net/netfilter/ipvs/ip_vs_proto_sctp.c

   1 #include <linux/kernel.h>
   2 #include <linux/ip.h>
   3 #include <linux/sctp.h>
   4 #include <net/ip.h>
   5 #include <net/ip6_checksum.h>
   6 #include <linux/netfilter.h>
   7 #include <linux/netfilter_ipv4.h>
   8 #include <net/sctp/checksum.h>
   9 #include <net/ip_vs.h>
  10
  11 static int
  12 sctp_conn_schedule(struct netns_ipvs *ipvs, int af, struct sk_buff *skb,
  13                    struct ip_vs_proto_data *pd,
  14                    int *verdict, struct ip_vs_conn **cpp,
  15                    struct ip_vs_iphdr *iph)
  16 {
  17         struct ip_vs_service *svc;
  18         struct sctp_chunkhdr _schunkh, *sch;
  19         struct sctphdr *sh, _sctph;
  20         __be16 _ports[2], *ports = NULL;
  21
  22         if (likely(!ip_vs_iph_icmp(iph))) {
  23                 sh = skb_header_pointer(skb, iph->len, sizeof(_sctph), &_sctph);
  24                 if (sh) {
  25                         sch = skb_header_pointer(skb, iph->len + sizeof(_sctph),
  26                                                  sizeof(_schunkh), &_schunkh);
  27                         if (sch) {
  28                                 if (sch->type == SCTP_CID_ABORT ||
  29                                     !(sysctl_sloppy_sctp(ipvs) ||
  30                                       sch->type == SCTP_CID_INIT))
  31                                         return 1;
  32                                 ports = &sh->source;
  33                         }
  34                 }
  35         } else {
  36                 ports = skb_header_pointer(
  37                         skb, iph->len, sizeof(_ports), &_ports);
  38         }
  39
  40         if (!ports) {
  41                 *verdict = NF_DROP;
  42                 return 0;
  43         }
  44
  45         if (likely(!ip_vs_iph_inverse(iph)))
  46                 svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
  47                                          &iph->daddr, ports[1]);
  48         else
  49                 svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
  50                                          &iph->saddr, ports[0]);
  51         if (svc) {
  52                 int ignored;
  53
  54                 if (ip_vs_todrop(ipvs)) {
  55                         /*
  56                          * It seems that we are very loaded.
  57                          * We have to drop this packet :(
  58                          */
  59                         *verdict = NF_DROP;
  60                         return 0;
  61                 }
  62                 /*
  63                  * Let the virtual server select a real server for the
  64                  * incoming connection, and create a connection entry.
  65                  */
  66                 *cpp = ip_vs_schedule(svc, skb, pd, &ignored, iph);
  67                 if (!*cpp && ignored <= 0) {
  68                         if (!ignored)
  69                                 *verdict = ip_vs_leave(svc, skb, pd, iph);
  70                         else
  71                                 *verdict = NF_DROP;
  72                         return 0;
  73                 }
  74         }
  75         /* NF_ACCEPT */
  76         return 1;
  77 }
  78
  79 static void sctp_nat_csum(struct sk_buff *skb, struct sctphdr *sctph,
  80                           unsigned int sctphoff)
  81 {
  82         sctph->checksum = sctp_compute_cksum(skb, sctphoff);
  83         skb->ip_summed = CHECKSUM_UNNECESSARY;
  84 }
  85
  86 static int
  87 sctp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
  88                   struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
  89 {
  90         struct sctphdr *sctph;
  91         unsigned int sctphoff = iph->len;
  92         bool payload_csum = false;
  93
  94 #ifdef CONFIG_IP_VS_IPV6
  95         if (cp->af == AF_INET6 && iph->fragoffs)
  96                 return 1;
  97 #endif
  98
  99         /* csum_check requires unshared skb */
 100         if (!skb_make_writable(skb, sctphoff + sizeof(*sctph)))
 101                 return 0;
 102
 103         if (unlikely(cp->app != NULL)) {
 104                 int ret;
 105
 106                 /* Some checks before mangling */
 107                 if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
 108                         return 0;
 109
 110                 /* Call application helper if needed */
 111                 ret = ip_vs_app_pkt_out(cp, skb);
 112                 if (ret == 0)
 113                         return 0;
 114                 /* ret=2: csum update is needed after payload mangling */
 115                 if (ret == 2)
 116                         payload_csum = true;
 117         }
 118
 119         sctph = (void *) skb_network_header(skb) + sctphoff;
 120
 121         /* Only update csum if we really have to */
 122         if (sctph->source != cp->vport || payload_csum ||
 123             skb->ip_summed == CHECKSUM_PARTIAL) {
 124                 sctph->source = cp->vport;
 125                 sctp_nat_csum(skb, sctph, sctphoff);
 126         } else {
 127                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 128         }
 129
 130         return 1;
 131 }
 132
 133 static int
 134 sctp_dnat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
 135                   struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
 136 {
 137         struct sctphdr *sctph;
 138         unsigned int sctphoff = iph->len;
 139         bool payload_csum = false;
 140
 141 #ifdef CONFIG_IP_VS_IPV6
 142         if (cp->af == AF_INET6 && iph->fragoffs)
 143                 return 1;
 144 #endif
 145
 146         /* csum_check requires unshared skb */
 147         if (!skb_make_writable(skb, sctphoff + sizeof(*sctph)))
 148                 return 0;
 149
 150         if (unlikely(cp->app != NULL)) {
 151                 int ret;
 152
 153                 /* Some checks before mangling */
 154                 if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
 155                         return 0;
 156
 157                 /* Call application helper if needed */
 158                 ret = ip_vs_app_pkt_in(cp, skb);
 159                 if (ret == 0)
 160                         return 0;
 161                 /* ret=2: csum update is needed after payload mangling */
 162                 if (ret == 2)
 163                         payload_csum = true;
 164         }
 165
 166         sctph = (void *) skb_network_header(skb) + sctphoff;
 167
 168         /* Only update csum if we really have to */
 169         if (sctph->dest != cp->dport || payload_csum ||
 170             (skb->ip_summed == CHECKSUM_PARTIAL &&
 171              !(skb_dst(skb)->dev->features & NETIF_F_SCTP_CRC))) {
 172                 sctph->dest = cp->dport;
 173                 sctp_nat_csum(skb, sctph, sctphoff);
 174         } else if (skb->ip_summed != CHECKSUM_PARTIAL) {
 175                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 176         }
 177
 178         return 1;
 179 }
 180
 181 static int
 182 sctp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp)
 183 {
 184         unsigned int sctphoff;
 185         struct sctphdr *sh, _sctph;
 186         __le32 cmp, val;
 187
 188 #ifdef CONFIG_IP_VS_IPV6
 189         if (af == AF_INET6)
 190                 sctphoff = sizeof(struct ipv6hdr);
 191         else
 192 #endif
 193                 sctphoff = ip_hdrlen(skb);
 194
 195         sh = skb_header_pointer(skb, sctphoff, sizeof(_sctph), &_sctph);
 196         if (sh == NULL)
 197                 return 0;
 198
 199         cmp = sh->checksum;
 200         val = sctp_compute_cksum(skb, sctphoff);
 201
 202         if (val != cmp) {
 203                 /* CRC failure, dump it. */
 204                 IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
 205                                 "Failed checksum for");
 206                 return 0;
 207         }
 208         return 1;
 209 }
 210
 211 enum ipvs_sctp_event_t {
 212         IP_VS_SCTP_DATA = 0,            /* DATA, SACK, HEARTBEATs */
 213         IP_VS_SCTP_INIT,
 214         IP_VS_SCTP_INIT_ACK,
 215         IP_VS_SCTP_COOKIE_ECHO,
 216         IP_VS_SCTP_COOKIE_ACK,
 217         IP_VS_SCTP_SHUTDOWN,
 218         IP_VS_SCTP_SHUTDOWN_ACK,
 219         IP_VS_SCTP_SHUTDOWN_COMPLETE,
 220         IP_VS_SCTP_ERROR,
 221         IP_VS_SCTP_ABORT,
 222         IP_VS_SCTP_EVENT_LAST
 223 };
 224
 225 /* RFC 2960, 3.2 Chunk Field Descriptions */
 226 static __u8 sctp_events[] = {
 227         [SCTP_CID_DATA]                 = IP_VS_SCTP_DATA,
 228         [SCTP_CID_INIT]                 = IP_VS_SCTP_INIT,
 229         [SCTP_CID_INIT_ACK]             = IP_VS_SCTP_INIT_ACK,
 230         [SCTP_CID_SACK]                 = IP_VS_SCTP_DATA,
 231         [SCTP_CID_HEARTBEAT]            = IP_VS_SCTP_DATA,
 232         [SCTP_CID_HEARTBEAT_ACK]        = IP_VS_SCTP_DATA,
 233         [SCTP_CID_ABORT]                = IP_VS_SCTP_ABORT,
 234         [SCTP_CID_SHUTDOWN]             = IP_VS_SCTP_SHUTDOWN,
 235         [SCTP_CID_SHUTDOWN_ACK]         = IP_VS_SCTP_SHUTDOWN_ACK,
 236         [SCTP_CID_ERROR]                = IP_VS_SCTP_ERROR,
 237         [SCTP_CID_COOKIE_ECHO]          = IP_VS_SCTP_COOKIE_ECHO,
 238         [SCTP_CID_COOKIE_ACK]           = IP_VS_SCTP_COOKIE_ACK,
 239         [SCTP_CID_ECN_ECNE]             = IP_VS_SCTP_DATA,
 240         [SCTP_CID_ECN_CWR]              = IP_VS_SCTP_DATA,
 241         [SCTP_CID_SHUTDOWN_COMPLETE]    = IP_VS_SCTP_SHUTDOWN_COMPLETE,
 242 };
 243
 244 /* SCTP States:
 245  * See RFC 2960, 4. SCTP Association State Diagram
 246  *
 247  * New states (not in diagram):
 248  * - INIT1 state: use shorter timeout for dropped INIT packets
 249  * - REJECTED state: use shorter timeout if INIT is rejected with ABORT
 250  * - INIT, COOKIE_SENT, COOKIE_REPLIED, COOKIE states: for better debugging
 251  *
 252  * The states are as seen in real server. In the diagram, INIT1, INIT,
 253  * COOKIE_SENT and COOKIE_REPLIED processing happens in CLOSED state.
 254  *
 255  * States as per packets from client (C) and server (S):
 256  *
 257  * Setup of client connection:
 258  * IP_VS_SCTP_S_INIT1: First C:INIT sent, wait for S:INIT-ACK
 259  * IP_VS_SCTP_S_INIT: Next C:INIT sent, wait for S:INIT-ACK
 260  * IP_VS_SCTP_S_COOKIE_SENT: S:INIT-ACK sent, wait for C:COOKIE-ECHO
 261  * IP_VS_SCTP_S_COOKIE_REPLIED: C:COOKIE-ECHO sent, wait for S:COOKIE-ACK
 262  *
 263  * Setup of server connection:
 264  * IP_VS_SCTP_S_COOKIE_WAIT: S:INIT sent, wait for C:INIT-ACK
 265  * IP_VS_SCTP_S_COOKIE: C:INIT-ACK sent, wait for S:COOKIE-ECHO
 266  * IP_VS_SCTP_S_COOKIE_ECHOED: S:COOKIE-ECHO sent, wait for C:COOKIE-ACK
 267  */
 268
 269 #define sNO IP_VS_SCTP_S_NONE
 270 #define sI1 IP_VS_SCTP_S_INIT1
 271 #define sIN IP_VS_SCTP_S_INIT
 272 #define sCS IP_VS_SCTP_S_COOKIE_SENT
 273 #define sCR IP_VS_SCTP_S_COOKIE_REPLIED
 274 #define sCW IP_VS_SCTP_S_COOKIE_WAIT
 275 #define sCO IP_VS_SCTP_S_COOKIE
 276 #define sCE IP_VS_SCTP_S_COOKIE_ECHOED
 277 #define sES IP_VS_SCTP_S_ESTABLISHED
 278 #define sSS IP_VS_SCTP_S_SHUTDOWN_SENT
 279 #define sSR IP_VS_SCTP_S_SHUTDOWN_RECEIVED
 280 #define sSA IP_VS_SCTP_S_SHUTDOWN_ACK_SENT
 281 #define sRJ IP_VS_SCTP_S_REJECTED
 282 #define sCL IP_VS_SCTP_S_CLOSED
 283
 284 static const __u8 sctp_states
 285         [IP_VS_DIR_LAST][IP_VS_SCTP_EVENT_LAST][IP_VS_SCTP_S_LAST] = {
 286         { /* INPUT */
 287 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
 288 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 289 /* i   */{sI1, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
 290 /* i_a */{sCW, sCW, sCW, sCS, sCR, sCO, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 291 /* c_e */{sCR, sIN, sIN, sCR, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 292 /* c_a */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sES, sES, sSS, sSR, sSA, sRJ, sCL},
 293 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
 294 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sES, sCL, sSR, sCL, sRJ, sCL},
 295 /* s_c */{sCL, sCL, sCL, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sCL, sRJ, sCL},
 296 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCL, sES, sSS, sSR, sSA, sRJ, sCL},
 297 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
 298         },
 299         { /* OUTPUT */
 300 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
 301 /* d   */{sES, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 302 /* i   */{sCW, sCW, sCW, sCW, sCW, sCW, sCW, sCW, sES, sCW, sCW, sCW, sCW, sCW},
 303 /* i_a */{sCS, sCS, sCS, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 304 /* c_e */{sCE, sCE, sCE, sCE, sCE, sCE, sCE, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 305 /* c_a */{sES, sES, sES, sES, sES, sES, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
 306 /* s   */{sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSS, sSR, sSA, sRJ, sCL},
 307 /* s_a */{sSA, sSA, sSA, sSA, sSA, sCW, sCO, sCE, sES, sSA, sSA, sSA, sRJ, sCL},
 308 /* s_c */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 309 /* err */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 310 /* ab  */{sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
 311         },
 312         { /* INPUT-ONLY */
 313 /*        sNO, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL*/
 314 /* d   */{sES, sI1, sIN, sCS, sCR, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 315 /* i   */{sI1, sIN, sIN, sIN, sIN, sIN, sCO, sCE, sES, sSS, sSR, sSA, sIN, sIN},
 316 /* i_a */{sCE, sCE, sCE, sCE, sCE, sCE, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 317 /* c_e */{sES, sES, sES, sES, sES, sES, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 318 /* c_a */{sES, sI1, sIN, sES, sES, sCW, sES, sES, sES, sSS, sSR, sSA, sRJ, sCL},
 319 /* s   */{sSR, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sSR, sSS, sSR, sSA, sRJ, sCL},
 320 /* s_a */{sCL, sIN, sIN, sCS, sCR, sCW, sCO, sCE, sCL, sCL, sSR, sCL, sRJ, sCL},
 321 /* s_c */{sCL, sCL, sCL, sCL, sCL, sCW, sCO, sCE, sES, sSS, sCL, sCL, sRJ, sCL},
 322 /* err */{sCL, sI1, sIN, sCS, sCR, sCW, sCO, sCE, sES, sSS, sSR, sSA, sRJ, sCL},
 323 /* ab  */{sCL, sCL, sCL, sCL, sCL, sRJ, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL},
 324         },
 325 };
 326
 327 #define IP_VS_SCTP_MAX_RTO      ((60 + 1) * HZ)
 328
 329 /* Timeout table[state] */
 330 static const int sctp_timeouts[IP_VS_SCTP_S_LAST + 1] = {
 331         [IP_VS_SCTP_S_NONE]                     = 2 * HZ,
 332         [IP_VS_SCTP_S_INIT1]                    = (0 + 3 + 1) * HZ,
 333         [IP_VS_SCTP_S_INIT]                     = IP_VS_SCTP_MAX_RTO,
 334         [IP_VS_SCTP_S_COOKIE_SENT]              = IP_VS_SCTP_MAX_RTO,
 335         [IP_VS_SCTP_S_COOKIE_REPLIED]           = IP_VS_SCTP_MAX_RTO,
 336         [IP_VS_SCTP_S_COOKIE_WAIT]              = IP_VS_SCTP_MAX_RTO,
 337         [IP_VS_SCTP_S_COOKIE]                   = IP_VS_SCTP_MAX_RTO,
 338         [IP_VS_SCTP_S_COOKIE_ECHOED]            = IP_VS_SCTP_MAX_RTO,
 339         [IP_VS_SCTP_S_ESTABLISHED]              = 15 * 60 * HZ,
 340         [IP_VS_SCTP_S_SHUTDOWN_SENT]            = IP_VS_SCTP_MAX_RTO,
 341         [IP_VS_SCTP_S_SHUTDOWN_RECEIVED]        = IP_VS_SCTP_MAX_RTO,
 342         [IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]        = IP_VS_SCTP_MAX_RTO,
 343         [IP_VS_SCTP_S_REJECTED]                 = (0 + 3 + 1) * HZ,
 344         [IP_VS_SCTP_S_CLOSED]                   = IP_VS_SCTP_MAX_RTO,
 345         [IP_VS_SCTP_S_LAST]                     = 2 * HZ,
 346 };
 347
 348 static const char *sctp_state_name_table[IP_VS_SCTP_S_LAST + 1] = {
 349         [IP_VS_SCTP_S_NONE]                     = "NONE",
 350         [IP_VS_SCTP_S_INIT1]                    = "INIT1",
 351         [IP_VS_SCTP_S_INIT]                     = "INIT",
 352         [IP_VS_SCTP_S_COOKIE_SENT]              = "C-SENT",
 353         [IP_VS_SCTP_S_COOKIE_REPLIED]           = "C-REPLIED",
 354         [IP_VS_SCTP_S_COOKIE_WAIT]              = "C-WAIT",
 355         [IP_VS_SCTP_S_COOKIE]                   = "COOKIE",
 356         [IP_VS_SCTP_S_COOKIE_ECHOED]            = "C-ECHOED",
 357         [IP_VS_SCTP_S_ESTABLISHED]              = "ESTABLISHED",
 358         [IP_VS_SCTP_S_SHUTDOWN_SENT]            = "S-SENT",
 359         [IP_VS_SCTP_S_SHUTDOWN_RECEIVED]        = "S-RECEIVED",
 360         [IP_VS_SCTP_S_SHUTDOWN_ACK_SENT]        = "S-ACK-SENT",
 361         [IP_VS_SCTP_S_REJECTED]                 = "REJECTED",
 362         [IP_VS_SCTP_S_CLOSED]                   = "CLOSED",
 363         [IP_VS_SCTP_S_LAST]                     = "BUG!",
 364 };
 365
 366
 367 static const char *sctp_state_name(int state)
 368 {
 369         if (state >= IP_VS_SCTP_S_LAST)
 370                 return "ERR!";
 371         if (sctp_state_name_table[state])
 372                 return sctp_state_name_table[state];
 373         return "?";
 374 }
 375
 376 static inline void
 377 set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
 378                 int direction, const struct sk_buff *skb)
 379 {
 380         struct sctp_chunkhdr _sctpch, *sch;
 381         unsigned char chunk_type;
 382         int event, next_state;
 383         int ihl, cofs;
 384
 385 #ifdef CONFIG_IP_VS_IPV6
 386         ihl = cp->af == AF_INET ? ip_hdrlen(skb) : sizeof(struct ipv6hdr);
 387 #else
 388         ihl = ip_hdrlen(skb);
 389 #endif
 390
 391         cofs = ihl + sizeof(struct sctphdr);
 392         sch = skb_header_pointer(skb, cofs, sizeof(_sctpch), &_sctpch);
 393         if (sch == NULL)
 394                 return;
 395
 396         chunk_type = sch->type;
 397         /*
 398          * Section 3: Multiple chunks can be bundled into one SCTP packet
 399          * up to the MTU size, except for the INIT, INIT ACK, and
 400          * SHUTDOWN COMPLETE chunks. These chunks MUST NOT be bundled with
 401          * any other chunk in a packet.
 402          *
 403          * Section 3.3.7: DATA chunks MUST NOT be bundled with ABORT. Control
 404          * chunks (except for INIT, INIT ACK, and SHUTDOWN COMPLETE) MAY be
 405          * bundled with an ABORT, but they MUST be placed before the ABORT
 406          * in the SCTP packet or they will be ignored by the receiver.
 407          */
 408         if ((sch->type == SCTP_CID_COOKIE_ECHO) ||
 409             (sch->type == SCTP_CID_COOKIE_ACK)) {
 410                 int clen = ntohs(sch->length);
 411
 412                 if (clen >= sizeof(_sctpch)) {
 413                         sch = skb_header_pointer(skb, cofs + ALIGN(clen, 4),
 414                                                  sizeof(_sctpch), &_sctpch);
 415                         if (sch && sch->type == SCTP_CID_ABORT)
 416                                 chunk_type = sch->type;
 417                 }
 418         }
 419
 420         event = (chunk_type < sizeof(sctp_events)) ?
 421                 sctp_events[chunk_type] : IP_VS_SCTP_DATA;
 422
 423         /* Update direction to INPUT_ONLY if necessary
 424          * or delete NO_OUTPUT flag if output packet detected
 425          */
 426         if (cp->flags & IP_VS_CONN_F_NOOUTPUT) {
 427                 if (direction == IP_VS_DIR_OUTPUT)
 428                         cp->flags &= ~IP_VS_CONN_F_NOOUTPUT;
 429                 else
 430                         direction = IP_VS_DIR_INPUT_ONLY;
 431         }
 432
 433         next_state = sctp_states[direction][event][cp->state];
 434
 435         if (next_state != cp->state) {
 436                 struct ip_vs_dest *dest = cp->dest;
 437
 438                 IP_VS_DBG_BUF(8, "%s %s  %s:%d->"
 439                                 "%s:%d state: %s->%s conn->refcnt:%d\n",
 440                                 pd->pp->name,
 441                                 ((direction == IP_VS_DIR_OUTPUT) ?
 442                                  "output " : "input "),
 443                                 IP_VS_DBG_ADDR(cp->daf, &cp->daddr),
 444                                 ntohs(cp->dport),
 445                                 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
 446                                 ntohs(cp->cport),
 447                                 sctp_state_name(cp->state),
 448                                 sctp_state_name(next_state),
 449                                 refcount_read(&cp->refcnt));
 450                 if (dest) {
 451                         if (!(cp->flags & IP_VS_CONN_F_INACTIVE) &&
 452                                 (next_state != IP_VS_SCTP_S_ESTABLISHED)) {
 453                                 atomic_dec(&dest->activeconns);
 454                                 atomic_inc(&dest->inactconns);
 455                                 cp->flags |= IP_VS_CONN_F_INACTIVE;
 456                         } else if ((cp->flags & IP_VS_CONN_F_INACTIVE) &&
 457                                    (next_state == IP_VS_SCTP_S_ESTABLISHED)) {
 458                                 atomic_inc(&dest->activeconns);
 459                                 atomic_dec(&dest->inactconns);
 460                                 cp->flags &= ~IP_VS_CONN_F_INACTIVE;
 461                         }
 462                 }
 463         }
 464         if (likely(pd))
 465                 cp->timeout = pd->timeout_table[cp->state = next_state];
 466         else    /* What to do ? */
 467                 cp->timeout = sctp_timeouts[cp->state = next_state];
 468 }
 469
 470 static void
 471 sctp_state_transition(struct ip_vs_conn *cp, int direction,
 472                 const struct sk_buff *skb, struct ip_vs_proto_data *pd)
 473 {
 474         spin_lock_bh(&cp->lock);
 475         set_sctp_state(pd, cp, direction, skb);
 476         spin_unlock_bh(&cp->lock);
 477 }
 478
 479 static inline __u16 sctp_app_hashkey(__be16 port)
 480 {
 481         return (((__force u16)port >> SCTP_APP_TAB_BITS) ^ (__force u16)port)
 482                 & SCTP_APP_TAB_MASK;
 483 }
 484
 485 static int sctp_register_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
 486 {
 487         struct ip_vs_app *i;
 488         __u16 hash;
 489         __be16 port = inc->port;
 490         int ret = 0;
 491         struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
 492
 493         hash = sctp_app_hashkey(port);
 494
 495         list_for_each_entry(i, &ipvs->sctp_apps[hash], p_list) {
 496                 if (i->port == port) {
 497                         ret = -EEXIST;
 498                         goto out;
 499                 }
 500         }
 501         list_add_rcu(&inc->p_list, &ipvs->sctp_apps[hash]);
 502         atomic_inc(&pd->appcnt);
 503 out:
 504
 505         return ret;
 506 }
 507
 508 static void sctp_unregister_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
 509 {
 510         struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_SCTP);
 511
 512         atomic_dec(&pd->appcnt);
 513         list_del_rcu(&inc->p_list);
 514 }
 515
 516 static int sctp_app_conn_bind(struct ip_vs_conn *cp)
 517 {
 518         struct netns_ipvs *ipvs = cp->ipvs;
 519         int hash;
 520         struct ip_vs_app *inc;
 521         int result = 0;
 522
 523         /* Default binding: bind app only for NAT */
 524         if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
 525                 return 0;
 526         /* Lookup application incarnations and bind the right one */
 527         hash = sctp_app_hashkey(cp->vport);
 528
 529         list_for_each_entry_rcu(inc, &ipvs->sctp_apps[hash], p_list) {
 530                 if (inc->port == cp->vport) {
 531                         if (unlikely(!ip_vs_app_inc_get(inc)))
 532                                 break;
 533
 534                         IP_VS_DBG_BUF(9, "%s: Binding conn %s:%u->"
 535                                         "%s:%u to app %s on port %u\n",
 536                                         __func__,
 537                                         IP_VS_DBG_ADDR(cp->af, &cp->caddr),
 538                                         ntohs(cp->cport),
 539                                         IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
 540                                         ntohs(cp->vport),
 541                                         inc->name, ntohs(inc->port));
 542                         cp->app = inc;
 543                         if (inc->init_conn)
 544                                 result = inc->init_conn(inc, cp);
 545                         break;
 546                 }
 547         }
 548
 549         return result;
 550 }
 551
 552 /* ---------------------------------------------
 553  *   timeouts is netns related now.
 554  * ---------------------------------------------
 555  */
 556 static int __ip_vs_sctp_init(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
 557 {
 558         ip_vs_init_hash_table(ipvs->sctp_apps, SCTP_APP_TAB_SIZE);
 559         pd->timeout_table = ip_vs_create_timeout_table((int *)sctp_timeouts,
 560                                                         sizeof(sctp_timeouts));
 561         if (!pd->timeout_table)
 562                 return -ENOMEM;
 563         return 0;
 564 }
 565
 566 static void __ip_vs_sctp_exit(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
 567 {
 568         kfree(pd->timeout_table);
 569 }
 570
 571 struct ip_vs_protocol ip_vs_protocol_sctp = {
 572         .name           = "SCTP",
 573         .protocol       = IPPROTO_SCTP,
 574         .num_states     = IP_VS_SCTP_S_LAST,
 575         .dont_defrag    = 0,
 576         .init           = NULL,
 577         .exit           = NULL,
 578         .init_netns     = __ip_vs_sctp_init,
 579         .exit_netns     = __ip_vs_sctp_exit,
 580         .register_app   = sctp_register_app,
 581         .unregister_app = sctp_unregister_app,
 582         .conn_schedule  = sctp_conn_schedule,
 583         .conn_in_get    = ip_vs_conn_in_get_proto,
 584         .conn_out_get   = ip_vs_conn_out_get_proto,
 585         .snat_handler   = sctp_snat_handler,
 586         .dnat_handler   = sctp_dnat_handler,
 587         .csum_check     = sctp_csum_check,
 588         .state_name     = sctp_state_name,
 589         .state_transition = sctp_state_transition,
 590         .app_conn_bind  = sctp_app_conn_bind,
 591         .debug_packet   = ip_vs_tcpudp_debug_packet,
 592         .timeout_change = NULL,
 593 };