net/dccp/minisocks.c

   1 /*
   2  *  net/dccp/minisocks.c
   3  *
   4  *  An implementation of the DCCP protocol
   5  *  Arnaldo Carvalho de Melo <acme@conectiva.com.br>
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License
   9  *      as published by the Free Software Foundation; either version
  10  *      2 of the License, or (at your option) any later version.
  11  */
  12
  13 #include <linux/dccp.h>
  14 #include <linux/kernel.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/timer.h>
  17
  18 #include <net/sock.h>
  19 #include <net/xfrm.h>
  20 #include <net/inet_timewait_sock.h>
  21
  22 #include "ackvec.h"
  23 #include "ccid.h"
  24 #include "dccp.h"
  25 #include "feat.h"
  26
  27 struct inet_timewait_death_row dccp_death_row = {
  28         .sysctl_max_tw_buckets = NR_FILE * 2,
  29         .period         = DCCP_TIMEWAIT_LEN / INET_TWDR_TWKILL_SLOTS,
  30         .death_lock     = __SPIN_LOCK_UNLOCKED(dccp_death_row.death_lock),
  31         .hashinfo       = &dccp_hashinfo,
  32         .tw_timer       = TIMER_INITIALIZER(inet_twdr_hangman, 0,
  33                                             (unsigned long)&dccp_death_row),
  34         .twkill_work    = __WORK_INITIALIZER(dccp_death_row.twkill_work,
  35                                              inet_twdr_twkill_work),
  36 /* Short-time timewait calendar */
  37
  38         .twcal_hand     = -1,
  39         .twcal_timer    = TIMER_INITIALIZER(inet_twdr_twcal_tick, 0,
  40                                             (unsigned long)&dccp_death_row),
  41 };
  42
  43 EXPORT_SYMBOL_GPL(dccp_death_row);
  44
  45 void dccp_minisock_init(struct dccp_minisock *dmsk)
  46 {
  47         dmsk->dccpms_sequence_window = sysctl_dccp_feat_sequence_window;
  48         dmsk->dccpms_rx_ccid         = sysctl_dccp_feat_rx_ccid;
  49         dmsk->dccpms_tx_ccid         = sysctl_dccp_feat_tx_ccid;
  50         dmsk->dccpms_send_ack_vector = sysctl_dccp_feat_send_ack_vector;
  51         dmsk->dccpms_send_ndp_count  = sysctl_dccp_feat_send_ndp_count;
  52 }
  53
  54 void dccp_time_wait(struct sock *sk, int state, int timeo)
  55 {
  56         struct inet_timewait_sock *tw = NULL;
  57
  58         if (dccp_death_row.tw_count < dccp_death_row.sysctl_max_tw_buckets)
  59                 tw = inet_twsk_alloc(sk, state);
  60
  61         if (tw != NULL) {
  62                 const struct inet_connection_sock *icsk = inet_csk(sk);
  63                 const int rto = (icsk->icsk_rto << 2) - (icsk->icsk_rto >> 1);
  64 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  65                 if (tw->tw_family == PF_INET6) {
  66                         const struct ipv6_pinfo *np = inet6_sk(sk);
  67                         struct inet6_timewait_sock *tw6;
  68
  69                         tw->tw_ipv6_offset = inet6_tw_offset(sk->sk_prot);
  70                         tw6 = inet6_twsk((struct sock *)tw);
  71                         ipv6_addr_copy(&tw6->tw_v6_daddr, &np->daddr);
  72                         ipv6_addr_copy(&tw6->tw_v6_rcv_saddr, &np->rcv_saddr);
  73                         tw->tw_ipv6only = np->ipv6only;
  74                 }
  75 #endif
  76                 /* Linkage updates. */
  77                 __inet_twsk_hashdance(tw, sk, &dccp_hashinfo);
  78
  79                 /* Get the TIME_WAIT timeout firing. */
  80                 if (timeo < rto)
  81                         timeo = rto;
  82
  83                 tw->tw_timeout = DCCP_TIMEWAIT_LEN;
  84                 if (state == DCCP_TIME_WAIT)
  85                         timeo = DCCP_TIMEWAIT_LEN;
  86
  87                 inet_twsk_schedule(tw, &dccp_death_row, timeo,
  88                                    DCCP_TIMEWAIT_LEN);
  89                 inet_twsk_put(tw);
  90         } else {
  91                 /* Sorry, if we're out of memory, just CLOSE this
  92                  * socket up.  We've got bigger problems than
  93                  * non-graceful socket closings.
  94                  */
  95                 DCCP_WARN("time wait bucket table overflow\n");
  96         }
  97
  98         dccp_done(sk);
  99 }
 100
 101 struct sock *dccp_create_openreq_child(struct sock *sk,
 102                                        const struct request_sock *req,
 103                                        const struct sk_buff *skb)
 104 {
 105         /*
 106          * Step 3: Process LISTEN state
 107          *
 108          *   (* Generate a new socket and switch to that socket *)
 109          *   Set S := new socket for this port pair
 110          */
 111         struct sock *newsk = inet_csk_clone(sk, req, GFP_ATOMIC);
 112
 113         if (newsk != NULL) {
 114                 struct dccp_request_sock *dreq = dccp_rsk(req);
 115                 struct inet_connection_sock *newicsk = inet_csk(newsk);
 116                 struct dccp_sock *newdp = dccp_sk(newsk);
 117                 struct dccp_minisock *newdmsk = dccp_msk(newsk);
 118
 119                 newdp->dccps_role           = DCCP_ROLE_SERVER;
 120                 newdp->dccps_hc_rx_ackvec   = NULL;
 121                 newdp->dccps_service_list   = NULL;
 122                 newdp->dccps_service        = dreq->dreq_service;
 123                 newdp->dccps_timestamp_echo = dreq->dreq_timestamp_echo;
 124                 newdp->dccps_timestamp_time = dreq->dreq_timestamp_time;
 125                 newicsk->icsk_rto           = DCCP_TIMEOUT_INIT;
 126
 127                 INIT_LIST_HEAD(&newdp->dccps_featneg);
 128                 /*
 129                  * Step 3: Process LISTEN state
 130                  *
 131                  *    Choose S.ISS (initial seqno) or set from Init Cookies
 132                  *    Initialize S.GAR := S.ISS
 133                  *    Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
 134                  */
 135
 136                 /* See dccp_v4_conn_request */
 137                 newdmsk->dccpms_sequence_window = req->rcv_wnd;
 138
 139                 newdp->dccps_gar = newdp->dccps_iss = dreq->dreq_iss;
 140                 dccp_update_gss(newsk, dreq->dreq_iss);
 141
 142                 newdp->dccps_isr = dreq->dreq_isr;
 143                 dccp_update_gsr(newsk, dreq->dreq_isr);
 144
 145                 /*
 146                  * SWL and AWL are initially adjusted so that they are not less than
 147                  * the initial Sequence Numbers received and sent, respectively:
 148                  *      SWL := max(GSR + 1 - floor(W/4), ISR),
 149                  *      AWL := max(GSS - W' + 1, ISS).
 150                  * These adjustments MUST be applied only at the beginning of the
 151                  * connection.
 152                  */
 153                 dccp_set_seqno(&newdp->dccps_swl,
 154                                max48(newdp->dccps_swl, newdp->dccps_isr));
 155                 dccp_set_seqno(&newdp->dccps_awl,
 156                                max48(newdp->dccps_awl, newdp->dccps_iss));
 157
 158                 /*
 159                  * Activate features after initialising the sequence numbers,
 160                  * since CCID initialisation may depend on GSS, ISR, ISS etc.
 161                  */
 162                 if (dccp_feat_activate_values(newsk, &dreq->dreq_featneg)) {
 163                         /* It is still raw copy of parent, so invalidate
 164                          * destructor and make plain sk_free() */
 165                         newsk->sk_destruct = NULL;
 166                         sk_free(newsk);
 167                         return NULL;
 168                 }
 169                 dccp_init_xmit_timers(newsk);
 170
 171                 DCCP_INC_STATS_BH(DCCP_MIB_PASSIVEOPENS);
 172         }
 173         return newsk;
 174 }
 175
 176 EXPORT_SYMBOL_GPL(dccp_create_openreq_child);
 177
 178 /*
 179  * Process an incoming packet for RESPOND sockets represented
 180  * as an request_sock.
 181  */
 182 struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
 183                             struct request_sock *req,
 184                             struct request_sock **prev)
 185 {
 186         struct sock *child = NULL;
 187         struct dccp_request_sock *dreq = dccp_rsk(req);
 188
 189         /* Check for retransmitted REQUEST */
 190         if (dccp_hdr(skb)->dccph_type == DCCP_PKT_REQUEST) {
 191
 192                 if (after48(DCCP_SKB_CB(skb)->dccpd_seq, dreq->dreq_isr)) {
 193                         dccp_pr_debug("Retransmitted REQUEST\n");
 194                         dreq->dreq_isr = DCCP_SKB_CB(skb)->dccpd_seq;
 195                         /*
 196                          * Send another RESPONSE packet
 197                          * To protect against Request floods, increment retrans
 198                          * counter (backoff, monitored by dccp_response_timer).
 199                          */
 200                         req->retrans++;
 201                         req->rsk_ops->rtx_syn_ack(sk, req);
 202                 }
 203                 /* Network Duplicate, discard packet */
 204                 return NULL;
 205         }
 206
 207         DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR;
 208
 209         if (dccp_hdr(skb)->dccph_type != DCCP_PKT_ACK &&
 210             dccp_hdr(skb)->dccph_type != DCCP_PKT_DATAACK)
 211                 goto drop;
 212
 213         /* Invalid ACK */
 214         if (DCCP_SKB_CB(skb)->dccpd_ack_seq != dreq->dreq_iss) {
 215                 dccp_pr_debug("Invalid ACK number: ack_seq=%llu, "
 216                               "dreq_iss=%llu\n",
 217                               (unsigned long long)
 218                               DCCP_SKB_CB(skb)->dccpd_ack_seq,
 219                               (unsigned long long) dreq->dreq_iss);
 220                 goto drop;
 221         }
 222
 223         if (dccp_parse_options(sk, dreq, skb))
 224                  goto drop;
 225
 226         child = inet_csk(sk)->icsk_af_ops->syn_recv_sock(sk, skb, req, NULL);
 227         if (child == NULL)
 228                 goto listen_overflow;
 229
 230         inet_csk_reqsk_queue_unlink(sk, req, prev);
 231         inet_csk_reqsk_queue_removed(sk, req);
 232         inet_csk_reqsk_queue_add(sk, req, child);
 233 out:
 234         return child;
 235 listen_overflow:
 236         dccp_pr_debug("listen_overflow!\n");
 237         DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
 238 drop:
 239         if (dccp_hdr(skb)->dccph_type != DCCP_PKT_RESET)
 240                 req->rsk_ops->send_reset(sk, skb);
 241
 242         inet_csk_reqsk_queue_drop(sk, req, prev);
 243         goto out;
 244 }
 245
 246 EXPORT_SYMBOL_GPL(dccp_check_req);
 247
 248 /*
 249  *  Queue segment on the new socket if the new socket is active,
 250  *  otherwise we just shortcircuit this and continue with
 251  *  the new socket.
 252  */
 253 int dccp_child_process(struct sock *parent, struct sock *child,
 254                        struct sk_buff *skb)
 255 {
 256         int ret = 0;
 257         const int state = child->sk_state;
 258
 259         if (!sock_owned_by_user(child)) {
 260                 ret = dccp_rcv_state_process(child, skb, dccp_hdr(skb),
 261                                              skb->len);
 262
 263                 /* Wakeup parent, send SIGIO */
 264                 if (state == DCCP_RESPOND && child->sk_state != state)
 265                         parent->sk_data_ready(parent, 0);
 266         } else {
 267                 /* Alas, it is possible again, because we do lookup
 268                  * in main socket hash table and lock on listening
 269                  * socket does not protect us more.
 270                  */
 271                 sk_add_backlog(child, skb);
 272         }
 273
 274         bh_unlock_sock(child);
 275         sock_put(child);
 276         return ret;
 277 }
 278
 279 EXPORT_SYMBOL_GPL(dccp_child_process);
 280
 281 void dccp_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
 282                          struct request_sock *rsk)
 283 {
 284         DCCP_BUG("DCCP-ACK packets are never sent in LISTEN/RESPOND state");
 285 }
 286
 287 EXPORT_SYMBOL_GPL(dccp_reqsk_send_ack);
 288
 289 int dccp_reqsk_init(struct request_sock *req,
 290                     struct dccp_sock const *dp, struct sk_buff const *skb)
 291 {
 292         struct dccp_request_sock *dreq = dccp_rsk(req);
 293
 294         inet_rsk(req)->rmt_port   = dccp_hdr(skb)->dccph_sport;
 295         inet_rsk(req)->acked      = 0;
 296         req->rcv_wnd              = sysctl_dccp_feat_sequence_window;
 297         dreq->dreq_timestamp_echo = 0;
 298
 299         /* inherit feature negotiation options from listening socket */
 300         return dccp_feat_clone_list(&dp->dccps_featneg, &dreq->dreq_featneg);
 301 }
 302
 303 EXPORT_SYMBOL_GPL(dccp_reqsk_init);