;
 {
-       register struct rtentry *rt, *rtmin;
-       register struct mbuf *m;
-       register int hash, (*match)();
-       struct afhash h;
-       struct sockaddr *dst = &ro->ro_dst;
-       int af = dst->sa_family;
-
-COUNT(RTALLOC);
-       if (ro->ro_rt && ro->ro_rt->rt_ifp)                     /* XXX */
-               return;
-       if (af >= AF_MAX)
-               return;
-       (*afswitch[af].af_hash)(dst, &h);
-       rtmin = 0, hash = h.afh_hosthash;
-       for (m = rthost[hash % RTHASHSIZ]; m; m = m->m_next) {
-               rt = mtod(m, struct rtentry *);
-               if (rt->rt_hash != hash)
-                       continue;
-               if ((rt->rt_flags & RTF_UP) == 0 ||
-                   (rt->rt_ifp->if_flags & IFF_UP) == 0)
-                       continue;
-               if (bcmp((caddr_t)&rt->rt_dst, (caddr_t)dst, sizeof (*dst)))
-                       continue;
-               if (rtmin == 0 || rt->rt_use < rtmin->rt_use)
-                       rtmin = rt;
-       }
-       if (rtmin) 
-               goto found;
-
-       hash = h.afh_nethash;
-       match = afswitch[af].af_netmatch;
-       for (m = rtnet[hash % RTHASHSIZ]; m; m = m->m_next) {
-               rt = mtod(m, struct rtentry *);
-               if (rt->rt_hash != hash)
-                       continue;
-               if ((rt->rt_flags & RTF_UP) == 0 ||
-                   (rt->rt_ifp->if_flags & IFF_UP) == 0)
-                       continue;
-               if (rt->rt_dst.sa_family != af || !(*match)(&rt->rt_dst, dst))
-                       continue;
-               if (rtmin == 0 || rt->rt_use < rtmin->rt_use)
-                       rtmin = rt;
+       if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
+               return;                          /* XXX */
+       ro->ro_rt = rtalloc1(&ro->ro_dst, 1);
+}
+
+struct rtentry *
+rtalloc1(dst, report)
+       register struct sockaddr *dst;
+       int  report;
+{
+       register struct radix_node_head *rnh;
+       register struct rtentry *rt;
+       register struct radix_node *rn;
+       struct rtentry *newrt = 0;
+       int  s = splnet(), err = 0;
+
+       for (rnh = radix_node_head; rnh && (dst->sa_family != rnh->rnh_af); )
+               rnh = rnh->rnh_next;
+       if (rnh && rnh->rnh_treetop &&
+           (rn = rn_match((caddr_t)dst, rnh->rnh_treetop)) &&
+           ((rn->rn_flags & RNF_ROOT) == 0)) {
+               newrt = rt = (struct rtentry *)rn;
+               if (report && (rt->rt_flags & RTF_CLONING)) {
+                       if (err = rtrequest(RTM_RESOLVE, dst, SA(0),
+                                             SA(0), 0, &newrt))
+                               goto miss;
+               } else
+                       rt->rt_refcnt++;
+       } else {
+               rtstat.rts_unreach++;
+       miss:   if (report)
+                       rt_missmsg(RTM_MISS, dst, SA(0), SA(0), SA(0), 0, err);
        }
        }
-found:
-       ro->ro_rt = rtmin;
-       if (rtmin)
-               rtmin->rt_refcnt++;
+       splx(s);
+       return (newrt);
 }
 
 rtfree(rt)
        register struct rtentry *rt;
 {
 }
 
 rtfree(rt)
        register struct rtentry *rt;
 {
-
+       register struct ifaddr *ifa;
        if (rt == 0)
        if (rt == 0)
-               panic("freeroute");
+               panic("rtfree");
        rt->rt_refcnt--;
        rt->rt_refcnt--;
-       /* on refcnt == 0 reclaim? notify someone? */
+       if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
+               rttrash--;
+               if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
+                       panic ("rtfree 2");
+               free((caddr_t)rt, M_RTABLE);
+       }
 }
 
 }
 
-#define        equal(a1, a2) \
-       (bcmp((caddr_t)(a1), (caddr_t)(a2), sizeof (struct sockaddr)) == 0)
 /*
 /*
- * Carry out a request to change the routing table.  Called by
- * interfaces at boot time to make their ``local routes'' known
- * and for ioctl's.
+ * Force a routing table entry to the specified
+ * destination to go through the given gateway.
+ * Normally called as a result of a routing redirect
+ * message from the network layer.
+ *
+ * N.B.: must be called at splnet
+ *
  */
  */
-rtrequest(req, new)
-       int req;
-       register struct rtentry *new;
+rtredirect(dst, gateway, netmask, flags, src, rtp)
+       struct sockaddr *dst, *gateway, *netmask, *src;
+       int flags;
+       struct rtentry **rtp;
 {
        register struct rtentry *rt;
 {
        register struct rtentry *rt;
-       register struct mbuf *m, **mprev;
-       register int hash, (*match)();
-       register struct sockaddr *sa = &new->rt_dst;
-       register struct sockaddr *gate = &new->rt_gateway;
-       struct afhash h;
-       struct mbuf **oldmprev;
-       int af = sa->sa_family, doinghost, s, error = 0;
-
-COUNT(RTREQUEST);
-       if (af >= AF_MAX)
-               return (EAFNOSUPPORT);
-       (*afswitch[af].af_hash)(sa, &h);
-       hash = h.afh_hosthash;
-       mprev = &rthost[hash % RTHASHSIZ];
-       doinghost = 1;
-       s = splimp();
-again:
-       for (; m = *mprev; mprev = &m->m_next) {
-               rt = mtod(m, struct rtentry *);
-               if (rt->rt_hash != hash)
-                       continue;
-               if (doinghost) {
-                       if (!equal(&rt->rt_dst, sa))
-                               continue;
+       int error = 0;
+       short *stat = 0;
+
+       /* verify the gateway is directly reachable */
+       if (ifa_ifwithnet(gateway) == 0) {
+               error = ENETUNREACH;
+               goto done;
+       }
+       rt = rtalloc1(dst, 0);
+       /*
+        * If the redirect isn't from our current router for this dst,
+        * it's either old or wrong.  If it redirects us to ourselves,
+        * we have a routing loop, perhaps as a result of an interface
+        * going down recently.
+        */
+#define        equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
+       if (!(flags & RTF_DONE) && rt && !equal(src, rt->rt_gateway))
+               error = EINVAL;
+       else if (ifa_ifwithaddr(gateway))
+               error = EHOSTUNREACH;
+       if (error)
+               goto done;
+       /*
+        * Create a new entry if we just got back a wildcard entry
+        * or the the lookup failed.  This is necessary for hosts
+        * which use routing redirects generated by smart gateways
+        * to dynamically build the routing tables.
+        */
+       if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
+               goto create;
+       /*
+        * Don't listen to the redirect if it's
+        * for a route to an interface. 
+        */
+       if (rt->rt_flags & RTF_GATEWAY) {
+               if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
+                       /*
+                        * Changing from route to net => route to host.
+                        * Create new route, rather than smashing route to net.
+                        */
+               create:
+                       flags |=  RTF_GATEWAY | RTF_DYNAMIC;
+                       error = rtrequest((int)RTM_ADD, dst, gateway,
+                                   SA(0), flags,
+                                   (struct rtentry **)0);
+                       stat = &rtstat.rts_dynamic;
                } else {
                } else {
-                       if (rt->rt_dst.sa_family != sa->sa_family ||
-                           (*match)(&rt->rt_dst, sa) == 0)
-                               continue;
-               }
-               /* require full match on deletions */
-               if (req == SIOCDELRT && !equal(&rt->rt_gateway, gate))
-                       continue;
-               /* don't keep multiple identical entries */
-               if (req == SIOCADDRT && equal(&rt->rt_gateway, gate)) {
-                       error = EEXIST;
-                       goto bad;
+                       /*
+                        * Smash the current notion of the gateway to
+                        * this destination.  Should check about netmask!!!
+                        */
+                       if (gateway->sa_len <= rt->rt_gateway->sa_len) {
+                               Bcopy(gateway, rt->rt_gateway, gateway->sa_len);
+                               rt->rt_flags |= RTF_MODIFIED;
+                               flags |= RTF_MODIFIED;
+                               stat = &rtstat.rts_newgateway;
+                       } else
+                               error = ENOSPC;
                }
                }
-               break;
+       } else
+               error = EHOSTUNREACH;
+done:
+       if (rt) {
+               if (rtp && !error)
+                       *rtp = rt;
+               else
+                       rtfree(rt);
        }
        }
-       if (m == 0 && doinghost) {
-               hash = h.afh_nethash;
-               oldmprev = mprev;
-               mprev = &rtnet[hash % RTHASHSIZ];
-               match = afswitch[af].af_netmatch;
-               doinghost = 0;
-               goto again;
+       if (error)
+               rtstat.rts_badredirect++;
+       else
+               (stat && (*stat)++);
+       rt_missmsg(RTM_REDIRECT, dst, gateway, netmask, src, flags, error);
+}
+
+/*
+* Routing table ioctl interface.
+*/
+rtioctl(req, data)
+       int req;
+       caddr_t data;
+{
+#ifndef COMPAT_43
+       return (EOPNOTSUPP);
+#else
+       register struct ortentry *entry = (struct ortentry *)data;
+       int error;
+       struct sockaddr *netmask = 0;
+
+       if (req == SIOCADDRT)
+               req = RTM_ADD;
+       else if (req == SIOCDELRT)
+               req = RTM_DELETE;
+       else
+               return (EINVAL);
+
+       if (error = suser(u.u_cred, &u.u_acflag))
+               return (error);
+#if BYTE_ORDER != BIG_ENDIAN
+       if (entry->rt_dst.sa_family == 0 && entry->rt_dst.sa_len < 16) {
+               entry->rt_dst.sa_family = entry->rt_dst.sa_len;
+               entry->rt_dst.sa_len = 16;
        }
        }
+       if (entry->rt_gateway.sa_family == 0 && entry->rt_gateway.sa_len < 16) {
+               entry->rt_gateway.sa_family = entry->rt_gateway.sa_len;
+               entry->rt_gateway.sa_len = 16;
+       }
+#else
+       if (entry->rt_dst.sa_len == 0)
+               entry->rt_dst.sa_len = 16;
+       if (entry->rt_gateway.sa_len == 0)
+               entry->rt_gateway.sa_len = 16;
+#endif
+       if ((entry->rt_flags & RTF_HOST) == 0)
+               switch (entry->rt_dst.sa_family) {
+#ifdef INET
+               case AF_INET:
+                       {
+                               extern struct sockaddr_in icmpmask;
+                               struct sockaddr_in *dst_in = 
+                                       (struct sockaddr_in *)&entry->rt_dst;
+
+                               in_sockmaskof(dst_in->sin_addr, &icmpmask);
+                               netmask = (struct sockaddr *)&icmpmask;
+                       }
+                       break;
+#endif
+#ifdef NS
+               case AF_NS:
+                       {
+                               extern struct sockaddr_ns ns_netmask;
+                               netmask = (struct sockaddr *)&ns_netmask;
+                       }
+#endif
+               }
+       error =  rtrequest(req, &(entry->rt_dst), &(entry->rt_gateway), netmask,
+                               entry->rt_flags, (struct rtentry **)0);
+       rt_missmsg((req == RTM_ADD ? RTM_OLDADD : RTM_OLDDEL),
+                  &(entry->rt_dst), &(entry->rt_gateway),
+                  netmask, SA(0), entry->rt_flags, error);
+       return (error);
+#endif
+}
 
 
-       if (m == 0 && req != SIOCADDRT) {
-               error = ESRCH;
-               goto bad;
+struct ifaddr *
+ifa_ifwithroute(flags, dst, gateway)
+int    flags;
+struct sockaddr        *dst, *gateway;
+{
+       struct ifaddr *ifa;
+       if ((flags & RTF_GATEWAY) == 0) {
+               /*
+                * If we are adding a route to an interface,
+                * and the interface is a pt to pt link
+                * we should search for the destination
+                * as our clue to the interface.  Otherwise
+                * we can use the local address.
+                */
+               ifa = 0;
+               if (flags & RTF_HOST) 
+                       ifa = ifa_ifwithdstaddr(dst);
+               if (ifa == 0)
+                       ifa = ifa_ifwithaddr(gateway);
+       } else {
+               /*
+                * If we are adding a route to a remote net
+                * or host, the gateway may still be on the
+                * other end of a pt to pt link.
+                */
+               ifa = ifa_ifwithdstaddr(gateway);
        }
        }
-found:
-       switch (req) {
+       if (ifa == 0)
+               ifa = ifa_ifwithnet(gateway);
+       return (ifa);
+}
+
+#define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
 
 
-       case SIOCDELRT:
+rtrequest(req, dst, gateway, netmask, flags, ret_nrt)
+       int req, flags;
+       struct sockaddr *dst, *gateway, *netmask;
+       struct rtentry **ret_nrt;
+{
+       int s = splnet(), len, error = 0;
+       register struct rtentry *rt;
+       register struct radix_node *rn;
+       register struct radix_node_head *rnh;
+       struct ifaddr *ifa, *ifa_ifwithdstaddr();
+       struct sockaddr *ndst;
+       u_char af = dst->sa_family;
+#define senderr(x) { error = x ; goto bad; }
+
+       if (rtinits_done == 0)
+               rtinitheads();
+       for (rnh = radix_node_head; rnh && (af != rnh->rnh_af); )
+               rnh = rnh->rnh_next;
+       if (rnh == 0)
+               senderr(ESRCH);
+       if (flags & RTF_HOST)
+               netmask = 0;
+       switch (req) {
+       case RTM_DELETE:
+               if (ret_nrt && (rt = *ret_nrt)) {
+                       RTFREE(rt);
+                       *ret_nrt = 0;
+               }
+               if ((rn = rn_delete((caddr_t)dst, (caddr_t)netmask, 
+                                       rnh->rnh_treetop)) == 0)
+                       senderr(ESRCH);
+               if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
+                       panic ("rtrequest delete");
+               rt = (struct rtentry *)rn;
                rt->rt_flags &= ~RTF_UP;
                rt->rt_flags &= ~RTF_UP;
-               if (rt->rt_refcnt > 0)  /* should we notify protocols? */
-                       error = EBUSY;
-               else
-                       *mprev = m_free(m);
+               if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
+                       ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0));
+               rttrash++;
+               if (rt->rt_refcnt <= 0)
+                       rtfree(rt);
                break;
 
                break;
 
-       case SIOCCHGRT:
-               rt->rt_flags = new->rt_flags;
-               if (rt->rt_refcnt > 0)
-                       error = EBUSY;
-               else if (!equal(&rt->rt_gateway, gate))
-                       goto newneighbor;
-               break;
+       case RTM_RESOLVE:
+               if (ret_nrt == 0 || (rt = *ret_nrt) == 0)
+                       senderr(EINVAL);
+               ifa = rt->rt_ifa;
+               flags = rt->rt_flags & ~RTF_CLONING;
+               gateway = rt->rt_gateway;
+               if ((netmask = rt->rt_genmask) == 0)
+                       flags |= RTF_HOST;
+               goto makeroute;
 
 
-       case SIOCADDRT:
-               m = m_get(M_DONTWAIT);
-               if (m == 0) {
-                       error = ENOBUFS;
-                       break;
+       case RTM_ADD:
+               if ((ifa = ifa_ifwithroute(flags, dst, gateway)) == 0)
+                       senderr(ENETUNREACH);
+       makeroute:
+               len = sizeof (*rt) + ROUNDUP(gateway->sa_len)
+                   + ROUNDUP(dst->sa_len);
+               R_Malloc(rt, struct rtentry *, len);
+               if (rt == 0)
+                       senderr(ENOBUFS);
+               Bzero(rt, len);
+               ndst = (struct sockaddr *)(rt + 1);
+               if (netmask) {
+                       rt_maskedcopy(dst, ndst, netmask);
+               } else
+                       Bcopy(dst, ndst, dst->sa_len);
+               rn = rn_addroute((caddr_t)ndst, (caddr_t)netmask,
+                                       rnh->rnh_treetop, rt->rt_nodes);
+               if (rn == 0) {
+                       free((caddr_t)rt, M_RTABLE);
+                       senderr(EEXIST);
                }
                }
-               m->m_off = MMINOFF;
-               m->m_len = sizeof (struct rtentry);
-               rt = mtod(m, struct rtentry *);
-               *rt = *new;
-               if (new->rt_flags & RTF_HOST) {
-                       rt->rt_hash = h.afh_hosthash;
-                       *oldmprev = m;
-               } else {
-                       rt->rt_hash = h.afh_nethash;
-                       *mprev = m;
+               rt->rt_ifa = ifa;
+               rt->rt_ifp = ifa->ifa_ifp;
+               rt->rt_flags = RTF_UP | flags;
+               rn->rn_key = (caddr_t) ndst; /* == rt_dst */
+               rt->rt_gateway = (struct sockaddr *)
+                                       (rn->rn_key + ROUNDUP(dst->sa_len));
+               Bcopy(gateway, rt->rt_gateway, gateway->sa_len);
+               if (req == RTM_RESOLVE)
+                       rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
+               if (ifa->ifa_rtrequest)
+                       ifa->ifa_rtrequest(req, rt, SA(ret_nrt ? *ret_nrt : 0));
+               if (ret_nrt) {
+                       *ret_nrt = rt;
+                       rt->rt_refcnt++;
                }
                }
-               rt->rt_use = 0;
-               rt->rt_refcnt = 0;
-newneighbor:
-               rt->rt_ifp = if_ifwithnet(gate);
-               if (rt->rt_ifp == 0)
-                       rt->rt_flags &= ~RTF_UP;
                break;
        }
 bad:
                break;
        }
 bad:
@@ -192,23 +406,32 @@ bad:
        return (error);
 }
 
        return (error);
 }
 
+rt_maskedcopy(src, dst, netmask)
+struct sockaddr *src, *dst, *netmask;
+{
+       register u_char *cp1 = (u_char *)src;
+       register u_char *cp2 = (u_char *)dst;
+       register u_char *cp3 = (u_char *)netmask;
+       u_char *cplim = cp2 + *cp3;
+       u_char *cplim2 = cp2 + *cp1;
+
+       *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
+       cp3 += 2;
+       if (cplim > cplim2)
+               cplim = cplim2;
+       while (cp2 < cplim)
+               *cp2++ = *cp1++ & *cp3++;
+       if (cp2 < cplim2)
+               bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
+}
 /*
  * Set up a routing table entry, normally
  * for an interface.
  */
 /*
  * Set up a routing table entry, normally
  * for an interface.
  */
-rtinit(dst, gateway, flags)
-       struct sockaddr *dst, *gateway;
-       int flags;
+rtinit(ifa, cmd, flags)
+       register struct ifaddr *ifa;
+       int cmd, flags;
 {
 {
-       struct rtentry route;
-       struct route ro;
-
-       route.rt_dst = *dst;
-       route.rt_gateway = *gateway;
-       route.rt_flags = flags;
-       route.rt_use = 0;
-       (void) rtrequest(SIOCADDRT, &route);
-       ro.ro_rt = 0;
-       ro.ro_dst = *dst;
-       rtalloc(&ro);
+       return rtrequest(cmd, ifa->ifa_dstaddr, ifa->ifa_addr,
+                   ifa->ifa_netmask, flags | ifa->ifa_flags, &ifa->ifa_rt);
 }
 }