xref: /freebsd/sys/net/route/nhop.c (revision fac71e4e09885bb2afa3d984a0c239a52e1a7418)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2020 Alexander V. Chernikov
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 #include "opt_inet.h"
31 #include "opt_route.h"
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/lock.h>
36 #include <sys/rwlock.h>
37 #include <sys/malloc.h>
38 #include <sys/mbuf.h>
39 #include <sys/socket.h>
40 #include <sys/kernel.h>
41 
42 #include <net/if.h>
43 #include <net/if_var.h>
44 #include <net/route.h>
45 #include <net/route/route_var.h>
46 #include <net/route/nhop_utils.h>
47 #include <net/route/nhop.h>
48 #include <net/route/nhop_var.h>
49 #include <net/vnet.h>
50 
51 #define	DEBUG_MOD_NAME	nhop
52 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
53 #include <net/route/route_debug.h>
54 _DECLARE_DEBUG(LOG_INFO);
55 
56 /*
57  * This file contains data structures management logic for the nexthop ("nhop")
58  *   route subsystem.
59  *
60  * Nexthops in the original sense are the objects containing all the necessary
61  * information to forward the packet to the selected destination.
62  * In particular, nexthop is defined by a combination of
63  *  ifp, ifa, aifp, mtu, gw addr(if set), nh_type, nh_family, mask of rt_flags and
64  *    NHF_DEFAULT
65  *
66  * All nexthops are stored in the resizable hash table.
67  * Additionally, each nexthop gets assigned its unique index (nexthop index)
68  * so userland programs can interact with the nexthops easier. Index allocation
69  * is backed by the bitmask array.
70  */
71 
72 MALLOC_DEFINE(M_NHOP, "nhops", "nexthops data");
73 
74 /* Hash management functions */
75 
76 int
77 nhops_init_rib(struct rib_head *rh)
78 {
79 	struct nh_control *ctl;
80 	size_t alloc_size;
81 	uint32_t num_buckets, num_items;
82 	void *ptr;
83 
84 	ctl = malloc(sizeof(struct nh_control), M_NHOP, M_WAITOK | M_ZERO);
85 
86 	/*
87 	 * Allocate nexthop hash. Start with 16 items by default (128 bytes).
88 	 * This will be enough for most of the cases.
89 	 */
90 	num_buckets = 16;
91 	alloc_size = CHT_SLIST_GET_RESIZE_SIZE(num_buckets);
92 	ptr = malloc(alloc_size, M_NHOP, M_WAITOK | M_ZERO);
93 	CHT_SLIST_INIT(&ctl->nh_head, ptr, num_buckets);
94 
95 	/*
96 	 * Allocate nexthop index bitmask.
97 	 */
98 	num_items = 128 * 8; /* 128 bytes */
99 	ptr = malloc(bitmask_get_size(num_items), M_NHOP, M_WAITOK | M_ZERO);
100 	bitmask_init(&ctl->nh_idx_head, ptr, num_items);
101 
102 	NHOPS_LOCK_INIT(ctl);
103 
104 	rh->nh_control = ctl;
105 	ctl->ctl_rh = rh;
106 
107 	FIB_CTL_LOG(LOG_DEBUG2, ctl, "nhops init: ctl %p rh %p", ctl, rh);
108 
109 	return (0);
110 }
111 
112 static void
113 destroy_ctl(struct nh_control *ctl)
114 {
115 
116 	NHOPS_LOCK_DESTROY(ctl);
117 	free(ctl->nh_head.ptr, M_NHOP);
118 	free(ctl->nh_idx_head.idx, M_NHOP);
119 #ifdef ROUTE_MPATH
120 	nhgrp_ctl_free(ctl);
121 #endif
122 	free(ctl, M_NHOP);
123 }
124 
125 /*
126  * Epoch callback indicating ctl is safe to destroy
127  */
128 static void
129 destroy_ctl_epoch(epoch_context_t ctx)
130 {
131 	struct nh_control *ctl;
132 
133 	ctl = __containerof(ctx, struct nh_control, ctl_epoch_ctx);
134 
135 	destroy_ctl(ctl);
136 }
137 
138 void
139 nhops_destroy_rib(struct rib_head *rh)
140 {
141 	struct nh_control *ctl;
142 	struct nhop_priv *nh_priv;
143 
144 	ctl = rh->nh_control;
145 
146 	/*
147 	 * All routes should have been deleted in rt_table_destroy().
148 	 * However, TCP stack or other consumers may store referenced
149 	 *  nexthop pointers. When these references go to zero,
150 	 *  nhop_free() will try to unlink these records from the
151 	 *  datastructures, most likely leading to panic.
152 	 *
153 	 * Avoid that by explicitly marking all of the remaining
154 	 *  nexthops as unlinked by removing a reference from a special
155 	 *  counter. Please see nhop_free() comments for more
156 	 *  details.
157 	 */
158 
159 	NHOPS_WLOCK(ctl);
160 	CHT_SLIST_FOREACH(&ctl->nh_head, nhops, nh_priv) {
161 		FIB_RH_LOG(LOG_DEBUG3, rh, "marking nhop %u unlinked", nh_priv->nh_idx);
162 		refcount_release(&nh_priv->nh_linked);
163 	} CHT_SLIST_FOREACH_END;
164 #ifdef ROUTE_MPATH
165 	nhgrp_ctl_unlink_all(ctl);
166 #endif
167 	NHOPS_WUNLOCK(ctl);
168 
169 	/*
170 	 * Postpone destruction till the end of current epoch
171 	 * so nhop_free() can safely use nh_control pointer.
172 	 */
173 	NET_EPOCH_CALL(destroy_ctl_epoch, &ctl->ctl_epoch_ctx);
174 }
175 
176 /*
177  * Nexhop hash calculation:
178  *
179  * Nexthops distribution:
180  * 2 "mandatory" nexthops per interface ("interface route", "loopback").
181  * For direct peering: 1 nexthop for the peering router per ifp/af.
182  * For Ix-like peering: tens to hundreds nexthops of neghbors per ifp/af.
183  * IGP control plane & broadcast segment: tens of nexthops per ifp/af.
184  *
185  * Each fib/af combination has its own hash table.
186  * With that in mind, hash nexthops by the combination of the interface
187  *  and GW IP address.
188  *
189  * To optimize hash calculation, ignore lower bits of ifnet pointer,
190  * as they  give very little entropy.
191  * Similarly, use lower 4 bytes of IPv6 address to distinguish between the
192  *  neighbors.
193  */
194 struct _hash_data {
195 	uint16_t	ifentropy;
196 	uint8_t		family;
197 	uint8_t		nh_type;
198 	uint32_t	gw_addr;
199 };
200 
201 static unsigned
202 djb_hash(const unsigned char *h, const int len)
203 {
204 	unsigned int result = 0;
205 	int i;
206 
207 	for (i = 0; i < len; i++)
208 		result = 33 * result ^ h[i];
209 
210 	return (result);
211 }
212 
213 static uint32_t
214 hash_priv(const struct nhop_priv *priv)
215 {
216 	struct nhop_object *nh = priv->nh;
217 	struct _hash_data key = {
218 	    .ifentropy = (uint16_t)((((uintptr_t)nh->nh_ifp) >> 6) & 0xFFFF),
219 	    .family = nh->gw_sa.sa_family,
220 	    .nh_type = priv->nh_type & 0xFF,
221 	    .gw_addr = (nh->gw_sa.sa_family == AF_INET6) ?
222 		nh->gw6_sa.sin6_addr.s6_addr32[3] :
223 		nh->gw4_sa.sin_addr.s_addr
224 	};
225 
226 	return (uint32_t)(djb_hash((const unsigned char *)&key, sizeof(key)));
227 }
228 
229 /*
230  * Checks if hash needs resizing and performs this resize if necessary
231  *
232  */
233 static void
234 consider_resize(struct nh_control *ctl, uint32_t new_nh_buckets, uint32_t new_idx_items)
235 {
236 	void *nh_ptr, *nh_idx_ptr;
237 	void *old_idx_ptr;
238 	size_t alloc_size;
239 
240 	nh_ptr = NULL;
241 	if (new_nh_buckets != 0) {
242 		alloc_size = CHT_SLIST_GET_RESIZE_SIZE(new_nh_buckets);
243 		nh_ptr = malloc(alloc_size, M_NHOP, M_NOWAIT | M_ZERO);
244 	}
245 
246 	nh_idx_ptr = NULL;
247 	if (new_idx_items != 0) {
248 		alloc_size = bitmask_get_size(new_idx_items);
249 		nh_idx_ptr = malloc(alloc_size, M_NHOP, M_NOWAIT | M_ZERO);
250 	}
251 
252 	if (nh_ptr == NULL && nh_idx_ptr == NULL) {
253 		/* Either resize is not required or allocations have failed. */
254 		return;
255 	}
256 
257 	FIB_CTL_LOG(LOG_DEBUG, ctl,
258 	    "going to resize: nh:[ptr:%p sz:%u] idx:[ptr:%p sz:%u]",
259 	    nh_ptr, new_nh_buckets, nh_idx_ptr, new_idx_items);
260 
261 	old_idx_ptr = NULL;
262 
263 	NHOPS_WLOCK(ctl);
264 	if (nh_ptr != NULL) {
265 		CHT_SLIST_RESIZE(&ctl->nh_head, nhops, nh_ptr, new_nh_buckets);
266 	}
267 	if (nh_idx_ptr != NULL) {
268 		if (bitmask_copy(&ctl->nh_idx_head, nh_idx_ptr, new_idx_items) == 0)
269 			bitmask_swap(&ctl->nh_idx_head, nh_idx_ptr, new_idx_items, &old_idx_ptr);
270 	}
271 	NHOPS_WUNLOCK(ctl);
272 
273 	if (nh_ptr != NULL)
274 		free(nh_ptr, M_NHOP);
275 	if (old_idx_ptr != NULL)
276 		free(old_idx_ptr, M_NHOP);
277 }
278 
279 /*
280  * Links nextop @nh_priv to the nexhop hash table and allocates
281  *  nexhop index.
282  * Returns allocated index or 0 on failure.
283  */
284 int
285 link_nhop(struct nh_control *ctl, struct nhop_priv *nh_priv)
286 {
287 	uint16_t idx;
288 	uint32_t num_buckets_new, num_items_new;
289 
290 	KASSERT((nh_priv->nh_idx == 0), ("nhop index is already allocated"));
291 	NHOPS_WLOCK(ctl);
292 
293 	/*
294 	 * Check if we need to resize hash and index.
295 	 * The following 2 functions returns either new size or 0
296 	 *  if resize is not required.
297 	 */
298 	num_buckets_new = CHT_SLIST_GET_RESIZE_BUCKETS(&ctl->nh_head);
299 	num_items_new = bitmask_get_resize_items(&ctl->nh_idx_head);
300 
301 	if (bitmask_alloc_idx(&ctl->nh_idx_head, &idx) != 0) {
302 		NHOPS_WUNLOCK(ctl);
303 		FIB_CTL_LOG(LOG_INFO, ctl, "Unable to allocate nhop index");
304 		RTSTAT_INC(rts_nh_idx_alloc_failure);
305 		consider_resize(ctl, num_buckets_new, num_items_new);
306 		return (0);
307 	}
308 
309 	nh_priv->nh_idx = idx;
310 	nh_priv->nh_control = ctl;
311 	nh_priv->nh_finalized = 1;
312 
313 	CHT_SLIST_INSERT_HEAD(&ctl->nh_head, nhops, nh_priv);
314 
315 	NHOPS_WUNLOCK(ctl);
316 
317 	FIB_RH_LOG(LOG_DEBUG2, ctl->ctl_rh,
318 	    "Linked nhop priv %p to %d, hash %u, ctl %p",
319 	    nh_priv, idx, hash_priv(nh_priv), ctl);
320 	consider_resize(ctl, num_buckets_new, num_items_new);
321 
322 	return (idx);
323 }
324 
325 /*
326  * Unlinks nexthop specified by @nh_priv data from the hash.
327  *
328  * Returns found nexthop or NULL.
329  */
330 struct nhop_priv *
331 unlink_nhop(struct nh_control *ctl, struct nhop_priv *nh_priv_del)
332 {
333 	struct nhop_priv *priv_ret;
334 	int idx;
335 	uint32_t num_buckets_new, num_items_new;
336 
337 	idx = 0;
338 
339 	NHOPS_WLOCK(ctl);
340 	CHT_SLIST_REMOVE(&ctl->nh_head, nhops, nh_priv_del, priv_ret);
341 
342 	if (priv_ret != NULL) {
343 		idx = priv_ret->nh_idx;
344 		priv_ret->nh_idx = 0;
345 
346 		KASSERT((idx != 0), ("bogus nhop index 0"));
347 		if ((bitmask_free_idx(&ctl->nh_idx_head, idx)) != 0) {
348 			FIB_CTL_LOG(LOG_DEBUG, ctl,
349 			    "Unable to remove index %d from fib %u af %d",
350 			    idx, ctl->ctl_rh->rib_fibnum, ctl->ctl_rh->rib_family);
351 		}
352 	}
353 
354 	/* Check if hash or index needs to be resized */
355 	num_buckets_new = CHT_SLIST_GET_RESIZE_BUCKETS(&ctl->nh_head);
356 	num_items_new = bitmask_get_resize_items(&ctl->nh_idx_head);
357 
358 	NHOPS_WUNLOCK(ctl);
359 
360 	if (priv_ret == NULL) {
361 		FIB_CTL_LOG(LOG_INFO, ctl,
362 		    "Unable to unlink nhop priv %p from hash, hash %u ctl %p",
363 		    nh_priv_del, hash_priv(nh_priv_del), ctl);
364 	} else {
365 		FIB_CTL_LOG(LOG_DEBUG2, ctl, "Unlinked nhop %p priv idx %d",
366 		    priv_ret, idx);
367 	}
368 
369 	consider_resize(ctl, num_buckets_new, num_items_new);
370 
371 	return (priv_ret);
372 }
373 
374 /*
375  * Searches for the nexthop by data specifcied in @nh_priv.
376  * Returns referenced nexthop or NULL.
377  */
378 struct nhop_priv *
379 find_nhop(struct nh_control *ctl, const struct nhop_priv *nh_priv)
380 {
381 	struct nhop_priv *nh_priv_ret;
382 
383 	NHOPS_RLOCK(ctl);
384 	CHT_SLIST_FIND_BYOBJ(&ctl->nh_head, nhops, nh_priv, nh_priv_ret);
385 	if (nh_priv_ret != NULL) {
386 		if (refcount_acquire_if_not_zero(&nh_priv_ret->nh_refcnt) == 0){
387 			/* refcount was 0 -> nhop is being deleted */
388 			nh_priv_ret = NULL;
389 		}
390 	}
391 	NHOPS_RUNLOCK(ctl);
392 
393 	return (nh_priv_ret);
394 }
395