1 /*
2  * Copyright (c) 2007 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Sepherosa Ziehau <sepherosa@gmail.com>
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 
36 #include <sys/param.h>
37 #include <sys/kernel.h>
38 #include <sys/mbuf.h>
39 #include <sys/msgport.h>
40 #include <sys/socketvar.h>
41 #include <sys/sysctl.h>
42 
43 #include <net/if.h>
44 #include <net/if_var.h>
45 #include <net/route.h>
46 #include <net/ethernet.h>
47 #include <net/netisr2.h>
48 #include <net/netmsg2.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_var.h>
52 #include <netinet/ip.h>
53 #include <netinet/ip_var.h>
54 
55 #include <net/ipfw3/ip_fw.h>
56 #include <net/dummynet3/ip_dummynet.h>
57 
58 static void	ip_dn_ether_output(netmsg_t);
59 static void	ip_dn_ether_demux(netmsg_t);
60 static void	ip_dn_ip_input(netmsg_t);
61 static void	ip_dn_ip_output(netmsg_t);
62 
63 static void	ip_dn_sockopt_dispatch(netmsg_t);
64 static void	ip_dn_freepkt_dispatch(netmsg_t);
65 static void	ip_dn_dispatch(netmsg_t);
66 
67 static void	ip_dn_freepkt(struct dn_pkt *);
68 
69 static int	ip_dn_sockopt_flush(struct sockopt *);
70 static int	ip_dn_sockopt_get(struct sockopt *);
71 static int	ip_dn_sockopt_config(struct sockopt *);
72 
73 ip_dn_io_t	*ip_dn_io_ptr;
74 int		ip_dn_cpu = 0;
75 
76 TUNABLE_INT("net.inet.ip.dummynet.cpu", &ip_dn_cpu);
77 
78 SYSCTL_NODE(_net_inet_ip, OID_AUTO, dummynet, CTLFLAG_RW, 0, "Dummynet");
79 SYSCTL_INT(_net_inet_ip_dummynet, OID_AUTO, cpu, CTLFLAG_RD,
80 	   &ip_dn_cpu, 0, "CPU to run dummynet");
81 
82 void
83 ip_dn_queue(struct mbuf *m)
84 {
85 	struct netmsg_packet *nmp;
86 	lwkt_port_t port;
87 
88 	M_ASSERTPKTHDR(m);
89 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
90 		("mbuf is not tagged for dummynet!"));
91 
92 	nmp = &m->m_hdr.mh_netmsg;
93 	netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
94 			0, ip_dn_dispatch);
95 	nmp->nm_packet = m;
96 
97 	port = netisr_cpuport(ip_dn_cpu);
98 	lwkt_sendmsg(port, &nmp->base.lmsg);
99 }
100 
101 void
102 ip_dn_packet_free(struct dn_pkt *pkt)
103 {
104 	struct netmsg_packet *nmp;
105 	struct mbuf *m = pkt->dn_m;
106 
107 	M_ASSERTPKTHDR(m);
108 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
109 		("mbuf is not tagged for dummynet!"));
110 
111 	nmp = &m->m_hdr.mh_netmsg;
112 	netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
113 			0, ip_dn_freepkt_dispatch);
114 	nmp->nm_packet = m;
115 
116 	lwkt_sendmsg(pkt->msgport, &nmp->base.lmsg);
117 }
118 
119 void
120 ip_dn_packet_redispatch(struct dn_pkt *pkt)
121 {
122 	static const netisr_fn_t dispatches[DN_TO_MAX] = {
123 	[DN_TO_IP_OUT]		= ip_dn_ip_output,
124 	[DN_TO_IP_IN]		= ip_dn_ip_input,
125 	[DN_TO_ETH_DEMUX]	= ip_dn_ether_demux,
126 	[DN_TO_ETH_OUT]		= ip_dn_ether_output
127 	};
128 
129 	struct netmsg_packet *nmp;
130 	struct mbuf *m;
131 	netisr_fn_t dispatch;
132 	int dir;
133 
134 	dir = (pkt->dn_flags & DN_FLAGS_DIR_MASK);
135 	KASSERT(dir < DN_TO_MAX,
136 		("unknown dummynet redispatch dir %d", dir));
137 
138 	dispatch = dispatches[dir];
139 	KASSERT(dispatch != NULL,
140 		("unsupported dummynet redispatch dir %d", dir));
141 
142 	m = pkt->dn_m;
143 	M_ASSERTPKTHDR(m);
144 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
145 		("mbuf is not tagged for dummynet!"));
146 
147 	nmp = &m->m_hdr.mh_netmsg;
148 	netmsg_init(&nmp->base, NULL, &netisr_apanic_rport, 0, dispatch);
149 	nmp->nm_packet = m;
150 
151 	lwkt_sendmsg(pkt->msgport, &nmp->base.lmsg);
152 }
153 
154 int
155 ip_dn_sockopt(struct sockopt *sopt)
156 {
157 	int error = 0;
158 
159 	/* Disallow sets in really-really secure mode. */
160 	if (sopt->sopt_dir == SOPT_SET) {
161 		if (securelevel >= 3)
162 			return EPERM;
163 	}
164 
165 	switch (sopt->sopt_name) {
166 	case IP_DUMMYNET_GET:
167 		error = ip_dn_sockopt_get(sopt);
168 		break;
169 
170 	case IP_DUMMYNET_FLUSH:
171 		error = ip_dn_sockopt_flush(sopt);
172 		break;
173 
174 	case IP_DUMMYNET_DEL:
175 	case IP_DUMMYNET_CONFIGURE:
176 		error = ip_dn_sockopt_config(sopt);
177 		break;
178 
179 	default:
180 		kprintf("%s -- unknown option %d\n", __func__, sopt->sopt_name);
181 		error = EINVAL;
182 		break;
183 	}
184 	return error;
185 }
186 
187 static void
188 ip_dn_freepkt(struct dn_pkt *pkt)
189 {
190 	struct rtentry *rt = pkt->ro.ro_rt;
191 
192 	/* Unreference route entry */
193 	if (rt != NULL) {
194 		if (rt->rt_refcnt <= 0) {	/* XXX assert? */
195 			kprintf("-- warning, refcnt now %ld, decreasing\n",
196 				rt->rt_refcnt);
197 		}
198 		RTFREE(rt);
199 	}
200 
201 	/* Unreference packet private data */
202 	if (pkt->dn_unref_priv)
203 		pkt->dn_unref_priv(pkt->dn_priv);
204 
205 	/* Free the parent mbuf, this will free 'pkt' as well */
206 	m_freem(pkt->dn_m);
207 }
208 
209 static void
210 ip_dn_freepkt_dispatch(netmsg_t nmsg)
211 {
212 	struct netmsg_packet *nmp;
213 	struct mbuf *m;
214 	struct m_tag *mtag;
215 	struct dn_pkt *pkt;
216 
217 	nmp = &nmsg->packet;
218 	m = nmp->nm_packet;
219 	M_ASSERTPKTHDR(m);
220 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
221 		("mbuf is not tagged for dummynet!"));
222 
223 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
224 	KKASSERT(mtag != NULL);
225 
226 	pkt = m_tag_data(mtag);
227 	KASSERT(pkt->cpuid == mycpuid,
228 		("%s: dummynet packet was delivered to wrong cpu! "
229 		 "target cpuid %d, mycpuid %d", __func__,
230 		 pkt->cpuid, mycpuid));
231 
232 	ip_dn_freepkt(pkt);
233 }
234 
235 static void
236 ip_dn_dispatch(netmsg_t nmsg)
237 {
238 	struct netmsg_packet *nmp;
239 	struct mbuf *m;
240 	struct m_tag *mtag;
241 	struct dn_pkt *pkt;
242 
243 	KASSERT(ip_dn_cpu == mycpuid,
244 		("%s: dummynet packet was delivered to wrong cpu! "
245 		 "dummynet cpuid %d, mycpuid %d", __func__,
246 		 ip_dn_cpu, mycpuid));
247 
248 	nmp = &nmsg->packet;
249 	m = nmp->nm_packet;
250 	M_ASSERTPKTHDR(m);
251 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
252 		("mbuf is not tagged for dummynet!"));
253 
254 	if (DUMMYNET_LOADED) {
255 		if (ip_dn_io_ptr(m) == 0)
256 			return;
257 	}
258 
259 	/*
260 	 * ip_dn_io_ptr() failed or dummynet(4) is not loaded
261 	 */
262 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
263 	KKASSERT(mtag != NULL);
264 
265 	pkt = m_tag_data(mtag);
266 	ip_dn_packet_free(pkt);
267 }
268 
269 static void
270 ip_dn_ip_output(netmsg_t nmsg)
271 {
272 	struct netmsg_packet *nmp;
273 	struct mbuf *m;
274 	struct m_tag *mtag;
275 	struct dn_pkt *pkt;
276 	struct rtentry *rt;
277 	ip_dn_unref_priv_t unref_priv;
278 	void *priv;
279 
280 	nmp = &nmsg->packet;
281 	m = nmp->nm_packet;
282 	M_ASSERTPKTHDR(m);
283 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
284 		("mbuf is not tagged for dummynet!"));
285 
286 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
287 	KKASSERT(mtag != NULL);
288 
289 	pkt = m_tag_data(mtag);
290 	KASSERT(pkt->cpuid == mycpuid,
291 		("%s: dummynet packet was delivered to wrong cpu! "
292 		 "target cpuid %d, mycpuid %d", __func__,
293 		 pkt->cpuid, mycpuid));
294 	KASSERT((pkt->dn_flags & DN_FLAGS_DIR_MASK) == DN_TO_IP_OUT,
295 		("wrong direction %d, should be %d",
296 		 (pkt->dn_flags & DN_FLAGS_DIR_MASK), DN_TO_IP_OUT));
297 
298 	priv = pkt->dn_priv;
299 	unref_priv = pkt->dn_unref_priv;
300 	rt = pkt->ro.ro_rt;
301 
302 	if (rt != NULL && !(rt->rt_flags & RTF_UP)) {
303 		/*
304 		 * Recorded rtentry is gone, when the packet
305 		 * was on delay line.
306 		 */
307 		ip_dn_freepkt(pkt);
308 		return;
309 	}
310 
311 	ip_output(pkt->dn_m, NULL, NULL, pkt->flags, NULL, NULL);
312 	/* 'rt' will be freed in ip_output */
313 
314 	if (unref_priv)
315 		unref_priv(priv);
316 }
317 
318 static void
319 ip_dn_ip_input(netmsg_t nmsg)
320 {
321 	struct netmsg_packet *nmp;
322 	struct mbuf *m;
323 	struct m_tag *mtag;
324 	struct dn_pkt *pkt;
325 	ip_dn_unref_priv_t unref_priv;
326 	void *priv;
327 	struct ip *ip;
328 
329 	nmp = &nmsg->packet;
330 	m = nmp->nm_packet;
331 	M_ASSERTPKTHDR(m);
332 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
333 		("mbuf is not tagged for dummynet!"));
334 
335 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
336 	KKASSERT(mtag != NULL);
337 
338 	pkt = m_tag_data(mtag);
339 	KASSERT(pkt->cpuid == mycpuid,
340 		("%s: dummynet packet was delivered to wrong cpu! "
341 		 "target cpuid %d, mycpuid %d", __func__,
342 		 pkt->cpuid, mycpuid));
343 	KASSERT(pkt->ro.ro_rt == NULL,
344 		("route entry is not NULL for ip_input"));
345 	KASSERT((pkt->dn_flags & DN_FLAGS_DIR_MASK) == DN_TO_IP_IN,
346 		("wrong direction %d, should be %d",
347 		 (pkt->dn_flags & DN_FLAGS_DIR_MASK), DN_TO_IP_IN));
348 
349 	priv = pkt->dn_priv;
350 	unref_priv = pkt->dn_unref_priv;
351 
352 	KKASSERT(m->m_len >= sizeof(*ip));
353 
354 	ip_input(m);
355 
356 	if (unref_priv)
357 		unref_priv(priv);
358 }
359 
360 static void
361 ip_dn_ether_demux(netmsg_t nmsg)
362 {
363 	struct netmsg_packet *nmp;
364 	struct mbuf *m;
365 	struct m_tag *mtag;
366 	struct dn_pkt *pkt;
367 	ip_dn_unref_priv_t unref_priv;
368 	void *priv;
369 
370 	nmp = &nmsg->packet;
371 	m = nmp->nm_packet;
372 	M_ASSERTPKTHDR(m);
373 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
374 		("mbuf is not tagged for dummynet!"));
375 
376 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
377 	KKASSERT(mtag != NULL);
378 
379 	pkt = m_tag_data(mtag);
380 	KASSERT(pkt->cpuid == mycpuid,
381 		("%s: dummynet packet was delivered to wrong cpu! "
382 		 "target cpuid %d, mycpuid %d", __func__,
383 		 pkt->cpuid, mycpuid));
384 	KASSERT(pkt->ro.ro_rt == NULL,
385 		("route entry is not NULL for ether_demux"));
386 	KASSERT((pkt->dn_flags & DN_FLAGS_DIR_MASK) == DN_TO_ETH_DEMUX,
387 		("wrong direction %d, should be %d",
388 		 (pkt->dn_flags & DN_FLAGS_DIR_MASK), DN_TO_ETH_DEMUX));
389 
390 	priv = pkt->dn_priv;
391 	unref_priv = pkt->dn_unref_priv;
392 
393 	/*
394 	 * Make sure that ether header is contiguous
395 	 */
396 	if (m->m_len < ETHER_HDR_LEN &&
397 		(m = m_pullup(m, ETHER_HDR_LEN)) == NULL) {
398 		kprintf("%s: pullup fail, dropping pkt\n", __func__);
399 		goto back;
400 	}
401 	ether_demux_oncpu(m->m_pkthdr.rcvif, m);
402 back:
403 	if (unref_priv)
404 		unref_priv(priv);
405 }
406 
407 static void
408 ip_dn_ether_output(netmsg_t nmsg)
409 {
410 	struct netmsg_packet *nmp;
411 	struct mbuf *m;
412 	struct m_tag *mtag;
413 	struct dn_pkt *pkt;
414 	ip_dn_unref_priv_t unref_priv;
415 	void *priv;
416 
417 	nmp = &nmsg->packet;
418 	m = nmp->nm_packet;
419 	M_ASSERTPKTHDR(m);
420 	KASSERT(m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED,
421 		("mbuf is not tagged for dummynet!"));
422 
423 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
424 	KKASSERT(mtag != NULL);
425 
426 	pkt = m_tag_data(mtag);
427 	KASSERT(pkt->cpuid == mycpuid,
428 		("%s: dummynet packet was delivered to wrong cpu! "
429 		 "target cpuid %d, mycpuid %d", __func__,
430 		 pkt->cpuid, mycpuid));
431 	KASSERT(pkt->ro.ro_rt == NULL,
432 		("route entry is not NULL for ether_output_frame"));
433 	KASSERT((pkt->dn_flags & DN_FLAGS_DIR_MASK) == DN_TO_ETH_OUT,
434 		("wrong direction %d, should be %d",
435 		 (pkt->dn_flags & DN_FLAGS_DIR_MASK), DN_TO_ETH_OUT));
436 
437 	priv = pkt->dn_priv;
438 	unref_priv = pkt->dn_unref_priv;
439 
440 	ether_output_frame(pkt->ifp, m);
441 
442 	if (unref_priv)
443 		unref_priv(priv);
444 }
445 
446 static void
447 ip_dn_sockopt_dispatch(netmsg_t nmsg)
448 {
449 	lwkt_msg *msg = &nmsg->lmsg;
450 	struct dn_sopt *dn_sopt = msg->u.ms_resultp;
451 	int error;
452 
453 	KASSERT(ip_dn_cpu == mycpuid,
454 		("%s: dummynet sockopt is done on wrong cpu! "
455 		 "dummynet cpuid %d, mycpuid %d", __func__,
456 		 ip_dn_cpu, mycpuid));
457 
458 	if (DUMMYNET_LOADED)
459 		error = ip_dn_ctl_ptr(dn_sopt);
460 	else
461 		error = ENOPROTOOPT;
462 	lwkt_replymsg(msg, error);
463 }
464 
465 static int
466 ip_dn_sockopt_flush(struct sockopt *sopt)
467 {
468 	struct dn_sopt dn_sopt;
469 	struct netmsg_base smsg;
470 
471 	bzero(&dn_sopt, sizeof(dn_sopt));
472 	dn_sopt.dn_sopt_name = sopt->sopt_name;
473 
474 	netmsg_init(&smsg, NULL, &curthread->td_msgport,
475 			0, ip_dn_sockopt_dispatch);
476 	smsg.lmsg.u.ms_resultp = &dn_sopt;
477 	lwkt_domsg(netisr_cpuport(ip_dn_cpu), &smsg.lmsg, 0);
478 
479 	return smsg.lmsg.ms_error;
480 }
481 
482 static int
483 ip_dn_sockopt_get(struct sockopt *sopt)
484 {
485 	struct dn_sopt dn_sopt;
486 	struct netmsg_base smsg;
487 	int error;
488 
489 	bzero(&dn_sopt, sizeof(dn_sopt));
490 	dn_sopt.dn_sopt_name = sopt->sopt_name;
491 
492 	netmsg_init(&smsg, NULL, &curthread->td_msgport,
493 			0, ip_dn_sockopt_dispatch);
494 	smsg.lmsg.u.ms_resultp = &dn_sopt;
495 	lwkt_domsg(netisr_cpuport(ip_dn_cpu), &smsg.lmsg, 0);
496 
497 	error = smsg.lmsg.ms_error;
498 	if (error) {
499 		KKASSERT(dn_sopt.dn_sopt_arg == NULL);
500 		KKASSERT(dn_sopt.dn_sopt_arglen == 0);
501 		return error;
502 	}
503 
504 	soopt_from_kbuf(sopt, dn_sopt.dn_sopt_arg, dn_sopt.dn_sopt_arglen);
505 	kfree(dn_sopt.dn_sopt_arg, M_TEMP);
506 	return 0;
507 }
508 
509 static int
510 ip_dn_sockopt_config(struct sockopt *sopt)
511 {
512 	struct dn_ioc_pipe tmp_ioc_pipe;
513 	struct dn_sopt dn_sopt;
514 	struct netmsg_base smsg;
515 	int error;
516 
517 	error = soopt_to_kbuf(sopt, &tmp_ioc_pipe, sizeof tmp_ioc_pipe,
518 				  sizeof tmp_ioc_pipe);
519 	if (error)
520 		return error;
521 
522 	bzero(&dn_sopt, sizeof(dn_sopt));
523 	dn_sopt.dn_sopt_name = sopt->sopt_name;
524 	dn_sopt.dn_sopt_arg = &tmp_ioc_pipe;
525 	dn_sopt.dn_sopt_arglen = sizeof(tmp_ioc_pipe);
526 
527 	netmsg_init(&smsg, NULL, &curthread->td_msgport,
528 			0, ip_dn_sockopt_dispatch);
529 	smsg.lmsg.u.ms_resultp = &dn_sopt;
530 	lwkt_domsg(netisr_cpuport(ip_dn_cpu), &smsg.lmsg, 0);
531 
532 	return smsg.lmsg.ms_error;
533 }
534