xref: /freebsd/sys/netgraph/netflow/netflow_v9.c (revision cab6a39d7b343596a5823e65c0f7b426551ec22d)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2010 Alexander V. Chernikov <melifaro@ipfw.ru>
5  * All rights reserved.
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  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * 	$FreeBSD$
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include "opt_inet6.h"
35 #include "opt_route.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/counter.h>
39 #include <sys/kernel.h>
40 #include <sys/ktr.h>
41 #include <sys/limits.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/syslog.h>
45 #include <sys/socket.h>
46 #include <vm/uma.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 #include <net/ethernet.h>
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/ip6.h>
55 #include <netinet/tcp.h>
56 #include <netinet/udp.h>
57 
58 #include <netgraph/ng_message.h>
59 #include <netgraph/netgraph.h>
60 
61 #include <netgraph/netflow/netflow.h>
62 #include <netgraph/netflow/ng_netflow.h>
63 #include <netgraph/netflow/netflow_v9.h>
64 
65 MALLOC_DECLARE(M_NETFLOW_GENERAL);
66 MALLOC_DEFINE(M_NETFLOW_GENERAL, "netflow_general", "plog, V9 templates data");
67 
68 /*
69  * Base V9 templates for L4+ IPv4/IPv6 protocols
70  */
71 struct netflow_v9_template _netflow_v9_record_ipv4_tcp[] =
72 {
73 	{ NETFLOW_V9_FIELD_IPV4_SRC_ADDR, 4},
74 	{ NETFLOW_V9_FIELD_IPV4_DST_ADDR, 4},
75 	{ NETFLOW_V9_FIELD_IPV4_NEXT_HOP, 4},
76 	{ NETFLOW_V9_FIELD_INPUT_SNMP, 2},
77 	{ NETFLOW_V9_FIELD_OUTPUT_SNMP, 2},
78 	{ NETFLOW_V9_FIELD_IN_PKTS, sizeof(CNTR)},
79 	{ NETFLOW_V9_FIELD_IN_BYTES, sizeof(CNTR)},
80 	{ NETFLOW_V9_FIELD_OUT_PKTS, sizeof(CNTR)},
81 	{ NETFLOW_V9_FIELD_OUT_BYTES, sizeof(CNTR)},
82 	{ NETFLOW_V9_FIELD_FIRST_SWITCHED, 4},
83 	{ NETFLOW_V9_FIELD_LAST_SWITCHED, 4},
84 	{ NETFLOW_V9_FIELD_L4_SRC_PORT, 2},
85 	{ NETFLOW_V9_FIELD_L4_DST_PORT, 2},
86 	{ NETFLOW_V9_FIELD_TCP_FLAGS, 1},
87 	{ NETFLOW_V9_FIELD_PROTOCOL, 1},
88 	{ NETFLOW_V9_FIELD_TOS, 1},
89 	{ NETFLOW_V9_FIELD_SRC_AS, 4},
90 	{ NETFLOW_V9_FIELD_DST_AS, 4},
91 	{ NETFLOW_V9_FIELD_SRC_MASK, 1},
92 	{ NETFLOW_V9_FIELD_DST_MASK, 1},
93 	{0, 0}
94 };
95 
96 struct netflow_v9_template _netflow_v9_record_ipv6_tcp[] =
97 {
98 	{ NETFLOW_V9_FIELD_IPV6_SRC_ADDR, 16},
99 	{ NETFLOW_V9_FIELD_IPV6_DST_ADDR, 16},
100 	{ NETFLOW_V9_FIELD_IPV6_NEXT_HOP, 16},
101 	{ NETFLOW_V9_FIELD_INPUT_SNMP, 2},
102 	{ NETFLOW_V9_FIELD_OUTPUT_SNMP, 2},
103 	{ NETFLOW_V9_FIELD_IN_PKTS, sizeof(CNTR)},
104 	{ NETFLOW_V9_FIELD_IN_BYTES, sizeof(CNTR)},
105 	{ NETFLOW_V9_FIELD_OUT_PKTS, sizeof(CNTR)},
106 	{ NETFLOW_V9_FIELD_OUT_BYTES, sizeof(CNTR)},
107 	{ NETFLOW_V9_FIELD_FIRST_SWITCHED, 4},
108 	{ NETFLOW_V9_FIELD_LAST_SWITCHED, 4},
109 	{ NETFLOW_V9_FIELD_L4_SRC_PORT, 2},
110 	{ NETFLOW_V9_FIELD_L4_DST_PORT, 2},
111 	{ NETFLOW_V9_FIELD_TCP_FLAGS, 1},
112 	{ NETFLOW_V9_FIELD_PROTOCOL, 1},
113 	{ NETFLOW_V9_FIELD_TOS, 1},
114 	{ NETFLOW_V9_FIELD_SRC_AS, 4},
115 	{ NETFLOW_V9_FIELD_DST_AS, 4},
116 	{ NETFLOW_V9_FIELD_SRC_MASK, 1},
117 	{ NETFLOW_V9_FIELD_DST_MASK, 1},
118 	{0, 0}
119 };
120 
121 /*
122  * Pre-compiles flow exporter for all possible FlowSets
123  * so we can add flowset to packet via simple memcpy()
124  */
125 static void
126 generate_v9_templates(priv_p priv)
127 {
128 	uint16_t *p, *template_fields_cnt;
129 	int cnt;
130 
131 	int flowset_size = sizeof(struct netflow_v9_flowset_header) +
132 		_NETFLOW_V9_TEMPLATE_SIZE(_netflow_v9_record_ipv4_tcp) + /* netflow_v9_record_ipv4_tcp */
133 		_NETFLOW_V9_TEMPLATE_SIZE(_netflow_v9_record_ipv6_tcp); /* netflow_v9_record_ipv6_tcp */
134 
135 	priv->v9_flowsets[0] = malloc(flowset_size, M_NETFLOW_GENERAL, M_WAITOK | M_ZERO);
136 
137 	if (flowset_size % 4)
138 		flowset_size += 4 - (flowset_size % 4); /* Padding to 4-byte boundary */
139 
140 	priv->flowsets_count = 1;
141 	p = (uint16_t *)priv->v9_flowsets[0];
142 	*p++ = 0; /* Flowset ID, 0 is reserved for Template FlowSets  */
143 	*p++ = htons(flowset_size); /* Total FlowSet length */
144 
145 	/*
146 	 * Most common TCP/UDP IPv4 template, ID = 256
147 	 */
148 	*p++ = htons(NETFLOW_V9_MAX_RESERVED_FLOWSET + NETFLOW_V9_FLOW_V4_L4);
149 	template_fields_cnt = p++;
150 	for (cnt = 0; _netflow_v9_record_ipv4_tcp[cnt].field_id != 0; cnt++) {
151 		*p++ = htons(_netflow_v9_record_ipv4_tcp[cnt].field_id);
152 		*p++ = htons(_netflow_v9_record_ipv4_tcp[cnt].field_length);
153 	}
154 	*template_fields_cnt = htons(cnt);
155 
156 	/*
157 	 * TCP/UDP IPv6 template, ID = 257
158 	 */
159 	*p++ = htons(NETFLOW_V9_MAX_RESERVED_FLOWSET + NETFLOW_V9_FLOW_V6_L4);
160 	template_fields_cnt = p++;
161 	for (cnt = 0; _netflow_v9_record_ipv6_tcp[cnt].field_id != 0; cnt++) {
162 		*p++ = htons(_netflow_v9_record_ipv6_tcp[cnt].field_id);
163 		*p++ = htons(_netflow_v9_record_ipv6_tcp[cnt].field_length);
164 	}
165 	*template_fields_cnt = htons(cnt);
166 
167 	priv->flowset_records[0] = 2;
168 }
169 
170 /* Closes current data flowset */
171 static void inline
172 close_flowset(struct mbuf *m, struct netflow_v9_packet_opt *t)
173 {
174 	struct mbuf *m_old;
175 	uint32_t zero = 0;
176 	int offset = 0;
177 	uint16_t *flowset_length, len;
178 
179 	/* Hack to ensure we are not crossing mbuf boundary, length is uint16_t  */
180 	m_old = m_getptr(m, t->flow_header + offsetof(struct netflow_v9_flowset_header, length), &offset);
181 	flowset_length = (uint16_t *)(mtod(m_old, char *) + offset);
182 
183 	len = (uint16_t)(m_pktlen(m) - t->flow_header);
184 	/* Align on 4-byte boundary (RFC 3954, Clause 5.3) */
185 	if (len % 4) {
186 		if (m_append(m, 4 - (len % 4), (void *)&zero) != 1)
187 			panic("ng_netflow: m_append() failed!");
188 
189 		len += 4 - (len % 4);
190 	}
191 
192 	*flowset_length = htons(len);
193 }
194 
195 /*
196  * Non-static functions called from ng_netflow.c
197  */
198 
199 /* We have full datagram in fib data. Send it to export hook. */
200 int
201 export9_send(priv_p priv, fib_export_p fe, item_p item, struct netflow_v9_packet_opt *t, int flags)
202 {
203 	struct mbuf *m = NGI_M(item);
204 	struct netflow_v9_export_dgram *dgram = mtod(m,
205 					struct netflow_v9_export_dgram *);
206 	struct netflow_v9_header *header = &dgram->header;
207 	struct timespec ts;
208 	int error = 0;
209 
210 	if (t == NULL) {
211 		CTR0(KTR_NET, "export9_send(): V9 export packet without tag");
212 		NG_FREE_ITEM(item);
213 		return (0);
214 	}
215 
216 	/* Close flowset if not closed already */
217 	if (m_pktlen(m) != t->flow_header)
218 		close_flowset(m, t);
219 
220 	/* Fill export header. */
221 	header->count = t->count;
222 	header->sys_uptime = htonl(MILLIUPTIME(time_uptime));
223 	getnanotime(&ts);
224 	header->unix_secs  = htonl(ts.tv_sec);
225 	header->seq_num = htonl(atomic_fetchadd_32(&fe->flow9_seq, 1));
226 	header->count = htons(t->count);
227 	header->source_id = htonl(fe->domain_id);
228 
229 	if (priv->export9 != NULL)
230 		NG_FWD_ITEM_HOOK_FLAGS(error, item, priv->export9, flags);
231 	else
232 		NG_FREE_ITEM(item);
233 
234 	free(t, M_NETFLOW_GENERAL);
235 
236 	return (error);
237 }
238 
239 /* Add V9 record to dgram. */
240 int
241 export9_add(item_p item, struct netflow_v9_packet_opt *t, struct flow_entry *fle)
242 {
243 	size_t len = 0;
244 	struct netflow_v9_flowset_header fsh;
245 	struct netflow_v9_record_general rg;
246 	struct mbuf *m = NGI_M(item);
247 	uint16_t flow_type;
248 	struct flow_entry_data *fed;
249 #ifdef INET6
250 	struct flow6_entry_data *fed6;
251 #endif
252 	if (t == NULL) {
253 		CTR0(KTR_NET, "ng_netflow: V9 export packet without tag!");
254 		return (0);
255 	}
256 
257 	/* Prepare flow record */
258 	fed = (struct flow_entry_data *)&fle->f;
259 #ifdef INET6
260 	fed6 = (struct flow6_entry_data *)&fle->f;
261 #endif
262 	/* We can use flow_type field since fle6 offset is equal to fle */
263 	flow_type = fed->r.flow_type;
264 
265 	switch (flow_type) {
266 	case NETFLOW_V9_FLOW_V4_L4:
267 	{
268 		/* IPv4 TCP/UDP/[SCTP] */
269 		struct netflow_v9_record_ipv4_tcp *rec = &rg.rec.v4_tcp;
270 
271 		rec->src_addr = fed->r.r_src.s_addr;
272 		rec->dst_addr = fed->r.r_dst.s_addr;
273 		rec->next_hop = fed->next_hop.s_addr;
274 		rec->i_ifx    = htons(fed->fle_i_ifx);
275 		rec->o_ifx    = htons(fed->fle_o_ifx);
276 		rec->i_packets  = htonl(fed->packets);
277 		rec->i_octets   = htonl(fed->bytes);
278 		rec->o_packets  = htonl(0);
279 		rec->o_octets   = htonl(0);
280 		rec->first    = htonl(MILLIUPTIME(fed->first));
281 		rec->last     = htonl(MILLIUPTIME(fed->last));
282 		rec->s_port   = fed->r.r_sport;
283 		rec->d_port   = fed->r.r_dport;
284 		rec->flags    = fed->tcp_flags;
285 		rec->prot     = fed->r.r_ip_p;
286 		rec->tos      = fed->r.r_tos;
287 		rec->dst_mask = fed->dst_mask;
288 		rec->src_mask = fed->src_mask;
289 
290 		/* Not supported fields. */
291 		rec->src_as = rec->dst_as = 0;
292 
293 		len = sizeof(struct netflow_v9_record_ipv4_tcp);
294 		break;
295 	}
296 #ifdef INET6
297 	case NETFLOW_V9_FLOW_V6_L4:
298 	{
299 		/* IPv6 TCP/UDP/[SCTP] */
300 		struct netflow_v9_record_ipv6_tcp *rec = &rg.rec.v6_tcp;
301 
302 		rec->src_addr = fed6->r.src.r_src6;
303 		rec->dst_addr = fed6->r.dst.r_dst6;
304 		rec->next_hop = fed6->n.next_hop6;
305 		rec->i_ifx    = htons(fed6->fle_i_ifx);
306 		rec->o_ifx    = htons(fed6->fle_o_ifx);
307 		rec->i_packets  = htonl(fed6->packets);
308 		rec->i_octets   = htonl(fed6->bytes);
309 		rec->o_packets  = htonl(0);
310 		rec->o_octets   = htonl(0);
311 		rec->first    = htonl(MILLIUPTIME(fed6->first));
312 		rec->last     = htonl(MILLIUPTIME(fed6->last));
313 		rec->s_port   = fed6->r.r_sport;
314 		rec->d_port   = fed6->r.r_dport;
315 		rec->flags    = fed6->tcp_flags;
316 		rec->prot     = fed6->r.r_ip_p;
317 		rec->tos      = fed6->r.r_tos;
318 		rec->dst_mask = fed6->dst_mask;
319 		rec->src_mask = fed6->src_mask;
320 
321 		/* Not supported fields. */
322 		rec->src_as = rec->dst_as = 0;
323 
324 		len = sizeof(struct netflow_v9_record_ipv6_tcp);
325 		break;
326 	}
327 #endif
328 	default:
329 	{
330 		CTR1(KTR_NET, "export9_add(): Don't know what to do with %d flow type!", flow_type);
331 		return (0);
332 	}
333 	}
334 
335 	/* Check if new records has the same template */
336 	if (flow_type != t->flow_type) {
337 		/* close old flowset */
338 		if (t->flow_type != 0)
339 			close_flowset(m, t);
340 
341 		t->flow_type = flow_type;
342 		t->flow_header = m_pktlen(m);
343 
344 		/* Generate data flowset ID */
345 		fsh.id = htons(NETFLOW_V9_MAX_RESERVED_FLOWSET + flow_type);
346 		fsh.length = 0;
347 
348 		/* m_append should not fail since all data is already allocated */
349 		if (m_append(m, sizeof(fsh), (void *)&fsh) != 1)
350 			panic("ng_netflow: m_append() failed");
351 
352 	}
353 
354 	if (m_append(m, len, (void *)&rg.rec) != 1)
355 		panic("ng_netflow: m_append() failed");
356 
357 	t->count++;
358 
359 	if (m_pktlen(m) + sizeof(struct netflow_v9_record_general) + sizeof(struct netflow_v9_flowset_header) >= _NETFLOW_V9_MAX_SIZE(t->mtu))
360 		return (1); /* end of datagram */
361 	return (0);
362 }
363 
364 /*
365  * Detach export datagram from fib instance, if there is any.
366  * If there is no, allocate a new one.
367  */
368 item_p
369 get_export9_dgram(priv_p priv, fib_export_p fe, struct netflow_v9_packet_opt **tt)
370 {
371 	item_p	item = NULL;
372 	struct netflow_v9_packet_opt *t = NULL;
373 
374 	mtx_lock(&fe->export9_mtx);
375 	if (fe->exp.item9 != NULL) {
376 		item = fe->exp.item9;
377 		fe->exp.item9 = NULL;
378 		t = fe->exp.item9_opt;
379 		fe->exp.item9_opt = NULL;
380 	}
381 	mtx_unlock(&fe->export9_mtx);
382 
383 	if (item == NULL) {
384 		struct netflow_v9_export_dgram *dgram;
385 		struct mbuf *m;
386 		uint16_t mtu = priv->mtu;
387 
388 		/* Allocate entire packet at once, allowing easy m_append() calls */
389 		m = m_getm(NULL, mtu, M_NOWAIT, MT_DATA);
390 		if (m == NULL)
391 			return (NULL);
392 
393 		t = malloc(sizeof(struct netflow_v9_packet_opt), M_NETFLOW_GENERAL, M_NOWAIT | M_ZERO);
394 		if (t == NULL) {
395 			m_free(m);
396 			return (NULL);
397 		}
398 
399 		item = ng_package_data(m, NG_NOFLAGS);
400 		if (item == NULL) {
401 			free(t, M_NETFLOW_GENERAL);
402 			return (NULL);
403 		}
404 
405 		dgram = mtod(m, struct netflow_v9_export_dgram *);
406 		dgram->header.count = 0;
407 		dgram->header.version = htons(NETFLOW_V9);
408 		/* Set mbuf current data length */
409 		m->m_len = m->m_pkthdr.len = sizeof(struct netflow_v9_header);
410 
411 		t->count = 0;
412 		t->mtu = mtu;
413 		t->flow_header = m->m_len;
414 
415 		/*
416 		 * Check if we need to insert templates into packet
417 		 */
418 
419 		struct netflow_v9_flowset_header	*fl;
420 
421 		if ((time_uptime >= priv->templ_time + fe->templ_last_ts) ||
422 				(fe->sent_packets >= priv->templ_packets + fe->templ_last_pkt)) {
423 			fe->templ_last_ts = time_uptime;
424 			fe->templ_last_pkt = fe->sent_packets;
425 
426 			fl = priv->v9_flowsets[0];
427 			m_append(m, ntohs(fl->length), (void *)fl);
428 			t->flow_header = m->m_len;
429 			t->count += priv->flowset_records[0];
430 		}
431 	}
432 
433 	*tt = t;
434 	return (item);
435 }
436 
437 /*
438  * Re-attach incomplete datagram back to fib instance.
439  * If there is already another one, then send incomplete.
440  */
441 void
442 return_export9_dgram(priv_p priv, fib_export_p fe, item_p item, struct netflow_v9_packet_opt *t, int flags)
443 {
444 	/*
445 	 * It may happen on SMP, that some thread has already
446 	 * put its item there, in this case we bail out and
447 	 * send what we have to collector.
448 	 */
449 	mtx_lock(&fe->export9_mtx);
450 	if (fe->exp.item9 == NULL) {
451 		fe->exp.item9 = item;
452 		fe->exp.item9_opt = t;
453 		mtx_unlock(&fe->export9_mtx);
454 	} else {
455 		mtx_unlock(&fe->export9_mtx);
456 		export9_send(priv, fe, item, t, flags);
457 	}
458 }
459 
460 /* Allocate memory and set up flow cache */
461 void
462 ng_netflow_v9_cache_init(priv_p priv)
463 {
464 	generate_v9_templates(priv);
465 
466 	priv->templ_time = NETFLOW_V9_MAX_TIME_TEMPL;
467 	priv->templ_packets = NETFLOW_V9_MAX_PACKETS_TEMPL;
468 	priv->mtu = BASE_MTU;
469 }
470 
471 /* Free all flow cache memory. Called from ng_netflow_cache_flush() */
472 void
473 ng_netflow_v9_cache_flush(priv_p priv)
474 {
475 	int i;
476 
477 	/* Free flowsets*/
478 	for (i = 0; i < priv->flowsets_count; i++)
479 		free(priv->v9_flowsets[i], M_NETFLOW_GENERAL);
480 }
481 
482 /* Get a snapshot of NetFlow v9 settings */
483 void
484 ng_netflow_copyv9info(priv_p priv, struct ng_netflow_v9info *i)
485 {
486 
487 	i->templ_time = priv->templ_time;
488 	i->templ_packets = priv->templ_packets;
489 	i->mtu = priv->mtu;
490 }
491