GNU Linux-libre 6.1.90-gnu
[releases.git] / net / netfilter / ipvs / ip_vs_proto_sctp.c
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
15 static int
16 sctp_conn_schedule(struct netns_ipvs *ipvs, int af, struct sk_buff *skb,
17                    struct ip_vs_proto_data *pd,
18                    int *verdict, struct ip_vs_conn **cpp,
19                    struct ip_vs_iphdr *iph)
20 {
21         struct ip_vs_service *svc;
22         struct sctp_chunkhdr _schunkh, *sch;
23         struct sctphdr *sh, _sctph;
24         __be16 _ports[2], *ports = NULL;
25
26         if (likely(!ip_vs_iph_icmp(iph))) {
27                 sh = skb_header_pointer(skb, iph->len, sizeof(_sctph), &_sctph);
28                 if (sh) {
29                         sch = skb_header_pointer(skb, iph->len + sizeof(_sctph),
30                                                  sizeof(_schunkh), &_schunkh);
31                         if (sch) {
32                                 if (sch->type == SCTP_CID_ABORT ||
33                                     !(sysctl_sloppy_sctp(ipvs) ||
34                                       sch->type == SCTP_CID_INIT))
35                                         return 1;
36                                 ports = &sh->source;
37                         }
38                 }
39         } else {
40                 ports = skb_header_pointer(
41                         skb, iph->len, sizeof(_ports), &_ports);
42         }
43
44         if (!ports) {
45                 *verdict = NF_DROP;
46                 return 0;
47         }
48
49         if (likely(!ip_vs_iph_inverse(iph)))
50                 svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
51                                          &iph->daddr, ports[1]);
52         else
53                 svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
54                                          &iph->saddr, ports[0]);
55         if (svc) {
56                 int ignored;
57
58                 if (ip_vs_todrop(ipvs)) {
59                         /*
60                          * It seems that we are very loaded.
61                          * We have to drop this packet :(
62                          */
63                         *verdict = NF_DROP;
64                         return 0;
65                 }
66                 /*
67                  * Let the virtual server select a real server for the
68                  * incoming connection, and create a connection entry.
69                  */
70                 *cpp = ip_vs_schedule(svc, skb, pd, &ignored, iph);
71                 if (!*cpp && ignored <= 0) {
72                         if (!ignored)
73                                 *verdict = ip_vs_leave(svc, skb, pd, iph);
74                         else
75                                 *verdict = NF_DROP;
76                         return 0;
77                 }
78         }
79         /* NF_ACCEPT */
80         return 1;
81 }
82
83 static void sctp_nat_csum(struct sk_buff *skb, struct sctphdr *sctph,
84                           unsigned int sctphoff)
85 {
86         sctph->checksum = sctp_compute_cksum(skb, sctphoff);
87         skb->ip_summed = CHECKSUM_UNNECESSARY;
88 }
89
90 static int
91 sctp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
92                   struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
93 {
94         struct sctphdr *sctph;
95         unsigned int sctphoff = iph->len;
96         bool payload_csum = false;
97
98 #ifdef CONFIG_IP_VS_IPV6
99         if (cp->af == AF_INET6 && iph->fragoffs)
100                 return 1;
101 #endif
102
103         /* csum_check requires unshared skb */
104         if (skb_ensure_writable(skb, sctphoff + sizeof(*sctph)))
105                 return 0;
106
107         if (unlikely(cp->app != NULL)) {
108                 int ret;
109
110                 /* Some checks before mangling */
111                 if (!sctp_csum_check(cp->af, skb, pp))
112                         return 0;
113
114                 /* Call application helper if needed */
115                 ret = ip_vs_app_pkt_out(cp, skb, iph);
116                 if (ret == 0)
117                         return 0;
118                 /* ret=2: csum update is needed after payload mangling */
119                 if (ret == 2)
120                         payload_csum = true;
121         }
122
123         sctph = (void *) skb_network_header(skb) + sctphoff;
124
125         /* Only update csum if we really have to */
126         if (sctph->source != cp->vport || payload_csum ||
127             skb->ip_summed == CHECKSUM_PARTIAL) {
128                 sctph->source = cp->vport;
129                 if (!skb_is_gso(skb) || !skb_is_gso_sctp(skb))
130                         sctp_nat_csum(skb, sctph, sctphoff);
131         } else {
132                 skb->ip_summed = CHECKSUM_UNNECESSARY;
133         }
134
135         return 1;
136 }
137
138 static int
139 sctp_dnat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
140                   struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
141 {
142         struct sctphdr *sctph;
143         unsigned int sctphoff = iph->len;
144         bool payload_csum = false;
145
146 #ifdef CONFIG_IP_VS_IPV6
147         if (cp->af == AF_INET6 && iph->fragoffs)
148                 return 1;
149 #endif
150
151         /* csum_check requires unshared skb */
152         if (skb_ensure_writable(skb, sctphoff + sizeof(*sctph)))
153                 return 0;
154
155         if (unlikely(cp->app != NULL)) {
156                 int ret;
157
158                 /* Some checks before mangling */
159                 if (!sctp_csum_check(cp->af, skb, pp))
160                         return 0;
161
162                 /* Call application helper if needed */
163                 ret = ip_vs_app_pkt_in(cp, skb, iph);
164                 if (ret == 0)
165                         return 0;
166                 /* ret=2: csum update is needed after payload mangling */
167                 if (ret == 2)
168                         payload_csum = true;
169         }
170
171         sctph = (void *) skb_network_header(skb) + sctphoff;
172
173         /* Only update csum if we really have to */
174         if (sctph->dest != cp->dport || payload_csum ||
175             (skb->ip_summed == CHECKSUM_PARTIAL &&
176              !(skb_dst(skb)->dev->features & NETIF_F_SCTP_CRC))) {
177                 sctph->dest = cp->dport;
178                 if (!skb_is_gso(skb) || !skb_is_gso_sctp(skb))
179                         sctp_nat_csum(skb, sctph, sctphoff);
180         } else if (skb->ip_summed != CHECKSUM_PARTIAL) {
181                 skb->ip_summed = CHECKSUM_UNNECESSARY;
182         }
183
184         return 1;
185 }
186
187 static int
188 sctp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp)
189 {
190         unsigned int sctphoff;
191         struct sctphdr *sh;
192         __le32 cmp, val;
193
194 #ifdef CONFIG_IP_VS_IPV6
195         if (af == AF_INET6)
196                 sctphoff = sizeof(struct ipv6hdr);
197         else
198 #endif
199                 sctphoff = ip_hdrlen(skb);
200
201         sh = (struct sctphdr *)(skb->data + sctphoff);
202         cmp = sh->checksum;
203         val = sctp_compute_cksum(skb, sctphoff);
204
205         if (val != cmp) {
206                 /* CRC failure, dump it. */
207                 IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
208                                 "Failed checksum for");
209                 return 0;
210         }
211         return 1;
212 }
213
214 enum ipvs_sctp_event_t {
215         IP_VS_SCTP_DATA = 0,            /* DATA, SACK, HEARTBEATs */
216         IP_VS_SCTP_INIT,
217         IP_VS_SCTP_INIT_ACK,
218         IP_VS_SCTP_COOKIE_ECHO,
219         IP_VS_SCTP_COOKIE_ACK,
220         IP_VS_SCTP_SHUTDOWN,
221         IP_VS_SCTP_SHUTDOWN_ACK,
222         IP_VS_SCTP_SHUTDOWN_COMPLETE,
223         IP_VS_SCTP_ERROR,
224         IP_VS_SCTP_ABORT,
225         IP_VS_SCTP_EVENT_LAST
226 };
227
228 /* RFC 2960, 3.2 Chunk Field Descriptions */
229 static __u8 sctp_events[] = {
230         [SCTP_CID_DATA]                 = IP_VS_SCTP_DATA,
231         [SCTP_CID_INIT]                 = IP_VS_SCTP_INIT,
232         [SCTP_CID_INIT_ACK]             = IP_VS_SCTP_INIT_ACK,
233         [SCTP_CID_SACK]                 = IP_VS_SCTP_DATA,
234         [SCTP_CID_HEARTBEAT]            = IP_VS_SCTP_DATA,
235         [SCTP_CID_HEARTBEAT_ACK]        = IP_VS_SCTP_DATA,
236         [SCTP_CID_ABORT]                = IP_VS_SCTP_ABORT,
237         [SCTP_CID_SHUTDOWN]             = IP_VS_SCTP_SHUTDOWN,
238         [SCTP_CID_SHUTDOWN_ACK]         = IP_VS_SCTP_SHUTDOWN_ACK,
239         [SCTP_CID_ERROR]                = IP_VS_SCTP_ERROR,
240         [SCTP_CID_COOKIE_ECHO]          = IP_VS_SCTP_COOKIE_ECHO,
241         [SCTP_CID_COOKIE_ACK]           = IP_VS_SCTP_COOKIE_ACK,
242         [SCTP_CID_ECN_ECNE]             = IP_VS_SCTP_DATA,
243         [SCTP_CID_ECN_CWR]              = IP_VS_SCTP_DATA,
244         [SCTP_CID_SHUTDOWN_COMPLETE]    = IP_VS_SCTP_SHUTDOWN_COMPLETE,
245 };
246
247 /* SCTP States:
248  * See RFC 2960, 4. SCTP Association State Diagram
249  *
250  * New states (not in diagram):
251  * - INIT1 state: use shorter timeout for dropped INIT packets
252  * - REJECTED state: use shorter timeout if INIT is rejected with ABORT
253  * - INIT, COOKIE_SENT, COOKIE_REPLIED, COOKIE states: for better debugging
254  *
255  * The states are as seen in real server. In the diagram, INIT1, INIT,
256  * COOKIE_SENT and COOKIE_REPLIED processing happens in CLOSED state.
257  *
258  * States as per packets from client (C) and server (S):
259  *
260  * Setup of client connection:
261  * IP_VS_SCTP_S_INIT1: First C:INIT sent, wait for S:INIT-ACK
262  * IP_VS_SCTP_S_INIT: Next C:INIT sent, wait for S:INIT-ACK
263  * IP_VS_SCTP_S_COOKIE_SENT: S:INIT-ACK sent, wait for C:COOKIE-ECHO
264  * IP_VS_SCTP_S_COOKIE_REPLIED: C:COOKIE-ECHO sent, wait for S:COOKIE-ACK
265  *
266  * Setup of server connection:
267  * IP_VS_SCTP_S_COOKIE_WAIT: S:INIT sent, wait for C:INIT-ACK
268  * IP_VS_SCTP_S_COOKIE: C:INIT-ACK sent, wait for S:COOKIE-ECHO
269  * IP_VS_SCTP_S_COOKIE_ECHOED: S:COOKIE-ECHO sent, wait for C:COOKIE-ACK
270  */
271
272 #define sNO IP_VS_SCTP_S_NONE
273 #define sI1 IP_VS_SCTP_S_INIT1
274 #define sIN IP_VS_SCTP_S_INIT
275 #define sCS IP_VS_SCTP_S_COOKIE_SENT
276 #define sCR IP_VS_SCTP_S_COOKIE_REPLIED
277 #define sCW IP_VS_SCTP_S_COOKIE_WAIT
278 #define sCO IP_VS_SCTP_S_COOKIE
279 #define sCE IP_VS_SCTP_S_COOKIE_ECHOED
280 #define sES IP_VS_SCTP_S_ESTABLISHED
281 #define sSS IP_VS_SCTP_S_SHUTDOWN_SENT
282 #define sSR IP_VS_SCTP_S_SHUTDOWN_RECEIVED
283 #define sSA IP_VS_SCTP_S_SHUTDOWN_ACK_SENT
284 #define sRJ IP_VS_SCTP_S_REJECTED
285 #define sCL IP_VS_SCTP_S_CLOSED
286
287 static const __u8 sctp_states
288         [IP_VS_DIR_LAST][IP_VS_SCTP_EVENT_LAST][IP_VS_SCTP_S_LAST] = {
289         { /* INPUT */
290 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
291 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
292 /* i   */{sI1, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
293 /* i_a */{sCW, sCW, sCW, sCS, sCR, sCO, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
294 /* c_e */{sCR, sIN, sIN, sCR, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
295 /* c_a */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sES, sES, sSS, sSR, sSA, sRJ, sCL},
296 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
297 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sCL, sSR, sCL, sRJ, sCL},
298 /* s_c */{sCL, sCL, sCL, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sCL, sRJ, sCL},
299 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCL, sES, sSS, sSR, sSA, sRJ, sCL},
300 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
301         },
302         { /* OUTPUT */
303 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
304 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
305 /* i   */{sCW, sCW, sCW, sCW, sCW, sCW, sCW, sCW, sES, sCW, sCW, sCW, sCW, sCW},
306 /* i_a */{sCS, sCS, sCS, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
307 /* c_e */{sCE, sCE, sCE, sCE, sCE, sCE, sCE, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
308 /* c_a */{sES, sES, sES, sES, sES, sES, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
309 /* s   */{sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSR, sSA, sRJ, sCL},
310 /* s_a */{sSA, sSA, sSA, sSA, sSA, sCW, sCO, sCE, sES, sSA, sSA, sSA, sRJ, sCL},
311 /* s_c */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
312 /* err */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
313 /* ab  */{sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
314         },
315         { /* INPUT-ONLY */
316 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
317 /* d   */{sES, sI1, sIN, sCS, sCR, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
318 /* i   */{sI1, sIN, sIN, sIN, sIN, sIN, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
319 /* i_a */{sCE, sCE, sCE, sCE, sCE, sCE, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
320 /* c_e */{sES, sES, sES, sES, sES, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
321 /* c_a */{sES, sI1, sIN, sES, sES, sCW, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
322 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
323 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sCL, sCL, sSR, sCL, sRJ, sCL},
324 /* s_c */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sCL, sCL, sRJ, sCL},
325 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
326 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
327         },
328 };
329
330 #define IP_VS_SCTP_MAX_RTO      ((60 + 1) * HZ)
331
332 /* Timeout table[state] */
333 static const int sctp_timeouts[IP_VS_SCTP_S_LAST + 1] = {
334         [IP_VS_SCTP_S_NONE]                     = 2 * HZ,
335         [IP_VS_SCTP_S_INIT1]                    = (0 + 3 + 1) * HZ,
336         [IP_VS_SCTP_S_INIT]                     = IP_VS_SCTP_MAX_RTO,
337         [IP_VS_SCTP_S_COOKIE_SENT]              = IP_VS_SCTP_MAX_RTO,
338         [IP_VS_SCTP_S_COOKIE_REPLIED]           = IP_VS_SCTP_MAX_RTO,
339         [IP_VS_SCTP_S_COOKIE_WAIT]              = IP_VS_SCTP_MAX_RTO,
340         [IP_VS_SCTP_S_COOKIE]                   = IP_VS_SCTP_MAX_RTO,
341         [IP_VS_SCTP_S_COOKIE_ECHOED]            = IP_VS_SCTP_MAX_RTO,
342         [IP_VS_SCTP_S_ESTABLISHED]              = 15 * 60 * HZ,
343         [IP_VS_SCTP_S_SHUTDOWN_SENT]            = IP_VS_SCTP_MAX_RTO,
344         [IP_VS_SCTP_S_SHUTDOWN_RECEIVED]        = IP_VS_SCTP_MAX_RTO,
345         [IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]        = IP_VS_SCTP_MAX_RTO,
346         [IP_VS_SCTP_S_REJECTED]                 = (0 + 3 + 1) * HZ,
347         [IP_VS_SCTP_S_CLOSED]                   = IP_VS_SCTP_MAX_RTO,
348         [IP_VS_SCTP_S_LAST]                     = 2 * HZ,
349 };
350
351 static const char *sctp_state_name_table[IP_VS_SCTP_S_LAST + 1] = {
352         [IP_VS_SCTP_S_NONE]                     = "NONE",
353         [IP_VS_SCTP_S_INIT1]                    = "INIT1",
354         [IP_VS_SCTP_S_INIT]                     = "INIT",
355         [IP_VS_SCTP_S_COOKIE_SENT]              = "C-SENT",
356         [IP_VS_SCTP_S_COOKIE_REPLIED]           = "C-REPLIED",
357         [IP_VS_SCTP_S_COOKIE_WAIT]              = "C-WAIT",
358         [IP_VS_SCTP_S_COOKIE]                   = "COOKIE",
359         [IP_VS_SCTP_S_COOKIE_ECHOED]            = "C-ECHOED",
360         [IP_VS_SCTP_S_ESTABLISHED]              = "ESTABLISHED",
361         [IP_VS_SCTP_S_SHUTDOWN_SENT]            = "S-SENT",
362         [IP_VS_SCTP_S_SHUTDOWN_RECEIVED]        = "S-RECEIVED",
363         [IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]        = "S-ACK-SENT",
364         [IP_VS_SCTP_S_REJECTED]                 = "REJECTED",
365         [IP_VS_SCTP_S_CLOSED]                   = "CLOSED",
366         [IP_VS_SCTP_S_LAST]                     = "BUG!",
367 };
368
369
370 static const char *sctp_state_name(int state)
371 {
372         if (state >= IP_VS_SCTP_S_LAST)
373                 return "ERR!";
374         if (sctp_state_name_table[state])
375                 return sctp_state_name_table[state];
376         return "?";
377 }
378
379 static inline void
380 set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
381                 int direction, const struct sk_buff *skb)
382 {
383         struct sctp_chunkhdr _sctpch, *sch;
384         unsigned char chunk_type;
385         int event, next_state;
386         int ihl, cofs;
387
388 #ifdef CONFIG_IP_VS_IPV6
389         ihl = cp->af == AF_INET ? ip_hdrlen(skb) : sizeof(struct ipv6hdr);
390 #else
391         ihl = ip_hdrlen(skb);
392 #endif
393
394         cofs = ihl + sizeof(struct sctphdr);
395         sch = skb_header_pointer(skb, cofs, sizeof(_sctpch), &_sctpch);
396         if (sch == NULL)
397                 return;
398
399         chunk_type = sch->type;
400         /*
401          * Section 3: Multiple chunks can be bundled into one SCTP packet
402          * up to the MTU size, except for the INIT, INIT ACK, and
403          * SHUTDOWN COMPLETE chunks. These chunks MUST NOT be bundled with
404          * any other chunk in a packet.
405          *
406          * Section 3.3.7: DATA chunks MUST NOT be bundled with ABORT. Control
407          * chunks (except for INIT, INIT ACK, and SHUTDOWN COMPLETE) MAY be
408          * bundled with an ABORT, but they MUST be placed before the ABORT
409          * in the SCTP packet or they will be ignored by the receiver.
410          */
411         if ((sch->type == SCTP_CID_COOKIE_ECHO) ||
412             (sch->type == SCTP_CID_COOKIE_ACK)) {
413                 int clen = ntohs(sch->length);
414
415                 if (clen >= sizeof(_sctpch)) {
416                         sch = skb_header_pointer(skb, cofs + ALIGN(clen, 4),
417                                                  sizeof(_sctpch), &_sctpch);
418                         if (sch && sch->type == SCTP_CID_ABORT)
419                                 chunk_type = sch->type;
420                 }
421         }
422
423         event = (chunk_type < sizeof(sctp_events)) ?
424                 sctp_events[chunk_type] : IP_VS_SCTP_DATA;
425
426         /* Update direction to INPUT_ONLY if necessary
427          * or delete NO_OUTPUT flag if output packet detected
428          */
429         if (cp->flags & IP_VS_CONN_F_NOOUTPUT) {
430                 if (direction == IP_VS_DIR_OUTPUT)
431                         cp->flags &= ~IP_VS_CONN_F_NOOUTPUT;
432                 else
433                         direction = IP_VS_DIR_INPUT_ONLY;
434         }
435
436         next_state = sctp_states[direction][event][cp->state];
437
438         if (next_state != cp->state) {
439                 struct ip_vs_dest *dest = cp->dest;
440
441                 IP_VS_DBG_BUF(8, "%s %s  %s:%d->"
442                                 "%s:%d state: %s->%s conn->refcnt:%d\n",
443                                 pd->pp->name,
444                                 ((direction == IP_VS_DIR_OUTPUT) ?
445                                  "output " : "input "),
446                                 IP_VS_DBG_ADDR(cp->daf, &cp->daddr),
447                                 ntohs(cp->dport),
448                                 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
449                                 ntohs(cp->cport),
450                                 sctp_state_name(cp->state),
451                                 sctp_state_name(next_state),
452                                 refcount_read(&cp->refcnt));
453                 if (dest) {
454                         if (!(cp->flags & IP_VS_CONN_F_INACTIVE) &&
455                                 (next_state != IP_VS_SCTP_S_ESTABLISHED)) {
456                                 atomic_dec(&dest->activeconns);
457                                 atomic_inc(&dest->inactconns);
458                                 cp->flags |= IP_VS_CONN_F_INACTIVE;
459                         } else if ((cp->flags & IP_VS_CONN_F_INACTIVE) &&
460                                    (next_state == IP_VS_SCTP_S_ESTABLISHED)) {
461                                 atomic_inc(&dest->activeconns);
462                                 atomic_dec(&dest->inactconns);
463                                 cp->flags &= ~IP_VS_CONN_F_INACTIVE;
464                         }
465                 }
466                 if (next_state == IP_VS_SCTP_S_ESTABLISHED)
467                         ip_vs_control_assure_ct(cp);
468         }
469         if (likely(pd))
470                 cp->timeout = pd->timeout_table[cp->state = next_state];
471         else    /* What to do ? */
472                 cp->timeout = sctp_timeouts[cp->state = next_state];
473 }
474
475 static void
476 sctp_state_transition(struct ip_vs_conn *cp, int direction,
477                 const struct sk_buff *skb, struct ip_vs_proto_data *pd)
478 {
479         spin_lock_bh(&cp->lock);
480         set_sctp_state(pd, cp, direction, skb);
481         spin_unlock_bh(&cp->lock);
482 }
483
484 static inline __u16 sctp_app_hashkey(__be16 port)
485 {
486         return (((__force u16)port >> SCTP_APP_TAB_BITS) ^ (__force u16)port)
487                 & SCTP_APP_TAB_MASK;
488 }
489
490 static int sctp_register_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
491 {
492         struct ip_vs_app *i;
493         __u16 hash;
494         __be16 port = inc->port;
495         int ret = 0;
496         struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
497
498         hash = sctp_app_hashkey(port);
499
500         list_for_each_entry(i, &ipvs->sctp_apps[hash], p_list) {
501                 if (i->port == port) {
502                         ret = -EEXIST;
503                         goto out;
504                 }
505         }
506         list_add_rcu(&inc->p_list, &ipvs->sctp_apps[hash]);
507         atomic_inc(&pd->appcnt);
508 out:
509
510         return ret;
511 }
512
513 static void sctp_unregister_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
514 {
515         struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
516
517         atomic_dec(&pd->appcnt);
518         list_del_rcu(&inc->p_list);
519 }
520
521 static int sctp_app_conn_bind(struct ip_vs_conn *cp)
522 {
523         struct netns_ipvs *ipvs = cp->ipvs;
524         int hash;
525         struct ip_vs_app *inc;
526         int result = 0;
527
528         /* Default binding: bind app only for NAT */
529         if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
530                 return 0;
531         /* Lookup application incarnations and bind the right one */
532         hash = sctp_app_hashkey(cp->vport);
533
534         list_for_each_entry_rcu(inc, &ipvs->sctp_apps[hash], p_list) {
535                 if (inc->port == cp->vport) {
536                         if (unlikely(!ip_vs_app_inc_get(inc)))
537                                 break;
538
539                         IP_VS_DBG_BUF(9, "%s: Binding conn %s:%u->"
540                                         "%s:%u to app %s on port %u\n",
541                                         __func__,
542                                         IP_VS_DBG_ADDR(cp->af, &cp->caddr),
543                                         ntohs(cp->cport),
544                                         IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
545                                         ntohs(cp->vport),
546                                         inc->name, ntohs(inc->port));
547                         cp->app = inc;
548                         if (inc->init_conn)
549                                 result = inc->init_conn(inc, cp);
550                         break;
551                 }
552         }
553
554         return result;
555 }
556
557 /* ---------------------------------------------
558  *   timeouts is netns related now.
559  * ---------------------------------------------
560  */
561 static int __ip_vs_sctp_init(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
562 {
563         ip_vs_init_hash_table(ipvs->sctp_apps, SCTP_APP_TAB_SIZE);
564         pd->timeout_table = ip_vs_create_timeout_table((int *)sctp_timeouts,
565                                                         sizeof(sctp_timeouts));
566         if (!pd->timeout_table)
567                 return -ENOMEM;
568         return 0;
569 }
570
571 static void __ip_vs_sctp_exit(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
572 {
573         kfree(pd->timeout_table);
574 }
575
576 struct ip_vs_protocol ip_vs_protocol_sctp = {
577         .name           = "SCTP",
578         .protocol       = IPPROTO_SCTP,
579         .num_states     = IP_VS_SCTP_S_LAST,
580         .dont_defrag    = 0,
581         .init           = NULL,
582         .exit           = NULL,
583         .init_netns     = __ip_vs_sctp_init,
584         .exit_netns     = __ip_vs_sctp_exit,
585         .register_app   = sctp_register_app,
586         .unregister_app = sctp_unregister_app,
587         .conn_schedule  = sctp_conn_schedule,
588         .conn_in_get    = ip_vs_conn_in_get_proto,
589         .conn_out_get   = ip_vs_conn_out_get_proto,
590         .snat_handler   = sctp_snat_handler,
591         .dnat_handler   = sctp_dnat_handler,
592         .state_name     = sctp_state_name,
593         .state_transition = sctp_state_transition,
594         .app_conn_bind  = sctp_app_conn_bind,
595         .debug_packet   = ip_vs_tcpudp_debug_packet,
596         .timeout_change = NULL,
597 };