diff --git a/sys/net/radix.c b/sys/net/radix.c index 7b4699d5f8e8..6b67f6642517 100644 --- a/sys/net/radix.c +++ b/sys/net/radix.c @@ -1,1203 +1,1203 @@ /*- * Copyright (c) 1988, 1989, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. 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 BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)radix.c 8.5 (Berkeley) 5/19/95 * $FreeBSD$ */ /* * Routines to build and maintain radix trees for routing lookups. */ #include #ifdef _KERNEL #include #include #include #include #include #include #include #include "opt_mpath.h" #ifdef RADIX_MPATH #include #endif #else /* !_KERNEL */ #include #include #include #define log(x, arg...) fprintf(stderr, ## arg) #define panic(x) fprintf(stderr, "PANIC: %s", x), exit(1) #define min(a, b) ((a) < (b) ? (a) : (b) ) #include #endif /* !_KERNEL */ static int rn_walktree_from(struct radix_node_head *h, void *a, void *m, walktree_f_t *f, void *w); static int rn_walktree(struct radix_node_head *, walktree_f_t *, void *); static struct radix_node *rn_insert(void *, struct radix_node_head *, int *, struct radix_node [2]), *rn_newpair(void *, int, struct radix_node[2]), *rn_search(void *, struct radix_node *), *rn_search_m(void *, struct radix_node *, void *); -static int max_keylen; -static struct radix_mask *rn_mkfreelist; -static struct radix_node_head *mask_rnhead; -/* - * Work area -- the following point to 3 buffers of size max_keylen, - * allocated in this order in a block of memory malloc'ed by rn_init. - * rn_zeros, rn_ones are set in rn_init and used in readonly afterwards. - * addmask_key is used in rn_addmask in rw mode and not thread-safe. - */ -static char *rn_zeros, *rn_ones, *addmask_key; +static void rn_detachhead_internal(void **head); +static int rn_inithead_internal(void **head, int off); + +#define RADIX_MAX_KEY_LEN 32 -#define MKGet(m) { \ - if (rn_mkfreelist) { \ - m = rn_mkfreelist; \ - rn_mkfreelist = (m)->rm_mklist; \ - } else \ - R_Malloc(m, struct radix_mask *, sizeof (struct radix_mask)); } - -#define MKFree(m) { (m)->rm_mklist = rn_mkfreelist; rn_mkfreelist = (m);} +static char rn_zeros[RADIX_MAX_KEY_LEN]; +static char rn_ones[RADIX_MAX_KEY_LEN] = { + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, +}; -#define rn_masktop (mask_rnhead->rnh_treetop) static int rn_lexobetter(void *m_arg, void *n_arg); static struct radix_mask * rn_new_radix_mask(struct radix_node *tt, struct radix_mask *next); static int rn_satisfies_leaf(char *trial, struct radix_node *leaf, int skip); /* * The data structure for the keys is a radix tree with one way * branching removed. The index rn_bit at an internal node n represents a bit * position to be tested. The tree is arranged so that all descendants * of a node n have keys whose bits all agree up to position rn_bit - 1. * (We say the index of n is rn_bit.) * * There is at least one descendant which has a one bit at position rn_bit, * and at least one with a zero there. * * A route is determined by a pair of key and mask. We require that the * bit-wise logical and of the key and mask to be the key. * We define the index of a route to associated with the mask to be * the first bit number in the mask where 0 occurs (with bit number 0 * representing the highest order bit). * * We say a mask is normal if every bit is 0, past the index of the mask. * If a node n has a descendant (k, m) with index(m) == index(n) == rn_bit, * and m is a normal mask, then the route applies to every descendant of n. * If the index(m) < rn_bit, this implies the trailing last few bits of k * before bit b are all 0, (and hence consequently true of every descendant * of n), so the route applies to all descendants of the node as well. * * Similar logic shows that a non-normal mask m such that * index(m) <= index(n) could potentially apply to many children of n. * Thus, for each non-host route, we attach its mask to a list at an internal * node as high in the tree as we can go. * * The present version of the code makes use of normal routes in short- * circuiting an explict mask and compare operation when testing whether * a key satisfies a normal route, and also in remembering the unique leaf * that governs a subtree. */ /* * Most of the functions in this code assume that the key/mask arguments * are sockaddr-like structures, where the first byte is an u_char * indicating the size of the entire structure. * * To make the assumption more explicit, we use the LEN() macro to access * this field. It is safe to pass an expression with side effects * to LEN() as the argument is evaluated only once. * We cast the result to int as this is the dominant usage. */ #define LEN(x) ( (int) (*(const u_char *)(x)) ) /* * XXX THIS NEEDS TO BE FIXED * In the code, pointers to keys and masks are passed as either * 'void *' (because callers use to pass pointers of various kinds), or * 'caddr_t' (which is fine for pointer arithmetics, but not very * clean when you dereference it to access data). Furthermore, caddr_t * is really 'char *', while the natural type to operate on keys and * masks would be 'u_char'. This mismatch require a lot of casts and * intermediate variables to adapt types that clutter the code. */ /* * Search a node in the tree matching the key. */ static struct radix_node * rn_search(v_arg, head) void *v_arg; struct radix_node *head; { register struct radix_node *x; register caddr_t v; for (x = head, v = v_arg; x->rn_bit >= 0;) { if (x->rn_bmask & v[x->rn_offset]) x = x->rn_right; else x = x->rn_left; } return (x); } /* * Same as above, but with an additional mask. * XXX note this function is used only once. */ static struct radix_node * rn_search_m(v_arg, head, m_arg) struct radix_node *head; void *v_arg, *m_arg; { register struct radix_node *x; register caddr_t v = v_arg, m = m_arg; for (x = head; x->rn_bit >= 0;) { if ((x->rn_bmask & m[x->rn_offset]) && (x->rn_bmask & v[x->rn_offset])) x = x->rn_right; else x = x->rn_left; } return x; } int rn_refines(m_arg, n_arg) void *m_arg, *n_arg; { register caddr_t m = m_arg, n = n_arg; register caddr_t lim, lim2 = lim = n + LEN(n); int longer = LEN(n++) - LEN(m++); int masks_are_equal = 1; if (longer > 0) lim -= longer; while (n < lim) { if (*n & ~(*m)) return 0; if (*n++ != *m++) masks_are_equal = 0; } while (n < lim2) if (*n++) return 0; if (masks_are_equal && (longer < 0)) for (lim2 = m - longer; m < lim2; ) if (*m++) return 1; return (!masks_are_equal); } struct radix_node * rn_lookup(v_arg, m_arg, head) void *v_arg, *m_arg; struct radix_node_head *head; { register struct radix_node *x; caddr_t netmask = 0; if (m_arg) { - x = rn_addmask(m_arg, 1, head->rnh_treetop->rn_offset); + x = rn_addmask(m_arg, head->rnh_masks, 1, + head->rnh_treetop->rn_offset); if (x == 0) return (0); netmask = x->rn_key; } x = rn_match(v_arg, head); if (x && netmask) { while (x && x->rn_mask != netmask) x = x->rn_dupedkey; } return x; } static int rn_satisfies_leaf(trial, leaf, skip) char *trial; register struct radix_node *leaf; int skip; { register char *cp = trial, *cp2 = leaf->rn_key, *cp3 = leaf->rn_mask; char *cplim; int length = min(LEN(cp), LEN(cp2)); if (cp3 == NULL) cp3 = rn_ones; else length = min(length, LEN(cp3)); cplim = cp + length; cp3 += skip; cp2 += skip; for (cp += skip; cp < cplim; cp++, cp2++, cp3++) if ((*cp ^ *cp2) & *cp3) return 0; return 1; } struct radix_node * rn_match(v_arg, head) void *v_arg; struct radix_node_head *head; { caddr_t v = v_arg; register struct radix_node *t = head->rnh_treetop, *x; register caddr_t cp = v, cp2; caddr_t cplim; struct radix_node *saved_t, *top = t; int off = t->rn_offset, vlen = LEN(cp), matched_off; register int test, b, rn_bit; /* * Open code rn_search(v, top) to avoid overhead of extra * subroutine call. */ for (; t->rn_bit >= 0; ) { if (t->rn_bmask & cp[t->rn_offset]) t = t->rn_right; else t = t->rn_left; } /* * See if we match exactly as a host destination * or at least learn how many bits match, for normal mask finesse. * * It doesn't hurt us to limit how many bytes to check * to the length of the mask, since if it matches we had a genuine * match and the leaf we have is the most specific one anyway; * if it didn't match with a shorter length it would fail * with a long one. This wins big for class B&C netmasks which * are probably the most common case... */ if (t->rn_mask) vlen = *(u_char *)t->rn_mask; cp += off; cp2 = t->rn_key + off; cplim = v + vlen; for (; cp < cplim; cp++, cp2++) if (*cp != *cp2) goto on1; /* * This extra grot is in case we are explicitly asked * to look up the default. Ugh! * * Never return the root node itself, it seems to cause a * lot of confusion. */ if (t->rn_flags & RNF_ROOT) t = t->rn_dupedkey; return t; on1: test = (*cp ^ *cp2) & 0xff; /* find first bit that differs */ for (b = 7; (test >>= 1) > 0;) b--; matched_off = cp - v; b += matched_off << 3; rn_bit = -1 - b; /* * If there is a host route in a duped-key chain, it will be first. */ if ((saved_t = t)->rn_mask == 0) t = t->rn_dupedkey; for (; t; t = t->rn_dupedkey) /* * Even if we don't match exactly as a host, * we may match if the leaf we wound up at is * a route to a net. */ if (t->rn_flags & RNF_NORMAL) { if (rn_bit <= t->rn_bit) return t; } else if (rn_satisfies_leaf(v, t, matched_off)) return t; t = saved_t; /* start searching up the tree */ do { register struct radix_mask *m; t = t->rn_parent; m = t->rn_mklist; /* * If non-contiguous masks ever become important * we can restore the masking and open coding of * the search and satisfaction test and put the * calculation of "off" back before the "do". */ while (m) { if (m->rm_flags & RNF_NORMAL) { if (rn_bit <= m->rm_bit) return (m->rm_leaf); } else { off = min(t->rn_offset, matched_off); x = rn_search_m(v, t, m->rm_mask); while (x && x->rn_mask != m->rm_mask) x = x->rn_dupedkey; if (x && rn_satisfies_leaf(v, x, off)) return x; } m = m->rm_mklist; } } while (t != top); return 0; } #ifdef RN_DEBUG int rn_nodenum; struct radix_node *rn_clist; int rn_saveinfo; int rn_debug = 1; #endif /* * Whenever we add a new leaf to the tree, we also add a parent node, * so we allocate them as an array of two elements: the first one must be * the leaf (see RNTORT() in route.c), the second one is the parent. * This routine initializes the relevant fields of the nodes, so that * the leaf is the left child of the parent node, and both nodes have * (almost) all all fields filled as appropriate. * (XXX some fields are left unset, see the '#if 0' section). * The function returns a pointer to the parent node. */ static struct radix_node * rn_newpair(v, b, nodes) void *v; int b; struct radix_node nodes[2]; { register struct radix_node *tt = nodes, *t = tt + 1; t->rn_bit = b; t->rn_bmask = 0x80 >> (b & 7); t->rn_left = tt; t->rn_offset = b >> 3; #if 0 /* XXX perhaps we should fill these fields as well. */ t->rn_parent = t->rn_right = NULL; tt->rn_mask = NULL; tt->rn_dupedkey = NULL; tt->rn_bmask = 0; #endif tt->rn_bit = -1; tt->rn_key = (caddr_t)v; tt->rn_parent = t; tt->rn_flags = t->rn_flags = RNF_ACTIVE; tt->rn_mklist = t->rn_mklist = 0; #ifdef RN_DEBUG tt->rn_info = rn_nodenum++; t->rn_info = rn_nodenum++; tt->rn_twin = t; tt->rn_ybro = rn_clist; rn_clist = tt; #endif return t; } static struct radix_node * rn_insert(v_arg, head, dupentry, nodes) void *v_arg; struct radix_node_head *head; int *dupentry; struct radix_node nodes[2]; { caddr_t v = v_arg; struct radix_node *top = head->rnh_treetop; int head_off = top->rn_offset, vlen = LEN(v); register struct radix_node *t = rn_search(v_arg, top); register caddr_t cp = v + head_off; register int b; struct radix_node *tt; /* * Find first bit at which v and t->rn_key differ */ { register caddr_t cp2 = t->rn_key + head_off; register int cmp_res; caddr_t cplim = v + vlen; while (cp < cplim) if (*cp2++ != *cp++) goto on1; *dupentry = 1; return t; on1: *dupentry = 0; cmp_res = (cp[-1] ^ cp2[-1]) & 0xff; for (b = (cp - v) << 3; cmp_res; b--) cmp_res >>= 1; } { register struct radix_node *p, *x = top; cp = v; do { p = x; if (cp[x->rn_offset] & x->rn_bmask) x = x->rn_right; else x = x->rn_left; } while (b > (unsigned) x->rn_bit); /* x->rn_bit < b && x->rn_bit >= 0 */ #ifdef RN_DEBUG if (rn_debug) log(LOG_DEBUG, "rn_insert: Going In:\n"), traverse(p); #endif t = rn_newpair(v_arg, b, nodes); tt = t->rn_left; if ((cp[p->rn_offset] & p->rn_bmask) == 0) p->rn_left = t; else p->rn_right = t; x->rn_parent = t; t->rn_parent = p; /* frees x, p as temp vars below */ if ((cp[t->rn_offset] & t->rn_bmask) == 0) { t->rn_right = x; } else { t->rn_right = tt; t->rn_left = x; } #ifdef RN_DEBUG if (rn_debug) log(LOG_DEBUG, "rn_insert: Coming Out:\n"), traverse(p); #endif } return (tt); } struct radix_node * -rn_addmask(n_arg, search, skip) - int search, skip; - void *n_arg; +rn_addmask(void *n_arg, struct radix_node_head *maskhead, int search, int skip) { caddr_t netmask = (caddr_t)n_arg; register struct radix_node *x; register caddr_t cp, cplim; register int b = 0, mlen, j; - int maskduplicated, m0, isnormal; + int maskduplicated, isnormal; struct radix_node *saved_x; - static int last_zeroed = 0; + char addmask_key[RADIX_MAX_KEY_LEN]; - if ((mlen = LEN(netmask)) > max_keylen) - mlen = max_keylen; + if ((mlen = LEN(netmask)) > RADIX_MAX_KEY_LEN) + mlen = RADIX_MAX_KEY_LEN; if (skip == 0) skip = 1; if (mlen <= skip) - return (mask_rnhead->rnh_nodes); + return (maskhead->rnh_nodes); + + bzero(addmask_key, RADIX_MAX_KEY_LEN); if (skip > 1) bcopy(rn_ones + 1, addmask_key + 1, skip - 1); - if ((m0 = mlen) > skip) - bcopy(netmask + skip, addmask_key + skip, mlen - skip); + bcopy(netmask + skip, addmask_key + skip, mlen - skip); /* * Trim trailing zeroes. */ for (cp = addmask_key + mlen; (cp > addmask_key) && cp[-1] == 0;) cp--; mlen = cp - addmask_key; - if (mlen <= skip) { - if (m0 >= last_zeroed) - last_zeroed = mlen; - return (mask_rnhead->rnh_nodes); - } - if (m0 < last_zeroed) - bzero(addmask_key + m0, last_zeroed - m0); - *addmask_key = last_zeroed = mlen; - x = rn_search(addmask_key, rn_masktop); + if (mlen <= skip) + return (maskhead->rnh_nodes); + *addmask_key = mlen; + x = rn_search(addmask_key, maskhead->rnh_treetop); if (bcmp(addmask_key, x->rn_key, mlen) != 0) x = 0; if (x || search) return (x); - R_Zalloc(x, struct radix_node *, max_keylen + 2 * sizeof (*x)); + R_Zalloc(x, struct radix_node *, RADIX_MAX_KEY_LEN + 2 * sizeof (*x)); if ((saved_x = x) == 0) return (0); netmask = cp = (caddr_t)(x + 2); bcopy(addmask_key, cp, mlen); - x = rn_insert(cp, mask_rnhead, &maskduplicated, x); + x = rn_insert(cp, maskhead, &maskduplicated, x); if (maskduplicated) { log(LOG_ERR, "rn_addmask: mask impossibly already in tree"); Free(saved_x); return (x); } /* * Calculate index of mask, and check for normalcy. * First find the first byte with a 0 bit, then if there are * more bits left (remember we already trimmed the trailing 0's), * the pattern must be one of those in normal_chars[], or we have * a non-contiguous mask. */ cplim = netmask + mlen; isnormal = 1; for (cp = netmask + skip; (cp < cplim) && *(u_char *)cp == 0xff;) cp++; if (cp != cplim) { static char normal_chars[] = { 0, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff}; for (j = 0x80; (j & *cp) != 0; j >>= 1) b++; if (*cp != normal_chars[b] || cp != (cplim - 1)) isnormal = 0; } b += (cp - netmask) << 3; x->rn_bit = -1 - b; if (isnormal) x->rn_flags |= RNF_NORMAL; return (x); } static int /* XXX: arbitrary ordering for non-contiguous masks */ rn_lexobetter(m_arg, n_arg) void *m_arg, *n_arg; { register u_char *mp = m_arg, *np = n_arg, *lim; if (LEN(mp) > LEN(np)) return 1; /* not really, but need to check longer one first */ if (LEN(mp) == LEN(np)) for (lim = mp + LEN(mp); mp < lim;) if (*mp++ > *np++) return 1; return 0; } static struct radix_mask * rn_new_radix_mask(tt, next) register struct radix_node *tt; register struct radix_mask *next; { register struct radix_mask *m; - MKGet(m); + R_Malloc(m, struct radix_mask *, sizeof (struct radix_mask)); if (m == 0) { - log(LOG_ERR, "Mask for route not entered\n"); + log(LOG_ERR, "Failed to allocate route mask\n"); return (0); } - bzero(m, sizeof *m); + bzero(m, sizeof(*m)); m->rm_bit = tt->rn_bit; m->rm_flags = tt->rn_flags; if (tt->rn_flags & RNF_NORMAL) m->rm_leaf = tt; else m->rm_mask = tt->rn_mask; m->rm_mklist = next; tt->rn_mklist = m; return m; } struct radix_node * rn_addroute(v_arg, n_arg, head, treenodes) void *v_arg, *n_arg; struct radix_node_head *head; struct radix_node treenodes[2]; { caddr_t v = (caddr_t)v_arg, netmask = (caddr_t)n_arg; register struct radix_node *t, *x = 0, *tt; struct radix_node *saved_tt, *top = head->rnh_treetop; short b = 0, b_leaf = 0; int keyduplicated; caddr_t mmask; struct radix_mask *m, **mp; /* * In dealing with non-contiguous masks, there may be * many different routes which have the same mask. * We will find it useful to have a unique pointer to * the mask to speed avoiding duplicate references at * nodes and possibly save time in calculating indices. */ if (netmask) { - if ((x = rn_addmask(netmask, 0, top->rn_offset)) == 0) + x = rn_addmask(netmask, head->rnh_masks, 0, top->rn_offset); + if (x == NULL) return (0); b_leaf = x->rn_bit; b = -1 - x->rn_bit; netmask = x->rn_key; } /* * Deal with duplicated keys: attach node to previous instance */ saved_tt = tt = rn_insert(v, head, &keyduplicated, treenodes); if (keyduplicated) { for (t = tt; tt; t = tt, tt = tt->rn_dupedkey) { #ifdef RADIX_MPATH /* permit multipath, if enabled for the family */ if (rn_mpath_capable(head) && netmask == tt->rn_mask) { /* * go down to the end of multipaths, so that * new entry goes into the end of rn_dupedkey * chain. */ do { t = tt; tt = tt->rn_dupedkey; } while (tt && t->rn_mask == tt->rn_mask); break; } #endif if (tt->rn_mask == netmask) return (0); if (netmask == 0 || (tt->rn_mask && ((b_leaf < tt->rn_bit) /* index(netmask) > node */ || rn_refines(netmask, tt->rn_mask) || rn_lexobetter(netmask, tt->rn_mask)))) break; } /* * If the mask is not duplicated, we wouldn't * find it among possible duplicate key entries * anyway, so the above test doesn't hurt. * * We sort the masks for a duplicated key the same way as * in a masklist -- most specific to least specific. * This may require the unfortunate nuisance of relocating * the head of the list. * * We also reverse, or doubly link the list through the * parent pointer. */ if (tt == saved_tt) { struct radix_node *xx = x; /* link in at head of list */ (tt = treenodes)->rn_dupedkey = t; tt->rn_flags = t->rn_flags; tt->rn_parent = x = t->rn_parent; t->rn_parent = tt; /* parent */ if (x->rn_left == t) x->rn_left = tt; else x->rn_right = tt; saved_tt = tt; x = xx; } else { (tt = treenodes)->rn_dupedkey = t->rn_dupedkey; t->rn_dupedkey = tt; tt->rn_parent = t; /* parent */ if (tt->rn_dupedkey) /* parent */ tt->rn_dupedkey->rn_parent = tt; /* parent */ } #ifdef RN_DEBUG t=tt+1; tt->rn_info = rn_nodenum++; t->rn_info = rn_nodenum++; tt->rn_twin = t; tt->rn_ybro = rn_clist; rn_clist = tt; #endif tt->rn_key = (caddr_t) v; tt->rn_bit = -1; tt->rn_flags = RNF_ACTIVE; } /* * Put mask in tree. */ if (netmask) { tt->rn_mask = netmask; tt->rn_bit = x->rn_bit; tt->rn_flags |= x->rn_flags & RNF_NORMAL; } t = saved_tt->rn_parent; if (keyduplicated) goto on2; b_leaf = -1 - t->rn_bit; if (t->rn_right == saved_tt) x = t->rn_left; else x = t->rn_right; /* Promote general routes from below */ if (x->rn_bit < 0) { for (mp = &t->rn_mklist; x; x = x->rn_dupedkey) if (x->rn_mask && (x->rn_bit >= b_leaf) && x->rn_mklist == 0) { *mp = m = rn_new_radix_mask(x, 0); if (m) mp = &m->rm_mklist; } } else if (x->rn_mklist) { /* * Skip over masks whose index is > that of new node */ for (mp = &x->rn_mklist; (m = *mp); mp = &m->rm_mklist) if (m->rm_bit >= b_leaf) break; t->rn_mklist = m; *mp = 0; } on2: /* Add new route to highest possible ancestor's list */ if ((netmask == 0) || (b > t->rn_bit )) return tt; /* can't lift at all */ b_leaf = tt->rn_bit; do { x = t; t = t->rn_parent; } while (b <= t->rn_bit && x != top); /* * Search through routes associated with node to * insert new route according to index. * Need same criteria as when sorting dupedkeys to avoid * double loop on deletion. */ for (mp = &x->rn_mklist; (m = *mp); mp = &m->rm_mklist) { if (m->rm_bit < b_leaf) continue; if (m->rm_bit > b_leaf) break; if (m->rm_flags & RNF_NORMAL) { mmask = m->rm_leaf->rn_mask; if (tt->rn_flags & RNF_NORMAL) { #if !defined(RADIX_MPATH) log(LOG_ERR, "Non-unique normal route, mask not entered\n"); #endif return tt; } } else mmask = m->rm_mask; if (mmask == netmask) { m->rm_refs++; tt->rn_mklist = m; return tt; } if (rn_refines(netmask, mmask) || rn_lexobetter(netmask, mmask)) break; } *mp = rn_new_radix_mask(tt, *mp); return tt; } struct radix_node * rn_delete(v_arg, netmask_arg, head) void *v_arg, *netmask_arg; struct radix_node_head *head; { register struct radix_node *t, *p, *x, *tt; struct radix_mask *m, *saved_m, **mp; struct radix_node *dupedkey, *saved_tt, *top; caddr_t v, netmask; int b, head_off, vlen; v = v_arg; netmask = netmask_arg; x = head->rnh_treetop; tt = rn_search(v, x); head_off = x->rn_offset; vlen = LEN(v); saved_tt = tt; top = x; if (tt == 0 || bcmp(v + head_off, tt->rn_key + head_off, vlen - head_off)) return (0); /* * Delete our route from mask lists. */ if (netmask) { - if ((x = rn_addmask(netmask, 1, head_off)) == 0) + x = rn_addmask(netmask, head->rnh_masks, 1, head_off); + if (x == NULL) return (0); netmask = x->rn_key; while (tt->rn_mask != netmask) if ((tt = tt->rn_dupedkey) == 0) return (0); } if (tt->rn_mask == 0 || (saved_m = m = tt->rn_mklist) == 0) goto on1; if (tt->rn_flags & RNF_NORMAL) { if (m->rm_leaf != tt || m->rm_refs > 0) { log(LOG_ERR, "rn_delete: inconsistent annotation\n"); return 0; /* dangling ref could cause disaster */ } } else { if (m->rm_mask != tt->rn_mask) { log(LOG_ERR, "rn_delete: inconsistent annotation\n"); goto on1; } if (--m->rm_refs >= 0) goto on1; } b = -1 - tt->rn_bit; t = saved_tt->rn_parent; if (b > t->rn_bit) goto on1; /* Wasn't lifted at all */ do { x = t; t = t->rn_parent; } while (b <= t->rn_bit && x != top); for (mp = &x->rn_mklist; (m = *mp); mp = &m->rm_mklist) if (m == saved_m) { *mp = m->rm_mklist; - MKFree(m); + Free(m); break; } if (m == 0) { log(LOG_ERR, "rn_delete: couldn't find our annotation\n"); if (tt->rn_flags & RNF_NORMAL) return (0); /* Dangling ref to us */ } on1: /* * Eliminate us from tree */ if (tt->rn_flags & RNF_ROOT) return (0); #ifdef RN_DEBUG /* Get us out of the creation list */ for (t = rn_clist; t && t->rn_ybro != tt; t = t->rn_ybro) {} if (t) t->rn_ybro = tt->rn_ybro; #endif t = tt->rn_parent; dupedkey = saved_tt->rn_dupedkey; if (dupedkey) { /* * Here, tt is the deletion target and * saved_tt is the head of the dupekey chain. */ if (tt == saved_tt) { /* remove from head of chain */ x = dupedkey; x->rn_parent = t; if (t->rn_left == tt) t->rn_left = x; else t->rn_right = x; } else { /* find node in front of tt on the chain */ for (x = p = saved_tt; p && p->rn_dupedkey != tt;) p = p->rn_dupedkey; if (p) { p->rn_dupedkey = tt->rn_dupedkey; if (tt->rn_dupedkey) /* parent */ tt->rn_dupedkey->rn_parent = p; /* parent */ } else log(LOG_ERR, "rn_delete: couldn't find us\n"); } t = tt + 1; if (t->rn_flags & RNF_ACTIVE) { #ifndef RN_DEBUG *++x = *t; p = t->rn_parent; #else b = t->rn_info; *++x = *t; t->rn_info = b; p = t->rn_parent; #endif if (p->rn_left == t) p->rn_left = x; else p->rn_right = x; x->rn_left->rn_parent = x; x->rn_right->rn_parent = x; } goto out; } if (t->rn_left == tt) x = t->rn_right; else x = t->rn_left; p = t->rn_parent; if (p->rn_right == t) p->rn_right = x; else p->rn_left = x; x->rn_parent = p; /* * Demote routes attached to us. */ if (t->rn_mklist) { if (x->rn_bit >= 0) { for (mp = &x->rn_mklist; (m = *mp);) mp = &m->rm_mklist; *mp = t->rn_mklist; } else { /* If there are any key,mask pairs in a sibling duped-key chain, some subset will appear sorted in the same order attached to our mklist */ for (m = t->rn_mklist; m && x; x = x->rn_dupedkey) if (m == x->rn_mklist) { struct radix_mask *mm = m->rm_mklist; x->rn_mklist = 0; if (--(m->rm_refs) < 0) - MKFree(m); + Free(m); m = mm; } if (m) log(LOG_ERR, "rn_delete: Orphaned Mask %p at %p\n", m, x); } } /* * We may be holding an active internal node in the tree. */ x = tt + 1; if (t != x) { #ifndef RN_DEBUG *t = *x; #else b = t->rn_info; *t = *x; t->rn_info = b; #endif t->rn_left->rn_parent = t; t->rn_right->rn_parent = t; p = x->rn_parent; if (p->rn_left == x) p->rn_left = t; else p->rn_right = t; } out: tt->rn_flags &= ~RNF_ACTIVE; tt[1].rn_flags &= ~RNF_ACTIVE; return (tt); } /* * This is the same as rn_walktree() except for the parameters and the * exit. */ static int rn_walktree_from(h, a, m, f, w) struct radix_node_head *h; void *a, *m; walktree_f_t *f; void *w; { int error; struct radix_node *base, *next; u_char *xa = (u_char *)a; u_char *xm = (u_char *)m; register struct radix_node *rn, *last = 0 /* shut up gcc */; int stopping = 0; int lastb; /* * rn_search_m is sort-of-open-coded here. We cannot use the * function because we need to keep track of the last node seen. */ /* printf("about to search\n"); */ for (rn = h->rnh_treetop; rn->rn_bit >= 0; ) { last = rn; /* printf("rn_bit %d, rn_bmask %x, xm[rn_offset] %x\n", rn->rn_bit, rn->rn_bmask, xm[rn->rn_offset]); */ if (!(rn->rn_bmask & xm[rn->rn_offset])) { break; } if (rn->rn_bmask & xa[rn->rn_offset]) { rn = rn->rn_right; } else { rn = rn->rn_left; } } /* printf("done searching\n"); */ /* * Two cases: either we stepped off the end of our mask, * in which case last == rn, or we reached a leaf, in which * case we want to start from the last node we looked at. * Either way, last is the node we want to start from. */ rn = last; lastb = rn->rn_bit; /* printf("rn %p, lastb %d\n", rn, lastb);*/ /* * This gets complicated because we may delete the node * while applying the function f to it, so we need to calculate * the successor node in advance. */ while (rn->rn_bit >= 0) rn = rn->rn_left; while (!stopping) { /* printf("node %p (%d)\n", rn, rn->rn_bit); */ base = rn; /* If at right child go back up, otherwise, go right */ while (rn->rn_parent->rn_right == rn && !(rn->rn_flags & RNF_ROOT)) { rn = rn->rn_parent; /* if went up beyond last, stop */ if (rn->rn_bit <= lastb) { stopping = 1; /* printf("up too far\n"); */ /* * XXX we should jump to the 'Process leaves' * part, because the values of 'rn' and 'next' * we compute will not be used. Not a big deal * because this loop will terminate, but it is * inefficient and hard to understand! */ } } /* * At the top of the tree, no need to traverse the right * half, prevent the traversal of the entire tree in the * case of default route. */ if (rn->rn_parent->rn_flags & RNF_ROOT) stopping = 1; /* Find the next *leaf* since next node might vanish, too */ for (rn = rn->rn_parent->rn_right; rn->rn_bit >= 0;) rn = rn->rn_left; next = rn; /* Process leaves */ while ((rn = base) != 0) { base = rn->rn_dupedkey; /* printf("leaf %p\n", rn); */ if (!(rn->rn_flags & RNF_ROOT) && (error = (*f)(rn, w))) return (error); } rn = next; if (rn->rn_flags & RNF_ROOT) { /* printf("root, stopping"); */ stopping = 1; } } return 0; } static int rn_walktree(h, f, w) struct radix_node_head *h; walktree_f_t *f; void *w; { int error; struct radix_node *base, *next; register struct radix_node *rn = h->rnh_treetop; /* * This gets complicated because we may delete the node * while applying the function f to it, so we need to calculate * the successor node in advance. */ /* First time through node, go left */ while (rn->rn_bit >= 0) rn = rn->rn_left; for (;;) { base = rn; /* If at right child go back up, otherwise, go right */ while (rn->rn_parent->rn_right == rn && (rn->rn_flags & RNF_ROOT) == 0) rn = rn->rn_parent; /* Find the next *leaf* since next node might vanish, too */ for (rn = rn->rn_parent->rn_right; rn->rn_bit >= 0;) rn = rn->rn_left; next = rn; /* Process leaves */ while ((rn = base)) { base = rn->rn_dupedkey; if (!(rn->rn_flags & RNF_ROOT) && (error = (*f)(rn, w))) return (error); } rn = next; if (rn->rn_flags & RNF_ROOT) return (0); } /* NOTREACHED */ } /* * Allocate and initialize an empty tree. This has 3 nodes, which are * part of the radix_node_head (in the order ) and are * marked RNF_ROOT so they cannot be freed. * The leaves have all-zero and all-one keys, with significant * bits starting at 'off'. * Return 1 on success, 0 on error. */ -int -rn_inithead(head, off) - void **head; - int off; +static int +rn_inithead_internal(void **head, int off) { register struct radix_node_head *rnh; register struct radix_node *t, *tt, *ttt; if (*head) return (1); R_Zalloc(rnh, struct radix_node_head *, sizeof (*rnh)); if (rnh == 0) return (0); #ifdef _KERNEL RADIX_NODE_HEAD_LOCK_INIT(rnh); #endif *head = rnh; t = rn_newpair(rn_zeros, off, rnh->rnh_nodes); ttt = rnh->rnh_nodes + 2; t->rn_right = ttt; t->rn_parent = t; tt = t->rn_left; /* ... which in turn is rnh->rnh_nodes */ tt->rn_flags = t->rn_flags = RNF_ROOT | RNF_ACTIVE; tt->rn_bit = -1 - off; *ttt = *tt; ttt->rn_key = rn_ones; rnh->rnh_addaddr = rn_addroute; rnh->rnh_deladdr = rn_delete; rnh->rnh_matchaddr = rn_match; rnh->rnh_lookup = rn_lookup; rnh->rnh_walktree = rn_walktree; rnh->rnh_walktree_from = rn_walktree_from; rnh->rnh_treetop = t; return (1); } -int -rn_detachhead(void **head) +static void +rn_detachhead_internal(void **head) { struct radix_node_head *rnh; KASSERT((head != NULL && *head != NULL), ("%s: head already freed", __func__)); rnh = *head; /* Free nodes. */ Free(rnh); *head = NULL; - return (1); } -void -rn_init(int maxk) +int +rn_inithead(void **head, int off) { - char *cp, *cplim; + struct radix_node_head *rnh; + + if (*head != NULL) + return (1); + + if (rn_inithead_internal(head, off) == 0) + return (0); + + rnh = (struct radix_node_head *)(*head); - max_keylen = maxk; - if (max_keylen == 0) { - log(LOG_ERR, - "rn_init: radix functions require max_keylen be set\n"); - return; + if (rn_inithead_internal((void **)&rnh->rnh_masks, 0) == 0) { + rn_detachhead_internal(head); + return (0); } - R_Malloc(rn_zeros, char *, 3 * max_keylen); - if (rn_zeros == NULL) - panic("rn_init"); - bzero(rn_zeros, 3 * max_keylen); - rn_ones = cp = rn_zeros + max_keylen; - addmask_key = cplim = rn_ones + max_keylen; - while (cp < cplim) - *cp++ = -1; - if (rn_inithead((void **)(void *)&mask_rnhead, 0) == 0) - panic("rn_init 2"); + + return (1); +} + +int +rn_detachhead(void **head) +{ + struct radix_node_head *rnh; + + KASSERT((head != NULL && *head != NULL), + ("%s: head already freed", __func__)); + + rnh = *head; + + rn_detachhead_internal((void **)&rnh->rnh_masks); + rn_detachhead_internal(head); + return (1); } + diff --git a/sys/net/radix.h b/sys/net/radix.h index f11c7d8900ea..06cd37b71e14 100644 --- a/sys/net/radix.h +++ b/sys/net/radix.h @@ -1,168 +1,168 @@ /*- * Copyright (c) 1988, 1989, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. 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 BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)radix.h 8.2 (Berkeley) 10/31/94 * $FreeBSD$ */ #ifndef _RADIX_H_ #define _RADIX_H_ #ifdef _KERNEL #include #include #include #endif #ifdef MALLOC_DECLARE MALLOC_DECLARE(M_RTABLE); #endif /* * Radix search tree node layout. */ struct radix_node { struct radix_mask *rn_mklist; /* list of masks contained in subtree */ struct radix_node *rn_parent; /* parent */ short rn_bit; /* bit offset; -1-index(netmask) */ char rn_bmask; /* node: mask for bit test*/ u_char rn_flags; /* enumerated next */ #define RNF_NORMAL 1 /* leaf contains normal route */ #define RNF_ROOT 2 /* leaf is root leaf for tree */ #define RNF_ACTIVE 4 /* This node is alive (for rtfree) */ union { struct { /* leaf only data: */ caddr_t rn_Key; /* object of search */ caddr_t rn_Mask; /* netmask, if present */ struct radix_node *rn_Dupedkey; } rn_leaf; struct { /* node only data: */ int rn_Off; /* where to start compare */ struct radix_node *rn_L;/* progeny */ struct radix_node *rn_R;/* progeny */ } rn_node; } rn_u; #ifdef RN_DEBUG int rn_info; struct radix_node *rn_twin; struct radix_node *rn_ybro; #endif }; #define rn_dupedkey rn_u.rn_leaf.rn_Dupedkey #define rn_key rn_u.rn_leaf.rn_Key #define rn_mask rn_u.rn_leaf.rn_Mask #define rn_offset rn_u.rn_node.rn_Off #define rn_left rn_u.rn_node.rn_L #define rn_right rn_u.rn_node.rn_R /* * Annotations to tree concerning potential routes applying to subtrees. */ struct radix_mask { short rm_bit; /* bit offset; -1-index(netmask) */ char rm_unused; /* cf. rn_bmask */ u_char rm_flags; /* cf. rn_flags */ struct radix_mask *rm_mklist; /* more masks to try */ union { caddr_t rmu_mask; /* the mask */ struct radix_node *rmu_leaf; /* for normal routes */ } rm_rmu; int rm_refs; /* # of references to this struct */ }; #define rm_mask rm_rmu.rmu_mask #define rm_leaf rm_rmu.rmu_leaf /* extra field would make 32 bytes */ typedef int walktree_f_t(struct radix_node *, void *); struct radix_node_head { struct radix_node *rnh_treetop; u_int rnh_gen; /* generation counter */ int rnh_multipath; /* multipath capable ? */ struct radix_node *(*rnh_addaddr) /* add based on sockaddr */ (void *v, void *mask, struct radix_node_head *head, struct radix_node nodes[]); struct radix_node *(*rnh_deladdr) /* remove based on sockaddr */ (void *v, void *mask, struct radix_node_head *head); struct radix_node *(*rnh_matchaddr) /* locate based on sockaddr */ (void *v, struct radix_node_head *head); struct radix_node *(*rnh_lookup) /* locate based on sockaddr */ (void *v, void *mask, struct radix_node_head *head); int (*rnh_walktree) /* traverse tree */ (struct radix_node_head *head, walktree_f_t *f, void *w); int (*rnh_walktree_from) /* traverse tree below a */ (struct radix_node_head *head, void *a, void *m, walktree_f_t *f, void *w); void (*rnh_close) /* do something when the last ref drops */ (struct radix_node *rn, struct radix_node_head *head); struct radix_node rnh_nodes[3]; /* empty tree for common case */ + struct radix_node_head *rnh_masks; /* Storage for our masks */ #ifdef _KERNEL struct rwlock rnh_lock; /* locks entire radix tree */ #endif }; #ifndef _KERNEL #define R_Malloc(p, t, n) (p = (t) malloc((unsigned int)(n))) #define R_Zalloc(p, t, n) (p = (t) calloc(1,(unsigned int)(n))) #define R_Free(p) free((char *)p); #else #define R_Malloc(p, t, n) (p = (t) malloc((unsigned long)(n), M_RTABLE, M_NOWAIT)) #define R_Zalloc(p, t, n) (p = (t) malloc((unsigned long)(n), M_RTABLE, M_NOWAIT | M_ZERO)) #define Free(p) free((caddr_t)p, M_RTABLE); #define RADIX_NODE_HEAD_LOCK_INIT(rnh) \ rw_init_flags(&(rnh)->rnh_lock, "radix node head", 0) #define RADIX_NODE_HEAD_LOCK(rnh) rw_wlock(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_UNLOCK(rnh) rw_wunlock(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_RLOCK(rnh) rw_rlock(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_RUNLOCK(rnh) rw_runlock(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_LOCK_TRY_UPGRADE(rnh) rw_try_upgrade(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_DESTROY(rnh) rw_destroy(&(rnh)->rnh_lock) #define RADIX_NODE_HEAD_LOCK_ASSERT(rnh) rw_assert(&(rnh)->rnh_lock, RA_LOCKED) #define RADIX_NODE_HEAD_WLOCK_ASSERT(rnh) rw_assert(&(rnh)->rnh_lock, RA_WLOCKED) #endif /* _KERNEL */ -void rn_init(int); int rn_inithead(void **, int); int rn_detachhead(void **); int rn_refines(void *, void *); struct radix_node - *rn_addmask(void *, int, int), + *rn_addmask(void *, struct radix_node_head *, int, int), *rn_addroute (void *, void *, struct radix_node_head *, struct radix_node [2]), *rn_delete(void *, void *, struct radix_node_head *), *rn_lookup (void *v_arg, void *m_arg, struct radix_node_head *head), *rn_match(void *, struct radix_node_head *); #endif /* _RADIX_H_ */ diff --git a/sys/net/route.c b/sys/net/route.c index 9b9835fa5291..55efd944b38f 100644 --- a/sys/net/route.c +++ b/sys/net/route.c @@ -1,1688 +1,1680 @@ /*- * Copyright (c) 1980, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. 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 BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)route.c 8.3.1.1 (Berkeley) 2/23/95 * $FreeBSD$ */ /************************************************************************ * Note: In this file a 'fib' is a "forwarding information base" * * Which is the new name for an in kernel routing (next hop) table. * ***********************************************************************/ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_route.h" #include "opt_mrouting.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #define RT_MAXFIBS UINT16_MAX /* Kernel config default option. */ #ifdef ROUTETABLES #if ROUTETABLES <= 0 #error "ROUTETABLES defined too low" #endif #if ROUTETABLES > RT_MAXFIBS #error "ROUTETABLES defined too big" #endif #define RT_NUMFIBS ROUTETABLES #endif /* ROUTETABLES */ /* Initialize to default if not otherwise set. */ #ifndef RT_NUMFIBS #define RT_NUMFIBS 1 #endif /* This is read-only.. */ u_int rt_numfibs = RT_NUMFIBS; SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RD, &rt_numfibs, 0, ""); /* and this can be set too big but will be fixed before it is used */ TUNABLE_INT("net.fibs", &rt_numfibs); /* * By default add routes to all fibs for new interfaces. * Once this is set to 0 then only allocate routes on interface * changes for the FIB of the caller when adding a new set of addresses * to an interface. XXX this is a shotgun aproach to a problem that needs * a more fine grained solution.. that will come. * XXX also has the problems getting the FIB from curthread which will not * always work given the fib can be overridden and prefixes can be added * from the network stack context. */ u_int rt_add_addr_allfibs = 1; SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RW, &rt_add_addr_allfibs, 0, ""); TUNABLE_INT("net.add_addr_allfibs", &rt_add_addr_allfibs); VNET_DEFINE(struct rtstat, rtstat); #define V_rtstat VNET(rtstat) VNET_DEFINE(struct radix_node_head *, rt_tables); #define V_rt_tables VNET(rt_tables) VNET_DEFINE(int, rttrash); /* routes not in table but not freed */ #define V_rttrash VNET(rttrash) /* compare two sockaddr structures */ #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0) /* * Convert a 'struct radix_node *' to a 'struct rtentry *'. * The operation can be done safely (in this code) because a * 'struct rtentry' starts with two 'struct radix_node''s, the first * one representing leaf nodes in the routing tree, which is * what the code in radix.c passes us as a 'struct radix_node'. * * But because there are a lot of assumptions in this conversion, * do not cast explicitly, but always use the macro below. */ #define RNTORT(p) ((struct rtentry *)(p)) static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */ #define V_rtzone VNET(rtzone) /* * handler for net.my_fibnum */ static int sysctl_my_fibnum(SYSCTL_HANDLER_ARGS) { int fibnum; int error; fibnum = curthread->td_proc->p_fibnum; error = sysctl_handle_int(oidp, &fibnum, 0, req); return (error); } SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD, NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller"); static __inline struct radix_node_head ** rt_tables_get_rnh_ptr(int table, int fam) { struct radix_node_head **rnh; KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.", __func__)); KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.", __func__)); /* rnh is [fib=0][af=0]. */ rnh = (struct radix_node_head **)V_rt_tables; /* Get the offset to the requested table and fam. */ rnh += table * (AF_MAX+1) + fam; return (rnh); } struct radix_node_head * rt_tables_get_rnh(int table, int fam) { return (*rt_tables_get_rnh_ptr(table, fam)); } /* * route initialization must occur before ip6_init2(), which happenas at * SI_ORDER_MIDDLE. */ static void route_init(void) { - struct domain *dom; - int max_keylen = 0; /* whack the tunable ints into line. */ if (rt_numfibs > RT_MAXFIBS) rt_numfibs = RT_MAXFIBS; if (rt_numfibs == 0) rt_numfibs = 1; - - for (dom = domains; dom; dom = dom->dom_next) - if (dom->dom_maxrtkey > max_keylen) - max_keylen = dom->dom_maxrtkey; - - rn_init(max_keylen); /* init all zeroes, all ones, mask table */ } SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); static void vnet_route_init(const void *unused __unused) { struct domain *dom; struct radix_node_head **rnh; int table; int fam; V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) * sizeof(struct radix_node_head *), M_RTABLE, M_WAITOK|M_ZERO); V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); for (dom = domains; dom; dom = dom->dom_next) { if (dom->dom_rtattach == NULL) continue; for (table = 0; table < rt_numfibs; table++) { fam = dom->dom_family; if (table != 0 && fam != AF_INET6 && fam != AF_INET) break; /* * XXX MRT rtattach will be also called from * vfs_export.c but the offset will be 0 (only for * AF_INET and AF_INET6 which don't need it anyhow). */ rnh = rt_tables_get_rnh_ptr(table, fam); if (rnh == NULL) panic("%s: rnh NULL", __func__); dom->dom_rtattach((void **)rnh, dom->dom_rtoffset); } } } VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, vnet_route_init, 0); #ifdef VIMAGE static void vnet_route_uninit(const void *unused __unused) { int table; int fam; struct domain *dom; struct radix_node_head **rnh; for (dom = domains; dom; dom = dom->dom_next) { if (dom->dom_rtdetach == NULL) continue; for (table = 0; table < rt_numfibs; table++) { fam = dom->dom_family; if (table != 0 && fam != AF_INET6 && fam != AF_INET) break; rnh = rt_tables_get_rnh_ptr(table, fam); if (rnh == NULL) panic("%s: rnh NULL", __func__); dom->dom_rtdetach((void **)rnh, dom->dom_rtoffset); } } } VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, vnet_route_uninit, 0); #endif #ifndef _SYS_SYSPROTO_H_ struct setfib_args { int fibnum; }; #endif int sys_setfib(struct thread *td, struct setfib_args *uap) { if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs) return EINVAL; td->td_proc->p_fibnum = uap->fibnum; return (0); } /* * Packet routing routines. */ void rtalloc(struct route *ro) { rtalloc_ign_fib(ro, 0UL, RT_DEFAULT_FIB); } void rtalloc_fib(struct route *ro, u_int fibnum) { rtalloc_ign_fib(ro, 0UL, fibnum); } void rtalloc_ign(struct route *ro, u_long ignore) { struct rtentry *rt; if ((rt = ro->ro_rt) != NULL) { if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP) return; RTFREE(rt); ro->ro_rt = NULL; } ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, RT_DEFAULT_FIB); if (ro->ro_rt) RT_UNLOCK(ro->ro_rt); } void rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum) { struct rtentry *rt; if ((rt = ro->ro_rt) != NULL) { if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP) return; RTFREE(rt); ro->ro_rt = NULL; } ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum); if (ro->ro_rt) RT_UNLOCK(ro->ro_rt); } /* * Look up the route that matches the address given * Or, at least try.. Create a cloned route if needed. * * The returned route, if any, is locked. */ struct rtentry * rtalloc1(struct sockaddr *dst, int report, u_long ignflags) { return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB)); } struct rtentry * rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags, u_int fibnum) { struct radix_node_head *rnh; struct radix_node *rn; struct rtentry *newrt; struct rt_addrinfo info; int err = 0, msgtype = RTM_MISS; int needlock; KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum")); switch (dst->sa_family) { case AF_INET6: case AF_INET: /* We support multiple FIBs. */ break; default: fibnum = RT_DEFAULT_FIB; break; } rnh = rt_tables_get_rnh(fibnum, dst->sa_family); newrt = NULL; if (rnh == NULL) goto miss; /* * Look up the address in the table for that Address Family */ needlock = !(ignflags & RTF_RNH_LOCKED); if (needlock) RADIX_NODE_HEAD_RLOCK(rnh); #ifdef INVARIANTS else RADIX_NODE_HEAD_LOCK_ASSERT(rnh); #endif rn = rnh->rnh_matchaddr(dst, rnh); if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) { newrt = RNTORT(rn); RT_LOCK(newrt); RT_ADDREF(newrt); if (needlock) RADIX_NODE_HEAD_RUNLOCK(rnh); goto done; } else if (needlock) RADIX_NODE_HEAD_RUNLOCK(rnh); /* * Either we hit the root or couldn't find any match, * Which basically means * "caint get there frm here" */ miss: V_rtstat.rts_unreach++; if (report) { /* * If required, report the failure to the supervising * Authorities. * For a delete, this is not an error. (report == 0) */ bzero(&info, sizeof(info)); info.rti_info[RTAX_DST] = dst; rt_missmsg_fib(msgtype, &info, 0, err, fibnum); } done: if (newrt) RT_LOCK_ASSERT(newrt); return (newrt); } /* * Remove a reference count from an rtentry. * If the count gets low enough, take it out of the routing table */ void rtfree(struct rtentry *rt) { struct radix_node_head *rnh; KASSERT(rt != NULL,("%s: NULL rt", __func__)); rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family); KASSERT(rnh != NULL,("%s: NULL rnh", __func__)); RT_LOCK_ASSERT(rt); /* * The callers should use RTFREE_LOCKED() or RTFREE(), so * we should come here exactly with the last reference. */ RT_REMREF(rt); if (rt->rt_refcnt > 0) { log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt); goto done; } /* * On last reference give the "close method" a chance * to cleanup private state. This also permits (for * IPv4 and IPv6) a chance to decide if the routing table * entry should be purged immediately or at a later time. * When an immediate purge is to happen the close routine * typically calls rtexpunge which clears the RTF_UP flag * on the entry so that the code below reclaims the storage. */ if (rt->rt_refcnt == 0 && rnh->rnh_close) rnh->rnh_close((struct radix_node *)rt, rnh); /* * If we are no longer "up" (and ref == 0) * then we can free the resources associated * with the route. */ if ((rt->rt_flags & RTF_UP) == 0) { if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) panic("rtfree 2"); /* * the rtentry must have been removed from the routing table * so it is represented in rttrash.. remove that now. */ V_rttrash--; #ifdef DIAGNOSTIC if (rt->rt_refcnt < 0) { printf("rtfree: %p not freed (neg refs)\n", rt); goto done; } #endif /* * release references on items we hold them on.. * e.g other routes and ifaddrs. */ if (rt->rt_ifa) ifa_free(rt->rt_ifa); /* * The key is separatly alloc'd so free it (see rt_setgate()). * This also frees the gateway, as they are always malloc'd * together. */ Free(rt_key(rt)); /* * and the rtentry itself of course */ RT_LOCK_DESTROY(rt); uma_zfree(V_rtzone, rt); return; } done: RT_UNLOCK(rt); } /* * 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. */ void rtredirect(struct sockaddr *dst, struct sockaddr *gateway, struct sockaddr *netmask, int flags, struct sockaddr *src) { rtredirect_fib(dst, gateway, netmask, flags, src, RT_DEFAULT_FIB); } void rtredirect_fib(struct sockaddr *dst, struct sockaddr *gateway, struct sockaddr *netmask, int flags, struct sockaddr *src, u_int fibnum) { struct rtentry *rt, *rt0 = NULL; int error = 0; short *stat = NULL; struct rt_addrinfo info; struct ifaddr *ifa; struct radix_node_head *rnh; ifa = NULL; rnh = rt_tables_get_rnh(fibnum, dst->sa_family); if (rnh == NULL) { error = EAFNOSUPPORT; goto out; } /* verify the gateway is directly reachable */ if ((ifa = ifa_ifwithnet(gateway, 0)) == NULL) { error = ENETUNREACH; goto out; } rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */ /* * 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. */ if (!(flags & RTF_DONE) && rt && (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) error = EINVAL; else if (ifa_ifwithaddr_check(gateway)) error = EHOSTUNREACH; if (error) goto done; /* * Create a new entry if we just got back a wildcard entry * or the lookup failed. This is necessary for hosts * which use routing redirects generated by smart gateways * to dynamically build the routing tables. */ if (rt == NULL || (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: rt0 = rt; rt = NULL; flags |= RTF_GATEWAY | RTF_DYNAMIC; bzero((caddr_t)&info, sizeof(info)); info.rti_info[RTAX_DST] = dst; info.rti_info[RTAX_GATEWAY] = gateway; info.rti_info[RTAX_NETMASK] = netmask; info.rti_ifa = ifa; info.rti_flags = flags; if (rt0 != NULL) RT_UNLOCK(rt0); /* drop lock to avoid LOR with RNH */ error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum); if (rt != NULL) { RT_LOCK(rt); if (rt0 != NULL) EVENTHANDLER_INVOKE(route_redirect_event, rt0, rt, dst); flags = rt->rt_flags; } if (rt0 != NULL) RTFREE(rt0); stat = &V_rtstat.rts_dynamic; } else { struct rtentry *gwrt; /* * Smash the current notion of the gateway to * this destination. Should check about netmask!!! */ rt->rt_flags |= RTF_MODIFIED; flags |= RTF_MODIFIED; stat = &V_rtstat.rts_newgateway; /* * add the key and gateway (in one malloc'd chunk). */ RT_UNLOCK(rt); RADIX_NODE_HEAD_LOCK(rnh); RT_LOCK(rt); rt_setgate(rt, rt_key(rt), gateway); gwrt = rtalloc1(gateway, 1, RTF_RNH_LOCKED); RADIX_NODE_HEAD_UNLOCK(rnh); EVENTHANDLER_INVOKE(route_redirect_event, rt, gwrt, dst); RTFREE_LOCKED(gwrt); } } else error = EHOSTUNREACH; done: if (rt) RTFREE_LOCKED(rt); out: if (error) V_rtstat.rts_badredirect++; else if (stat != NULL) (*stat)++; bzero((caddr_t)&info, sizeof(info)); info.rti_info[RTAX_DST] = dst; info.rti_info[RTAX_GATEWAY] = gateway; info.rti_info[RTAX_NETMASK] = netmask; info.rti_info[RTAX_AUTHOR] = src; rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum); if (ifa != NULL) ifa_free(ifa); } int rtioctl(u_long req, caddr_t data) { return (rtioctl_fib(req, data, RT_DEFAULT_FIB)); } /* * Routing table ioctl interface. */ int rtioctl_fib(u_long req, caddr_t data, u_int fibnum) { /* * If more ioctl commands are added here, make sure the proper * super-user checks are being performed because it is possible for * prison-root to make it this far if raw sockets have been enabled * in jails. */ #ifdef INET /* Multicast goop, grrr... */ return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP; #else /* INET */ return ENXIO; #endif /* INET */ } /* * For both ifa_ifwithroute() routines, 'ifa' is returned referenced. */ struct ifaddr * ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway) { return (ifa_ifwithroute_fib(flags, dst, gateway, RT_DEFAULT_FIB)); } struct ifaddr * ifa_ifwithroute_fib(int flags, struct sockaddr *dst, struct sockaddr *gateway, u_int fibnum) { register struct ifaddr *ifa; int not_found = 0; 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 = NULL; if (flags & RTF_HOST) ifa = ifa_ifwithdstaddr(dst); if (ifa == NULL) 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); } if (ifa == NULL) ifa = ifa_ifwithnet(gateway, 0); if (ifa == NULL) { struct rtentry *rt = rtalloc1_fib(gateway, 0, RTF_RNH_LOCKED, fibnum); if (rt == NULL) return (NULL); /* * dismiss a gateway that is reachable only * through the default router */ switch (gateway->sa_family) { case AF_INET: if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY) not_found = 1; break; case AF_INET6: if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr)) not_found = 1; break; default: break; } if (!not_found && rt->rt_ifa != NULL) { ifa = rt->rt_ifa; ifa_ref(ifa); } RT_REMREF(rt); RT_UNLOCK(rt); if (not_found || ifa == NULL) return (NULL); } if (ifa->ifa_addr->sa_family != dst->sa_family) { struct ifaddr *oifa = ifa; ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); if (ifa == NULL) ifa = oifa; else ifa_free(oifa); } return (ifa); } /* * Do appropriate manipulations of a routing tree given * all the bits of info needed */ int rtrequest(int req, struct sockaddr *dst, struct sockaddr *gateway, struct sockaddr *netmask, int flags, struct rtentry **ret_nrt) { return (rtrequest_fib(req, dst, gateway, netmask, flags, ret_nrt, RT_DEFAULT_FIB)); } int rtrequest_fib(int req, struct sockaddr *dst, struct sockaddr *gateway, struct sockaddr *netmask, int flags, struct rtentry **ret_nrt, u_int fibnum) { struct rt_addrinfo info; if (dst->sa_len == 0) return(EINVAL); bzero((caddr_t)&info, sizeof(info)); info.rti_flags = flags; info.rti_info[RTAX_DST] = dst; info.rti_info[RTAX_GATEWAY] = gateway; info.rti_info[RTAX_NETMASK] = netmask; return rtrequest1_fib(req, &info, ret_nrt, fibnum); } /* * These (questionable) definitions of apparent local variables apply * to the next two functions. XXXXXX!!! */ #define dst info->rti_info[RTAX_DST] #define gateway info->rti_info[RTAX_GATEWAY] #define netmask info->rti_info[RTAX_NETMASK] #define ifaaddr info->rti_info[RTAX_IFA] #define ifpaddr info->rti_info[RTAX_IFP] #define flags info->rti_flags int rt_getifa(struct rt_addrinfo *info) { return (rt_getifa_fib(info, RT_DEFAULT_FIB)); } /* * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined, * it will be referenced so the caller must free it. */ int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum) { struct ifaddr *ifa; int error = 0; /* * ifp may be specified by sockaddr_dl * when protocol address is ambiguous. */ if (info->rti_ifp == NULL && ifpaddr != NULL && ifpaddr->sa_family == AF_LINK && (ifa = ifa_ifwithnet(ifpaddr, 0)) != NULL) { info->rti_ifp = ifa->ifa_ifp; ifa_free(ifa); } if (info->rti_ifa == NULL && ifaaddr != NULL) info->rti_ifa = ifa_ifwithaddr(ifaaddr); if (info->rti_ifa == NULL) { struct sockaddr *sa; sa = ifaaddr != NULL ? ifaaddr : (gateway != NULL ? gateway : dst); if (sa != NULL && info->rti_ifp != NULL) info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp); else if (dst != NULL && gateway != NULL) info->rti_ifa = ifa_ifwithroute_fib(flags, dst, gateway, fibnum); else if (sa != NULL) info->rti_ifa = ifa_ifwithroute_fib(flags, sa, sa, fibnum); } if ((ifa = info->rti_ifa) != NULL) { if (info->rti_ifp == NULL) info->rti_ifp = ifa->ifa_ifp; } else error = ENETUNREACH; return (error); } /* * Expunges references to a route that's about to be reclaimed. * The route must be locked. */ int rtexpunge(struct rtentry *rt) { #if !defined(RADIX_MPATH) struct radix_node *rn; #else struct rt_addrinfo info; int fib; struct rtentry *rt0; #endif struct radix_node_head *rnh; struct ifaddr *ifa; int error = 0; /* * Find the correct routing tree to use for this Address Family */ rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family); RT_LOCK_ASSERT(rt); if (rnh == NULL) return (EAFNOSUPPORT); RADIX_NODE_HEAD_LOCK_ASSERT(rnh); #ifdef RADIX_MPATH fib = rt->rt_fibnum; bzero(&info, sizeof(info)); info.rti_ifp = rt->rt_ifp; info.rti_flags = RTF_RNH_LOCKED; info.rti_info[RTAX_DST] = rt_key(rt); info.rti_info[RTAX_GATEWAY] = rt->rt_ifa->ifa_addr; RT_UNLOCK(rt); error = rtrequest1_fib(RTM_DELETE, &info, &rt0, fib); if (error == 0 && rt0 != NULL) { rt = rt0; RT_LOCK(rt); } else if (error != 0) { RT_LOCK(rt); return (error); } #else /* * Remove the item from the tree; it should be there, * but when callers invoke us blindly it may not (sigh). */ rn = rnh->rnh_deladdr(rt_key(rt), rt_mask(rt), rnh); if (rn == NULL) { error = ESRCH; goto bad; } KASSERT((rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) == 0, ("unexpected flags 0x%x", rn->rn_flags)); KASSERT(rt == RNTORT(rn), ("lookup mismatch, rt %p rn %p", rt, rn)); #endif /* RADIX_MPATH */ rt->rt_flags &= ~RTF_UP; /* * Give the protocol a chance to keep things in sync. */ if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) { struct rt_addrinfo info; bzero((caddr_t)&info, sizeof(info)); info.rti_flags = rt->rt_flags; info.rti_info[RTAX_DST] = rt_key(rt); info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; info.rti_info[RTAX_NETMASK] = rt_mask(rt); ifa->ifa_rtrequest(RTM_DELETE, rt, &info); } /* * one more rtentry floating around that is not * linked to the routing table. */ V_rttrash++; #if !defined(RADIX_MPATH) bad: #endif return (error); } #ifdef RADIX_MPATH static int rn_mpath_update(int req, struct rt_addrinfo *info, struct radix_node_head *rnh, struct rtentry **ret_nrt) { /* * if we got multipath routes, we require users to specify * a matching RTAX_GATEWAY. */ struct rtentry *rt, *rto = NULL; register struct radix_node *rn; int error = 0; rn = rnh->rnh_matchaddr(dst, rnh); if (rn == NULL) return (ESRCH); rto = rt = RNTORT(rn); rt = rt_mpath_matchgate(rt, gateway); if (rt == NULL) return (ESRCH); /* * this is the first entry in the chain */ if (rto == rt) { rn = rn_mpath_next((struct radix_node *)rt); /* * there is another entry, now it's active */ if (rn) { rto = RNTORT(rn); RT_LOCK(rto); rto->rt_flags |= RTF_UP; RT_UNLOCK(rto); } else if (rt->rt_flags & RTF_GATEWAY) { /* * For gateway routes, we need to * make sure that we we are deleting * the correct gateway. * rt_mpath_matchgate() does not * check the case when there is only * one route in the chain. */ if (gateway && (rt->rt_gateway->sa_len != gateway->sa_len || memcmp(rt->rt_gateway, gateway, gateway->sa_len))) error = ESRCH; else { /* * remove from tree before returning it * to the caller */ rn = rnh->rnh_deladdr(dst, netmask, rnh); KASSERT(rt == RNTORT(rn), ("radix node disappeared")); goto gwdelete; } } /* * use the normal delete code to remove * the first entry */ if (req != RTM_DELETE) goto nondelete; error = ENOENT; goto done; } /* * if the entry is 2nd and on up */ if ((req == RTM_DELETE) && !rt_mpath_deldup(rto, rt)) panic ("rtrequest1: rt_mpath_deldup"); gwdelete: RT_LOCK(rt); RT_ADDREF(rt); if (req == RTM_DELETE) { rt->rt_flags &= ~RTF_UP; /* * One more rtentry floating around that is not * linked to the routing table. rttrash will be decremented * when RTFREE(rt) is eventually called. */ V_rttrash++; } nondelete: if (req != RTM_DELETE) panic("unrecognized request %d", req); /* * If the caller wants it, then it can have it, * but it's up to it to free the rtentry as we won't be * doing it. */ if (ret_nrt) { *ret_nrt = rt; RT_UNLOCK(rt); } else RTFREE_LOCKED(rt); done: return (error); } #endif int rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt, u_int fibnum) { int error = 0, needlock = 0; register struct rtentry *rt; #ifdef FLOWTABLE register struct rtentry *rt0; #endif register struct radix_node *rn; register struct radix_node_head *rnh; struct ifaddr *ifa; struct sockaddr *ndst; struct sockaddr_storage mdst; #define senderr(x) { error = x ; goto bad; } KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum")); switch (dst->sa_family) { case AF_INET6: case AF_INET: /* We support multiple FIBs. */ break; default: fibnum = RT_DEFAULT_FIB; break; } /* * Find the correct routing tree to use for this Address Family */ rnh = rt_tables_get_rnh(fibnum, dst->sa_family); if (rnh == NULL) return (EAFNOSUPPORT); needlock = ((flags & RTF_RNH_LOCKED) == 0); flags &= ~RTF_RNH_LOCKED; if (needlock) RADIX_NODE_HEAD_LOCK(rnh); else RADIX_NODE_HEAD_LOCK_ASSERT(rnh); /* * If we are adding a host route then we don't want to put * a netmask in the tree, nor do we want to clone it. */ if (flags & RTF_HOST) netmask = NULL; switch (req) { case RTM_DELETE: if (netmask) { rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask); dst = (struct sockaddr *)&mdst; } #ifdef RADIX_MPATH if (rn_mpath_capable(rnh)) { error = rn_mpath_update(req, info, rnh, ret_nrt); /* * "bad" holds true for the success case * as well */ if (error != ENOENT) goto bad; error = 0; } #endif if ((flags & RTF_PINNED) == 0) { /* Check if target route can be deleted */ rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, rnh); if ((rt != NULL) && (rt->rt_flags & RTF_PINNED)) senderr(EADDRINUSE); } /* * Remove the item from the tree and return it. * Complain if it is not there and do no more processing. */ rn = rnh->rnh_deladdr(dst, netmask, rnh); if (rn == NULL) senderr(ESRCH); if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) panic ("rtrequest delete"); rt = RNTORT(rn); RT_LOCK(rt); RT_ADDREF(rt); rt->rt_flags &= ~RTF_UP; /* * give the protocol a chance to keep things in sync. */ if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) ifa->ifa_rtrequest(RTM_DELETE, rt, info); /* * One more rtentry floating around that is not * linked to the routing table. rttrash will be decremented * when RTFREE(rt) is eventually called. */ V_rttrash++; /* * If the caller wants it, then it can have it, * but it's up to it to free the rtentry as we won't be * doing it. */ if (ret_nrt) { *ret_nrt = rt; RT_UNLOCK(rt); } else RTFREE_LOCKED(rt); break; case RTM_RESOLVE: /* * resolve was only used for route cloning * here for compat */ break; case RTM_ADD: if ((flags & RTF_GATEWAY) && !gateway) senderr(EINVAL); if (dst && gateway && (dst->sa_family != gateway->sa_family) && (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK)) senderr(EINVAL); if (info->rti_ifa == NULL) { error = rt_getifa_fib(info, fibnum); if (error) senderr(error); } else ifa_ref(info->rti_ifa); ifa = info->rti_ifa; rt = uma_zalloc(V_rtzone, M_NOWAIT | M_ZERO); if (rt == NULL) { ifa_free(ifa); senderr(ENOBUFS); } RT_LOCK_INIT(rt); rt->rt_flags = RTF_UP | flags; rt->rt_fibnum = fibnum; /* * Add the gateway. Possibly re-malloc-ing the storage for it. */ RT_LOCK(rt); if ((error = rt_setgate(rt, dst, gateway)) != 0) { RT_LOCK_DESTROY(rt); ifa_free(ifa); uma_zfree(V_rtzone, rt); senderr(error); } /* * point to the (possibly newly malloc'd) dest address. */ ndst = (struct sockaddr *)rt_key(rt); /* * make sure it contains the value we want (masked if needed). */ if (netmask) { rt_maskedcopy(dst, ndst, netmask); } else bcopy(dst, ndst, dst->sa_len); /* * We use the ifa reference returned by rt_getifa_fib(). * This moved from below so that rnh->rnh_addaddr() can * examine the ifa and ifa->ifa_ifp if it so desires. */ rt->rt_ifa = ifa; rt->rt_ifp = ifa->ifa_ifp; rt->rt_rmx.rmx_weight = 1; #ifdef RADIX_MPATH /* do not permit exactly the same dst/mask/gw pair */ if (rn_mpath_capable(rnh) && rt_mpath_conflict(rnh, rt, netmask)) { ifa_free(rt->rt_ifa); Free(rt_key(rt)); RT_LOCK_DESTROY(rt); uma_zfree(V_rtzone, rt); senderr(EEXIST); } #endif #ifdef FLOWTABLE rt0 = NULL; /* "flow-table" only supports IPv6 and IPv4 at the moment. */ switch (dst->sa_family) { #ifdef INET6 case AF_INET6: #endif #ifdef INET case AF_INET: #endif #if defined(INET6) || defined(INET) rn = rnh->rnh_matchaddr(dst, rnh); if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) { struct sockaddr *mask; u_char *m, *n; int len; /* * compare mask to see if the new route is * more specific than the existing one */ rt0 = RNTORT(rn); RT_LOCK(rt0); RT_ADDREF(rt0); RT_UNLOCK(rt0); /* * A host route is already present, so * leave the flow-table entries as is. */ if (rt0->rt_flags & RTF_HOST) { RTFREE(rt0); rt0 = NULL; } else if (!(flags & RTF_HOST) && netmask) { mask = rt_mask(rt0); len = mask->sa_len; m = (u_char *)mask; n = (u_char *)netmask; while (len-- > 0) { if (*n != *m) break; n++; m++; } if (len == 0 || (*n < *m)) { RTFREE(rt0); rt0 = NULL; } } } #endif/* INET6 || INET */ } #endif /* FLOWTABLE */ /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes); /* * If it still failed to go into the tree, * then un-make it (this should be a function) */ if (rn == NULL) { ifa_free(rt->rt_ifa); Free(rt_key(rt)); RT_LOCK_DESTROY(rt); uma_zfree(V_rtzone, rt); #ifdef FLOWTABLE if (rt0 != NULL) RTFREE(rt0); #endif senderr(EEXIST); } #ifdef FLOWTABLE else if (rt0 != NULL) { switch (dst->sa_family) { #ifdef INET6 case AF_INET6: flowtable_route_flush(V_ip6_ft, rt0); break; #endif #ifdef INET case AF_INET: flowtable_route_flush(V_ip_ft, rt0); break; #endif } RTFREE(rt0); } #endif /* * If this protocol has something to add to this then * allow it to do that as well. */ if (ifa->ifa_rtrequest) ifa->ifa_rtrequest(req, rt, info); /* * actually return a resultant rtentry and * give the caller a single reference. */ if (ret_nrt) { *ret_nrt = rt; RT_ADDREF(rt); } RT_UNLOCK(rt); break; default: error = EOPNOTSUPP; } bad: if (needlock) RADIX_NODE_HEAD_UNLOCK(rnh); return (error); #undef senderr } #undef dst #undef gateway #undef netmask #undef ifaaddr #undef ifpaddr #undef flags int rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate) { /* XXX dst may be overwritten, can we move this to below */ int dlen = SA_SIZE(dst), glen = SA_SIZE(gate); #ifdef INVARIANTS struct radix_node_head *rnh; rnh = rt_tables_get_rnh(rt->rt_fibnum, dst->sa_family); #endif RT_LOCK_ASSERT(rt); RADIX_NODE_HEAD_LOCK_ASSERT(rnh); /* * Prepare to store the gateway in rt->rt_gateway. * Both dst and gateway are stored one after the other in the same * malloc'd chunk. If we have room, we can reuse the old buffer, * rt_gateway already points to the right place. * Otherwise, malloc a new block and update the 'dst' address. */ if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) { caddr_t new; R_Malloc(new, caddr_t, dlen + glen); if (new == NULL) return ENOBUFS; /* * XXX note, we copy from *dst and not *rt_key(rt) because * rt_setgate() can be called to initialize a newly * allocated route entry, in which case rt_key(rt) == NULL * (and also rt->rt_gateway == NULL). * Free()/free() handle a NULL argument just fine. */ bcopy(dst, new, dlen); Free(rt_key(rt)); /* free old block, if any */ rt_key(rt) = (struct sockaddr *)new; rt->rt_gateway = (struct sockaddr *)(new + dlen); } /* * Copy the new gateway value into the memory chunk. */ bcopy(gate, rt->rt_gateway, glen); return (0); } void rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *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. */ #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */ static inline int rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum) { struct sockaddr *dst; struct sockaddr *netmask; struct rtentry *rt = NULL; struct rt_addrinfo info; int error = 0; int startfib, endfib; char tempbuf[_SOCKADDR_TMPSIZE]; int didwork = 0; int a_failure = 0; static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; struct radix_node_head *rnh; if (flags & RTF_HOST) { dst = ifa->ifa_dstaddr; netmask = NULL; } else { dst = ifa->ifa_addr; netmask = ifa->ifa_netmask; } if (dst->sa_len == 0) return(EINVAL); switch (dst->sa_family) { case AF_INET6: case AF_INET: /* We support multiple FIBs. */ break; default: fibnum = RT_DEFAULT_FIB; break; } if (fibnum == -1) { if (rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD) { startfib = endfib = curthread->td_proc->p_fibnum; } else { startfib = 0; endfib = rt_numfibs - 1; } } else { KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum")); startfib = fibnum; endfib = fibnum; } /* * If it's a delete, check that if it exists, * it's on the correct interface or we might scrub * a route to another ifa which would * be confusing at best and possibly worse. */ if (cmd == RTM_DELETE) { /* * It's a delete, so it should already exist.. * If it's a net, mask off the host bits * (Assuming we have a mask) * XXX this is kinda inet specific.. */ if (netmask != NULL) { rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask); dst = (struct sockaddr *)tempbuf; } } /* * Now go through all the requested tables (fibs) and do the * requested action. Realistically, this will either be fib 0 * for protocols that don't do multiple tables or all the * tables for those that do. */ for ( fibnum = startfib; fibnum <= endfib; fibnum++) { if (cmd == RTM_DELETE) { struct radix_node *rn; /* * Look up an rtentry that is in the routing tree and * contains the correct info. */ rnh = rt_tables_get_rnh(fibnum, dst->sa_family); if (rnh == NULL) /* this table doesn't exist but others might */ continue; RADIX_NODE_HEAD_RLOCK(rnh); #ifdef RADIX_MPATH if (rn_mpath_capable(rnh)) { rn = rnh->rnh_matchaddr(dst, rnh); if (rn == NULL) error = ESRCH; else { rt = RNTORT(rn); /* * for interface route the * rt->rt_gateway is sockaddr_intf * for cloning ARP entries, so * rt_mpath_matchgate must use the * interface address */ rt = rt_mpath_matchgate(rt, ifa->ifa_addr); if (!rt) error = ESRCH; } } else #endif rn = rnh->rnh_lookup(dst, netmask, rnh); error = (rn == NULL || (rn->rn_flags & RNF_ROOT) || RNTORT(rn)->rt_ifa != ifa || !sa_equal((struct sockaddr *)rn->rn_key, dst)); RADIX_NODE_HEAD_RUNLOCK(rnh); if (error) { /* this is only an error if bad on ALL tables */ continue; } } /* * Do the actual request */ bzero((caddr_t)&info, sizeof(info)); info.rti_ifa = ifa; info.rti_flags = flags | (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED; info.rti_info[RTAX_DST] = dst; /* * doing this for compatibility reasons */ if (cmd == RTM_ADD) info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; else info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; info.rti_info[RTAX_NETMASK] = netmask; error = rtrequest1_fib(cmd, &info, &rt, fibnum); if ((error == EEXIST) && (cmd == RTM_ADD)) { /* * Interface route addition failed. * Atomically delete current prefix generating * RTM_DELETE message, and retry adding * interface prefix. */ rnh = rt_tables_get_rnh(fibnum, dst->sa_family); RADIX_NODE_HEAD_LOCK(rnh); /* Delete old prefix */ info.rti_ifa = NULL; info.rti_flags = RTF_RNH_LOCKED; error = rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum); if (error == 0) { info.rti_ifa = ifa; info.rti_flags = flags | RTF_RNH_LOCKED | (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED; error = rtrequest1_fib(cmd, &info, &rt, fibnum); } RADIX_NODE_HEAD_UNLOCK(rnh); } if (error == 0 && rt != NULL) { /* * notify any listening routing agents of the change */ RT_LOCK(rt); #ifdef RADIX_MPATH /* * in case address alias finds the first address * e.g. ifconfig bge0 192.0.2.246/24 * e.g. ifconfig bge0 192.0.2.247/24 * the address set in the route is 192.0.2.246 * so we need to replace it with 192.0.2.247 */ if (memcmp(rt->rt_ifa->ifa_addr, ifa->ifa_addr, ifa->ifa_addr->sa_len)) { ifa_free(rt->rt_ifa); ifa_ref(ifa); rt->rt_ifp = ifa->ifa_ifp; rt->rt_ifa = ifa; } #endif /* * doing this for compatibility reasons */ if (cmd == RTM_ADD) { ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = rt->rt_ifp->if_type; ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = rt->rt_ifp->if_index; } RT_ADDREF(rt); RT_UNLOCK(rt); rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum); RT_LOCK(rt); RT_REMREF(rt); if (cmd == RTM_DELETE) { /* * If we are deleting, and we found an entry, * then it's been removed from the tree.. * now throw it away. */ RTFREE_LOCKED(rt); } else { if (cmd == RTM_ADD) { /* * We just wanted to add it.. * we don't actually need a reference. */ RT_REMREF(rt); } RT_UNLOCK(rt); } didwork = 1; } if (error) a_failure = error; } if (cmd == RTM_DELETE) { if (didwork) { error = 0; } else { /* we only give an error if it wasn't in any table */ error = ((flags & RTF_HOST) ? EHOSTUNREACH : ENETUNREACH); } } else { if (a_failure) { /* return an error if any of them failed */ error = a_failure; } } return (error); } #ifndef BURN_BRIDGES /* special one for inet internal use. may not use. */ int rtinit_fib(struct ifaddr *ifa, int cmd, int flags) { return (rtinit1(ifa, cmd, flags, -1)); } #endif /* * Set up a routing table entry, normally * for an interface. */ int rtinit(struct ifaddr *ifa, int cmd, int flags) { struct sockaddr *dst; int fib = RT_DEFAULT_FIB; if (flags & RTF_HOST) { dst = ifa->ifa_dstaddr; } else { dst = ifa->ifa_addr; } switch (dst->sa_family) { case AF_INET6: case AF_INET: /* We do support multiple FIBs. */ fib = -1; break; } return (rtinit1(ifa, cmd, flags, fib)); }