summaryrefslogtreecommitdiffstats
path: root/sys/netinet/in_rmx.c
blob: baa8c71805dc7861de937a9a004f07e87ba5e013 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
/*
 * Copyright 1994, 1995 Massachusetts Institute of Technology
 *
 * Permission to use, copy, modify, and distribute this software and
 * its documentation for any purpose and without fee is hereby
 * granted, provided that both the above copyright notice and this
 * permission notice appear in all copies, that both the above
 * copyright notice and this permission notice appear in all
 * supporting documentation, and that the name of M.I.T. not be used
 * in advertising or publicity pertaining to distribution of the
 * software without specific, written prior permission.  M.I.T. makes
 * no representations about the suitability of this software for any
 * purpose.  It is provided "as is" without express or implied
 * warranty.
 *
 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
 * SHALL M.I.T. 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.
 *
 * $FreeBSD$
 */

/*
 * This code does two things necessary for the enhanced TCP metrics to
 * function in a useful manner:
 *  1) It marks all non-host routes as `cloning', thus ensuring that
 *     every actual reference to such a route actually gets turned
 *     into a reference to a host route to the specific destination
 *     requested.
 *  2) When such routes lose all their references, it arranges for them
 *     to be deleted in some random collection of circumstances, so that
 *     a large quantity of stale routing data is not kept in kernel memory
 *     indefinitely.  See in_rtqtimo() below for the exact mechanism.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/socket.h>
#include <sys/mbuf.h>
#include <sys/syslog.h>
#include <sys/callout.h>

#include <net/if.h>
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/in_var.h>
#include <netinet/ip_var.h>

extern int	in_inithead(void **head, int off);

#define RTPRF_OURS		RTF_PROTO3	/* set on routes we manage */

/*
 * Do what we need to do when inserting a route.
 */
static struct radix_node *
in_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
	    struct radix_node *treenodes)
{
	struct rtentry *rt = (struct rtentry *)treenodes;
	struct sockaddr_in *sin = (struct sockaddr_in *)rt_key(rt);
	struct radix_node *ret;

	/*
	 * For IP, all unicast non-host routes are automatically cloning.
	 */
	if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
		rt->rt_flags |= RTF_MULTICAST;

	if (!(rt->rt_flags & (RTF_HOST | RTF_CLONING | RTF_MULTICAST)))
		rt->rt_flags |= RTF_PRCLONING;

	/*
	 * A little bit of help for both IP output and input:
	 *   For host routes, we make sure that RTF_BROADCAST
	 *   is set for anything that looks like a broadcast address.
	 *   This way, we can avoid an expensive call to in_broadcast()
	 *   in ip_output() most of the time (because the route passed
	 *   to ip_output() is almost always a host route).
	 *
	 *   We also do the same for local addresses, with the thought
	 *   that this might one day be used to speed up ip_input().
	 *
	 * We also mark routes to multicast addresses as such, because
	 * it's easy to do and might be useful (but this is much more
	 * dubious since it's so easy to inspect the address).  (This
	 * is done above.)
	 */
	if (rt->rt_flags & RTF_HOST) {
		if (in_broadcast(sin->sin_addr, rt->rt_ifp)) {
			rt->rt_flags |= RTF_BROADCAST;
		} else if (satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr ==
		    sin->sin_addr.s_addr) {
			rt->rt_flags |= RTF_LOCAL;
		}
	}

	if (!rt->rt_rmx.rmx_mtu && !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
	    rt->rt_ifp)
		rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;

	ret = rn_addroute(v_arg, n_arg, head, treenodes);
	if (ret == NULL && rt->rt_flags & RTF_HOST) {
		struct rtentry *rt2;
		/*
		 * We are trying to add a host route, but can't.
		 * Find out if it is because of an
		 * ARP entry and delete it if so.
		 */
		rt2 = rtalloc1((struct sockaddr *)sin, 0,
				RTF_CLONING | RTF_PRCLONING);
		if (rt2) {
			if (rt2->rt_flags & RTF_LLINFO &&
			    rt2->rt_flags & RTF_HOST &&
			    rt2->rt_gateway &&
			    rt2->rt_gateway->sa_family == AF_LINK) {
				/* NB: must unlock to avoid recursion */
				RT_UNLOCK(rt2);
				rtrequest(RTM_DELETE,
					  (struct sockaddr *)rt_key(rt2),
					  rt2->rt_gateway, rt_mask(rt2),
					  rt2->rt_flags, 0);
				ret = rn_addroute(v_arg, n_arg, head,
						  treenodes);
				RT_LOCK(rt2);
			}
			RTFREE_LOCKED(rt2);
		}
	}

	/*
	 * If the new route created successfully, and we are forwarding,
	 * flush any cached routes to avoid using a stale value.
	 */
	if (ret != NULL && ipforwarding)
		ip_forward_cacheinval();

	return ret;
}

/*
 * This code is the inverse of in_clsroute: on first reference, if we
 * were managing the route, stop doing so and set the expiration timer
 * back off again.
 */
static struct radix_node *
in_matroute(void *v_arg, struct radix_node_head *head)
{
	struct radix_node *rn = rn_match(v_arg, head);
	struct rtentry *rt = (struct rtentry *)rn;

	/*XXX locking? */
	if (rt && rt->rt_refcnt == 0) {		/* this is first reference */
		if (rt->rt_flags & RTPRF_OURS) {
			rt->rt_flags &= ~RTPRF_OURS;
			rt->rt_rmx.rmx_expire = 0;
		}
	}
	return rn;
}

static int rtq_reallyold = 60*60;		/* one hour is "really old" */
SYSCTL_INT(_net_inet_ip, IPCTL_RTEXPIRE, rtexpire, CTLFLAG_RW,
    &rtq_reallyold, 0, "Default expiration time on dynamically learned routes");

static int rtq_minreallyold = 10;  /* never automatically crank down to less */
SYSCTL_INT(_net_inet_ip, IPCTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW,
    &rtq_minreallyold, 0,
    "Minimum time to attempt to hold onto dynamically learned routes");

static int rtq_toomany = 128;		/* 128 cached routes is "too many" */
SYSCTL_INT(_net_inet_ip, IPCTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW,
    &rtq_toomany, 0, "Upper limit on dynamically learned routes");

/*
 * On last reference drop, mark the route as belong to us so that it can be
 * timed out.
 */
static void
in_clsroute(struct radix_node *rn, struct radix_node_head *head)
{
	struct rtentry *rt = (struct rtentry *)rn;

	RT_LOCK_ASSERT(rt);

	if (!(rt->rt_flags & RTF_UP))
		return;			/* prophylactic measures */

	if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST)
		return;

	if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED)
		return;

	/*
	 * If rtq_reallyold is 0, just delete the route without
	 * waiting for a timeout cycle to kill it.
	 */
	if (rtq_reallyold != 0) {
		rt->rt_flags |= RTPRF_OURS;
		rt->rt_rmx.rmx_expire = time_second + rtq_reallyold;
	} else {
		/* NB: must unlock to avoid recursion */
		RT_UNLOCK(rt);
		rtrequest(RTM_DELETE,
			  (struct sockaddr *)rt_key(rt),
			  rt->rt_gateway, rt_mask(rt),
			  rt->rt_flags, 0);
		RT_LOCK(rt);
	}
}

struct rtqk_arg {
	struct radix_node_head *rnh;
	int draining;
	int killed;
	int found;
	int updating;
	time_t nextstop;
};

/*
 * Get rid of old routes.  When draining, this deletes everything, even when
 * the timeout is not expired yet.  When updating, this makes sure that
 * nothing has a timeout longer than the current value of rtq_reallyold.
 */
static int
in_rtqkill(struct radix_node *rn, void *rock)
{
	struct rtqk_arg *ap = rock;
	struct rtentry *rt = (struct rtentry *)rn;
	int err;

	if (rt->rt_flags & RTPRF_OURS) {
		ap->found++;

		if (ap->draining || rt->rt_rmx.rmx_expire <= time_second) {
			if (rt->rt_refcnt > 0)
				panic("rtqkill route really not free");

			err = rtrequest(RTM_DELETE,
					(struct sockaddr *)rt_key(rt),
					rt->rt_gateway, rt_mask(rt),
					rt->rt_flags, 0);
			if (err) {
				log(LOG_WARNING, "in_rtqkill: error %d\n", err);
			} else {
				ap->killed++;
			}
		} else {
			if (ap->updating &&
			    (rt->rt_rmx.rmx_expire - time_second >
			     rtq_reallyold)) {
				rt->rt_rmx.rmx_expire =
				    time_second + rtq_reallyold;
			}
			ap->nextstop = lmin(ap->nextstop,
					    rt->rt_rmx.rmx_expire);
		}
	}

	return 0;
}

#define RTQ_TIMEOUT	60*10	/* run no less than once every ten minutes */
static int rtq_timeout = RTQ_TIMEOUT;
static struct callout rtq_timer;

static void
in_rtqtimo(void *rock)
{
	struct radix_node_head *rnh = rock;
	struct rtqk_arg arg;
	struct timeval atv;
	static time_t last_adjusted_timeout = 0;

	arg.found = arg.killed = 0;
	arg.rnh = rnh;
	arg.nextstop = time_second + rtq_timeout;
	arg.draining = arg.updating = 0;
	RADIX_NODE_HEAD_LOCK(rnh);
	rnh->rnh_walktree(rnh, in_rtqkill, &arg);
	RADIX_NODE_HEAD_UNLOCK(rnh);

	/*
	 * Attempt to be somewhat dynamic about this:
	 * If there are ``too many'' routes sitting around taking up space,
	 * then crank down the timeout, and see if we can't make some more
	 * go away.  However, we make sure that we will never adjust more
	 * than once in rtq_timeout seconds, to keep from cranking down too
	 * hard.
	 */
	if ((arg.found - arg.killed > rtq_toomany) &&
	    (time_second - last_adjusted_timeout >= rtq_timeout) &&
	    rtq_reallyold > rtq_minreallyold) {
		rtq_reallyold = 2 * rtq_reallyold / 3;
		if (rtq_reallyold < rtq_minreallyold) {
			rtq_reallyold = rtq_minreallyold;
		}

		last_adjusted_timeout = time_second;
#ifdef DIAGNOSTIC
		log(LOG_DEBUG, "in_rtqtimo: adjusted rtq_reallyold to %d\n",
		    rtq_reallyold);
#endif
		arg.found = arg.killed = 0;
		arg.updating = 1;
		RADIX_NODE_HEAD_LOCK(rnh);
		rnh->rnh_walktree(rnh, in_rtqkill, &arg);
		RADIX_NODE_HEAD_UNLOCK(rnh);
	}

	atv.tv_usec = 0;
	atv.tv_sec = arg.nextstop - time_second;
	callout_reset(&rtq_timer, tvtohz(&atv), in_rtqtimo, rock);
}

void
in_rtqdrain(void)
{
	struct radix_node_head *rnh = rt_tables[AF_INET];
	struct rtqk_arg arg;

	arg.found = arg.killed = 0;
	arg.rnh = rnh;
	arg.nextstop = 0;
	arg.draining = 1;
	arg.updating = 0;
	RADIX_NODE_HEAD_LOCK(rnh);
	rnh->rnh_walktree(rnh, in_rtqkill, &arg);
	RADIX_NODE_HEAD_UNLOCK(rnh);
}

/*
 * Initialize our routing tree.
 */
int
in_inithead(void **head, int off)
{
	struct radix_node_head *rnh;

	if (!rn_inithead(head, off))
		return 0;

	if (head != (void **)&rt_tables[AF_INET])	/* BOGUS! */
		return 1;	/* only do this for the real routing table */

	rnh = *head;
	rnh->rnh_addaddr = in_addroute;
	rnh->rnh_matchaddr = in_matroute;
	rnh->rnh_close = in_clsroute;
	callout_init(&rtq_timer, CALLOUT_MPSAFE);
	in_rtqtimo(rnh);	/* kick off timeout first time */
	return 1;
}

/*
 * This zaps old routes when the interface goes down or interface
 * address is deleted.  In the latter case, it deletes static routes
 * that point to this address.  If we don't do this, we may end up
 * using the old address in the future.  The ones we always want to
 * get rid of are things like ARP entries, since the user might down
 * the interface, walk over to a completely different network, and
 * plug back in.
 */
struct in_ifadown_arg {
	struct radix_node_head *rnh;
	struct ifaddr *ifa;
	int del;
};

static int
in_ifadownkill(struct radix_node *rn, void *xap)
{
	struct in_ifadown_arg *ap = xap;
	struct rtentry *rt = (struct rtentry *)rn;
	int err;

	if (rt->rt_ifa == ap->ifa &&
	    (ap->del || !(rt->rt_flags & RTF_STATIC))) {
		/*
		 * We need to disable the automatic prune that happens
		 * in this case in rtrequest() because it will blow
		 * away the pointers that rn_walktree() needs in order
		 * continue our descent.  We will end up deleting all
		 * the routes that rtrequest() would have in any case,
		 * so that behavior is not needed there.
		 */
		RT_LOCK(rt);
		rt->rt_flags &= ~(RTF_CLONING | RTF_PRCLONING);
		RT_UNLOCK(rt);
		err = rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt),
				rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
		if (err) {
			log(LOG_WARNING, "in_ifadownkill: error %d\n", err);
		}
	}
	return 0;
}

int
in_ifadown(struct ifaddr *ifa, int delete)
{
	struct in_ifadown_arg arg;
	struct radix_node_head *rnh;

	if (ifa->ifa_addr->sa_family != AF_INET)
		return 1;

	arg.rnh = rnh = rt_tables[AF_INET];
	arg.ifa = ifa;
	arg.del = delete;
	RADIX_NODE_HEAD_LOCK(rnh);
	rnh->rnh_walktree(rnh, in_ifadownkill, &arg);
	RADIX_NODE_HEAD_UNLOCK(rnh);
	ifa->ifa_flags &= ~IFA_ROUTE;		/* XXXlocking? */
	return 0;
}
OpenPOWER on IntegriCloud