xref: /linux/net/netfilter/ipvs/ip_vs_proto_sctp.c (revision 8bf22c33e7a172fbc72464f4cc484d23a6b412ba)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/kernel.h>
3 #include <linux/ip.h>
4 #include <linux/sctp.h>
5 #include <net/ip.h>
6 #include <net/ip6_checksum.h>
7 #include <linux/netfilter.h>
8 #include <linux/netfilter_ipv4.h>
9 #include <net/sctp/checksum.h>
10 #include <net/ip_vs.h>
11 
12 static int
13 sctp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp,
14 		unsigned int sctphoff);
15 
16 static int
sctp_conn_schedule(struct netns_ipvs * ipvs,int af,struct sk_buff * skb,struct ip_vs_proto_data * pd,int * verdict,struct ip_vs_conn ** cpp,struct ip_vs_iphdr * iph)17 sctp_conn_schedule(struct netns_ipvs *ipvs, int af, struct sk_buff *skb,
18 		   struct ip_vs_proto_data *pd,
19 		   int *verdict, struct ip_vs_conn **cpp,
20 		   struct ip_vs_iphdr *iph)
21 {
22 	struct ip_vs_service *svc;
23 	struct sctp_chunkhdr _schunkh, *sch;
24 	struct sctphdr *sh, _sctph;
25 	__be16 _ports[2], *ports = NULL;
26 
27 	if (likely(!ip_vs_iph_icmp(iph))) {
28 		sh = skb_header_pointer(skb, iph->len, sizeof(_sctph), &_sctph);
29 		if (sh) {
30 			sch = skb_header_pointer(skb, iph->len + sizeof(_sctph),
31 						 sizeof(_schunkh), &_schunkh);
32 			if (sch) {
33 				if (sch->type == SCTP_CID_ABORT ||
34 				    !(sysctl_sloppy_sctp(ipvs) ||
35 				      sch->type == SCTP_CID_INIT))
36 					return 1;
37 				ports = &sh->source;
38 			}
39 		}
40 	} else {
41 		ports = skb_header_pointer(
42 			skb, iph->len, sizeof(_ports), &_ports);
43 	}
44 
45 	if (!ports) {
46 		*verdict = NF_DROP;
47 		return 0;
48 	}
49 
50 	if (likely(!ip_vs_iph_inverse(iph)))
51 		svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
52 					 &iph->daddr, ports[1]);
53 	else
54 		svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
55 					 &iph->saddr, ports[0]);
56 	if (svc) {
57 		int ignored;
58 
59 		if (ip_vs_todrop(ipvs)) {
60 			/*
61 			 * It seems that we are very loaded.
62 			 * We have to drop this packet :(
63 			 */
64 			*verdict = NF_DROP;
65 			return 0;
66 		}
67 		/*
68 		 * Let the virtual server select a real server for the
69 		 * incoming connection, and create a connection entry.
70 		 */
71 		*cpp = ip_vs_schedule(svc, skb, pd, &ignored, iph);
72 		if (!*cpp && ignored <= 0) {
73 			if (!ignored)
74 				*verdict = ip_vs_leave(svc, skb, pd, iph);
75 			else
76 				*verdict = NF_DROP;
77 			return 0;
78 		}
79 	}
80 	/* NF_ACCEPT */
81 	return 1;
82 }
83 
sctp_nat_csum(struct sk_buff * skb,struct sctphdr * sctph,unsigned int sctphoff)84 static void sctp_nat_csum(struct sk_buff *skb, struct sctphdr *sctph,
85 			  unsigned int sctphoff)
86 {
87 	sctph->checksum = sctp_compute_cksum(skb, sctphoff);
88 	skb->ip_summed = CHECKSUM_UNNECESSARY;
89 }
90 
91 static int
sctp_snat_handler(struct sk_buff * skb,struct ip_vs_protocol * pp,struct ip_vs_conn * cp,struct ip_vs_iphdr * iph)92 sctp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
93 		  struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
94 {
95 	struct sctphdr *sctph;
96 	unsigned int sctphoff = iph->len;
97 	bool payload_csum = false;
98 
99 #ifdef CONFIG_IP_VS_IPV6
100 	if (cp->af == AF_INET6 && iph->fragoffs)
101 		return 1;
102 #endif
103 
104 	/* csum_check requires unshared skb */
105 	if (skb_ensure_writable(skb, sctphoff + sizeof(*sctph)))
106 		return 0;
107 
108 	if (unlikely(cp->app != NULL)) {
109 		int ret;
110 
111 		/* Some checks before mangling */
112 		if (!sctp_csum_check(cp->af, skb, pp, sctphoff))
113 			return 0;
114 
115 		/* Call application helper if needed */
116 		ret = ip_vs_app_pkt_out(cp, skb, iph);
117 		if (ret == 0)
118 			return 0;
119 		/* ret=2: csum update is needed after payload mangling */
120 		if (ret == 2)
121 			payload_csum = true;
122 	}
123 
124 	sctph = (void *) skb_network_header(skb) + sctphoff;
125 
126 	/* Only update csum if we really have to */
127 	if (sctph->source != cp->vport || payload_csum ||
128 	    skb->ip_summed == CHECKSUM_PARTIAL) {
129 		sctph->source = cp->vport;
130 		if (!skb_is_gso(skb))
131 			sctp_nat_csum(skb, sctph, sctphoff);
132 	} else {
133 		skb->ip_summed = CHECKSUM_UNNECESSARY;
134 	}
135 
136 	return 1;
137 }
138 
139 static int
sctp_dnat_handler(struct sk_buff * skb,struct ip_vs_protocol * pp,struct ip_vs_conn * cp,struct ip_vs_iphdr * iph)140 sctp_dnat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
141 		  struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
142 {
143 	struct sctphdr *sctph;
144 	unsigned int sctphoff = iph->len;
145 	bool payload_csum = false;
146 
147 #ifdef CONFIG_IP_VS_IPV6
148 	if (cp->af == AF_INET6 && iph->fragoffs)
149 		return 1;
150 #endif
151 
152 	/* csum_check requires unshared skb */
153 	if (skb_ensure_writable(skb, sctphoff + sizeof(*sctph)))
154 		return 0;
155 
156 	if (unlikely(cp->app != NULL)) {
157 		int ret;
158 
159 		/* Some checks before mangling */
160 		if (!sctp_csum_check(cp->af, skb, pp, sctphoff))
161 			return 0;
162 
163 		/* Call application helper if needed */
164 		ret = ip_vs_app_pkt_in(cp, skb, iph);
165 		if (ret == 0)
166 			return 0;
167 		/* ret=2: csum update is needed after payload mangling */
168 		if (ret == 2)
169 			payload_csum = true;
170 	}
171 
172 	sctph = (void *) skb_network_header(skb) + sctphoff;
173 
174 	/* Only update csum if we really have to */
175 	if (sctph->dest != cp->dport || payload_csum ||
176 	    (skb->ip_summed == CHECKSUM_PARTIAL &&
177 	     !(skb_dst(skb)->dev->features & NETIF_F_SCTP_CRC))) {
178 		sctph->dest = cp->dport;
179 		if (!skb_is_gso(skb))
180 			sctp_nat_csum(skb, sctph, sctphoff);
181 	} else if (skb->ip_summed != CHECKSUM_PARTIAL) {
182 		skb->ip_summed = CHECKSUM_UNNECESSARY;
183 	}
184 
185 	return 1;
186 }
187 
188 static int
sctp_csum_check(int af,struct sk_buff * skb,struct ip_vs_protocol * pp,unsigned int sctphoff)189 sctp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp,
190 		unsigned int sctphoff)
191 {
192 	struct sctphdr *sh;
193 	__le32 cmp, val;
194 
195 	sh = (struct sctphdr *)(skb->data + sctphoff);
196 	cmp = sh->checksum;
197 	val = sctp_compute_cksum(skb, sctphoff);
198 
199 	if (val != cmp) {
200 		/* CRC failure, dump it. */
201 		IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
202 				"Failed checksum for");
203 		return 0;
204 	}
205 	return 1;
206 }
207 
208 enum ipvs_sctp_event_t {
209 	IP_VS_SCTP_DATA = 0,		/* DATA, SACK, HEARTBEATs */
210 	IP_VS_SCTP_INIT,
211 	IP_VS_SCTP_INIT_ACK,
212 	IP_VS_SCTP_COOKIE_ECHO,
213 	IP_VS_SCTP_COOKIE_ACK,
214 	IP_VS_SCTP_SHUTDOWN,
215 	IP_VS_SCTP_SHUTDOWN_ACK,
216 	IP_VS_SCTP_SHUTDOWN_COMPLETE,
217 	IP_VS_SCTP_ERROR,
218 	IP_VS_SCTP_ABORT,
219 	IP_VS_SCTP_EVENT_LAST
220 };
221 
222 /* RFC 2960, 3.2 Chunk Field Descriptions */
223 static __u8 sctp_events[] = {
224 	[SCTP_CID_DATA]			= IP_VS_SCTP_DATA,
225 	[SCTP_CID_INIT]			= IP_VS_SCTP_INIT,
226 	[SCTP_CID_INIT_ACK]		= IP_VS_SCTP_INIT_ACK,
227 	[SCTP_CID_SACK]			= IP_VS_SCTP_DATA,
228 	[SCTP_CID_HEARTBEAT]		= IP_VS_SCTP_DATA,
229 	[SCTP_CID_HEARTBEAT_ACK]	= IP_VS_SCTP_DATA,
230 	[SCTP_CID_ABORT]		= IP_VS_SCTP_ABORT,
231 	[SCTP_CID_SHUTDOWN]		= IP_VS_SCTP_SHUTDOWN,
232 	[SCTP_CID_SHUTDOWN_ACK]		= IP_VS_SCTP_SHUTDOWN_ACK,
233 	[SCTP_CID_ERROR]		= IP_VS_SCTP_ERROR,
234 	[SCTP_CID_COOKIE_ECHO]		= IP_VS_SCTP_COOKIE_ECHO,
235 	[SCTP_CID_COOKIE_ACK]		= IP_VS_SCTP_COOKIE_ACK,
236 	[SCTP_CID_ECN_ECNE]		= IP_VS_SCTP_DATA,
237 	[SCTP_CID_ECN_CWR]		= IP_VS_SCTP_DATA,
238 	[SCTP_CID_SHUTDOWN_COMPLETE]	= IP_VS_SCTP_SHUTDOWN_COMPLETE,
239 };
240 
241 /* SCTP States:
242  * See RFC 2960, 4. SCTP Association State Diagram
243  *
244  * New states (not in diagram):
245  * - INIT1 state: use shorter timeout for dropped INIT packets
246  * - REJECTED state: use shorter timeout if INIT is rejected with ABORT
247  * - INIT, COOKIE_SENT, COOKIE_REPLIED, COOKIE states: for better debugging
248  *
249  * The states are as seen in real server. In the diagram, INIT1, INIT,
250  * COOKIE_SENT and COOKIE_REPLIED processing happens in CLOSED state.
251  *
252  * States as per packets from client (C) and server (S):
253  *
254  * Setup of client connection:
255  * IP_VS_SCTP_S_INIT1: First C:INIT sent, wait for S:INIT-ACK
256  * IP_VS_SCTP_S_INIT: Next C:INIT sent, wait for S:INIT-ACK
257  * IP_VS_SCTP_S_COOKIE_SENT: S:INIT-ACK sent, wait for C:COOKIE-ECHO
258  * IP_VS_SCTP_S_COOKIE_REPLIED: C:COOKIE-ECHO sent, wait for S:COOKIE-ACK
259  *
260  * Setup of server connection:
261  * IP_VS_SCTP_S_COOKIE_WAIT: S:INIT sent, wait for C:INIT-ACK
262  * IP_VS_SCTP_S_COOKIE: C:INIT-ACK sent, wait for S:COOKIE-ECHO
263  * IP_VS_SCTP_S_COOKIE_ECHOED: S:COOKIE-ECHO sent, wait for C:COOKIE-ACK
264  */
265 
266 #define sNO IP_VS_SCTP_S_NONE
267 #define sI1 IP_VS_SCTP_S_INIT1
268 #define sIN IP_VS_SCTP_S_INIT
269 #define sCS IP_VS_SCTP_S_COOKIE_SENT
270 #define sCR IP_VS_SCTP_S_COOKIE_REPLIED
271 #define sCW IP_VS_SCTP_S_COOKIE_WAIT
272 #define sCO IP_VS_SCTP_S_COOKIE
273 #define sCE IP_VS_SCTP_S_COOKIE_ECHOED
274 #define sES IP_VS_SCTP_S_ESTABLISHED
275 #define sSS IP_VS_SCTP_S_SHUTDOWN_SENT
276 #define sSR IP_VS_SCTP_S_SHUTDOWN_RECEIVED
277 #define sSA IP_VS_SCTP_S_SHUTDOWN_ACK_SENT
278 #define sRJ IP_VS_SCTP_S_REJECTED
279 #define sCL IP_VS_SCTP_S_CLOSED
280 
281 static const __u8 sctp_states
282 	[IP_VS_DIR_LAST][IP_VS_SCTP_EVENT_LAST][IP_VS_SCTP_S_LAST] = {
283 	{ /* INPUT */
284 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
285 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
286 /* i   */{sI1, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
287 /* i_a */{sCW, sCW, sCW, sCS, sCR, sCO, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
288 /* c_e */{sCR, sIN, sIN, sCR, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
289 /* c_a */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sES, sES, sSS, sSR, sSA, sRJ, sCL},
290 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
291 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sCL, sSR, sCL, sRJ, sCL},
292 /* s_c */{sCL, sCL, sCL, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sCL, sRJ, sCL},
293 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCL, sES, sSS, sSR, sSA, sRJ, sCL},
294 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
295 	},
296 	{ /* OUTPUT */
297 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
298 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
299 /* i   */{sCW, sCW, sCW, sCW, sCW, sCW, sCW, sCW, sES, sCW, sCW, sCW, sCW, sCW},
300 /* i_a */{sCS, sCS, sCS, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
301 /* c_e */{sCE, sCE, sCE, sCE, sCE, sCE, sCE, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
302 /* c_a */{sES, sES, sES, sES, sES, sES, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
303 /* s   */{sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSR, sSA, sRJ, sCL},
304 /* s_a */{sSA, sSA, sSA, sSA, sSA, sCW, sCO, sCE, sES, sSA, sSA, sSA, sRJ, sCL},
305 /* s_c */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
306 /* err */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
307 /* ab  */{sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
308 	},
309 	{ /* INPUT-ONLY */
310 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
311 /* d   */{sES, sI1, sIN, sCS, sCR, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
312 /* i   */{sI1, sIN, sIN, sIN, sIN, sIN, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
313 /* i_a */{sCE, sCE, sCE, sCE, sCE, sCE, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
314 /* c_e */{sES, sES, sES, sES, sES, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
315 /* c_a */{sES, sI1, sIN, sES, sES, sCW, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
316 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
317 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sCL, sCL, sSR, sCL, sRJ, sCL},
318 /* s_c */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sCL, sCL, sRJ, sCL},
319 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
320 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
321 	},
322 };
323 
324 #define IP_VS_SCTP_MAX_RTO	((60 + 1) * HZ)
325 
326 /* Timeout table[state] */
327 static const int sctp_timeouts[IP_VS_SCTP_S_LAST + 1] = {
328 	[IP_VS_SCTP_S_NONE]			= 2 * HZ,
329 	[IP_VS_SCTP_S_INIT1]			= (0 + 3 + 1) * HZ,
330 	[IP_VS_SCTP_S_INIT]			= IP_VS_SCTP_MAX_RTO,
331 	[IP_VS_SCTP_S_COOKIE_SENT]		= IP_VS_SCTP_MAX_RTO,
332 	[IP_VS_SCTP_S_COOKIE_REPLIED]		= IP_VS_SCTP_MAX_RTO,
333 	[IP_VS_SCTP_S_COOKIE_WAIT]		= IP_VS_SCTP_MAX_RTO,
334 	[IP_VS_SCTP_S_COOKIE]			= IP_VS_SCTP_MAX_RTO,
335 	[IP_VS_SCTP_S_COOKIE_ECHOED]		= IP_VS_SCTP_MAX_RTO,
336 	[IP_VS_SCTP_S_ESTABLISHED]		= 15 * 60 * HZ,
337 	[IP_VS_SCTP_S_SHUTDOWN_SENT]		= IP_VS_SCTP_MAX_RTO,
338 	[IP_VS_SCTP_S_SHUTDOWN_RECEIVED]	= IP_VS_SCTP_MAX_RTO,
339 	[IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]	= IP_VS_SCTP_MAX_RTO,
340 	[IP_VS_SCTP_S_REJECTED]			= (0 + 3 + 1) * HZ,
341 	[IP_VS_SCTP_S_CLOSED]			= IP_VS_SCTP_MAX_RTO,
342 	[IP_VS_SCTP_S_LAST]			= 2 * HZ,
343 };
344 
345 static const char *sctp_state_name_table[IP_VS_SCTP_S_LAST + 1] = {
346 	[IP_VS_SCTP_S_NONE]			= "NONE",
347 	[IP_VS_SCTP_S_INIT1]			= "INIT1",
348 	[IP_VS_SCTP_S_INIT]			= "INIT",
349 	[IP_VS_SCTP_S_COOKIE_SENT]		= "C-SENT",
350 	[IP_VS_SCTP_S_COOKIE_REPLIED]		= "C-REPLIED",
351 	[IP_VS_SCTP_S_COOKIE_WAIT]		= "C-WAIT",
352 	[IP_VS_SCTP_S_COOKIE]			= "COOKIE",
353 	[IP_VS_SCTP_S_COOKIE_ECHOED]		= "C-ECHOED",
354 	[IP_VS_SCTP_S_ESTABLISHED]		= "ESTABLISHED",
355 	[IP_VS_SCTP_S_SHUTDOWN_SENT]		= "S-SENT",
356 	[IP_VS_SCTP_S_SHUTDOWN_RECEIVED]	= "S-RECEIVED",
357 	[IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]	= "S-ACK-SENT",
358 	[IP_VS_SCTP_S_REJECTED]			= "REJECTED",
359 	[IP_VS_SCTP_S_CLOSED]			= "CLOSED",
360 	[IP_VS_SCTP_S_LAST]			= "BUG!",
361 };
362 
363 
sctp_state_name(int state)364 static const char *sctp_state_name(int state)
365 {
366 	if (state >= IP_VS_SCTP_S_LAST)
367 		return "ERR!";
368 	if (sctp_state_name_table[state])
369 		return sctp_state_name_table[state];
370 	return "?";
371 }
372 
373 static inline void
set_sctp_state(struct ip_vs_proto_data * pd,struct ip_vs_conn * cp,int direction,const struct sk_buff * skb)374 set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
375 		int direction, const struct sk_buff *skb)
376 {
377 	struct sctp_chunkhdr _sctpch, *sch;
378 	unsigned char chunk_type;
379 	int event, next_state;
380 	int ihl, cofs;
381 
382 #ifdef CONFIG_IP_VS_IPV6
383 	ihl = cp->af == AF_INET ? ip_hdrlen(skb) : sizeof(struct ipv6hdr);
384 #else
385 	ihl = ip_hdrlen(skb);
386 #endif
387 
388 	cofs = ihl + sizeof(struct sctphdr);
389 	sch = skb_header_pointer(skb, cofs, sizeof(_sctpch), &_sctpch);
390 	if (sch == NULL)
391 		return;
392 
393 	chunk_type = sch->type;
394 	/*
395 	 * Section 3: Multiple chunks can be bundled into one SCTP packet
396 	 * up to the MTU size, except for the INIT, INIT ACK, and
397 	 * SHUTDOWN COMPLETE chunks. These chunks MUST NOT be bundled with
398 	 * any other chunk in a packet.
399 	 *
400 	 * Section 3.3.7: DATA chunks MUST NOT be bundled with ABORT. Control
401 	 * chunks (except for INIT, INIT ACK, and SHUTDOWN COMPLETE) MAY be
402 	 * bundled with an ABORT, but they MUST be placed before the ABORT
403 	 * in the SCTP packet or they will be ignored by the receiver.
404 	 */
405 	if ((sch->type == SCTP_CID_COOKIE_ECHO) ||
406 	    (sch->type == SCTP_CID_COOKIE_ACK)) {
407 		int clen = ntohs(sch->length);
408 
409 		if (clen >= sizeof(_sctpch)) {
410 			sch = skb_header_pointer(skb, cofs + ALIGN(clen, 4),
411 						 sizeof(_sctpch), &_sctpch);
412 			if (sch && sch->type == SCTP_CID_ABORT)
413 				chunk_type = sch->type;
414 		}
415 	}
416 
417 	event = (chunk_type < sizeof(sctp_events)) ?
418 		sctp_events[chunk_type] : IP_VS_SCTP_DATA;
419 
420 	/* Update direction to INPUT_ONLY if necessary
421 	 * or delete NO_OUTPUT flag if output packet detected
422 	 */
423 	if (cp->flags & IP_VS_CONN_F_NOOUTPUT) {
424 		if (direction == IP_VS_DIR_OUTPUT)
425 			cp->flags &= ~IP_VS_CONN_F_NOOUTPUT;
426 		else
427 			direction = IP_VS_DIR_INPUT_ONLY;
428 	}
429 
430 	next_state = sctp_states[direction][event][cp->state];
431 
432 	if (next_state != cp->state) {
433 		struct ip_vs_dest *dest = cp->dest;
434 
435 		IP_VS_DBG_BUF(8, "%s %s  %s:%d->"
436 				"%s:%d state: %s->%s conn->refcnt:%d\n",
437 				pd->pp->name,
438 				((direction == IP_VS_DIR_OUTPUT) ?
439 				 "output " : "input "),
440 				IP_VS_DBG_ADDR(cp->daf, &cp->daddr),
441 				ntohs(cp->dport),
442 				IP_VS_DBG_ADDR(cp->af, &cp->caddr),
443 				ntohs(cp->cport),
444 				sctp_state_name(cp->state),
445 				sctp_state_name(next_state),
446 				refcount_read(&cp->refcnt));
447 		if (dest) {
448 			if (!(cp->flags & IP_VS_CONN_F_INACTIVE) &&
449 				(next_state != IP_VS_SCTP_S_ESTABLISHED)) {
450 				atomic_dec(&dest->activeconns);
451 				atomic_inc(&dest->inactconns);
452 				cp->flags |= IP_VS_CONN_F_INACTIVE;
453 			} else if ((cp->flags & IP_VS_CONN_F_INACTIVE) &&
454 				   (next_state == IP_VS_SCTP_S_ESTABLISHED)) {
455 				atomic_inc(&dest->activeconns);
456 				atomic_dec(&dest->inactconns);
457 				cp->flags &= ~IP_VS_CONN_F_INACTIVE;
458 			}
459 		}
460 		if (next_state == IP_VS_SCTP_S_ESTABLISHED)
461 			ip_vs_control_assure_ct(cp);
462 	}
463 	if (likely(pd))
464 		cp->timeout = pd->timeout_table[cp->state = next_state];
465 	else	/* What to do ? */
466 		cp->timeout = sctp_timeouts[cp->state = next_state];
467 }
468 
469 static void
sctp_state_transition(struct ip_vs_conn * cp,int direction,const struct sk_buff * skb,struct ip_vs_proto_data * pd)470 sctp_state_transition(struct ip_vs_conn *cp, int direction,
471 		const struct sk_buff *skb, struct ip_vs_proto_data *pd)
472 {
473 	spin_lock_bh(&cp->lock);
474 	set_sctp_state(pd, cp, direction, skb);
475 	spin_unlock_bh(&cp->lock);
476 }
477 
sctp_app_hashkey(__be16 port)478 static inline __u16 sctp_app_hashkey(__be16 port)
479 {
480 	return (((__force u16)port >> SCTP_APP_TAB_BITS) ^ (__force u16)port)
481 		& SCTP_APP_TAB_MASK;
482 }
483 
sctp_register_app(struct netns_ipvs * ipvs,struct ip_vs_app * inc)484 static int sctp_register_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
485 {
486 	struct ip_vs_app *i;
487 	__u16 hash;
488 	__be16 port = inc->port;
489 	int ret = 0;
490 	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
491 
492 	hash = sctp_app_hashkey(port);
493 
494 	list_for_each_entry(i, &ipvs->sctp_apps[hash], p_list) {
495 		if (i->port == port) {
496 			ret = -EEXIST;
497 			goto out;
498 		}
499 	}
500 	list_add_rcu(&inc->p_list, &ipvs->sctp_apps[hash]);
501 	atomic_inc(&pd->appcnt);
502 out:
503 
504 	return ret;
505 }
506 
sctp_unregister_app(struct netns_ipvs * ipvs,struct ip_vs_app * inc)507 static void sctp_unregister_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
508 {
509 	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
510 
511 	atomic_dec(&pd->appcnt);
512 	list_del_rcu(&inc->p_list);
513 }
514 
sctp_app_conn_bind(struct ip_vs_conn * cp)515 static int sctp_app_conn_bind(struct ip_vs_conn *cp)
516 {
517 	struct netns_ipvs *ipvs = cp->ipvs;
518 	int hash;
519 	struct ip_vs_app *inc;
520 	int result = 0;
521 
522 	/* Default binding: bind app only for NAT */
523 	if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
524 		return 0;
525 	/* Lookup application incarnations and bind the right one */
526 	hash = sctp_app_hashkey(cp->vport);
527 
528 	list_for_each_entry_rcu(inc, &ipvs->sctp_apps[hash], p_list) {
529 		if (inc->port == cp->vport) {
530 			if (unlikely(!ip_vs_app_inc_get(inc)))
531 				break;
532 
533 			IP_VS_DBG_BUF(9, "%s: Binding conn %s:%u->"
534 					"%s:%u to app %s on port %u\n",
535 					__func__,
536 					IP_VS_DBG_ADDR(cp->af, &cp->caddr),
537 					ntohs(cp->cport),
538 					IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
539 					ntohs(cp->vport),
540 					inc->name, ntohs(inc->port));
541 			cp->app = inc;
542 			if (inc->init_conn)
543 				result = inc->init_conn(inc, cp);
544 			break;
545 		}
546 	}
547 
548 	return result;
549 }
550 
551 /* ---------------------------------------------
552  *   timeouts is netns related now.
553  * ---------------------------------------------
554  */
__ip_vs_sctp_init(struct netns_ipvs * ipvs,struct ip_vs_proto_data * pd)555 static int __ip_vs_sctp_init(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
556 {
557 	ip_vs_init_hash_table(ipvs->sctp_apps, SCTP_APP_TAB_SIZE);
558 	pd->timeout_table = ip_vs_create_timeout_table((int *)sctp_timeouts,
559 							sizeof(sctp_timeouts));
560 	if (!pd->timeout_table)
561 		return -ENOMEM;
562 	return 0;
563 }
564 
__ip_vs_sctp_exit(struct netns_ipvs * ipvs,struct ip_vs_proto_data * pd)565 static void __ip_vs_sctp_exit(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
566 {
567 	kfree(pd->timeout_table);
568 }
569 
570 struct ip_vs_protocol ip_vs_protocol_sctp = {
571 	.name		= "SCTP",
572 	.protocol	= IPPROTO_SCTP,
573 	.num_states	= IP_VS_SCTP_S_LAST,
574 	.dont_defrag	= 0,
575 	.init		= NULL,
576 	.exit		= NULL,
577 	.init_netns	= __ip_vs_sctp_init,
578 	.exit_netns	= __ip_vs_sctp_exit,
579 	.register_app	= sctp_register_app,
580 	.unregister_app = sctp_unregister_app,
581 	.conn_schedule	= sctp_conn_schedule,
582 	.conn_in_get	= ip_vs_conn_in_get_proto,
583 	.conn_out_get	= ip_vs_conn_out_get_proto,
584 	.snat_handler	= sctp_snat_handler,
585 	.dnat_handler	= sctp_dnat_handler,
586 	.state_name	= sctp_state_name,
587 	.state_transition = sctp_state_transition,
588 	.app_conn_bind	= sctp_app_conn_bind,
589 	.debug_packet	= ip_vs_tcpudp_debug_packet,
590 	.timeout_change	= NULL,
591 };
592